]> git.sur5r.net Git - u-boot/commitdiff
mtd, ubi, ubifs: resync with Linux-3.14
authorHeiko Schocher <hs@denx.de>
Tue, 24 Jun 2014 08:10:04 +0000 (10:10 +0200)
committerTom Rini <trini@ti.com>
Mon, 25 Aug 2014 23:25:55 +0000 (19:25 -0400)
resync ubi subsystem with linux:

commit 455c6fdbd219161bd09b1165f11699d6d73de11c
Author: Linus Torvalds <torvalds@linux-foundation.org>
Date:   Sun Mar 30 20:40:15 2014 -0700

    Linux 3.14

A nice side effect of this, is we introduce UBI Fastmap support
to U-Boot.

Signed-off-by: Heiko Schocher <hs@denx.de>
Signed-off-by: Tom Rini <trini@ti.com>
Cc: Marek Vasut <marex@denx.de>
Cc: Sergey Lapin <slapin@ossfans.org>
Cc: Scott Wood <scottwood@freescale.com>
Cc: Joerg Krause <jkrause@posteo.de>
75 files changed:
README
board/prodrive/alpr/nand.c
board/socrates/nand.c
board/tqc/tqm8272/nand.c
common/cmd_ubi.c
common/cmd_ubifs.c
drivers/mtd/mtdconcat.c
drivers/mtd/mtdcore.c
drivers/mtd/mtdcore.h [new file with mode: 0644]
drivers/mtd/mtdpart.c
drivers/mtd/nand/fsl_elbc_nand.c
drivers/mtd/nand/fsl_ifc_nand.c
drivers/mtd/nand/fsl_upm.c
drivers/mtd/nand/mpc5121_nfc.c
drivers/mtd/nand/mxc_nand.c
drivers/mtd/nand/nand_base.c
drivers/mtd/nand/nand_bbt.c
drivers/mtd/nand/nand_ids.c
drivers/mtd/nand/nand_util.c
drivers/mtd/nand/ndfc.c
drivers/mtd/onenand/onenand_base.c
drivers/mtd/onenand/onenand_bbt.c
drivers/mtd/onenand/samsung.c
drivers/mtd/ubi/Makefile
drivers/mtd/ubi/attach.c [new file with mode: 0644]
drivers/mtd/ubi/build.c
drivers/mtd/ubi/crc32.c
drivers/mtd/ubi/crc32table.h
drivers/mtd/ubi/debug.c
drivers/mtd/ubi/debug.h
drivers/mtd/ubi/eba.c
drivers/mtd/ubi/fastmap.c [new file with mode: 0644]
drivers/mtd/ubi/io.c
drivers/mtd/ubi/kapi.c
drivers/mtd/ubi/misc.c
drivers/mtd/ubi/scan.c [deleted file]
drivers/mtd/ubi/scan.h [deleted file]
drivers/mtd/ubi/ubi-media.h
drivers/mtd/ubi/ubi.h
drivers/mtd/ubi/upd.c
drivers/mtd/ubi/vmt.c
drivers/mtd/ubi/vtbl.c
drivers/mtd/ubi/wl.c
fs/ubifs/budget.c
fs/ubifs/debug.c
fs/ubifs/debug.h
fs/ubifs/io.c
fs/ubifs/key.h
fs/ubifs/log.c
fs/ubifs/lprops.c
fs/ubifs/lpt.c
fs/ubifs/lpt_commit.c
fs/ubifs/master.c
fs/ubifs/misc.h
fs/ubifs/orphan.c
fs/ubifs/recovery.c
fs/ubifs/replay.c
fs/ubifs/sb.c
fs/ubifs/scan.c
fs/ubifs/super.c
fs/ubifs/tnc.c
fs/ubifs/tnc_misc.c
fs/ubifs/ubifs-media.h
fs/ubifs/ubifs.c
fs/ubifs/ubifs.h
include/linux/mtd/bbm.h
include/linux/mtd/concat.h
include/linux/mtd/flashchip.h [new file with mode: 0644]
include/linux/mtd/mtd.h
include/linux/mtd/nand.h
include/linux/mtd/partitions.h
include/linux/mtd/ubi.h
include/mtd/mtd-abi.h
include/mtd/ubi-user.h
include/usb/lin_gadget_compat.h

diff --git a/README b/README
index 14d6b227d689825025f9dfc98fb305021882446d..1e63f048efe6fb6a9e004e530832c80bb3520fe8 100644 (file)
--- a/README
+++ b/README
@@ -3338,6 +3338,9 @@ FIT uImage format:
                Adds the MTD partitioning infrastructure from the Linux
                kernel. Needed for UBI support.
 
+               CONFIG_MTD_NAND_VERIFY_WRITE
+               verify if the written data is correct reread.
+
 - UBI support
                CONFIG_CMD_UBI
 
@@ -3351,6 +3354,64 @@ FIT uImage format:
                Make the verbose messages from UBI stop printing.  This leaves
                warnings and errors enabled.
 
+
+               CONFIG_MTD_UBI_WL_THRESHOLD
+               This parameter defines the maximum difference between the highest
+               erase counter value and the lowest erase counter value of eraseblocks
+               of UBI devices. When this threshold is exceeded, UBI starts performing
+               wear leveling by means of moving data from eraseblock with low erase
+               counter to eraseblocks with high erase counter.
+
+               The default value should be OK for SLC NAND flashes, NOR flashes and
+               other flashes which have eraseblock life-cycle 100000 or more.
+               However, in case of MLC NAND flashes which typically have eraseblock
+               life-cycle less than 10000, the threshold should be lessened (e.g.,
+               to 128 or 256, although it does not have to be power of 2).
+
+               default: 4096
+               
+               CONFIG_MTD_UBI_BEB_LIMIT
+               This option specifies the maximum bad physical eraseblocks UBI
+               expects on the MTD device (per 1024 eraseblocks). If the
+               underlying flash does not admit of bad eraseblocks (e.g. NOR
+               flash), this value is ignored.
+
+               NAND datasheets often specify the minimum and maximum NVM
+               (Number of Valid Blocks) for the flashes' endurance lifetime.
+               The maximum expected bad eraseblocks per 1024 eraseblocks
+               then can be calculated as "1024 * (1 - MinNVB / MaxNVB)",
+               which gives 20 for most NANDs (MaxNVB is basically the total
+               count of eraseblocks on the chip).
+
+               To put it differently, if this value is 20, UBI will try to
+               reserve about 1.9% of physical eraseblocks for bad blocks
+               handling. And that will be 1.9% of eraseblocks on the entire
+               NAND chip, not just the MTD partition UBI attaches. This means
+               that if you have, say, a NAND flash chip admits maximum 40 bad
+               eraseblocks, and it is split on two MTD partitions of the same
+               size, UBI will reserve 40 eraseblocks when attaching a
+               partition.
+
+               default: 20
+
+               CONFIG_MTD_UBI_FASTMAP
+               Fastmap is a mechanism which allows attaching an UBI device
+               in nearly constant time. Instead of scanning the whole MTD device it
+               only has to locate a checkpoint (called fastmap) on the device.
+               The on-flash fastmap contains all information needed to attach
+               the device. Using fastmap makes only sense on large devices where
+               attaching by scanning takes long. UBI will not automatically install
+               a fastmap on old images, but you can set the UBI parameter
+               CONFIG_MTD_UBI_FASTMAP_AUTOCONVERT to 1 if you want so. Please note
+               that fastmap-enabled images are still usable with UBI implementations
+               without fastmap support. On typical flash devices the whole fastmap
+               fits into one PEB. UBI will reserve PEBs to hold two fastmaps.
+
+               CONFIG_MTD_UBI_FASTMAP_AUTOCONVERT
+               Set this parameter to enable fastmap automatically on images
+               without a fastmap.
+               default: 0
+
 - UBIFS support
                CONFIG_CMD_UBIFS
 
index 50e8d82b4b38efcbe9af59981fd05a1fbbe3650b..5427de5634213ba427257ff07fd18cff79949a69 100644 (file)
@@ -93,6 +93,7 @@ static void alpr_nand_read_buf(struct mtd_info *mtd, u_char *buf, int len)
        }
 }
 
+#if defined(CONFIG_MTD_NAND_VERIFY_WRITE)
 static int alpr_nand_verify_buf(struct mtd_info *mtd, const u_char *buf, int len)
 {
        int i;
@@ -103,6 +104,7 @@ static int alpr_nand_verify_buf(struct mtd_info *mtd, const u_char *buf, int len
 
        return 0;
 }
+#endif
 
 static int alpr_nand_dev_ready(struct mtd_info *mtd)
 {
@@ -128,7 +130,9 @@ int board_nand_init(struct nand_chip *nand)
        nand->read_byte  = alpr_nand_read_byte;
        nand->write_buf  = alpr_nand_write_buf;
        nand->read_buf   = alpr_nand_read_buf;
+#if defined(CONFIG_MTD_NAND_VERIFY_WRITE)
        nand->verify_buf = alpr_nand_verify_buf;
+#endif
        nand->dev_ready  = alpr_nand_dev_ready;
 
        return 0;
index 3802c7ec7abcd6bd2ff314c90d2adc450edc991a..73944780bfe38fe635898f8181aad6614e09a4aa 100644 (file)
@@ -18,7 +18,9 @@ static void sc_nand_write_buf(struct mtd_info *mtd, const u_char *buf, int len);
 static u_char sc_nand_read_byte(struct mtd_info *mtd);
 static u16 sc_nand_read_word(struct mtd_info *mtd);
 static void sc_nand_read_buf(struct mtd_info *mtd, u_char *buf, int len);
+#if defined(CONFIG_MTD_NAND_VERIFY_WRITE)
 static int sc_nand_verify_buf(struct mtd_info *mtd, const u_char *buf, int len);
+#endif
 static int sc_nand_device_ready(struct mtd_info *mtdinfo);
 
 #define FPGA_NAND_CMD_MASK             (0x7 << 28)
@@ -100,6 +102,7 @@ static void sc_nand_read_buf(struct mtd_info *mtd, u_char *buf, int len)
        }
 }
 
+#if defined(CONFIG_MTD_NAND_VERIFY_WRITE)
 /**
  * sc_nand_verify_buf -  Verify chip data against buffer
  * @mtd:       MTD device structure
@@ -116,6 +119,7 @@ static int sc_nand_verify_buf(struct mtd_info *mtd, const u_char *buf, int len)
        }
        return 0;
 }
+#endif
 
 /**
  * sc_nand_device_ready - Check the NAND device is ready for next command.
@@ -174,7 +178,9 @@ int board_nand_init(struct nand_chip *nand)
        nand->read_word = sc_nand_read_word;
        nand->write_buf = sc_nand_write_buf;
        nand->read_buf = sc_nand_read_buf;
+#if defined(CONFIG_MTD_NAND_VERIFY_WRITE)
        nand->verify_buf = sc_nand_verify_buf;
+#endif
 
        return 0;
 }
index 4925b8dda3055b1d6651e46004551fe136f49756..7fb2dfabc17e79f2fc1ebac2b07628bd169e4fdc 100644 (file)
@@ -188,6 +188,7 @@ static void tqm8272_write_buf(struct mtd_info *mtdinfo, const uint8_t *buf, int
                *base = buf[i];
 }
 
+#if defined(CONFIG_MTD_NAND_VERIFY_WRITE)
 static int tqm8272_verify_buf(struct mtd_info *mtdinfo, const uint8_t *buf, int len)
 {
        struct nand_chip *this = mtdinfo->priv;
@@ -199,6 +200,7 @@ static int tqm8272_verify_buf(struct mtd_info *mtdinfo, const uint8_t *buf, int
                        return -1;
        return 0;
 }
+#endif
 #endif /* #ifndef CONFIG_NAND_SPL */
 
 void board_nand_select_device(struct nand_chip *nand, int chip)
@@ -247,7 +249,9 @@ int board_nand_init(struct nand_chip *nand)
 #ifndef CONFIG_NAND_SPL
        nand->write_buf  = tqm8272_write_buf;
        nand->read_buf   = tqm8272_read_buf;
+#if defined(CONFIG_MTD_NAND_VERIFY_WRITE)
        nand->verify_buf = tqm8272_verify_buf;
+#endif
 #endif
 
        /*
index 3c37c93f100e2390c5b5d2177bc94bbff3786c7b..6c8570377e31d4726424eea67b4e53e417a08469 100644 (file)
@@ -19,6 +19,7 @@
 #include <onenand_uboot.h>
 #include <linux/mtd/mtd.h>
 #include <linux/mtd/partitions.h>
+#include <linux/err.h>
 #include <ubi_uboot.h>
 #include <asm/errno.h>
 #include <jffs2/load_kernel.h>
@@ -50,33 +51,6 @@ int ubifs_is_mounted(void);
 void cmd_ubifs_umount(void);
 #endif
 
-static void ubi_dump_vol_info(const struct ubi_volume *vol)
-{
-       ubi_msg("volume information dump:");
-       ubi_msg("vol_id          %d", vol->vol_id);
-       ubi_msg("reserved_pebs   %d", vol->reserved_pebs);
-       ubi_msg("alignment       %d", vol->alignment);
-       ubi_msg("data_pad        %d", vol->data_pad);
-       ubi_msg("vol_type        %d", vol->vol_type);
-       ubi_msg("name_len        %d", vol->name_len);
-       ubi_msg("usable_leb_size %d", vol->usable_leb_size);
-       ubi_msg("used_ebs        %d", vol->used_ebs);
-       ubi_msg("used_bytes      %lld", vol->used_bytes);
-       ubi_msg("last_eb_bytes   %d", vol->last_eb_bytes);
-       ubi_msg("corrupted       %d", vol->corrupted);
-       ubi_msg("upd_marker      %d", vol->upd_marker);
-
-       if (vol->name_len <= UBI_VOL_NAME_MAX &&
-               strnlen(vol->name, vol->name_len + 1) == vol->name_len) {
-               ubi_msg("name            %s", vol->name);
-       } else {
-               ubi_msg("the 1st 5 characters of the name: %c%c%c%c%c",
-                               vol->name[0], vol->name[1], vol->name[2],
-                               vol->name[3], vol->name[4]);
-       }
-       printf("\n");
-}
-
 static void display_volume_info(struct ubi_device *ubi)
 {
        int i;
index 19c8a43ce4fceeb788a509c844b45aa3339b9375..8e9a4e5038d87c19cfda9850dab0dbfa3d7446b0 100644 (file)
@@ -38,7 +38,7 @@ static int do_ubifs_mount(cmd_tbl_t *cmdtp, int flag, int argc,
                ubifs_initialized = 1;
        }
 
-       ret = ubifs_mount(vol_name);
+       ret = uboot_ubifs_mount(vol_name);
        if (ret)
                return -1;
 
index 31e4289b1601a4bb9de7cb2df4862258dc34bae8..39daeabd9f1e578741fc2a2403ae7d353c8fef05 100644 (file)
@@ -1,16 +1,32 @@
 /*
  * MTD device concatenation layer
  *
- * (C) 2002 Robert Kaiser <rkaiser@sysgo.de>
+ * Copyright © 2002 Robert Kaiser <rkaiser@sysgo.de>
+ * Copyright © 2002-2010 David Woodhouse <dwmw2@infradead.org>
  *
  * NAND support by Christian Gan <cgan@iders.ca>
  *
- * This code is GPL
+ * SPDX-License-Identifier:    GPL-2.0+
+ *
  */
 
-#include <linux/mtd/mtd.h>
+#define __UBOOT__
+#ifndef __UBOOT__
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/slab.h>
+#include <linux/sched.h>
+#include <linux/types.h>
+#include <linux/backing-dev.h>
+#include <asm/div64.h>
+#else
+#include <div64.h>
 #include <linux/compat.h>
+#endif
+
+#include <linux/mtd/mtd.h>
 #include <linux/mtd/concat.h>
+
 #include <ubi_uboot.h>
 
 /*
@@ -51,7 +67,9 @@ concat_read(struct mtd_info *mtd, loff_t from, size_t len,
        int ret = 0, err;
        int i;
 
+#ifdef __UBOOT__
        *retlen = 0;
+#endif
 
        for (i = 0; i < concat->num_subdev; i++) {
                struct mtd_info *subdev = concat->subdev[i];
@@ -105,7 +123,9 @@ concat_write(struct mtd_info *mtd, loff_t to, size_t len,
        int err = -EINVAL;
        int i;
 
+#ifdef __UBOOT__
        *retlen = 0;
+#endif
 
        for (i = 0; i < concat->num_subdev; i++) {
                struct mtd_info *subdev = concat->subdev[i];
@@ -137,6 +157,83 @@ concat_write(struct mtd_info *mtd, loff_t to, size_t len,
        return err;
 }
 
+#ifndef __UBOOT__
+static int
+concat_writev(struct mtd_info *mtd, const struct kvec *vecs,
+               unsigned long count, loff_t to, size_t * retlen)
+{
+       struct mtd_concat *concat = CONCAT(mtd);
+       struct kvec *vecs_copy;
+       unsigned long entry_low, entry_high;
+       size_t total_len = 0;
+       int i;
+       int err = -EINVAL;
+
+       /* Calculate total length of data */
+       for (i = 0; i < count; i++)
+               total_len += vecs[i].iov_len;
+
+       /* Check alignment */
+       if (mtd->writesize > 1) {
+               uint64_t __to = to;
+               if (do_div(__to, mtd->writesize) || (total_len % mtd->writesize))
+                       return -EINVAL;
+       }
+
+       /* make a copy of vecs */
+       vecs_copy = kmemdup(vecs, sizeof(struct kvec) * count, GFP_KERNEL);
+       if (!vecs_copy)
+               return -ENOMEM;
+
+       entry_low = 0;
+       for (i = 0; i < concat->num_subdev; i++) {
+               struct mtd_info *subdev = concat->subdev[i];
+               size_t size, wsize, retsize, old_iov_len;
+
+               if (to >= subdev->size) {
+                       to -= subdev->size;
+                       continue;
+               }
+
+               size = min_t(uint64_t, total_len, subdev->size - to);
+               wsize = size; /* store for future use */
+
+               entry_high = entry_low;
+               while (entry_high < count) {
+                       if (size <= vecs_copy[entry_high].iov_len)
+                               break;
+                       size -= vecs_copy[entry_high++].iov_len;
+               }
+
+               old_iov_len = vecs_copy[entry_high].iov_len;
+               vecs_copy[entry_high].iov_len = size;
+
+               err = mtd_writev(subdev, &vecs_copy[entry_low],
+                                entry_high - entry_low + 1, to, &retsize);
+
+               vecs_copy[entry_high].iov_len = old_iov_len - size;
+               vecs_copy[entry_high].iov_base += size;
+
+               entry_low = entry_high;
+
+               if (err)
+                       break;
+
+               *retlen += retsize;
+               total_len -= wsize;
+
+               if (total_len == 0)
+                       break;
+
+               err = -EINVAL;
+               to = 0;
+       }
+
+       kfree(vecs_copy);
+       return err;
+}
+#endif
+
 static int
 concat_read_oob(struct mtd_info *mtd, loff_t from, struct mtd_oob_ops *ops)
 {
@@ -204,7 +301,7 @@ concat_write_oob(struct mtd_info *mtd, loff_t to, struct mtd_oob_ops *ops)
        if (!(mtd->flags & MTD_WRITEABLE))
                return -EROFS;
 
-       ops->retlen = 0;
+       ops->retlen = ops->oobretlen = 0;
 
        for (i = 0; i < concat->num_subdev; i++) {
                struct mtd_info *subdev = concat->subdev[i];
@@ -219,7 +316,7 @@ concat_write_oob(struct mtd_info *mtd, loff_t to, struct mtd_oob_ops *ops)
                        devops.len = subdev->size - to;
 
                err = mtd_write_oob(subdev, to, &devops);
-               ops->retlen += devops.retlen;
+               ops->retlen += devops.oobretlen;
                if (err)
                        return err;
 
@@ -243,6 +340,9 @@ concat_write_oob(struct mtd_info *mtd, loff_t to, struct mtd_oob_ops *ops)
 static void concat_erase_callback(struct erase_info *instr)
 {
        /* Nothing to do here in U-Boot */
+#ifndef __UBOOT__
+       wake_up((wait_queue_head_t *) instr->priv);
+#endif
 }
 
 static int concat_dev_erase(struct mtd_info *mtd, struct erase_info *erase)
@@ -316,7 +416,7 @@ static int concat_erase(struct mtd_info *mtd, struct erase_info *instr)
                 * to-be-erased area begins. Verify that the starting
                 * offset is aligned to this region's erase size:
                 */
-               if (instr->addr & (erase_regions[i].erasesize - 1))
+               if (i < 0 || instr->addr & (erase_regions[i].erasesize - 1))
                        return -EINVAL;
 
                /*
@@ -329,8 +429,8 @@ static int concat_erase(struct mtd_info *mtd, struct erase_info *instr)
                /*
                 * check if the ending offset is aligned to this region's erase size
                 */
-               if ((instr->addr + instr->len) & (erase_regions[i].erasesize -
-                                                 1))
+               if (i < 0 || ((instr->addr + instr->len) &
+                                       (erase_regions[i].erasesize - 1)))
                        return -EINVAL;
        }
 
@@ -422,7 +522,6 @@ static int concat_lock(struct mtd_info *mtd, loff_t ofs, uint64_t len)
                        size = len;
 
                err = mtd_lock(subdev, ofs, size);
-
                if (err)
                        break;
 
@@ -457,7 +556,6 @@ static int concat_unlock(struct mtd_info *mtd, loff_t ofs, uint64_t len)
                        size = len;
 
                err = mtd_unlock(subdev, ofs, size);
-
                if (err)
                        break;
 
@@ -483,6 +581,32 @@ static void concat_sync(struct mtd_info *mtd)
        }
 }
 
+#ifndef __UBOOT__
+static int concat_suspend(struct mtd_info *mtd)
+{
+       struct mtd_concat *concat = CONCAT(mtd);
+       int i, rc = 0;
+
+       for (i = 0; i < concat->num_subdev; i++) {
+               struct mtd_info *subdev = concat->subdev[i];
+               if ((rc = mtd_suspend(subdev)) < 0)
+                       return rc;
+       }
+       return rc;
+}
+
+static void concat_resume(struct mtd_info *mtd)
+{
+       struct mtd_concat *concat = CONCAT(mtd);
+       int i;
+
+       for (i = 0; i < concat->num_subdev; i++) {
+               struct mtd_info *subdev = concat->subdev[i];
+               mtd_resume(subdev);
+       }
+}
+#endif
+
 static int concat_block_isbad(struct mtd_info *mtd, loff_t ofs)
 {
        struct mtd_concat *concat = CONCAT(mtd);
@@ -511,9 +635,6 @@ static int concat_block_markbad(struct mtd_info *mtd, loff_t ofs)
        struct mtd_concat *concat = CONCAT(mtd);
        int i, err = -EINVAL;
 
-       if (!mtd_can_have_bb(concat->subdev[0]))
-               return 0;
-
        for (i = 0; i < concat->num_subdev; i++) {
                struct mtd_info *subdev = concat->subdev[i];
 
@@ -531,6 +652,32 @@ static int concat_block_markbad(struct mtd_info *mtd, loff_t ofs)
        return err;
 }
 
+/*
+ * try to support NOMMU mmaps on concatenated devices
+ * - we don't support subdev spanning as we can't guarantee it'll work
+ */
+static unsigned long concat_get_unmapped_area(struct mtd_info *mtd,
+                                             unsigned long len,
+                                             unsigned long offset,
+                                             unsigned long flags)
+{
+       struct mtd_concat *concat = CONCAT(mtd);
+       int i;
+
+       for (i = 0; i < concat->num_subdev; i++) {
+               struct mtd_info *subdev = concat->subdev[i];
+
+               if (offset >= subdev->size) {
+                       offset -= subdev->size;
+                       continue;
+               }
+
+               return mtd_get_unmapped_area(subdev, len, offset, flags);
+       }
+
+       return (unsigned long) -ENOSYS;
+}
+
 /*
  * This function constructs a virtual MTD device by concatenating
  * num_devs MTD devices. A pointer to the new device object is
@@ -539,17 +686,22 @@ static int concat_block_markbad(struct mtd_info *mtd, loff_t ofs)
  */
 struct mtd_info *mtd_concat_create(struct mtd_info *subdev[],  /* subdevices to concatenate */
                                   int num_devs,        /* number of subdevices      */
+#ifndef __UBOOT__
                                   const char *name)
+#else
+                                  char *name)
+#endif
 {                              /* name for the new device   */
        int i;
        size_t size;
        struct mtd_concat *concat;
        uint32_t max_erasesize, curr_erasesize;
        int num_erase_region;
+       int max_writebufsize = 0;
 
        debug("Concatenating MTD devices:\n");
        for (i = 0; i < num_devs; i++)
-               debug("(%d): \"%s\"\n", i, subdev[i]->name);
+               printk(KERN_NOTICE "(%d): \"%s\"\n", i, subdev[i]->name);
        debug("into device \"%s\"\n", name);
 
        /* allocate the device structure */
@@ -565,16 +717,26 @@ struct mtd_info *mtd_concat_create(struct mtd_info *subdev[],     /* subdevices to c
 
        /*
         * Set up the new "super" device's MTD object structure, check for
-        * incompatibilites between the subdevices.
+        * incompatibilities between the subdevices.
         */
        concat->mtd.type = subdev[0]->type;
        concat->mtd.flags = subdev[0]->flags;
        concat->mtd.size = subdev[0]->size;
        concat->mtd.erasesize = subdev[0]->erasesize;
        concat->mtd.writesize = subdev[0]->writesize;
+
+       for (i = 0; i < num_devs; i++)
+               if (max_writebufsize < subdev[i]->writebufsize)
+                       max_writebufsize = subdev[i]->writebufsize;
+       concat->mtd.writebufsize = max_writebufsize;
+
        concat->mtd.subpage_sft = subdev[0]->subpage_sft;
        concat->mtd.oobsize = subdev[0]->oobsize;
        concat->mtd.oobavail = subdev[0]->oobavail;
+#ifndef __UBOOT__
+       if (subdev[0]->_writev)
+               concat->mtd._writev = concat_writev;
+#endif
        if (subdev[0]->_read_oob)
                concat->mtd._read_oob = concat_read_oob;
        if (subdev[0]->_write_oob)
@@ -586,6 +748,10 @@ struct mtd_info *mtd_concat_create(struct mtd_info *subdev[],      /* subdevices to c
 
        concat->mtd.ecc_stats.badblocks = subdev[0]->ecc_stats.badblocks;
 
+#ifndef __UBOOT__
+       concat->mtd.backing_dev_info = subdev[0]->backing_dev_info;
+#endif
+
        concat->subdev[0] = subdev[0];
 
        for (i = 1; i < num_devs; i++) {
@@ -613,6 +779,16 @@ struct mtd_info *mtd_concat_create(struct mtd_info *subdev[],      /* subdevices to c
                                    subdev[i]->flags & MTD_WRITEABLE;
                }
 
+#ifndef __UBOOT__
+               /* only permit direct mapping if the BDIs are all the same
+                * - copy-mapping is still permitted
+                */
+               if (concat->mtd.backing_dev_info !=
+                   subdev[i]->backing_dev_info)
+                       concat->mtd.backing_dev_info =
+                               &default_backing_dev_info;
+#endif
+
                concat->mtd.size += subdev[i]->size;
                concat->mtd.ecc_stats.badblocks +=
                        subdev[i]->ecc_stats.badblocks;
@@ -641,6 +817,11 @@ struct mtd_info *mtd_concat_create(struct mtd_info *subdev[],      /* subdevices to c
        concat->mtd._sync = concat_sync;
        concat->mtd._lock = concat_lock;
        concat->mtd._unlock = concat_unlock;
+#ifndef __UBOOT__
+       concat->mtd._suspend = concat_suspend;
+       concat->mtd._resume = concat_resume;
+#endif
+       concat->mtd._get_unmapped_area = concat_get_unmapped_area;
 
        /*
         * Combine the erase block size info of the subdevices:
@@ -771,3 +952,22 @@ struct mtd_info *mtd_concat_create(struct mtd_info *subdev[],      /* subdevices to c
 
        return &concat->mtd;
 }
+
+/*
+ * This function destroys an MTD object obtained from concat_mtd_devs()
+ */
+
+void mtd_concat_destroy(struct mtd_info *mtd)
+{
+       struct mtd_concat *concat = CONCAT(mtd);
+       if (concat->mtd.numeraseregions)
+               kfree(concat->mtd.eraseregions);
+       kfree(concat);
+}
+
+EXPORT_SYMBOL(mtd_concat_create);
+EXPORT_SYMBOL(mtd_concat_destroy);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Robert Kaiser <rkaiser@sysgo.de>");
+MODULE_DESCRIPTION("Generic support for concatenating of MTD devices");
index 0a38fbef1414e9c03669bd0391c650bc43ee1dda..796ac0729e3a3f84545736f09f849ae18f7d0984 100644 (file)
  * Core registration and callback routines for MTD
  * drivers and users.
  *
- * This program is free software; you can redistribute it and/or modify
- * it under the terms of the GNU General Public License version 2 as
- * published by the Free Software Foundation.
+ * Copyright © 1999-2010 David Woodhouse <dwmw2@infradead.org>
+ * Copyright © 2006      Red Hat UK Limited 
+ *
+ * SPDX-License-Identifier:    GPL-2.0+
+ *
  */
 
-#include <linux/mtd/mtd.h>
+#define __UBOOT__
+#ifndef __UBOOT__
+#include <linux/module.h>
+#include <linux/kernel.h>
+#include <linux/ptrace.h>
+#include <linux/seq_file.h>
+#include <linux/string.h>
+#include <linux/timer.h>
+#include <linux/major.h>
+#include <linux/fs.h>
+#include <linux/err.h>
+#include <linux/ioctl.h>
+#include <linux/init.h>
+#include <linux/proc_fs.h>
+#include <linux/idr.h>
+#include <linux/backing-dev.h>
+#include <linux/gfp.h>
+#include <linux/slab.h>
+#else
 #include <linux/compat.h>
+#include <linux/err.h>
 #include <ubi_uboot.h>
+#endif
+
+#include <linux/mtd/mtd.h>
+#include <linux/mtd/partitions.h>
+
+#include "mtdcore.h"
+
+#ifndef __UBOOT__
+/*
+ * backing device capabilities for non-mappable devices (such as NAND flash)
+ * - permits private mappings, copies are taken of the data
+ */
+static struct backing_dev_info mtd_bdi_unmappable = {
+       .capabilities   = BDI_CAP_MAP_COPY,
+};
+
+/*
+ * backing device capabilities for R/O mappable devices (such as ROM)
+ * - permits private mappings, copies are taken of the data
+ * - permits non-writable shared mappings
+ */
+static struct backing_dev_info mtd_bdi_ro_mappable = {
+       .capabilities   = (BDI_CAP_MAP_COPY | BDI_CAP_MAP_DIRECT |
+                          BDI_CAP_EXEC_MAP | BDI_CAP_READ_MAP),
+};
+
+/*
+ * backing device capabilities for writable mappable devices (such as RAM)
+ * - permits private mappings, copies are taken of the data
+ * - permits non-writable shared mappings
+ */
+static struct backing_dev_info mtd_bdi_rw_mappable = {
+       .capabilities   = (BDI_CAP_MAP_COPY | BDI_CAP_MAP_DIRECT |
+                          BDI_CAP_EXEC_MAP | BDI_CAP_READ_MAP |
+                          BDI_CAP_WRITE_MAP),
+};
+
+static int mtd_cls_suspend(struct device *dev, pm_message_t state);
+static int mtd_cls_resume(struct device *dev);
 
+static struct class mtd_class = {
+       .name = "mtd",
+       .owner = THIS_MODULE,
+       .suspend = mtd_cls_suspend,
+       .resume = mtd_cls_resume,
+};
+#else
 struct mtd_info *mtd_table[MAX_MTD_DEVICES];
 
+#define MAX_IDR_ID     64
+
+struct idr_layer {
+       int     used;
+       void    *ptr;
+};
+
+struct idr {
+       struct idr_layer id[MAX_IDR_ID];
+};
+
+#define DEFINE_IDR(name)       struct idr name;
+
+void idr_remove(struct idr *idp, int id)
+{
+       if (idp->id[id].used)
+               idp->id[id].used = 0;
+
+       return;
+}
+void *idr_find(struct idr *idp, int id)
+{
+       if (idp->id[id].used)
+               return idp->id[id].ptr;
+
+       return NULL;
+}
+
+void *idr_get_next(struct idr *idp, int *next)
+{
+       void *ret;
+       int id = *next;
+
+       ret = idr_find(idp, id);
+       if (ret) {
+               id ++;
+               if (!idp->id[id].used)
+                       id = 0;
+               *next = id;
+       } else {
+               *next = 0;
+       }
+       
+       return ret;
+}
+
+int idr_alloc(struct idr *idp, void *ptr, int start, int end, gfp_t gfp_mask)
+{
+       struct idr_layer *idl;
+       int i = 0;
+
+       while (i < MAX_IDR_ID) {
+               idl = &idp->id[i];
+               if (idl->used == 0) {
+                       idl->used = 1;
+                       idl->ptr = ptr;
+                       return i;
+               }
+               i++;
+       }
+       return -ENOSPC;
+}
+#endif
+
+static DEFINE_IDR(mtd_idr);
+
+/* These are exported solely for the purpose of mtd_blkdevs.c. You
+   should not use them for _anything_ else */
+DEFINE_MUTEX(mtd_table_mutex);
+EXPORT_SYMBOL_GPL(mtd_table_mutex);
+
+struct mtd_info *__mtd_next_device(int i)
+{
+       return idr_get_next(&mtd_idr, &i);
+}
+EXPORT_SYMBOL_GPL(__mtd_next_device);
+
+#ifndef __UBOOT__
+static LIST_HEAD(mtd_notifiers);
+
+
+#define MTD_DEVT(index) MKDEV(MTD_CHAR_MAJOR, (index)*2)
+
+/* REVISIT once MTD uses the driver model better, whoever allocates
+ * the mtd_info will probably want to use the release() hook...
+ */
+static void mtd_release(struct device *dev)
+{
+       struct mtd_info __maybe_unused *mtd = dev_get_drvdata(dev);
+       dev_t index = MTD_DEVT(mtd->index);
+
+       /* remove /dev/mtdXro node if needed */
+       if (index)
+               device_destroy(&mtd_class, index + 1);
+}
+
+static int mtd_cls_suspend(struct device *dev, pm_message_t state)
+{
+       struct mtd_info *mtd = dev_get_drvdata(dev);
+
+       return mtd ? mtd_suspend(mtd) : 0;
+}
+
+static int mtd_cls_resume(struct device *dev)
+{
+       struct mtd_info *mtd = dev_get_drvdata(dev);
+
+       if (mtd)
+               mtd_resume(mtd);
+       return 0;
+}
+
+static ssize_t mtd_type_show(struct device *dev,
+               struct device_attribute *attr, char *buf)
+{
+       struct mtd_info *mtd = dev_get_drvdata(dev);
+       char *type;
+
+       switch (mtd->type) {
+       case MTD_ABSENT:
+               type = "absent";
+               break;
+       case MTD_RAM:
+               type = "ram";
+               break;
+       case MTD_ROM:
+               type = "rom";
+               break;
+       case MTD_NORFLASH:
+               type = "nor";
+               break;
+       case MTD_NANDFLASH:
+               type = "nand";
+               break;
+       case MTD_DATAFLASH:
+               type = "dataflash";
+               break;
+       case MTD_UBIVOLUME:
+               type = "ubi";
+               break;
+       case MTD_MLCNANDFLASH:
+               type = "mlc-nand";
+               break;
+       default:
+               type = "unknown";
+       }
+
+       return snprintf(buf, PAGE_SIZE, "%s\n", type);
+}
+static DEVICE_ATTR(type, S_IRUGO, mtd_type_show, NULL);
+
+static ssize_t mtd_flags_show(struct device *dev,
+               struct device_attribute *attr, char *buf)
+{
+       struct mtd_info *mtd = dev_get_drvdata(dev);
+
+       return snprintf(buf, PAGE_SIZE, "0x%lx\n", (unsigned long)mtd->flags);
+
+}
+static DEVICE_ATTR(flags, S_IRUGO, mtd_flags_show, NULL);
+
+static ssize_t mtd_size_show(struct device *dev,
+               struct device_attribute *attr, char *buf)
+{
+       struct mtd_info *mtd = dev_get_drvdata(dev);
+
+       return snprintf(buf, PAGE_SIZE, "%llu\n",
+               (unsigned long long)mtd->size);
+
+}
+static DEVICE_ATTR(size, S_IRUGO, mtd_size_show, NULL);
+
+static ssize_t mtd_erasesize_show(struct device *dev,
+               struct device_attribute *attr, char *buf)
+{
+       struct mtd_info *mtd = dev_get_drvdata(dev);
+
+       return snprintf(buf, PAGE_SIZE, "%lu\n", (unsigned long)mtd->erasesize);
+
+}
+static DEVICE_ATTR(erasesize, S_IRUGO, mtd_erasesize_show, NULL);
+
+static ssize_t mtd_writesize_show(struct device *dev,
+               struct device_attribute *attr, char *buf)
+{
+       struct mtd_info *mtd = dev_get_drvdata(dev);
+
+       return snprintf(buf, PAGE_SIZE, "%lu\n", (unsigned long)mtd->writesize);
+
+}
+static DEVICE_ATTR(writesize, S_IRUGO, mtd_writesize_show, NULL);
+
+static ssize_t mtd_subpagesize_show(struct device *dev,
+               struct device_attribute *attr, char *buf)
+{
+       struct mtd_info *mtd = dev_get_drvdata(dev);
+       unsigned int subpagesize = mtd->writesize >> mtd->subpage_sft;
+
+       return snprintf(buf, PAGE_SIZE, "%u\n", subpagesize);
+
+}
+static DEVICE_ATTR(subpagesize, S_IRUGO, mtd_subpagesize_show, NULL);
+
+static ssize_t mtd_oobsize_show(struct device *dev,
+               struct device_attribute *attr, char *buf)
+{
+       struct mtd_info *mtd = dev_get_drvdata(dev);
+
+       return snprintf(buf, PAGE_SIZE, "%lu\n", (unsigned long)mtd->oobsize);
+
+}
+static DEVICE_ATTR(oobsize, S_IRUGO, mtd_oobsize_show, NULL);
+
+static ssize_t mtd_numeraseregions_show(struct device *dev,
+               struct device_attribute *attr, char *buf)
+{
+       struct mtd_info *mtd = dev_get_drvdata(dev);
+
+       return snprintf(buf, PAGE_SIZE, "%u\n", mtd->numeraseregions);
+
+}
+static DEVICE_ATTR(numeraseregions, S_IRUGO, mtd_numeraseregions_show,
+       NULL);
+
+static ssize_t mtd_name_show(struct device *dev,
+               struct device_attribute *attr, char *buf)
+{
+       struct mtd_info *mtd = dev_get_drvdata(dev);
+
+       return snprintf(buf, PAGE_SIZE, "%s\n", mtd->name);
+
+}
+static DEVICE_ATTR(name, S_IRUGO, mtd_name_show, NULL);
+
+static ssize_t mtd_ecc_strength_show(struct device *dev,
+                                    struct device_attribute *attr, char *buf)
+{
+       struct mtd_info *mtd = dev_get_drvdata(dev);
+
+       return snprintf(buf, PAGE_SIZE, "%u\n", mtd->ecc_strength);
+}
+static DEVICE_ATTR(ecc_strength, S_IRUGO, mtd_ecc_strength_show, NULL);
+
+static ssize_t mtd_bitflip_threshold_show(struct device *dev,
+                                         struct device_attribute *attr,
+                                         char *buf)
+{
+       struct mtd_info *mtd = dev_get_drvdata(dev);
+
+       return snprintf(buf, PAGE_SIZE, "%u\n", mtd->bitflip_threshold);
+}
+
+static ssize_t mtd_bitflip_threshold_store(struct device *dev,
+                                          struct device_attribute *attr,
+                                          const char *buf, size_t count)
+{
+       struct mtd_info *mtd = dev_get_drvdata(dev);
+       unsigned int bitflip_threshold;
+       int retval;
+
+       retval = kstrtouint(buf, 0, &bitflip_threshold);
+       if (retval)
+               return retval;
+
+       mtd->bitflip_threshold = bitflip_threshold;
+       return count;
+}
+static DEVICE_ATTR(bitflip_threshold, S_IRUGO | S_IWUSR,
+                  mtd_bitflip_threshold_show,
+                  mtd_bitflip_threshold_store);
+
+static ssize_t mtd_ecc_step_size_show(struct device *dev,
+               struct device_attribute *attr, char *buf)
+{
+       struct mtd_info *mtd = dev_get_drvdata(dev);
+
+       return snprintf(buf, PAGE_SIZE, "%u\n", mtd->ecc_step_size);
+
+}
+static DEVICE_ATTR(ecc_step_size, S_IRUGO, mtd_ecc_step_size_show, NULL);
+
+static struct attribute *mtd_attrs[] = {
+       &dev_attr_type.attr,
+       &dev_attr_flags.attr,
+       &dev_attr_size.attr,
+       &dev_attr_erasesize.attr,
+       &dev_attr_writesize.attr,
+       &dev_attr_subpagesize.attr,
+       &dev_attr_oobsize.attr,
+       &dev_attr_numeraseregions.attr,
+       &dev_attr_name.attr,
+       &dev_attr_ecc_strength.attr,
+       &dev_attr_ecc_step_size.attr,
+       &dev_attr_bitflip_threshold.attr,
+       NULL,
+};
+ATTRIBUTE_GROUPS(mtd);
+
+static struct device_type mtd_devtype = {
+       .name           = "mtd",
+       .groups         = mtd_groups,
+       .release        = mtd_release,
+};
+#endif
+
+/**
+ *     add_mtd_device - register an MTD device
+ *     @mtd: pointer to new MTD device info structure
+ *
+ *     Add a device to the list of MTD devices present in the system, and
+ *     notify each currently active MTD 'user' of its arrival. Returns
+ *     zero on success or 1 on failure, which currently will only happen
+ *     if there is insufficient memory or a sysfs error.
+ */
+
 int add_mtd_device(struct mtd_info *mtd)
 {
-       int i;
+#ifndef __UBOOT__
+       struct mtd_notifier *not;
+#endif
+       int i, error;
+
+#ifndef __UBOOT__
+       if (!mtd->backing_dev_info) {
+               switch (mtd->type) {
+               case MTD_RAM:
+                       mtd->backing_dev_info = &mtd_bdi_rw_mappable;
+                       break;
+               case MTD_ROM:
+                       mtd->backing_dev_info = &mtd_bdi_ro_mappable;
+                       break;
+               default:
+                       mtd->backing_dev_info = &mtd_bdi_unmappable;
+                       break;
+               }
+       }
+#endif
 
        BUG_ON(mtd->writesize == 0);
+       mutex_lock(&mtd_table_mutex);
 
-       for (i = 0; i < MAX_MTD_DEVICES; i++)
-               if (!mtd_table[i]) {
-                       mtd_table[i] = mtd;
-                       mtd->index = i;
-                       mtd->usecount = 0;
+       i = idr_alloc(&mtd_idr, mtd, 0, 0, GFP_KERNEL);
+       if (i < 0)
+               goto fail_locked;
 
-                       /* default value if not set by driver */
-                       if (mtd->bitflip_threshold == 0)
-                               mtd->bitflip_threshold = mtd->ecc_strength;
+       mtd->index = i;
+       mtd->usecount = 0;
 
+       /* default value if not set by driver */
+       if (mtd->bitflip_threshold == 0)
+               mtd->bitflip_threshold = mtd->ecc_strength;
 
-                       /* No need to get a refcount on the module containing
-                          the notifier, since we hold the mtd_table_mutex */
+       if (is_power_of_2(mtd->erasesize))
+               mtd->erasesize_shift = ffs(mtd->erasesize) - 1;
+       else
+               mtd->erasesize_shift = 0;
 
-                       /* We _know_ we aren't being removed, because
-                          our caller is still holding us here. So none
-                          of this try_ nonsense, and no bitching about it
-                          either. :) */
-                       return 0;
-               }
+       if (is_power_of_2(mtd->writesize))
+               mtd->writesize_shift = ffs(mtd->writesize) - 1;
+       else
+               mtd->writesize_shift = 0;
+
+       mtd->erasesize_mask = (1 << mtd->erasesize_shift) - 1;
+       mtd->writesize_mask = (1 << mtd->writesize_shift) - 1;
+
+       /* Some chips always power up locked. Unlock them now */
+       if ((mtd->flags & MTD_WRITEABLE) && (mtd->flags & MTD_POWERUP_LOCK)) {
+               error = mtd_unlock(mtd, 0, mtd->size);
+               if (error && error != -EOPNOTSUPP)
+                       printk(KERN_WARNING
+                              "%s: unlock failed, writes may not work\n",
+                              mtd->name);
+       }
+
+#ifndef __UBOOT__
+       /* Caller should have set dev.parent to match the
+        * physical device.
+        */
+       mtd->dev.type = &mtd_devtype;
+       mtd->dev.class = &mtd_class;
+       mtd->dev.devt = MTD_DEVT(i);
+       dev_set_name(&mtd->dev, "mtd%d", i);
+       dev_set_drvdata(&mtd->dev, mtd);
+       if (device_register(&mtd->dev) != 0)
+               goto fail_added;
 
+       if (MTD_DEVT(i))
+               device_create(&mtd_class, mtd->dev.parent,
+                             MTD_DEVT(i) + 1,
+                             NULL, "mtd%dro", i);
+
+       pr_debug("mtd: Giving out device %d to %s\n", i, mtd->name);
+       /* No need to get a refcount on the module containing
+          the notifier, since we hold the mtd_table_mutex */
+       list_for_each_entry(not, &mtd_notifiers, list)
+               not->add(mtd);
+#endif
+
+       mutex_unlock(&mtd_table_mutex);
+       /* We _know_ we aren't being removed, because
+          our caller is still holding us here. So none
+          of this try_ nonsense, and no bitching about it
+          either. :) */
+       __module_get(THIS_MODULE);
+       return 0;
+
+#ifndef __UBOOT__
+fail_added:
+       idr_remove(&mtd_idr, i);
+#endif
+fail_locked:
+       mutex_unlock(&mtd_table_mutex);
        return 1;
 }
 
 /**
- *      del_mtd_device - unregister an MTD device
- *      @mtd: pointer to MTD device info structure
+ *     del_mtd_device - unregister an MTD device
+ *     @mtd: pointer to MTD device info structure
  *
- *      Remove a device from the list of MTD devices present in the system,
- *      and notify each currently active MTD 'user' of its departure.
- *      Returns zero on success or 1 on failure, which currently will happen
- *      if the requested device does not appear to be present in the list.
+ *     Remove a device from the list of MTD devices present in the system,
+ *     and notify each currently active MTD 'user' of its departure.
+ *     Returns zero on success or 1 on failure, which currently will happen
+ *     if the requested device does not appear to be present in the list.
  */
+
 int del_mtd_device(struct mtd_info *mtd)
 {
        int ret;
+#ifndef __UBOOT__
+       struct mtd_notifier *not;
+#endif
+
+       mutex_lock(&mtd_table_mutex);
 
-       if (mtd_table[mtd->index] != mtd) {
+       if (idr_find(&mtd_idr, mtd->index) != mtd) {
                ret = -ENODEV;
-       } else if (mtd->usecount) {
-               printk(KERN_NOTICE "Removing MTD device #%d (%s)"
-                               " with use count %d\n",
-                               mtd->index, mtd->name, mtd->usecount);
+               goto out_error;
+       }
+
+#ifndef __UBOOT__
+       /* No need to get a refcount on the module containing
+               the notifier, since we hold the mtd_table_mutex */
+       list_for_each_entry(not, &mtd_notifiers, list)
+               not->remove(mtd);
+#endif
+
+       if (mtd->usecount) {
+               printk(KERN_NOTICE "Removing MTD device #%d (%s) with use count %d\n",
+                      mtd->index, mtd->name, mtd->usecount);
                ret = -EBUSY;
        } else {
-               /* No need to get a refcount on the module containing
-                * the notifier, since we hold the mtd_table_mutex */
-               mtd_table[mtd->index] = NULL;
+#ifndef __UBOOT__
+               device_unregister(&mtd->dev);
+#endif
+
+               idr_remove(&mtd_idr, mtd->index);
 
+               module_put(THIS_MODULE);
                ret = 0;
        }
 
+out_error:
+       mutex_unlock(&mtd_table_mutex);
        return ret;
 }
 
+#ifndef __UBOOT__
+/**
+ * mtd_device_parse_register - parse partitions and register an MTD device.
+ *
+ * @mtd: the MTD device to register
+ * @types: the list of MTD partition probes to try, see
+ *         'parse_mtd_partitions()' for more information
+ * @parser_data: MTD partition parser-specific data
+ * @parts: fallback partition information to register, if parsing fails;
+ *         only valid if %nr_parts > %0
+ * @nr_parts: the number of partitions in parts, if zero then the full
+ *            MTD device is registered if no partition info is found
+ *
+ * This function aggregates MTD partitions parsing (done by
+ * 'parse_mtd_partitions()') and MTD device and partitions registering. It
+ * basically follows the most common pattern found in many MTD drivers:
+ *
+ * * It first tries to probe partitions on MTD device @mtd using parsers
+ *   specified in @types (if @types is %NULL, then the default list of parsers
+ *   is used, see 'parse_mtd_partitions()' for more information). If none are
+ *   found this functions tries to fallback to information specified in
+ *   @parts/@nr_parts.
+ * * If any partitioning info was found, this function registers the found
+ *   partitions.
+ * * If no partitions were found this function just registers the MTD device
+ *   @mtd and exits.
+ *
+ * Returns zero in case of success and a negative error code in case of failure.
+ */
+int mtd_device_parse_register(struct mtd_info *mtd, const char * const *types,
+                             struct mtd_part_parser_data *parser_data,
+                             const struct mtd_partition *parts,
+                             int nr_parts)
+{
+       int err;
+       struct mtd_partition *real_parts;
+
+       err = parse_mtd_partitions(mtd, types, &real_parts, parser_data);
+       if (err <= 0 && nr_parts && parts) {
+               real_parts = kmemdup(parts, sizeof(*parts) * nr_parts,
+                                    GFP_KERNEL);
+               if (!real_parts)
+                       err = -ENOMEM;
+               else
+                       err = nr_parts;
+       }
+
+       if (err > 0) {
+               err = add_mtd_partitions(mtd, real_parts, err);
+               kfree(real_parts);
+       } else if (err == 0) {
+               err = add_mtd_device(mtd);
+               if (err == 1)
+                       err = -ENODEV;
+       }
+
+       return err;
+}
+EXPORT_SYMBOL_GPL(mtd_device_parse_register);
+
+/**
+ * mtd_device_unregister - unregister an existing MTD device.
+ *
+ * @master: the MTD device to unregister.  This will unregister both the master
+ *          and any partitions if registered.
+ */
+int mtd_device_unregister(struct mtd_info *master)
+{
+       int err;
+
+       err = del_mtd_partitions(master);
+       if (err)
+               return err;
+
+       if (!device_is_registered(&master->dev))
+               return 0;
+
+       return del_mtd_device(master);
+}
+EXPORT_SYMBOL_GPL(mtd_device_unregister);
+
+/**
+ *     register_mtd_user - register a 'user' of MTD devices.
+ *     @new: pointer to notifier info structure
+ *
+ *     Registers a pair of callbacks function to be called upon addition
+ *     or removal of MTD devices. Causes the 'add' callback to be immediately
+ *     invoked for each MTD device currently present in the system.
+ */
+void register_mtd_user (struct mtd_notifier *new)
+{
+       struct mtd_info *mtd;
+
+       mutex_lock(&mtd_table_mutex);
+
+       list_add(&new->list, &mtd_notifiers);
+
+       __module_get(THIS_MODULE);
+
+       mtd_for_each_device(mtd)
+               new->add(mtd);
+
+       mutex_unlock(&mtd_table_mutex);
+}
+EXPORT_SYMBOL_GPL(register_mtd_user);
+
+/**
+ *     unregister_mtd_user - unregister a 'user' of MTD devices.
+ *     @old: pointer to notifier info structure
+ *
+ *     Removes a callback function pair from the list of 'users' to be
+ *     notified upon addition or removal of MTD devices. Causes the
+ *     'remove' callback to be immediately invoked for each MTD device
+ *     currently present in the system.
+ */
+int unregister_mtd_user (struct mtd_notifier *old)
+{
+       struct mtd_info *mtd;
+
+       mutex_lock(&mtd_table_mutex);
+
+       module_put(THIS_MODULE);
+
+       mtd_for_each_device(mtd)
+               old->remove(mtd);
+
+       list_del(&old->list);
+       mutex_unlock(&mtd_table_mutex);
+       return 0;
+}
+EXPORT_SYMBOL_GPL(unregister_mtd_user);
+#endif
+
 /**
  *     get_mtd_device - obtain a validated handle for an MTD device
  *     @mtd: last known address of the required MTD device
  *     @num: internal device number of the required MTD device
  *
  *     Given a number and NULL address, return the num'th entry in the device
- *      table, if any.  Given an address and num == -1, search the device table
- *      for a device with that address and return if it's still present. Given
- *      both, return the num'th driver only if its address matches. Return
- *      error code if not.
+ *     table, if any.  Given an address and num == -1, search the device table
+ *     for a device with that address and return if it's still present. Given
+ *     both, return the num'th driver only if its address matches. Return
+ *     error code if not.
  */
 struct mtd_info *get_mtd_device(struct mtd_info *mtd, int num)
 {
-       struct mtd_info *ret = NULL;
-       int i, err = -ENODEV;
+       struct mtd_info *ret = NULL, *other;
+       int err = -ENODEV;
+
+       mutex_lock(&mtd_table_mutex);
 
        if (num == -1) {
-               for (i = 0; i < MAX_MTD_DEVICES; i++)
-                       if (mtd_table[i] == mtd)
-                               ret = mtd_table[i];
-       } else if (num < MAX_MTD_DEVICES) {
-               ret = mtd_table[num];
+               mtd_for_each_device(other) {
+                       if (other == mtd) {
+                               ret = mtd;
+                               break;
+                       }
+               }
+       } else if (num >= 0) {
+               ret = idr_find(&mtd_idr, num);
                if (mtd && mtd != ret)
                        ret = NULL;
        }
 
-       if (!ret)
-               goto out_unlock;
+       if (!ret) {
+               ret = ERR_PTR(err);
+               goto out;
+       }
 
-       ret->usecount++;
+       err = __get_mtd_device(ret);
+       if (err)
+               ret = ERR_PTR(err);
+out:
+       mutex_unlock(&mtd_table_mutex);
        return ret;
+}
+EXPORT_SYMBOL_GPL(get_mtd_device);
 
-out_unlock:
-       return ERR_PTR(err);
+
+int __get_mtd_device(struct mtd_info *mtd)
+{
+       int err;
+
+       if (!try_module_get(mtd->owner))
+               return -ENODEV;
+
+       if (mtd->_get_device) {
+               err = mtd->_get_device(mtd);
+
+               if (err) {
+                       module_put(mtd->owner);
+                       return err;
+               }
+       }
+       mtd->usecount++;
+       return 0;
 }
+EXPORT_SYMBOL_GPL(__get_mtd_device);
 
 /**
- *      get_mtd_device_nm - obtain a validated handle for an MTD device by
- *      device name
- *      @name: MTD device name to open
+ *     get_mtd_device_nm - obtain a validated handle for an MTD device by
+ *     device name
+ *     @name: MTD device name to open
  *
- *      This function returns MTD device description structure in case of
- *      success and an error code in case of failure.
+ *     This function returns MTD device description structure in case of
+ *     success and an error code in case of failure.
  */
 struct mtd_info *get_mtd_device_nm(const char *name)
 {
-       int i, err = -ENODEV;
-       struct mtd_info *mtd = NULL;
+       int err = -ENODEV;
+       struct mtd_info *mtd = NULL, *other;
+
+       mutex_lock(&mtd_table_mutex);
 
-       for (i = 0; i < MAX_MTD_DEVICES; i++) {
-               if (mtd_table[i] && !strcmp(name, mtd_table[i]->name)) {
-                       mtd = mtd_table[i];
+       mtd_for_each_device(other) {
+               if (!strcmp(name, other->name)) {
+                       mtd = other;
                        break;
                }
        }
@@ -133,20 +770,18 @@ struct mtd_info *get_mtd_device_nm(const char *name)
        if (!mtd)
                goto out_unlock;
 
-       mtd->usecount++;
+       err = __get_mtd_device(mtd);
+       if (err)
+               goto out_unlock;
+
+       mutex_unlock(&mtd_table_mutex);
        return mtd;
 
 out_unlock:
+       mutex_unlock(&mtd_table_mutex);
        return ERR_PTR(err);
 }
-
-void put_mtd_device(struct mtd_info *mtd)
-{
-       int c;
-
-       c = --mtd->usecount;
-       BUG_ON(c < 0);
-}
+EXPORT_SYMBOL_GPL(get_mtd_device_nm);
 
 #if defined(CONFIG_CMD_MTDPARTS_SPREAD)
 /**
@@ -192,7 +827,28 @@ void mtd_get_len_incl_bad(struct mtd_info *mtd, uint64_t offset,
 }
 #endif /* defined(CONFIG_CMD_MTDPARTS_SPREAD) */
 
- /*
+void put_mtd_device(struct mtd_info *mtd)
+{
+       mutex_lock(&mtd_table_mutex);
+       __put_mtd_device(mtd);
+       mutex_unlock(&mtd_table_mutex);
+
+}
+EXPORT_SYMBOL_GPL(put_mtd_device);
+
+void __put_mtd_device(struct mtd_info *mtd)
+{
+       --mtd->usecount;
+       BUG_ON(mtd->usecount < 0);
+
+       if (mtd->_put_device)
+               mtd->_put_device(mtd);
+
+       module_put(mtd->owner);
+}
+EXPORT_SYMBOL_GPL(__put_mtd_device);
+
+/*
  * Erase is an asynchronous operation.  Device drivers are supposed
  * to call instr->callback() whenever the operation completes, even
  * if it completes with a failure.
@@ -213,11 +869,64 @@ int mtd_erase(struct mtd_info *mtd, struct erase_info *instr)
        }
        return mtd->_erase(mtd, instr);
 }
+EXPORT_SYMBOL_GPL(mtd_erase);
+
+#ifndef __UBOOT__
+/*
+ * This stuff for eXecute-In-Place. phys is optional and may be set to NULL.
+ */
+int mtd_point(struct mtd_info *mtd, loff_t from, size_t len, size_t *retlen,
+             void **virt, resource_size_t *phys)
+{
+       *retlen = 0;
+       *virt = NULL;
+       if (phys)
+               *phys = 0;
+       if (!mtd->_point)
+               return -EOPNOTSUPP;
+       if (from < 0 || from > mtd->size || len > mtd->size - from)
+               return -EINVAL;
+       if (!len)
+               return 0;
+       return mtd->_point(mtd, from, len, retlen, virt, phys);
+}
+EXPORT_SYMBOL_GPL(mtd_point);
+
+/* We probably shouldn't allow XIP if the unpoint isn't a NULL */
+int mtd_unpoint(struct mtd_info *mtd, loff_t from, size_t len)
+{
+       if (!mtd->_point)
+               return -EOPNOTSUPP;
+       if (from < 0 || from > mtd->size || len > mtd->size - from)
+               return -EINVAL;
+       if (!len)
+               return 0;
+       return mtd->_unpoint(mtd, from, len);
+}
+EXPORT_SYMBOL_GPL(mtd_unpoint);
+#endif
+
+/*
+ * Allow NOMMU mmap() to directly map the device (if not NULL)
+ * - return the address to which the offset maps
+ * - return -ENOSYS to indicate refusal to do the mapping
+ */
+unsigned long mtd_get_unmapped_area(struct mtd_info *mtd, unsigned long len,
+                                   unsigned long offset, unsigned long flags)
+{
+       if (!mtd->_get_unmapped_area)
+               return -EOPNOTSUPP;
+       if (offset > mtd->size || len > mtd->size - offset)
+               return -EINVAL;
+       return mtd->_get_unmapped_area(mtd, len, offset, flags);
+}
+EXPORT_SYMBOL_GPL(mtd_get_unmapped_area);
 
 int mtd_read(struct mtd_info *mtd, loff_t from, size_t len, size_t *retlen,
             u_char *buf)
 {
        int ret_code;
+       *retlen = 0;
        if (from < 0 || from > mtd->size || len > mtd->size - from)
                return -EINVAL;
        if (!len)
@@ -235,6 +944,7 @@ int mtd_read(struct mtd_info *mtd, loff_t from, size_t len, size_t *retlen,
                return 0;       /* device lacks ecc */
        return ret_code >= mtd->bitflip_threshold ? -EUCLEAN : 0;
 }
+EXPORT_SYMBOL_GPL(mtd_read);
 
 int mtd_write(struct mtd_info *mtd, loff_t to, size_t len, size_t *retlen,
              const u_char *buf)
@@ -248,6 +958,7 @@ int mtd_write(struct mtd_info *mtd, loff_t to, size_t len, size_t *retlen,
                return 0;
        return mtd->_write(mtd, to, len, retlen, buf);
 }
+EXPORT_SYMBOL_GPL(mtd_write);
 
 /*
  * In blackbox flight recorder like scenarios we want to make successful writes
@@ -270,14 +981,28 @@ int mtd_panic_write(struct mtd_info *mtd, loff_t to, size_t len, size_t *retlen,
                return 0;
        return mtd->_panic_write(mtd, to, len, retlen, buf);
 }
+EXPORT_SYMBOL_GPL(mtd_panic_write);
 
 int mtd_read_oob(struct mtd_info *mtd, loff_t from, struct mtd_oob_ops *ops)
 {
+       int ret_code;
        ops->retlen = ops->oobretlen = 0;
        if (!mtd->_read_oob)
                return -EOPNOTSUPP;
-       return mtd->_read_oob(mtd, from, ops);
+       /*
+        * In cases where ops->datbuf != NULL, mtd->_read_oob() has semantics
+        * similar to mtd->_read(), returning a non-negative integer
+        * representing max bitflips. In other cases, mtd->_read_oob() may
+        * return -EUCLEAN. In all cases, perform similar logic to mtd_read().
+        */
+       ret_code = mtd->_read_oob(mtd, from, ops);
+       if (unlikely(ret_code < 0))
+               return ret_code;
+       if (mtd->ecc_strength == 0)
+               return 0;       /* device lacks ecc */
+       return ret_code >= mtd->bitflip_threshold ? -EUCLEAN : 0;
 }
+EXPORT_SYMBOL_GPL(mtd_read_oob);
 
 /*
  * Method to access the protection register area, present in some flash
@@ -293,6 +1018,7 @@ int mtd_get_fact_prot_info(struct mtd_info *mtd, struct otp_info *buf,
                return 0;
        return mtd->_get_fact_prot_info(mtd, buf, len);
 }
+EXPORT_SYMBOL_GPL(mtd_get_fact_prot_info);
 
 int mtd_read_fact_prot_reg(struct mtd_info *mtd, loff_t from, size_t len,
                           size_t *retlen, u_char *buf)
@@ -304,6 +1030,7 @@ int mtd_read_fact_prot_reg(struct mtd_info *mtd, loff_t from, size_t len,
                return 0;
        return mtd->_read_fact_prot_reg(mtd, from, len, retlen, buf);
 }
+EXPORT_SYMBOL_GPL(mtd_read_fact_prot_reg);
 
 int mtd_get_user_prot_info(struct mtd_info *mtd, struct otp_info *buf,
                           size_t len)
@@ -314,6 +1041,7 @@ int mtd_get_user_prot_info(struct mtd_info *mtd, struct otp_info *buf,
                return 0;
        return mtd->_get_user_prot_info(mtd, buf, len);
 }
+EXPORT_SYMBOL_GPL(mtd_get_user_prot_info);
 
 int mtd_read_user_prot_reg(struct mtd_info *mtd, loff_t from, size_t len,
                           size_t *retlen, u_char *buf)
@@ -325,6 +1053,7 @@ int mtd_read_user_prot_reg(struct mtd_info *mtd, loff_t from, size_t len,
                return 0;
        return mtd->_read_user_prot_reg(mtd, from, len, retlen, buf);
 }
+EXPORT_SYMBOL_GPL(mtd_read_user_prot_reg);
 
 int mtd_write_user_prot_reg(struct mtd_info *mtd, loff_t to, size_t len,
                            size_t *retlen, u_char *buf)
@@ -336,6 +1065,7 @@ int mtd_write_user_prot_reg(struct mtd_info *mtd, loff_t to, size_t len,
                return 0;
        return mtd->_write_user_prot_reg(mtd, to, len, retlen, buf);
 }
+EXPORT_SYMBOL_GPL(mtd_write_user_prot_reg);
 
 int mtd_lock_user_prot_reg(struct mtd_info *mtd, loff_t from, size_t len)
 {
@@ -345,6 +1075,7 @@ int mtd_lock_user_prot_reg(struct mtd_info *mtd, loff_t from, size_t len)
                return 0;
        return mtd->_lock_user_prot_reg(mtd, from, len);
 }
+EXPORT_SYMBOL_GPL(mtd_lock_user_prot_reg);
 
 /* Chip-supported device locking */
 int mtd_lock(struct mtd_info *mtd, loff_t ofs, uint64_t len)
@@ -357,6 +1088,7 @@ int mtd_lock(struct mtd_info *mtd, loff_t ofs, uint64_t len)
                return 0;
        return mtd->_lock(mtd, ofs, len);
 }
+EXPORT_SYMBOL_GPL(mtd_lock);
 
 int mtd_unlock(struct mtd_info *mtd, loff_t ofs, uint64_t len)
 {
@@ -368,6 +1100,19 @@ int mtd_unlock(struct mtd_info *mtd, loff_t ofs, uint64_t len)
                return 0;
        return mtd->_unlock(mtd, ofs, len);
 }
+EXPORT_SYMBOL_GPL(mtd_unlock);
+
+int mtd_is_locked(struct mtd_info *mtd, loff_t ofs, uint64_t len)
+{
+       if (!mtd->_is_locked)
+               return -EOPNOTSUPP;
+       if (ofs < 0 || ofs > mtd->size || len > mtd->size - ofs)
+               return -EINVAL;
+       if (!len)
+               return 0;
+       return mtd->_is_locked(mtd, ofs, len);
+}
+EXPORT_SYMBOL_GPL(mtd_is_locked);
 
 int mtd_block_isbad(struct mtd_info *mtd, loff_t ofs)
 {
@@ -377,6 +1122,7 @@ int mtd_block_isbad(struct mtd_info *mtd, loff_t ofs)
                return -EINVAL;
        return mtd->_block_isbad(mtd, ofs);
 }
+EXPORT_SYMBOL_GPL(mtd_block_isbad);
 
 int mtd_block_markbad(struct mtd_info *mtd, loff_t ofs)
 {
@@ -388,3 +1134,225 @@ int mtd_block_markbad(struct mtd_info *mtd, loff_t ofs)
                return -EROFS;
        return mtd->_block_markbad(mtd, ofs);
 }
+EXPORT_SYMBOL_GPL(mtd_block_markbad);
+
+#ifndef __UBOOT__
+/*
+ * default_mtd_writev - the default writev method
+ * @mtd: mtd device description object pointer
+ * @vecs: the vectors to write
+ * @count: count of vectors in @vecs
+ * @to: the MTD device offset to write to
+ * @retlen: on exit contains the count of bytes written to the MTD device.
+ *
+ * This function returns zero in case of success and a negative error code in
+ * case of failure.
+ */
+static int default_mtd_writev(struct mtd_info *mtd, const struct kvec *vecs,
+                             unsigned long count, loff_t to, size_t *retlen)
+{
+       unsigned long i;
+       size_t totlen = 0, thislen;
+       int ret = 0;
+
+       for (i = 0; i < count; i++) {
+               if (!vecs[i].iov_len)
+                       continue;
+               ret = mtd_write(mtd, to, vecs[i].iov_len, &thislen,
+                               vecs[i].iov_base);
+               totlen += thislen;
+               if (ret || thislen != vecs[i].iov_len)
+                       break;
+               to += vecs[i].iov_len;
+       }
+       *retlen = totlen;
+       return ret;
+}
+
+/*
+ * mtd_writev - the vector-based MTD write method
+ * @mtd: mtd device description object pointer
+ * @vecs: the vectors to write
+ * @count: count of vectors in @vecs
+ * @to: the MTD device offset to write to
+ * @retlen: on exit contains the count of bytes written to the MTD device.
+ *
+ * This function returns zero in case of success and a negative error code in
+ * case of failure.
+ */
+int mtd_writev(struct mtd_info *mtd, const struct kvec *vecs,
+              unsigned long count, loff_t to, size_t *retlen)
+{
+       *retlen = 0;
+       if (!(mtd->flags & MTD_WRITEABLE))
+               return -EROFS;
+       if (!mtd->_writev)
+               return default_mtd_writev(mtd, vecs, count, to, retlen);
+       return mtd->_writev(mtd, vecs, count, to, retlen);
+}
+EXPORT_SYMBOL_GPL(mtd_writev);
+
+/**
+ * mtd_kmalloc_up_to - allocate a contiguous buffer up to the specified size
+ * @mtd: mtd device description object pointer
+ * @size: a pointer to the ideal or maximum size of the allocation, points
+ *        to the actual allocation size on success.
+ *
+ * This routine attempts to allocate a contiguous kernel buffer up to
+ * the specified size, backing off the size of the request exponentially
+ * until the request succeeds or until the allocation size falls below
+ * the system page size. This attempts to make sure it does not adversely
+ * impact system performance, so when allocating more than one page, we
+ * ask the memory allocator to avoid re-trying, swapping, writing back
+ * or performing I/O.
+ *
+ * Note, this function also makes sure that the allocated buffer is aligned to
+ * the MTD device's min. I/O unit, i.e. the "mtd->writesize" value.
+ *
+ * This is called, for example by mtd_{read,write} and jffs2_scan_medium,
+ * to handle smaller (i.e. degraded) buffer allocations under low- or
+ * fragmented-memory situations where such reduced allocations, from a
+ * requested ideal, are allowed.
+ *
+ * Returns a pointer to the allocated buffer on success; otherwise, NULL.
+ */
+void *mtd_kmalloc_up_to(const struct mtd_info *mtd, size_t *size)
+{
+       gfp_t flags = __GFP_NOWARN | __GFP_WAIT |
+                      __GFP_NORETRY | __GFP_NO_KSWAPD;
+       size_t min_alloc = max_t(size_t, mtd->writesize, PAGE_SIZE);
+       void *kbuf;
+
+       *size = min_t(size_t, *size, KMALLOC_MAX_SIZE);
+
+       while (*size > min_alloc) {
+               kbuf = kmalloc(*size, flags);
+               if (kbuf)
+                       return kbuf;
+
+               *size >>= 1;
+               *size = ALIGN(*size, mtd->writesize);
+       }
+
+       /*
+        * For the last resort allocation allow 'kmalloc()' to do all sorts of
+        * things (write-back, dropping caches, etc) by using GFP_KERNEL.
+        */
+       return kmalloc(*size, GFP_KERNEL);
+}
+EXPORT_SYMBOL_GPL(mtd_kmalloc_up_to);
+#endif
+
+#ifdef CONFIG_PROC_FS
+
+/*====================================================================*/
+/* Support for /proc/mtd */
+
+static int mtd_proc_show(struct seq_file *m, void *v)
+{
+       struct mtd_info *mtd;
+
+       seq_puts(m, "dev:    size   erasesize  name\n");
+       mutex_lock(&mtd_table_mutex);
+       mtd_for_each_device(mtd) {
+               seq_printf(m, "mtd%d: %8.8llx %8.8x \"%s\"\n",
+                          mtd->index, (unsigned long long)mtd->size,
+                          mtd->erasesize, mtd->name);
+       }
+       mutex_unlock(&mtd_table_mutex);
+       return 0;
+}
+
+static int mtd_proc_open(struct inode *inode, struct file *file)
+{
+       return single_open(file, mtd_proc_show, NULL);
+}
+
+static const struct file_operations mtd_proc_ops = {
+       .open           = mtd_proc_open,
+       .read           = seq_read,
+       .llseek         = seq_lseek,
+       .release        = single_release,
+};
+#endif /* CONFIG_PROC_FS */
+
+/*====================================================================*/
+/* Init code */
+
+#ifndef __UBOOT__
+static int __init mtd_bdi_init(struct backing_dev_info *bdi, const char *name)
+{
+       int ret;
+
+       ret = bdi_init(bdi);
+       if (!ret)
+               ret = bdi_register(bdi, NULL, "%s", name);
+
+       if (ret)
+               bdi_destroy(bdi);
+
+       return ret;
+}
+
+static struct proc_dir_entry *proc_mtd;
+
+static int __init init_mtd(void)
+{
+       int ret;
+
+       ret = class_register(&mtd_class);
+       if (ret)
+               goto err_reg;
+
+       ret = mtd_bdi_init(&mtd_bdi_unmappable, "mtd-unmap");
+       if (ret)
+               goto err_bdi1;
+
+       ret = mtd_bdi_init(&mtd_bdi_ro_mappable, "mtd-romap");
+       if (ret)
+               goto err_bdi2;
+
+       ret = mtd_bdi_init(&mtd_bdi_rw_mappable, "mtd-rwmap");
+       if (ret)
+               goto err_bdi3;
+
+       proc_mtd = proc_create("mtd", 0, NULL, &mtd_proc_ops);
+
+       ret = init_mtdchar();
+       if (ret)
+               goto out_procfs;
+
+       return 0;
+
+out_procfs:
+       if (proc_mtd)
+               remove_proc_entry("mtd", NULL);
+err_bdi3:
+       bdi_destroy(&mtd_bdi_ro_mappable);
+err_bdi2:
+       bdi_destroy(&mtd_bdi_unmappable);
+err_bdi1:
+       class_unregister(&mtd_class);
+err_reg:
+       pr_err("Error registering mtd class or bdi: %d\n", ret);
+       return ret;
+}
+
+static void __exit cleanup_mtd(void)
+{
+       cleanup_mtdchar();
+       if (proc_mtd)
+               remove_proc_entry("mtd", NULL);
+       class_unregister(&mtd_class);
+       bdi_destroy(&mtd_bdi_unmappable);
+       bdi_destroy(&mtd_bdi_ro_mappable);
+       bdi_destroy(&mtd_bdi_rw_mappable);
+}
+
+module_init(init_mtd);
+module_exit(cleanup_mtd);
+#endif
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("David Woodhouse <dwmw2@infradead.org>");
+MODULE_DESCRIPTION("Core MTD registration and access routines");
diff --git a/drivers/mtd/mtdcore.h b/drivers/mtd/mtdcore.h
new file mode 100644 (file)
index 0000000..7b03533
--- /dev/null
@@ -0,0 +1,23 @@
+/*
+ * These are exported solely for the purpose of mtd_blkdevs.c and mtdchar.c.
+ * You should not use them for _anything_ else.
+ */
+
+extern struct mutex mtd_table_mutex;
+
+struct mtd_info *__mtd_next_device(int i);
+int add_mtd_device(struct mtd_info *mtd);
+int del_mtd_device(struct mtd_info *mtd);
+int add_mtd_partitions(struct mtd_info *, const struct mtd_partition *, int);
+int del_mtd_partitions(struct mtd_info *);
+int parse_mtd_partitions(struct mtd_info *master, const char * const *types,
+                        struct mtd_partition **pparts,
+                        struct mtd_part_parser_data *data);
+
+int __init init_mtdchar(void);
+void __exit cleanup_mtdchar(void);
+
+#define mtd_for_each_device(mtd)                       \
+       for ((mtd) = __mtd_next_device(0);              \
+            (mtd) != NULL;                             \
+            (mtd) = __mtd_next_device(mtd->index + 1))
index 146ce11eb1bafcfbb8f0add516c5675d0f166bc7..d20b8578e7644ce1cba2323ae489c6c4db7d3206 100644 (file)
@@ -1,35 +1,50 @@
 /*
  * Simple MTD partitioning layer
  *
- * (C) 2000 Nicolas Pitre <nico@cam.org>
+ * Copyright © 2000 Nicolas Pitre <nico@fluxnic.net>
+ * Copyright © 2002 Thomas Gleixner <gleixner@linutronix.de>
+ * Copyright © 2000-2010 David Woodhouse <dwmw2@infradead.org>
  *
- * This code is GPL
+ * SPDX-License-Identifier:    GPL-2.0+
  *
- *     02-21-2002      Thomas Gleixner <gleixner@autronix.de>
- *                     added support for read_oob, write_oob
  */
 
+#define __UBOOT__
+#ifndef __UBOOT__
+#include <linux/module.h>
+#include <linux/types.h>
+#include <linux/kernel.h>
+#include <linux/slab.h>
+#include <linux/list.h>
+#include <linux/kmod.h>
+#endif
+
 #include <common.h>
 #include <malloc.h>
 #include <asm/errno.h>
+#include <linux/compat.h>
+#include <ubi_uboot.h>
 
-#include <linux/types.h>
-#include <linux/list.h>
 #include <linux/mtd/mtd.h>
 #include <linux/mtd/partitions.h>
-#include <linux/compat.h>
+#include <linux/err.h>
+
+#include "mtdcore.h"
 
 /* Our partition linked list */
-struct list_head mtd_partitions;
+static LIST_HEAD(mtd_partitions);
+#ifndef __UBOOT__
+static DEFINE_MUTEX(mtd_partitions_mutex);
+#else
+DEFINE_MUTEX(mtd_partitions_mutex);
+#endif
 
 /* Our partition node structure */
 struct mtd_part {
        struct mtd_info mtd;
        struct mtd_info *master;
        uint64_t offset;
-       int index;
        struct list_head list;
-       int registered;
 };
 
 /*
@@ -39,6 +54,30 @@ struct mtd_part {
 #define PART(x)  ((struct mtd_part *)(x))
 
 
+#ifdef __UBOOT__
+/* from mm/util.c */
+
+/**
+ * kstrdup - allocate space for and copy an existing string
+ * @s: the string to duplicate
+ * @gfp: the GFP mask used in the kmalloc() call when allocating memory
+ */
+char *kstrdup(const char *s, gfp_t gfp)
+{
+       size_t len;
+       char *buf;
+
+       if (!s)
+               return NULL;
+
+       len = strlen(s) + 1;
+       buf = kmalloc(len, gfp);
+       if (buf)
+               memcpy(buf, s, len);
+       return buf;
+}
+#endif
+
 /*
  * MTD methods which simply translate the effective address and pass through
  * to the _real_ device.
@@ -52,7 +91,8 @@ static int part_read(struct mtd_info *mtd, loff_t from, size_t len,
        int res;
 
        stats = part->master->ecc_stats;
-       res = mtd_read(part->master, from + part->offset, len, retlen, buf);
+       res = part->master->_read(part->master, from + part->offset, len,
+                                 retlen, buf);
        if (unlikely(mtd_is_eccerr(res)))
                mtd->ecc_stats.failed +=
                        part->master->ecc_stats.failed - stats.failed;
@@ -62,6 +102,36 @@ static int part_read(struct mtd_info *mtd, loff_t from, size_t len,
        return res;
 }
 
+#ifndef __UBOOT__
+static int part_point(struct mtd_info *mtd, loff_t from, size_t len,
+               size_t *retlen, void **virt, resource_size_t *phys)
+{
+       struct mtd_part *part = PART(mtd);
+
+       return part->master->_point(part->master, from + part->offset, len,
+                                   retlen, virt, phys);
+}
+
+static int part_unpoint(struct mtd_info *mtd, loff_t from, size_t len)
+{
+       struct mtd_part *part = PART(mtd);
+
+       return part->master->_unpoint(part->master, from + part->offset, len);
+}
+#endif
+
+static unsigned long part_get_unmapped_area(struct mtd_info *mtd,
+                                           unsigned long len,
+                                           unsigned long offset,
+                                           unsigned long flags)
+{
+       struct mtd_part *part = PART(mtd);
+
+       offset += part->offset;
+       return part->master->_get_unmapped_area(part->master, len, offset,
+                                               flags);
+}
+
 static int part_read_oob(struct mtd_info *mtd, loff_t from,
                struct mtd_oob_ops *ops)
 {
@@ -72,8 +142,25 @@ static int part_read_oob(struct mtd_info *mtd, loff_t from,
                return -EINVAL;
        if (ops->datbuf && from + ops->len > mtd->size)
                return -EINVAL;
-       res = mtd_read_oob(part->master, from + part->offset, ops);
 
+       /*
+        * If OOB is also requested, make sure that we do not read past the end
+        * of this partition.
+        */
+       if (ops->oobbuf) {
+               size_t len, pages;
+
+               if (ops->mode == MTD_OPS_AUTO_OOB)
+                       len = mtd->oobavail;
+               else
+                       len = mtd->oobsize;
+               pages = mtd_div_by_ws(mtd->size, mtd);
+               pages -= mtd_div_by_ws(from, mtd);
+               if (ops->ooboffs + ops->ooblen > pages * len)
+                       return -EINVAL;
+       }
+
+       res = part->master->_read_oob(part->master, from + part->offset, ops);
        if (unlikely(res)) {
                if (mtd_is_bitflip(res))
                        mtd->ecc_stats.corrected++;
@@ -87,35 +174,46 @@ static int part_read_user_prot_reg(struct mtd_info *mtd, loff_t from,
                size_t len, size_t *retlen, u_char *buf)
 {
        struct mtd_part *part = PART(mtd);
-       return mtd_read_user_prot_reg(part->master, from, len, retlen, buf);
+       return part->master->_read_user_prot_reg(part->master, from, len,
+                                                retlen, buf);
 }
 
 static int part_get_user_prot_info(struct mtd_info *mtd,
                struct otp_info *buf, size_t len)
 {
        struct mtd_part *part = PART(mtd);
-       return mtd_get_user_prot_info(part->master, buf, len);
+       return part->master->_get_user_prot_info(part->master, buf, len);
 }
 
 static int part_read_fact_prot_reg(struct mtd_info *mtd, loff_t from,
                size_t len, size_t *retlen, u_char *buf)
 {
        struct mtd_part *part = PART(mtd);
-       return mtd_read_fact_prot_reg(part->master, from, len, retlen, buf);
+       return part->master->_read_fact_prot_reg(part->master, from, len,
+                                                retlen, buf);
 }
 
 static int part_get_fact_prot_info(struct mtd_info *mtd, struct otp_info *buf,
                size_t len)
 {
        struct mtd_part *part = PART(mtd);
-       return mtd_get_fact_prot_info(part->master, buf, len);
+       return part->master->_get_fact_prot_info(part->master, buf, len);
 }
 
 static int part_write(struct mtd_info *mtd, loff_t to, size_t len,
                size_t *retlen, const u_char *buf)
 {
        struct mtd_part *part = PART(mtd);
-       return mtd_write(part->master, to + part->offset, len, retlen, buf);
+       return part->master->_write(part->master, to + part->offset, len,
+                                   retlen, buf);
+}
+
+static int part_panic_write(struct mtd_info *mtd, loff_t to, size_t len,
+               size_t *retlen, const u_char *buf)
+{
+       struct mtd_part *part = PART(mtd);
+       return part->master->_panic_write(part->master, to + part->offset, len,
+                                         retlen, buf);
 }
 
 static int part_write_oob(struct mtd_info *mtd, loff_t to,
@@ -127,30 +225,41 @@ static int part_write_oob(struct mtd_info *mtd, loff_t to,
                return -EINVAL;
        if (ops->datbuf && to + ops->len > mtd->size)
                return -EINVAL;
-       return mtd_write_oob(part->master, to + part->offset, ops);
+       return part->master->_write_oob(part->master, to + part->offset, ops);
 }
 
 static int part_write_user_prot_reg(struct mtd_info *mtd, loff_t from,
                size_t len, size_t *retlen, u_char *buf)
 {
        struct mtd_part *part = PART(mtd);
-       return mtd_write_user_prot_reg(part->master, from, len, retlen, buf);
+       return part->master->_write_user_prot_reg(part->master, from, len,
+                                                 retlen, buf);
 }
 
 static int part_lock_user_prot_reg(struct mtd_info *mtd, loff_t from,
                size_t len)
 {
        struct mtd_part *part = PART(mtd);
-       return mtd_lock_user_prot_reg(part->master, from, len);
+       return part->master->_lock_user_prot_reg(part->master, from, len);
 }
 
+#ifndef __UBOOT__
+static int part_writev(struct mtd_info *mtd, const struct kvec *vecs,
+               unsigned long count, loff_t to, size_t *retlen)
+{
+       struct mtd_part *part = PART(mtd);
+       return part->master->_writev(part->master, vecs, count,
+                                    to + part->offset, retlen);
+}
+#endif
+
 static int part_erase(struct mtd_info *mtd, struct erase_info *instr)
 {
        struct mtd_part *part = PART(mtd);
        int ret;
 
        instr->addr += part->offset;
-       ret = mtd_erase(part->master, instr);
+       ret = part->master->_erase(part->master, instr);
        if (ret) {
                if (instr->fail_addr != MTD_FAIL_ADDR_UNKNOWN)
                        instr->fail_addr -= part->offset;
@@ -171,30 +280,51 @@ void mtd_erase_callback(struct erase_info *instr)
        if (instr->callback)
                instr->callback(instr);
 }
+EXPORT_SYMBOL_GPL(mtd_erase_callback);
 
 static int part_lock(struct mtd_info *mtd, loff_t ofs, uint64_t len)
 {
        struct mtd_part *part = PART(mtd);
-       return mtd_lock(part->master, ofs + part->offset, len);
+       return part->master->_lock(part->master, ofs + part->offset, len);
 }
 
 static int part_unlock(struct mtd_info *mtd, loff_t ofs, uint64_t len)
 {
        struct mtd_part *part = PART(mtd);
-       return mtd_unlock(part->master, ofs + part->offset, len);
+       return part->master->_unlock(part->master, ofs + part->offset, len);
+}
+
+static int part_is_locked(struct mtd_info *mtd, loff_t ofs, uint64_t len)
+{
+       struct mtd_part *part = PART(mtd);
+       return part->master->_is_locked(part->master, ofs + part->offset, len);
 }
 
 static void part_sync(struct mtd_info *mtd)
 {
        struct mtd_part *part = PART(mtd);
-       mtd_sync(part->master);
+       part->master->_sync(part->master);
 }
 
+#ifndef __UBOOT__
+static int part_suspend(struct mtd_info *mtd)
+{
+       struct mtd_part *part = PART(mtd);
+       return part->master->_suspend(part->master);
+}
+
+static void part_resume(struct mtd_info *mtd)
+{
+       struct mtd_part *part = PART(mtd);
+       part->master->_resume(part->master);
+}
+#endif
+
 static int part_block_isbad(struct mtd_info *mtd, loff_t ofs)
 {
        struct mtd_part *part = PART(mtd);
        ofs += part->offset;
-       return mtd_block_isbad(part->master, ofs);
+       return part->master->_block_isbad(part->master, ofs);
 }
 
 static int part_block_markbad(struct mtd_info *mtd, loff_t ofs)
@@ -203,12 +333,18 @@ static int part_block_markbad(struct mtd_info *mtd, loff_t ofs)
        int res;
 
        ofs += part->offset;
-       res = mtd_block_markbad(part->master, ofs);
+       res = part->master->_block_markbad(part->master, ofs);
        if (!res)
                mtd->ecc_stats.badblocks++;
        return res;
 }
 
+static inline void free_partition(struct mtd_part *p)
+{
+       kfree(p->mtd.name);
+       kfree(p);
+}
+
 /*
  * This function unregisters and destroy all slave MTD objects which are
  * attached to the given master MTD object.
@@ -217,49 +353,78 @@ static int part_block_markbad(struct mtd_info *mtd, loff_t ofs)
 int del_mtd_partitions(struct mtd_info *master)
 {
        struct mtd_part *slave, *next;
+       int ret, err = 0;
 
+       mutex_lock(&mtd_partitions_mutex);
        list_for_each_entry_safe(slave, next, &mtd_partitions, list)
                if (slave->master == master) {
+                       ret = del_mtd_device(&slave->mtd);
+                       if (ret < 0) {
+                               err = ret;
+                               continue;
+                       }
                        list_del(&slave->list);
-                       if (slave->registered)
-                               del_mtd_device(&slave->mtd);
-                       kfree(slave);
+                       free_partition(slave);
                }
+       mutex_unlock(&mtd_partitions_mutex);
 
-       return 0;
+       return err;
 }
 
-static struct mtd_part *add_one_partition(struct mtd_info *master,
-               const struct mtd_partition *part, int partno,
-               uint64_t cur_offset)
+static struct mtd_part *allocate_partition(struct mtd_info *master,
+                       const struct mtd_partition *part, int partno,
+                       uint64_t cur_offset)
 {
        struct mtd_part *slave;
+       char *name;
 
        /* allocate the partition structure */
        slave = kzalloc(sizeof(*slave), GFP_KERNEL);
-       if (!slave) {
+       name = kstrdup(part->name, GFP_KERNEL);
+       if (!name || !slave) {
                printk(KERN_ERR"memory allocation error while creating partitions for \"%s\"\n",
-                       master->name);
-               del_mtd_partitions(master);
-               return NULL;
+                      master->name);
+               kfree(name);
+               kfree(slave);
+               return ERR_PTR(-ENOMEM);
        }
-       list_add(&slave->list, &mtd_partitions);
 
        /* set up the MTD object for this partition */
        slave->mtd.type = master->type;
        slave->mtd.flags = master->flags & ~part->mask_flags;
        slave->mtd.size = part->size;
        slave->mtd.writesize = master->writesize;
+       slave->mtd.writebufsize = master->writebufsize;
        slave->mtd.oobsize = master->oobsize;
        slave->mtd.oobavail = master->oobavail;
        slave->mtd.subpage_sft = master->subpage_sft;
 
-       slave->mtd.name = part->name;
+       slave->mtd.name = name;
        slave->mtd.owner = master->owner;
+#ifndef __UBOOT__
+       slave->mtd.backing_dev_info = master->backing_dev_info;
+
+       /* NOTE:  we don't arrange MTDs as a tree; it'd be error-prone
+        * to have the same data be in two different partitions.
+        */
+       slave->mtd.dev.parent = master->dev.parent;
+#endif
 
        slave->mtd._read = part_read;
        slave->mtd._write = part_write;
 
+       if (master->_panic_write)
+               slave->mtd._panic_write = part_panic_write;
+
+#ifndef __UBOOT__
+       if (master->_point && master->_unpoint) {
+               slave->mtd._point = part_point;
+               slave->mtd._unpoint = part_unpoint;
+       }
+#endif
+
+       if (master->_get_unmapped_area)
+               slave->mtd._get_unmapped_area = part_get_unmapped_area;
        if (master->_read_oob)
                slave->mtd._read_oob = part_read_oob;
        if (master->_write_oob)
@@ -278,10 +443,21 @@ static struct mtd_part *add_one_partition(struct mtd_info *master,
                slave->mtd._get_fact_prot_info = part_get_fact_prot_info;
        if (master->_sync)
                slave->mtd._sync = part_sync;
+#ifndef __UBOOT__
+       if (!partno && !master->dev.class && master->_suspend &&
+           master->_resume) {
+                       slave->mtd._suspend = part_suspend;
+                       slave->mtd._resume = part_resume;
+       }
+       if (master->_writev)
+               slave->mtd._writev = part_writev;
+#endif
        if (master->_lock)
                slave->mtd._lock = part_lock;
        if (master->_unlock)
                slave->mtd._unlock = part_unlock;
+       if (master->_is_locked)
+               slave->mtd._is_locked = part_is_locked;
        if (master->_block_isbad)
                slave->mtd._block_isbad = part_block_isbad;
        if (master->_block_markbad)
@@ -289,7 +465,6 @@ static struct mtd_part *add_one_partition(struct mtd_info *master,
        slave->mtd._erase = part_erase;
        slave->master = master;
        slave->offset = part->offset;
-       slave->index = partno;
 
        if (slave->offset == MTDPART_OFS_APPEND)
                slave->offset = cur_offset;
@@ -298,18 +473,29 @@ static struct mtd_part *add_one_partition(struct mtd_info *master,
                if (mtd_mod_by_eb(cur_offset, master) != 0) {
                        /* Round up to next erasesize */
                        slave->offset = (mtd_div_by_eb(cur_offset, master) + 1) * master->erasesize;
-                       debug("Moving partition %d: 0x%012llx -> 0x%012llx\n",
-                             partno, (unsigned long long)cur_offset,
-                             (unsigned long long)slave->offset);
+                       debug("Moving partition %d: "
+                              "0x%012llx -> 0x%012llx\n", partno,
+                              (unsigned long long)cur_offset, (unsigned long long)slave->offset);
+               }
+       }
+       if (slave->offset == MTDPART_OFS_RETAIN) {
+               slave->offset = cur_offset;
+               if (master->size - slave->offset >= slave->mtd.size) {
+                       slave->mtd.size = master->size - slave->offset
+                                                       - slave->mtd.size;
+               } else {
+                       debug("mtd partition \"%s\" doesn't have enough space: %#llx < %#llx, disabled\n",
+                               part->name, master->size - slave->offset,
+                               slave->mtd.size);
+                       /* register to preserve ordering */
+                       goto out_register;
                }
        }
        if (slave->mtd.size == MTDPART_SIZ_FULL)
                slave->mtd.size = master->size - slave->offset;
 
-       debug("0x%012llx-0x%012llx : \"%s\"\n",
-             (unsigned long long)slave->offset,
-             (unsigned long long)(slave->offset + slave->mtd.size),
-             slave->mtd.name);
+       debug("0x%012llx-0x%012llx : \"%s\"\n", (unsigned long long)slave->offset,
+               (unsigned long long)(slave->offset + slave->mtd.size), slave->mtd.name);
 
        /* let's do some sanity checks */
        if (slave->offset >= master->size) {
@@ -336,7 +522,8 @@ static struct mtd_part *add_one_partition(struct mtd_info *master,
                for (i = 0; i < max && regions[i].offset <= slave->offset; i++)
                        ;
                /* The loop searched for the region _behind_ the first one */
-               i--;
+               if (i > 0)
+                       i--;
 
                /* Pick biggest erasesize */
                for (; i < max && regions[i].offset < end; i++) {
@@ -367,6 +554,10 @@ static struct mtd_part *add_one_partition(struct mtd_info *master,
        }
 
        slave->mtd.ecclayout = master->ecclayout;
+       slave->mtd.ecc_step_size = master->ecc_step_size;
+       slave->mtd.ecc_strength = master->ecc_strength;
+       slave->mtd.bitflip_threshold = master->bitflip_threshold;
+
        if (master->_block_isbad) {
                uint64_t offs = 0;
 
@@ -378,18 +569,89 @@ static struct mtd_part *add_one_partition(struct mtd_info *master,
        }
 
 out_register:
-       if (part->mtdp) {
-               /* store the object pointer (caller may or may not register it*/
-               *part->mtdp = &slave->mtd;
-               slave->registered = 0;
-       } else {
-               /* register our partition */
-               add_mtd_device(&slave->mtd);
-               slave->registered = 1;
-       }
        return slave;
 }
 
+int mtd_add_partition(struct mtd_info *master, const char *name,
+                     long long offset, long long length)
+{
+       struct mtd_partition part;
+       struct mtd_part *p, *new;
+       uint64_t start, end;
+       int ret = 0;
+
+       /* the direct offset is expected */
+       if (offset == MTDPART_OFS_APPEND ||
+           offset == MTDPART_OFS_NXTBLK)
+               return -EINVAL;
+
+       if (length == MTDPART_SIZ_FULL)
+               length = master->size - offset;
+
+       if (length <= 0)
+               return -EINVAL;
+
+       part.name = name;
+       part.size = length;
+       part.offset = offset;
+       part.mask_flags = 0;
+       part.ecclayout = NULL;
+
+       new = allocate_partition(master, &part, -1, offset);
+       if (IS_ERR(new))
+               return PTR_ERR(new);
+
+       start = offset;
+       end = offset + length;
+
+       mutex_lock(&mtd_partitions_mutex);
+       list_for_each_entry(p, &mtd_partitions, list)
+               if (p->master == master) {
+                       if ((start >= p->offset) &&
+                           (start < (p->offset + p->mtd.size)))
+                               goto err_inv;
+
+                       if ((end >= p->offset) &&
+                           (end < (p->offset + p->mtd.size)))
+                               goto err_inv;
+               }
+
+       list_add(&new->list, &mtd_partitions);
+       mutex_unlock(&mtd_partitions_mutex);
+
+       add_mtd_device(&new->mtd);
+
+       return ret;
+err_inv:
+       mutex_unlock(&mtd_partitions_mutex);
+       free_partition(new);
+       return -EINVAL;
+}
+EXPORT_SYMBOL_GPL(mtd_add_partition);
+
+int mtd_del_partition(struct mtd_info *master, int partno)
+{
+       struct mtd_part *slave, *next;
+       int ret = -EINVAL;
+
+       mutex_lock(&mtd_partitions_mutex);
+       list_for_each_entry_safe(slave, next, &mtd_partitions, list)
+               if ((slave->master == master) &&
+                   (slave->mtd.index == partno)) {
+                       ret = del_mtd_device(&slave->mtd);
+                       if (ret < 0)
+                               break;
+
+                       list_del(&slave->list);
+                       free_partition(slave);
+                       break;
+               }
+       mutex_unlock(&mtd_partitions_mutex);
+
+       return ret;
+}
+EXPORT_SYMBOL_GPL(mtd_del_partition);
+
 /*
  * This function, given a master MTD object and a partition table, creates
  * and registers slave MTD objects which are bound to the master according to
@@ -407,6 +669,7 @@ int add_mtd_partitions(struct mtd_info *master,
        uint64_t cur_offset = 0;
        int i;
 
+#ifdef __UBOOT__
        /*
         * Need to init the list here, since LIST_INIT() does not
         * work on platforms where relocation has problems (like MIPS
@@ -414,15 +677,147 @@ int add_mtd_partitions(struct mtd_info *master,
         */
        if (mtd_partitions.next == NULL)
                INIT_LIST_HEAD(&mtd_partitions);
+#endif
 
        debug("Creating %d MTD partitions on \"%s\":\n", nbparts, master->name);
 
        for (i = 0; i < nbparts; i++) {
-               slave = add_one_partition(master, parts + i, i, cur_offset);
-               if (!slave)
-                       return -ENOMEM;
+               slave = allocate_partition(master, parts + i, i, cur_offset);
+               if (IS_ERR(slave))
+                       return PTR_ERR(slave);
+
+               mutex_lock(&mtd_partitions_mutex);
+               list_add(&slave->list, &mtd_partitions);
+               mutex_unlock(&mtd_partitions_mutex);
+
+               add_mtd_device(&slave->mtd);
+
                cur_offset = slave->offset + slave->mtd.size;
        }
 
        return 0;
 }
+
+#ifndef __UBOOT__
+static DEFINE_SPINLOCK(part_parser_lock);
+static LIST_HEAD(part_parsers);
+
+static struct mtd_part_parser *get_partition_parser(const char *name)
+{
+       struct mtd_part_parser *p, *ret = NULL;
+
+       spin_lock(&part_parser_lock);
+
+       list_for_each_entry(p, &part_parsers, list)
+               if (!strcmp(p->name, name) && try_module_get(p->owner)) {
+                       ret = p;
+                       break;
+               }
+
+       spin_unlock(&part_parser_lock);
+
+       return ret;
+}
+
+#define put_partition_parser(p) do { module_put((p)->owner); } while (0)
+
+void register_mtd_parser(struct mtd_part_parser *p)
+{
+       spin_lock(&part_parser_lock);
+       list_add(&p->list, &part_parsers);
+       spin_unlock(&part_parser_lock);
+}
+EXPORT_SYMBOL_GPL(register_mtd_parser);
+
+void deregister_mtd_parser(struct mtd_part_parser *p)
+{
+       spin_lock(&part_parser_lock);
+       list_del(&p->list);
+       spin_unlock(&part_parser_lock);
+}
+EXPORT_SYMBOL_GPL(deregister_mtd_parser);
+
+/*
+ * Do not forget to update 'parse_mtd_partitions()' kerneldoc comment if you
+ * are changing this array!
+ */
+static const char * const default_mtd_part_types[] = {
+       "cmdlinepart",
+       "ofpart",
+       NULL
+};
+
+/**
+ * parse_mtd_partitions - parse MTD partitions
+ * @master: the master partition (describes whole MTD device)
+ * @types: names of partition parsers to try or %NULL
+ * @pparts: array of partitions found is returned here
+ * @data: MTD partition parser-specific data
+ *
+ * This function tries to find partition on MTD device @master. It uses MTD
+ * partition parsers, specified in @types. However, if @types is %NULL, then
+ * the default list of parsers is used. The default list contains only the
+ * "cmdlinepart" and "ofpart" parsers ATM.
+ * Note: If there are more then one parser in @types, the kernel only takes the
+ * partitions parsed out by the first parser.
+ *
+ * This function may return:
+ * o a negative error code in case of failure
+ * o zero if no partitions were found
+ * o a positive number of found partitions, in which case on exit @pparts will
+ *   point to an array containing this number of &struct mtd_info objects.
+ */
+int parse_mtd_partitions(struct mtd_info *master, const char *const *types,
+                        struct mtd_partition **pparts,
+                        struct mtd_part_parser_data *data)
+{
+       struct mtd_part_parser *parser;
+       int ret = 0;
+
+       if (!types)
+               types = default_mtd_part_types;
+
+       for ( ; ret <= 0 && *types; types++) {
+               parser = get_partition_parser(*types);
+               if (!parser && !request_module("%s", *types))
+                       parser = get_partition_parser(*types);
+               if (!parser)
+                       continue;
+               ret = (*parser->parse_fn)(master, pparts, data);
+               put_partition_parser(parser);
+               if (ret > 0) {
+                       printk(KERN_NOTICE "%d %s partitions found on MTD device %s\n",
+                              ret, parser->name, master->name);
+                       break;
+               }
+       }
+       return ret;
+}
+#endif
+
+int mtd_is_partition(const struct mtd_info *mtd)
+{
+       struct mtd_part *part;
+       int ispart = 0;
+
+       mutex_lock(&mtd_partitions_mutex);
+       list_for_each_entry(part, &mtd_partitions, list)
+               if (&part->mtd == mtd) {
+                       ispart = 1;
+                       break;
+               }
+       mutex_unlock(&mtd_partitions_mutex);
+
+       return ispart;
+}
+EXPORT_SYMBOL_GPL(mtd_is_partition);
+
+/* Returns the size of the entire flash chip */
+uint64_t mtd_get_device_size(const struct mtd_info *mtd)
+{
+       if (!mtd_is_partition(mtd))
+               return mtd->size;
+
+       return PART(mtd)->master->size;
+}
+EXPORT_SYMBOL_GPL(mtd_get_device_size);
index 2f31fc96ad2e4a0ed4f0e5db62365a5d5ca145d9..7e1e6ec78b20081dc4b48d586bc0ece55e1fcebf 100644 (file)
@@ -561,6 +561,7 @@ static void fsl_elbc_read_buf(struct mtd_info *mtd, u8 *buf, int len)
                       len, avail);
 }
 
+#if defined(CONFIG_MTD_NAND_VERIFY_WRITE)
 /*
  * Verify buffer against the FCM Controller Data Buffer
  */
@@ -593,6 +594,7 @@ static int fsl_elbc_verify_buf(struct mtd_info *mtd,
        ctrl->index += len;
        return i == len && ctrl->status == LTESR_CC ? 0 : -EIO;
 }
+#endif
 
 /* This function is called after Program and Erase Operations to
  * check for success or failure.
@@ -725,7 +727,9 @@ static int fsl_elbc_chip_init(int devnum, u8 *addr)
        nand->read_byte = fsl_elbc_read_byte;
        nand->write_buf = fsl_elbc_write_buf;
        nand->read_buf = fsl_elbc_read_buf;
+#if defined(CONFIG_MTD_NAND_VERIFY_WRITE)
        nand->verify_buf = fsl_elbc_verify_buf;
+#endif
        nand->select_chip = fsl_elbc_select_chip;
        nand->cmdfunc = fsl_elbc_cmdfunc;
        nand->waitfunc = fsl_elbc_wait;
index 8b453cb383c206f2e9acc4d29c425343a3703052..2f04c698d3b850142218c31989715e97d8b6b746 100644 (file)
@@ -684,6 +684,7 @@ static void fsl_ifc_read_buf(struct mtd_info *mtd, u8 *buf, int len)
                       __func__, len, avail);
 }
 
+#if defined(CONFIG_MTD_NAND_VERIFY_WRITE)
 /*
  * Verify buffer against the IFC Controller Data Buffer
  */
@@ -716,6 +717,7 @@ static int fsl_ifc_verify_buf(struct mtd_info *mtd,
        ctrl->index += len;
        return i == len && ctrl->status == IFC_NAND_EVTER_STAT_OPC ? 0 : -EIO;
 }
+#endif
 
 /* This function is called after Program and Erase Operations to
  * check for success or failure.
@@ -939,7 +941,9 @@ static int fsl_ifc_chip_init(int devnum, u8 *addr)
 
        nand->write_buf = fsl_ifc_write_buf;
        nand->read_buf = fsl_ifc_read_buf;
+#if defined(CONFIG_MTD_NAND_VERIFY_WRITE)
        nand->verify_buf = fsl_ifc_verify_buf;
+#endif
        nand->select_chip = fsl_ifc_select_chip;
        nand->cmdfunc = fsl_ifc_cmdfunc;
        nand->waitfunc = fsl_ifc_wait;
index 3ae0044f26e8ba42bbb429cc542e68a2e9fa4a45..65ce98ad5e332b8704344c990fa21e7da4d6827e 100644 (file)
@@ -153,6 +153,7 @@ static void upm_nand_read_buf(struct mtd_info *mtd, u_char *buf, int len)
                buf[i] = in_8(chip->IO_ADDR_R);
 }
 
+#if defined(CONFIG_MTD_NAND_VERIFY_WRITE)
 static int upm_nand_verify_buf(struct mtd_info *mtd, const u_char *buf, int len)
 {
        int i;
@@ -165,6 +166,7 @@ static int upm_nand_verify_buf(struct mtd_info *mtd, const u_char *buf, int len)
 
        return 0;
 }
+#endif
 
 static int nand_dev_ready(struct mtd_info *mtd)
 {
@@ -191,7 +193,9 @@ int fsl_upm_nand_init(struct nand_chip *chip, struct fsl_upm_nand *fun)
        chip->read_byte = upm_nand_read_byte;
        chip->read_buf = upm_nand_read_buf;
        chip->write_buf = upm_nand_write_buf;
+#if defined(CONFIG_MTD_NAND_VERIFY_WRITE)
        chip->verify_buf = upm_nand_verify_buf;
+#endif
        if (fun->dev_ready)
                chip->dev_ready = nand_dev_ready;
 
index d0f3a35329c275d60ff6bb9210e5bcf0f8b2e369..7233bfc12796c10d18ccf877c969fc8e5b953aaa 100644 (file)
@@ -459,6 +459,7 @@ static void mpc5121_nfc_write_buf(struct mtd_info *mtd,
        mpc5121_nfc_buf_copy(mtd, (u_char *) buf, len, 1);
 }
 
+#if defined(CONFIG_MTD_NAND_VERIFY_WRITE)
 /* Compare buffer with NAND flash */
 static int mpc5121_nfc_verify_buf(struct mtd_info *mtd,
                                  const u_char * buf, int len)
@@ -479,6 +480,7 @@ static int mpc5121_nfc_verify_buf(struct mtd_info *mtd,
 
        return 0;
 }
+#endif
 
 /* Read byte from NFC buffers */
 static u8 mpc5121_nfc_read_byte(struct mtd_info *mtd)
@@ -607,7 +609,9 @@ int board_nand_init(struct nand_chip *chip)
        chip->read_word = mpc5121_nfc_read_word;
        chip->read_buf = mpc5121_nfc_read_buf;
        chip->write_buf = mpc5121_nfc_write_buf;
+#if defined(CONFIG_MTD_NAND_VERIFY_WRITE)
        chip->verify_buf = mpc5121_nfc_verify_buf;
+#endif
        chip->select_chip = mpc5121_nfc_select_chip;
        chip->bbt_options = NAND_BBT_USE_FLASH;
        chip->ecc.mode = NAND_ECC_SOFT;
index ed0ca3aca85ae65546a3b6eaee84ae7b3d3e4c7a..2e5b5b9bf985ef59a77d3010aee0ad2992eb9cfe 100644 (file)
@@ -949,6 +949,8 @@ static void mxc_nand_read_buf(struct mtd_info *mtd, u_char *buf, int len)
        host->col_addr = col;
 }
 
+#ifdef __UBOOT__
+#if defined(CONFIG_MTD_NAND_VERIFY_WRITE)
 /*
  * Used by the upper layer to verify the data in NAND Flash
  * with the data in the buf.
@@ -972,6 +974,8 @@ static int mxc_nand_verify_buf(struct mtd_info *mtd,
 
        return 0;
 }
+#endif
+#endif
 
 /*
  * This function is used by upper layer for select and
@@ -1203,7 +1207,11 @@ int board_nand_init(struct nand_chip *this)
        this->read_word = mxc_nand_read_word;
        this->write_buf = mxc_nand_write_buf;
        this->read_buf = mxc_nand_read_buf;
+#ifdef __UBOOT__
+#if defined(CONFIG_MTD_NAND_VERIFY_WRITE)
        this->verify_buf = mxc_nand_verify_buf;
+#endif
+#endif
 
        host->regs = (struct mxc_nand_regs __iomem *)CONFIG_MXC_NAND_REGS_BASE;
 #ifdef MXC_NFC_V3_2
index 376976d5795bbdb95e5e680dbe7a8e8f818fa0ef..ae61cca440ef090021eeda0baa4d5f039ea385c1 100644 (file)
@@ -4,7 +4,6 @@
  *  Overview:
  *   This is the generic MTD driver for NAND flash devices. It should be
  *   capable of working with almost all NAND chips currently available.
- *   Basic support for AG-AND chips is provided.
  *
  *     Additional technical information is available on
  *     http://www.linux-mtd.infradead.org/doc/nand.html
@@ -22,8 +21,6 @@
  *     Enable cached programming for 2k page size chips
  *     Check, if mtd->ecctype should be set to MTD_ECC_HW
  *     if we have HW ECC support.
- *     The AG-AND chips have nice features for speed improvement,
- *     which are not supported yet. Read / program 4 pages in one go.
  *     BBT table is not serialized, has to be fixed
  *
  * This program is free software; you can redistribute it and/or modify
  *
  */
 
-#include <common.h>
-
-#define ENOTSUPP       524     /* Operation is not supported */
+#define __UBOOT__
+#ifndef __UBOOT__
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
 
+#include <linux/module.h>
+#include <linux/delay.h>
+#include <linux/errno.h>
+#include <linux/err.h>
+#include <linux/sched.h>
+#include <linux/slab.h>
+#include <linux/types.h>
+#include <linux/mtd/mtd.h>
+#include <linux/mtd/nand.h>
+#include <linux/mtd/nand_ecc.h>
+#include <linux/mtd/nand_bch.h>
+#include <linux/interrupt.h>
+#include <linux/bitops.h>
+#include <linux/leds.h>
+#include <linux/io.h>
+#include <linux/mtd/partitions.h>
+#else
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+#include <common.h>
 #include <malloc.h>
 #include <watchdog.h>
 #include <linux/err.h>
 #include <linux/mtd/nand.h>
 #include <linux/mtd/nand_ecc.h>
 #include <linux/mtd/nand_bch.h>
-
 #ifdef CONFIG_MTD_PARTITIONS
 #include <linux/mtd/partitions.h>
 #endif
-
 #include <asm/io.h>
 #include <asm/errno.h>
 
@@ -63,6 +77,9 @@
 #define CONFIG_SYS_NAND_RESET_CNT 200000
 #endif
 
+static bool is_module_text_address(unsigned long addr) {return 0;}
+#endif
+
 /* Define default oob placement schemes for large and small page devices */
 static struct nand_ecclayout nand_oob_8 = {
        .eccbytes = 3,
@@ -107,13 +124,16 @@ static struct nand_ecclayout nand_oob_128 = {
                 .length = 78} }
 };
 
-static int nand_get_device(struct nand_chip *chip, struct mtd_info *mtd,
-                          int new_state);
+static int nand_get_device(struct mtd_info *mtd, int new_state);
 
 static int nand_do_write_oob(struct mtd_info *mtd, loff_t to,
                             struct mtd_oob_ops *ops);
 
-static int nand_wait(struct mtd_info *mtd, struct nand_chip *this);
+/*
+ * For devices which display every fart in the system on a separate LED. Is
+ * compiled away when LED support is disabled.
+ */
+DEFINE_LED_TRIGGER(nand_led_trigger);
 
 static int check_offs_len(struct mtd_info *mtd,
                                        loff_t ofs, uint64_t len)
@@ -122,15 +142,14 @@ static int check_offs_len(struct mtd_info *mtd,
        int ret = 0;
 
        /* Start address must align on block boundary */
-       if (ofs & ((1 << chip->phys_erase_shift) - 1)) {
-               MTDDEBUG(MTD_DEBUG_LEVEL0, "%s: Unaligned address\n", __func__);
+       if (ofs & ((1ULL << chip->phys_erase_shift) - 1)) {
+               pr_debug("%s: unaligned address\n", __func__);
                ret = -EINVAL;
        }
 
        /* Length must align on block boundary */
-       if (len & ((1 << chip->phys_erase_shift) - 1)) {
-               MTDDEBUG(MTD_DEBUG_LEVEL0, "%s: Length not block aligned\n",
-                                       __func__);
+       if (len & ((1ULL << chip->phys_erase_shift) - 1)) {
+               pr_debug("%s: length not block aligned\n", __func__);
                ret = -EINVAL;
        }
 
@@ -141,30 +160,43 @@ static int check_offs_len(struct mtd_info *mtd,
  * nand_release_device - [GENERIC] release chip
  * @mtd: MTD device structure
  *
- * Deselect, release chip lock and wake up anyone waiting on the device.
+ * Release chip lock and wake up anyone waiting on the device.
  */
 static void nand_release_device(struct mtd_info *mtd)
 {
        struct nand_chip *chip = mtd->priv;
 
+#ifndef __UBOOT__
+       /* Release the controller and the chip */
+       spin_lock(&chip->controller->lock);
+       chip->controller->active = NULL;
+       chip->state = FL_READY;
+       wake_up(&chip->controller->wq);
+       spin_unlock(&chip->controller->lock);
+#else
        /* De-select the NAND device */
        chip->select_chip(mtd, -1);
+#endif
 }
 
 /**
  * nand_read_byte - [DEFAULT] read one byte from the chip
  * @mtd: MTD device structure
  *
- * Default read function for 8bit buswidth.
+ * Default read function for 8bit buswidth
  */
+#ifndef __UBOOT__
+static uint8_t nand_read_byte(struct mtd_info *mtd)
+#else
 uint8_t nand_read_byte(struct mtd_info *mtd)
+#endif
 {
        struct nand_chip *chip = mtd->priv;
        return readb(chip->IO_ADDR_R);
 }
 
 /**
- * nand_read_byte16 - [DEFAULT] read one byte endianess aware from the chip
+ * nand_read_byte16 - [DEFAULT] read one byte endianness aware from the chip
  * nand_read_byte16 - [DEFAULT] read one byte endianness aware from the chip
  * @mtd: MTD device structure
  *
@@ -212,6 +244,88 @@ static void nand_select_chip(struct mtd_info *mtd, int chipnr)
        }
 }
 
+/**
+ * nand_write_byte - [DEFAULT] write single byte to chip
+ * @mtd: MTD device structure
+ * @byte: value to write
+ *
+ * Default function to write a byte to I/O[7:0]
+ */
+static void nand_write_byte(struct mtd_info *mtd, uint8_t byte)
+{
+       struct nand_chip *chip = mtd->priv;
+
+       chip->write_buf(mtd, &byte, 1);
+}
+
+/**
+ * nand_write_byte16 - [DEFAULT] write single byte to a chip with width 16
+ * @mtd: MTD device structure
+ * @byte: value to write
+ *
+ * Default function to write a byte to I/O[7:0] on a 16-bit wide chip.
+ */
+static void nand_write_byte16(struct mtd_info *mtd, uint8_t byte)
+{
+       struct nand_chip *chip = mtd->priv;
+       uint16_t word = byte;
+
+       /*
+        * It's not entirely clear what should happen to I/O[15:8] when writing
+        * a byte. The ONFi spec (Revision 3.1; 2012-09-19, Section 2.16) reads:
+        *
+        *    When the host supports a 16-bit bus width, only data is
+        *    transferred at the 16-bit width. All address and command line
+        *    transfers shall use only the lower 8-bits of the data bus. During
+        *    command transfers, the host may place any value on the upper
+        *    8-bits of the data bus. During address transfers, the host shall
+        *    set the upper 8-bits of the data bus to 00h.
+        *
+        * One user of the write_byte callback is nand_onfi_set_features. The
+        * four parameters are specified to be written to I/O[7:0], but this is
+        * neither an address nor a command transfer. Let's assume a 0 on the
+        * upper I/O lines is OK.
+        */
+       chip->write_buf(mtd, (uint8_t *)&word, 2);
+}
+
+#if defined(__UBOOT__) && !defined(CONFIG_BLACKFIN)
+static void iowrite8_rep(void *addr, const uint8_t *buf, int len)
+{
+       int i;
+
+       for (i = 0; i < len; i++)
+               writeb(buf[i], addr);
+}
+static void ioread8_rep(void *addr, uint8_t *buf, int len)
+{
+       int i;
+
+       for (i = 0; i < len; i++)
+               buf[i] = readb(addr);
+}
+
+static void ioread16_rep(void *addr, void *buf, int len)
+{
+       int i;
+       u16 *p = (u16 *) buf;
+       len >>= 1;
+       for (i = 0; i < len; i++)
+               p[i] = readw(addr);
+}
+
+static void iowrite16_rep(void *addr, void *buf, int len)
+{
+       int i;
+        u16 *p = (u16 *) buf;
+        len >>= 1;
+
+        for (i = 0; i < len; i++)
+                writew(p[i], addr);
+}
+#endif
+
 /**
  * nand_write_buf - [DEFAULT] write buffer to chip
  * @mtd: MTD device structure
@@ -220,13 +334,15 @@ static void nand_select_chip(struct mtd_info *mtd, int chipnr)
  *
  * Default write function for 8bit buswidth.
  */
+#ifndef __UBOOT__
+static void nand_write_buf(struct mtd_info *mtd, const uint8_t *buf, int len)
+#else
 void nand_write_buf(struct mtd_info *mtd, const uint8_t *buf, int len)
+#endif
 {
-       int i;
        struct nand_chip *chip = mtd->priv;
 
-       for (i = 0; i < len; i++)
-               writeb(buf[i], chip->IO_ADDR_W);
+       iowrite8_rep(chip->IO_ADDR_W, buf, len);
 }
 
 /**
@@ -237,15 +353,19 @@ void nand_write_buf(struct mtd_info *mtd, const uint8_t *buf, int len)
  *
  * Default read function for 8bit buswidth.
  */
+#ifndef __UBOOT__
+static void nand_read_buf(struct mtd_info *mtd, uint8_t *buf, int len)
+#else
 void nand_read_buf(struct mtd_info *mtd, uint8_t *buf, int len)
+#endif
 {
-       int i;
        struct nand_chip *chip = mtd->priv;
 
-       for (i = 0; i < len; i++)
-               buf[i] = readb(chip->IO_ADDR_R);
+       ioread8_rep(chip->IO_ADDR_R, buf, len);
 }
 
+#ifdef __UBOOT__
+#if defined(CONFIG_MTD_NAND_VERIFY_WRITE)
 /**
  * nand_verify_buf - [DEFAULT] Verify chip data against buffer
  * @mtd: MTD device structure
@@ -266,14 +386,14 @@ static int nand_verify_buf(struct mtd_info *mtd, const uint8_t *buf, int len)
 }
 
 /**
- * nand_write_buf16 - [DEFAULT] write buffer to chip
+ * nand_verify_buf16 - [DEFAULT] Verify chip data against buffer
  * @mtd: MTD device structure
- * @buf: data buffer
- * @len: number of bytes to write
+ * @buf: buffer containing the data to compare
+ * @len: number of bytes to compare
  *
- * Default write function for 16bit buswidth.
+ * Default verify function for 16bit buswidth.
  */
-void nand_write_buf16(struct mtd_info *mtd, const uint8_t *buf, int len)
+static int nand_verify_buf16(struct mtd_info *mtd, const uint8_t *buf, int len)
 {
        int i;
        struct nand_chip *chip = mtd->priv;
@@ -281,49 +401,52 @@ void nand_write_buf16(struct mtd_info *mtd, const uint8_t *buf, int len)
        len >>= 1;
 
        for (i = 0; i < len; i++)
-               writew(p[i], chip->IO_ADDR_W);
+               if (p[i] != readw(chip->IO_ADDR_R))
+                       return -EFAULT;
 
+       return 0;
 }
+#endif
+#endif
 
 /**
- * nand_read_buf16 - [DEFAULT] read chip data into buffer
+ * nand_write_buf16 - [DEFAULT] write buffer to chip
  * @mtd: MTD device structure
- * @buf: buffer to store date
- * @len: number of bytes to read
+ * @buf: data buffer
+ * @len: number of bytes to write
  *
- * Default read function for 16bit buswidth.
+ * Default write function for 16bit buswidth.
  */
-void nand_read_buf16(struct mtd_info *mtd, uint8_t *buf, int len)
+#ifndef __UBOOT__
+static void nand_write_buf16(struct mtd_info *mtd, const uint8_t *buf, int len)
+#else
+void nand_write_buf16(struct mtd_info *mtd, const uint8_t *buf, int len)
+#endif
 {
-       int i;
        struct nand_chip *chip = mtd->priv;
        u16 *p = (u16 *) buf;
-       len >>= 1;
 
-       for (i = 0; i < len; i++)
-               p[i] = readw(chip->IO_ADDR_R);
+       iowrite16_rep(chip->IO_ADDR_W, p, len >> 1);
 }
 
 /**
- * nand_verify_buf16 - [DEFAULT] Verify chip data against buffer
+ * nand_read_buf16 - [DEFAULT] read chip data into buffer
  * @mtd: MTD device structure
- * @buf: buffer containing the data to compare
- * @len: number of bytes to compare
+ * @buf: buffer to store date
+ * @len: number of bytes to read
  *
- * Default verify function for 16bit buswidth.
+ * Default read function for 16bit buswidth.
  */
-static int nand_verify_buf16(struct mtd_info *mtd, const uint8_t *buf, int len)
+#ifndef __UBOOT__
+static void nand_read_buf16(struct mtd_info *mtd, uint8_t *buf, int len)
+#else
+void nand_read_buf16(struct mtd_info *mtd, uint8_t *buf, int len)
+#endif
 {
-       int i;
        struct nand_chip *chip = mtd->priv;
        u16 *p = (u16 *) buf;
-       len >>= 1;
-
-       for (i = 0; i < len; i++)
-               if (p[i] != readw(chip->IO_ADDR_R))
-                       return -EFAULT;
 
-       return 0;
+       ioread16_rep(chip->IO_ADDR_R, p, len >> 1);
 }
 
 /**
@@ -348,7 +471,7 @@ static int nand_block_bad(struct mtd_info *mtd, loff_t ofs, int getchip)
        if (getchip) {
                chipnr = (int)(ofs >> chip->chip_shift);
 
-               nand_get_device(chip, mtd, FL_READING);
+               nand_get_device(mtd, FL_READING);
 
                /* Select the NAND device */
                chip->select_chip(mtd, chipnr);
@@ -378,87 +501,97 @@ static int nand_block_bad(struct mtd_info *mtd, loff_t ofs, int getchip)
                i++;
        } while (!res && i < 2 && (chip->bbt_options & NAND_BBT_SCAN2NDPAGE));
 
-       if (getchip)
+       if (getchip) {
+               chip->select_chip(mtd, -1);
                nand_release_device(mtd);
+       }
 
        return res;
 }
 
 /**
- * nand_default_block_markbad - [DEFAULT] mark a block bad
+ * nand_default_block_markbad - [DEFAULT] mark a block bad via bad block marker
  * @mtd: MTD device structure
  * @ofs: offset from device start
  *
  * This is the default implementation, which can be overridden by a hardware
- * specific driver. We try operations in the following order, according to our
- * bbt_options (NAND_BBT_NO_OOB_BBM and NAND_BBT_USE_FLASH):
+ * specific driver. It provides the details for writing a bad block marker to a
+ * block.
+ */
+static int nand_default_block_markbad(struct mtd_info *mtd, loff_t ofs)
+{
+       struct nand_chip *chip = mtd->priv;
+       struct mtd_oob_ops ops;
+       uint8_t buf[2] = { 0, 0 };
+       int ret = 0, res, i = 0;
+
+       ops.datbuf = NULL;
+       ops.oobbuf = buf;
+       ops.ooboffs = chip->badblockpos;
+       if (chip->options & NAND_BUSWIDTH_16) {
+               ops.ooboffs &= ~0x01;
+               ops.len = ops.ooblen = 2;
+       } else {
+               ops.len = ops.ooblen = 1;
+       }
+       ops.mode = MTD_OPS_PLACE_OOB;
+
+       /* Write to first/last page(s) if necessary */
+       if (chip->bbt_options & NAND_BBT_SCANLASTPAGE)
+               ofs += mtd->erasesize - mtd->writesize;
+       do {
+               res = nand_do_write_oob(mtd, ofs, &ops);
+               if (!ret)
+                       ret = res;
+
+               i++;
+               ofs += mtd->writesize;
+       } while ((chip->bbt_options & NAND_BBT_SCAN2NDPAGE) && i < 2);
+
+       return ret;
+}
+
+/**
+ * nand_block_markbad_lowlevel - mark a block bad
+ * @mtd: MTD device structure
+ * @ofs: offset from device start
+ *
+ * This function performs the generic NAND bad block marking steps (i.e., bad
+ * block table(s) and/or marker(s)). We only allow the hardware driver to
+ * specify how to write bad block markers to OOB (chip->block_markbad).
+ *
+ * We try operations in the following order:
  *  (1) erase the affected block, to allow OOB marker to be written cleanly
- *  (2) update in-memory BBT
- *  (3) write bad block marker to OOB area of affected block
- *  (4) update flash-based BBT
- * Note that we retain the first error encountered in (3) or (4), finish the
+ *  (2) write bad block marker to OOB area of affected block (unless flag
+ *      NAND_BBT_NO_OOB_BBM is present)
+ *  (3) update the BBT
+ * Note that we retain the first error encountered in (2) or (3), finish the
  * procedures, and dump the error in the end.
 */
-static int nand_default_block_markbad(struct mtd_info *mtd, loff_t ofs)
+static int nand_block_markbad_lowlevel(struct mtd_info *mtd, loff_t ofs)
 {
        struct nand_chip *chip = mtd->priv;
-       uint8_t buf[2] = { 0, 0 };
-       int block, res, ret = 0, i = 0;
-       int write_oob = !(chip->bbt_options & NAND_BBT_NO_OOB_BBM);
+       int res, ret = 0;
 
-       if (write_oob) {
+       if (!(chip->bbt_options & NAND_BBT_NO_OOB_BBM)) {
                struct erase_info einfo;
 
                /* Attempt erase before marking OOB */
                memset(&einfo, 0, sizeof(einfo));
                einfo.mtd = mtd;
                einfo.addr = ofs;
-               einfo.len = 1 << chip->phys_erase_shift;
+               einfo.len = 1ULL << chip->phys_erase_shift;
                nand_erase_nand(mtd, &einfo, 0);
-       }
-
-       /* Get block number */
-       block = (int)(ofs >> chip->bbt_erase_shift);
-       /* Mark block bad in memory-based BBT */
-       if (chip->bbt)
-               chip->bbt[block >> 2] |= 0x01 << ((block & 0x03) << 1);
-
-       /* Write bad block marker to OOB */
-       if (write_oob) {
-               struct mtd_oob_ops ops;
-               loff_t wr_ofs = ofs;
-
-               nand_get_device(chip, mtd, FL_WRITING);
-
-               ops.datbuf = NULL;
-               ops.oobbuf = buf;
-               ops.ooboffs = chip->badblockpos;
-               if (chip->options & NAND_BUSWIDTH_16) {
-                       ops.ooboffs &= ~0x01;
-                       ops.len = ops.ooblen = 2;
-               } else {
-                       ops.len = ops.ooblen = 1;
-               }
-               ops.mode = MTD_OPS_PLACE_OOB;
-
-               /* Write to first/last page(s) if necessary */
-               if (chip->bbt_options & NAND_BBT_SCANLASTPAGE)
-                       wr_ofs += mtd->erasesize - mtd->writesize;
-               do {
-                       res = nand_do_write_oob(mtd, wr_ofs, &ops);
-                       if (!ret)
-                               ret = res;
-
-                       i++;
-                       wr_ofs += mtd->writesize;
-               } while ((chip->bbt_options & NAND_BBT_SCAN2NDPAGE) && i < 2);
 
+               /* Write bad block marker to OOB */
+               nand_get_device(mtd, FL_WRITING);
+               ret = chip->block_markbad(mtd, ofs);
                nand_release_device(mtd);
        }
 
-       /* Update flash-based bad block table */
-       if (chip->bbt_options & NAND_BBT_USE_FLASH) {
-               res = nand_update_bbt(mtd, ofs);
+       /* Mark block bad in BBT */
+       if (chip->bbt) {
+               res = nand_markbad_bbt(mtd, ofs);
                if (!ret)
                        ret = res;
        }
@@ -504,11 +637,6 @@ static int nand_block_checkbad(struct mtd_info *mtd, loff_t ofs, int getchip,
 {
        struct nand_chip *chip = mtd->priv;
 
-       if (!(chip->options & NAND_BBT_SCANNED)) {
-               chip->options |= NAND_BBT_SCANNED;
-               chip->scan_bbt(mtd);
-       }
-
        if (!chip->bbt)
                return chip->block_bad(mtd, ofs, getchip);
 
@@ -516,22 +644,63 @@ static int nand_block_checkbad(struct mtd_info *mtd, loff_t ofs, int getchip,
        return nand_isbad_bbt(mtd, ofs, allowbbt);
 }
 
+#ifndef __UBOOT__
+/**
+ * panic_nand_wait_ready - [GENERIC] Wait for the ready pin after commands.
+ * @mtd: MTD device structure
+ * @timeo: Timeout
+ *
+ * Helper function for nand_wait_ready used when needing to wait in interrupt
+ * context.
+ */
+static void panic_nand_wait_ready(struct mtd_info *mtd, unsigned long timeo)
+{
+       struct nand_chip *chip = mtd->priv;
+       int i;
+
+       /* Wait for the device to get ready */
+       for (i = 0; i < timeo; i++) {
+               if (chip->dev_ready(mtd))
+                       break;
+               touch_softlockup_watchdog();
+               mdelay(1);
+       }
+}
+#endif
+
 /* Wait for the ready pin, after a command. The timeout is caught later. */
 void nand_wait_ready(struct mtd_info *mtd)
 {
        struct nand_chip *chip = mtd->priv;
+#ifndef __UBOOT__
+       unsigned long timeo = jiffies + msecs_to_jiffies(20);
+
+       /* 400ms timeout */
+       if (in_interrupt() || oops_in_progress)
+               return panic_nand_wait_ready(mtd, 400);
+
+       led_trigger_event(nand_led_trigger, LED_FULL);
+       /* Wait until command is processed or timeout occurs */
+       do {
+               if (chip->dev_ready(mtd))
+                       break;
+               touch_softlockup_watchdog();
+       } while (time_before(jiffies, timeo));
+       led_trigger_event(nand_led_trigger, LED_OFF);
+#else
        u32 timeo = (CONFIG_SYS_HZ * 20) / 1000;
        u32 time_start;
 
        time_start = get_timer(0);
-
        /* Wait until command is processed or timeout occurs */
        while (get_timer(time_start) < timeo) {
                if (chip->dev_ready)
                        if (chip->dev_ready(mtd))
                                break;
        }
+#endif
 }
+EXPORT_SYMBOL_GPL(nand_wait_ready);
 
 /**
  * nand_command - [DEFAULT] Send command to NAND device
@@ -541,7 +710,7 @@ void nand_wait_ready(struct mtd_info *mtd)
  * @page_addr: the page address for this command, -1 if none
  *
  * Send command to NAND device. This function is used for small page devices
- * (256/512 Bytes per page).
+ * (512 Bytes per page).
  */
 static void nand_command(struct mtd_info *mtd, unsigned int command,
                         int column, int page_addr)
@@ -660,8 +829,7 @@ static void nand_command_lp(struct mtd_info *mtd, unsigned int command,
        }
 
        /* Command latch cycle */
-       chip->cmd_ctrl(mtd, command & 0xff,
-                      NAND_NCE | NAND_CLE | NAND_CTRL_CHANGE);
+       chip->cmd_ctrl(mtd, command, NAND_NCE | NAND_CLE | NAND_CTRL_CHANGE);
 
        if (column != -1 || page_addr != -1) {
                int ctrl = NAND_CTRL_CHANGE | NAND_NCE | NAND_ALE;
@@ -701,16 +869,6 @@ static void nand_command_lp(struct mtd_info *mtd, unsigned int command,
        case NAND_CMD_SEQIN:
        case NAND_CMD_RNDIN:
        case NAND_CMD_STATUS:
-       case NAND_CMD_DEPLETE1:
-               return;
-
-       case NAND_CMD_STATUS_ERROR:
-       case NAND_CMD_STATUS_ERROR0:
-       case NAND_CMD_STATUS_ERROR1:
-       case NAND_CMD_STATUS_ERROR2:
-       case NAND_CMD_STATUS_ERROR3:
-               /* Read error status commands require only a short delay */
-               udelay(chip->chip_delay);
                return;
 
        case NAND_CMD_RESET:
@@ -761,18 +919,91 @@ static void nand_command_lp(struct mtd_info *mtd, unsigned int command,
 }
 
 /**
- * nand_get_device - [GENERIC] Get chip for selected access
+ * panic_nand_get_device - [GENERIC] Get chip for selected access
  * @chip: the nand chip descriptor
  * @mtd: MTD device structure
  * @new_state: the state which is requested
  *
+ * Used when in panic, no locks are taken.
+ */
+static void panic_nand_get_device(struct nand_chip *chip,
+                     struct mtd_info *mtd, int new_state)
+{
+       /* Hardware controller shared among independent devices */
+       chip->controller->active = chip;
+       chip->state = new_state;
+}
+
+/**
+ * nand_get_device - [GENERIC] Get chip for selected access
+ * @mtd: MTD device structure
+ * @new_state: the state which is requested
+ *
  * Get the device and lock it for exclusive access
  */
 static int
-nand_get_device(struct nand_chip *chip, struct mtd_info *mtd, int new_state)
+nand_get_device(struct mtd_info *mtd, int new_state)
 {
+       struct nand_chip *chip = mtd->priv;
+#ifndef __UBOOT__
+       spinlock_t *lock = &chip->controller->lock;
+       wait_queue_head_t *wq = &chip->controller->wq;
+       DECLARE_WAITQUEUE(wait, current);
+retry:
+       spin_lock(lock);
+
+       /* Hardware controller shared among independent devices */
+       if (!chip->controller->active)
+               chip->controller->active = chip;
+
+       if (chip->controller->active == chip && chip->state == FL_READY) {
+               chip->state = new_state;
+               spin_unlock(lock);
+               return 0;
+       }
+       if (new_state == FL_PM_SUSPENDED) {
+               if (chip->controller->active->state == FL_PM_SUSPENDED) {
+                       chip->state = FL_PM_SUSPENDED;
+                       spin_unlock(lock);
+                       return 0;
+               }
+       }
+       set_current_state(TASK_UNINTERRUPTIBLE);
+       add_wait_queue(wq, &wait);
+       spin_unlock(lock);
+       schedule();
+       remove_wait_queue(wq, &wait);
+       goto retry;
+#else
        chip->state = new_state;
        return 0;
+#endif
+}
+
+/**
+ * panic_nand_wait - [GENERIC] wait until the command is done
+ * @mtd: MTD device structure
+ * @chip: NAND chip structure
+ * @timeo: timeout
+ *
+ * Wait for command done. This is a helper function for nand_wait used when
+ * we are in interrupt context. May happen when in panic and trying to write
+ * an oops through mtdoops.
+ */
+static void panic_nand_wait(struct mtd_info *mtd, struct nand_chip *chip,
+                           unsigned long timeo)
+{
+       int i;
+       for (i = 0; i < timeo; i++) {
+               if (chip->dev_ready) {
+                       if (chip->dev_ready(mtd))
+                               break;
+               } else {
+                       if (chip->read_byte(mtd) & NAND_STATUS_READY)
+                               break;
+               }
+               mdelay(1);
+       }
 }
 
 /**
@@ -786,28 +1017,42 @@ nand_get_device(struct nand_chip *chip, struct mtd_info *mtd, int new_state)
  */
 static int nand_wait(struct mtd_info *mtd, struct nand_chip *chip)
 {
-       unsigned long   timeo;
-       int state = chip->state;
-       u32 time_start;
 
-       if (state == FL_ERASING)
-               timeo = (CONFIG_SYS_HZ * 400) / 1000;
-       else
-               timeo = (CONFIG_SYS_HZ * 20) / 1000;
+       int status, state = chip->state;
+       unsigned long timeo = (state == FL_ERASING ? 400 : 20);
 
-       if ((state == FL_ERASING) && (chip->options & NAND_IS_AND))
-               chip->cmdfunc(mtd, NAND_CMD_STATUS_MULTI, -1, -1);
-       else
-               chip->cmdfunc(mtd, NAND_CMD_STATUS, -1, -1);
+       led_trigger_event(nand_led_trigger, LED_FULL);
 
-       time_start = get_timer(0);
+       /*
+        * Apply this short delay always to ensure that we do wait tWB in any
+        * case on any machine.
+        */
+       ndelay(100);
 
-       while (1) {
-               if (get_timer(time_start) > timeo) {
-                       printf("Timeout!");
-                       return 0x01;
-               }
+       chip->cmdfunc(mtd, NAND_CMD_STATUS, -1, -1);
 
+#ifndef __UBOOT__
+       if (in_interrupt() || oops_in_progress)
+               panic_nand_wait(mtd, chip, timeo);
+       else {
+               timeo = jiffies + msecs_to_jiffies(timeo);
+               while (time_before(jiffies, timeo)) {
+                       if (chip->dev_ready) {
+                               if (chip->dev_ready(mtd))
+                                       break;
+                       } else {
+                               if (chip->read_byte(mtd) & NAND_STATUS_READY)
+                                       break;
+                       }
+                       cond_resched();
+               }
+       }
+#else
+       u32 timer = (CONFIG_SYS_HZ * timeo) / 1000;
+       u32 time_start;
+       time_start = get_timer(0);
+       while (get_timer(time_start) < timer) {
                if (chip->dev_ready) {
                        if (chip->dev_ready(mtd))
                                break;
@@ -816,14 +1061,175 @@ static int nand_wait(struct mtd_info *mtd, struct nand_chip *chip)
                                break;
                }
        }
+#endif
 #ifdef PPCHAMELON_NAND_TIMER_HACK
        time_start = get_timer(0);
        while (get_timer(time_start) < 10)
                ;
 #endif /*  PPCHAMELON_NAND_TIMER_HACK */
+       led_trigger_event(nand_led_trigger, LED_OFF);
+
+       status = (int)chip->read_byte(mtd);
+       /* This can happen if in case of timeout or buggy dev_ready */
+       WARN_ON(!(status & NAND_STATUS_READY));
+       return status;
+}
+
+#ifndef __UBOOT__
+/**
+ * __nand_unlock - [REPLACEABLE] unlocks specified locked blocks
+ * @mtd: mtd info
+ * @ofs: offset to start unlock from
+ * @len: length to unlock
+ * @invert: when = 0, unlock the range of blocks within the lower and
+ *                    upper boundary address
+ *          when = 1, unlock the range of blocks outside the boundaries
+ *                    of the lower and upper boundary address
+ *
+ * Returs unlock status.
+ */
+static int __nand_unlock(struct mtd_info *mtd, loff_t ofs,
+                                       uint64_t len, int invert)
+{
+       int ret = 0;
+       int status, page;
+       struct nand_chip *chip = mtd->priv;
+
+       /* Submit address of first page to unlock */
+       page = ofs >> chip->page_shift;
+       chip->cmdfunc(mtd, NAND_CMD_UNLOCK1, -1, page & chip->pagemask);
+
+       /* Submit address of last page to unlock */
+       page = (ofs + len) >> chip->page_shift;
+       chip->cmdfunc(mtd, NAND_CMD_UNLOCK2, -1,
+                               (page | invert) & chip->pagemask);
+
+       /* Call wait ready function */
+       status = chip->waitfunc(mtd, chip);
+       /* See if device thinks it succeeded */
+       if (status & NAND_STATUS_FAIL) {
+               pr_debug("%s: error status = 0x%08x\n",
+                                       __func__, status);
+               ret = -EIO;
+       }
+
+       return ret;
+}
+
+/**
+ * nand_unlock - [REPLACEABLE] unlocks specified locked blocks
+ * @mtd: mtd info
+ * @ofs: offset to start unlock from
+ * @len: length to unlock
+ *
+ * Returns unlock status.
+ */
+int nand_unlock(struct mtd_info *mtd, loff_t ofs, uint64_t len)
+{
+       int ret = 0;
+       int chipnr;
+       struct nand_chip *chip = mtd->priv;
+
+       pr_debug("%s: start = 0x%012llx, len = %llu\n",
+                       __func__, (unsigned long long)ofs, len);
+
+       if (check_offs_len(mtd, ofs, len))
+               ret = -EINVAL;
+
+       /* Align to last block address if size addresses end of the device */
+       if (ofs + len == mtd->size)
+               len -= mtd->erasesize;
+
+       nand_get_device(mtd, FL_UNLOCKING);
+
+       /* Shift to get chip number */
+       chipnr = ofs >> chip->chip_shift;
+
+       chip->select_chip(mtd, chipnr);
+
+       /* Check, if it is write protected */
+       if (nand_check_wp(mtd)) {
+               pr_debug("%s: device is write protected!\n",
+                                       __func__);
+               ret = -EIO;
+               goto out;
+       }
+
+       ret = __nand_unlock(mtd, ofs, len, 0);
+
+out:
+       chip->select_chip(mtd, -1);
+       nand_release_device(mtd);
+
+       return ret;
+}
+EXPORT_SYMBOL(nand_unlock);
+
+/**
+ * nand_lock - [REPLACEABLE] locks all blocks present in the device
+ * @mtd: mtd info
+ * @ofs: offset to start unlock from
+ * @len: length to unlock
+ *
+ * This feature is not supported in many NAND parts. 'Micron' NAND parts do
+ * have this feature, but it allows only to lock all blocks, not for specified
+ * range for block. Implementing 'lock' feature by making use of 'unlock', for
+ * now.
+ *
+ * Returns lock status.
+ */
+int nand_lock(struct mtd_info *mtd, loff_t ofs, uint64_t len)
+{
+       int ret = 0;
+       int chipnr, status, page;
+       struct nand_chip *chip = mtd->priv;
+
+       pr_debug("%s: start = 0x%012llx, len = %llu\n",
+                       __func__, (unsigned long long)ofs, len);
+
+       if (check_offs_len(mtd, ofs, len))
+               ret = -EINVAL;
+
+       nand_get_device(mtd, FL_LOCKING);
+
+       /* Shift to get chip number */
+       chipnr = ofs >> chip->chip_shift;
+
+       chip->select_chip(mtd, chipnr);
+
+       /* Check, if it is write protected */
+       if (nand_check_wp(mtd)) {
+               pr_debug("%s: device is write protected!\n",
+                                       __func__);
+               status = MTD_ERASE_FAILED;
+               ret = -EIO;
+               goto out;
+       }
+
+       /* Submit address of first page to lock */
+       page = ofs >> chip->page_shift;
+       chip->cmdfunc(mtd, NAND_CMD_LOCK, -1, page & chip->pagemask);
 
-       return (int)chip->read_byte(mtd);
+       /* Call wait ready function */
+       status = chip->waitfunc(mtd, chip);
+       /* See if device thinks it succeeded */
+       if (status & NAND_STATUS_FAIL) {
+               pr_debug("%s: error status = 0x%08x\n",
+                                       __func__, status);
+               ret = -EIO;
+               goto out;
+       }
+
+       ret = __nand_unlock(mtd, ofs, len, 0x1);
+
+out:
+       chip->select_chip(mtd, -1);
+       nand_release_device(mtd);
+
+       return ret;
 }
+EXPORT_SYMBOL(nand_lock);
+#endif
 
 /**
  * nand_read_page_raw - [INTERN] read raw page data without ecc
@@ -906,6 +1312,7 @@ static int nand_read_page_swecc(struct mtd_info *mtd, struct nand_chip *chip,
        uint8_t *ecc_calc = chip->buffers->ecccalc;
        uint8_t *ecc_code = chip->buffers->ecccode;
        uint32_t *eccpos = chip->ecc.layout->eccpos;
+       unsigned int max_bitflips = 0;
 
        chip->ecc.read_page_raw(mtd, chip, buf, 1, page);
 
@@ -922,16 +1329,18 @@ static int nand_read_page_swecc(struct mtd_info *mtd, struct nand_chip *chip,
                int stat;
 
                stat = chip->ecc.correct(mtd, p, &ecc_code[i], &ecc_calc[i]);
-               if (stat < 0)
+               if (stat < 0) {
                        mtd->ecc_stats.failed++;
-               else
+               } else {
                        mtd->ecc_stats.corrected += stat;
+                       max_bitflips = max_t(unsigned int, max_bitflips, stat);
+               }
        }
-       return 0;
+       return max_bitflips;
 }
 
 /**
- * nand_read_subpage - [REPLACEABLE] software ECC based sub-page read function
+ * nand_read_subpage - [REPLACEABLE] ECC based sub-page read function
  * @mtd: mtd info structure
  * @chip: nand chip info structure
  * @data_offs: offset of requested data within the page
@@ -948,6 +1357,7 @@ static int nand_read_subpage(struct mtd_info *mtd, struct nand_chip *chip,
        int datafrag_len, eccfrag_len, aligned_len, aligned_pos;
        int busw = (chip->options & NAND_BUSWIDTH_16) ? 2 : 1;
        int index = 0;
+       unsigned int max_bitflips = 0;
 
        /* Column address within the page aligned to ECC size (256bytes) */
        start_step = data_offs / chip->ecc.size;
@@ -1012,12 +1422,14 @@ static int nand_read_subpage(struct mtd_info *mtd, struct nand_chip *chip,
 
                stat = chip->ecc.correct(mtd, p,
                        &chip->buffers->ecccode[i], &chip->buffers->ecccalc[i]);
-               if (stat < 0)
+               if (stat < 0) {
                        mtd->ecc_stats.failed++;
-               else
+               } else {
                        mtd->ecc_stats.corrected += stat;
+                       max_bitflips = max_t(unsigned int, max_bitflips, stat);
+               }
        }
-       return 0;
+       return max_bitflips;
 }
 
 /**
@@ -1040,6 +1452,7 @@ static int nand_read_page_hwecc(struct mtd_info *mtd, struct nand_chip *chip,
        uint8_t *ecc_calc = chip->buffers->ecccalc;
        uint8_t *ecc_code = chip->buffers->ecccode;
        uint32_t *eccpos = chip->ecc.layout->eccpos;
+       unsigned int max_bitflips = 0;
 
        for (i = 0; eccsteps; eccsteps--, i += eccbytes, p += eccsize) {
                chip->ecc.hwctl(mtd, NAND_ECC_READ);
@@ -1058,12 +1471,14 @@ static int nand_read_page_hwecc(struct mtd_info *mtd, struct nand_chip *chip,
                int stat;
 
                stat = chip->ecc.correct(mtd, p, &ecc_code[i], &ecc_calc[i]);
-               if (stat < 0)
+               if (stat < 0) {
                        mtd->ecc_stats.failed++;
-               else
+               } else {
                        mtd->ecc_stats.corrected += stat;
+                       max_bitflips = max_t(unsigned int, max_bitflips, stat);
+               }
        }
-       return 0;
+       return max_bitflips;
 }
 
 /**
@@ -1090,6 +1505,7 @@ static int nand_read_page_hwecc_oob_first(struct mtd_info *mtd,
        uint8_t *ecc_code = chip->buffers->ecccode;
        uint32_t *eccpos = chip->ecc.layout->eccpos;
        uint8_t *ecc_calc = chip->buffers->ecccalc;
+       unsigned int max_bitflips = 0;
 
        /* Read the OOB area first */
        chip->cmdfunc(mtd, NAND_CMD_READOOB, 0, page);
@@ -1107,12 +1523,14 @@ static int nand_read_page_hwecc_oob_first(struct mtd_info *mtd,
                chip->ecc.calculate(mtd, p, &ecc_calc[i]);
 
                stat = chip->ecc.correct(mtd, p, &ecc_code[i], NULL);
-               if (stat < 0)
+               if (stat < 0) {
                        mtd->ecc_stats.failed++;
-               else
+               } else {
                        mtd->ecc_stats.corrected += stat;
+                       max_bitflips = max_t(unsigned int, max_bitflips, stat);
+               }
        }
-       return 0;
+       return max_bitflips;
 }
 
 /**
@@ -1134,6 +1552,7 @@ static int nand_read_page_syndrome(struct mtd_info *mtd, struct nand_chip *chip,
        int eccsteps = chip->ecc.steps;
        uint8_t *p = buf;
        uint8_t *oob = chip->oob_poi;
+       unsigned int max_bitflips = 0;
 
        for (i = 0; eccsteps; eccsteps--, i += eccbytes, p += eccsize) {
                int stat;
@@ -1150,10 +1569,12 @@ static int nand_read_page_syndrome(struct mtd_info *mtd, struct nand_chip *chip,
                chip->read_buf(mtd, oob, eccbytes);
                stat = chip->ecc.correct(mtd, p, oob, NULL);
 
-               if (stat < 0)
+               if (stat < 0) {
                        mtd->ecc_stats.failed++;
-               else
+               } else {
                        mtd->ecc_stats.corrected += stat;
+                       max_bitflips = max_t(unsigned int, max_bitflips, stat);
+               }
 
                oob += eccbytes;
 
@@ -1168,7 +1589,7 @@ static int nand_read_page_syndrome(struct mtd_info *mtd, struct nand_chip *chip,
        if (i)
                chip->read_buf(mtd, oob, i);
 
-       return 0;
+       return max_bitflips;
 }
 
 /**
@@ -1219,6 +1640,30 @@ static uint8_t *nand_transfer_oob(struct nand_chip *chip, uint8_t *oob,
        return NULL;
 }
 
+/**
+ * nand_setup_read_retry - [INTERN] Set the READ RETRY mode
+ * @mtd: MTD device structure
+ * @retry_mode: the retry mode to use
+ *
+ * Some vendors supply a special command to shift the Vt threshold, to be used
+ * when there are too many bitflips in a page (i.e., ECC error). After setting
+ * a new threshold, the host should retry reading the page.
+ */
+static int nand_setup_read_retry(struct mtd_info *mtd, int retry_mode)
+{
+       struct nand_chip *chip = mtd->priv;
+
+       pr_debug("setting READ RETRY mode %d\n", retry_mode);
+
+       if (retry_mode >= chip->read_retries)
+               return -EINVAL;
+
+       if (!chip->setup_read_retry)
+               return -EOPNOTSUPP;
+
+       return chip->setup_read_retry(mtd, retry_mode);
+}
+
 /**
  * nand_do_read_ops - [INTERN] Read data with ECC
  * @mtd: MTD device structure
@@ -1232,7 +1677,6 @@ static int nand_do_read_ops(struct mtd_info *mtd, loff_t from,
 {
        int chipnr, page, realpage, col, bytes, aligned, oob_required;
        struct nand_chip *chip = mtd->priv;
-       struct mtd_ecc_stats stats;
        int ret = 0;
        uint32_t readlen = ops->len;
        uint32_t oobreadlen = ops->ooblen;
@@ -1241,8 +1685,8 @@ static int nand_do_read_ops(struct mtd_info *mtd, loff_t from,
 
        uint8_t *bufpoi, *oob, *buf;
        unsigned int max_bitflips = 0;
-
-       stats = mtd->ecc_stats;
+       int retry_mode = 0;
+       bool ecc_fail = false;
 
        chipnr = (int)(from >> chip->chip_shift);
        chip->select_chip(mtd, chipnr);
@@ -1257,8 +1701,9 @@ static int nand_do_read_ops(struct mtd_info *mtd, loff_t from,
        oob_required = oob ? 1 : 0;
 
        while (1) {
-               WATCHDOG_RESET();
+               unsigned int ecc_failures = mtd->ecc_stats.failed;
 
+               WATCHDOG_RESET();
                bytes = min(mtd->writesize - col, readlen);
                aligned = (bytes == mtd->writesize);
 
@@ -1266,6 +1711,7 @@ static int nand_do_read_ops(struct mtd_info *mtd, loff_t from,
                if (realpage != chip->pagebuf || oob) {
                        bufpoi = aligned ? buf : chip->buffers->databuf;
 
+read_retry:
                        chip->cmdfunc(mtd, NAND_CMD_READ0, 0x00, page);
 
                        /*
@@ -1277,7 +1723,7 @@ static int nand_do_read_ops(struct mtd_info *mtd, loff_t from,
                                                              oob_required,
                                                              page);
                        else if (!aligned && NAND_HAS_SUBPAGE_READ(chip) &&
-                           !oob)
+                                !oob)
                                ret = chip->ecc.read_subpage(mtd, chip,
                                                        col, bytes, bufpoi);
                        else
@@ -1295,7 +1741,7 @@ static int nand_do_read_ops(struct mtd_info *mtd, loff_t from,
                        /* Transfer not aligned data */
                        if (!aligned) {
                                if (!NAND_HAS_SUBPAGE_READ(chip) && !oob &&
-                                   !(mtd->ecc_stats.failed - stats.failed) &&
+                                   !(mtd->ecc_stats.failed - ecc_failures) &&
                                    (ops->mode != MTD_OPS_RAW)) {
                                        chip->pagebuf = realpage;
                                        chip->pagebuf_bitflips = ret;
@@ -1306,8 +1752,6 @@ static int nand_do_read_ops(struct mtd_info *mtd, loff_t from,
                                memcpy(buf, chip->buffers->databuf + col, bytes);
                        }
 
-                       buf += bytes;
-
                        if (unlikely(oob)) {
                                int toread = min(oobreadlen, max_oobsize);
 
@@ -1317,6 +1761,33 @@ static int nand_do_read_ops(struct mtd_info *mtd, loff_t from,
                                        oobreadlen -= toread;
                                }
                        }
+
+                       if (chip->options & NAND_NEED_READRDY) {
+                               /* Apply delay or wait for ready/busy pin */
+                               if (!chip->dev_ready)
+                                       udelay(chip->chip_delay);
+                               else
+                                       nand_wait_ready(mtd);
+                       }
+
+                       if (mtd->ecc_stats.failed - ecc_failures) {
+                               if (retry_mode + 1 < chip->read_retries) {
+                                       retry_mode++;
+                                       ret = nand_setup_read_retry(mtd,
+                                                       retry_mode);
+                                       if (ret < 0)
+                                               break;
+
+                                       /* Reset failures; retry */
+                                       mtd->ecc_stats.failed = ecc_failures;
+                                       goto read_retry;
+                               } else {
+                                       /* No more retry modes; real failure */
+                                       ecc_fail = true;
+                               }
+                       }
+
+                       buf += bytes;
                } else {
                        memcpy(buf, chip->buffers->databuf + col, bytes);
                        buf += bytes;
@@ -1326,6 +1797,14 @@ static int nand_do_read_ops(struct mtd_info *mtd, loff_t from,
 
                readlen -= bytes;
 
+               /* Reset to retry mode 0 */
+               if (retry_mode) {
+                       ret = nand_setup_read_retry(mtd, 0);
+                       if (ret < 0)
+                               break;
+                       retry_mode = 0;
+               }
+
                if (!readlen)
                        break;
 
@@ -1342,15 +1821,16 @@ static int nand_do_read_ops(struct mtd_info *mtd, loff_t from,
                        chip->select_chip(mtd, chipnr);
                }
        }
+       chip->select_chip(mtd, -1);
 
        ops->retlen = ops->len - (size_t) readlen;
        if (oob)
                ops->oobretlen = ops->ooblen - oobreadlen;
 
-       if (ret)
+       if (ret < 0)
                return ret;
 
-       if (mtd->ecc_stats.failed - stats.failed)
+       if (ecc_fail)
                return -EBADMSG;
 
        return max_bitflips;
@@ -1369,11 +1849,10 @@ static int nand_do_read_ops(struct mtd_info *mtd, loff_t from,
 static int nand_read(struct mtd_info *mtd, loff_t from, size_t len,
                     size_t *retlen, uint8_t *buf)
 {
-       struct nand_chip *chip = mtd->priv;
        struct mtd_oob_ops ops;
        int ret;
 
-       nand_get_device(chip, mtd, FL_READING);
+       nand_get_device(mtd, FL_READING);
        ops.len = len;
        ops.datbuf = buf;
        ops.oobbuf = NULL;
@@ -1537,7 +2016,7 @@ static int nand_do_read_oob(struct mtd_info *mtd, loff_t from,
        uint8_t *buf = ops->oobbuf;
        int ret = 0;
 
-       MTDDEBUG(MTD_DEBUG_LEVEL3, "%s: from = 0x%08Lx, len = %i\n",
+       pr_debug("%s: from = 0x%08Lx, len = %i\n",
                        __func__, (unsigned long long)from, readlen);
 
        stats = mtd->ecc_stats;
@@ -1548,8 +2027,8 @@ static int nand_do_read_oob(struct mtd_info *mtd, loff_t from,
                len = mtd->oobsize;
 
        if (unlikely(ops->ooboffs >= len)) {
-               MTDDEBUG(MTD_DEBUG_LEVEL0, "%s: Attempt to start read "
-                                       "outside oob\n", __func__);
+               pr_debug("%s: attempt to start read outside oob\n",
+                               __func__);
                return -EINVAL;
        }
 
@@ -1557,8 +2036,8 @@ static int nand_do_read_oob(struct mtd_info *mtd, loff_t from,
        if (unlikely(from >= mtd->size ||
                     ops->ooboffs + readlen > ((mtd->size >> chip->page_shift) -
                                        (from >> chip->page_shift)) * len)) {
-               MTDDEBUG(MTD_DEBUG_LEVEL0, "%s: Attempt read beyond end "
-                                       "of device\n", __func__);
+               pr_debug("%s: attempt to read beyond end of device\n",
+                               __func__);
                return -EINVAL;
        }
 
@@ -1571,6 +2050,7 @@ static int nand_do_read_oob(struct mtd_info *mtd, loff_t from,
 
        while (1) {
                WATCHDOG_RESET();
+
                if (ops->mode == MTD_OPS_RAW)
                        ret = chip->ecc.read_oob_raw(mtd, chip, page);
                else
@@ -1582,6 +2062,14 @@ static int nand_do_read_oob(struct mtd_info *mtd, loff_t from,
                len = min(len, readlen);
                buf = nand_transfer_oob(chip, buf, ops, len);
 
+               if (chip->options & NAND_NEED_READRDY) {
+                       /* Apply delay or wait for ready/busy pin */
+                       if (!chip->dev_ready)
+                               udelay(chip->chip_delay);
+                       else
+                               nand_wait_ready(mtd);
+               }
+
                readlen -= len;
                if (!readlen)
                        break;
@@ -1597,6 +2085,7 @@ static int nand_do_read_oob(struct mtd_info *mtd, loff_t from,
                        chip->select_chip(mtd, chipnr);
                }
        }
+       chip->select_chip(mtd, -1);
 
        ops->oobretlen = ops->ooblen - readlen;
 
@@ -1620,19 +2109,18 @@ static int nand_do_read_oob(struct mtd_info *mtd, loff_t from,
 static int nand_read_oob(struct mtd_info *mtd, loff_t from,
                         struct mtd_oob_ops *ops)
 {
-       struct nand_chip *chip = mtd->priv;
        int ret = -ENOTSUPP;
 
        ops->retlen = 0;
 
        /* Do not allow reads past end of device */
        if (ops->datbuf && (from + ops->len) > mtd->size) {
-               MTDDEBUG(MTD_DEBUG_LEVEL0, "%s: Attempt read "
-                               "beyond end of device\n", __func__);
+               pr_debug("%s: attempt to read beyond end of device\n",
+                               __func__);
                return -EINVAL;
        }
 
-       nand_get_device(chip, mtd, FL_READING);
+       nand_get_device(mtd, FL_READING);
 
        switch (ops->mode) {
        case MTD_OPS_PLACE_OOB:
@@ -1774,6 +2262,68 @@ static int nand_write_page_hwecc(struct mtd_info *mtd, struct nand_chip *chip,
        return 0;
 }
 
+
+/**
+ * nand_write_subpage_hwecc - [REPLACABLE] hardware ECC based subpage write
+ * @mtd:       mtd info structure
+ * @chip:      nand chip info structure
+ * @offset:    column address of subpage within the page
+ * @data_len:  data length
+ * @buf:       data buffer
+ * @oob_required: must write chip->oob_poi to OOB
+ */
+static int nand_write_subpage_hwecc(struct mtd_info *mtd,
+                               struct nand_chip *chip, uint32_t offset,
+                               uint32_t data_len, const uint8_t *buf,
+                               int oob_required)
+{
+       uint8_t *oob_buf  = chip->oob_poi;
+       uint8_t *ecc_calc = chip->buffers->ecccalc;
+       int ecc_size      = chip->ecc.size;
+       int ecc_bytes     = chip->ecc.bytes;
+       int ecc_steps     = chip->ecc.steps;
+       uint32_t *eccpos  = chip->ecc.layout->eccpos;
+       uint32_t start_step = offset / ecc_size;
+       uint32_t end_step   = (offset + data_len - 1) / ecc_size;
+       int oob_bytes       = mtd->oobsize / ecc_steps;
+       int step, i;
+
+       for (step = 0; step < ecc_steps; step++) {
+               /* configure controller for WRITE access */
+               chip->ecc.hwctl(mtd, NAND_ECC_WRITE);
+
+               /* write data (untouched subpages already masked by 0xFF) */
+               chip->write_buf(mtd, buf, ecc_size);
+
+               /* mask ECC of un-touched subpages by padding 0xFF */
+               if ((step < start_step) || (step > end_step))
+                       memset(ecc_calc, 0xff, ecc_bytes);
+               else
+                       chip->ecc.calculate(mtd, buf, ecc_calc);
+
+               /* mask OOB of un-touched subpages by padding 0xFF */
+               /* if oob_required, preserve OOB metadata of written subpage */
+               if (!oob_required || (step < start_step) || (step > end_step))
+                       memset(oob_buf, 0xff, oob_bytes);
+
+               buf += ecc_size;
+               ecc_calc += ecc_bytes;
+               oob_buf  += oob_bytes;
+       }
+
+       /* copy calculated ECC for whole page to chip->buffer->oob */
+       /* this include masked-value(0xFF) for unwritten subpages */
+       ecc_calc = chip->buffers->ecccalc;
+       for (i = 0; i < chip->ecc.total; i++)
+               chip->oob_poi[eccpos[i]] = ecc_calc[i];
+
+       /* write OOB buffer to NAND device */
+       chip->write_buf(mtd, chip->oob_poi, mtd->oobsize);
+
+       return 0;
+}
+
+
 /**
  * nand_write_page_syndrome - [REPLACEABLE] hardware ECC syndrome based page write
  * @mtd: mtd info structure
@@ -1826,6 +2376,8 @@ static int nand_write_page_syndrome(struct mtd_info *mtd,
  * nand_write_page - [REPLACEABLE] write one page
  * @mtd: MTD device structure
  * @chip: NAND chip descriptor
+ * @offset: address offset within the page
+ * @data_len: length of actual data to be written
  * @buf: the data to write
  * @oob_required: must write chip->oob_poi to OOB
  * @page: page number to write
@@ -1833,15 +2385,25 @@ static int nand_write_page_syndrome(struct mtd_info *mtd,
  * @raw: use _raw version of write_page
  */
 static int nand_write_page(struct mtd_info *mtd, struct nand_chip *chip,
-                          const uint8_t *buf, int oob_required, int page,
-                          int cached, int raw)
+               uint32_t offset, int data_len, const uint8_t *buf,
+               int oob_required, int page, int cached, int raw)
 {
-       int status;
+       int status, subpage;
+
+       if (!(chip->options & NAND_NO_SUBPAGE_WRITE) &&
+               chip->ecc.write_subpage)
+               subpage = offset || (data_len < mtd->writesize);
+       else
+               subpage = 0;
 
        chip->cmdfunc(mtd, NAND_CMD_SEQIN, 0x00, page);
 
        if (unlikely(raw))
-               status = chip->ecc.write_page_raw(mtd, chip, buf, oob_required);
+               status = chip->ecc.write_page_raw(mtd, chip, buf,
+                                                       oob_required);
+       else if (subpage)
+               status = chip->ecc.write_subpage(mtd, chip, offset, data_len,
+                                                        buf, oob_required);
        else
                status = chip->ecc.write_page(mtd, chip, buf, oob_required);
 
@@ -1854,7 +2416,7 @@ static int nand_write_page(struct mtd_info *mtd, struct nand_chip *chip,
         */
        cached = 0;
 
-       if (!cached || !(chip->options & NAND_CACHEPRG)) {
+       if (!cached || !NAND_HAS_CACHEPROG(chip)) {
 
                chip->cmdfunc(mtd, NAND_CMD_PAGEPROG, -1, -1);
                status = chip->waitfunc(mtd, chip);
@@ -1873,7 +2435,9 @@ static int nand_write_page(struct mtd_info *mtd, struct nand_chip *chip,
                status = chip->waitfunc(mtd, chip);
        }
 
-#ifdef CONFIG_MTD_NAND_VERIFY_WRITE
+
+#ifdef __UBOOT__
+#if defined(CONFIG_MTD_NAND_VERIFY_WRITE)
        /* Send command to read back the data */
        chip->cmdfunc(mtd, NAND_CMD_READ0, 0, page);
 
@@ -1883,6 +2447,8 @@ static int nand_write_page(struct mtd_info *mtd, struct nand_chip *chip,
        /* Make sure the next page prog is preceded by a status read */
        chip->cmdfunc(mtd, NAND_CMD_STATUS, -1, -1);
 #endif
+#endif
+
        return 0;
 }
 
@@ -1965,26 +2531,34 @@ static int nand_do_write_ops(struct mtd_info *mtd, loff_t to,
 
        uint8_t *oob = ops->oobbuf;
        uint8_t *buf = ops->datbuf;
-       int ret, subpage;
+       int ret;
        int oob_required = oob ? 1 : 0;
 
        ops->retlen = 0;
        if (!writelen)
                return 0;
 
-       column = to & (mtd->writesize - 1);
-       subpage = column || (writelen & (mtd->writesize - 1));
-
-       if (subpage && oob)
+#ifndef __UBOOT__
+       /* Reject writes, which are not page aligned */
+       if (NOTALIGNED(to) || NOTALIGNED(ops->len)) {
+#else
+       /* Reject writes, which are not page aligned */
+       if (NOTALIGNED(to)) {
+#endif
+               pr_notice("%s: attempt to write non page aligned data\n",
+                          __func__);
                return -EINVAL;
+       }
+
+       column = to & (mtd->writesize - 1);
 
        chipnr = (int)(to >> chip->chip_shift);
        chip->select_chip(mtd, chipnr);
 
        /* Check, if it is write protected */
        if (nand_check_wp(mtd)) {
-               printk (KERN_NOTICE "nand_do_write_ops: Device is write protected\n");
-               return -EIO;
+               ret = -EIO;
+               goto err_out;
        }
 
        realpage = (int)(to >> chip->page_shift);
@@ -1997,18 +2571,19 @@ static int nand_do_write_ops(struct mtd_info *mtd, loff_t to,
                chip->pagebuf = -1;
 
        /* Don't allow multipage oob writes with offset */
-       if (oob && ops->ooboffs && (ops->ooboffs + ops->ooblen > oobmaxlen))
-               return -EINVAL;
+       if (oob && ops->ooboffs && (ops->ooboffs + ops->ooblen > oobmaxlen)) {
+               ret = -EINVAL;
+               goto err_out;
+       }
 
        while (1) {
-               WATCHDOG_RESET();
-
                int bytes = mtd->writesize;
                int cached = writelen > bytes && page != blockmask;
                uint8_t *wbuf = buf;
 
+               WATCHDOG_RESET();
                /* Partial page write? */
-               if (unlikely(column || writelen < mtd->writesize)) {
+               if (unlikely(column || writelen < (mtd->writesize - 1))) {
                        cached = 0;
                        bytes = min_t(int, bytes - column, (int) writelen);
                        chip->pagebuf = -1;
@@ -2025,9 +2600,9 @@ static int nand_do_write_ops(struct mtd_info *mtd, loff_t to,
                        /* We still need to erase leftover OOB data */
                        memset(chip->oob_poi, 0xff, mtd->oobsize);
                }
-
-               ret = chip->write_page(mtd, chip, wbuf, oob_required, page,
-                                      cached, (ops->mode == MTD_OPS_RAW));
+               ret = chip->write_page(mtd, chip, column, bytes, wbuf,
+                                       oob_required, page, cached,
+                                       (ops->mode == MTD_OPS_RAW));
                if (ret)
                        break;
 
@@ -2035,22 +2610,60 @@ static int nand_do_write_ops(struct mtd_info *mtd, loff_t to,
                if (!writelen)
                        break;
 
-               column = 0;
-               buf += bytes;
-               realpage++;
+               column = 0;
+               buf += bytes;
+               realpage++;
+
+               page = realpage & chip->pagemask;
+               /* Check, if we cross a chip boundary */
+               if (!page) {
+                       chipnr++;
+                       chip->select_chip(mtd, -1);
+                       chip->select_chip(mtd, chipnr);
+               }
+       }
+
+       ops->retlen = ops->len - writelen;
+       if (unlikely(oob))
+               ops->oobretlen = ops->ooblen;
+
+err_out:
+       chip->select_chip(mtd, -1);
+       return ret;
+}
+
+/**
+ * panic_nand_write - [MTD Interface] NAND write with ECC
+ * @mtd: MTD device structure
+ * @to: offset to write to
+ * @len: number of bytes to write
+ * @retlen: pointer to variable to store the number of written bytes
+ * @buf: the data to write
+ *
+ * NAND write with ECC. Used when performing writes in interrupt context, this
+ * may for example be called by mtdoops when writing an oops while in panic.
+ */
+static int panic_nand_write(struct mtd_info *mtd, loff_t to, size_t len,
+                           size_t *retlen, const uint8_t *buf)
+{
+       struct nand_chip *chip = mtd->priv;
+       struct mtd_oob_ops ops;
+       int ret;
+
+       /* Wait for the device to get ready */
+       panic_nand_wait(mtd, chip, 400);
+
+       /* Grab the device */
+       panic_nand_get_device(chip, mtd, FL_WRITING);
+
+       ops.len = len;
+       ops.datbuf = (uint8_t *)buf;
+       ops.oobbuf = NULL;
+       ops.mode = MTD_OPS_PLACE_OOB;
 
-               page = realpage & chip->pagemask;
-               /* Check, if we cross a chip boundary */
-               if (!page) {
-                       chipnr++;
-                       chip->select_chip(mtd, -1);
-                       chip->select_chip(mtd, chipnr);
-               }
-       }
+       ret = nand_do_write_ops(mtd, to, &ops);
 
-       ops->retlen = ops->len - writelen;
-       if (unlikely(oob))
-               ops->oobretlen = ops->ooblen;
+       *retlen = ops.retlen;
        return ret;
 }
 
@@ -2067,11 +2680,10 @@ static int nand_do_write_ops(struct mtd_info *mtd, loff_t to,
 static int nand_write(struct mtd_info *mtd, loff_t to, size_t len,
                          size_t *retlen, const uint8_t *buf)
 {
-       struct nand_chip *chip = mtd->priv;
        struct mtd_oob_ops ops;
        int ret;
 
-       nand_get_device(chip, mtd, FL_WRITING);
+       nand_get_device(mtd, FL_WRITING);
        ops.len = len;
        ops.datbuf = (uint8_t *)buf;
        ops.oobbuf = NULL;
@@ -2096,7 +2708,7 @@ static int nand_do_write_oob(struct mtd_info *mtd, loff_t to,
        int chipnr, page, status, len;
        struct nand_chip *chip = mtd->priv;
 
-       MTDDEBUG(MTD_DEBUG_LEVEL3, "%s: to = 0x%08x, len = %i\n",
+       pr_debug("%s: to = 0x%08x, len = %i\n",
                         __func__, (unsigned int)to, (int)ops->ooblen);
 
        if (ops->mode == MTD_OPS_AUTO_OOB)
@@ -2106,14 +2718,14 @@ static int nand_do_write_oob(struct mtd_info *mtd, loff_t to,
 
        /* Do not allow write past end of page */
        if ((ops->ooboffs + ops->ooblen) > len) {
-               MTDDEBUG(MTD_DEBUG_LEVEL0, "%s: Attempt to write "
-                               "past end of page\n", __func__);
+               pr_debug("%s: attempt to write past end of page\n",
+                               __func__);
                return -EINVAL;
        }
 
        if (unlikely(ops->ooboffs >= len)) {
-               MTDDEBUG(MTD_DEBUG_LEVEL0, "%s: Attempt to start "
-                               "write outside oob\n", __func__);
+               pr_debug("%s: attempt to start write outside oob\n",
+                               __func__);
                return -EINVAL;
        }
 
@@ -2122,8 +2734,8 @@ static int nand_do_write_oob(struct mtd_info *mtd, loff_t to,
                     ops->ooboffs + ops->ooblen >
                        ((mtd->size >> chip->page_shift) -
                         (to >> chip->page_shift)) * len)) {
-               MTDDEBUG(MTD_DEBUG_LEVEL0, "%s: Attempt write beyond "
-                               "end of device\n", __func__);
+               pr_debug("%s: attempt to write beyond end of device\n",
+                               __func__);
                return -EINVAL;
        }
 
@@ -2142,8 +2754,10 @@ static int nand_do_write_oob(struct mtd_info *mtd, loff_t to,
        chip->cmdfunc(mtd, NAND_CMD_RESET, -1, -1);
 
        /* Check, if it is write protected */
-       if (nand_check_wp(mtd))
+       if (nand_check_wp(mtd)) {
+               chip->select_chip(mtd, -1);
                return -EROFS;
+       }
 
        /* Invalidate the page cache, if we write to the cached page */
        if (page == chip->pagebuf)
@@ -2156,6 +2770,8 @@ static int nand_do_write_oob(struct mtd_info *mtd, loff_t to,
        else
                status = chip->ecc.write_oob(mtd, chip, page & chip->pagemask);
 
+       chip->select_chip(mtd, -1);
+
        if (status)
                return status;
 
@@ -2173,19 +2789,18 @@ static int nand_do_write_oob(struct mtd_info *mtd, loff_t to,
 static int nand_write_oob(struct mtd_info *mtd, loff_t to,
                          struct mtd_oob_ops *ops)
 {
-       struct nand_chip *chip = mtd->priv;
        int ret = -ENOTSUPP;
 
        ops->retlen = 0;
 
        /* Do not allow writes past end of device */
        if (ops->datbuf && (to + ops->len) > mtd->size) {
-               MTDDEBUG(MTD_DEBUG_LEVEL0, "%s: Attempt write beyond "
-                               "end of device\n", __func__);
+               pr_debug("%s: attempt to write beyond end of device\n",
+                               __func__);
                return -EINVAL;
        }
 
-       nand_get_device(chip, mtd, FL_WRITING);
+       nand_get_device(mtd, FL_WRITING);
 
        switch (ops->mode) {
        case MTD_OPS_PLACE_OOB:
@@ -2222,24 +2837,6 @@ static void single_erase_cmd(struct mtd_info *mtd, int page)
        chip->cmdfunc(mtd, NAND_CMD_ERASE2, -1, -1);
 }
 
-/**
- * multi_erase_cmd - [GENERIC] AND specific block erase command function
- * @mtd: MTD device structure
- * @page: the page address of the block which will be erased
- *
- * AND multi block erase command function. Erase 4 consecutive blocks.
- */
-static void multi_erase_cmd(struct mtd_info *mtd, int page)
-{
-       struct nand_chip *chip = mtd->priv;
-       /* Send commands to erase a block */
-       chip->cmdfunc(mtd, NAND_CMD_ERASE1, -1, page++);
-       chip->cmdfunc(mtd, NAND_CMD_ERASE1, -1, page++);
-       chip->cmdfunc(mtd, NAND_CMD_ERASE1, -1, page++);
-       chip->cmdfunc(mtd, NAND_CMD_ERASE1, -1, page);
-       chip->cmdfunc(mtd, NAND_CMD_ERASE2, -1, -1);
-}
-
 /**
  * nand_erase - [MTD Interface] erase block(s)
  * @mtd: MTD device structure
@@ -2252,7 +2849,6 @@ static int nand_erase(struct mtd_info *mtd, struct erase_info *instr)
        return nand_erase_nand(mtd, instr, 0);
 }
 
-#define BBT_PAGE_MASK  0xffffff3f
 /**
  * nand_erase_nand - [INTERN] erase block(s)
  * @mtd: MTD device structure
@@ -2266,19 +2862,17 @@ int nand_erase_nand(struct mtd_info *mtd, struct erase_info *instr,
 {
        int page, status, pages_per_block, ret, chipnr;
        struct nand_chip *chip = mtd->priv;
-       loff_t rewrite_bbt[CONFIG_SYS_NAND_MAX_CHIPS] = {0};
-       unsigned int bbt_masked_page = 0xffffffff;
        loff_t len;
 
-       MTDDEBUG(MTD_DEBUG_LEVEL3, "%s: start = 0x%012llx, len = %llu\n",
-                               __func__, (unsigned long long)instr->addr,
-                               (unsigned long long)instr->len);
+       pr_debug("%s: start = 0x%012llx, len = %llu\n",
+                       __func__, (unsigned long long)instr->addr,
+                       (unsigned long long)instr->len);
 
        if (check_offs_len(mtd, instr->addr, instr->len))
                return -EINVAL;
 
        /* Grab the lock and see if the device is available */
-       nand_get_device(chip, mtd, FL_ERASING);
+       nand_get_device(mtd, FL_ERASING);
 
        /* Shift to get first page */
        page = (int)(instr->addr >> chip->page_shift);
@@ -2292,21 +2886,12 @@ int nand_erase_nand(struct mtd_info *mtd, struct erase_info *instr,
 
        /* Check, if it is write protected */
        if (nand_check_wp(mtd)) {
-               MTDDEBUG(MTD_DEBUG_LEVEL0, "%s: Device is write protected!!!\n",
-                                       __func__);
+               pr_debug("%s: device is write protected!\n",
+                               __func__);
                instr->state = MTD_ERASE_FAILED;
                goto erase_exit;
        }
 
-       /*
-        * If BBT requires refresh, set the BBT page mask to see if the BBT
-        * should be rewritten. Otherwise the mask is set to 0xffffffff which
-        * can not be matched. This is also done when the bbt is actually
-        * erased to avoid recursive updates.
-        */
-       if (chip->options & BBT_AUTO_REFRESH && !allowbbt)
-               bbt_masked_page = chip->bbt_td->pages[chipnr] & BBT_PAGE_MASK;
-
        /* Loop through the pages */
        len = instr->len;
 
@@ -2314,11 +2899,12 @@ int nand_erase_nand(struct mtd_info *mtd, struct erase_info *instr,
 
        while (len) {
                WATCHDOG_RESET();
+
                /* Check if we have a bad block, we do not erase bad blocks! */
-               if (!instr->scrub && nand_block_checkbad(mtd, ((loff_t) page) <<
+               if (nand_block_checkbad(mtd, ((loff_t) page) <<
                                        chip->page_shift, 0, allowbbt)) {
                        pr_warn("%s: attempt to erase a bad block at page 0x%08x\n",
-                                  __func__, page);
+                                   __func__, page);
                        instr->state = MTD_ERASE_FAILED;
                        goto erase_exit;
                }
@@ -2345,25 +2931,16 @@ int nand_erase_nand(struct mtd_info *mtd, struct erase_info *instr,
 
                /* See if block erase succeeded */
                if (status & NAND_STATUS_FAIL) {
-                       MTDDEBUG(MTD_DEBUG_LEVEL0, "%s: Failed erase, "
-                                       "page 0x%08x\n", __func__, page);
+                       pr_debug("%s: failed erase, page 0x%08x\n",
+                                       __func__, page);
                        instr->state = MTD_ERASE_FAILED;
                        instr->fail_addr =
                                ((loff_t)page << chip->page_shift);
                        goto erase_exit;
                }
 
-               /*
-                * If BBT requires refresh, set the BBT rewrite flag to the
-                * page being erased.
-                */
-               if (bbt_masked_page != 0xffffffff &&
-                   (page & BBT_PAGE_MASK) == bbt_masked_page)
-                       rewrite_bbt[chipnr] =
-                               ((loff_t)page << chip->page_shift);
-
                /* Increment page address and decrement length */
-               len -= (1 << chip->phys_erase_shift);
+               len -= (1ULL << chip->phys_erase_shift);
                page += pages_per_block;
 
                /* Check, if we cross a chip boundary */
@@ -2371,15 +2948,6 @@ int nand_erase_nand(struct mtd_info *mtd, struct erase_info *instr,
                        chipnr++;
                        chip->select_chip(mtd, -1);
                        chip->select_chip(mtd, chipnr);
-
-                       /*
-                        * If BBT requires refresh and BBT-PERCHIP, set the BBT
-                        * page mask to see if this BBT should be rewritten.
-                        */
-                       if (bbt_masked_page != 0xffffffff &&
-                           (chip->bbt_td->options & NAND_BBT_PERCHIP))
-                               bbt_masked_page = chip->bbt_td->pages[chipnr] &
-                                       BBT_PAGE_MASK;
                }
        }
        instr->state = MTD_ERASE_DONE;
@@ -2389,29 +2957,13 @@ erase_exit:
        ret = instr->state == MTD_ERASE_DONE ? 0 : -EIO;
 
        /* Deselect and wake up anyone waiting on the device */
+       chip->select_chip(mtd, -1);
        nand_release_device(mtd);
 
        /* Do call back function */
        if (!ret)
                mtd_erase_callback(instr);
 
-       /*
-        * If BBT requires refresh and erase was successful, rewrite any
-        * selected bad block tables.
-        */
-       if (bbt_masked_page == 0xffffffff || ret)
-               return ret;
-
-       for (chipnr = 0; chipnr < chip->numchips; chipnr++) {
-               if (!rewrite_bbt[chipnr])
-                       continue;
-               /* Update the BBT for chip */
-               MTDDEBUG(MTD_DEBUG_LEVEL0, "%s: nand_update_bbt "
-                       "(%d:0x%0llx 0x%0x)\n", __func__, chipnr,
-                       rewrite_bbt[chipnr], chip->bbt_td->pages[chipnr]);
-               nand_update_bbt(mtd, rewrite_bbt[chipnr]);
-       }
-
        /* Return more or less happy */
        return ret;
 }
@@ -2424,12 +2976,10 @@ erase_exit:
  */
 static void nand_sync(struct mtd_info *mtd)
 {
-       struct nand_chip *chip = mtd->priv;
-
-       MTDDEBUG(MTD_DEBUG_LEVEL3, "%s: called\n", __func__);
+       pr_debug("%s: called\n", __func__);
 
        /* Grab the lock and see if the device is available */
-       nand_get_device(chip, mtd, FL_SYNCING);
+       nand_get_device(mtd, FL_SYNCING);
        /* Release it and go back */
        nand_release_device(mtd);
 }
@@ -2451,7 +3001,6 @@ static int nand_block_isbad(struct mtd_info *mtd, loff_t offs)
  */
 static int nand_block_markbad(struct mtd_info *mtd, loff_t ofs)
 {
-       struct nand_chip *chip = mtd->priv;
        int ret;
 
        ret = nand_block_isbad(mtd, ofs);
@@ -2462,10 +3011,10 @@ static int nand_block_markbad(struct mtd_info *mtd, loff_t ofs)
                return ret;
        }
 
-       return chip->block_markbad(mtd, ofs);
+       return nand_block_markbad_lowlevel(mtd, ofs);
 }
 
- /**
+/**
  * nand_onfi_set_features- [REPLACEABLE] set features for ONFI nand
  * @mtd: MTD device structure
  * @chip: nand chip info structure
@@ -2476,12 +3025,19 @@ static int nand_onfi_set_features(struct mtd_info *mtd, struct nand_chip *chip,
                        int addr, uint8_t *subfeature_param)
 {
        int status;
+       int i;
 
-       if (!chip->onfi_version)
+#ifdef CONFIG_SYS_NAND_ONFI_DETECTION
+       if (!chip->onfi_version ||
+           !(le16_to_cpu(chip->onfi_params.opt_cmd)
+             & ONFI_OPT_CMD_SET_GET_FEATURES))
                return -EINVAL;
+#endif
 
        chip->cmdfunc(mtd, NAND_CMD_SET_FEATURES, addr, -1);
-       chip->write_buf(mtd, subfeature_param, ONFI_SUBFEATURE_PARAM_LEN);
+       for (i = 0; i < ONFI_SUBFEATURE_PARAM_LEN; ++i)
+               chip->write_byte(mtd, subfeature_param[i]);
+
        status = chip->waitfunc(mtd, chip);
        if (status & NAND_STATUS_FAIL)
                return -EIO;
@@ -2498,17 +3054,50 @@ static int nand_onfi_set_features(struct mtd_info *mtd, struct nand_chip *chip,
 static int nand_onfi_get_features(struct mtd_info *mtd, struct nand_chip *chip,
                        int addr, uint8_t *subfeature_param)
 {
-       if (!chip->onfi_version)
+       int i;
+
+#ifdef CONFIG_SYS_NAND_ONFI_DETECTION
+       if (!chip->onfi_version ||
+           !(le16_to_cpu(chip->onfi_params.opt_cmd)
+             & ONFI_OPT_CMD_SET_GET_FEATURES))
                return -EINVAL;
+#endif
 
        /* clear the sub feature parameters */
        memset(subfeature_param, 0, ONFI_SUBFEATURE_PARAM_LEN);
 
        chip->cmdfunc(mtd, NAND_CMD_GET_FEATURES, addr, -1);
-       chip->read_buf(mtd, subfeature_param, ONFI_SUBFEATURE_PARAM_LEN);
+       for (i = 0; i < ONFI_SUBFEATURE_PARAM_LEN; ++i)
+               *subfeature_param++ = chip->read_byte(mtd);
        return 0;
 }
 
+#ifndef __UBOOT__
+/**
+ * nand_suspend - [MTD Interface] Suspend the NAND flash
+ * @mtd: MTD device structure
+ */
+static int nand_suspend(struct mtd_info *mtd)
+{
+       return nand_get_device(mtd, FL_PM_SUSPENDED);
+}
+
+/**
+ * nand_resume - [MTD Interface] Resume the NAND flash
+ * @mtd: MTD device structure
+ */
+static void nand_resume(struct mtd_info *mtd)
+{
+       struct nand_chip *chip = mtd->priv;
+
+       if (chip->state == FL_PM_SUSPENDED)
+               nand_release_device(mtd);
+       else
+               pr_err("%s called for a chip which is not in suspended state\n",
+                       __func__);
+}
+#endif
+
 /* Set default functions */
 static void nand_set_defaults(struct nand_chip *chip, int busw)
 {
@@ -2526,7 +3115,15 @@ static void nand_set_defaults(struct nand_chip *chip, int busw)
 
        if (!chip->select_chip)
                chip->select_chip = nand_select_chip;
-       if (!chip->read_byte)
+
+       /* set for ONFI nand */
+       if (!chip->onfi_set_features)
+               chip->onfi_set_features = nand_onfi_set_features;
+       if (!chip->onfi_get_features)
+               chip->onfi_get_features = nand_onfi_get_features;
+
+       /* If called twice, pointers that depend on busw may need to be reset */
+       if (!chip->read_byte || chip->read_byte == nand_read_byte)
                chip->read_byte = busw ? nand_read_byte16 : nand_read_byte;
        if (!chip->read_word)
                chip->read_word = nand_read_word;
@@ -2534,21 +3131,36 @@ static void nand_set_defaults(struct nand_chip *chip, int busw)
                chip->block_bad = nand_block_bad;
        if (!chip->block_markbad)
                chip->block_markbad = nand_default_block_markbad;
-       if (!chip->write_buf)
+       if (!chip->write_buf || chip->write_buf == nand_write_buf)
                chip->write_buf = busw ? nand_write_buf16 : nand_write_buf;
-       if (!chip->read_buf)
+       if (!chip->write_byte || chip->write_byte == nand_write_byte)
+               chip->write_byte = busw ? nand_write_byte16 : nand_write_byte;
+       if (!chip->read_buf || chip->read_buf == nand_read_buf)
                chip->read_buf = busw ? nand_read_buf16 : nand_read_buf;
-       if (!chip->verify_buf)
-               chip->verify_buf = busw ? nand_verify_buf16 : nand_verify_buf;
        if (!chip->scan_bbt)
                chip->scan_bbt = nand_default_bbt;
-       if (!chip->controller)
+#ifdef __UBOOT__
+#if defined(CONFIG_MTD_NAND_VERIFY_WRITE)
+       if (!chip->verify_buf)
+               chip->verify_buf = busw ? nand_verify_buf16 : nand_verify_buf;
+#endif
+#endif
+
+       if (!chip->controller) {
                chip->controller = &chip->hwcontrol;
+               spin_lock_init(&chip->controller->lock);
+               init_waitqueue_head(&chip->controller->wq);
+       }
+
 }
 
 #ifdef CONFIG_SYS_NAND_ONFI_DETECTION
 /* Sanitize ONFI strings so we can safely print them */
+#ifndef __UBOOT__
+static void sanitize_string(uint8_t *s, size_t len)
+#else
 static void sanitize_string(char *s, size_t len)
+#endif
 {
        ssize_t i;
 
@@ -2577,6 +3189,105 @@ static u16 onfi_crc16(u16 crc, u8 const *p, size_t len)
        return crc;
 }
 
+/* Parse the Extended Parameter Page. */
+static int nand_flash_detect_ext_param_page(struct mtd_info *mtd,
+               struct nand_chip *chip, struct nand_onfi_params *p)
+{
+       struct onfi_ext_param_page *ep;
+       struct onfi_ext_section *s;
+       struct onfi_ext_ecc_info *ecc;
+       uint8_t *cursor;
+       int ret = -EINVAL;
+       int len;
+       int i;
+
+       len = le16_to_cpu(p->ext_param_page_length) * 16;
+       ep = kmalloc(len, GFP_KERNEL);
+       if (!ep)
+               return -ENOMEM;
+
+       /* Send our own NAND_CMD_PARAM. */
+       chip->cmdfunc(mtd, NAND_CMD_PARAM, 0, -1);
+
+       /* Use the Change Read Column command to skip the ONFI param pages. */
+       chip->cmdfunc(mtd, NAND_CMD_RNDOUT,
+                       sizeof(*p) * p->num_of_param_pages , -1);
+
+       /* Read out the Extended Parameter Page. */
+       chip->read_buf(mtd, (uint8_t *)ep, len);
+       if ((onfi_crc16(ONFI_CRC_BASE, ((uint8_t *)ep) + 2, len - 2)
+               != le16_to_cpu(ep->crc))) {
+               pr_debug("fail in the CRC.\n");
+               goto ext_out;
+       }
+
+       /*
+        * Check the signature.
+        * Do not strictly follow the ONFI spec, maybe changed in future.
+        */
+#ifndef __UBOOT__
+       if (strncmp(ep->sig, "EPPS", 4)) {
+#else
+       if (strncmp((char *)ep->sig, "EPPS", 4)) {
+#endif
+               pr_debug("The signature is invalid.\n");
+               goto ext_out;
+       }
+
+       /* find the ECC section. */
+       cursor = (uint8_t *)(ep + 1);
+       for (i = 0; i < ONFI_EXT_SECTION_MAX; i++) {
+               s = ep->sections + i;
+               if (s->type == ONFI_SECTION_TYPE_2)
+                       break;
+               cursor += s->length * 16;
+       }
+       if (i == ONFI_EXT_SECTION_MAX) {
+               pr_debug("We can not find the ECC section.\n");
+               goto ext_out;
+       }
+
+       /* get the info we want. */
+       ecc = (struct onfi_ext_ecc_info *)cursor;
+
+       if (!ecc->codeword_size) {
+               pr_debug("Invalid codeword size\n");
+               goto ext_out;
+       }
+
+       chip->ecc_strength_ds = ecc->ecc_bits;
+       chip->ecc_step_ds = 1 << ecc->codeword_size;
+       ret = 0;
+
+ext_out:
+       kfree(ep);
+       return ret;
+}
+
+static int nand_setup_read_retry_micron(struct mtd_info *mtd, int retry_mode)
+{
+       struct nand_chip *chip = mtd->priv;
+       uint8_t feature[ONFI_SUBFEATURE_PARAM_LEN] = {retry_mode};
+
+       return chip->onfi_set_features(mtd, chip, ONFI_FEATURE_ADDR_READ_RETRY,
+                       feature);
+}
+
+/*
+ * Configure chip properties from Micron vendor-specific ONFI table
+ */
+static void nand_onfi_detect_micron(struct nand_chip *chip,
+               struct nand_onfi_params *p)
+{
+       struct nand_onfi_vendor_micron *micron = (void *)p->vendor;
+
+       if (le16_to_cpu(p->vendor_revision) < 1)
+               return;
+
+       chip->read_retries = micron->read_retry_options;
+       chip->setup_read_retry = nand_setup_read_retry_micron;
+}
+
 /*
  * Check if the NAND chip is ONFI compliant, returns 1 if it is, 0 otherwise.
  */
@@ -2593,19 +3304,29 @@ static int nand_flash_detect_onfi(struct mtd_info *mtd, struct nand_chip *chip,
                chip->read_byte(mtd) != 'F' || chip->read_byte(mtd) != 'I')
                return 0;
 
+       /*
+        * ONFI must be probed in 8-bit mode or with NAND_BUSWIDTH_AUTO, not
+        * with NAND_BUSWIDTH_16
+        */
+       if (chip->options & NAND_BUSWIDTH_16) {
+               pr_err("ONFI cannot be probed in 16-bit mode; aborting\n");
+               return 0;
+       }
+
        chip->cmdfunc(mtd, NAND_CMD_PARAM, 0, -1);
        for (i = 0; i < 3; i++) {
                for (j = 0; j < sizeof(*p); j++)
                        ((uint8_t *)p)[j] = chip->read_byte(mtd);
                if (onfi_crc16(ONFI_CRC_BASE, (uint8_t *)p, 254) ==
                                le16_to_cpu(p->crc)) {
-                       pr_info("ONFI param page %d valid\n", i);
                        break;
                }
        }
 
-       if (i == 3)
+       if (i == 3) {
+               pr_err("Could not find valid ONFI parameter page; aborting\n");
                return 0;
+       }
 
        /* Check version */
        val = le16_to_cpu(p->revision);
@@ -2619,11 +3340,9 @@ static int nand_flash_detect_onfi(struct mtd_info *mtd, struct nand_chip *chip,
                chip->onfi_version = 20;
        else if (val & (1 << 1))
                chip->onfi_version = 10;
-       else
-               chip->onfi_version = 0;
 
        if (!chip->onfi_version) {
-               pr_info("%s: unsupported ONFI version: %d\n", __func__, val);
+               pr_info("unsupported ONFI version: %d\n", val);
                return 0;
        }
 
@@ -2631,21 +3350,58 @@ static int nand_flash_detect_onfi(struct mtd_info *mtd, struct nand_chip *chip,
        sanitize_string(p->model, sizeof(p->model));
        if (!mtd->name)
                mtd->name = p->model;
+
        mtd->writesize = le32_to_cpu(p->byte_per_page);
-       mtd->erasesize = le32_to_cpu(p->pages_per_block) * mtd->writesize;
+
+       /*
+        * pages_per_block and blocks_per_lun may not be a power-of-2 size
+        * (don't ask me who thought of this...). MTD assumes that these
+        * dimensions will be power-of-2, so just truncate the remaining area.
+        */
+       mtd->erasesize = 1 << (fls(le32_to_cpu(p->pages_per_block)) - 1);
+       mtd->erasesize *= mtd->writesize;
+
        mtd->oobsize = le16_to_cpu(p->spare_bytes_per_page);
-       chip->chipsize = le32_to_cpu(p->blocks_per_lun);
+
+       /* See erasesize comment */
+       chip->chipsize = 1 << (fls(le32_to_cpu(p->blocks_per_lun)) - 1);
        chip->chipsize *= (uint64_t)mtd->erasesize * p->lun_count;
-       *busw = 0;
-       if (le16_to_cpu(p->features) & 1)
+       chip->bits_per_cell = p->bits_per_cell;
+
+       if (onfi_feature(chip) & ONFI_FEATURE_16_BIT_BUS)
                *busw = NAND_BUSWIDTH_16;
+       else
+               *busw = 0;
+
+       if (p->ecc_bits != 0xff) {
+               chip->ecc_strength_ds = p->ecc_bits;
+               chip->ecc_step_ds = 512;
+       } else if (chip->onfi_version >= 21 &&
+               (onfi_feature(chip) & ONFI_FEATURE_EXT_PARAM_PAGE)) {
+
+               /*
+                * The nand_flash_detect_ext_param_page() uses the
+                * Change Read Column command which maybe not supported
+                * by the chip->cmdfunc. So try to update the chip->cmdfunc
+                * now. We do not replace user supplied command function.
+                */
+               if (mtd->writesize > 512 && chip->cmdfunc == nand_command)
+                       chip->cmdfunc = nand_command_lp;
+
+               /* The Extended Parameter Page is supported since ONFI 2.1. */
+               if (nand_flash_detect_ext_param_page(mtd, chip, p))
+                       pr_warn("Failed to detect ONFI extended param page\n");
+       } else {
+               pr_warn("Could not retrieve ONFI ECC requirements\n");
+       }
+
+       if (p->jedec_id == NAND_MFR_MICRON)
+               nand_onfi_detect_micron(chip, p);
 
-       pr_info("ONFI flash detected\n");
        return 1;
 }
 #else
-static inline int nand_flash_detect_onfi(struct mtd_info *mtd,
-                                       struct nand_chip *chip,
+static int nand_flash_detect_onfi(struct mtd_info *mtd, struct nand_chip *chip,
                                        int *busw)
 {
        return 0;
@@ -2660,7 +3416,7 @@ static inline int nand_flash_detect_onfi(struct mtd_info *mtd,
  *
  * Check if an ID string is repeated within a given sequence of bytes at
  * specific repetition interval period (e.g., {0x20,0x01,0x7F,0x20} has a
- * period of 2). This is a helper function for nand_id_len(). Returns non-zero
+ * period of 3). This is a helper function for nand_id_len(). Returns non-zero
  * if the repetition has a period of @period; otherwise, returns zero.
  */
 static int nand_id_has_period(u8 *id_data, int arrlen, int period)
@@ -2711,6 +3467,16 @@ static int nand_id_len(u8 *id_data, int arrlen)
        return arrlen;
 }
 
+/* Extract the bits of per cell from the 3rd byte of the extended ID */
+static int nand_get_bits_per_cell(u8 cellinfo)
+{
+       int bits;
+
+       bits = cellinfo & NAND_CI_CELLTYPE_MSK;
+       bits >>= NAND_CI_CELLTYPE_SHIFT;
+       return bits + 1;
+}
+
 /*
  * Many new NAND share similar device ID codes, which represent the size of the
  * chip. The rest of the parameters must be decoded according to generic or
@@ -2721,7 +3487,7 @@ static void nand_decode_ext_id(struct mtd_info *mtd, struct nand_chip *chip,
 {
        int extid, id_len;
        /* The 3rd id byte holds MLC / multichip data */
-       chip->cellinfo = id_data[2];
+       chip->bits_per_cell = nand_get_bits_per_cell(id_data[2]);
        /* The 4th id byte is the important one */
        extid = id_data[3];
 
@@ -2737,8 +3503,7 @@ static void nand_decode_ext_id(struct mtd_info *mtd, struct nand_chip *chip,
         * ID to decide what to do.
         */
        if (id_len == 6 && id_data[0] == NAND_MFR_SAMSUNG &&
-                       (chip->cellinfo & NAND_CI_CELLTYPE_MSK) &&
-                       id_data[5] != 0x00) {
+                       !nand_is_slc(chip) && id_data[5] != 0x00) {
                /* Calc pagesize */
                mtd->writesize = 2048 << (extid & 0x03);
                extid >>= 2;
@@ -2760,9 +3525,12 @@ static void nand_decode_ext_id(struct mtd_info *mtd, struct nand_chip *chip,
                        mtd->oobsize = 512;
                        break;
                case 6:
-               default: /* Other cases are "reserved" (unknown) */
                        mtd->oobsize = 640;
                        break;
+               case 7:
+               default: /* Other cases are "reserved" (unknown) */
+                       mtd->oobsize = 1024;
+                       break;
                }
                extid >>= 2;
                /* Calc blocksize */
@@ -2770,7 +3538,7 @@ static void nand_decode_ext_id(struct mtd_info *mtd, struct nand_chip *chip,
                        (((extid >> 1) & 0x04) | (extid & 0x03));
                *busw = 0;
        } else if (id_len == 6 && id_data[0] == NAND_MFR_HYNIX &&
-                       (chip->cellinfo & NAND_CI_CELLTYPE_MSK)) {
+                       !nand_is_slc(chip)) {
                unsigned int tmp;
 
                /* Calc pagesize */
@@ -2823,16 +3591,32 @@ static void nand_decode_ext_id(struct mtd_info *mtd, struct nand_chip *chip,
                extid >>= 2;
                /* Get buswidth information */
                *busw = (extid & 0x01) ? NAND_BUSWIDTH_16 : 0;
+
+               /*
+                * Toshiba 24nm raw SLC (i.e., not BENAND) have 32B OOB per
+                * 512B page. For Toshiba SLC, we decode the 5th/6th byte as
+                * follows:
+                * - ID byte 6, bits[2:0]: 100b -> 43nm, 101b -> 32nm,
+                *                         110b -> 24nm
+                * - ID byte 5, bit[7]:    1 -> BENAND, 0 -> raw SLC
+                */
+               if (id_len >= 6 && id_data[0] == NAND_MFR_TOSHIBA &&
+                               nand_is_slc(chip) &&
+                               (id_data[5] & 0x7) == 0x6 /* 24nm */ &&
+                               !(id_data[4] & 0x80) /* !BENAND */) {
+                       mtd->oobsize = 32 * mtd->writesize >> 9;
+               }
+
        }
 }
 
- /*
+/*
  * Old devices have chip data hardcoded in the device ID table. nand_decode_id
  * decodes a matching ID table entry and assigns the MTD size parameters for
  * the chip.
  */
 static void nand_decode_id(struct mtd_info *mtd, struct nand_chip *chip,
-                               const struct nand_flash_dev *type, u8 id_data[8],
+                               struct nand_flash_dev *type, u8 id_data[8],
                                int *busw)
 {
        int maf_id = id_data[0];
@@ -2842,6 +3626,9 @@ static void nand_decode_id(struct mtd_info *mtd, struct nand_chip *chip,
        mtd->oobsize = mtd->writesize / 32;
        *busw = type->options & NAND_BUSWIDTH_16;
 
+       /* All legacy ID NAND are small-page, SLC */
+       chip->bits_per_cell = 1;
+
        /*
         * Check for Spansion/AMD ID + repeating 5th, 6th byte since
         * some Spansion chips have erasesize that conflicts with size
@@ -2856,7 +3643,7 @@ static void nand_decode_id(struct mtd_info *mtd, struct nand_chip *chip,
        }
 }
 
- /*
+/*
  * Set the bad block marker/indicator (BBM/BBI) patterns according to some
  * heuristic patterns using various detected parameters (e.g., manufacturer,
  * page size, cell-type information).
@@ -2878,11 +3665,11 @@ static void nand_decode_bbm_options(struct mtd_info *mtd,
         * Micron devices with 2KiB pages and on SLC Samsung, Hynix, Toshiba,
         * AMD/Spansion, and Macronix.  All others scan only the first page.
         */
-       if ((chip->cellinfo & NAND_CI_CELLTYPE_MSK) &&
+       if (!nand_is_slc(chip) &&
                        (maf_id == NAND_MFR_SAMSUNG ||
                         maf_id == NAND_MFR_HYNIX))
                chip->bbt_options |= NAND_BBT_SCANLASTPAGE;
-       else if ((!(chip->cellinfo & NAND_CI_CELLTYPE_MSK) &&
+       else if ((nand_is_slc(chip) &&
                                (maf_id == NAND_MFR_SAMSUNG ||
                                 maf_id == NAND_MFR_HYNIX ||
                                 maf_id == NAND_MFR_TOSHIBA ||
@@ -2893,16 +3680,48 @@ static void nand_decode_bbm_options(struct mtd_info *mtd,
                chip->bbt_options |= NAND_BBT_SCAN2NDPAGE;
 }
 
+static inline bool is_full_id_nand(struct nand_flash_dev *type)
+{
+       return type->id_len;
+}
+
+static bool find_full_id_nand(struct mtd_info *mtd, struct nand_chip *chip,
+                  struct nand_flash_dev *type, u8 *id_data, int *busw)
+{
+#ifndef __UBOOT__
+       if (!strncmp(type->id, id_data, type->id_len)) {
+#else
+       if (!strncmp((char *)type->id, (char *)id_data, type->id_len)) {
+#endif
+               mtd->writesize = type->pagesize;
+               mtd->erasesize = type->erasesize;
+               mtd->oobsize = type->oobsize;
+
+               chip->bits_per_cell = nand_get_bits_per_cell(id_data[2]);
+               chip->chipsize = (uint64_t)type->chipsize << 20;
+               chip->options |= type->options;
+               chip->ecc_strength_ds = NAND_ECC_STRENGTH(type);
+               chip->ecc_step_ds = NAND_ECC_STEP(type);
+
+               *busw = type->options & NAND_BUSWIDTH_16;
+
+               if (!mtd->name)
+                       mtd->name = type->name;
+
+               return true;
+       }
+       return false;
+}
+
 /*
  * Get the flash and manufacturer id and lookup if the type is supported.
  */
-static const struct nand_flash_dev *nand_get_flash_type(struct mtd_info *mtd,
+static struct nand_flash_dev *nand_get_flash_type(struct mtd_info *mtd,
                                                  struct nand_chip *chip,
                                                  int busw,
                                                  int *maf_id, int *dev_id,
-                                                 const struct nand_flash_dev *type)
+                                                 struct nand_flash_dev *type)
 {
-       const char *name;
        int i, maf_idx;
        u8 id_data[8];
 
@@ -2936,8 +3755,7 @@ static const struct nand_flash_dev *nand_get_flash_type(struct mtd_info *mtd,
                id_data[i] = chip->read_byte(mtd);
 
        if (id_data[0] != *maf_id || id_data[1] != *dev_id) {
-               pr_info("%s: second ID read did not match "
-                       "%02x,%02x against %02x,%02x\n", __func__,
+               pr_info("second ID read did not match %02x,%02x against %02x,%02x\n",
                        *maf_id, *dev_id, id_data[0], id_data[1]);
                return ERR_PTR(-ENODEV);
        }
@@ -2945,9 +3763,14 @@ static const struct nand_flash_dev *nand_get_flash_type(struct mtd_info *mtd,
        if (!type)
                type = nand_flash_ids;
 
-       for (; type->name != NULL; type++)
-               if (*dev_id == type->id)
-                       break;
+       for (; type->name != NULL; type++) {
+               if (is_full_id_nand(type)) {
+                       if (find_full_id_nand(mtd, chip, type, id_data, &busw))
+                               goto ident_done;
+               } else if (*dev_id == type->dev_id) {
+                               break;
+               }
+       }
 
        chip->onfi_version = 0;
        if (!type->name || !type->pagesize) {
@@ -2973,7 +3796,7 @@ static const struct nand_flash_dev *nand_get_flash_type(struct mtd_info *mtd,
        } else {
                nand_decode_id(mtd, chip, type, id_data, &busw);
        }
-       /* Get chip options, preserve non chip based options */
+       /* Get chip options */
        chip->options |= type->options;
 
        /*
@@ -2990,15 +3813,19 @@ ident_done:
                        break;
        }
 
-       /*
-        * Check, if buswidth is correct. Hardware drivers should set
-        * chip correct!
-        */
-       if (busw != (chip->options & NAND_BUSWIDTH_16)) {
-               pr_info("NAND device: Manufacturer ID:"
-                       " 0x%02x, Chip ID: 0x%02x (%s %s)\n", *maf_id,
-                       *dev_id, nand_manuf_ids[maf_idx].name, mtd->name);
-               pr_warn("NAND bus width %d instead %d bit\n",
+       if (chip->options & NAND_BUSWIDTH_AUTO) {
+               WARN_ON(chip->options & NAND_BUSWIDTH_16);
+               chip->options |= busw;
+               nand_set_defaults(chip, busw);
+       } else if (busw != (chip->options & NAND_BUSWIDTH_16)) {
+               /*
+                * Check, if buswidth is correct. Hardware drivers should set
+                * chip correct!
+                */
+               pr_info("device found, Manufacturer ID: 0x%02x, Chip ID: 0x%02x\n",
+                       *maf_id, *dev_id);
+               pr_info("%s %s\n", nand_manuf_ids[maf_idx].name, mtd->name);
+               pr_warn("bus width %d instead %d bit\n",
                           (chip->options & NAND_BUSWIDTH_16) ? 16 : 8,
                           busw ? 16 : 8);
                return ERR_PTR(-EINVAL);
@@ -3021,28 +3848,23 @@ ident_done:
        }
 
        chip->badblockbits = 8;
-
-       /* Check for AND chips with 4 page planes */
-       if (chip->options & NAND_4PAGE_ARRAY)
-               chip->erase_cmd = multi_erase_cmd;
-       else
-               chip->erase_cmd = single_erase_cmd;
+       chip->erase_cmd = single_erase_cmd;
 
        /* Do not replace user supplied command function! */
        if (mtd->writesize > 512 && chip->cmdfunc == nand_command)
                chip->cmdfunc = nand_command_lp;
 
-       name = type->name;
+       pr_info("device found, Manufacturer ID: 0x%02x, Chip ID: 0x%02x\n",
+               *maf_id, *dev_id);
 #ifdef CONFIG_SYS_NAND_ONFI_DETECTION
-       if (chip->onfi_version)
-               name = chip->onfi_params.model;
+       pr_info("%s %s\n", nand_manuf_ids[maf_idx].name,
+               chip->onfi_version ? chip->onfi_params.model : type->name);
+#else
+       pr_info("%s %s\n", nand_manuf_ids[maf_idx].name, type->name);
 #endif
-       pr_info("NAND device: Manufacturer ID: 0x%02x, Chip ID: 0x%02x (%s %s),"
-               " page size: %d, OOB size: %d\n",
-               *maf_id, *dev_id, nand_manuf_ids[maf_idx].name,
-               name,
+       pr_info("%dMiB, %s, page size: %d, OOB size: %d\n",
+               (int)(chip->chipsize >> 20), nand_is_slc(chip) ? "SLC" : "MLC",
                mtd->writesize, mtd->oobsize);
-
        return type;
 }
 
@@ -3058,11 +3880,11 @@ ident_done:
  * The mtd->owner field must be set to the module of the caller.
  */
 int nand_scan_ident(struct mtd_info *mtd, int maxchips,
-                   const struct nand_flash_dev *table)
+                   struct nand_flash_dev *table)
 {
        int i, busw, nand_maf_id, nand_dev_id;
        struct nand_chip *chip = mtd->priv;
-       const struct nand_flash_dev *type;
+       struct nand_flash_dev *type;
 
        /* Get buswidth to select the correct functions */
        busw = chip->options & NAND_BUSWIDTH_16;
@@ -3074,13 +3896,14 @@ int nand_scan_ident(struct mtd_info *mtd, int maxchips,
                                &nand_maf_id, &nand_dev_id, table);
 
        if (IS_ERR(type)) {
-#ifndef CONFIG_SYS_NAND_QUIET_TEST
-               pr_warn("No NAND device found\n");
-#endif
+               if (!(chip->options & NAND_SCAN_SILENT_NODEV))
+                       pr_warn("No NAND device found\n");
                chip->select_chip(mtd, -1);
                return PTR_ERR(type);
        }
 
+       chip->select_chip(mtd, -1);
+
        /* Check for a chip array */
        for (i = 1; i < maxchips; i++) {
                chip->select_chip(mtd, i);
@@ -3090,12 +3913,16 @@ int nand_scan_ident(struct mtd_info *mtd, int maxchips,
                chip->cmdfunc(mtd, NAND_CMD_READID, 0x00, -1);
                /* Read manufacturer and device IDs */
                if (nand_maf_id != chip->read_byte(mtd) ||
-                   nand_dev_id != chip->read_byte(mtd))
+                   nand_dev_id != chip->read_byte(mtd)) {
+                       chip->select_chip(mtd, -1);
                        break;
+               }
+               chip->select_chip(mtd, -1);
        }
+
 #ifdef DEBUG
        if (i > 1)
-               pr_info("%d NAND chips detected\n", i);
+               pr_info("%d chips detected\n", i);
 #endif
 
        /* Store the number of chips and calc total size for mtd */
@@ -3104,6 +3931,7 @@ int nand_scan_ident(struct mtd_info *mtd, int maxchips,
 
        return 0;
 }
+EXPORT_SYMBOL(nand_scan_ident);
 
 
 /**
@@ -3118,14 +3946,14 @@ int nand_scan_tail(struct mtd_info *mtd)
 {
        int i;
        struct nand_chip *chip = mtd->priv;
+       struct nand_ecc_ctrl *ecc = &chip->ecc;
 
        /* New bad blocks should be marked in OOB, flash-based BBT, or both */
        BUG_ON((chip->bbt_options & NAND_BBT_NO_OOB_BBM) &&
                        !(chip->bbt_options & NAND_BBT_USE_FLASH));
 
        if (!(chip->options & NAND_OWN_BUFFERS))
-               chip->buffers = memalign(ARCH_DMA_MINALIGN,
-                                        sizeof(*chip->buffers));
+               chip->buffers = kmalloc(sizeof(*chip->buffers), GFP_KERNEL);
        if (!chip->buffers)
                return -ENOMEM;
 
@@ -3135,94 +3963,91 @@ int nand_scan_tail(struct mtd_info *mtd)
        /*
         * If no default placement scheme is given, select an appropriate one.
         */
-       if (!chip->ecc.layout && (chip->ecc.mode != NAND_ECC_SOFT_BCH)) {
+       if (!ecc->layout && (ecc->mode != NAND_ECC_SOFT_BCH)) {
                switch (mtd->oobsize) {
                case 8:
-                       chip->ecc.layout = &nand_oob_8;
+                       ecc->layout = &nand_oob_8;
                        break;
                case 16:
-                       chip->ecc.layout = &nand_oob_16;
+                       ecc->layout = &nand_oob_16;
                        break;
                case 64:
-                       chip->ecc.layout = &nand_oob_64;
+                       ecc->layout = &nand_oob_64;
                        break;
                case 128:
-                       chip->ecc.layout = &nand_oob_128;
+                       ecc->layout = &nand_oob_128;
                        break;
                default:
                        pr_warn("No oob scheme defined for oobsize %d\n",
                                   mtd->oobsize);
+                       BUG();
                }
        }
 
        if (!chip->write_page)
                chip->write_page = nand_write_page;
 
-       /* set for ONFI nand */
-       if (!chip->onfi_set_features)
-               chip->onfi_set_features = nand_onfi_set_features;
-       if (!chip->onfi_get_features)
-               chip->onfi_get_features = nand_onfi_get_features;
-
        /*
         * Check ECC mode, default to software if 3byte/512byte hardware ECC is
         * selected and we have 256 byte pagesize fallback to software ECC
         */
 
-       switch (chip->ecc.mode) {
+       switch (ecc->mode) {
        case NAND_ECC_HW_OOB_FIRST:
                /* Similar to NAND_ECC_HW, but a separate read_page handle */
-               if (!chip->ecc.calculate || !chip->ecc.correct ||
-                    !chip->ecc.hwctl) {
+               if (!ecc->calculate || !ecc->correct || !ecc->hwctl) {
                        pr_warn("No ECC functions supplied; "
                                   "hardware ECC not possible\n");
                        BUG();
                }
-               if (!chip->ecc.read_page)
-                       chip->ecc.read_page = nand_read_page_hwecc_oob_first;
+               if (!ecc->read_page)
+                       ecc->read_page = nand_read_page_hwecc_oob_first;
 
        case NAND_ECC_HW:
                /* Use standard hwecc read page function? */
-               if (!chip->ecc.read_page)
-                       chip->ecc.read_page = nand_read_page_hwecc;
-               if (!chip->ecc.write_page)
-                       chip->ecc.write_page = nand_write_page_hwecc;
-               if (!chip->ecc.read_page_raw)
-                       chip->ecc.read_page_raw = nand_read_page_raw;
-               if (!chip->ecc.write_page_raw)
-                       chip->ecc.write_page_raw = nand_write_page_raw;
-               if (!chip->ecc.read_oob)
-                       chip->ecc.read_oob = nand_read_oob_std;
-               if (!chip->ecc.write_oob)
-                       chip->ecc.write_oob = nand_write_oob_std;
+               if (!ecc->read_page)
+                       ecc->read_page = nand_read_page_hwecc;
+               if (!ecc->write_page)
+                       ecc->write_page = nand_write_page_hwecc;
+               if (!ecc->read_page_raw)
+                       ecc->read_page_raw = nand_read_page_raw;
+               if (!ecc->write_page_raw)
+                       ecc->write_page_raw = nand_write_page_raw;
+               if (!ecc->read_oob)
+                       ecc->read_oob = nand_read_oob_std;
+               if (!ecc->write_oob)
+                       ecc->write_oob = nand_write_oob_std;
+               if (!ecc->read_subpage)
+                       ecc->read_subpage = nand_read_subpage;
+               if (!ecc->write_subpage)
+                       ecc->write_subpage = nand_write_subpage_hwecc;
 
        case NAND_ECC_HW_SYNDROME:
-               if ((!chip->ecc.calculate || !chip->ecc.correct ||
-                    !chip->ecc.hwctl) &&
-                   (!chip->ecc.read_page ||
-                    chip->ecc.read_page == nand_read_page_hwecc ||
-                    !chip->ecc.write_page ||
-                    chip->ecc.write_page == nand_write_page_hwecc)) {
+               if ((!ecc->calculate || !ecc->correct || !ecc->hwctl) &&
+                   (!ecc->read_page ||
+                    ecc->read_page == nand_read_page_hwecc ||
+                    !ecc->write_page ||
+                    ecc->write_page == nand_write_page_hwecc)) {
                        pr_warn("No ECC functions supplied; "
                                   "hardware ECC not possible\n");
                        BUG();
                }
                /* Use standard syndrome read/write page function? */
-               if (!chip->ecc.read_page)
-                       chip->ecc.read_page = nand_read_page_syndrome;
-               if (!chip->ecc.write_page)
-                       chip->ecc.write_page = nand_write_page_syndrome;
-               if (!chip->ecc.read_page_raw)
-                       chip->ecc.read_page_raw = nand_read_page_raw_syndrome;
-               if (!chip->ecc.write_page_raw)
-                       chip->ecc.write_page_raw = nand_write_page_raw_syndrome;
-               if (!chip->ecc.read_oob)
-                       chip->ecc.read_oob = nand_read_oob_syndrome;
-               if (!chip->ecc.write_oob)
-                       chip->ecc.write_oob = nand_write_oob_syndrome;
-
-               if (mtd->writesize >= chip->ecc.size) {
-                       if (!chip->ecc.strength) {
+               if (!ecc->read_page)
+                       ecc->read_page = nand_read_page_syndrome;
+               if (!ecc->write_page)
+                       ecc->write_page = nand_write_page_syndrome;
+               if (!ecc->read_page_raw)
+                       ecc->read_page_raw = nand_read_page_raw_syndrome;
+               if (!ecc->write_page_raw)
+                       ecc->write_page_raw = nand_write_page_raw_syndrome;
+               if (!ecc->read_oob)
+                       ecc->read_oob = nand_read_oob_syndrome;
+               if (!ecc->write_oob)
+                       ecc->write_oob = nand_write_oob_syndrome;
+
+               if (mtd->writesize >= ecc->size) {
+                       if (!ecc->strength) {
                                pr_warn("Driver must set ecc.strength when using hardware ECC\n");
                                BUG();
                        }
@@ -3230,109 +4055,107 @@ int nand_scan_tail(struct mtd_info *mtd)
                }
                pr_warn("%d byte HW ECC not possible on "
                           "%d byte page size, fallback to SW ECC\n",
-                          chip->ecc.size, mtd->writesize);
-               chip->ecc.mode = NAND_ECC_SOFT;
+                          ecc->size, mtd->writesize);
+               ecc->mode = NAND_ECC_SOFT;
 
        case NAND_ECC_SOFT:
-               chip->ecc.calculate = nand_calculate_ecc;
-               chip->ecc.correct = nand_correct_data;
-               chip->ecc.read_page = nand_read_page_swecc;
-               chip->ecc.read_subpage = nand_read_subpage;
-               chip->ecc.write_page = nand_write_page_swecc;
-               chip->ecc.read_page_raw = nand_read_page_raw;
-               chip->ecc.write_page_raw = nand_write_page_raw;
-               chip->ecc.read_oob = nand_read_oob_std;
-               chip->ecc.write_oob = nand_write_oob_std;
-               if (!chip->ecc.size)
-                       chip->ecc.size = 256;
-               chip->ecc.bytes = 3;
-               chip->ecc.strength = 1;
+               ecc->calculate = nand_calculate_ecc;
+               ecc->correct = nand_correct_data;
+               ecc->read_page = nand_read_page_swecc;
+               ecc->read_subpage = nand_read_subpage;
+               ecc->write_page = nand_write_page_swecc;
+               ecc->read_page_raw = nand_read_page_raw;
+               ecc->write_page_raw = nand_write_page_raw;
+               ecc->read_oob = nand_read_oob_std;
+               ecc->write_oob = nand_write_oob_std;
+               if (!ecc->size)
+                       ecc->size = 256;
+               ecc->bytes = 3;
+               ecc->strength = 1;
                break;
 
        case NAND_ECC_SOFT_BCH:
                if (!mtd_nand_has_bch()) {
                        pr_warn("CONFIG_MTD_ECC_BCH not enabled\n");
-                       return -EINVAL;
+                       BUG();
                }
-               chip->ecc.calculate = nand_bch_calculate_ecc;
-               chip->ecc.correct = nand_bch_correct_data;
-               chip->ecc.read_page = nand_read_page_swecc;
-               chip->ecc.read_subpage = nand_read_subpage;
-               chip->ecc.write_page = nand_write_page_swecc;
-               chip->ecc.read_page_raw = nand_read_page_raw;
-               chip->ecc.write_page_raw = nand_write_page_raw;
-               chip->ecc.read_oob = nand_read_oob_std;
-               chip->ecc.write_oob = nand_write_oob_std;
+               ecc->calculate = nand_bch_calculate_ecc;
+               ecc->correct = nand_bch_correct_data;
+               ecc->read_page = nand_read_page_swecc;
+               ecc->read_subpage = nand_read_subpage;
+               ecc->write_page = nand_write_page_swecc;
+               ecc->read_page_raw = nand_read_page_raw;
+               ecc->write_page_raw = nand_write_page_raw;
+               ecc->read_oob = nand_read_oob_std;
+               ecc->write_oob = nand_write_oob_std;
                /*
                 * Board driver should supply ecc.size and ecc.bytes values to
                 * select how many bits are correctable; see nand_bch_init()
                 * for details. Otherwise, default to 4 bits for large page
                 * devices.
                 */
-               if (!chip->ecc.size && (mtd->oobsize >= 64)) {
-                       chip->ecc.size = 512;
-                       chip->ecc.bytes = 7;
+               if (!ecc->size && (mtd->oobsize >= 64)) {
+                       ecc->size = 512;
+                       ecc->bytes = 7;
                }
-               chip->ecc.priv = nand_bch_init(mtd,
-                                              chip->ecc.size,
-                                              chip->ecc.bytes,
-                                              &chip->ecc.layout);
-               if (!chip->ecc.priv)
+               ecc->priv = nand_bch_init(mtd, ecc->size, ecc->bytes,
+                                              &ecc->layout);
+               if (!ecc->priv) {
                        pr_warn("BCH ECC initialization failed!\n");
-               chip->ecc.strength =
-                       chip->ecc.bytes * 8 / fls(8 * chip->ecc.size);
+                       BUG();
+               }
+               ecc->strength = ecc->bytes * 8 / fls(8 * ecc->size);
                break;
 
        case NAND_ECC_NONE:
                pr_warn("NAND_ECC_NONE selected by board driver. "
-                       "This is not recommended !!\n");
-               chip->ecc.read_page = nand_read_page_raw;
-               chip->ecc.write_page = nand_write_page_raw;
-               chip->ecc.read_oob = nand_read_oob_std;
-               chip->ecc.read_page_raw = nand_read_page_raw;
-               chip->ecc.write_page_raw = nand_write_page_raw;
-               chip->ecc.write_oob = nand_write_oob_std;
-               chip->ecc.size = mtd->writesize;
-               chip->ecc.bytes = 0;
+                          "This is not recommended!\n");
+               ecc->read_page = nand_read_page_raw;
+               ecc->write_page = nand_write_page_raw;
+               ecc->read_oob = nand_read_oob_std;
+               ecc->read_page_raw = nand_read_page_raw;
+               ecc->write_page_raw = nand_write_page_raw;
+               ecc->write_oob = nand_write_oob_std;
+               ecc->size = mtd->writesize;
+               ecc->bytes = 0;
+               ecc->strength = 0;
                break;
 
        default:
-               pr_warn("Invalid NAND_ECC_MODE %d\n", chip->ecc.mode);
+               pr_warn("Invalid NAND_ECC_MODE %d\n", ecc->mode);
                BUG();
        }
 
        /* For many systems, the standard OOB write also works for raw */
-       if (!chip->ecc.read_oob_raw)
-               chip->ecc.read_oob_raw = chip->ecc.read_oob;
-       if (!chip->ecc.write_oob_raw)
-               chip->ecc.write_oob_raw = chip->ecc.write_oob;
+       if (!ecc->read_oob_raw)
+               ecc->read_oob_raw = ecc->read_oob;
+       if (!ecc->write_oob_raw)
+               ecc->write_oob_raw = ecc->write_oob;
 
        /*
         * The number of bytes available for a client to place data into
         * the out of band area.
         */
-       chip->ecc.layout->oobavail = 0;
-       for (i = 0; chip->ecc.layout->oobfree[i].length
-                       && i < ARRAY_SIZE(chip->ecc.layout->oobfree); i++)
-               chip->ecc.layout->oobavail +=
-                       chip->ecc.layout->oobfree[i].length;
-       mtd->oobavail = chip->ecc.layout->oobavail;
+       ecc->layout->oobavail = 0;
+       for (i = 0; ecc->layout->oobfree[i].length
+                       && i < ARRAY_SIZE(ecc->layout->oobfree); i++)
+               ecc->layout->oobavail += ecc->layout->oobfree[i].length;
+       mtd->oobavail = ecc->layout->oobavail;
 
        /*
         * Set the number of read / write steps for one page depending on ECC
         * mode.
         */
-       chip->ecc.steps = mtd->writesize / chip->ecc.size;
-       if (chip->ecc.steps * chip->ecc.size != mtd->writesize) {
+       ecc->steps = mtd->writesize / ecc->size;
+       if (ecc->steps * ecc->size != mtd->writesize) {
                pr_warn("Invalid ECC parameters\n");
                BUG();
        }
-       chip->ecc.total = chip->ecc.steps * chip->ecc.bytes;
+       ecc->total = ecc->steps * ecc->bytes;
 
        /* Allow subpage writes up to ecc.steps. Not possible for MLC flash */
-       if (!(chip->options & NAND_NO_SUBPAGE_WRITE) &&
-           !(chip->cellinfo & NAND_CI_CELLTYPE_MSK)) {
-               switch (chip->ecc.steps) {
+       if (!(chip->options & NAND_NO_SUBPAGE_WRITE) && nand_is_slc(chip)) {
+               switch (ecc->steps) {
                case 2:
                        mtd->subpage_sft = 1;
                        break;
@@ -3348,36 +4171,42 @@ int nand_scan_tail(struct mtd_info *mtd)
        /* Initialize state */
        chip->state = FL_READY;
 
-       /* De-select the device */
-       chip->select_chip(mtd, -1);
-
        /* Invalidate the pagebuffer reference */
        chip->pagebuf = -1;
 
        /* Large page NAND with SOFT_ECC should support subpage reads */
-       if ((chip->ecc.mode == NAND_ECC_SOFT) && (chip->page_shift > 9))
+       if ((ecc->mode == NAND_ECC_SOFT) && (chip->page_shift > 9))
                chip->options |= NAND_SUBPAGE_READ;
 
        /* Fill in remaining MTD driver data */
-       mtd->type = MTD_NANDFLASH;
+       mtd->type = nand_is_slc(chip) ? MTD_NANDFLASH : MTD_MLCNANDFLASH;
        mtd->flags = (chip->options & NAND_ROM) ? MTD_CAP_ROM :
                                                MTD_CAP_NANDFLASH;
        mtd->_erase = nand_erase;
+#ifndef __UBOOT__
        mtd->_point = NULL;
        mtd->_unpoint = NULL;
+#endif
        mtd->_read = nand_read;
        mtd->_write = nand_write;
+       mtd->_panic_write = panic_nand_write;
        mtd->_read_oob = nand_read_oob;
        mtd->_write_oob = nand_write_oob;
        mtd->_sync = nand_sync;
        mtd->_lock = NULL;
        mtd->_unlock = NULL;
+#ifndef __UBOOT__
+       mtd->_suspend = nand_suspend;
+       mtd->_resume = nand_resume;
+#endif
        mtd->_block_isbad = nand_block_isbad;
        mtd->_block_markbad = nand_block_markbad;
+       mtd->writebufsize = mtd->writesize;
 
        /* propagate ecc info to mtd_info */
-       mtd->ecclayout = chip->ecc.layout;
-       mtd->ecc_strength = chip->ecc.strength;
+       mtd->ecclayout = ecc->layout;
+       mtd->ecc_strength = ecc->strength;
+       mtd->ecc_step_size = ecc->size;
        /*
         * Initialize bitflip_threshold to its default prior scan_bbt() call.
         * scan_bbt() might invoke mtd_read(), thus bitflip_threshold must be
@@ -3388,10 +4217,24 @@ int nand_scan_tail(struct mtd_info *mtd)
 
        /* Check, if we should skip the bad block table scan */
        if (chip->options & NAND_SKIP_BBTSCAN)
-               chip->options |= NAND_BBT_SCANNED;
+               return 0;
 
-       return 0;
+       /* Build bad block table */
+       return chip->scan_bbt(mtd);
 }
+EXPORT_SYMBOL(nand_scan_tail);
+
+/*
+ * is_module_text_address() isn't exported, and it's mostly a pointless
+ * test if this is a module _anyway_ -- they'd have to try _really_ hard
+ * to call us from in-kernel code if the core NAND support is modular.
+ */
+#ifdef MODULE
+#define caller_is_module() (1)
+#else
+#define caller_is_module() \
+       is_module_text_address((unsigned long)__builtin_return_address(0))
+#endif
 
 /**
  * nand_scan - [NAND Interface] Scan for the NAND device
@@ -3407,12 +4250,20 @@ int nand_scan(struct mtd_info *mtd, int maxchips)
 {
        int ret;
 
+       /* Many callers got this wrong, so check for it for a while... */
+       if (!mtd->owner && caller_is_module()) {
+               pr_crit("%s called with NULL mtd->owner!\n", __func__);
+               BUG();
+       }
+
        ret = nand_scan_ident(mtd, maxchips, NULL);
        if (!ret)
                ret = nand_scan_tail(mtd);
        return ret;
 }
+EXPORT_SYMBOL(nand_scan);
 
+#ifndef __UBOOT__
 /**
  * nand_release - [NAND Interface] Free resources held by the NAND device
  * @mtd: MTD device structure
@@ -3424,10 +4275,7 @@ void nand_release(struct mtd_info *mtd)
        if (chip->ecc.mode == NAND_ECC_SOFT_BCH)
                nand_bch_free((struct nand_bch_control *)chip->ecc.priv);
 
-#ifdef CONFIG_MTD_PARTITIONS
-       /* Deregister partitions */
-       del_mtd_partitions(mtd);
-#endif
+       mtd_device_unregister(mtd);
 
        /* Free bad block table memory */
        kfree(chip->bbt);
@@ -3439,3 +4287,24 @@ void nand_release(struct mtd_info *mtd)
                        & NAND_BBT_DYNAMICSTRUCT)
                kfree(chip->badblock_pattern);
 }
+EXPORT_SYMBOL_GPL(nand_release);
+
+static int __init nand_base_init(void)
+{
+       led_trigger_register_simple("nand-disk", &nand_led_trigger);
+       return 0;
+}
+
+static void __exit nand_base_exit(void)
+{
+       led_trigger_unregister_simple(nand_led_trigger);
+}
+#endif
+
+module_init(nand_base_init);
+module_exit(nand_base_exit);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Steven J. Hill <sjhill@realitydiluted.com>");
+MODULE_AUTHOR("Thomas Gleixner <tglx@linutronix.de>");
+MODULE_DESCRIPTION("Generic NAND flash driver code");
index 8ef58451d522131ba3b9bec3c116531c94217681..c8f28c792b2a3e21c919be0a725aca19fe794a28 100644 (file)
  *
  */
 
-#include <common.h>
-#include <malloc.h>
-#include <linux/compat.h>
+#define __UBOOT__
+#ifndef __UBOOT__
+#include <linux/slab.h>
+#include <linux/types.h>
 #include <linux/mtd/mtd.h>
 #include <linux/mtd/bbm.h>
 #include <linux/mtd/nand.h>
 #include <linux/mtd/nand_ecc.h>
 #include <linux/bitops.h>
+#include <linux/delay.h>
+#include <linux/vmalloc.h>
+#include <linux/export.h>
 #include <linux/string.h>
+#else
+#include <common.h>
+#include <malloc.h>
+#include <linux/compat.h>
+
+ #include <linux/mtd/mtd.h>
+ #include <linux/mtd/bbm.h>
+ #include <linux/mtd/nand.h>
+ #include <linux/mtd/nand_ecc.h>
+ #include <linux/bitops.h>
+ #include <linux/string.h>
+#endif
+
+#define BBT_BLOCK_GOOD         0x00
+#define BBT_BLOCK_WORN         0x01
+#define BBT_BLOCK_RESERVED     0x02
+#define BBT_BLOCK_FACTORY_BAD  0x03
 
-#include <asm/errno.h>
+#define BBT_ENTRY_MASK         0x03
+#define BBT_ENTRY_SHIFT                2
+
+static int nand_update_bbt(struct mtd_info *mtd, loff_t offs);
+
+static inline uint8_t bbt_get_entry(struct nand_chip *chip, int block)
+{
+       uint8_t entry = chip->bbt[block >> BBT_ENTRY_SHIFT];
+       entry >>= (block & BBT_ENTRY_MASK) * 2;
+       return entry & BBT_ENTRY_MASK;
+}
+
+static inline void bbt_mark_entry(struct nand_chip *chip, int block,
+               uint8_t mark)
+{
+       uint8_t msk = (mark & BBT_ENTRY_MASK) << ((block & BBT_ENTRY_MASK) * 2);
+       chip->bbt[block >> BBT_ENTRY_SHIFT] |= msk;
+}
 
 static int check_pattern_no_oob(uint8_t *buf, struct nand_bbt_descr *td)
 {
@@ -86,33 +124,17 @@ static int check_pattern_no_oob(uint8_t *buf, struct nand_bbt_descr *td)
  * @td: search pattern descriptor
  *
  * Check for a pattern at the given place. Used to search bad block tables and
- * good / bad block identifiers. If the SCAN_EMPTY option is set then check, if
- * all bytes except the pattern area contain 0xff.
+ * good / bad block identifiers.
  */
 static int check_pattern(uint8_t *buf, int len, int paglen, struct nand_bbt_descr *td)
 {
-       int end = 0;
-       uint8_t *p = buf;
-
        if (td->options & NAND_BBT_NO_OOB)
                return check_pattern_no_oob(buf, td);
 
-       end = paglen + td->offs;
-       if (td->options & NAND_BBT_SCANEMPTY)
-               if (memchr_inv(p, 0xff, end))
-                       return -1;
-       p += end;
-
        /* Compare the pattern */
-       if (memcmp(p, td->pattern, td->len))
+       if (memcmp(buf + paglen + td->offs, td->pattern, td->len))
                return -1;
 
-       if (td->options & NAND_BBT_SCANEMPTY) {
-               p += td->len;
-               end += td->len;
-               if (memchr_inv(p, 0xff, len - end))
-                       return -1;
-       }
        return 0;
 }
 
@@ -159,7 +181,7 @@ static u32 add_marker_len(struct nand_bbt_descr *td)
  * @page: the starting page
  * @num: the number of bbt descriptors to read
  * @td: the bbt describtion table
- * @offs: offset in the memory table
+ * @offs: block number offset in the table
  *
  * Read the bad block table starting from page.
  */
@@ -209,25 +231,33 @@ static int read_bbt(struct mtd_info *mtd, uint8_t *buf, int page, int num,
                /* Analyse data */
                for (i = 0; i < len; i++) {
                        uint8_t dat = buf[i];
-                       for (j = 0; j < 8; j += bits, act += 2) {
+                       for (j = 0; j < 8; j += bits, act++) {
                                uint8_t tmp = (dat >> j) & msk;
                                if (tmp == msk)
                                        continue;
                                if (reserved_block_code && (tmp == reserved_block_code)) {
                                        pr_info("nand_read_bbt: reserved block at 0x%012llx\n",
-                                                (loff_t)((offs << 2) + (act >> 1)) << this->bbt_erase_shift);
-                                       this->bbt[offs + (act >> 3)] |= 0x2 << (act & 0x06);
+                                                (loff_t)(offs + act) <<
+                                                this->bbt_erase_shift);
+                                       bbt_mark_entry(this, offs + act,
+                                                       BBT_BLOCK_RESERVED);
                                        mtd->ecc_stats.bbtblocks++;
                                        continue;
                                }
-                               pr_info("nand_read_bbt: Bad block at 0x%012llx\n",
-                                       (loff_t)((offs << 2) + (act >> 1))
-                                       << this->bbt_erase_shift);
+                               /*
+                                * Leave it for now, if it's matured we can
+                                * move this message to pr_debug.
+                                */
+                               pr_info("nand_read_bbt: bad block at 0x%012llx\n",
+                                        (loff_t)(offs + act) <<
+                                        this->bbt_erase_shift);
                                /* Factory marked bad or worn out? */
                                if (tmp == 0)
-                                       this->bbt[offs + (act >> 3)] |= 0x3 << (act & 0x06);
+                                       bbt_mark_entry(this, offs + act,
+                                                       BBT_BLOCK_FACTORY_BAD);
                                else
-                                       this->bbt[offs + (act >> 3)] |= 0x1 << (act & 0x06);
+                                       bbt_mark_entry(this, offs + act,
+                                                       BBT_BLOCK_WORN);
                                mtd->ecc_stats.badblocks++;
                        }
                }
@@ -262,7 +292,7 @@ static int read_abs_bbt(struct mtd_info *mtd, uint8_t *buf, struct nand_bbt_desc
                                        td, offs);
                        if (res)
                                return res;
-                       offs += this->chipsize >> (this->bbt_erase_shift + 2);
+                       offs += this->chipsize >> this->bbt_erase_shift;
                }
        } else {
                res = read_bbt(mtd, buf, td->pages[0],
@@ -396,25 +426,6 @@ static void read_abs_bbts(struct mtd_info *mtd, uint8_t *buf,
        }
 }
 
-/* Scan a given block full */
-static int scan_block_full(struct mtd_info *mtd, struct nand_bbt_descr *bd,
-                          loff_t offs, uint8_t *buf, size_t readlen,
-                          int scanlen, int numpages)
-{
-       int ret, j;
-
-       ret = scan_read_oob(mtd, buf, offs, readlen);
-       /* Ignore ECC errors when checking for BBM */
-       if (ret && !mtd_is_bitflip_or_eccerr(ret))
-               return ret;
-
-       for (j = 0; j < numpages; j++, buf += scanlen) {
-               if (check_pattern(buf, scanlen, mtd->writesize, bd))
-                       return 1;
-       }
-       return 0;
-}
-
 /* Scan a given block partially */
 static int scan_block_fast(struct mtd_info *mtd, struct nand_bbt_descr *bd,
                           loff_t offs, uint8_t *buf, int numpages)
@@ -461,36 +472,19 @@ static int create_bbt(struct mtd_info *mtd, uint8_t *buf,
        struct nand_bbt_descr *bd, int chip)
 {
        struct nand_chip *this = mtd->priv;
-       int i, numblocks, numpages, scanlen;
+       int i, numblocks, numpages;
        int startblock;
        loff_t from;
-       size_t readlen;
 
        pr_info("Scanning device for bad blocks\n");
 
-       if (bd->options & NAND_BBT_SCANALLPAGES)
-               numpages = 1 << (this->bbt_erase_shift - this->page_shift);
-       else if (bd->options & NAND_BBT_SCAN2NDPAGE)
+       if (bd->options & NAND_BBT_SCAN2NDPAGE)
                numpages = 2;
        else
                numpages = 1;
 
-       if (!(bd->options & NAND_BBT_SCANEMPTY)) {
-               /* We need only read few bytes from the OOB area */
-               scanlen = 0;
-               readlen = bd->len;
-       } else {
-               /* Full page content should be read */
-               scanlen = mtd->writesize + mtd->oobsize;
-               readlen = numpages * mtd->writesize;
-       }
-
        if (chip == -1) {
-               /*
-                * Note that numblocks is 2 * (real numblocks) here, see i+=2
-                * below as it makes shifting and masking less painful
-                */
-               numblocks = mtd->size >> (this->bbt_erase_shift - 1);
+               numblocks = mtd->size >> this->bbt_erase_shift;
                startblock = 0;
                from = 0;
        } else {
@@ -499,37 +493,31 @@ static int create_bbt(struct mtd_info *mtd, uint8_t *buf,
                               chip + 1, this->numchips);
                        return -EINVAL;
                }
-               numblocks = this->chipsize >> (this->bbt_erase_shift - 1);
+               numblocks = this->chipsize >> this->bbt_erase_shift;
                startblock = chip * numblocks;
                numblocks += startblock;
-               from = (loff_t)startblock << (this->bbt_erase_shift - 1);
+               from = (loff_t)startblock << this->bbt_erase_shift;
        }
 
        if (this->bbt_options & NAND_BBT_SCANLASTPAGE)
                from += mtd->erasesize - (mtd->writesize * numpages);
 
-       for (i = startblock; i < numblocks;) {
+       for (i = startblock; i < numblocks; i++) {
                int ret;
 
                BUG_ON(bd->options & NAND_BBT_NO_OOB);
 
-               if (bd->options & NAND_BBT_SCANALLPAGES)
-                       ret = scan_block_full(mtd, bd, from, buf, readlen,
-                                             scanlen, numpages);
-               else
-                       ret = scan_block_fast(mtd, bd, from, buf, numpages);
-
+               ret = scan_block_fast(mtd, bd, from, buf, numpages);
                if (ret < 0)
                        return ret;
 
                if (ret) {
-                       this->bbt[i >> 3] |= 0x03 << (i & 0x6);
+                       bbt_mark_entry(this, i, BBT_BLOCK_FACTORY_BAD);
                        pr_warn("Bad eraseblock %d at 0x%012llx\n",
-                                 i >> 1, (unsigned long long)from);
+                               i, (unsigned long long)from);
                        mtd->ecc_stats.badblocks++;
                }
 
-               i += 2;
                from += (1 << this->bbt_erase_shift);
        }
        return 0;
@@ -554,7 +542,11 @@ static int search_bbt(struct mtd_info *mtd, uint8_t *buf, struct nand_bbt_descr
 {
        struct nand_chip *this = mtd->priv;
        int i, chips;
+#ifndef __UBOOT__
+       int bits, startblock, block, dir;
+#else
        int startblock, block, dir;
+#endif
        int scanlen = mtd->writesize + mtd->oobsize;
        int bbtblocks;
        int blocktopage = this->bbt_erase_shift - this->page_shift;
@@ -578,6 +570,11 @@ static int search_bbt(struct mtd_info *mtd, uint8_t *buf, struct nand_bbt_descr
                bbtblocks = mtd->size >> this->bbt_erase_shift;
        }
 
+#ifndef __UBOOT__
+       /* Number of bits for each erase block in the bbt */
+       bits = td->options & NAND_BBT_NRBITS_MSK;
+#endif
+
        for (i = 0; i < chips; i++) {
                /* Reset version information */
                td->version[i] = 0;
@@ -606,8 +603,8 @@ static int search_bbt(struct mtd_info *mtd, uint8_t *buf, struct nand_bbt_descr
                if (td->pages[i] == -1)
                        pr_warn("Bad block table not found for chip %d\n", i);
                else
-                       pr_info("Bad block table found at page %d, version 0x%02X\n", td->pages[i],
-                               td->version[i]);
+                       pr_info("Bad block table found at page %d, version "
+                                "0x%02X\n", td->pages[i], td->version[i]);
        }
        return 0;
 }
@@ -649,9 +646,9 @@ static int write_bbt(struct mtd_info *mtd, uint8_t *buf,
 {
        struct nand_chip *this = mtd->priv;
        struct erase_info einfo;
-       int i, j, res, chip = 0;
+       int i, res, chip = 0;
        int bits, startblock, dir, page, offs, numblocks, sft, sftmsk;
-       int nrchips, bbtoffs, pageoffs, ooboffs;
+       int nrchips, pageoffs, ooboffs;
        uint8_t msk[4];
        uint8_t rcode = td->reserved_block_code;
        size_t retlen, len = 0;
@@ -707,10 +704,9 @@ static int write_bbt(struct mtd_info *mtd, uint8_t *buf,
                for (i = 0; i < td->maxblocks; i++) {
                        int block = startblock + dir * i;
                        /* Check, if the block is bad */
-                       switch ((this->bbt[block >> 2] >>
-                                (2 * (block & 0x03))) & 0x03) {
-                       case 0x01:
-                       case 0x03:
+                       switch (bbt_get_entry(this, block)) {
+                       case BBT_BLOCK_WORN:
+                       case BBT_BLOCK_FACTORY_BAD:
                                continue;
                        }
                        page = block <<
@@ -742,8 +738,6 @@ static int write_bbt(struct mtd_info *mtd, uint8_t *buf,
                default: return -EINVAL;
                }
 
-               bbtoffs = chip * (numblocks >> 2);
-
                to = ((loff_t)page) << this->page_shift;
 
                /* Must we save the block contents? */
@@ -808,16 +802,12 @@ static int write_bbt(struct mtd_info *mtd, uint8_t *buf,
                        buf[ooboffs + td->veroffs] = td->version[chip];
 
                /* Walk through the memory table */
-               for (i = 0; i < numblocks;) {
+               for (i = 0; i < numblocks; i++) {
                        uint8_t dat;
-                       dat = this->bbt[bbtoffs + (i >> 2)];
-                       for (j = 0; j < 4; j++, i++) {
-                               int sftcnt = (i << (3 - sft)) & sftmsk;
-                               /* Do not store the reserved bbt blocks! */
-                               buf[offs + (i >> sft)] &=
-                                       ~(msk[dat & 0x03] << sftcnt);
-                               dat >>= 2;
-                       }
+                       int sftcnt = (i << (3 - sft)) & sftmsk;
+                       dat = bbt_get_entry(this, chip * numblocks + i);
+                       /* Do not store the reserved bbt blocks! */
+                       buf[offs + (i >> sft)] &= ~(msk[dat] << sftcnt);
                }
 
                memset(&einfo, 0, sizeof(einfo));
@@ -859,7 +849,6 @@ static inline int nand_memory_bbt(struct mtd_info *mtd, struct nand_bbt_descr *b
 {
        struct nand_chip *this = mtd->priv;
 
-       bd->options &= ~NAND_BBT_SCANEMPTY;
        return create_bbt(mtd, this->buffers->databuf, bd, -1);
 }
 
@@ -1003,7 +992,7 @@ static void mark_bbt_region(struct mtd_info *mtd, struct nand_bbt_descr *td)
 {
        struct nand_chip *this = mtd->priv;
        int i, j, chips, block, nrblocks, update;
-       uint8_t oldval, newval;
+       uint8_t oldval;
 
        /* Do we have a bbt per chip? */
        if (td->options & NAND_BBT_PERCHIP) {
@@ -1020,12 +1009,12 @@ static void mark_bbt_region(struct mtd_info *mtd, struct nand_bbt_descr *td)
                        if (td->pages[i] == -1)
                                continue;
                        block = td->pages[i] >> (this->bbt_erase_shift - this->page_shift);
-                       block <<= 1;
-                       oldval = this->bbt[(block >> 3)];
-                       newval = oldval | (0x2 << (block & 0x06));
-                       this->bbt[(block >> 3)] = newval;
-                       if ((oldval != newval) && td->reserved_block_code)
-                               nand_update_bbt(mtd, (loff_t)block << (this->bbt_erase_shift - 1));
+                       oldval = bbt_get_entry(this, block);
+                       bbt_mark_entry(this, block, BBT_BLOCK_RESERVED);
+                       if ((oldval != BBT_BLOCK_RESERVED) &&
+                                       td->reserved_block_code)
+                               nand_update_bbt(mtd, (loff_t)block <<
+                                               this->bbt_erase_shift);
                        continue;
                }
                update = 0;
@@ -1033,14 +1022,12 @@ static void mark_bbt_region(struct mtd_info *mtd, struct nand_bbt_descr *td)
                        block = ((i + 1) * nrblocks) - td->maxblocks;
                else
                        block = i * nrblocks;
-               block <<= 1;
                for (j = 0; j < td->maxblocks; j++) {
-                       oldval = this->bbt[(block >> 3)];
-                       newval = oldval | (0x2 << (block & 0x06));
-                       this->bbt[(block >> 3)] = newval;
-                       if (oldval != newval)
+                       oldval = bbt_get_entry(this, block);
+                       bbt_mark_entry(this, block, BBT_BLOCK_RESERVED);
+                       if (oldval != BBT_BLOCK_RESERVED)
                                update = 1;
-                       block += 2;
+                       block++;
                }
                /*
                 * If we want reserved blocks to be recorded to flash, and some
@@ -1048,7 +1035,8 @@ static void mark_bbt_region(struct mtd_info *mtd, struct nand_bbt_descr *td)
                 * bbts.  This should only happen once.
                 */
                if (update && td->reserved_block_code)
-                       nand_update_bbt(mtd, (loff_t)(block - 2) << (this->bbt_erase_shift - 1));
+                       nand_update_bbt(mtd, (loff_t)(block - 1) <<
+                                       this->bbt_erase_shift);
        }
 }
 
@@ -1174,13 +1162,13 @@ int nand_scan_bbt(struct mtd_info *mtd, struct nand_bbt_descr *bd)
 }
 
 /**
- * nand_update_bbt - [NAND Interface] update bad block table(s)
+ * nand_update_bbt - update bad block table(s)
  * @mtd: MTD device structure
  * @offs: the offset of the newly marked block
  *
  * The function updates the bad block table(s).
  */
-int nand_update_bbt(struct mtd_info *mtd, loff_t offs)
+static int nand_update_bbt(struct mtd_info *mtd, loff_t offs)
 {
        struct nand_chip *this = mtd->priv;
        int len, res = 0;
@@ -1234,15 +1222,6 @@ int nand_update_bbt(struct mtd_info *mtd, loff_t offs)
  */
 static uint8_t scan_ff_pattern[] = { 0xff, 0xff };
 
-static uint8_t scan_agand_pattern[] = { 0x1C, 0x71, 0xC7, 0x1C, 0x71, 0xC7 };
-
-static struct nand_bbt_descr agand_flashbased = {
-       .options = NAND_BBT_SCANEMPTY | NAND_BBT_SCANALLPAGES,
-       .offs = 0x20,
-       .len = 6,
-       .pattern = scan_agand_pattern
-};
-
 /* Generic flash bbt descriptors */
 static uint8_t bbt_pattern[] = {'B', 'b', 't', '0' };
 static uint8_t mirror_pattern[] = {'1', 't', 'b', 'B' };
@@ -1327,22 +1306,6 @@ int nand_default_bbt(struct mtd_info *mtd)
 {
        struct nand_chip *this = mtd->priv;
 
-       /*
-        * Default for AG-AND. We must use a flash based bad block table as the
-        * devices have factory marked _good_ blocks. Erasing those blocks
-        * leads to loss of the good / bad information, so we _must_ store this
-        * information in a good / bad table during startup.
-        */
-       if (this->options & NAND_IS_AND) {
-               /* Use the default pattern descriptors */
-               if (!this->bbt_td) {
-                       this->bbt_td = &bbt_main_descr;
-                       this->bbt_md = &bbt_mirror_descr;
-               }
-               this->bbt_options |= NAND_BBT_USE_FLASH;
-               return nand_scan_bbt(mtd, &agand_flashbased);
-       }
-
        /* Is a flash based bad block table requested? */
        if (this->bbt_options & NAND_BBT_USE_FLASH) {
                /* Use the default pattern descriptors */
@@ -1375,23 +1338,46 @@ int nand_default_bbt(struct mtd_info *mtd)
 int nand_isbad_bbt(struct mtd_info *mtd, loff_t offs, int allowbbt)
 {
        struct nand_chip *this = mtd->priv;
-       int block;
-       uint8_t res;
+       int block, res;
 
-       /* Get block number * 2 */
-       block = (int)(offs >> (this->bbt_erase_shift - 1));
-       res = (this->bbt[block >> 3] >> (block & 0x06)) & 0x03;
+       block = (int)(offs >> this->bbt_erase_shift);
+       res = bbt_get_entry(this, block);
 
-       MTDDEBUG(MTD_DEBUG_LEVEL2, "nand_isbad_bbt(): bbt info for offs 0x%08x: (block %d) 0x%02x\n",
-             (unsigned int)offs, block >> 1, res);
+       pr_debug("nand_isbad_bbt(): bbt info for offs 0x%08x: "
+                       "(block %d) 0x%02x\n",
+                       (unsigned int)offs, block, res);
 
-       switch ((int)res) {
-       case 0x00:
+       switch (res) {
+       case BBT_BLOCK_GOOD:
                return 0;
-       case 0x01:
+       case BBT_BLOCK_WORN:
                return 1;
-       case 0x02:
+       case BBT_BLOCK_RESERVED:
                return allowbbt ? 0 : 1;
        }
        return 1;
 }
+
+/**
+ * nand_markbad_bbt - [NAND Interface] Mark a block bad in the BBT
+ * @mtd: MTD device structure
+ * @offs: offset of the bad block
+ */
+int nand_markbad_bbt(struct mtd_info *mtd, loff_t offs)
+{
+       struct nand_chip *this = mtd->priv;
+       int block, ret = 0;
+
+       block = (int)(offs >> this->bbt_erase_shift);
+
+       /* Mark bad block in memory */
+       bbt_mark_entry(this, block, BBT_BLOCK_WORN);
+
+       /* Update flash-based bad block table */
+       if (this->bbt_options & NAND_BBT_USE_FLASH)
+               ret = nand_update_bbt(mtd, offs);
+
+       return ret;
+}
+
+EXPORT_SYMBOL(nand_scan_bbt);
index f3f0cb676dc96021005a608a9ff8d52c3bc2256a..2da8d08ad490971f2305553d39d12074f76acf7f 100644 (file)
  * published by the Free Software Foundation.
  *
  */
-
+#define __UBOOT__
+#ifndef __UBOOT__
+#include <linux/module.h>
+#include <linux/mtd/nand.h>
+#else
 #include <common.h>
 #include <linux/mtd/nand.h>
-/*
-*      Chip ID list
-*
-*      Name. ID code, pagesize, chipsize in MegaByte, eraseblock size,
-*      options
-*
-*      Pagesize; 0, 256, 512
-*      0       get this information from the extended chip ID
-+      256     256 Byte page size
-*      512     512 Byte page size
-*/
-const struct nand_flash_dev nand_flash_ids[] = {
-
-#ifdef CONFIG_MTD_NAND_MUSEUM_IDS
-       {"NAND 1MiB 5V 8-bit",          0x6e, 256, 1, 0x1000, 0},
-       {"NAND 2MiB 5V 8-bit",          0x64, 256, 2, 0x1000, 0},
-       {"NAND 4MiB 5V 8-bit",          0x6b, 512, 4, 0x2000, 0},
-       {"NAND 1MiB 3,3V 8-bit",        0xe8, 256, 1, 0x1000, 0},
-       {"NAND 1MiB 3,3V 8-bit",        0xec, 256, 1, 0x1000, 0},
-       {"NAND 2MiB 3,3V 8-bit",        0xea, 256, 2, 0x1000, 0},
-       {"NAND 4MiB 3,3V 8-bit",        0xd5, 512, 4, 0x2000, 0},
-       {"NAND 4MiB 3,3V 8-bit",        0xe3, 512, 4, 0x2000, 0},
-       {"NAND 4MiB 3,3V 8-bit",        0xe5, 512, 4, 0x2000, 0},
-       {"NAND 8MiB 3,3V 8-bit",        0xd6, 512, 8, 0x2000, 0},
-
-       {"NAND 8MiB 1,8V 8-bit",        0x39, 512, 8, 0x2000, 0},
-       {"NAND 8MiB 3,3V 8-bit",        0xe6, 512, 8, 0x2000, 0},
-       {"NAND 8MiB 1,8V 16-bit",       0x49, 512, 8, 0x2000, NAND_BUSWIDTH_16},
-       {"NAND 8MiB 3,3V 16-bit",       0x59, 512, 8, 0x2000, NAND_BUSWIDTH_16},
 #endif
+#include <linux/sizes.h>
 
-       {"NAND 16MiB 1,8V 8-bit",       0x33, 512, 16, 0x4000, 0},
-       {"NAND 16MiB 3,3V 8-bit",       0x73, 512, 16, 0x4000, 0},
-       {"NAND 16MiB 1,8V 16-bit",      0x43, 512, 16, 0x4000, NAND_BUSWIDTH_16},
-       {"NAND 16MiB 3,3V 16-bit",      0x53, 512, 16, 0x4000, NAND_BUSWIDTH_16},
-
-       {"NAND 32MiB 1,8V 8-bit",       0x35, 512, 32, 0x4000, 0},
-       {"NAND 32MiB 3,3V 8-bit",       0x75, 512, 32, 0x4000, 0},
-       {"NAND 32MiB 1,8V 16-bit",      0x45, 512, 32, 0x4000, NAND_BUSWIDTH_16},
-       {"NAND 32MiB 3,3V 16-bit",      0x55, 512, 32, 0x4000, NAND_BUSWIDTH_16},
-
-       {"NAND 64MiB 1,8V 8-bit",       0x36, 512, 64, 0x4000, 0},
-       {"NAND 64MiB 3,3V 8-bit",       0x76, 512, 64, 0x4000, 0},
-       {"NAND 64MiB 1,8V 16-bit",      0x46, 512, 64, 0x4000, NAND_BUSWIDTH_16},
-       {"NAND 64MiB 3,3V 16-bit",      0x56, 512, 64, 0x4000, NAND_BUSWIDTH_16},
+#define LP_OPTIONS NAND_SAMSUNG_LP_OPTIONS
+#define LP_OPTIONS16 (LP_OPTIONS | NAND_BUSWIDTH_16)
 
-       {"NAND 128MiB 1,8V 8-bit",      0x78, 512, 128, 0x4000, 0},
-       {"NAND 128MiB 1,8V 8-bit",      0x39, 512, 128, 0x4000, 0},
-       {"NAND 128MiB 3,3V 8-bit",      0x79, 512, 128, 0x4000, 0},
-       {"NAND 128MiB 1,8V 16-bit",     0x72, 512, 128, 0x4000, NAND_BUSWIDTH_16},
-       {"NAND 128MiB 1,8V 16-bit",     0x49, 512, 128, 0x4000, NAND_BUSWIDTH_16},
-       {"NAND 128MiB 3,3V 16-bit",     0x74, 512, 128, 0x4000, NAND_BUSWIDTH_16},
-       {"NAND 128MiB 3,3V 16-bit",     0x59, 512, 128, 0x4000, NAND_BUSWIDTH_16},
+#define SP_OPTIONS NAND_NEED_READRDY
+#define SP_OPTIONS16 (SP_OPTIONS | NAND_BUSWIDTH_16)
 
-       {"NAND 256MiB 3,3V 8-bit",      0x71, 512, 256, 0x4000, 0},
+/*
+ * The chip ID list:
+ *    name, device ID, page size, chip size in MiB, eraseblock size, options
+ *
+ * If page size and eraseblock size are 0, the sizes are taken from the
+ * extended chip ID.
+ */
+struct nand_flash_dev nand_flash_ids[] = {
+#ifdef CONFIG_MTD_NAND_MUSEUM_IDS
+       LEGACY_ID_NAND("NAND 1MiB 5V 8-bit",    0x6e, 1, SZ_4K, SP_OPTIONS),
+       LEGACY_ID_NAND("NAND 2MiB 5V 8-bit",    0x64, 2, SZ_4K, SP_OPTIONS),
+       LEGACY_ID_NAND("NAND 1MiB 3,3V 8-bit",  0xe8, 1, SZ_4K, SP_OPTIONS),
+       LEGACY_ID_NAND("NAND 1MiB 3,3V 8-bit",  0xec, 1, SZ_4K, SP_OPTIONS),
+       LEGACY_ID_NAND("NAND 2MiB 3,3V 8-bit",  0xea, 2, SZ_4K, SP_OPTIONS),
+       LEGACY_ID_NAND("NAND 4MiB 3,3V 8-bit",  0xd5, 4, SZ_8K, SP_OPTIONS),
+
+       LEGACY_ID_NAND("NAND 8MiB 3,3V 8-bit",  0xe6, 8, SZ_8K, SP_OPTIONS),
+#endif
+       /*
+        * Some incompatible NAND chips share device ID's and so must be
+        * listed by full ID. We list them first so that we can easily identify
+        * the most specific match.
+        */
+       {"TC58NVG2S0F 4G 3.3V 8-bit",
+               { .id = {0x98, 0xdc, 0x90, 0x26, 0x76, 0x15, 0x01, 0x08} },
+                 SZ_4K, SZ_512, SZ_256K, 0, 8, 224, NAND_ECC_INFO(4, SZ_512) },
+       {"TC58NVG3S0F 8G 3.3V 8-bit",
+               { .id = {0x98, 0xd3, 0x90, 0x26, 0x76, 0x15, 0x02, 0x08} },
+                 SZ_4K, SZ_1K, SZ_256K, 0, 8, 232, NAND_ECC_INFO(4, SZ_512) },
+       {"TC58NVG5D2 32G 3.3V 8-bit",
+               { .id = {0x98, 0xd7, 0x94, 0x32, 0x76, 0x56, 0x09, 0x00} },
+                 SZ_8K, SZ_4K, SZ_1M, 0, 8, 640, NAND_ECC_INFO(40, SZ_1K) },
+       {"TC58NVG6D2 64G 3.3V 8-bit",
+               { .id = {0x98, 0xde, 0x94, 0x82, 0x76, 0x56, 0x04, 0x20} },
+                 SZ_8K, SZ_8K, SZ_2M, 0, 8, 640, NAND_ECC_INFO(40, SZ_1K) },
+
+       LEGACY_ID_NAND("NAND 4MiB 5V 8-bit",   0x6B, 4, SZ_8K, SP_OPTIONS),
+       LEGACY_ID_NAND("NAND 4MiB 3,3V 8-bit", 0xE3, 4, SZ_8K, SP_OPTIONS),
+       LEGACY_ID_NAND("NAND 4MiB 3,3V 8-bit", 0xE5, 4, SZ_8K, SP_OPTIONS),
+       LEGACY_ID_NAND("NAND 8MiB 3,3V 8-bit", 0xD6, 8, SZ_8K, SP_OPTIONS),
+       LEGACY_ID_NAND("NAND 8MiB 3,3V 8-bit", 0xE6, 8, SZ_8K, SP_OPTIONS),
+
+       LEGACY_ID_NAND("NAND 16MiB 1,8V 8-bit",  0x33, 16, SZ_16K, SP_OPTIONS),
+       LEGACY_ID_NAND("NAND 16MiB 3,3V 8-bit",  0x73, 16, SZ_16K, SP_OPTIONS),
+       LEGACY_ID_NAND("NAND 16MiB 1,8V 16-bit", 0x43, 16, SZ_16K, SP_OPTIONS16),
+       LEGACY_ID_NAND("NAND 16MiB 3,3V 16-bit", 0x53, 16, SZ_16K, SP_OPTIONS16),
+
+       LEGACY_ID_NAND("NAND 32MiB 1,8V 8-bit",  0x35, 32, SZ_16K, SP_OPTIONS),
+       LEGACY_ID_NAND("NAND 32MiB 3,3V 8-bit",  0x75, 32, SZ_16K, SP_OPTIONS),
+       LEGACY_ID_NAND("NAND 32MiB 1,8V 16-bit", 0x45, 32, SZ_16K, SP_OPTIONS16),
+       LEGACY_ID_NAND("NAND 32MiB 3,3V 16-bit", 0x55, 32, SZ_16K, SP_OPTIONS16),
+
+       LEGACY_ID_NAND("NAND 64MiB 1,8V 8-bit",  0x36, 64, SZ_16K, SP_OPTIONS),
+       LEGACY_ID_NAND("NAND 64MiB 3,3V 8-bit",  0x76, 64, SZ_16K, SP_OPTIONS),
+       LEGACY_ID_NAND("NAND 64MiB 1,8V 16-bit", 0x46, 64, SZ_16K, SP_OPTIONS16),
+       LEGACY_ID_NAND("NAND 64MiB 3,3V 16-bit", 0x56, 64, SZ_16K, SP_OPTIONS16),
+
+       LEGACY_ID_NAND("NAND 128MiB 1,8V 8-bit",  0x78, 128, SZ_16K, SP_OPTIONS),
+       LEGACY_ID_NAND("NAND 128MiB 1,8V 8-bit",  0x39, 128, SZ_16K, SP_OPTIONS),
+       LEGACY_ID_NAND("NAND 128MiB 3,3V 8-bit",  0x79, 128, SZ_16K, SP_OPTIONS),
+       LEGACY_ID_NAND("NAND 128MiB 1,8V 16-bit", 0x72, 128, SZ_16K, SP_OPTIONS16),
+       LEGACY_ID_NAND("NAND 128MiB 1,8V 16-bit", 0x49, 128, SZ_16K, SP_OPTIONS16),
+       LEGACY_ID_NAND("NAND 128MiB 3,3V 16-bit", 0x74, 128, SZ_16K, SP_OPTIONS16),
+       LEGACY_ID_NAND("NAND 128MiB 3,3V 16-bit", 0x59, 128, SZ_16K, SP_OPTIONS16),
+
+       LEGACY_ID_NAND("NAND 256MiB 3,3V 8-bit", 0x71, 256, SZ_16K, SP_OPTIONS),
 
        /*
-        * These are the new chips with large page size. The pagesize and the
-        * erasesize is determined from the extended id bytes
+        * These are the new chips with large page size. Their page size and
+        * eraseblock size are determined from the extended ID bytes.
         */
-#define LP_OPTIONS NAND_SAMSUNG_LP_OPTIONS
-#define LP_OPTIONS16 (LP_OPTIONS | NAND_BUSWIDTH_16)
 
        /* 512 Megabit */
-       {"NAND 64MiB 1,8V 8-bit",       0xA2, 0,  64, 0, LP_OPTIONS},
-       {"NAND 64MiB 1,8V 8-bit",       0xA0, 0,  64, 0, LP_OPTIONS},
-       {"NAND 64MiB 3,3V 8-bit",       0xF2, 0,  64, 0, LP_OPTIONS},
-       {"NAND 64MiB 3,3V 8-bit",       0xD0, 0,  64, 0, LP_OPTIONS},
-       {"NAND 64MiB 3,3V 8-bit",       0xF0, 0,  64, 0, LP_OPTIONS},
-       {"NAND 64MiB 1,8V 16-bit",      0xB2, 0,  64, 0, LP_OPTIONS16},
-       {"NAND 64MiB 1,8V 16-bit",      0xB0, 0,  64, 0, LP_OPTIONS16},
-       {"NAND 64MiB 3,3V 16-bit",      0xC2, 0,  64, 0, LP_OPTIONS16},
-       {"NAND 64MiB 3,3V 16-bit",      0xC0, 0,  64, 0, LP_OPTIONS16},
+       EXTENDED_ID_NAND("NAND 64MiB 1,8V 8-bit",  0xA2,  64, LP_OPTIONS),
+       EXTENDED_ID_NAND("NAND 64MiB 1,8V 8-bit",  0xA0,  64, LP_OPTIONS),
+       EXTENDED_ID_NAND("NAND 64MiB 3,3V 8-bit",  0xF2,  64, LP_OPTIONS),
+       EXTENDED_ID_NAND("NAND 64MiB 3,3V 8-bit",  0xD0,  64, LP_OPTIONS),
+       EXTENDED_ID_NAND("NAND 64MiB 3,3V 8-bit",  0xF0,  64, LP_OPTIONS),
+       EXTENDED_ID_NAND("NAND 64MiB 1,8V 16-bit", 0xB2,  64, LP_OPTIONS16),
+       EXTENDED_ID_NAND("NAND 64MiB 1,8V 16-bit", 0xB0,  64, LP_OPTIONS16),
+       EXTENDED_ID_NAND("NAND 64MiB 3,3V 16-bit", 0xC2,  64, LP_OPTIONS16),
+       EXTENDED_ID_NAND("NAND 64MiB 3,3V 16-bit", 0xC0,  64, LP_OPTIONS16),
 
        /* 1 Gigabit */
-       {"NAND 128MiB 1,8V 8-bit",      0xA1, 0, 128, 0, LP_OPTIONS},
-       {"NAND 128MiB 3,3V 8-bit",      0xF1, 0, 128, 0, LP_OPTIONS},
-       {"NAND 128MiB 3,3V 8-bit",      0xD1, 0, 128, 0, LP_OPTIONS},
-       {"NAND 128MiB 1,8V 16-bit",     0xB1, 0, 128, 0, LP_OPTIONS16},
-       {"NAND 128MiB 3,3V 16-bit",     0xC1, 0, 128, 0, LP_OPTIONS16},
-       {"NAND 128MiB 1,8V 16-bit",     0xAD, 0, 128, 0, LP_OPTIONS16},
+       EXTENDED_ID_NAND("NAND 128MiB 1,8V 8-bit",  0xA1, 128, LP_OPTIONS),
+       EXTENDED_ID_NAND("NAND 128MiB 3,3V 8-bit",  0xF1, 128, LP_OPTIONS),
+       EXTENDED_ID_NAND("NAND 128MiB 3,3V 8-bit",  0xD1, 128, LP_OPTIONS),
+       EXTENDED_ID_NAND("NAND 128MiB 1,8V 16-bit", 0xB1, 128, LP_OPTIONS16),
+       EXTENDED_ID_NAND("NAND 128MiB 3,3V 16-bit", 0xC1, 128, LP_OPTIONS16),
+       EXTENDED_ID_NAND("NAND 128MiB 1,8V 16-bit", 0xAD, 128, LP_OPTIONS16),
 
        /* 2 Gigabit */
-       {"NAND 256MiB 1,8V 8-bit",      0xAA, 0, 256, 0, LP_OPTIONS},
-       {"NAND 256MiB 3,3V 8-bit",      0xDA, 0, 256, 0, LP_OPTIONS},
-       {"NAND 256MiB 1,8V 16-bit",     0xBA, 0, 256, 0, LP_OPTIONS16},
-       {"NAND 256MiB 3,3V 16-bit",     0xCA, 0, 256, 0, LP_OPTIONS16},
+       EXTENDED_ID_NAND("NAND 256MiB 1,8V 8-bit",  0xAA, 256, LP_OPTIONS),
+       EXTENDED_ID_NAND("NAND 256MiB 3,3V 8-bit",  0xDA, 256, LP_OPTIONS),
+       EXTENDED_ID_NAND("NAND 256MiB 1,8V 16-bit", 0xBA, 256, LP_OPTIONS16),
+       EXTENDED_ID_NAND("NAND 256MiB 3,3V 16-bit", 0xCA, 256, LP_OPTIONS16),
 
        /* 4 Gigabit */
-       {"NAND 512MiB 1,8V 8-bit",      0xAC, 0, 512, 0, LP_OPTIONS},
-       {"NAND 512MiB 3,3V 8-bit",      0xDC, 0, 512, 0, LP_OPTIONS},
-       {"NAND 512MiB 1,8V 16-bit",     0xBC, 0, 512, 0, LP_OPTIONS16},
-       {"NAND 512MiB 3,3V 16-bit",     0xCC, 0, 512, 0, LP_OPTIONS16},
+       EXTENDED_ID_NAND("NAND 512MiB 1,8V 8-bit",  0xAC, 512, LP_OPTIONS),
+       EXTENDED_ID_NAND("NAND 512MiB 3,3V 8-bit",  0xDC, 512, LP_OPTIONS),
+       EXTENDED_ID_NAND("NAND 512MiB 1,8V 16-bit", 0xBC, 512, LP_OPTIONS16),
+       EXTENDED_ID_NAND("NAND 512MiB 3,3V 16-bit", 0xCC, 512, LP_OPTIONS16),
 
        /* 8 Gigabit */
-       {"NAND 1GiB 1,8V 8-bit",        0xA3, 0, 1024, 0, LP_OPTIONS},
-       {"NAND 1GiB 3,3V 8-bit",        0xD3, 0, 1024, 0, LP_OPTIONS},
-       {"NAND 1GiB 1,8V 16-bit",       0xB3, 0, 1024, 0, LP_OPTIONS16},
-       {"NAND 1GiB 3,3V 16-bit",       0xC3, 0, 1024, 0, LP_OPTIONS16},
+       EXTENDED_ID_NAND("NAND 1GiB 1,8V 8-bit",  0xA3, 1024, LP_OPTIONS),
+       EXTENDED_ID_NAND("NAND 1GiB 3,3V 8-bit",  0xD3, 1024, LP_OPTIONS),
+       EXTENDED_ID_NAND("NAND 1GiB 1,8V 16-bit", 0xB3, 1024, LP_OPTIONS16),
+       EXTENDED_ID_NAND("NAND 1GiB 3,3V 16-bit", 0xC3, 1024, LP_OPTIONS16),
 
        /* 16 Gigabit */
-       {"NAND 2GiB 1,8V 8-bit",        0xA5, 0, 2048, 0, LP_OPTIONS},
-       {"NAND 2GiB 3,3V 8-bit",        0xD5, 0, 2048, 0, LP_OPTIONS},
-       {"NAND 2GiB 1,8V 16-bit",       0xB5, 0, 2048, 0, LP_OPTIONS16},
-       {"NAND 2GiB 3,3V 16-bit",       0xC5, 0, 2048, 0, LP_OPTIONS16},
+       EXTENDED_ID_NAND("NAND 2GiB 1,8V 8-bit",  0xA5, 2048, LP_OPTIONS),
+       EXTENDED_ID_NAND("NAND 2GiB 3,3V 8-bit",  0xD5, 2048, LP_OPTIONS),
+       EXTENDED_ID_NAND("NAND 2GiB 1,8V 16-bit", 0xB5, 2048, LP_OPTIONS16),
+       EXTENDED_ID_NAND("NAND 2GiB 3,3V 16-bit", 0xC5, 2048, LP_OPTIONS16),
 
        /* 32 Gigabit */
-       {"NAND 4GiB 1,8V 8-bit",        0xA7, 0, 4096, 0, LP_OPTIONS},
-       {"NAND 4GiB 3,3V 8-bit",        0xD7, 0, 4096, 0, LP_OPTIONS},
-       {"NAND 4GiB 1,8V 16-bit",       0xB7, 0, 4096, 0, LP_OPTIONS16},
-       {"NAND 4GiB 3,3V 16-bit",       0xC7, 0, 4096, 0, LP_OPTIONS16},
+       EXTENDED_ID_NAND("NAND 4GiB 1,8V 8-bit",  0xA7, 4096, LP_OPTIONS),
+       EXTENDED_ID_NAND("NAND 4GiB 3,3V 8-bit",  0xD7, 4096, LP_OPTIONS),
+       EXTENDED_ID_NAND("NAND 4GiB 1,8V 16-bit", 0xB7, 4096, LP_OPTIONS16),
+       EXTENDED_ID_NAND("NAND 4GiB 3,3V 16-bit", 0xC7, 4096, LP_OPTIONS16),
 
        /* 64 Gigabit */
-       {"NAND 8GiB 1,8V 8-bit",        0xAE, 0, 8192, 0, LP_OPTIONS},
-       {"NAND 8GiB 3,3V 8-bit",        0xDE, 0, 8192, 0, LP_OPTIONS},
-       {"NAND 8GiB 1,8V 16-bit",       0xBE, 0, 8192, 0, LP_OPTIONS16},
-       {"NAND 8GiB 3,3V 16-bit",       0xCE, 0, 8192, 0, LP_OPTIONS16},
+       EXTENDED_ID_NAND("NAND 8GiB 1,8V 8-bit",  0xAE, 8192, LP_OPTIONS),
+       EXTENDED_ID_NAND("NAND 8GiB 3,3V 8-bit",  0xDE, 8192, LP_OPTIONS),
+       EXTENDED_ID_NAND("NAND 8GiB 1,8V 16-bit", 0xBE, 8192, LP_OPTIONS16),
+       EXTENDED_ID_NAND("NAND 8GiB 3,3V 16-bit", 0xCE, 8192, LP_OPTIONS16),
 
        /* 128 Gigabit */
-       {"NAND 16GiB 1,8V 8-bit",       0x1A, 0, 16384, 0, LP_OPTIONS},
-       {"NAND 16GiB 3,3V 8-bit",       0x3A, 0, 16384, 0, LP_OPTIONS},
-       {"NAND 16GiB 1,8V 16-bit",      0x2A, 0, 16384, 0, LP_OPTIONS16},
-       {"NAND 16GiB 3,3V 16-bit",      0x4A, 0, 16384, 0, LP_OPTIONS16},
+       EXTENDED_ID_NAND("NAND 16GiB 1,8V 8-bit",  0x1A, 16384, LP_OPTIONS),
+       EXTENDED_ID_NAND("NAND 16GiB 3,3V 8-bit",  0x3A, 16384, LP_OPTIONS),
+       EXTENDED_ID_NAND("NAND 16GiB 1,8V 16-bit", 0x2A, 16384, LP_OPTIONS16),
+       EXTENDED_ID_NAND("NAND 16GiB 3,3V 16-bit", 0x4A, 16384, LP_OPTIONS16),
 
        /* 256 Gigabit */
-       {"NAND 32GiB 1,8V 8-bit",       0x1C, 0, 32768, 0, LP_OPTIONS},
-       {"NAND 32GiB 3,3V 8-bit",       0x3C, 0, 32768, 0, LP_OPTIONS},
-       {"NAND 32GiB 1,8V 16-bit",      0x2C, 0, 32768, 0, LP_OPTIONS16},
-       {"NAND 32GiB 3,3V 16-bit",      0x4C, 0, 32768, 0, LP_OPTIONS16},
+       EXTENDED_ID_NAND("NAND 32GiB 1,8V 8-bit",  0x1C, 32768, LP_OPTIONS),
+       EXTENDED_ID_NAND("NAND 32GiB 3,3V 8-bit",  0x3C, 32768, LP_OPTIONS),
+       EXTENDED_ID_NAND("NAND 32GiB 1,8V 16-bit", 0x2C, 32768, LP_OPTIONS16),
+       EXTENDED_ID_NAND("NAND 32GiB 3,3V 16-bit", 0x4C, 32768, LP_OPTIONS16),
 
        /* 512 Gigabit */
-       {"NAND 64GiB 1,8V 8-bit",       0x1E, 0, 65536, 0, LP_OPTIONS},
-       {"NAND 64GiB 3,3V 8-bit",       0x3E, 0, 65536, 0, LP_OPTIONS},
-       {"NAND 64GiB 1,8V 16-bit",      0x2E, 0, 65536, 0, LP_OPTIONS16},
-       {"NAND 64GiB 3,3V 16-bit",      0x4E, 0, 65536, 0, LP_OPTIONS16},
+       EXTENDED_ID_NAND("NAND 64GiB 1,8V 8-bit",  0x1E, 65536, LP_OPTIONS),
+       EXTENDED_ID_NAND("NAND 64GiB 3,3V 8-bit",  0x3E, 65536, LP_OPTIONS),
+       EXTENDED_ID_NAND("NAND 64GiB 1,8V 16-bit", 0x2E, 65536, LP_OPTIONS16),
+       EXTENDED_ID_NAND("NAND 64GiB 3,3V 16-bit", 0x4E, 65536, LP_OPTIONS16),
 
-       /*
-        * Renesas AND 1 Gigabit. Those chips do not support extended id and
-        * have a strange page/block layout !  The chosen minimum erasesize is
-        * 4 * 2 * 2048 = 16384 Byte, as those chips have an array of 4 page
-        * planes 1 block = 2 pages, but due to plane arrangement the blocks
-        * 0-3 consists of page 0 + 4,1 + 5, 2 + 6, 3 + 7 Anyway JFFS2 would
-        * increase the eraseblock size so we chose a combined one which can be
-        * erased in one go There are more speed improvements for reads and
-        * writes possible, but not implemented now
-        */
-       {"AND 128MiB 3,3V 8-bit",       0x01, 2048, 128, 0x4000,
-        NAND_IS_AND | NAND_4PAGE_ARRAY | BBT_AUTO_REFRESH},
-
-       {NULL,}
+       {NULL}
 };
 
-/*
-*      Manufacturer ID list
-*/
-const struct nand_manufacturers nand_manuf_ids[] = {
+/* Manufacturer IDs */
+struct nand_manufacturers nand_manuf_ids[] = {
        {NAND_MFR_TOSHIBA, "Toshiba"},
        {NAND_MFR_SAMSUNG, "Samsung"},
        {NAND_MFR_FUJITSU, "Fujitsu"},
@@ -178,5 +185,14 @@ const struct nand_manufacturers nand_manuf_ids[] = {
        {NAND_MFR_AMD, "AMD/Spansion"},
        {NAND_MFR_MACRONIX, "Macronix"},
        {NAND_MFR_EON, "Eon"},
+       {NAND_MFR_SANDISK, "SanDisk"},
+       {NAND_MFR_INTEL, "Intel"},
        {0x0, "Unknown"}
 };
+
+EXPORT_SYMBOL(nand_manuf_ids);
+EXPORT_SYMBOL(nand_flash_ids);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Thomas Gleixner <tglx@linutronix.de>");
+MODULE_DESCRIPTION("Nand device & manufacturer IDs");
index b292826034f9f3964ae4278ed77314b953bcb761..024f6fb4402485bc6337a00af308e1eeb5a3e013 100644 (file)
@@ -187,6 +187,9 @@ int nand_erase_opts(nand_info_t *meminfo, const nand_erase_options_t *opts)
 
 #ifdef CONFIG_CMD_NAND_LOCK_UNLOCK
 
+#define NAND_CMD_LOCK_TIGHT     0x2c
+#define NAND_CMD_LOCK_STATUS    0x7a
 /******************************************************************************
  * Support for locking / unlocking operations of some NAND devices
  *****************************************************************************/
index 5510b13c01269c0236c02f3b0e3ea47461715e55..265959502d8320a0533d59eb4be88528984518a8 100644 (file)
@@ -118,6 +118,7 @@ static void ndfc_write_buf(struct mtd_info *mtdinfo, const uint8_t *buf, int len
                out_be32((u32 *)(base + NDFC_DATA), *p++);
 }
 
+#if defined(CONFIG_MTD_NAND_VERIFY_WRITE)
 static int ndfc_verify_buf(struct mtd_info *mtdinfo, const uint8_t *buf, int len)
 {
        struct nand_chip *this = mtdinfo->priv;
@@ -130,6 +131,7 @@ static int ndfc_verify_buf(struct mtd_info *mtdinfo, const uint8_t *buf, int len
 
        return 0;
 }
+#endif
 
 /*
  * Read a byte from the NDFC.
@@ -205,7 +207,9 @@ int board_nand_init(struct nand_chip *nand)
 #endif
 
        nand->write_buf  = ndfc_write_buf;
+#if defined(CONFIG_MTD_NAND_VERIFY_WRITE)
        nand->verify_buf = ndfc_verify_buf;
+#endif
        nand->read_byte = ndfc_read_byte;
 
        chip++;
index e33e8d38e7a48808d9e8c982a7470b2b78eea01c..03deabce102341881a7b430f35164eea9c455ce6 100644 (file)
@@ -22,6 +22,7 @@
 #include <common.h>
 #include <linux/compat.h>
 #include <linux/mtd/mtd.h>
+#include "linux/mtd/flashchip.h"
 #include <linux/mtd/onenand.h>
 
 #include <asm/io.h>
index 0267c2c5c90a8b7f8b0bba69422638e625c22c0c..52509f1ae9ab4747a80f14304b05231c9466ced1 100644 (file)
@@ -140,7 +140,6 @@ static inline int onenand_memory_bbt(struct mtd_info *mtd,
 {
        unsigned char data_buf[MAX_ONENAND_PAGESIZE];
 
-       bd->options &= ~NAND_BBT_SCANEMPTY;
        return create_bbt(mtd, data_buf, bd, -1);
 }
 
index df04c2bb481b4e4bc0dc841100d3e7b3a1c0be64..5e56a2954df6e480ce4b3b9a3b19d7a2485c1142 100644 (file)
 #include <linux/compat.h>
 #include <linux/mtd/mtd.h>
 #include <linux/mtd/onenand.h>
+#include <linux/mtd/flashchip.h>
 #include <linux/mtd/samsung_onenand.h>
 
 #include <asm/io.h>
 #include <asm/errno.h>
 
-#ifdef ONENAND_DEBUG
-#define DPRINTK(format, args...)                                       \
-do {                                                                   \
-       printf("%s[%d]: " format "\n", __func__, __LINE__, ##args);     \
-} while (0)
-#else
-#define DPRINTK(...)                   do { } while (0)
-#endif
-
 #define ONENAND_ERASE_STATUS           0x00
 #define ONENAND_MULTI_ERASE_SET                0x01
 #define ONENAND_ERASE_START            0x03
index 56c2823477da3388d455b23b18b7253a29f776be..4807f94fab69fb53e39ed012b58de3a4f11a1da0 100644 (file)
@@ -5,6 +5,7 @@
 # SPDX-License-Identifier:     GPL-2.0+
 #
 
-obj-y += build.o vtbl.o vmt.o upd.o kapi.o eba.o io.o wl.o scan.o crc32.o
+obj-y += attach.o build.o vtbl.o vmt.o upd.o kapi.o eba.o io.o wl.o crc32.o
+obj-$(CONFIG_MTD_UBI_FASTMAP) += fastmap.o
 obj-y += misc.o
 obj-y += debug.o
diff --git a/drivers/mtd/ubi/attach.c b/drivers/mtd/ubi/attach.c
new file mode 100644 (file)
index 0000000..9fce02e
--- /dev/null
@@ -0,0 +1,1754 @@
+/*
+ * Copyright (c) International Business Machines Corp., 2006
+ *
+ * SPDX-License-Identifier:    GPL-2.0+
+ *
+ * Author: Artem Bityutskiy (Битюцкий Артём)
+ */
+
+/*
+ * UBI attaching sub-system.
+ *
+ * This sub-system is responsible for attaching MTD devices and it also
+ * implements flash media scanning.
+ *
+ * The attaching information is represented by a &struct ubi_attach_info'
+ * object. Information about volumes is represented by &struct ubi_ainf_volume
+ * objects which are kept in volume RB-tree with root at the @volumes field.
+ * The RB-tree is indexed by the volume ID.
+ *
+ * Logical eraseblocks are represented by &struct ubi_ainf_peb objects. These
+ * objects are kept in per-volume RB-trees with the root at the corresponding
+ * &struct ubi_ainf_volume object. To put it differently, we keep an RB-tree of
+ * per-volume objects and each of these objects is the root of RB-tree of
+ * per-LEB objects.
+ *
+ * Corrupted physical eraseblocks are put to the @corr list, free physical
+ * eraseblocks are put to the @free list and the physical eraseblock to be
+ * erased are put to the @erase list.
+ *
+ * About corruptions
+ * ~~~~~~~~~~~~~~~~~
+ *
+ * UBI protects EC and VID headers with CRC-32 checksums, so it can detect
+ * whether the headers are corrupted or not. Sometimes UBI also protects the
+ * data with CRC-32, e.g., when it executes the atomic LEB change operation, or
+ * when it moves the contents of a PEB for wear-leveling purposes.
+ *
+ * UBI tries to distinguish between 2 types of corruptions.
+ *
+ * 1. Corruptions caused by power cuts. These are expected corruptions and UBI
+ * tries to handle them gracefully, without printing too many warnings and
+ * error messages. The idea is that we do not lose important data in these
+ * cases - we may lose only the data which were being written to the media just
+ * before the power cut happened, and the upper layers (e.g., UBIFS) are
+ * supposed to handle such data losses (e.g., by using the FS journal).
+ *
+ * When UBI detects a corruption (CRC-32 mismatch) in a PEB, and it looks like
+ * the reason is a power cut, UBI puts this PEB to the @erase list, and all
+ * PEBs in the @erase list are scheduled for erasure later.
+ *
+ * 2. Unexpected corruptions which are not caused by power cuts. During
+ * attaching, such PEBs are put to the @corr list and UBI preserves them.
+ * Obviously, this lessens the amount of available PEBs, and if at some  point
+ * UBI runs out of free PEBs, it switches to R/O mode. UBI also loudly informs
+ * about such PEBs every time the MTD device is attached.
+ *
+ * However, it is difficult to reliably distinguish between these types of
+ * corruptions and UBI's strategy is as follows (in case of attaching by
+ * scanning). UBI assumes corruption type 2 if the VID header is corrupted and
+ * the data area does not contain all 0xFFs, and there were no bit-flips or
+ * integrity errors (e.g., ECC errors in case of NAND) while reading the data
+ * area.  Otherwise UBI assumes corruption type 1. So the decision criteria
+ * are as follows.
+ *   o If the data area contains only 0xFFs, there are no data, and it is safe
+ *     to just erase this PEB - this is corruption type 1.
+ *   o If the data area has bit-flips or data integrity errors (ECC errors on
+ *     NAND), it is probably a PEB which was being erased when power cut
+ *     happened, so this is corruption type 1. However, this is just a guess,
+ *     which might be wrong.
+ *   o Otherwise this is corruption type 2.
+ */
+
+#define __UBOOT__
+#ifndef __UBOOT__
+#include <linux/err.h>
+#include <linux/slab.h>
+#include <linux/crc32.h>
+#include <linux/random.h>
+#else
+#include <div64.h>
+#include <linux/err.h>
+#endif
+
+#include <linux/math64.h>
+
+#include <ubi_uboot.h>
+#include "ubi.h"
+
+static int self_check_ai(struct ubi_device *ubi, struct ubi_attach_info *ai);
+
+/* Temporary variables used during scanning */
+static struct ubi_ec_hdr *ech;
+static struct ubi_vid_hdr *vidh;
+
+/**
+ * add_to_list - add physical eraseblock to a list.
+ * @ai: attaching information
+ * @pnum: physical eraseblock number to add
+ * @vol_id: the last used volume id for the PEB
+ * @lnum: the last used LEB number for the PEB
+ * @ec: erase counter of the physical eraseblock
+ * @to_head: if not zero, add to the head of the list
+ * @list: the list to add to
+ *
+ * This function allocates a 'struct ubi_ainf_peb' object for physical
+ * eraseblock @pnum and adds it to the "free", "erase", or "alien" lists.
+ * It stores the @lnum and @vol_id alongside, which can both be
+ * %UBI_UNKNOWN if they are not available, not readable, or not assigned.
+ * If @to_head is not zero, PEB will be added to the head of the list, which
+ * basically means it will be processed first later. E.g., we add corrupted
+ * PEBs (corrupted due to power cuts) to the head of the erase list to make
+ * sure we erase them first and get rid of corruptions ASAP. This function
+ * returns zero in case of success and a negative error code in case of
+ * failure.
+ */
+static int add_to_list(struct ubi_attach_info *ai, int pnum, int vol_id,
+                      int lnum, int ec, int to_head, struct list_head *list)
+{
+       struct ubi_ainf_peb *aeb;
+
+       if (list == &ai->free) {
+               dbg_bld("add to free: PEB %d, EC %d", pnum, ec);
+       } else if (list == &ai->erase) {
+               dbg_bld("add to erase: PEB %d, EC %d", pnum, ec);
+       } else if (list == &ai->alien) {
+               dbg_bld("add to alien: PEB %d, EC %d", pnum, ec);
+               ai->alien_peb_count += 1;
+       } else
+               BUG();
+
+       aeb = kmem_cache_alloc(ai->aeb_slab_cache, GFP_KERNEL);
+       if (!aeb)
+               return -ENOMEM;
+
+       aeb->pnum = pnum;
+       aeb->vol_id = vol_id;
+       aeb->lnum = lnum;
+       aeb->ec = ec;
+       if (to_head)
+               list_add(&aeb->u.list, list);
+       else
+               list_add_tail(&aeb->u.list, list);
+       return 0;
+}
+
+/**
+ * add_corrupted - add a corrupted physical eraseblock.
+ * @ai: attaching information
+ * @pnum: physical eraseblock number to add
+ * @ec: erase counter of the physical eraseblock
+ *
+ * This function allocates a 'struct ubi_ainf_peb' object for a corrupted
+ * physical eraseblock @pnum and adds it to the 'corr' list.  The corruption
+ * was presumably not caused by a power cut. Returns zero in case of success
+ * and a negative error code in case of failure.
+ */
+static int add_corrupted(struct ubi_attach_info *ai, int pnum, int ec)
+{
+       struct ubi_ainf_peb *aeb;
+
+       dbg_bld("add to corrupted: PEB %d, EC %d", pnum, ec);
+
+       aeb = kmem_cache_alloc(ai->aeb_slab_cache, GFP_KERNEL);
+       if (!aeb)
+               return -ENOMEM;
+
+       ai->corr_peb_count += 1;
+       aeb->pnum = pnum;
+       aeb->ec = ec;
+       list_add(&aeb->u.list, &ai->corr);
+       return 0;
+}
+
+/**
+ * validate_vid_hdr - check volume identifier header.
+ * @vid_hdr: the volume identifier header to check
+ * @av: information about the volume this logical eraseblock belongs to
+ * @pnum: physical eraseblock number the VID header came from
+ *
+ * This function checks that data stored in @vid_hdr is consistent. Returns
+ * non-zero if an inconsistency was found and zero if not.
+ *
+ * Note, UBI does sanity check of everything it reads from the flash media.
+ * Most of the checks are done in the I/O sub-system. Here we check that the
+ * information in the VID header is consistent to the information in other VID
+ * headers of the same volume.
+ */
+static int validate_vid_hdr(const struct ubi_vid_hdr *vid_hdr,
+                           const struct ubi_ainf_volume *av, int pnum)
+{
+       int vol_type = vid_hdr->vol_type;
+       int vol_id = be32_to_cpu(vid_hdr->vol_id);
+       int used_ebs = be32_to_cpu(vid_hdr->used_ebs);
+       int data_pad = be32_to_cpu(vid_hdr->data_pad);
+
+       if (av->leb_count != 0) {
+               int av_vol_type;
+
+               /*
+                * This is not the first logical eraseblock belonging to this
+                * volume. Ensure that the data in its VID header is consistent
+                * to the data in previous logical eraseblock headers.
+                */
+
+               if (vol_id != av->vol_id) {
+                       ubi_err("inconsistent vol_id");
+                       goto bad;
+               }
+
+               if (av->vol_type == UBI_STATIC_VOLUME)
+                       av_vol_type = UBI_VID_STATIC;
+               else
+                       av_vol_type = UBI_VID_DYNAMIC;
+
+               if (vol_type != av_vol_type) {
+                       ubi_err("inconsistent vol_type");
+                       goto bad;
+               }
+
+               if (used_ebs != av->used_ebs) {
+                       ubi_err("inconsistent used_ebs");
+                       goto bad;
+               }
+
+               if (data_pad != av->data_pad) {
+                       ubi_err("inconsistent data_pad");
+                       goto bad;
+               }
+       }
+
+       return 0;
+
+bad:
+       ubi_err("inconsistent VID header at PEB %d", pnum);
+       ubi_dump_vid_hdr(vid_hdr);
+       ubi_dump_av(av);
+       return -EINVAL;
+}
+
+/**
+ * add_volume - add volume to the attaching information.
+ * @ai: attaching information
+ * @vol_id: ID of the volume to add
+ * @pnum: physical eraseblock number
+ * @vid_hdr: volume identifier header
+ *
+ * If the volume corresponding to the @vid_hdr logical eraseblock is already
+ * present in the attaching information, this function does nothing. Otherwise
+ * it adds corresponding volume to the attaching information. Returns a pointer
+ * to the allocated "av" object in case of success and a negative error code in
+ * case of failure.
+ */
+static struct ubi_ainf_volume *add_volume(struct ubi_attach_info *ai,
+                                         int vol_id, int pnum,
+                                         const struct ubi_vid_hdr *vid_hdr)
+{
+       struct ubi_ainf_volume *av;
+       struct rb_node **p = &ai->volumes.rb_node, *parent = NULL;
+
+       ubi_assert(vol_id == be32_to_cpu(vid_hdr->vol_id));
+
+       /* Walk the volume RB-tree to look if this volume is already present */
+       while (*p) {
+               parent = *p;
+               av = rb_entry(parent, struct ubi_ainf_volume, rb);
+
+               if (vol_id == av->vol_id)
+                       return av;
+
+               if (vol_id > av->vol_id)
+                       p = &(*p)->rb_left;
+               else
+                       p = &(*p)->rb_right;
+       }
+
+       /* The volume is absent - add it */
+       av = kmalloc(sizeof(struct ubi_ainf_volume), GFP_KERNEL);
+       if (!av)
+               return ERR_PTR(-ENOMEM);
+
+       av->highest_lnum = av->leb_count = 0;
+       av->vol_id = vol_id;
+       av->root = RB_ROOT;
+       av->used_ebs = be32_to_cpu(vid_hdr->used_ebs);
+       av->data_pad = be32_to_cpu(vid_hdr->data_pad);
+       av->compat = vid_hdr->compat;
+       av->vol_type = vid_hdr->vol_type == UBI_VID_DYNAMIC ? UBI_DYNAMIC_VOLUME
+                                                           : UBI_STATIC_VOLUME;
+       if (vol_id > ai->highest_vol_id)
+               ai->highest_vol_id = vol_id;
+
+       rb_link_node(&av->rb, parent, p);
+       rb_insert_color(&av->rb, &ai->volumes);
+       ai->vols_found += 1;
+       dbg_bld("added volume %d", vol_id);
+       return av;
+}
+
+/**
+ * ubi_compare_lebs - find out which logical eraseblock is newer.
+ * @ubi: UBI device description object
+ * @aeb: first logical eraseblock to compare
+ * @pnum: physical eraseblock number of the second logical eraseblock to
+ * compare
+ * @vid_hdr: volume identifier header of the second logical eraseblock
+ *
+ * This function compares 2 copies of a LEB and informs which one is newer. In
+ * case of success this function returns a positive value, in case of failure, a
+ * negative error code is returned. The success return codes use the following
+ * bits:
+ *     o bit 0 is cleared: the first PEB (described by @aeb) is newer than the
+ *       second PEB (described by @pnum and @vid_hdr);
+ *     o bit 0 is set: the second PEB is newer;
+ *     o bit 1 is cleared: no bit-flips were detected in the newer LEB;
+ *     o bit 1 is set: bit-flips were detected in the newer LEB;
+ *     o bit 2 is cleared: the older LEB is not corrupted;
+ *     o bit 2 is set: the older LEB is corrupted.
+ */
+int ubi_compare_lebs(struct ubi_device *ubi, const struct ubi_ainf_peb *aeb,
+                       int pnum, const struct ubi_vid_hdr *vid_hdr)
+{
+       int len, err, second_is_newer, bitflips = 0, corrupted = 0;
+       uint32_t data_crc, crc;
+       struct ubi_vid_hdr *vh = NULL;
+       unsigned long long sqnum2 = be64_to_cpu(vid_hdr->sqnum);
+
+       if (sqnum2 == aeb->sqnum) {
+               /*
+                * This must be a really ancient UBI image which has been
+                * created before sequence numbers support has been added. At
+                * that times we used 32-bit LEB versions stored in logical
+                * eraseblocks. That was before UBI got into mainline. We do not
+                * support these images anymore. Well, those images still work,
+                * but only if no unclean reboots happened.
+                */
+               ubi_err("unsupported on-flash UBI format");
+               return -EINVAL;
+       }
+
+       /* Obviously the LEB with lower sequence counter is older */
+       second_is_newer = (sqnum2 > aeb->sqnum);
+
+       /*
+        * Now we know which copy is newer. If the copy flag of the PEB with
+        * newer version is not set, then we just return, otherwise we have to
+        * check data CRC. For the second PEB we already have the VID header,
+        * for the first one - we'll need to re-read it from flash.
+        *
+        * Note: this may be optimized so that we wouldn't read twice.
+        */
+
+       if (second_is_newer) {
+               if (!vid_hdr->copy_flag) {
+                       /* It is not a copy, so it is newer */
+                       dbg_bld("second PEB %d is newer, copy_flag is unset",
+                               pnum);
+                       return 1;
+               }
+       } else {
+               if (!aeb->copy_flag) {
+                       /* It is not a copy, so it is newer */
+                       dbg_bld("first PEB %d is newer, copy_flag is unset",
+                               pnum);
+                       return bitflips << 1;
+               }
+
+               vh = ubi_zalloc_vid_hdr(ubi, GFP_KERNEL);
+               if (!vh)
+                       return -ENOMEM;
+
+               pnum = aeb->pnum;
+               err = ubi_io_read_vid_hdr(ubi, pnum, vh, 0);
+               if (err) {
+                       if (err == UBI_IO_BITFLIPS)
+                               bitflips = 1;
+                       else {
+                               ubi_err("VID of PEB %d header is bad, but it was OK earlier, err %d",
+                                       pnum, err);
+                               if (err > 0)
+                                       err = -EIO;
+
+                               goto out_free_vidh;
+                       }
+               }
+
+               vid_hdr = vh;
+       }
+
+       /* Read the data of the copy and check the CRC */
+
+       len = be32_to_cpu(vid_hdr->data_size);
+
+       mutex_lock(&ubi->buf_mutex);
+       err = ubi_io_read_data(ubi, ubi->peb_buf, pnum, 0, len);
+       if (err && err != UBI_IO_BITFLIPS && !mtd_is_eccerr(err))
+               goto out_unlock;
+
+       data_crc = be32_to_cpu(vid_hdr->data_crc);
+       crc = crc32(UBI_CRC32_INIT, ubi->peb_buf, len);
+       if (crc != data_crc) {
+               dbg_bld("PEB %d CRC error: calculated %#08x, must be %#08x",
+                       pnum, crc, data_crc);
+               corrupted = 1;
+               bitflips = 0;
+               second_is_newer = !second_is_newer;
+       } else {
+               dbg_bld("PEB %d CRC is OK", pnum);
+               bitflips = !!err;
+       }
+       mutex_unlock(&ubi->buf_mutex);
+
+       ubi_free_vid_hdr(ubi, vh);
+
+       if (second_is_newer)
+               dbg_bld("second PEB %d is newer, copy_flag is set", pnum);
+       else
+               dbg_bld("first PEB %d is newer, copy_flag is set", pnum);
+
+       return second_is_newer | (bitflips << 1) | (corrupted << 2);
+
+out_unlock:
+       mutex_unlock(&ubi->buf_mutex);
+out_free_vidh:
+       ubi_free_vid_hdr(ubi, vh);
+       return err;
+}
+
+/**
+ * ubi_add_to_av - add used physical eraseblock to the attaching information.
+ * @ubi: UBI device description object
+ * @ai: attaching information
+ * @pnum: the physical eraseblock number
+ * @ec: erase counter
+ * @vid_hdr: the volume identifier header
+ * @bitflips: if bit-flips were detected when this physical eraseblock was read
+ *
+ * This function adds information about a used physical eraseblock to the
+ * 'used' tree of the corresponding volume. The function is rather complex
+ * because it has to handle cases when this is not the first physical
+ * eraseblock belonging to the same logical eraseblock, and the newer one has
+ * to be picked, while the older one has to be dropped. This function returns
+ * zero in case of success and a negative error code in case of failure.
+ */
+int ubi_add_to_av(struct ubi_device *ubi, struct ubi_attach_info *ai, int pnum,
+                 int ec, const struct ubi_vid_hdr *vid_hdr, int bitflips)
+{
+       int err, vol_id, lnum;
+       unsigned long long sqnum;
+       struct ubi_ainf_volume *av;
+       struct ubi_ainf_peb *aeb;
+       struct rb_node **p, *parent = NULL;
+
+       vol_id = be32_to_cpu(vid_hdr->vol_id);
+       lnum = be32_to_cpu(vid_hdr->lnum);
+       sqnum = be64_to_cpu(vid_hdr->sqnum);
+
+       dbg_bld("PEB %d, LEB %d:%d, EC %d, sqnum %llu, bitflips %d",
+               pnum, vol_id, lnum, ec, sqnum, bitflips);
+
+       av = add_volume(ai, vol_id, pnum, vid_hdr);
+       if (IS_ERR(av))
+               return PTR_ERR(av);
+
+       if (ai->max_sqnum < sqnum)
+               ai->max_sqnum = sqnum;
+
+       /*
+        * Walk the RB-tree of logical eraseblocks of volume @vol_id to look
+        * if this is the first instance of this logical eraseblock or not.
+        */
+       p = &av->root.rb_node;
+       while (*p) {
+               int cmp_res;
+
+               parent = *p;
+               aeb = rb_entry(parent, struct ubi_ainf_peb, u.rb);
+               if (lnum != aeb->lnum) {
+                       if (lnum < aeb->lnum)
+                               p = &(*p)->rb_left;
+                       else
+                               p = &(*p)->rb_right;
+                       continue;
+               }
+
+               /*
+                * There is already a physical eraseblock describing the same
+                * logical eraseblock present.
+                */
+
+               dbg_bld("this LEB already exists: PEB %d, sqnum %llu, EC %d",
+                       aeb->pnum, aeb->sqnum, aeb->ec);
+
+               /*
+                * Make sure that the logical eraseblocks have different
+                * sequence numbers. Otherwise the image is bad.
+                *
+                * However, if the sequence number is zero, we assume it must
+                * be an ancient UBI image from the era when UBI did not have
+                * sequence numbers. We still can attach these images, unless
+                * there is a need to distinguish between old and new
+                * eraseblocks, in which case we'll refuse the image in
+                * 'ubi_compare_lebs()'. In other words, we attach old clean
+                * images, but refuse attaching old images with duplicated
+                * logical eraseblocks because there was an unclean reboot.
+                */
+               if (aeb->sqnum == sqnum && sqnum != 0) {
+                       ubi_err("two LEBs with same sequence number %llu",
+                               sqnum);
+                       ubi_dump_aeb(aeb, 0);
+                       ubi_dump_vid_hdr(vid_hdr);
+                       return -EINVAL;
+               }
+
+               /*
+                * Now we have to drop the older one and preserve the newer
+                * one.
+                */
+               cmp_res = ubi_compare_lebs(ubi, aeb, pnum, vid_hdr);
+               if (cmp_res < 0)
+                       return cmp_res;
+
+               if (cmp_res & 1) {
+                       /*
+                        * This logical eraseblock is newer than the one
+                        * found earlier.
+                        */
+                       err = validate_vid_hdr(vid_hdr, av, pnum);
+                       if (err)
+                               return err;
+
+                       err = add_to_list(ai, aeb->pnum, aeb->vol_id,
+                                         aeb->lnum, aeb->ec, cmp_res & 4,
+                                         &ai->erase);
+                       if (err)
+                               return err;
+
+                       aeb->ec = ec;
+                       aeb->pnum = pnum;
+                       aeb->vol_id = vol_id;
+                       aeb->lnum = lnum;
+                       aeb->scrub = ((cmp_res & 2) || bitflips);
+                       aeb->copy_flag = vid_hdr->copy_flag;
+                       aeb->sqnum = sqnum;
+
+                       if (av->highest_lnum == lnum)
+                               av->last_data_size =
+                                       be32_to_cpu(vid_hdr->data_size);
+
+                       return 0;
+               } else {
+                       /*
+                        * This logical eraseblock is older than the one found
+                        * previously.
+                        */
+                       return add_to_list(ai, pnum, vol_id, lnum, ec,
+                                          cmp_res & 4, &ai->erase);
+               }
+       }
+
+       /*
+        * We've met this logical eraseblock for the first time, add it to the
+        * attaching information.
+        */
+
+       err = validate_vid_hdr(vid_hdr, av, pnum);
+       if (err)
+               return err;
+
+       aeb = kmem_cache_alloc(ai->aeb_slab_cache, GFP_KERNEL);
+       if (!aeb)
+               return -ENOMEM;
+
+       aeb->ec = ec;
+       aeb->pnum = pnum;
+       aeb->vol_id = vol_id;
+       aeb->lnum = lnum;
+       aeb->scrub = bitflips;
+       aeb->copy_flag = vid_hdr->copy_flag;
+       aeb->sqnum = sqnum;
+
+       if (av->highest_lnum <= lnum) {
+               av->highest_lnum = lnum;
+               av->last_data_size = be32_to_cpu(vid_hdr->data_size);
+       }
+
+       av->leb_count += 1;
+       rb_link_node(&aeb->u.rb, parent, p);
+       rb_insert_color(&aeb->u.rb, &av->root);
+       return 0;
+}
+
+/**
+ * ubi_find_av - find volume in the attaching information.
+ * @ai: attaching information
+ * @vol_id: the requested volume ID
+ *
+ * This function returns a pointer to the volume description or %NULL if there
+ * are no data about this volume in the attaching information.
+ */
+struct ubi_ainf_volume *ubi_find_av(const struct ubi_attach_info *ai,
+                                   int vol_id)
+{
+       struct ubi_ainf_volume *av;
+       struct rb_node *p = ai->volumes.rb_node;
+
+       while (p) {
+               av = rb_entry(p, struct ubi_ainf_volume, rb);
+
+               if (vol_id == av->vol_id)
+                       return av;
+
+               if (vol_id > av->vol_id)
+                       p = p->rb_left;
+               else
+                       p = p->rb_right;
+       }
+
+       return NULL;
+}
+
+/**
+ * ubi_remove_av - delete attaching information about a volume.
+ * @ai: attaching information
+ * @av: the volume attaching information to delete
+ */
+void ubi_remove_av(struct ubi_attach_info *ai, struct ubi_ainf_volume *av)
+{
+       struct rb_node *rb;
+       struct ubi_ainf_peb *aeb;
+
+       dbg_bld("remove attaching information about volume %d", av->vol_id);
+
+       while ((rb = rb_first(&av->root))) {
+               aeb = rb_entry(rb, struct ubi_ainf_peb, u.rb);
+               rb_erase(&aeb->u.rb, &av->root);
+               list_add_tail(&aeb->u.list, &ai->erase);
+       }
+
+       rb_erase(&av->rb, &ai->volumes);
+       kfree(av);
+       ai->vols_found -= 1;
+}
+
+/**
+ * early_erase_peb - erase a physical eraseblock.
+ * @ubi: UBI device description object
+ * @ai: attaching information
+ * @pnum: physical eraseblock number to erase;
+ * @ec: erase counter value to write (%UBI_UNKNOWN if it is unknown)
+ *
+ * This function erases physical eraseblock 'pnum', and writes the erase
+ * counter header to it. This function should only be used on UBI device
+ * initialization stages, when the EBA sub-system had not been yet initialized.
+ * This function returns zero in case of success and a negative error code in
+ * case of failure.
+ */
+static int early_erase_peb(struct ubi_device *ubi,
+                          const struct ubi_attach_info *ai, int pnum, int ec)
+{
+       int err;
+       struct ubi_ec_hdr *ec_hdr;
+
+       if ((long long)ec >= UBI_MAX_ERASECOUNTER) {
+               /*
+                * Erase counter overflow. Upgrade UBI and use 64-bit
+                * erase counters internally.
+                */
+               ubi_err("erase counter overflow at PEB %d, EC %d", pnum, ec);
+               return -EINVAL;
+       }
+
+       ec_hdr = kzalloc(ubi->ec_hdr_alsize, GFP_KERNEL);
+       if (!ec_hdr)
+               return -ENOMEM;
+
+       ec_hdr->ec = cpu_to_be64(ec);
+
+       err = ubi_io_sync_erase(ubi, pnum, 0);
+       if (err < 0)
+               goto out_free;
+
+       err = ubi_io_write_ec_hdr(ubi, pnum, ec_hdr);
+
+out_free:
+       kfree(ec_hdr);
+       return err;
+}
+
+/**
+ * ubi_early_get_peb - get a free physical eraseblock.
+ * @ubi: UBI device description object
+ * @ai: attaching information
+ *
+ * This function returns a free physical eraseblock. It is supposed to be
+ * called on the UBI initialization stages when the wear-leveling sub-system is
+ * not initialized yet. This function picks a physical eraseblocks from one of
+ * the lists, writes the EC header if it is needed, and removes it from the
+ * list.
+ *
+ * This function returns a pointer to the "aeb" of the found free PEB in case
+ * of success and an error code in case of failure.
+ */
+struct ubi_ainf_peb *ubi_early_get_peb(struct ubi_device *ubi,
+                                      struct ubi_attach_info *ai)
+{
+       int err = 0;
+       struct ubi_ainf_peb *aeb, *tmp_aeb;
+
+       if (!list_empty(&ai->free)) {
+               aeb = list_entry(ai->free.next, struct ubi_ainf_peb, u.list);
+               list_del(&aeb->u.list);
+               dbg_bld("return free PEB %d, EC %d", aeb->pnum, aeb->ec);
+               return aeb;
+       }
+
+       /*
+        * We try to erase the first physical eraseblock from the erase list
+        * and pick it if we succeed, or try to erase the next one if not. And
+        * so forth. We don't want to take care about bad eraseblocks here -
+        * they'll be handled later.
+        */
+       list_for_each_entry_safe(aeb, tmp_aeb, &ai->erase, u.list) {
+               if (aeb->ec == UBI_UNKNOWN)
+                       aeb->ec = ai->mean_ec;
+
+               err = early_erase_peb(ubi, ai, aeb->pnum, aeb->ec+1);
+               if (err)
+                       continue;
+
+               aeb->ec += 1;
+               list_del(&aeb->u.list);
+               dbg_bld("return PEB %d, EC %d", aeb->pnum, aeb->ec);
+               return aeb;
+       }
+
+       ubi_err("no free eraseblocks");
+       return ERR_PTR(-ENOSPC);
+}
+
+/**
+ * check_corruption - check the data area of PEB.
+ * @ubi: UBI device description object
+ * @vid_hdr: the (corrupted) VID header of this PEB
+ * @pnum: the physical eraseblock number to check
+ *
+ * This is a helper function which is used to distinguish between VID header
+ * corruptions caused by power cuts and other reasons. If the PEB contains only
+ * 0xFF bytes in the data area, the VID header is most probably corrupted
+ * because of a power cut (%0 is returned in this case). Otherwise, it was
+ * probably corrupted for some other reasons (%1 is returned in this case). A
+ * negative error code is returned if a read error occurred.
+ *
+ * If the corruption reason was a power cut, UBI can safely erase this PEB.
+ * Otherwise, it should preserve it to avoid possibly destroying important
+ * information.
+ */
+static int check_corruption(struct ubi_device *ubi, struct ubi_vid_hdr *vid_hdr,
+                           int pnum)
+{
+       int err;
+
+       mutex_lock(&ubi->buf_mutex);
+       memset(ubi->peb_buf, 0x00, ubi->leb_size);
+
+       err = ubi_io_read(ubi, ubi->peb_buf, pnum, ubi->leb_start,
+                         ubi->leb_size);
+       if (err == UBI_IO_BITFLIPS || mtd_is_eccerr(err)) {
+               /*
+                * Bit-flips or integrity errors while reading the data area.
+                * It is difficult to say for sure what type of corruption is
+                * this, but presumably a power cut happened while this PEB was
+                * erased, so it became unstable and corrupted, and should be
+                * erased.
+                */
+               err = 0;
+               goto out_unlock;
+       }
+
+       if (err)
+               goto out_unlock;
+
+       if (ubi_check_pattern(ubi->peb_buf, 0xFF, ubi->leb_size))
+               goto out_unlock;
+
+       ubi_err("PEB %d contains corrupted VID header, and the data does not contain all 0xFF",
+               pnum);
+       ubi_err("this may be a non-UBI PEB or a severe VID header corruption which requires manual inspection");
+       ubi_dump_vid_hdr(vid_hdr);
+       pr_err("hexdump of PEB %d offset %d, length %d",
+              pnum, ubi->leb_start, ubi->leb_size);
+       ubi_dbg_print_hex_dump(KERN_DEBUG, "", DUMP_PREFIX_OFFSET, 32, 1,
+                              ubi->peb_buf, ubi->leb_size, 1);
+       err = 1;
+
+out_unlock:
+       mutex_unlock(&ubi->buf_mutex);
+       return err;
+}
+
+/**
+ * scan_peb - scan and process UBI headers of a PEB.
+ * @ubi: UBI device description object
+ * @ai: attaching information
+ * @pnum: the physical eraseblock number
+ * @vid: The volume ID of the found volume will be stored in this pointer
+ * @sqnum: The sqnum of the found volume will be stored in this pointer
+ *
+ * This function reads UBI headers of PEB @pnum, checks them, and adds
+ * information about this PEB to the corresponding list or RB-tree in the
+ * "attaching info" structure. Returns zero if the physical eraseblock was
+ * successfully handled and a negative error code in case of failure.
+ */
+static int scan_peb(struct ubi_device *ubi, struct ubi_attach_info *ai,
+                   int pnum, int *vid, unsigned long long *sqnum)
+{
+       long long uninitialized_var(ec);
+       int err, bitflips = 0, vol_id = -1, ec_err = 0;
+
+       dbg_bld("scan PEB %d", pnum);
+
+       /* Skip bad physical eraseblocks */
+       err = ubi_io_is_bad(ubi, pnum);
+       if (err < 0)
+               return err;
+       else if (err) {
+               ai->bad_peb_count += 1;
+               return 0;
+       }
+
+       err = ubi_io_read_ec_hdr(ubi, pnum, ech, 0);
+       if (err < 0)
+               return err;
+       switch (err) {
+       case 0:
+               break;
+       case UBI_IO_BITFLIPS:
+               bitflips = 1;
+               break;
+       case UBI_IO_FF:
+               ai->empty_peb_count += 1;
+               return add_to_list(ai, pnum, UBI_UNKNOWN, UBI_UNKNOWN,
+                                  UBI_UNKNOWN, 0, &ai->erase);
+       case UBI_IO_FF_BITFLIPS:
+               ai->empty_peb_count += 1;
+               return add_to_list(ai, pnum, UBI_UNKNOWN, UBI_UNKNOWN,
+                                  UBI_UNKNOWN, 1, &ai->erase);
+       case UBI_IO_BAD_HDR_EBADMSG:
+       case UBI_IO_BAD_HDR:
+               /*
+                * We have to also look at the VID header, possibly it is not
+                * corrupted. Set %bitflips flag in order to make this PEB be
+                * moved and EC be re-created.
+                */
+               ec_err = err;
+               ec = UBI_UNKNOWN;
+               bitflips = 1;
+               break;
+       default:
+               ubi_err("'ubi_io_read_ec_hdr()' returned unknown code %d", err);
+               return -EINVAL;
+       }
+
+       if (!ec_err) {
+               int image_seq;
+
+               /* Make sure UBI version is OK */
+               if (ech->version != UBI_VERSION) {
+                       ubi_err("this UBI version is %d, image version is %d",
+                               UBI_VERSION, (int)ech->version);
+                       return -EINVAL;
+               }
+
+               ec = be64_to_cpu(ech->ec);
+               if (ec > UBI_MAX_ERASECOUNTER) {
+                       /*
+                        * Erase counter overflow. The EC headers have 64 bits
+                        * reserved, but we anyway make use of only 31 bit
+                        * values, as this seems to be enough for any existing
+                        * flash. Upgrade UBI and use 64-bit erase counters
+                        * internally.
+                        */
+                       ubi_err("erase counter overflow, max is %d",
+                               UBI_MAX_ERASECOUNTER);
+                       ubi_dump_ec_hdr(ech);
+                       return -EINVAL;
+               }
+
+               /*
+                * Make sure that all PEBs have the same image sequence number.
+                * This allows us to detect situations when users flash UBI
+                * images incorrectly, so that the flash has the new UBI image
+                * and leftovers from the old one. This feature was added
+                * relatively recently, and the sequence number was always
+                * zero, because old UBI implementations always set it to zero.
+                * For this reasons, we do not panic if some PEBs have zero
+                * sequence number, while other PEBs have non-zero sequence
+                * number.
+                */
+               image_seq = be32_to_cpu(ech->image_seq);
+               if (!ubi->image_seq)
+                       ubi->image_seq = image_seq;
+               if (image_seq && ubi->image_seq != image_seq) {
+                       ubi_err("bad image sequence number %d in PEB %d, expected %d",
+                               image_seq, pnum, ubi->image_seq);
+                       ubi_dump_ec_hdr(ech);
+                       return -EINVAL;
+               }
+       }
+
+       /* OK, we've done with the EC header, let's look at the VID header */
+
+       err = ubi_io_read_vid_hdr(ubi, pnum, vidh, 0);
+       if (err < 0)
+               return err;
+       switch (err) {
+       case 0:
+               break;
+       case UBI_IO_BITFLIPS:
+               bitflips = 1;
+               break;
+       case UBI_IO_BAD_HDR_EBADMSG:
+               if (ec_err == UBI_IO_BAD_HDR_EBADMSG)
+                       /*
+                        * Both EC and VID headers are corrupted and were read
+                        * with data integrity error, probably this is a bad
+                        * PEB, bit it is not marked as bad yet. This may also
+                        * be a result of power cut during erasure.
+                        */
+                       ai->maybe_bad_peb_count += 1;
+       case UBI_IO_BAD_HDR:
+               if (ec_err)
+                       /*
+                        * Both headers are corrupted. There is a possibility
+                        * that this a valid UBI PEB which has corresponding
+                        * LEB, but the headers are corrupted. However, it is
+                        * impossible to distinguish it from a PEB which just
+                        * contains garbage because of a power cut during erase
+                        * operation. So we just schedule this PEB for erasure.
+                        *
+                        * Besides, in case of NOR flash, we deliberately
+                        * corrupt both headers because NOR flash erasure is
+                        * slow and can start from the end.
+                        */
+                       err = 0;
+               else
+                       /*
+                        * The EC was OK, but the VID header is corrupted. We
+                        * have to check what is in the data area.
+                        */
+                       err = check_corruption(ubi, vidh, pnum);
+
+               if (err < 0)
+                       return err;
+               else if (!err)
+                       /* This corruption is caused by a power cut */
+                       err = add_to_list(ai, pnum, UBI_UNKNOWN,
+                                         UBI_UNKNOWN, ec, 1, &ai->erase);
+               else
+                       /* This is an unexpected corruption */
+                       err = add_corrupted(ai, pnum, ec);
+               if (err)
+                       return err;
+               goto adjust_mean_ec;
+       case UBI_IO_FF_BITFLIPS:
+               err = add_to_list(ai, pnum, UBI_UNKNOWN, UBI_UNKNOWN,
+                                 ec, 1, &ai->erase);
+               if (err)
+                       return err;
+               goto adjust_mean_ec;
+       case UBI_IO_FF:
+               if (ec_err || bitflips)
+                       err = add_to_list(ai, pnum, UBI_UNKNOWN,
+                                         UBI_UNKNOWN, ec, 1, &ai->erase);
+               else
+                       err = add_to_list(ai, pnum, UBI_UNKNOWN,
+                                         UBI_UNKNOWN, ec, 0, &ai->free);
+               if (err)
+                       return err;
+               goto adjust_mean_ec;
+       default:
+               ubi_err("'ubi_io_read_vid_hdr()' returned unknown code %d",
+                       err);
+               return -EINVAL;
+       }
+
+       vol_id = be32_to_cpu(vidh->vol_id);
+       if (vid)
+               *vid = vol_id;
+       if (sqnum)
+               *sqnum = be64_to_cpu(vidh->sqnum);
+       if (vol_id > UBI_MAX_VOLUMES && vol_id != UBI_LAYOUT_VOLUME_ID) {
+               int lnum = be32_to_cpu(vidh->lnum);
+
+               /* Unsupported internal volume */
+               switch (vidh->compat) {
+               case UBI_COMPAT_DELETE:
+                       if (vol_id != UBI_FM_SB_VOLUME_ID
+                           && vol_id != UBI_FM_DATA_VOLUME_ID) {
+                               ubi_msg("\"delete\" compatible internal volume %d:%d found, will remove it",
+                                       vol_id, lnum);
+                       }
+                       err = add_to_list(ai, pnum, vol_id, lnum,
+                                         ec, 1, &ai->erase);
+                       if (err)
+                               return err;
+                       return 0;
+
+               case UBI_COMPAT_RO:
+                       ubi_msg("read-only compatible internal volume %d:%d found, switch to read-only mode",
+                               vol_id, lnum);
+                       ubi->ro_mode = 1;
+                       break;
+
+               case UBI_COMPAT_PRESERVE:
+                       ubi_msg("\"preserve\" compatible internal volume %d:%d found",
+                               vol_id, lnum);
+                       err = add_to_list(ai, pnum, vol_id, lnum,
+                                         ec, 0, &ai->alien);
+                       if (err)
+                               return err;
+                       return 0;
+
+               case UBI_COMPAT_REJECT:
+                       ubi_err("incompatible internal volume %d:%d found",
+                               vol_id, lnum);
+                       return -EINVAL;
+               }
+       }
+
+       if (ec_err)
+               ubi_warn("valid VID header but corrupted EC header at PEB %d",
+                        pnum);
+       err = ubi_add_to_av(ubi, ai, pnum, ec, vidh, bitflips);
+       if (err)
+               return err;
+
+adjust_mean_ec:
+       if (!ec_err) {
+               ai->ec_sum += ec;
+               ai->ec_count += 1;
+               if (ec > ai->max_ec)
+                       ai->max_ec = ec;
+               if (ec < ai->min_ec)
+                       ai->min_ec = ec;
+       }
+
+       return 0;
+}
+
+/**
+ * late_analysis - analyze the overall situation with PEB.
+ * @ubi: UBI device description object
+ * @ai: attaching information
+ *
+ * This is a helper function which takes a look what PEBs we have after we
+ * gather information about all of them ("ai" is compete). It decides whether
+ * the flash is empty and should be formatted of whether there are too many
+ * corrupted PEBs and we should not attach this MTD device. Returns zero if we
+ * should proceed with attaching the MTD device, and %-EINVAL if we should not.
+ */
+static int late_analysis(struct ubi_device *ubi, struct ubi_attach_info *ai)
+{
+       struct ubi_ainf_peb *aeb;
+       int max_corr, peb_count;
+
+       peb_count = ubi->peb_count - ai->bad_peb_count - ai->alien_peb_count;
+       max_corr = peb_count / 20 ?: 8;
+
+       /*
+        * Few corrupted PEBs is not a problem and may be just a result of
+        * unclean reboots. However, many of them may indicate some problems
+        * with the flash HW or driver.
+        */
+       if (ai->corr_peb_count) {
+               ubi_err("%d PEBs are corrupted and preserved",
+                       ai->corr_peb_count);
+               pr_err("Corrupted PEBs are:");
+               list_for_each_entry(aeb, &ai->corr, u.list)
+                       pr_cont(" %d", aeb->pnum);
+               pr_cont("\n");
+
+               /*
+                * If too many PEBs are corrupted, we refuse attaching,
+                * otherwise, only print a warning.
+                */
+               if (ai->corr_peb_count >= max_corr) {
+                       ubi_err("too many corrupted PEBs, refusing");
+                       return -EINVAL;
+               }
+       }
+
+       if (ai->empty_peb_count + ai->maybe_bad_peb_count == peb_count) {
+               /*
+                * All PEBs are empty, or almost all - a couple PEBs look like
+                * they may be bad PEBs which were not marked as bad yet.
+                *
+                * This piece of code basically tries to distinguish between
+                * the following situations:
+                *
+                * 1. Flash is empty, but there are few bad PEBs, which are not
+                *    marked as bad so far, and which were read with error. We
+                *    want to go ahead and format this flash. While formatting,
+                *    the faulty PEBs will probably be marked as bad.
+                *
+                * 2. Flash contains non-UBI data and we do not want to format
+                *    it and destroy possibly important information.
+                */
+               if (ai->maybe_bad_peb_count <= 2) {
+                       ai->is_empty = 1;
+                       ubi_msg("empty MTD device detected");
+                       get_random_bytes(&ubi->image_seq,
+                                        sizeof(ubi->image_seq));
+               } else {
+                       ubi_err("MTD device is not UBI-formatted and possibly contains non-UBI data - refusing it");
+                       return -EINVAL;
+               }
+
+       }
+
+       return 0;
+}
+
+/**
+ * destroy_av - free volume attaching information.
+ * @av: volume attaching information
+ * @ai: attaching information
+ *
+ * This function destroys the volume attaching information.
+ */
+static void destroy_av(struct ubi_attach_info *ai, struct ubi_ainf_volume *av)
+{
+       struct ubi_ainf_peb *aeb;
+       struct rb_node *this = av->root.rb_node;
+
+       while (this) {
+               if (this->rb_left)
+                       this = this->rb_left;
+               else if (this->rb_right)
+                       this = this->rb_right;
+               else {
+                       aeb = rb_entry(this, struct ubi_ainf_peb, u.rb);
+                       this = rb_parent(this);
+                       if (this) {
+                               if (this->rb_left == &aeb->u.rb)
+                                       this->rb_left = NULL;
+                               else
+                                       this->rb_right = NULL;
+                       }
+
+                       kmem_cache_free(ai->aeb_slab_cache, aeb);
+               }
+       }
+       kfree(av);
+}
+
+/**
+ * destroy_ai - destroy attaching information.
+ * @ai: attaching information
+ */
+static void destroy_ai(struct ubi_attach_info *ai)
+{
+       struct ubi_ainf_peb *aeb, *aeb_tmp;
+       struct ubi_ainf_volume *av;
+       struct rb_node *rb;
+
+       list_for_each_entry_safe(aeb, aeb_tmp, &ai->alien, u.list) {
+               list_del(&aeb->u.list);
+               kmem_cache_free(ai->aeb_slab_cache, aeb);
+       }
+       list_for_each_entry_safe(aeb, aeb_tmp, &ai->erase, u.list) {
+               list_del(&aeb->u.list);
+               kmem_cache_free(ai->aeb_slab_cache, aeb);
+       }
+       list_for_each_entry_safe(aeb, aeb_tmp, &ai->corr, u.list) {
+               list_del(&aeb->u.list);
+               kmem_cache_free(ai->aeb_slab_cache, aeb);
+       }
+       list_for_each_entry_safe(aeb, aeb_tmp, &ai->free, u.list) {
+               list_del(&aeb->u.list);
+               kmem_cache_free(ai->aeb_slab_cache, aeb);
+       }
+
+       /* Destroy the volume RB-tree */
+       rb = ai->volumes.rb_node;
+       while (rb) {
+               if (rb->rb_left)
+                       rb = rb->rb_left;
+               else if (rb->rb_right)
+                       rb = rb->rb_right;
+               else {
+                       av = rb_entry(rb, struct ubi_ainf_volume, rb);
+
+                       rb = rb_parent(rb);
+                       if (rb) {
+                               if (rb->rb_left == &av->rb)
+                                       rb->rb_left = NULL;
+                               else
+                                       rb->rb_right = NULL;
+                       }
+
+                       destroy_av(ai, av);
+               }
+       }
+
+       if (ai->aeb_slab_cache)
+               kmem_cache_destroy(ai->aeb_slab_cache);
+
+       kfree(ai);
+}
+
+/**
+ * scan_all - scan entire MTD device.
+ * @ubi: UBI device description object
+ * @ai: attach info object
+ * @start: start scanning at this PEB
+ *
+ * This function does full scanning of an MTD device and returns complete
+ * information about it in form of a "struct ubi_attach_info" object. In case
+ * of failure, an error code is returned.
+ */
+static int scan_all(struct ubi_device *ubi, struct ubi_attach_info *ai,
+                   int start)
+{
+       int err, pnum;
+       struct rb_node *rb1, *rb2;
+       struct ubi_ainf_volume *av;
+       struct ubi_ainf_peb *aeb;
+
+       err = -ENOMEM;
+
+       ech = kzalloc(ubi->ec_hdr_alsize, GFP_KERNEL);
+       if (!ech)
+               return err;
+
+       vidh = ubi_zalloc_vid_hdr(ubi, GFP_KERNEL);
+       if (!vidh)
+               goto out_ech;
+
+       for (pnum = start; pnum < ubi->peb_count; pnum++) {
+               cond_resched();
+
+               dbg_gen("process PEB %d", pnum);
+               err = scan_peb(ubi, ai, pnum, NULL, NULL);
+               if (err < 0)
+                       goto out_vidh;
+       }
+
+       ubi_msg("scanning is finished");
+
+       /* Calculate mean erase counter */
+       if (ai->ec_count)
+               ai->mean_ec = div_u64(ai->ec_sum, ai->ec_count);
+
+       err = late_analysis(ubi, ai);
+       if (err)
+               goto out_vidh;
+
+       /*
+        * In case of unknown erase counter we use the mean erase counter
+        * value.
+        */
+       ubi_rb_for_each_entry(rb1, av, &ai->volumes, rb) {
+               ubi_rb_for_each_entry(rb2, aeb, &av->root, u.rb)
+                       if (aeb->ec == UBI_UNKNOWN)
+                               aeb->ec = ai->mean_ec;
+       }
+
+       list_for_each_entry(aeb, &ai->free, u.list) {
+               if (aeb->ec == UBI_UNKNOWN)
+                       aeb->ec = ai->mean_ec;
+       }
+
+       list_for_each_entry(aeb, &ai->corr, u.list)
+               if (aeb->ec == UBI_UNKNOWN)
+                       aeb->ec = ai->mean_ec;
+
+       list_for_each_entry(aeb, &ai->erase, u.list)
+               if (aeb->ec == UBI_UNKNOWN)
+                       aeb->ec = ai->mean_ec;
+
+       err = self_check_ai(ubi, ai);
+       if (err)
+               goto out_vidh;
+
+       ubi_free_vid_hdr(ubi, vidh);
+       kfree(ech);
+
+       return 0;
+
+out_vidh:
+       ubi_free_vid_hdr(ubi, vidh);
+out_ech:
+       kfree(ech);
+       return err;
+}
+
+#ifdef CONFIG_MTD_UBI_FASTMAP
+
+/**
+ * scan_fastmap - try to find a fastmap and attach from it.
+ * @ubi: UBI device description object
+ * @ai: attach info object
+ *
+ * Returns 0 on success, negative return values indicate an internal
+ * error.
+ * UBI_NO_FASTMAP denotes that no fastmap was found.
+ * UBI_BAD_FASTMAP denotes that the found fastmap was invalid.
+ */
+static int scan_fast(struct ubi_device *ubi, struct ubi_attach_info *ai)
+{
+       int err, pnum, fm_anchor = -1;
+       unsigned long long max_sqnum = 0;
+
+       err = -ENOMEM;
+
+       ech = kzalloc(ubi->ec_hdr_alsize, GFP_KERNEL);
+       if (!ech)
+               goto out;
+
+       vidh = ubi_zalloc_vid_hdr(ubi, GFP_KERNEL);
+       if (!vidh)
+               goto out_ech;
+
+       for (pnum = 0; pnum < UBI_FM_MAX_START; pnum++) {
+               int vol_id = -1;
+               unsigned long long sqnum = -1;
+               cond_resched();
+
+               dbg_gen("process PEB %d", pnum);
+               err = scan_peb(ubi, ai, pnum, &vol_id, &sqnum);
+               if (err < 0)
+                       goto out_vidh;
+
+               if (vol_id == UBI_FM_SB_VOLUME_ID && sqnum > max_sqnum) {
+                       max_sqnum = sqnum;
+                       fm_anchor = pnum;
+               }
+       }
+
+       ubi_free_vid_hdr(ubi, vidh);
+       kfree(ech);
+
+       if (fm_anchor < 0)
+               return UBI_NO_FASTMAP;
+
+       return ubi_scan_fastmap(ubi, ai, fm_anchor);
+
+out_vidh:
+       ubi_free_vid_hdr(ubi, vidh);
+out_ech:
+       kfree(ech);
+out:
+       return err;
+}
+
+#endif
+
+static struct ubi_attach_info *alloc_ai(const char *slab_name)
+{
+       struct ubi_attach_info *ai;
+
+       ai = kzalloc(sizeof(struct ubi_attach_info), GFP_KERNEL);
+       if (!ai)
+               return ai;
+
+       INIT_LIST_HEAD(&ai->corr);
+       INIT_LIST_HEAD(&ai->free);
+       INIT_LIST_HEAD(&ai->erase);
+       INIT_LIST_HEAD(&ai->alien);
+       ai->volumes = RB_ROOT;
+       ai->aeb_slab_cache = kmem_cache_create(slab_name,
+                                              sizeof(struct ubi_ainf_peb),
+                                              0, 0, NULL);
+       if (!ai->aeb_slab_cache) {
+               kfree(ai);
+               ai = NULL;
+       }
+
+       return ai;
+}
+
+/**
+ * ubi_attach - attach an MTD device.
+ * @ubi: UBI device descriptor
+ * @force_scan: if set to non-zero attach by scanning
+ *
+ * This function returns zero in case of success and a negative error code in
+ * case of failure.
+ */
+int ubi_attach(struct ubi_device *ubi, int force_scan)
+{
+       int err;
+       struct ubi_attach_info *ai;
+
+       ai = alloc_ai("ubi_aeb_slab_cache");
+       if (!ai)
+               return -ENOMEM;
+
+#ifdef CONFIG_MTD_UBI_FASTMAP
+       /* On small flash devices we disable fastmap in any case. */
+       if ((int)mtd_div_by_eb(ubi->mtd->size, ubi->mtd) <= UBI_FM_MAX_START) {
+               ubi->fm_disabled = 1;
+               force_scan = 1;
+       }
+
+       if (force_scan)
+               err = scan_all(ubi, ai, 0);
+       else {
+               err = scan_fast(ubi, ai);
+               if (err > 0) {
+                       if (err != UBI_NO_FASTMAP) {
+                               destroy_ai(ai);
+                               ai = alloc_ai("ubi_aeb_slab_cache2");
+                               if (!ai)
+                                       return -ENOMEM;
+
+                               err = scan_all(ubi, ai, 0);
+                       } else {
+                               err = scan_all(ubi, ai, UBI_FM_MAX_START);
+                       }
+               }
+       }
+#else
+       err = scan_all(ubi, ai, 0);
+#endif
+       if (err)
+               goto out_ai;
+
+       ubi->bad_peb_count = ai->bad_peb_count;
+       ubi->good_peb_count = ubi->peb_count - ubi->bad_peb_count;
+       ubi->corr_peb_count = ai->corr_peb_count;
+       ubi->max_ec = ai->max_ec;
+       ubi->mean_ec = ai->mean_ec;
+       dbg_gen("max. sequence number:       %llu", ai->max_sqnum);
+
+       err = ubi_read_volume_table(ubi, ai);
+       if (err)
+               goto out_ai;
+
+       err = ubi_wl_init(ubi, ai);
+       if (err)
+               goto out_vtbl;
+
+       err = ubi_eba_init(ubi, ai);
+       if (err)
+               goto out_wl;
+
+#ifdef CONFIG_MTD_UBI_FASTMAP
+       if (ubi->fm && ubi_dbg_chk_gen(ubi)) {
+               struct ubi_attach_info *scan_ai;
+
+               scan_ai = alloc_ai("ubi_ckh_aeb_slab_cache");
+               if (!scan_ai) {
+                       err = -ENOMEM;
+                       goto out_wl;
+               }
+
+               err = scan_all(ubi, scan_ai, 0);
+               if (err) {
+                       destroy_ai(scan_ai);
+                       goto out_wl;
+               }
+
+               err = self_check_eba(ubi, ai, scan_ai);
+               destroy_ai(scan_ai);
+
+               if (err)
+                       goto out_wl;
+       }
+#endif
+
+       destroy_ai(ai);
+       return 0;
+
+out_wl:
+       ubi_wl_close(ubi);
+out_vtbl:
+       ubi_free_internal_volumes(ubi);
+       vfree(ubi->vtbl);
+out_ai:
+       destroy_ai(ai);
+       return err;
+}
+
+/**
+ * self_check_ai - check the attaching information.
+ * @ubi: UBI device description object
+ * @ai: attaching information
+ *
+ * This function returns zero if the attaching information is all right, and a
+ * negative error code if not or if an error occurred.
+ */
+static int self_check_ai(struct ubi_device *ubi, struct ubi_attach_info *ai)
+{
+       int pnum, err, vols_found = 0;
+       struct rb_node *rb1, *rb2;
+       struct ubi_ainf_volume *av;
+       struct ubi_ainf_peb *aeb, *last_aeb;
+       uint8_t *buf;
+
+       if (!ubi_dbg_chk_gen(ubi))
+               return 0;
+
+       /*
+        * At first, check that attaching information is OK.
+        */
+       ubi_rb_for_each_entry(rb1, av, &ai->volumes, rb) {
+               int leb_count = 0;
+
+               cond_resched();
+
+               vols_found += 1;
+
+               if (ai->is_empty) {
+                       ubi_err("bad is_empty flag");
+                       goto bad_av;
+               }
+
+               if (av->vol_id < 0 || av->highest_lnum < 0 ||
+                   av->leb_count < 0 || av->vol_type < 0 || av->used_ebs < 0 ||
+                   av->data_pad < 0 || av->last_data_size < 0) {
+                       ubi_err("negative values");
+                       goto bad_av;
+               }
+
+               if (av->vol_id >= UBI_MAX_VOLUMES &&
+                   av->vol_id < UBI_INTERNAL_VOL_START) {
+                       ubi_err("bad vol_id");
+                       goto bad_av;
+               }
+
+               if (av->vol_id > ai->highest_vol_id) {
+                       ubi_err("highest_vol_id is %d, but vol_id %d is there",
+                               ai->highest_vol_id, av->vol_id);
+                       goto out;
+               }
+
+               if (av->vol_type != UBI_DYNAMIC_VOLUME &&
+                   av->vol_type != UBI_STATIC_VOLUME) {
+                       ubi_err("bad vol_type");
+                       goto bad_av;
+               }
+
+               if (av->data_pad > ubi->leb_size / 2) {
+                       ubi_err("bad data_pad");
+                       goto bad_av;
+               }
+
+               last_aeb = NULL;
+               ubi_rb_for_each_entry(rb2, aeb, &av->root, u.rb) {
+                       cond_resched();
+
+                       last_aeb = aeb;
+                       leb_count += 1;
+
+                       if (aeb->pnum < 0 || aeb->ec < 0) {
+                               ubi_err("negative values");
+                               goto bad_aeb;
+                       }
+
+                       if (aeb->ec < ai->min_ec) {
+                               ubi_err("bad ai->min_ec (%d), %d found",
+                                       ai->min_ec, aeb->ec);
+                               goto bad_aeb;
+                       }
+
+                       if (aeb->ec > ai->max_ec) {
+                               ubi_err("bad ai->max_ec (%d), %d found",
+                                       ai->max_ec, aeb->ec);
+                               goto bad_aeb;
+                       }
+
+                       if (aeb->pnum >= ubi->peb_count) {
+                               ubi_err("too high PEB number %d, total PEBs %d",
+                                       aeb->pnum, ubi->peb_count);
+                               goto bad_aeb;
+                       }
+
+                       if (av->vol_type == UBI_STATIC_VOLUME) {
+                               if (aeb->lnum >= av->used_ebs) {
+                                       ubi_err("bad lnum or used_ebs");
+                                       goto bad_aeb;
+                               }
+                       } else {
+                               if (av->used_ebs != 0) {
+                                       ubi_err("non-zero used_ebs");
+                                       goto bad_aeb;
+                               }
+                       }
+
+                       if (aeb->lnum > av->highest_lnum) {
+                               ubi_err("incorrect highest_lnum or lnum");
+                               goto bad_aeb;
+                       }
+               }
+
+               if (av->leb_count != leb_count) {
+                       ubi_err("bad leb_count, %d objects in the tree",
+                               leb_count);
+                       goto bad_av;
+               }
+
+               if (!last_aeb)
+                       continue;
+
+               aeb = last_aeb;
+
+               if (aeb->lnum != av->highest_lnum) {
+                       ubi_err("bad highest_lnum");
+                       goto bad_aeb;
+               }
+       }
+
+       if (vols_found != ai->vols_found) {
+               ubi_err("bad ai->vols_found %d, should be %d",
+                       ai->vols_found, vols_found);
+               goto out;
+       }
+
+       /* Check that attaching information is correct */
+       ubi_rb_for_each_entry(rb1, av, &ai->volumes, rb) {
+               last_aeb = NULL;
+               ubi_rb_for_each_entry(rb2, aeb, &av->root, u.rb) {
+                       int vol_type;
+
+                       cond_resched();
+
+                       last_aeb = aeb;
+
+                       err = ubi_io_read_vid_hdr(ubi, aeb->pnum, vidh, 1);
+                       if (err && err != UBI_IO_BITFLIPS) {
+                               ubi_err("VID header is not OK (%d)", err);
+                               if (err > 0)
+                                       err = -EIO;
+                               return err;
+                       }
+
+                       vol_type = vidh->vol_type == UBI_VID_DYNAMIC ?
+                                  UBI_DYNAMIC_VOLUME : UBI_STATIC_VOLUME;
+                       if (av->vol_type != vol_type) {
+                               ubi_err("bad vol_type");
+                               goto bad_vid_hdr;
+                       }
+
+                       if (aeb->sqnum != be64_to_cpu(vidh->sqnum)) {
+                               ubi_err("bad sqnum %llu", aeb->sqnum);
+                               goto bad_vid_hdr;
+                       }
+
+                       if (av->vol_id != be32_to_cpu(vidh->vol_id)) {
+                               ubi_err("bad vol_id %d", av->vol_id);
+                               goto bad_vid_hdr;
+                       }
+
+                       if (av->compat != vidh->compat) {
+                               ubi_err("bad compat %d", vidh->compat);
+                               goto bad_vid_hdr;
+                       }
+
+                       if (aeb->lnum != be32_to_cpu(vidh->lnum)) {
+                               ubi_err("bad lnum %d", aeb->lnum);
+                               goto bad_vid_hdr;
+                       }
+
+                       if (av->used_ebs != be32_to_cpu(vidh->used_ebs)) {
+                               ubi_err("bad used_ebs %d", av->used_ebs);
+                               goto bad_vid_hdr;
+                       }
+
+                       if (av->data_pad != be32_to_cpu(vidh->data_pad)) {
+                               ubi_err("bad data_pad %d", av->data_pad);
+                               goto bad_vid_hdr;
+                       }
+               }
+
+               if (!last_aeb)
+                       continue;
+
+               if (av->highest_lnum != be32_to_cpu(vidh->lnum)) {
+                       ubi_err("bad highest_lnum %d", av->highest_lnum);
+                       goto bad_vid_hdr;
+               }
+
+               if (av->last_data_size != be32_to_cpu(vidh->data_size)) {
+                       ubi_err("bad last_data_size %d", av->last_data_size);
+                       goto bad_vid_hdr;
+               }
+       }
+
+       /*
+        * Make sure that all the physical eraseblocks are in one of the lists
+        * or trees.
+        */
+       buf = kzalloc(ubi->peb_count, GFP_KERNEL);
+       if (!buf)
+               return -ENOMEM;
+
+       for (pnum = 0; pnum < ubi->peb_count; pnum++) {
+               err = ubi_io_is_bad(ubi, pnum);
+               if (err < 0) {
+                       kfree(buf);
+                       return err;
+               } else if (err)
+                       buf[pnum] = 1;
+       }
+
+       ubi_rb_for_each_entry(rb1, av, &ai->volumes, rb)
+               ubi_rb_for_each_entry(rb2, aeb, &av->root, u.rb)
+                       buf[aeb->pnum] = 1;
+
+       list_for_each_entry(aeb, &ai->free, u.list)
+               buf[aeb->pnum] = 1;
+
+       list_for_each_entry(aeb, &ai->corr, u.list)
+               buf[aeb->pnum] = 1;
+
+       list_for_each_entry(aeb, &ai->erase, u.list)
+               buf[aeb->pnum] = 1;
+
+       list_for_each_entry(aeb, &ai->alien, u.list)
+               buf[aeb->pnum] = 1;
+
+       err = 0;
+       for (pnum = 0; pnum < ubi->peb_count; pnum++)
+               if (!buf[pnum]) {
+                       ubi_err("PEB %d is not referred", pnum);
+                       err = 1;
+               }
+
+       kfree(buf);
+       if (err)
+               goto out;
+       return 0;
+
+bad_aeb:
+       ubi_err("bad attaching information about LEB %d", aeb->lnum);
+       ubi_dump_aeb(aeb, 0);
+       ubi_dump_av(av);
+       goto out;
+
+bad_av:
+       ubi_err("bad attaching information about volume %d", av->vol_id);
+       ubi_dump_av(av);
+       goto out;
+
+bad_vid_hdr:
+       ubi_err("bad attaching information about volume %d", av->vol_id);
+       ubi_dump_av(av);
+       ubi_dump_vid_hdr(vidh);
+
+out:
+       dump_stack();
+       return -EINVAL;
+}
index 6d86c0b6bcfed229152f009701a4a228c14e51b4..7094b9c0e81bc1b2c6c2782233f8732a19726233 100644 (file)
  * module load parameters or the kernel boot parameters. If MTD devices were
  * specified, UBI does not attach any MTD device, but it is possible to do
  * later using the "UBI control device".
- *
- * At the moment we only attach UBI devices by scanning, which will become a
- * bottleneck when flashes reach certain large size. Then one may improve UBI
- * and add other methods, although it does not seem to be easy to do.
  */
 
-#ifdef UBI_LINUX
-#include <linux/err.h>
+#define __UBOOT__
+#ifndef __UBOOT__
 #include <linux/module.h>
 #include <linux/moduleparam.h>
 #include <linux/stringify.h>
+#include <linux/namei.h>
 #include <linux/stat.h>
 #include <linux/miscdevice.h>
 #include <linux/log2.h>
 #include <linux/kthread.h>
+#include <linux/kernel.h>
+#include <linux/slab.h>
+#include <linux/major.h>
+#else
+#include <linux/compat.h>
 #endif
+#include <linux/err.h>
 #include <ubi_uboot.h>
+#include <linux/mtd/partitions.h>
+
 #include "ubi.h"
 
+/* Maximum length of the 'mtd=' parameter */
+#define MTD_PARAM_LEN_MAX 64
+
+/* Maximum number of comma-separated items in the 'mtd=' parameter */
+#define MTD_PARAM_MAX_COUNT 4
+
+/* Maximum value for the number of bad PEBs per 1024 PEBs */
+#define MAX_MTD_UBI_BEB_LIMIT 768
+
+#ifdef CONFIG_MTD_UBI_MODULE
+#define ubi_is_module() 1
+#else
+#define ubi_is_module() 0
+#endif
+
 #if (CONFIG_SYS_MALLOC_LEN < (512 << 10))
 #error Malloc area too small for UBI, increase CONFIG_SYS_MALLOC_LEN to >= 512k
 #endif
 
-/* Maximum length of the 'mtd=' parameter */
-#define MTD_PARAM_LEN_MAX 64
-
 /**
  * struct mtd_dev_param - MTD device parameter description data structure.
- * @name: MTD device name or number string
+ * @name: MTD character device node path, MTD device name, or MTD device number
+ *        string
  * @vid_hdr_offs: VID header offset
+ * @max_beb_per1024: maximum expected number of bad PEBs per 1024 PEBs
  */
-struct mtd_dev_param
-{
+struct mtd_dev_param {
        char name[MTD_PARAM_LEN_MAX];
+       int ubi_num;
        int vid_hdr_offs;
+       int max_beb_per1024;
 };
 
 /* Numbers of elements set in the @mtd_dev_param array */
-static int mtd_devs = 0;
+static int __initdata mtd_devs;
 
 /* MTD devices specification parameters */
-static struct mtd_dev_param mtd_dev_param[UBI_MAX_DEVICES];
-
+static struct mtd_dev_param __initdata mtd_dev_param[UBI_MAX_DEVICES];
+#ifndef __UBOOT__
+#ifdef CONFIG_MTD_UBI_FASTMAP
+/* UBI module parameter to enable fastmap automatically on non-fastmap images */
+static bool fm_autoconvert;
+#endif
+#else
+#ifdef CONFIG_MTD_UBI_FASTMAP
+#if !defined(CONFIG_MTD_UBI_FASTMAP_AUTOCONVERT)
+#define CONFIG_MTD_UBI_FASTMAP_AUTOCONVERT 0
+#endif
+static bool fm_autoconvert = CONFIG_MTD_UBI_FASTMAP_AUTOCONVERT;
+#endif
+#endif
 /* Root UBI "class" object (corresponds to '/<sysfs>/class/ubi/') */
 struct class *ubi_class;
 
-#ifdef UBI_LINUX
 /* Slab cache for wear-leveling entries */
 struct kmem_cache *ubi_wl_entry_slab;
 
+#ifndef __UBOOT__
 /* UBI control character device */
 static struct miscdevice ubi_ctrl_cdev = {
        .minor = MISC_DYNAMIC_MINOR,
@@ -74,9 +106,13 @@ static struct miscdevice ubi_ctrl_cdev = {
 #endif
 
 /* All UBI devices in system */
+#ifndef __UBOOT__
+static struct ubi_device *ubi_devices[UBI_MAX_DEVICES];
+#else
 struct ubi_device *ubi_devices[UBI_MAX_DEVICES];
+#endif
 
-#ifdef UBI_LINUX
+#ifndef __UBOOT__
 /* Serializes UBI devices creations and removals */
 DEFINE_MUTEX(ubi_devices_mutex);
 
@@ -84,7 +120,8 @@ DEFINE_MUTEX(ubi_devices_mutex);
 static DEFINE_SPINLOCK(ubi_devices_lock);
 
 /* "Show" method for files in '/<sysfs>/class/ubi/' */
-static ssize_t ubi_version_show(struct class *class, char *buf)
+static ssize_t ubi_version_show(struct class *class,
+                               struct class_attribute *attr, char *buf)
 {
        return sprintf(buf, "%d\n", UBI_VERSION);
 }
@@ -121,6 +158,112 @@ static struct device_attribute dev_mtd_num =
        __ATTR(mtd_num, S_IRUGO, dev_attribute_show, NULL);
 #endif
 
+/**
+ * ubi_volume_notify - send a volume change notification.
+ * @ubi: UBI device description object
+ * @vol: volume description object of the changed volume
+ * @ntype: notification type to send (%UBI_VOLUME_ADDED, etc)
+ *
+ * This is a helper function which notifies all subscribers about a volume
+ * change event (creation, removal, re-sizing, re-naming, updating). Returns
+ * zero in case of success and a negative error code in case of failure.
+ */
+int ubi_volume_notify(struct ubi_device *ubi, struct ubi_volume *vol, int ntype)
+{
+       struct ubi_notification nt;
+
+       ubi_do_get_device_info(ubi, &nt.di);
+       ubi_do_get_volume_info(ubi, vol, &nt.vi);
+
+#ifdef CONFIG_MTD_UBI_FASTMAP
+       switch (ntype) {
+       case UBI_VOLUME_ADDED:
+       case UBI_VOLUME_REMOVED:
+       case UBI_VOLUME_RESIZED:
+       case UBI_VOLUME_RENAMED:
+               if (ubi_update_fastmap(ubi)) {
+                       ubi_err("Unable to update fastmap!");
+                       ubi_ro_mode(ubi);
+               }
+       }
+#endif
+       return blocking_notifier_call_chain(&ubi_notifiers, ntype, &nt);
+}
+
+/**
+ * ubi_notify_all - send a notification to all volumes.
+ * @ubi: UBI device description object
+ * @ntype: notification type to send (%UBI_VOLUME_ADDED, etc)
+ * @nb: the notifier to call
+ *
+ * This function walks all volumes of UBI device @ubi and sends the @ntype
+ * notification for each volume. If @nb is %NULL, then all registered notifiers
+ * are called, otherwise only the @nb notifier is called. Returns the number of
+ * sent notifications.
+ */
+int ubi_notify_all(struct ubi_device *ubi, int ntype, struct notifier_block *nb)
+{
+       struct ubi_notification nt;
+       int i, count = 0;
+#ifndef __UBOOT__
+       int ret;
+#endif
+
+       ubi_do_get_device_info(ubi, &nt.di);
+
+       mutex_lock(&ubi->device_mutex);
+       for (i = 0; i < ubi->vtbl_slots; i++) {
+               /*
+                * Since the @ubi->device is locked, and we are not going to
+                * change @ubi->volumes, we do not have to lock
+                * @ubi->volumes_lock.
+                */
+               if (!ubi->volumes[i])
+                       continue;
+
+               ubi_do_get_volume_info(ubi, ubi->volumes[i], &nt.vi);
+#ifndef __UBOOT__
+               if (nb)
+                       nb->notifier_call(nb, ntype, &nt);
+               else
+                       ret = blocking_notifier_call_chain(&ubi_notifiers, ntype,
+                                                    &nt);
+#endif
+               count += 1;
+       }
+       mutex_unlock(&ubi->device_mutex);
+
+       return count;
+}
+
+/**
+ * ubi_enumerate_volumes - send "add" notification for all existing volumes.
+ * @nb: the notifier to call
+ *
+ * This function walks all UBI devices and volumes and sends the
+ * %UBI_VOLUME_ADDED notification for each volume. If @nb is %NULL, then all
+ * registered notifiers are called, otherwise only the @nb notifier is called.
+ * Returns the number of sent notifications.
+ */
+int ubi_enumerate_volumes(struct notifier_block *nb)
+{
+       int i, count = 0;
+
+       /*
+        * Since the @ubi_devices_mutex is locked, and we are not going to
+        * change @ubi_devices, we do not have to lock @ubi_devices_lock.
+        */
+       for (i = 0; i < UBI_MAX_DEVICES; i++) {
+               struct ubi_device *ubi = ubi_devices[i];
+
+               if (!ubi)
+                       continue;
+               count += ubi_notify_all(ubi, UBI_VOLUME_ADDED, nb);
+       }
+
+       return count;
+}
+
 /**
  * ubi_get_device - get UBI device.
  * @ubi_num: UBI device number
@@ -159,8 +302,7 @@ void ubi_put_device(struct ubi_device *ubi)
 }
 
 /**
- * ubi_get_by_major - get UBI device description object by character device
- *                    major number.
+ * ubi_get_by_major - get UBI device by character device major number.
  * @major: major number
  *
  * This function is similar to 'ubi_get_device()', but it searches the device
@@ -213,7 +355,7 @@ int ubi_major2num(int major)
        return ubi_num;
 }
 
-#ifdef UBI_LINUX
+#ifndef __UBOOT__
 /* "Show" method for files in '/<sysfs>/class/ubi/ubiX/' */
 static ssize_t dev_attribute_show(struct device *dev,
                                  struct device_attribute *attr, char *buf)
@@ -265,28 +407,35 @@ static ssize_t dev_attribute_show(struct device *dev,
        return ret;
 }
 
-/* Fake "release" method for UBI devices */
-static void dev_release(struct device *dev) { }
+static void dev_release(struct device *dev)
+{
+       struct ubi_device *ubi = container_of(dev, struct ubi_device, dev);
+
+       kfree(ubi);
+}
 
 /**
  * ubi_sysfs_init - initialize sysfs for an UBI device.
  * @ubi: UBI device description object
+ * @ref: set to %1 on exit in case of failure if a reference to @ubi->dev was
+ *       taken
  *
  * This function returns zero in case of success and a negative error code in
  * case of failure.
  */
-static int ubi_sysfs_init(struct ubi_device *ubi)
+static int ubi_sysfs_init(struct ubi_device *ubi, int *ref)
 {
        int err;
 
        ubi->dev.release = dev_release;
        ubi->dev.devt = ubi->cdev.dev;
        ubi->dev.class = ubi_class;
-       sprintf(&ubi->dev.bus_id[0], UBI_NAME_STR"%d", ubi->ubi_num);
+       dev_set_name(&ubi->dev, UBI_NAME_STR"%d", ubi->ubi_num);
        err = device_register(&ubi->dev);
        if (err)
                return err;
 
+       *ref = 1;
        err = device_create_file(&ubi->dev, &dev_eraseblock_size);
        if (err)
                return err;
@@ -343,7 +492,7 @@ static void ubi_sysfs_close(struct ubi_device *ubi)
 #endif
 
 /**
- * kill_volumes - destroy all volumes.
+ * kill_volumes - destroy all user volumes.
  * @ubi: UBI device description object
  */
 static void kill_volumes(struct ubi_device *ubi)
@@ -358,17 +507,29 @@ static void kill_volumes(struct ubi_device *ubi)
 /**
  * uif_init - initialize user interfaces for an UBI device.
  * @ubi: UBI device description object
+ * @ref: set to %1 on exit in case of failure if a reference to @ubi->dev was
+ *       taken, otherwise set to %0
+ *
+ * This function initializes various user interfaces for an UBI device. If the
+ * initialization fails at an early stage, this function frees all the
+ * resources it allocated, returns an error, and @ref is set to %0. However,
+ * if the initialization fails after the UBI device was registered in the
+ * driver core subsystem, this function takes a reference to @ubi->dev, because
+ * otherwise the release function ('dev_release()') would free whole @ubi
+ * object. The @ref argument is set to %1 in this case. The caller has to put
+ * this reference.
  *
  * This function returns zero in case of success and a negative error code in
  * case of failure.
  */
-static int uif_init(struct ubi_device *ubi)
+static int uif_init(struct ubi_device *ubi, int *ref)
 {
        int i, err;
-#ifdef UBI_LINUX
+#ifndef __UBOOT__
        dev_t dev;
 #endif
 
+       *ref = 0;
        sprintf(ubi->ubi_name, UBI_NAME_STR "%d", ubi->ubi_num);
 
        /*
@@ -387,7 +548,7 @@ static int uif_init(struct ubi_device *ubi)
 
        ubi_assert(MINOR(dev) == 0);
        cdev_init(&ubi->cdev, &ubi_cdev_operations);
-       dbg_msg("%s major is %u", ubi->ubi_name, MAJOR(dev));
+       dbg_gen("%s major is %u", ubi->ubi_name, MAJOR(dev));
        ubi->cdev.owner = THIS_MODULE;
 
        err = cdev_add(&ubi->cdev, dev, 1);
@@ -396,7 +557,7 @@ static int uif_init(struct ubi_device *ubi)
                goto out_unreg;
        }
 
-       err = ubi_sysfs_init(ubi);
+       err = ubi_sysfs_init(ubi, ref);
        if (err)
                goto out_sysfs;
 
@@ -414,6 +575,8 @@ static int uif_init(struct ubi_device *ubi)
 out_volumes:
        kill_volumes(ubi);
 out_sysfs:
+       if (*ref)
+               get_device(&ubi->dev);
        ubi_sysfs_close(ubi);
        cdev_del(&ubi->cdev);
 out_unreg:
@@ -425,6 +588,10 @@ out_unreg:
 /**
  * uif_close - close user interfaces for an UBI device.
  * @ubi: UBI device description object
+ *
+ * Note, since this function un-registers UBI volume device objects (@vol->dev),
+ * the memory allocated voe the volumes is freed as well (in the release
+ * function).
  */
 static void uif_close(struct ubi_device *ubi)
 {
@@ -435,58 +602,52 @@ static void uif_close(struct ubi_device *ubi)
 }
 
 /**
- * attach_by_scanning - attach an MTD device using scanning method.
- * @ubi: UBI device descriptor
- *
- * This function returns zero in case of success and a negative error code in
- * case of failure.
- *
- * Note, currently this is the only method to attach UBI devices. Hopefully in
- * the future we'll have more scalable attaching methods and avoid full media
- * scanning. But even in this case scanning will be needed as a fall-back
- * attaching method if there are some on-flash table corruptions.
+ * ubi_free_internal_volumes - free internal volumes.
+ * @ubi: UBI device description object
  */
-static int attach_by_scanning(struct ubi_device *ubi)
+void ubi_free_internal_volumes(struct ubi_device *ubi)
 {
-       int err;
-       struct ubi_scan_info *si;
+       int i;
 
-       si = ubi_scan(ubi);
-       if (IS_ERR(si))
-               return PTR_ERR(si);
+       for (i = ubi->vtbl_slots;
+            i < ubi->vtbl_slots + UBI_INT_VOL_COUNT; i++) {
+               kfree(ubi->volumes[i]->eba_tbl);
+               kfree(ubi->volumes[i]);
+       }
+}
 
-       ubi->bad_peb_count = si->bad_peb_count;
-       ubi->good_peb_count = ubi->peb_count - ubi->bad_peb_count;
-       ubi->max_ec = si->max_ec;
-       ubi->mean_ec = si->mean_ec;
+static int get_bad_peb_limit(const struct ubi_device *ubi, int max_beb_per1024)
+{
+       int limit, device_pebs;
+       uint64_t device_size;
 
-       err = ubi_read_volume_table(ubi, si);
-       if (err)
-               goto out_si;
+       if (!max_beb_per1024)
+               return 0;
 
-       err = ubi_eba_init_scan(ubi, si);
-       if (err)
-               goto out_vtbl;
+       /*
+        * Here we are using size of the entire flash chip and
+        * not just the MTD partition size because the maximum
+        * number of bad eraseblocks is a percentage of the
+        * whole device and bad eraseblocks are not fairly
+        * distributed over the flash chip. So the worst case
+        * is that all the bad eraseblocks of the chip are in
+        * the MTD partition we are attaching (ubi->mtd).
+        */
+       device_size = mtd_get_device_size(ubi->mtd);
+       device_pebs = mtd_div_by_eb(device_size, ubi->mtd);
+       limit = mult_frac(device_pebs, max_beb_per1024, 1024);
 
-       err = ubi_wl_init_scan(ubi, si);
-       if (err)
-               goto out_eba;
+       /* Round it up */
+       if (mult_frac(limit, 1024, max_beb_per1024) < device_pebs)
+               limit += 1;
 
-       ubi_scan_destroy_si(si);
-       return 0;
-
-out_eba:
-       ubi_eba_close(ubi);
-out_vtbl:
-       vfree(ubi->vtbl);
-out_si:
-       ubi_scan_destroy_si(si);
-       return err;
+       return limit;
 }
 
 /**
- * io_init - initialize I/O unit for a given UBI device.
+ * io_init - initialize I/O sub-system for a given UBI device.
  * @ubi: UBI device description object
+ * @max_beb_per1024: maximum expected number of bad PEB per 1024 PEBs
  *
  * If @ubi->vid_hdr_offset or @ubi->leb_start is zero, default offsets are
  * assumed:
@@ -499,8 +660,11 @@ out_si:
  * This function returns zero in case of success and a negative error code in
  * case of failure.
  */
-static int io_init(struct ubi_device *ubi)
+static int io_init(struct ubi_device *ubi, int max_beb_per1024)
 {
+       dbg_gen("sizeof(struct ubi_ainf_peb) %zu", sizeof(struct ubi_ainf_peb));
+       dbg_gen("sizeof(struct ubi_wl_entry) %zu", sizeof(struct ubi_wl_entry));
+
        if (ubi->mtd->numeraseregions != 0) {
                /*
                 * Some flashes have several erase regions. Different regions
@@ -527,8 +691,15 @@ static int io_init(struct ubi_device *ubi)
        ubi->peb_count  = mtd_div_by_eb(ubi->mtd->size, ubi->mtd);
        ubi->flash_size = ubi->mtd->size;
 
-       if (mtd_can_have_bb(ubi->mtd))
+       if (mtd_can_have_bb(ubi->mtd)) {
                ubi->bad_allowed = 1;
+               ubi->bad_peb_limit = get_bad_peb_limit(ubi, max_beb_per1024);
+       }
+
+       if (ubi->mtd->type == MTD_NORFLASH) {
+               ubi_assert(ubi->mtd->writesize == 1);
+               ubi->nor_flash = 1;
+       }
 
        ubi->min_io_size = ubi->mtd->writesize;
        ubi->hdrs_min_io_size = ubi->mtd->writesize >> ubi->mtd->subpage_sft;
@@ -548,14 +719,28 @@ static int io_init(struct ubi_device *ubi)
        ubi_assert(ubi->hdrs_min_io_size <= ubi->min_io_size);
        ubi_assert(ubi->min_io_size % ubi->hdrs_min_io_size == 0);
 
+       ubi->max_write_size = ubi->mtd->writebufsize;
+       /*
+        * Maximum write size has to be greater or equivalent to min. I/O
+        * size, and be multiple of min. I/O size.
+        */
+       if (ubi->max_write_size < ubi->min_io_size ||
+           ubi->max_write_size % ubi->min_io_size ||
+           !is_power_of_2(ubi->max_write_size)) {
+               ubi_err("bad write buffer size %d for %d min. I/O unit",
+                       ubi->max_write_size, ubi->min_io_size);
+               return -EINVAL;
+       }
+
        /* Calculate default aligned sizes of EC and VID headers */
        ubi->ec_hdr_alsize = ALIGN(UBI_EC_HDR_SIZE, ubi->hdrs_min_io_size);
        ubi->vid_hdr_alsize = ALIGN(UBI_VID_HDR_SIZE, ubi->hdrs_min_io_size);
 
-       dbg_msg("min_io_size      %d", ubi->min_io_size);
-       dbg_msg("hdrs_min_io_size %d", ubi->hdrs_min_io_size);
-       dbg_msg("ec_hdr_alsize    %d", ubi->ec_hdr_alsize);
-       dbg_msg("vid_hdr_alsize   %d", ubi->vid_hdr_alsize);
+       dbg_gen("min_io_size      %d", ubi->min_io_size);
+       dbg_gen("max_write_size   %d", ubi->max_write_size);
+       dbg_gen("hdrs_min_io_size %d", ubi->hdrs_min_io_size);
+       dbg_gen("ec_hdr_alsize    %d", ubi->ec_hdr_alsize);
+       dbg_gen("vid_hdr_alsize   %d", ubi->vid_hdr_alsize);
 
        if (ubi->vid_hdr_offset == 0)
                /* Default offset */
@@ -569,13 +754,13 @@ static int io_init(struct ubi_device *ubi)
        }
 
        /* Similar for the data offset */
-       ubi->leb_start = ubi->vid_hdr_offset + UBI_EC_HDR_SIZE;
+       ubi->leb_start = ubi->vid_hdr_offset + UBI_VID_HDR_SIZE;
        ubi->leb_start = ALIGN(ubi->leb_start, ubi->min_io_size);
 
-       dbg_msg("vid_hdr_offset   %d", ubi->vid_hdr_offset);
-       dbg_msg("vid_hdr_aloffset %d", ubi->vid_hdr_aloffset);
-       dbg_msg("vid_hdr_shift    %d", ubi->vid_hdr_shift);
-       dbg_msg("leb_start        %d", ubi->leb_start);
+       dbg_gen("vid_hdr_offset   %d", ubi->vid_hdr_offset);
+       dbg_gen("vid_hdr_aloffset %d", ubi->vid_hdr_aloffset);
+       dbg_gen("vid_hdr_shift    %d", ubi->vid_hdr_shift);
+       dbg_gen("leb_start        %d", ubi->leb_start);
 
        /* The shift must be aligned to 32-bit boundary */
        if (ubi->vid_hdr_shift % 4) {
@@ -594,42 +779,39 @@ static int io_init(struct ubi_device *ubi)
                return -EINVAL;
        }
 
+       /*
+        * Set maximum amount of physical erroneous eraseblocks to be 10%.
+        * Erroneous PEB are those which have read errors.
+        */
+       ubi->max_erroneous = ubi->peb_count / 10;
+       if (ubi->max_erroneous < 16)
+               ubi->max_erroneous = 16;
+       dbg_gen("max_erroneous    %d", ubi->max_erroneous);
+
        /*
         * It may happen that EC and VID headers are situated in one minimal
         * I/O unit. In this case we can only accept this UBI image in
         * read-only mode.
         */
        if (ubi->vid_hdr_offset + UBI_VID_HDR_SIZE <= ubi->hdrs_min_io_size) {
-               ubi_warn("EC and VID headers are in the same minimal I/O unit, "
-                        "switch to read-only mode");
+               ubi_warn("EC and VID headers are in the same minimal I/O unit, switch to read-only mode");
                ubi->ro_mode = 1;
        }
 
        ubi->leb_size = ubi->peb_size - ubi->leb_start;
 
        if (!(ubi->mtd->flags & MTD_WRITEABLE)) {
-               ubi_msg("MTD device %d is write-protected, attach in "
-                       "read-only mode", ubi->mtd->index);
+               ubi_msg("MTD device %d is write-protected, attach in read-only mode",
+                       ubi->mtd->index);
                ubi->ro_mode = 1;
        }
 
-       ubi_msg("physical eraseblock size:   %d bytes (%d KiB)",
-               ubi->peb_size, ubi->peb_size >> 10);
-       ubi_msg("logical eraseblock size:    %d bytes", ubi->leb_size);
-       ubi_msg("smallest flash I/O unit:    %d", ubi->min_io_size);
-       if (ubi->hdrs_min_io_size != ubi->min_io_size)
-               ubi_msg("sub-page size:              %d",
-                       ubi->hdrs_min_io_size);
-       ubi_msg("VID header offset:          %d (aligned %d)",
-               ubi->vid_hdr_offset, ubi->vid_hdr_aloffset);
-       ubi_msg("data offset:                %d", ubi->leb_start);
-
        /*
-        * Note, ideally, we have to initialize ubi->bad_peb_count here. But
+        * Note, ideally, we have to initialize @ubi->bad_peb_count here. But
         * unfortunately, MTD does not provide this information. We should loop
         * over all physical eraseblocks and invoke mtd->block_is_bad() for
-        * each physical eraseblock. So, we skip ubi->bad_peb_count
-        * uninitialized and initialize it after scanning.
+        * each physical eraseblock. So, we leave @ubi->bad_peb_count
+        * uninitialized so far.
         */
 
        return 0;
@@ -640,7 +822,7 @@ static int io_init(struct ubi_device *ubi)
  * @ubi: UBI device description object
  * @vol_id: ID of the volume to re-size
  *
- * This function re-sizes the volume marked by the @UBI_VTBL_AUTORESIZE_FLG in
+ * This function re-sizes the volume marked by the %UBI_VTBL_AUTORESIZE_FLG in
  * the volume table to the largest possible size. See comments in ubi-header.h
  * for more description of the flag. Returns zero in case of success and a
  * negative error code in case of failure.
@@ -651,9 +833,14 @@ static int autoresize(struct ubi_device *ubi, int vol_id)
        struct ubi_volume *vol = ubi->volumes[vol_id];
        int err, old_reserved_pebs = vol->reserved_pebs;
 
+       if (ubi->ro_mode) {
+               ubi_warn("skip auto-resize because of R/O mode");
+               return 0;
+       }
+
        /*
         * Clear the auto-resize flag in the volume in-memory copy of the
-        * volume table, and 'ubi_resize_volume()' will propogate this change
+        * volume table, and 'ubi_resize_volume()' will propagate this change
         * to the flash.
         */
        ubi->vtbl[vol_id].flags &= ~UBI_VTBL_AUTORESIZE_FLG;
@@ -662,11 +849,10 @@ static int autoresize(struct ubi_device *ubi, int vol_id)
                struct ubi_vtbl_record vtbl_rec;
 
                /*
-                * No avalilable PEBs to re-size the volume, clear the flag on
+                * No available PEBs to re-size the volume, clear the flag on
                 * flash and exit.
                 */
-               memcpy(&vtbl_rec, &ubi->vtbl[vol_id],
-                      sizeof(struct ubi_vtbl_record));
+               vtbl_rec = ubi->vtbl[vol_id];
                err = ubi_change_vtbl_record(ubi, vol_id, &vtbl_rec);
                if (err)
                        ubi_err("cannot clean auto-resize flag for volume %d",
@@ -689,23 +875,31 @@ static int autoresize(struct ubi_device *ubi, int vol_id)
 
 /**
  * ubi_attach_mtd_dev - attach an MTD device.
- * @mtd_dev: MTD device description object
+ * @mtd: MTD device description object
  * @ubi_num: number to assign to the new UBI device
  * @vid_hdr_offset: VID header offset
+ * @max_beb_per1024: maximum expected number of bad PEB per 1024 PEBs
  *
  * This function attaches MTD device @mtd_dev to UBI and assign @ubi_num number
  * to the newly created UBI device, unless @ubi_num is %UBI_DEV_NUM_AUTO, in
- * which case this function finds a vacant device nubert and assings it
+ * which case this function finds a vacant device number and assigns it
  * automatically. Returns the new UBI device number in case of success and a
  * negative error code in case of failure.
  *
  * Note, the invocations of this function has to be serialized by the
  * @ubi_devices_mutex.
  */
-int ubi_attach_mtd_dev(struct mtd_info *mtd, int ubi_num, int vid_hdr_offset)
+int ubi_attach_mtd_dev(struct mtd_info *mtd, int ubi_num,
+                      int vid_hdr_offset, int max_beb_per1024)
 {
        struct ubi_device *ubi;
-       int i, err;
+       int i, err, ref = 0;
+
+       if (max_beb_per1024 < 0 || max_beb_per1024 > MAX_MTD_UBI_BEB_LIMIT)
+               return -EINVAL;
+
+       if (!max_beb_per1024)
+               max_beb_per1024 = CONFIG_MTD_UBI_BEB_LIMIT;
 
        /*
         * Check if we already have the same MTD device attached.
@@ -716,7 +910,7 @@ int ubi_attach_mtd_dev(struct mtd_info *mtd, int ubi_num, int vid_hdr_offset)
        for (i = 0; i < UBI_MAX_DEVICES; i++) {
                ubi = ubi_devices[i];
                if (ubi && mtd->index == ubi->mtd->index) {
-                       dbg_err("mtd%d is already attached to ubi%d",
+                       ubi_err("mtd%d is already attached to ubi%d",
                                mtd->index, i);
                        return -EEXIST;
                }
@@ -731,8 +925,8 @@ int ubi_attach_mtd_dev(struct mtd_info *mtd, int ubi_num, int vid_hdr_offset)
         * no sense to attach emulated MTD devices, so we prohibit this.
         */
        if (mtd->type == MTD_UBIVOLUME) {
-               ubi_err("refuse attaching mtd%d - it is already emulated on "
-                       "top of UBI", mtd->index);
+               ubi_err("refuse attaching mtd%d - it is already emulated on top of UBI",
+                       mtd->index);
                return -EINVAL;
        }
 
@@ -742,7 +936,8 @@ int ubi_attach_mtd_dev(struct mtd_info *mtd, int ubi_num, int vid_hdr_offset)
                        if (!ubi_devices[ubi_num])
                                break;
                if (ubi_num == UBI_MAX_DEVICES) {
-                       dbg_err("only %d UBI devices may be created", UBI_MAX_DEVICES);
+                       ubi_err("only %d UBI devices may be created",
+                               UBI_MAX_DEVICES);
                        return -ENFILE;
                }
        } else {
@@ -751,7 +946,7 @@ int ubi_attach_mtd_dev(struct mtd_info *mtd, int ubi_num, int vid_hdr_offset)
 
                /* Make sure ubi_num is not busy */
                if (ubi_devices[ubi_num]) {
-                       dbg_err("ubi%d already exists", ubi_num);
+                       ubi_err("ubi%d already exists", ubi_num);
                        return -EEXIST;
                }
        }
@@ -765,36 +960,61 @@ int ubi_attach_mtd_dev(struct mtd_info *mtd, int ubi_num, int vid_hdr_offset)
        ubi->vid_hdr_offset = vid_hdr_offset;
        ubi->autoresize_vol_id = -1;
 
+#ifdef CONFIG_MTD_UBI_FASTMAP
+       ubi->fm_pool.used = ubi->fm_pool.size = 0;
+       ubi->fm_wl_pool.used = ubi->fm_wl_pool.size = 0;
+
+       /*
+        * fm_pool.max_size is 5% of the total number of PEBs but it's also
+        * between UBI_FM_MAX_POOL_SIZE and UBI_FM_MIN_POOL_SIZE.
+        */
+       ubi->fm_pool.max_size = min(((int)mtd_div_by_eb(ubi->mtd->size,
+               ubi->mtd) / 100) * 5, UBI_FM_MAX_POOL_SIZE);
+       if (ubi->fm_pool.max_size < UBI_FM_MIN_POOL_SIZE)
+               ubi->fm_pool.max_size = UBI_FM_MIN_POOL_SIZE;
+
+       ubi->fm_wl_pool.max_size = UBI_FM_WL_POOL_SIZE;
+       ubi->fm_disabled = !fm_autoconvert;
+
+       if (!ubi->fm_disabled && (int)mtd_div_by_eb(ubi->mtd->size, ubi->mtd)
+           <= UBI_FM_MAX_START) {
+               ubi_err("More than %i PEBs are needed for fastmap, sorry.",
+                       UBI_FM_MAX_START);
+               ubi->fm_disabled = 1;
+       }
+
+       ubi_msg("default fastmap pool size: %d", ubi->fm_pool.max_size);
+       ubi_msg("default fastmap WL pool size: %d", ubi->fm_wl_pool.max_size);
+#else
+       ubi->fm_disabled = 1;
+#endif
        mutex_init(&ubi->buf_mutex);
        mutex_init(&ubi->ckvol_mutex);
-       mutex_init(&ubi->volumes_mutex);
+       mutex_init(&ubi->device_mutex);
        spin_lock_init(&ubi->volumes_lock);
+       mutex_init(&ubi->fm_mutex);
+       init_rwsem(&ubi->fm_sem);
 
        ubi_msg("attaching mtd%d to ubi%d", mtd->index, ubi_num);
 
-       err = io_init(ubi);
+       err = io_init(ubi, max_beb_per1024);
        if (err)
                goto out_free;
 
        err = -ENOMEM;
-       ubi->peb_buf1 = vmalloc(ubi->peb_size);
-       if (!ubi->peb_buf1)
+       ubi->peb_buf = vmalloc(ubi->peb_size);
+       if (!ubi->peb_buf)
                goto out_free;
 
-       ubi->peb_buf2 = vmalloc(ubi->peb_size);
-       if (!ubi->peb_buf2)
-               goto out_free;
-
-#ifdef CONFIG_MTD_UBI_DEBUG
-       mutex_init(&ubi->dbg_buf_mutex);
-       ubi->dbg_peb_buf = vmalloc(ubi->peb_size);
-       if (!ubi->dbg_peb_buf)
+#ifdef CONFIG_MTD_UBI_FASTMAP
+       ubi->fm_size = ubi_calc_fm_size(ubi);
+       ubi->fm_buf = vzalloc(ubi->fm_size);
+       if (!ubi->fm_buf)
                goto out_free;
 #endif
-
-       err = attach_by_scanning(ubi);
+       err = ubi_attach(ubi, 0);
        if (err) {
-               dbg_err("failed to attach by scanning, error %d", err);
+               ubi_err("failed to attach mtd%d, error %d", mtd->index, err);
                goto out_free;
        }
 
@@ -804,56 +1024,71 @@ int ubi_attach_mtd_dev(struct mtd_info *mtd, int ubi_num, int vid_hdr_offset)
                        goto out_detach;
        }
 
-       err = uif_init(ubi);
+       err = uif_init(ubi, &ref);
        if (err)
                goto out_detach;
 
-       ubi->bgt_thread = kthread_create(ubi_thread, ubi, ubi->bgt_name);
+       err = ubi_debugfs_init_dev(ubi);
+       if (err)
+               goto out_uif;
+
+       ubi->bgt_thread = kthread_create(ubi_thread, ubi, "%s", ubi->bgt_name);
        if (IS_ERR(ubi->bgt_thread)) {
                err = PTR_ERR(ubi->bgt_thread);
                ubi_err("cannot spawn \"%s\", error %d", ubi->bgt_name,
                        err);
-               goto out_uif;
+               goto out_debugfs;
        }
 
-       ubi_msg("attached mtd%d to ubi%d", mtd->index, ubi_num);
-       ubi_msg("MTD device name:            \"%s\"", mtd->name);
-       ubi_msg("MTD device size:            %llu MiB", ubi->flash_size >> 20);
-       ubi_msg("number of good PEBs:        %d", ubi->good_peb_count);
-       ubi_msg("number of bad PEBs:         %d", ubi->bad_peb_count);
-       ubi_msg("max. allowed volumes:       %d", ubi->vtbl_slots);
-       ubi_msg("wear-leveling threshold:    %d", CONFIG_MTD_UBI_WL_THRESHOLD);
-       ubi_msg("number of internal volumes: %d", UBI_INT_VOL_COUNT);
-       ubi_msg("number of user volumes:     %d",
-               ubi->vol_count - UBI_INT_VOL_COUNT);
-       ubi_msg("available PEBs:             %d", ubi->avail_pebs);
-       ubi_msg("total number of reserved PEBs: %d", ubi->rsvd_pebs);
-       ubi_msg("number of PEBs reserved for bad PEB handling: %d",
-               ubi->beb_rsvd_pebs);
-       ubi_msg("max/mean erase counter: %d/%d", ubi->max_ec, ubi->mean_ec);
-
-       /* Enable the background thread */
-       if (!DBG_DISABLE_BGT) {
-               ubi->thread_enabled = 1;
-               wake_up_process(ubi->bgt_thread);
-       }
+       ubi_msg("attached mtd%d (name \"%s\", size %llu MiB) to ubi%d",
+               mtd->index, mtd->name, ubi->flash_size >> 20, ubi_num);
+       ubi_msg("PEB size: %d bytes (%d KiB), LEB size: %d bytes",
+               ubi->peb_size, ubi->peb_size >> 10, ubi->leb_size);
+       ubi_msg("min./max. I/O unit sizes: %d/%d, sub-page size %d",
+               ubi->min_io_size, ubi->max_write_size, ubi->hdrs_min_io_size);
+       ubi_msg("VID header offset: %d (aligned %d), data offset: %d",
+               ubi->vid_hdr_offset, ubi->vid_hdr_aloffset, ubi->leb_start);
+       ubi_msg("good PEBs: %d, bad PEBs: %d, corrupted PEBs: %d",
+               ubi->good_peb_count, ubi->bad_peb_count, ubi->corr_peb_count);
+       ubi_msg("user volume: %d, internal volumes: %d, max. volumes count: %d",
+               ubi->vol_count - UBI_INT_VOL_COUNT, UBI_INT_VOL_COUNT,
+               ubi->vtbl_slots);
+       ubi_msg("max/mean erase counter: %d/%d, WL threshold: %d, image sequence number: %u",
+               ubi->max_ec, ubi->mean_ec, CONFIG_MTD_UBI_WL_THRESHOLD,
+               ubi->image_seq);
+       ubi_msg("available PEBs: %d, total reserved PEBs: %d, PEBs reserved for bad PEB handling: %d",
+               ubi->avail_pebs, ubi->rsvd_pebs, ubi->beb_rsvd_pebs);
+
+       /*
+        * The below lock makes sure we do not race with 'ubi_thread()' which
+        * checks @ubi->thread_enabled. Otherwise we may fail to wake it up.
+        */
+       spin_lock(&ubi->wl_lock);
+       ubi->thread_enabled = 1;
+       wake_up_process(ubi->bgt_thread);
+       spin_unlock(&ubi->wl_lock);
 
        ubi_devices[ubi_num] = ubi;
+       ubi_notify_all(ubi, UBI_VOLUME_ADDED, NULL);
        return ubi_num;
 
+out_debugfs:
+       ubi_debugfs_exit_dev(ubi);
 out_uif:
+       get_device(&ubi->dev);
+       ubi_assert(ref);
        uif_close(ubi);
 out_detach:
-       ubi_eba_close(ubi);
        ubi_wl_close(ubi);
+       ubi_free_internal_volumes(ubi);
        vfree(ubi->vtbl);
 out_free:
-       vfree(ubi->peb_buf1);
-       vfree(ubi->peb_buf2);
-#ifdef CONFIG_MTD_UBI_DEBUG
-       vfree(ubi->dbg_peb_buf);
-#endif
-       kfree(ubi);
+       vfree(ubi->peb_buf);
+       vfree(ubi->fm_buf);
+       if (ref)
+               put_device(&ubi->dev);
+       else
+               kfree(ubi);
        return err;
 }
 
@@ -877,13 +1112,13 @@ int ubi_detach_mtd_dev(int ubi_num, int anyway)
        if (ubi_num < 0 || ubi_num >= UBI_MAX_DEVICES)
                return -EINVAL;
 
-       spin_lock(&ubi_devices_lock);
-       ubi = ubi_devices[ubi_num];
-       if (!ubi) {
-               spin_unlock(&ubi_devices_lock);
+       ubi = ubi_get_device(ubi_num);
+       if (!ubi)
                return -EINVAL;
-       }
 
+       spin_lock(&ubi_devices_lock);
+       put_device(&ubi->dev);
+       ubi->ref_count -= 1;
        if (ubi->ref_count) {
                if (!anyway) {
                        spin_unlock(&ubi_devices_lock);
@@ -897,8 +1132,13 @@ int ubi_detach_mtd_dev(int ubi_num, int anyway)
        spin_unlock(&ubi_devices_lock);
 
        ubi_assert(ubi_num == ubi->ubi_num);
-       dbg_msg("detaching mtd%d from ubi%d", ubi->mtd->index, ubi_num);
-
+       ubi_notify_all(ubi, UBI_VOLUME_REMOVED, NULL);
+       ubi_msg("detaching mtd%d from ubi%d", ubi->mtd->index, ubi_num);
+#ifdef CONFIG_MTD_UBI_FASTMAP
+       /* If we don't write a new fastmap at detach time we lose all
+        * EC updates that have been made since the last written fastmap. */
+       ubi_update_fastmap(ubi);
+#endif
        /*
         * Before freeing anything, we have to stop the background thread to
         * prevent it from doing anything on this device while we are freeing.
@@ -906,29 +1146,73 @@ int ubi_detach_mtd_dev(int ubi_num, int anyway)
        if (ubi->bgt_thread)
                kthread_stop(ubi->bgt_thread);
 
+       /*
+        * Get a reference to the device in order to prevent 'dev_release()'
+        * from freeing the @ubi object.
+        */
+       get_device(&ubi->dev);
+
+       ubi_debugfs_exit_dev(ubi);
        uif_close(ubi);
-       ubi_eba_close(ubi);
+
        ubi_wl_close(ubi);
+       ubi_free_internal_volumes(ubi);
        vfree(ubi->vtbl);
        put_mtd_device(ubi->mtd);
-       vfree(ubi->peb_buf1);
-       vfree(ubi->peb_buf2);
-#ifdef CONFIG_MTD_UBI_DEBUG
-       vfree(ubi->dbg_peb_buf);
-#endif
+       vfree(ubi->peb_buf);
+       vfree(ubi->fm_buf);
        ubi_msg("mtd%d is detached from ubi%d", ubi->mtd->index, ubi->ubi_num);
-       kfree(ubi);
+       put_device(&ubi->dev);
        return 0;
 }
 
+#ifndef __UBOOT__
 /**
- * find_mtd_device - open an MTD device by its name or number.
- * @mtd_dev: name or number of the device
+ * open_mtd_by_chdev - open an MTD device by its character device node path.
+ * @mtd_dev: MTD character device node path
+ *
+ * This helper function opens an MTD device by its character node device path.
+ * Returns MTD device description object in case of success and a negative
+ * error code in case of failure.
+ */
+static struct mtd_info * __init open_mtd_by_chdev(const char *mtd_dev)
+{
+       int err, major, minor, mode;
+       struct path path;
+
+       /* Probably this is an MTD character device node path */
+       err = kern_path(mtd_dev, LOOKUP_FOLLOW, &path);
+       if (err)
+               return ERR_PTR(err);
+
+       /* MTD device number is defined by the major / minor numbers */
+       major = imajor(path.dentry->d_inode);
+       minor = iminor(path.dentry->d_inode);
+       mode = path.dentry->d_inode->i_mode;
+       path_put(&path);
+       if (major != MTD_CHAR_MAJOR || !S_ISCHR(mode))
+               return ERR_PTR(-EINVAL);
+
+       if (minor & 1)
+               /*
+                * Just do not think the "/dev/mtdrX" devices support is need,
+                * so do not support them to avoid doing extra work.
+                */
+               return ERR_PTR(-EINVAL);
+
+       return get_mtd_device(NULL, minor / 2);
+}
+#endif
+
+/**
+ * open_mtd_device - open MTD device by name, character device path, or number.
+ * @mtd_dev: name, character device node path, or MTD device device number
  *
  * This function tries to open and MTD device described by @mtd_dev string,
- * which is first treated as an ASCII number, and if it is not true, it is
- * treated as MTD device name. Returns MTD device description object in case of
- * success and a negative error code in case of failure.
+ * which is first treated as ASCII MTD device number, and if it is not true, it
+ * is treated as MTD device name, and if that is also not true, it is treated
+ * as MTD character device node path. Returns MTD device description object in
+ * case of success and a negative error code in case of failure.
  */
 static struct mtd_info * __init open_mtd_device(const char *mtd_dev)
 {
@@ -943,13 +1227,22 @@ static struct mtd_info * __init open_mtd_device(const char *mtd_dev)
                 * MTD device name.
                 */
                mtd = get_mtd_device_nm(mtd_dev);
+#ifndef __UBOOT__
+               if (IS_ERR(mtd) && PTR_ERR(mtd) == -ENODEV)
+                       /* Probably this is an MTD character device node path */
+                       mtd = open_mtd_by_chdev(mtd_dev);
+#endif
        } else
                mtd = get_mtd_device(NULL, mtd_num);
 
        return mtd;
 }
 
-int __init ubi_init(void)
+#ifndef __UBOOT__
+static int __init ubi_init(void)
+#else
+int ubi_init(void)
+#endif
 {
        int err, i, k;
 
@@ -982,13 +1275,18 @@ int __init ubi_init(void)
                goto out_version;
        }
 
-#ifdef UBI_LINUX
        ubi_wl_entry_slab = kmem_cache_create("ubi_wl_entry_slab",
                                              sizeof(struct ubi_wl_entry),
                                              0, 0, NULL);
-       if (!ubi_wl_entry_slab)
+       if (!ubi_wl_entry_slab) {
+               err = -ENOMEM;
                goto out_dev_unreg;
-#endif
+       }
+
+       err = ubi_debugfs_init();
+       if (err)
+               goto out_slab;
+
 
        /* Attach MTD devices */
        for (i = 0; i < mtd_devs; i++) {
@@ -1000,17 +1298,36 @@ int __init ubi_init(void)
                mtd = open_mtd_device(p->name);
                if (IS_ERR(mtd)) {
                        err = PTR_ERR(mtd);
-                       goto out_detach;
+                       ubi_err("cannot open mtd %s, error %d", p->name, err);
+                       /* See comment below re-ubi_is_module(). */
+                       if (ubi_is_module())
+                               goto out_detach;
+                       continue;
                }
 
                mutex_lock(&ubi_devices_mutex);
-               err = ubi_attach_mtd_dev(mtd, UBI_DEV_NUM_AUTO,
-                                        p->vid_hdr_offs);
+               err = ubi_attach_mtd_dev(mtd, p->ubi_num,
+                                        p->vid_hdr_offs, p->max_beb_per1024);
                mutex_unlock(&ubi_devices_mutex);
                if (err < 0) {
-                       put_mtd_device(mtd);
                        ubi_err("cannot attach mtd%d", mtd->index);
-                       goto out_detach;
+                       put_mtd_device(mtd);
+
+                       /*
+                        * Originally UBI stopped initializing on any error.
+                        * However, later on it was found out that this
+                        * behavior is not very good when UBI is compiled into
+                        * the kernel and the MTD devices to attach are passed
+                        * through the command line. Indeed, UBI failure
+                        * stopped whole boot sequence.
+                        *
+                        * To fix this, we changed the behavior for the
+                        * non-module case, but preserved the old behavior for
+                        * the module case, just for compatibility. This is a
+                        * little inconsistent, though.
+                        */
+                       if (ubi_is_module())
+                               goto out_detach;
                }
        }
 
@@ -1023,23 +1340,26 @@ out_detach:
                        ubi_detach_mtd_dev(ubi_devices[k]->ubi_num, 1);
                        mutex_unlock(&ubi_devices_mutex);
                }
-#ifdef UBI_LINUX
+       ubi_debugfs_exit();
+out_slab:
        kmem_cache_destroy(ubi_wl_entry_slab);
 out_dev_unreg:
-#endif
        misc_deregister(&ubi_ctrl_cdev);
 out_version:
        class_remove_file(ubi_class, &ubi_version);
 out_class:
        class_destroy(ubi_class);
 out:
-       mtd_devs = 0;
-       ubi_err("UBI error: cannot initialize UBI, error %d", err);
+       ubi_err("cannot initialize UBI, error %d", err);
        return err;
 }
-module_init(ubi_init);
+late_initcall(ubi_init);
 
-void __exit ubi_exit(void)
+#ifndef __UBOOT__
+static void __exit ubi_exit(void)
+#else
+void ubi_exit(void)
+#endif
 {
        int i;
 
@@ -1049,17 +1369,16 @@ void __exit ubi_exit(void)
                        ubi_detach_mtd_dev(ubi_devices[i]->ubi_num, 1);
                        mutex_unlock(&ubi_devices_mutex);
                }
+       ubi_debugfs_exit();
        kmem_cache_destroy(ubi_wl_entry_slab);
        misc_deregister(&ubi_ctrl_cdev);
        class_remove_file(ubi_class, &ubi_version);
        class_destroy(ubi_class);
-       mtd_devs = 0;
 }
 module_exit(ubi_exit);
 
 /**
- * bytes_str_to_int - convert a string representing number of bytes to an
- * integer.
+ * bytes_str_to_int - convert a number of bytes string into an integer.
  * @str: the string to convert
  *
  * This function returns positive resulting integer in case of success and a
@@ -1071,9 +1390,8 @@ static int __init bytes_str_to_int(const char *str)
        unsigned long result;
 
        result = simple_strtoul(str, &endp, 0);
-       if (str == endp || result < 0) {
-               printk(KERN_ERR "UBI error: incorrect bytes count: \"%s\"\n",
-                      str);
+       if (str == endp || result >= INT_MAX) {
+               ubi_err("incorrect bytes count: \"%s\"\n", str);
                return -EINVAL;
        }
 
@@ -1089,14 +1407,24 @@ static int __init bytes_str_to_int(const char *str)
        case '\0':
                break;
        default:
-               printk(KERN_ERR "UBI error: incorrect bytes count: \"%s\"\n",
-                      str);
+               ubi_err("incorrect bytes count: \"%s\"\n", str);
                return -EINVAL;
        }
 
        return result;
 }
 
+int kstrtoint(const char *s, unsigned int base, int *res)
+{
+       unsigned long long tmp;
+
+       tmp = simple_strtoull(s, NULL, base);
+       if (tmp != (unsigned long long)(int)tmp)
+               return -ERANGE;
+
+       return (int)tmp;
+}
+
 /**
  * ubi_mtd_param_parse - parse the 'mtd=' UBI parameter.
  * @val: the parameter value to parse
@@ -1105,33 +1433,36 @@ static int __init bytes_str_to_int(const char *str)
  * This function returns zero in case of success and a negative error code in
  * case of error.
  */
-int __init ubi_mtd_param_parse(const char *val, struct kernel_param *kp)
+#ifndef __UBOOT__
+static int __init ubi_mtd_param_parse(const char *val, struct kernel_param *kp)
+#else
+int ubi_mtd_param_parse(const char *val, struct kernel_param *kp)
+#endif
 {
        int i, len;
        struct mtd_dev_param *p;
        char buf[MTD_PARAM_LEN_MAX];
        char *pbuf = &buf[0];
-       char *tokens[2] = {NULL, NULL};
+       char *tokens[MTD_PARAM_MAX_COUNT], *token;
 
        if (!val)
                return -EINVAL;
 
        if (mtd_devs == UBI_MAX_DEVICES) {
-               printk(KERN_ERR "UBI error: too many parameters, max. is %d\n",
-                      UBI_MAX_DEVICES);
+               ubi_err("too many parameters, max. is %d\n",
+                       UBI_MAX_DEVICES);
                return -EINVAL;
        }
 
        len = strnlen(val, MTD_PARAM_LEN_MAX);
        if (len == MTD_PARAM_LEN_MAX) {
-               printk(KERN_ERR "UBI error: parameter \"%s\" is too long, "
-                      "max. is %d\n", val, MTD_PARAM_LEN_MAX);
+               ubi_err("parameter \"%s\" is too long, max. is %d\n",
+                       val, MTD_PARAM_LEN_MAX);
                return -EINVAL;
        }
 
        if (len == 0) {
-               printk(KERN_WARNING "UBI warning: empty 'mtd=' parameter - "
-                      "ignored\n");
+               pr_warn("UBI warning: empty 'mtd=' parameter - ignored\n");
                return 0;
        }
 
@@ -1141,40 +1472,69 @@ int __init ubi_mtd_param_parse(const char *val, struct kernel_param *kp)
        if (buf[len - 1] == '\n')
                buf[len - 1] = '\0';
 
-       for (i = 0; i < 2; i++)
+       for (i = 0; i < MTD_PARAM_MAX_COUNT; i++)
                tokens[i] = strsep(&pbuf, ",");
 
        if (pbuf) {
-               printk(KERN_ERR "UBI error: too many arguments at \"%s\"\n",
-                      val);
+               ubi_err("too many arguments at \"%s\"\n", val);
                return -EINVAL;
        }
 
        p = &mtd_dev_param[mtd_devs];
        strcpy(&p->name[0], tokens[0]);
 
-       if (tokens[1])
-               p->vid_hdr_offs = bytes_str_to_int(tokens[1]);
+       token = tokens[1];
+       if (token) {
+               p->vid_hdr_offs = bytes_str_to_int(token);
+
+               if (p->vid_hdr_offs < 0)
+                       return p->vid_hdr_offs;
+       }
 
-       if (p->vid_hdr_offs < 0)
-               return p->vid_hdr_offs;
+       token = tokens[2];
+       if (token) {
+               int err = kstrtoint(token, 10, &p->max_beb_per1024);
+
+               if (err) {
+                       ubi_err("bad value for max_beb_per1024 parameter: %s",
+                               token);
+                       return -EINVAL;
+               }
+       }
+
+       token = tokens[3];
+       if (token) {
+               int err = kstrtoint(token, 10, &p->ubi_num);
+
+               if (err) {
+                       ubi_err("bad value for ubi_num parameter: %s", token);
+                       return -EINVAL;
+               }
+       } else
+               p->ubi_num = UBI_DEV_NUM_AUTO;
 
        mtd_devs += 1;
        return 0;
 }
 
 module_param_call(mtd, ubi_mtd_param_parse, NULL, NULL, 000);
-MODULE_PARM_DESC(mtd, "MTD devices to attach. Parameter format: "
-                     "mtd=<name|num>[,<vid_hdr_offs>].\n"
+MODULE_PARM_DESC(mtd, "MTD devices to attach. Parameter format: mtd=<name|num|path>[,<vid_hdr_offs>[,max_beb_per1024[,ubi_num]]].\n"
                      "Multiple \"mtd\" parameters may be specified.\n"
-                     "MTD devices may be specified by their number or name.\n"
-                     "Optional \"vid_hdr_offs\" parameter specifies UBI VID "
-                     "header position and data starting position to be used "
-                     "by UBI.\n"
-                     "Example: mtd=content,1984 mtd=4 - attach MTD device"
-                     "with name \"content\" using VID header offset 1984, and "
-                     "MTD device number 4 with default VID header offset.");
-
+                     "MTD devices may be specified by their number, name, or path to the MTD character device node.\n"
+                     "Optional \"vid_hdr_offs\" parameter specifies UBI VID header position to be used by UBI. (default value if 0)\n"
+                     "Optional \"max_beb_per1024\" parameter specifies the maximum expected bad eraseblock per 1024 eraseblocks. (default value ("
+                     __stringify(CONFIG_MTD_UBI_BEB_LIMIT) ") if 0)\n"
+                     "Optional \"ubi_num\" parameter specifies UBI device number which have to be assigned to the newly created UBI device (assigned automatically by default)\n"
+                     "\n"
+                     "Example 1: mtd=/dev/mtd0 - attach MTD device /dev/mtd0.\n"
+                     "Example 2: mtd=content,1984 mtd=4 - attach MTD device with name \"content\" using VID header offset 1984, and MTD device number 4 with default VID header offset.\n"
+                     "Example 3: mtd=/dev/mtd1,0,25 - attach MTD device /dev/mtd1 using default VID header offset and reserve 25*nand_size_in_blocks/1024 erase blocks for bad block handling.\n"
+                     "Example 4: mtd=/dev/mtd1,0,0,5 - attach MTD device /dev/mtd1 to UBI 5 and using default values for the other fields.\n"
+                     "\t(e.g. if the NAND *chipset* has 4096 PEB, 100 will be reserved for this UBI device).");
+#ifdef CONFIG_MTD_UBI_FASTMAP
+module_param(fm_autoconvert, bool, 0644);
+MODULE_PARM_DESC(fm_autoconvert, "Set this parameter to enable fastmap automatically on images without a fastmap.");
+#endif
 MODULE_VERSION(__stringify(UBI_VERSION));
 MODULE_DESCRIPTION("UBI - Unsorted Block Images");
 MODULE_AUTHOR("Artem Bityutskiy");
index f1bebf58c12816ff5e33890f3afc561437affbcd..0d65bf4b8a562f5cbc4549520cb0deedd40b21e7 100644 (file)
@@ -20,7 +20,8 @@
  * Version 2.  See the file COPYING for more details.
  */
 
-#ifdef UBI_LINUX
+#define __UBOOT__
+#ifndef __UBOOT__
 #include <linux/crc32.h>
 #include <linux/kernel.h>
 #include <linux/module.h>
@@ -30,7 +31,7 @@
 
 #include <asm/byteorder.h>
 
-#ifdef UBI_LINUX
+#ifndef __UBOOT__
 #include <linux/slab.h>
 #include <linux/init.h>
 #include <asm/atomic.h>
@@ -46,7 +47,7 @@
 #define tobe(x) (x)
 #endif
 #include "crc32table.h"
-#ifdef UBI_LINUX
+#ifndef __UBOOT__
 MODULE_AUTHOR("Matt Domsch <Matt_Domsch@dell.com>");
 MODULE_DESCRIPTION("Ethernet CRC32 calculations");
 MODULE_LICENSE("GPL");
@@ -146,7 +147,7 @@ u32 crc32_le(u32 crc, unsigned char const *p, size_t len)
 # endif
 }
 #endif
-#ifdef UBI_LINUX
+#ifndef __UBOOT__
 /**
  * crc32_be() - Calculate bitwise big-endian Ethernet AUTODIN II CRC32
  * @crc: seed value for computation.  ~0 for Ethernet, sometimes 0 for
@@ -379,7 +380,7 @@ EXPORT_SYMBOL(crc32_be);
 #include <stdlib.h>
 #include <stdio.h>
 
-#ifdef UBI_LINUX                               /*Not used at present */
+#ifndef __UBOOT__
 static void
 buf_dump(char const *prefix, unsigned char const *buf, size_t len)
 {
@@ -405,7 +406,7 @@ static void random_garbage(unsigned char *buf, size_t len)
                *buf++ = (unsigned char) random();
 }
 
-#ifdef UBI_LINUX                               /* Not used at present */
+#ifndef __UBOOT__
 static void store_le(u32 x, unsigned char *buf)
 {
        buf[0] = (unsigned char) x;
index 0438af43500e9d502b65e238fad3b9c69955bdea..02ce6fd90173658fa452259b9ec5f88f40f5eee9 100644 (file)
@@ -66,7 +66,7 @@ tole(0xbad03605L), tole(0xcdd70693L), tole(0x54de5729L), tole(0x23d967bfL),
 tole(0xb3667a2eL), tole(0xc4614ab8L), tole(0x5d681b02L), tole(0x2a6f2b94L),
 tole(0xb40bbe37L), tole(0xc30c8ea1L), tole(0x5a05df1bL), tole(0x2d02ef8dL)
 };
-#ifdef UBI_LINUX
+#ifndef __UBOOT__
 static const u32 crc32table_be[] = {
 tobe(0x00000000L), tobe(0x04c11db7L), tobe(0x09823b6eL), tobe(0x0d4326d9L),
 tobe(0x130476dcL), tobe(0x17c56b6bL), tobe(0x1a864db2L), tobe(0x1e475005L),
index 6c22301d932cc47d5e702a388e507b0f0518d9c7..af254da4883c375cc9dd39efdb4df43c6454ea46 100644 (file)
  * Author: Artem Bityutskiy (Битюцкий Артём)
  */
 
-/*
- * Here we keep all the UBI debugging stuff which should normally be disabled
- * and compiled-out, but it is extremely helpful when hunting bugs or doing big
- * changes.
- */
 #include <ubi_uboot.h>
+#include "ubi.h"
+#define __UBOOT__
+#ifndef __UBOOT__
+#include <linux/debugfs.h>
+#include <linux/uaccess.h>
+#include <linux/module.h>
+#endif
 
-#ifdef CONFIG_MTD_UBI_DEBUG_MSG
+/**
+ * ubi_dump_flash - dump a region of flash.
+ * @ubi: UBI device description object
+ * @pnum: the physical eraseblock number to dump
+ * @offset: the starting offset within the physical eraseblock to dump
+ * @len: the length of the region to dump
+ */
+void ubi_dump_flash(struct ubi_device *ubi, int pnum, int offset, int len)
+{
+       int err;
+       size_t read;
+       void *buf;
+       loff_t addr = (loff_t)pnum * ubi->peb_size + offset;
 
-#include "ubi.h"
+       buf = vmalloc(len);
+       if (!buf)
+               return;
+       err = mtd_read(ubi->mtd, addr, len, &read, buf);
+       if (err && err != -EUCLEAN) {
+               ubi_err("error %d while reading %d bytes from PEB %d:%d, read %zd bytes",
+                       err, len, pnum, offset, read);
+               goto out;
+       }
+
+       ubi_msg("dumping %d bytes of data from PEB %d, offset %d",
+               len, pnum, offset);
+       print_hex_dump(KERN_DEBUG, "", DUMP_PREFIX_OFFSET, 32, 1, buf, len, 1);
+out:
+       vfree(buf);
+       return;
+}
 
 /**
- * ubi_dbg_dump_ec_hdr - dump an erase counter header.
+ * ubi_dump_ec_hdr - dump an erase counter header.
  * @ec_hdr: the erase counter header to dump
  */
-void ubi_dbg_dump_ec_hdr(const struct ubi_ec_hdr *ec_hdr)
-{
-       dbg_msg("erase counter header dump:");
-       dbg_msg("magic          %#08x", be32_to_cpu(ec_hdr->magic));
-       dbg_msg("version        %d",    (int)ec_hdr->version);
-       dbg_msg("ec             %llu",  (long long)be64_to_cpu(ec_hdr->ec));
-       dbg_msg("vid_hdr_offset %d",    be32_to_cpu(ec_hdr->vid_hdr_offset));
-       dbg_msg("data_offset    %d",    be32_to_cpu(ec_hdr->data_offset));
-       dbg_msg("hdr_crc        %#08x", be32_to_cpu(ec_hdr->hdr_crc));
-       dbg_msg("erase counter header hexdump:");
+void ubi_dump_ec_hdr(const struct ubi_ec_hdr *ec_hdr)
+{
+       pr_err("Erase counter header dump:\n");
+       pr_err("\tmagic          %#08x\n", be32_to_cpu(ec_hdr->magic));
+       pr_err("\tversion        %d\n", (int)ec_hdr->version);
+       pr_err("\tec             %llu\n", (long long)be64_to_cpu(ec_hdr->ec));
+       pr_err("\tvid_hdr_offset %d\n", be32_to_cpu(ec_hdr->vid_hdr_offset));
+       pr_err("\tdata_offset    %d\n", be32_to_cpu(ec_hdr->data_offset));
+       pr_err("\timage_seq      %d\n", be32_to_cpu(ec_hdr->image_seq));
+       pr_err("\thdr_crc        %#08x\n", be32_to_cpu(ec_hdr->hdr_crc));
+       pr_err("erase counter header hexdump:\n");
        print_hex_dump(KERN_DEBUG, "", DUMP_PREFIX_OFFSET, 32, 1,
                       ec_hdr, UBI_EC_HDR_SIZE, 1);
 }
 
 /**
- * ubi_dbg_dump_vid_hdr - dump a volume identifier header.
+ * ubi_dump_vid_hdr - dump a volume identifier header.
  * @vid_hdr: the volume identifier header to dump
  */
-void ubi_dbg_dump_vid_hdr(const struct ubi_vid_hdr *vid_hdr)
-{
-       dbg_msg("volume identifier header dump:");
-       dbg_msg("magic     %08x", be32_to_cpu(vid_hdr->magic));
-       dbg_msg("version   %d",   (int)vid_hdr->version);
-       dbg_msg("vol_type  %d",   (int)vid_hdr->vol_type);
-       dbg_msg("copy_flag %d",   (int)vid_hdr->copy_flag);
-       dbg_msg("compat    %d",   (int)vid_hdr->compat);
-       dbg_msg("vol_id    %d",   be32_to_cpu(vid_hdr->vol_id));
-       dbg_msg("lnum      %d",   be32_to_cpu(vid_hdr->lnum));
-       dbg_msg("leb_ver   %u",   be32_to_cpu(vid_hdr->leb_ver));
-       dbg_msg("data_size %d",   be32_to_cpu(vid_hdr->data_size));
-       dbg_msg("used_ebs  %d",   be32_to_cpu(vid_hdr->used_ebs));
-       dbg_msg("data_pad  %d",   be32_to_cpu(vid_hdr->data_pad));
-       dbg_msg("sqnum     %llu",
+void ubi_dump_vid_hdr(const struct ubi_vid_hdr *vid_hdr)
+{
+       pr_err("Volume identifier header dump:\n");
+       pr_err("\tmagic     %08x\n", be32_to_cpu(vid_hdr->magic));
+       pr_err("\tversion   %d\n",  (int)vid_hdr->version);
+       pr_err("\tvol_type  %d\n",  (int)vid_hdr->vol_type);
+       pr_err("\tcopy_flag %d\n",  (int)vid_hdr->copy_flag);
+       pr_err("\tcompat    %d\n",  (int)vid_hdr->compat);
+       pr_err("\tvol_id    %d\n",  be32_to_cpu(vid_hdr->vol_id));
+       pr_err("\tlnum      %d\n",  be32_to_cpu(vid_hdr->lnum));
+       pr_err("\tdata_size %d\n",  be32_to_cpu(vid_hdr->data_size));
+       pr_err("\tused_ebs  %d\n",  be32_to_cpu(vid_hdr->used_ebs));
+       pr_err("\tdata_pad  %d\n",  be32_to_cpu(vid_hdr->data_pad));
+       pr_err("\tsqnum     %llu\n",
                (unsigned long long)be64_to_cpu(vid_hdr->sqnum));
-       dbg_msg("hdr_crc   %08x", be32_to_cpu(vid_hdr->hdr_crc));
-       dbg_msg("volume identifier header hexdump:");
+       pr_err("\thdr_crc   %08x\n", be32_to_cpu(vid_hdr->hdr_crc));
+       pr_err("Volume identifier header hexdump:\n");
+       print_hex_dump(KERN_DEBUG, "", DUMP_PREFIX_OFFSET, 32, 1,
+                      vid_hdr, UBI_VID_HDR_SIZE, 1);
 }
 
 /**
- * ubi_dbg_dump_vol_info- dump volume information.
+ * ubi_dump_vol_info - dump volume information.
  * @vol: UBI volume description object
  */
-void ubi_dbg_dump_vol_info(const struct ubi_volume *vol)
-{
-       dbg_msg("volume information dump:");
-       dbg_msg("vol_id          %d", vol->vol_id);
-       dbg_msg("reserved_pebs   %d", vol->reserved_pebs);
-       dbg_msg("alignment       %d", vol->alignment);
-       dbg_msg("data_pad        %d", vol->data_pad);
-       dbg_msg("vol_type        %d", vol->vol_type);
-       dbg_msg("name_len        %d", vol->name_len);
-       dbg_msg("usable_leb_size %d", vol->usable_leb_size);
-       dbg_msg("used_ebs        %d", vol->used_ebs);
-       dbg_msg("used_bytes      %lld", vol->used_bytes);
-       dbg_msg("last_eb_bytes   %d", vol->last_eb_bytes);
-       dbg_msg("corrupted       %d", vol->corrupted);
-       dbg_msg("upd_marker      %d", vol->upd_marker);
+void ubi_dump_vol_info(const struct ubi_volume *vol)
+{
+       printf("Volume information dump:\n");
+       printf("\tvol_id          %d\n", vol->vol_id);
+       printf("\treserved_pebs   %d\n", vol->reserved_pebs);
+       printf("\talignment       %d\n", vol->alignment);
+       printf("\tdata_pad        %d\n", vol->data_pad);
+       printf("\tvol_type        %d\n", vol->vol_type);
+       printf("\tname_len        %d\n", vol->name_len);
+       printf("\tusable_leb_size %d\n", vol->usable_leb_size);
+       printf("\tused_ebs        %d\n", vol->used_ebs);
+       printf("\tused_bytes      %lld\n", vol->used_bytes);
+       printf("\tlast_eb_bytes   %d\n", vol->last_eb_bytes);
+       printf("\tcorrupted       %d\n", vol->corrupted);
+       printf("\tupd_marker      %d\n", vol->upd_marker);
 
        if (vol->name_len <= UBI_VOL_NAME_MAX &&
            strnlen(vol->name, vol->name_len + 1) == vol->name_len) {
-               dbg_msg("name            %s", vol->name);
+               printf("\tname            %s\n", vol->name);
        } else {
-               dbg_msg("the 1st 5 characters of the name: %c%c%c%c%c",
-                       vol->name[0], vol->name[1], vol->name[2],
-                       vol->name[3], vol->name[4]);
+               printf("\t1st 5 characters of name: %c%c%c%c%c\n",
+                      vol->name[0], vol->name[1], vol->name[2],
+                      vol->name[3], vol->name[4]);
        }
 }
 
 /**
- * ubi_dbg_dump_vtbl_record - dump a &struct ubi_vtbl_record object.
+ * ubi_dump_vtbl_record - dump a &struct ubi_vtbl_record object.
  * @r: the object to dump
  * @idx: volume table index
  */
-void ubi_dbg_dump_vtbl_record(const struct ubi_vtbl_record *r, int idx)
+void ubi_dump_vtbl_record(const struct ubi_vtbl_record *r, int idx)
 {
        int name_len = be16_to_cpu(r->name_len);
 
-       dbg_msg("volume table record %d dump:", idx);
-       dbg_msg("reserved_pebs   %d", be32_to_cpu(r->reserved_pebs));
-       dbg_msg("alignment       %d", be32_to_cpu(r->alignment));
-       dbg_msg("data_pad        %d", be32_to_cpu(r->data_pad));
-       dbg_msg("vol_type        %d", (int)r->vol_type);
-       dbg_msg("upd_marker      %d", (int)r->upd_marker);
-       dbg_msg("name_len        %d", name_len);
+       pr_err("Volume table record %d dump:\n", idx);
+       pr_err("\treserved_pebs   %d\n", be32_to_cpu(r->reserved_pebs));
+       pr_err("\talignment       %d\n", be32_to_cpu(r->alignment));
+       pr_err("\tdata_pad        %d\n", be32_to_cpu(r->data_pad));
+       pr_err("\tvol_type        %d\n", (int)r->vol_type);
+       pr_err("\tupd_marker      %d\n", (int)r->upd_marker);
+       pr_err("\tname_len        %d\n", name_len);
 
        if (r->name[0] == '\0') {
-               dbg_msg("name            NULL");
+               pr_err("\tname            NULL\n");
                return;
        }
 
        if (name_len <= UBI_VOL_NAME_MAX &&
            strnlen(&r->name[0], name_len + 1) == name_len) {
-               dbg_msg("name            %s", &r->name[0]);
+               pr_err("\tname            %s\n", &r->name[0]);
        } else {
-               dbg_msg("1st 5 characters of the name: %c%c%c%c%c",
+               pr_err("\t1st 5 characters of name: %c%c%c%c%c\n",
                        r->name[0], r->name[1], r->name[2], r->name[3],
                        r->name[4]);
        }
-       dbg_msg("crc             %#08x", be32_to_cpu(r->crc));
+       pr_err("\tcrc             %#08x\n", be32_to_cpu(r->crc));
 }
 
 /**
- * ubi_dbg_dump_sv - dump a &struct ubi_scan_volume object.
- * @sv: the object to dump
+ * ubi_dump_av - dump a &struct ubi_ainf_volume object.
+ * @av: the object to dump
  */
-void ubi_dbg_dump_sv(const struct ubi_scan_volume *sv)
+void ubi_dump_av(const struct ubi_ainf_volume *av)
 {
-       dbg_msg("volume scanning information dump:");
-       dbg_msg("vol_id         %d", sv->vol_id);
-       dbg_msg("highest_lnum   %d", sv->highest_lnum);
-       dbg_msg("leb_count      %d", sv->leb_count);
-       dbg_msg("compat         %d", sv->compat);
-       dbg_msg("vol_type       %d", sv->vol_type);
-       dbg_msg("used_ebs       %d", sv->used_ebs);
-       dbg_msg("last_data_size %d", sv->last_data_size);
-       dbg_msg("data_pad       %d", sv->data_pad);
+       pr_err("Volume attaching information dump:\n");
+       pr_err("\tvol_id         %d\n", av->vol_id);
+       pr_err("\thighest_lnum   %d\n", av->highest_lnum);
+       pr_err("\tleb_count      %d\n", av->leb_count);
+       pr_err("\tcompat         %d\n", av->compat);
+       pr_err("\tvol_type       %d\n", av->vol_type);
+       pr_err("\tused_ebs       %d\n", av->used_ebs);
+       pr_err("\tlast_data_size %d\n", av->last_data_size);
+       pr_err("\tdata_pad       %d\n", av->data_pad);
 }
 
 /**
- * ubi_dbg_dump_seb - dump a &struct ubi_scan_leb object.
- * @seb: the object to dump
+ * ubi_dump_aeb - dump a &struct ubi_ainf_peb object.
+ * @aeb: the object to dump
  * @type: object type: 0 - not corrupted, 1 - corrupted
  */
-void ubi_dbg_dump_seb(const struct ubi_scan_leb *seb, int type)
+void ubi_dump_aeb(const struct ubi_ainf_peb *aeb, int type)
 {
-       dbg_msg("eraseblock scanning information dump:");
-       dbg_msg("ec       %d", seb->ec);
-       dbg_msg("pnum     %d", seb->pnum);
+       pr_err("eraseblock attaching information dump:\n");
+       pr_err("\tec       %d\n", aeb->ec);
+       pr_err("\tpnum     %d\n", aeb->pnum);
        if (type == 0) {
-               dbg_msg("lnum     %d", seb->lnum);
-               dbg_msg("scrub    %d", seb->scrub);
-               dbg_msg("sqnum    %llu", seb->sqnum);
-               dbg_msg("leb_ver  %u", seb->leb_ver);
+               pr_err("\tlnum     %d\n", aeb->lnum);
+               pr_err("\tscrub    %d\n", aeb->scrub);
+               pr_err("\tsqnum    %llu\n", aeb->sqnum);
        }
 }
 
 /**
- * ubi_dbg_dump_mkvol_req - dump a &struct ubi_mkvol_req object.
+ * ubi_dump_mkvol_req - dump a &struct ubi_mkvol_req object.
  * @req: the object to dump
  */
-void ubi_dbg_dump_mkvol_req(const struct ubi_mkvol_req *req)
+void ubi_dump_mkvol_req(const struct ubi_mkvol_req *req)
 {
        char nm[17];
 
-       dbg_msg("volume creation request dump:");
-       dbg_msg("vol_id    %d",   req->vol_id);
-       dbg_msg("alignment %d",   req->alignment);
-       dbg_msg("bytes     %lld", (long long)req->bytes);
-       dbg_msg("vol_type  %d",   req->vol_type);
-       dbg_msg("name_len  %d",   req->name_len);
+       pr_err("Volume creation request dump:\n");
+       pr_err("\tvol_id    %d\n",   req->vol_id);
+       pr_err("\talignment %d\n",   req->alignment);
+       pr_err("\tbytes     %lld\n", (long long)req->bytes);
+       pr_err("\tvol_type  %d\n",   req->vol_type);
+       pr_err("\tname_len  %d\n",   req->name_len);
 
        memcpy(nm, req->name, 16);
        nm[16] = 0;
-       dbg_msg("the 1st 16 characters of the name: %s", nm);
+       pr_err("\t1st 16 characters of name: %s\n", nm);
 }
 
-#endif /* CONFIG_MTD_UBI_DEBUG_MSG */
+#ifndef __UBOOT__
+/*
+ * Root directory for UBI stuff in debugfs. Contains sub-directories which
+ * contain the stuff specific to particular UBI devices.
+ */
+static struct dentry *dfs_rootdir;
+
+/**
+ * ubi_debugfs_init - create UBI debugfs directory.
+ *
+ * Create UBI debugfs directory. Returns zero in case of success and a negative
+ * error code in case of failure.
+ */
+int ubi_debugfs_init(void)
+{
+       if (!IS_ENABLED(CONFIG_DEBUG_FS))
+               return 0;
+
+       dfs_rootdir = debugfs_create_dir("ubi", NULL);
+       if (IS_ERR_OR_NULL(dfs_rootdir)) {
+               int err = dfs_rootdir ? -ENODEV : PTR_ERR(dfs_rootdir);
+
+               ubi_err("cannot create \"ubi\" debugfs directory, error %d\n",
+                       err);
+               return err;
+       }
+
+       return 0;
+}
+
+/**
+ * ubi_debugfs_exit - remove UBI debugfs directory.
+ */
+void ubi_debugfs_exit(void)
+{
+       if (IS_ENABLED(CONFIG_DEBUG_FS))
+               debugfs_remove(dfs_rootdir);
+}
+
+/* Read an UBI debugfs file */
+static ssize_t dfs_file_read(struct file *file, char __user *user_buf,
+                            size_t count, loff_t *ppos)
+{
+       unsigned long ubi_num = (unsigned long)file->private_data;
+       struct dentry *dent = file->f_path.dentry;
+       struct ubi_device *ubi;
+       struct ubi_debug_info *d;
+       char buf[3];
+       int val;
+
+       ubi = ubi_get_device(ubi_num);
+       if (!ubi)
+               return -ENODEV;
+       d = &ubi->dbg;
+
+       if (dent == d->dfs_chk_gen)
+               val = d->chk_gen;
+       else if (dent == d->dfs_chk_io)
+               val = d->chk_io;
+       else if (dent == d->dfs_disable_bgt)
+               val = d->disable_bgt;
+       else if (dent == d->dfs_emulate_bitflips)
+               val = d->emulate_bitflips;
+       else if (dent == d->dfs_emulate_io_failures)
+               val = d->emulate_io_failures;
+       else {
+               count = -EINVAL;
+               goto out;
+       }
+
+       if (val)
+               buf[0] = '1';
+       else
+               buf[0] = '0';
+       buf[1] = '\n';
+       buf[2] = 0x00;
+
+       count = simple_read_from_buffer(user_buf, count, ppos, buf, 2);
+
+out:
+       ubi_put_device(ubi);
+       return count;
+}
+
+/* Write an UBI debugfs file */
+static ssize_t dfs_file_write(struct file *file, const char __user *user_buf,
+                             size_t count, loff_t *ppos)
+{
+       unsigned long ubi_num = (unsigned long)file->private_data;
+       struct dentry *dent = file->f_path.dentry;
+       struct ubi_device *ubi;
+       struct ubi_debug_info *d;
+       size_t buf_size;
+       char buf[8];
+       int val;
+
+       ubi = ubi_get_device(ubi_num);
+       if (!ubi)
+               return -ENODEV;
+       d = &ubi->dbg;
+
+       buf_size = min_t(size_t, count, (sizeof(buf) - 1));
+       if (copy_from_user(buf, user_buf, buf_size)) {
+               count = -EFAULT;
+               goto out;
+       }
+
+       if (buf[0] == '1')
+               val = 1;
+       else if (buf[0] == '0')
+               val = 0;
+       else {
+               count = -EINVAL;
+               goto out;
+       }
+
+       if (dent == d->dfs_chk_gen)
+               d->chk_gen = val;
+       else if (dent == d->dfs_chk_io)
+               d->chk_io = val;
+       else if (dent == d->dfs_disable_bgt)
+               d->disable_bgt = val;
+       else if (dent == d->dfs_emulate_bitflips)
+               d->emulate_bitflips = val;
+       else if (dent == d->dfs_emulate_io_failures)
+               d->emulate_io_failures = val;
+       else
+               count = -EINVAL;
+
+out:
+       ubi_put_device(ubi);
+       return count;
+}
+
+/* File operations for all UBI debugfs files */
+static const struct file_operations dfs_fops = {
+       .read   = dfs_file_read,
+       .write  = dfs_file_write,
+       .open   = simple_open,
+       .llseek = no_llseek,
+       .owner  = THIS_MODULE,
+};
+
+/**
+ * ubi_debugfs_init_dev - initialize debugfs for an UBI device.
+ * @ubi: UBI device description object
+ *
+ * This function creates all debugfs files for UBI device @ubi. Returns zero in
+ * case of success and a negative error code in case of failure.
+ */
+int ubi_debugfs_init_dev(struct ubi_device *ubi)
+{
+       int err, n;
+       unsigned long ubi_num = ubi->ubi_num;
+       const char *fname;
+       struct dentry *dent;
+       struct ubi_debug_info *d = &ubi->dbg;
+
+       if (!IS_ENABLED(CONFIG_DEBUG_FS))
+               return 0;
+
+       n = snprintf(d->dfs_dir_name, UBI_DFS_DIR_LEN + 1, UBI_DFS_DIR_NAME,
+                    ubi->ubi_num);
+       if (n == UBI_DFS_DIR_LEN) {
+               /* The array size is too small */
+               fname = UBI_DFS_DIR_NAME;
+               dent = ERR_PTR(-EINVAL);
+               goto out;
+       }
+
+       fname = d->dfs_dir_name;
+       dent = debugfs_create_dir(fname, dfs_rootdir);
+       if (IS_ERR_OR_NULL(dent))
+               goto out;
+       d->dfs_dir = dent;
+
+       fname = "chk_gen";
+       dent = debugfs_create_file(fname, S_IWUSR, d->dfs_dir, (void *)ubi_num,
+                                  &dfs_fops);
+       if (IS_ERR_OR_NULL(dent))
+               goto out_remove;
+       d->dfs_chk_gen = dent;
+
+       fname = "chk_io";
+       dent = debugfs_create_file(fname, S_IWUSR, d->dfs_dir, (void *)ubi_num,
+                                  &dfs_fops);
+       if (IS_ERR_OR_NULL(dent))
+               goto out_remove;
+       d->dfs_chk_io = dent;
+
+       fname = "tst_disable_bgt";
+       dent = debugfs_create_file(fname, S_IWUSR, d->dfs_dir, (void *)ubi_num,
+                                  &dfs_fops);
+       if (IS_ERR_OR_NULL(dent))
+               goto out_remove;
+       d->dfs_disable_bgt = dent;
+
+       fname = "tst_emulate_bitflips";
+       dent = debugfs_create_file(fname, S_IWUSR, d->dfs_dir, (void *)ubi_num,
+                                  &dfs_fops);
+       if (IS_ERR_OR_NULL(dent))
+               goto out_remove;
+       d->dfs_emulate_bitflips = dent;
+
+       fname = "tst_emulate_io_failures";
+       dent = debugfs_create_file(fname, S_IWUSR, d->dfs_dir, (void *)ubi_num,
+                                  &dfs_fops);
+       if (IS_ERR_OR_NULL(dent))
+               goto out_remove;
+       d->dfs_emulate_io_failures = dent;
+
+       return 0;
+
+out_remove:
+       debugfs_remove_recursive(d->dfs_dir);
+out:
+       err = dent ? PTR_ERR(dent) : -ENODEV;
+       ubi_err("cannot create \"%s\" debugfs file or directory, error %d\n",
+               fname, err);
+       return err;
+}
+
+/**
+ * dbg_debug_exit_dev - free all debugfs files corresponding to device @ubi
+ * @ubi: UBI device description object
+ */
+void ubi_debugfs_exit_dev(struct ubi_device *ubi)
+{
+       if (IS_ENABLED(CONFIG_DEBUG_FS))
+               debugfs_remove_recursive(ubi->dbg.dfs_dir);
+}
+#else
+int ubi_debugfs_init(void)
+{
+       return 0;
+}
+
+void ubi_debugfs_exit(void)
+{
+}
+
+int ubi_debugfs_init_dev(struct ubi_device *ubi)
+{
+       return 0;
+}
+
+void ubi_debugfs_exit_dev(struct ubi_device *ubi)
+{
+}
+#endif
index 222b2b8ae969504ea400226f759a8ceea15ed51a..980eb11ed235c50df8e1a6a1b129df22b3ac2435 100644 (file)
 #ifndef __UBI_DEBUG_H__
 #define __UBI_DEBUG_H__
 
-#ifdef CONFIG_MTD_UBI_DEBUG
-#ifdef UBI_LINUX
-#include <linux/random.h>
-#endif
-
-#define ubi_assert(expr)  BUG_ON(!(expr))
-#define dbg_err(fmt, ...) ubi_err(fmt, ##__VA_ARGS__)
-#else
-#define ubi_assert(expr)  ({})
-#define dbg_err(fmt, ...) ({})
-#endif
-
-#ifdef CONFIG_MTD_UBI_DEBUG_DISABLE_BGT
-#define DBG_DISABLE_BGT 1
-#else
-#define DBG_DISABLE_BGT 0
-#endif
-
-#ifdef CONFIG_MTD_UBI_DEBUG_MSG
-/* Generic debugging message */
-#define dbg_msg(fmt, ...)                                    \
-       printk(KERN_DEBUG "UBI DBG: %s: " fmt "\n", \
-              __FUNCTION__, ##__VA_ARGS__)
-
-#define ubi_dbg_dump_stack() dump_stack()
-
-struct ubi_ec_hdr;
-struct ubi_vid_hdr;
-struct ubi_volume;
-struct ubi_vtbl_record;
-struct ubi_scan_volume;
-struct ubi_scan_leb;
-struct ubi_mkvol_req;
-
-void ubi_dbg_dump_ec_hdr(const struct ubi_ec_hdr *ec_hdr);
-void ubi_dbg_dump_vid_hdr(const struct ubi_vid_hdr *vid_hdr);
-void ubi_dbg_dump_vol_info(const struct ubi_volume *vol);
-void ubi_dbg_dump_vtbl_record(const struct ubi_vtbl_record *r, int idx);
-void ubi_dbg_dump_sv(const struct ubi_scan_volume *sv);
-void ubi_dbg_dump_seb(const struct ubi_scan_leb *seb, int type);
-void ubi_dbg_dump_mkvol_req(const struct ubi_mkvol_req *req);
-
-#else
-
-#define dbg_msg(fmt, ...)    ({})
-#define ubi_dbg_dump_stack() ({})
-#define ubi_dbg_dump_ec_hdr(ec_hdr)      ({})
-#define ubi_dbg_dump_vid_hdr(vid_hdr)    ({})
-#define ubi_dbg_dump_vol_info(vol)       ({})
-#define ubi_dbg_dump_vtbl_record(r, idx) ({})
-#define ubi_dbg_dump_sv(sv)              ({})
-#define ubi_dbg_dump_seb(seb, type)      ({})
-#define ubi_dbg_dump_mkvol_req(req)      ({})
-
-#endif /* CONFIG_MTD_UBI_DEBUG_MSG */
-
-#ifdef CONFIG_MTD_UBI_DEBUG_MSG_EBA
-/* Messages from the eraseblock association unit */
-#define dbg_eba(fmt, ...) dbg_msg(fmt, ##__VA_ARGS__)
-#else
-#define dbg_eba(fmt, ...) ({})
-#endif
-
-#ifdef CONFIG_MTD_UBI_DEBUG_MSG_WL
-/* Messages from the wear-leveling unit */
-#define dbg_wl(fmt, ...) dbg_msg(fmt, ##__VA_ARGS__)
-#else
-#define dbg_wl(fmt, ...) ({})
-#endif
+void ubi_dump_flash(struct ubi_device *ubi, int pnum, int offset, int len);
+void ubi_dump_ec_hdr(const struct ubi_ec_hdr *ec_hdr);
+void ubi_dump_vid_hdr(const struct ubi_vid_hdr *vid_hdr);
 
-#ifdef CONFIG_MTD_UBI_DEBUG_MSG_IO
-/* Messages from the input/output unit */
-#define dbg_io(fmt, ...) dbg_msg(fmt, ##__VA_ARGS__)
-#else
-#define dbg_io(fmt, ...) ({})
+#define __UBOOT__
+#ifndef __UBOOT__
+#include <linux/random.h>
 #endif
 
-#ifdef CONFIG_MTD_UBI_DEBUG_MSG_BLD
+#define ubi_assert(expr)  do {                                               \
+       if (unlikely(!(expr))) {                                             \
+               pr_crit("UBI assert failed in %s at %u (pid %d)\n",          \
+                      __func__, __LINE__, current->pid);                    \
+               dump_stack();                                                \
+       }                                                                    \
+} while (0)
+
+#define ubi_dbg_print_hex_dump(l, ps, pt, r, g, b, len, a)                   \
+               print_hex_dump(l, ps, pt, r, g, b, len, a)
+
+#define ubi_dbg_msg(type, fmt, ...) \
+       pr_debug("UBI DBG " type " (pid %d): " fmt "\n", current->pid,       \
+                ##__VA_ARGS__)
+
+/* General debugging messages */
+#define dbg_gen(fmt, ...) ubi_dbg_msg("gen", fmt, ##__VA_ARGS__)
+/* Messages from the eraseblock association sub-system */
+#define dbg_eba(fmt, ...) ubi_dbg_msg("eba", fmt, ##__VA_ARGS__)
+/* Messages from the wear-leveling sub-system */
+#define dbg_wl(fmt, ...)  ubi_dbg_msg("wl", fmt, ##__VA_ARGS__)
+/* Messages from the input/output sub-system */
+#define dbg_io(fmt, ...)  ubi_dbg_msg("io", fmt, ##__VA_ARGS__)
 /* Initialization and build messages */
-#define dbg_bld(fmt, ...) dbg_msg(fmt, ##__VA_ARGS__)
-#else
-#define dbg_bld(fmt, ...) ({})
-#endif
+#define dbg_bld(fmt, ...) ubi_dbg_msg("bld", fmt, ##__VA_ARGS__)
+
+void ubi_dump_vol_info(const struct ubi_volume *vol);
+void ubi_dump_vtbl_record(const struct ubi_vtbl_record *r, int idx);
+void ubi_dump_av(const struct ubi_ainf_volume *av);
+void ubi_dump_aeb(const struct ubi_ainf_peb *aeb, int type);
+void ubi_dump_mkvol_req(const struct ubi_mkvol_req *req);
+int ubi_self_check_all_ff(struct ubi_device *ubi, int pnum, int offset,
+                         int len);
+int ubi_debugfs_init(void);
+void ubi_debugfs_exit(void);
+int ubi_debugfs_init_dev(struct ubi_device *ubi);
+void ubi_debugfs_exit_dev(struct ubi_device *ubi);
+
+/**
+ * ubi_dbg_is_bgt_disabled - if the background thread is disabled.
+ * @ubi: UBI device description object
+ *
+ * Returns non-zero if the UBI background thread is disabled for testing
+ * purposes.
+ */
+static inline int ubi_dbg_is_bgt_disabled(const struct ubi_device *ubi)
+{
+       return ubi->dbg.disable_bgt;
+}
 
-#ifdef CONFIG_MTD_UBI_DEBUG_EMULATE_BITFLIPS
 /**
  * ubi_dbg_is_bitflip - if it is time to emulate a bit-flip.
+ * @ubi: UBI device description object
  *
  * Returns non-zero if a bit-flip should be emulated, otherwise returns zero.
  */
-static inline int ubi_dbg_is_bitflip(void)
+static inline int ubi_dbg_is_bitflip(const struct ubi_device *ubi)
 {
-       return !(random32() % 200);
+       if (ubi->dbg.emulate_bitflips)
+               return !(prandom_u32() % 200);
+       return 0;
 }
-#else
-#define ubi_dbg_is_bitflip() 0
-#endif
 
-#ifdef CONFIG_MTD_UBI_DEBUG_EMULATE_WRITE_FAILURES
 /**
  * ubi_dbg_is_write_failure - if it is time to emulate a write failure.
+ * @ubi: UBI device description object
  *
  * Returns non-zero if a write failure should be emulated, otherwise returns
  * zero.
  */
-static inline int ubi_dbg_is_write_failure(void)
+static inline int ubi_dbg_is_write_failure(const struct ubi_device *ubi)
 {
-       return !(random32() % 500);
+       if (ubi->dbg.emulate_io_failures)
+               return !(prandom_u32() % 500);
+       return 0;
 }
-#else
-#define ubi_dbg_is_write_failure() 0
-#endif
 
-#ifdef CONFIG_MTD_UBI_DEBUG_EMULATE_ERASE_FAILURES
 /**
  * ubi_dbg_is_erase_failure - if its time to emulate an erase failure.
+ * @ubi: UBI device description object
  *
  * Returns non-zero if an erase failure should be emulated, otherwise returns
  * zero.
  */
-static inline int ubi_dbg_is_erase_failure(void)
+static inline int ubi_dbg_is_erase_failure(const struct ubi_device *ubi)
 {
-               return !(random32() % 400);
+       if (ubi->dbg.emulate_io_failures)
+               return !(prandom_u32() % 400);
+       return 0;
+}
+
+static inline int ubi_dbg_chk_io(const struct ubi_device *ubi)
+{
+       return ubi->dbg.chk_io;
 }
-#else
-#define ubi_dbg_is_erase_failure() 0
-#endif
 
+static inline int ubi_dbg_chk_gen(const struct ubi_device *ubi)
+{
+       return ubi->dbg.chk_gen;
+}
 #endif /* !__UBI_DEBUG_H__ */
index 7d27edaee2a4adc084e0648bed406bd2852064fc..3c2a7e69e1841aa587b6fd6a2e7bf2d4f3f7650b 100644 (file)
@@ -7,20 +7,20 @@
  */
 
 /*
- * The UBI Eraseblock Association (EBA) unit.
+ * The UBI Eraseblock Association (EBA) sub-system.
  *
- * This unit is responsible for I/O to/from logical eraseblock.
+ * This sub-system is responsible for I/O to/from logical eraseblock.
  *
  * Although in this implementation the EBA table is fully kept and managed in
  * RAM, which assumes poor scalability, it might be (partially) maintained on
  * flash in future implementations.
  *
- * The EBA unit implements per-logical eraseblock locking. Before accessing a
- * logical eraseblock it is locked for reading or writing. The per-logical
- * eraseblock locking is implemented by means of the lock tree. The lock tree
- * is an RB-tree which refers all the currently locked logical eraseblocks. The
- * lock tree elements are &struct ubi_ltree_entry objects. They are indexed by
- * (@vol_id, @lnum) pairs.
+ * The EBA sub-system implements per-logical eraseblock locking. Before
+ * accessing a logical eraseblock it is locked for reading or writing. The
+ * per-logical eraseblock locking is implemented by means of the lock tree. The
+ * lock tree is an RB-tree which refers all the currently locked logical
+ * eraseblocks. The lock tree elements are &struct ubi_ltree_entry objects.
+ * They are indexed by (@vol_id, @lnum) pairs.
  *
  * EBA also maintains the global sequence counter which is incremented each
  * time a logical eraseblock is mapped to a physical eraseblock and it is
  * 64 bits is enough to never overflow.
  */
 
-#ifdef UBI_LINUX
+#define __UBOOT__
+#ifndef __UBOOT__
 #include <linux/slab.h>
 #include <linux/crc32.h>
-#include <linux/err.h>
+#else
+#include <ubi_uboot.h>
 #endif
 
-#include <ubi_uboot.h>
+#include <linux/err.h>
 #include "ubi.h"
 
 /* Number of physical eraseblocks reserved for atomic LEB change operation */
@@ -49,7 +51,7 @@
  * global sequence counter value. It also increases the global sequence
  * counter.
  */
-static unsigned long long next_sqnum(struct ubi_device *ubi)
+unsigned long long ubi_next_sqnum(struct ubi_device *ubi)
 {
        unsigned long long sqnum;
 
@@ -181,9 +183,7 @@ static struct ubi_ltree_entry *ltree_add_entry(struct ubi_device *ubi,
        le->users += 1;
        spin_unlock(&ubi->ltree_lock);
 
-       if (le_free)
-               kfree(le_free);
-
+       kfree(le_free);
        return le;
 }
 
@@ -215,22 +215,18 @@ static int leb_read_lock(struct ubi_device *ubi, int vol_id, int lnum)
  */
 static void leb_read_unlock(struct ubi_device *ubi, int vol_id, int lnum)
 {
-       int _free = 0;
        struct ubi_ltree_entry *le;
 
        spin_lock(&ubi->ltree_lock);
        le = ltree_lookup(ubi, vol_id, lnum);
        le->users -= 1;
        ubi_assert(le->users >= 0);
+       up_read(&le->mutex);
        if (le->users == 0) {
                rb_erase(&le->rb, &ubi->ltree);
-               _free = 1;
+               kfree(le);
        }
        spin_unlock(&ubi->ltree_lock);
-
-       up_read(&le->mutex);
-       if (_free)
-               kfree(le);
 }
 
 /**
@@ -266,7 +262,6 @@ static int leb_write_lock(struct ubi_device *ubi, int vol_id, int lnum)
  */
 static int leb_write_trylock(struct ubi_device *ubi, int vol_id, int lnum)
 {
-       int _free;
        struct ubi_ltree_entry *le;
 
        le = ltree_add_entry(ubi, vol_id, lnum);
@@ -281,12 +276,9 @@ static int leb_write_trylock(struct ubi_device *ubi, int vol_id, int lnum)
        ubi_assert(le->users >= 0);
        if (le->users == 0) {
                rb_erase(&le->rb, &ubi->ltree);
-               _free = 1;
-       } else
-               _free = 0;
-       spin_unlock(&ubi->ltree_lock);
-       if (_free)
                kfree(le);
+       }
+       spin_unlock(&ubi->ltree_lock);
 
        return 1;
 }
@@ -299,23 +291,18 @@ static int leb_write_trylock(struct ubi_device *ubi, int vol_id, int lnum)
  */
 static void leb_write_unlock(struct ubi_device *ubi, int vol_id, int lnum)
 {
-       int _free;
        struct ubi_ltree_entry *le;
 
        spin_lock(&ubi->ltree_lock);
        le = ltree_lookup(ubi, vol_id, lnum);
        le->users -= 1;
        ubi_assert(le->users >= 0);
+       up_write(&le->mutex);
        if (le->users == 0) {
                rb_erase(&le->rb, &ubi->ltree);
-               _free = 1;
-       } else
-               _free = 0;
-       spin_unlock(&ubi->ltree_lock);
-
-       up_write(&le->mutex);
-       if (_free)
                kfree(le);
+       }
+       spin_unlock(&ubi->ltree_lock);
 }
 
 /**
@@ -347,8 +334,10 @@ int ubi_eba_unmap_leb(struct ubi_device *ubi, struct ubi_volume *vol,
 
        dbg_eba("erase LEB %d:%d, PEB %d", vol_id, lnum, pnum);
 
+       down_read(&ubi->fm_sem);
        vol->eba_tbl[lnum] = UBI_LEB_UNMAPPED;
-       err = ubi_wl_put_peb(ubi, pnum, 0);
+       up_read(&ubi->fm_sem);
+       err = ubi_wl_put_peb(ubi, vol_id, lnum, pnum, 0);
 
 out_unlock:
        leb_write_unlock(ubi, vol_id, lnum);
@@ -425,9 +414,10 @@ retry:
                                 * may try to recover data. FIXME: but this is
                                 * not implemented.
                                 */
-                               if (err == UBI_IO_BAD_VID_HDR) {
-                                       ubi_warn("bad VID header at PEB %d, LEB"
-                                                "%d:%d", pnum, vol_id, lnum);
+                               if (err == UBI_IO_BAD_HDR_EBADMSG ||
+                                   err == UBI_IO_BAD_HDR) {
+                                       ubi_warn("corrupted VID header at PEB %d, LEB %d:%d",
+                                                pnum, vol_id, lnum);
                                        err = -EBADMSG;
                                } else
                                        ubi_ro_mode(ubi);
@@ -508,16 +498,12 @@ static int recover_peb(struct ubi_device *ubi, int pnum, int vol_id, int lnum,
        struct ubi_vid_hdr *vid_hdr;
 
        vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS);
-       if (!vid_hdr) {
+       if (!vid_hdr)
                return -ENOMEM;
-       }
-
-       mutex_lock(&ubi->buf_mutex);
 
 retry:
-       new_pnum = ubi_wl_get_peb(ubi, UBI_UNKNOWN);
+       new_pnum = ubi_wl_get_peb(ubi);
        if (new_pnum < 0) {
-               mutex_unlock(&ubi->buf_mutex);
                ubi_free_vid_hdr(ubi, vid_hdr);
                return new_pnum;
        }
@@ -531,39 +517,45 @@ retry:
                goto out_put;
        }
 
-       vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi));
+       vid_hdr->sqnum = cpu_to_be64(ubi_next_sqnum(ubi));
        err = ubi_io_write_vid_hdr(ubi, new_pnum, vid_hdr);
        if (err)
                goto write_error;
 
        data_size = offset + len;
-       memset(ubi->peb_buf1 + offset, 0xFF, len);
+       mutex_lock(&ubi->buf_mutex);
+       memset(ubi->peb_buf + offset, 0xFF, len);
 
        /* Read everything before the area where the write failure happened */
        if (offset > 0) {
-               err = ubi_io_read_data(ubi, ubi->peb_buf1, pnum, 0, offset);
+               err = ubi_io_read_data(ubi, ubi->peb_buf, pnum, 0, offset);
                if (err && err != UBI_IO_BITFLIPS)
-                       goto out_put;
+                       goto out_unlock;
        }
 
-       memcpy(ubi->peb_buf1 + offset, buf, len);
+       memcpy(ubi->peb_buf + offset, buf, len);
 
-       err = ubi_io_write_data(ubi, ubi->peb_buf1, new_pnum, 0, data_size);
-       if (err)
+       err = ubi_io_write_data(ubi, ubi->peb_buf, new_pnum, 0, data_size);
+       if (err) {
+               mutex_unlock(&ubi->buf_mutex);
                goto write_error;
+       }
 
        mutex_unlock(&ubi->buf_mutex);
        ubi_free_vid_hdr(ubi, vid_hdr);
 
+       down_read(&ubi->fm_sem);
        vol->eba_tbl[lnum] = new_pnum;
-       ubi_wl_put_peb(ubi, pnum, 1);
+       up_read(&ubi->fm_sem);
+       ubi_wl_put_peb(ubi, vol_id, lnum, pnum, 1);
 
        ubi_msg("data was successfully recovered");
        return 0;
 
-out_put:
+out_unlock:
        mutex_unlock(&ubi->buf_mutex);
-       ubi_wl_put_peb(ubi, new_pnum, 1);
+out_put:
+       ubi_wl_put_peb(ubi, vol_id, lnum, new_pnum, 1);
        ubi_free_vid_hdr(ubi, vid_hdr);
        return err;
 
@@ -573,9 +565,8 @@ write_error:
         * get another one.
         */
        ubi_warn("failed to write to PEB %d", new_pnum);
-       ubi_wl_put_peb(ubi, new_pnum, 1);
+       ubi_wl_put_peb(ubi, vol_id, lnum, new_pnum, 1);
        if (++tries > UBI_IO_RETRIES) {
-               mutex_unlock(&ubi->buf_mutex);
                ubi_free_vid_hdr(ubi, vid_hdr);
                return err;
        }
@@ -591,7 +582,6 @@ write_error:
  * @buf: the data to write
  * @offset: offset within the logical eraseblock where to write
  * @len: how many bytes to write
- * @dtype: data type
  *
  * This function writes data to logical eraseblock @lnum of a dynamic volume
  * @vol. Returns zero in case of success and a negative error code in case
@@ -599,7 +589,7 @@ write_error:
  * written to the flash media, but may be some garbage.
  */
 int ubi_eba_write_leb(struct ubi_device *ubi, struct ubi_volume *vol, int lnum,
-                     const void *buf, int offset, int len, int dtype)
+                     const void *buf, int offset, int len)
 {
        int err, pnum, tries = 0, vol_id = vol->vol_id;
        struct ubi_vid_hdr *vid_hdr;
@@ -640,14 +630,14 @@ int ubi_eba_write_leb(struct ubi_device *ubi, struct ubi_volume *vol, int lnum,
        }
 
        vid_hdr->vol_type = UBI_VID_DYNAMIC;
-       vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi));
+       vid_hdr->sqnum = cpu_to_be64(ubi_next_sqnum(ubi));
        vid_hdr->vol_id = cpu_to_be32(vol_id);
        vid_hdr->lnum = cpu_to_be32(lnum);
        vid_hdr->compat = ubi_get_compat(ubi, vol_id);
        vid_hdr->data_pad = cpu_to_be32(vol->data_pad);
 
 retry:
-       pnum = ubi_wl_get_peb(ubi, dtype);
+       pnum = ubi_wl_get_peb(ubi);
        if (pnum < 0) {
                ubi_free_vid_hdr(ubi, vid_hdr);
                leb_write_unlock(ubi, vol_id, lnum);
@@ -667,14 +657,15 @@ retry:
        if (len) {
                err = ubi_io_write_data(ubi, buf, pnum, offset, len);
                if (err) {
-                       ubi_warn("failed to write %d bytes at offset %d of "
-                                "LEB %d:%d, PEB %d", len, offset, vol_id,
-                                lnum, pnum);
+                       ubi_warn("failed to write %d bytes at offset %d of LEB %d:%d, PEB %d",
+                                len, offset, vol_id, lnum, pnum);
                        goto write_error;
                }
        }
 
+       down_read(&ubi->fm_sem);
        vol->eba_tbl[lnum] = pnum;
+       up_read(&ubi->fm_sem);
 
        leb_write_unlock(ubi, vol_id, lnum);
        ubi_free_vid_hdr(ubi, vid_hdr);
@@ -693,7 +684,7 @@ write_error:
         * eraseblock, so just put it and request a new one. We assume that if
         * this physical eraseblock went bad, the erase code will handle that.
         */
-       err = ubi_wl_put_peb(ubi, pnum, 1);
+       err = ubi_wl_put_peb(ubi, vol_id, lnum, pnum, 1);
        if (err || ++tries > UBI_IO_RETRIES) {
                ubi_ro_mode(ubi);
                leb_write_unlock(ubi, vol_id, lnum);
@@ -701,7 +692,7 @@ write_error:
                return err;
        }
 
-       vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi));
+       vid_hdr->sqnum = cpu_to_be64(ubi_next_sqnum(ubi));
        ubi_msg("try another PEB");
        goto retry;
 }
@@ -713,7 +704,6 @@ write_error:
  * @lnum: logical eraseblock number
  * @buf: data to write
  * @len: how many bytes to write
- * @dtype: data type
  * @used_ebs: how many logical eraseblocks will this volume contain
  *
  * This function writes data to logical eraseblock @lnum of static volume
@@ -725,13 +715,12 @@ write_error:
  * to the real data size, although the @buf buffer has to contain the
  * alignment. In all other cases, @len has to be aligned.
  *
- * It is prohibited to write more then once to logical eraseblocks of static
+ * It is prohibited to write more than once to logical eraseblocks of static
  * volumes. This function returns zero in case of success and a negative error
  * code in case of failure.
  */
 int ubi_eba_write_leb_st(struct ubi_device *ubi, struct ubi_volume *vol,
-                        int lnum, const void *buf, int len, int dtype,
-                        int used_ebs)
+                        int lnum, const void *buf, int len, int used_ebs)
 {
        int err, pnum, tries = 0, data_size = len, vol_id = vol->vol_id;
        struct ubi_vid_hdr *vid_hdr;
@@ -756,7 +745,7 @@ int ubi_eba_write_leb_st(struct ubi_device *ubi, struct ubi_volume *vol,
                return err;
        }
 
-       vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi));
+       vid_hdr->sqnum = cpu_to_be64(ubi_next_sqnum(ubi));
        vid_hdr->vol_id = cpu_to_be32(vol_id);
        vid_hdr->lnum = cpu_to_be32(lnum);
        vid_hdr->compat = ubi_get_compat(ubi, vol_id);
@@ -769,7 +758,7 @@ int ubi_eba_write_leb_st(struct ubi_device *ubi, struct ubi_volume *vol,
        vid_hdr->data_crc = cpu_to_be32(crc);
 
 retry:
-       pnum = ubi_wl_get_peb(ubi, dtype);
+       pnum = ubi_wl_get_peb(ubi);
        if (pnum < 0) {
                ubi_free_vid_hdr(ubi, vid_hdr);
                leb_write_unlock(ubi, vol_id, lnum);
@@ -794,7 +783,9 @@ retry:
        }
 
        ubi_assert(vol->eba_tbl[lnum] < 0);
+       down_read(&ubi->fm_sem);
        vol->eba_tbl[lnum] = pnum;
+       up_read(&ubi->fm_sem);
 
        leb_write_unlock(ubi, vol_id, lnum);
        ubi_free_vid_hdr(ubi, vid_hdr);
@@ -813,7 +804,7 @@ write_error:
                return err;
        }
 
-       err = ubi_wl_put_peb(ubi, pnum, 1);
+       err = ubi_wl_put_peb(ubi, vol_id, lnum, pnum, 1);
        if (err || ++tries > UBI_IO_RETRIES) {
                ubi_ro_mode(ubi);
                leb_write_unlock(ubi, vol_id, lnum);
@@ -821,7 +812,7 @@ write_error:
                return err;
        }
 
-       vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi));
+       vid_hdr->sqnum = cpu_to_be64(ubi_next_sqnum(ubi));
        ubi_msg("try another PEB");
        goto retry;
 }
@@ -833,7 +824,6 @@ write_error:
  * @lnum: logical eraseblock number
  * @buf: data to write
  * @len: how many bytes to write
- * @dtype: data type
  *
  * This function changes the contents of a logical eraseblock atomically. @buf
  * has to contain new logical eraseblock data, and @len - the length of the
@@ -845,7 +835,7 @@ write_error:
  * LEB change may be done at a time. This is ensured by @ubi->alc_mutex.
  */
 int ubi_eba_atomic_leb_change(struct ubi_device *ubi, struct ubi_volume *vol,
-                             int lnum, const void *buf, int len, int dtype)
+                             int lnum, const void *buf, int len)
 {
        int err, pnum, tries = 0, vol_id = vol->vol_id;
        struct ubi_vid_hdr *vid_hdr;
@@ -862,7 +852,7 @@ int ubi_eba_atomic_leb_change(struct ubi_device *ubi, struct ubi_volume *vol,
                err = ubi_eba_unmap_leb(ubi, vol, lnum);
                if (err)
                        return err;
-               return ubi_eba_write_leb(ubi, vol, lnum, NULL, 0, 0, dtype);
+               return ubi_eba_write_leb(ubi, vol, lnum, NULL, 0, 0);
        }
 
        vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS);
@@ -874,7 +864,7 @@ int ubi_eba_atomic_leb_change(struct ubi_device *ubi, struct ubi_volume *vol,
        if (err)
                goto out_mutex;
 
-       vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi));
+       vid_hdr->sqnum = cpu_to_be64(ubi_next_sqnum(ubi));
        vid_hdr->vol_id = cpu_to_be32(vol_id);
        vid_hdr->lnum = cpu_to_be32(lnum);
        vid_hdr->compat = ubi_get_compat(ubi, vol_id);
@@ -887,7 +877,7 @@ int ubi_eba_atomic_leb_change(struct ubi_device *ubi, struct ubi_volume *vol,
        vid_hdr->data_crc = cpu_to_be32(crc);
 
 retry:
-       pnum = ubi_wl_get_peb(ubi, dtype);
+       pnum = ubi_wl_get_peb(ubi);
        if (pnum < 0) {
                err = pnum;
                goto out_leb_unlock;
@@ -911,12 +901,14 @@ retry:
        }
 
        if (vol->eba_tbl[lnum] >= 0) {
-               err = ubi_wl_put_peb(ubi, vol->eba_tbl[lnum], 1);
+               err = ubi_wl_put_peb(ubi, vol_id, lnum, vol->eba_tbl[lnum], 0);
                if (err)
                        goto out_leb_unlock;
        }
 
+       down_read(&ubi->fm_sem);
        vol->eba_tbl[lnum] = pnum;
+       up_read(&ubi->fm_sem);
 
 out_leb_unlock:
        leb_write_unlock(ubi, vol_id, lnum);
@@ -936,17 +928,44 @@ write_error:
                goto out_leb_unlock;
        }
 
-       err = ubi_wl_put_peb(ubi, pnum, 1);
+       err = ubi_wl_put_peb(ubi, vol_id, lnum, pnum, 1);
        if (err || ++tries > UBI_IO_RETRIES) {
                ubi_ro_mode(ubi);
                goto out_leb_unlock;
        }
 
-       vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi));
+       vid_hdr->sqnum = cpu_to_be64(ubi_next_sqnum(ubi));
        ubi_msg("try another PEB");
        goto retry;
 }
 
+/**
+ * is_error_sane - check whether a read error is sane.
+ * @err: code of the error happened during reading
+ *
+ * This is a helper function for 'ubi_eba_copy_leb()' which is called when we
+ * cannot read data from the target PEB (an error @err happened). If the error
+ * code is sane, then we treat this error as non-fatal. Otherwise the error is
+ * fatal and UBI will be switched to R/O mode later.
+ *
+ * The idea is that we try not to switch to R/O mode if the read error is
+ * something which suggests there was a real read problem. E.g., %-EIO. Or a
+ * memory allocation failed (-%ENOMEM). Otherwise, it is safer to switch to R/O
+ * mode, simply because we do not know what happened at the MTD level, and we
+ * cannot handle this. E.g., the underlying driver may have become crazy, and
+ * it is safer to switch to R/O mode to preserve the data.
+ *
+ * And bear in mind, this is about reading from the target PEB, i.e. the PEB
+ * which we have just written.
+ */
+static int is_error_sane(int err)
+{
+       if (err == -EIO || err == -ENOMEM || err == UBI_IO_BAD_HDR ||
+           err == UBI_IO_BAD_HDR_EBADMSG || err == -ETIMEDOUT)
+               return 0;
+       return 1;
+}
+
 /**
  * ubi_eba_copy_leb - copy logical eraseblock.
  * @ubi: UBI device description object
@@ -957,10 +976,9 @@ write_error:
  * This function copies logical eraseblock from physical eraseblock @from to
  * physical eraseblock @to. The @vid_hdr buffer may be changed by this
  * function. Returns:
- *   o %0  in case of success;
- *   o %1 if the operation was canceled and should be tried later (e.g.,
- *     because a bit-flip was detected at the target PEB);
- *   o %2 if the volume is being deleted and this LEB should not be moved.
+ *   o %0 in case of success;
+ *   o %MOVE_CANCEL_RACE, %MOVE_TARGET_WR_ERR, %MOVE_TARGET_BITFLIPS, etc;
+ *   o a negative error code in case of failure.
  */
 int ubi_eba_copy_leb(struct ubi_device *ubi, int from, int to,
                     struct ubi_vid_hdr *vid_hdr)
@@ -972,7 +990,7 @@ int ubi_eba_copy_leb(struct ubi_device *ubi, int from, int to,
        vol_id = be32_to_cpu(vid_hdr->vol_id);
        lnum = be32_to_cpu(vid_hdr->lnum);
 
-       dbg_eba("copy LEB %d:%d, PEB %d to PEB %d", vol_id, lnum, from, to);
+       dbg_wl("copy LEB %d:%d, PEB %d to PEB %d", vol_id, lnum, from, to);
 
        if (vid_hdr->vol_type == UBI_VID_STATIC) {
                data_size = be32_to_cpu(vid_hdr->data_size);
@@ -986,17 +1004,16 @@ int ubi_eba_copy_leb(struct ubi_device *ubi, int from, int to,
        /*
         * Note, we may race with volume deletion, which means that the volume
         * this logical eraseblock belongs to might be being deleted. Since the
-        * volume deletion unmaps all the volume's logical eraseblocks, it will
+        * volume deletion un-maps all the volume's logical eraseblocks, it will
         * be locked in 'ubi_wl_put_peb()' and wait for the WL worker to finish.
         */
        vol = ubi->volumes[idx];
+       spin_unlock(&ubi->volumes_lock);
        if (!vol) {
                /* No need to do further work, cancel */
-               dbg_eba("volume %d is being removed, cancel", vol_id);
-               spin_unlock(&ubi->volumes_lock);
-               return 2;
+               dbg_wl("volume %d is being removed, cancel", vol_id);
+               return MOVE_CANCEL_RACE;
        }
-       spin_unlock(&ubi->volumes_lock);
 
        /*
         * We do not want anybody to write to this logical eraseblock while we
@@ -1008,12 +1025,15 @@ int ubi_eba_copy_leb(struct ubi_device *ubi, int from, int to,
         * (@from). This task locks the LEB and goes sleep in the
         * 'ubi_wl_put_peb()' function on the @ubi->move_mutex. In turn, we are
         * holding @ubi->move_mutex and go sleep on the LEB lock. So, if the
-        * LEB is already locked, we just do not move it and return %1.
+        * LEB is already locked, we just do not move it and return
+        * %MOVE_RETRY. Note, we do not return %MOVE_CANCEL_RACE here because
+        * we do not know the reasons of the contention - it may be just a
+        * normal I/O on this LEB, so we want to re-try.
         */
        err = leb_write_trylock(ubi, vol_id, lnum);
        if (err) {
-               dbg_eba("contention on LEB %d:%d, cancel", vol_id, lnum);
-               return err;
+               dbg_wl("contention on LEB %d:%d, cancel", vol_id, lnum);
+               return MOVE_RETRY;
        }
 
        /*
@@ -1022,30 +1042,30 @@ int ubi_eba_copy_leb(struct ubi_device *ubi, int from, int to,
         * cancel it.
         */
        if (vol->eba_tbl[lnum] != from) {
-               dbg_eba("LEB %d:%d is no longer mapped to PEB %d, mapped to "
-                       "PEB %d, cancel", vol_id, lnum, from,
-                       vol->eba_tbl[lnum]);
-               err = 1;
+               dbg_wl("LEB %d:%d is no longer mapped to PEB %d, mapped to PEB %d, cancel",
+                      vol_id, lnum, from, vol->eba_tbl[lnum]);
+               err = MOVE_CANCEL_RACE;
                goto out_unlock_leb;
        }
 
        /*
-        * OK, now the LEB is locked and we can safely start moving iy. Since
-        * this function utilizes thie @ubi->peb1_buf buffer which is shared
-        * with some other functions, so lock the buffer by taking the
+        * OK, now the LEB is locked and we can safely start moving it. Since
+        * this function utilizes the @ubi->peb_buf buffer which is shared
+        * with some other functions - we lock the buffer by taking the
         * @ubi->buf_mutex.
         */
        mutex_lock(&ubi->buf_mutex);
-       dbg_eba("read %d bytes of data", aldata_size);
-       err = ubi_io_read_data(ubi, ubi->peb_buf1, from, 0, aldata_size);
+       dbg_wl("read %d bytes of data", aldata_size);
+       err = ubi_io_read_data(ubi, ubi->peb_buf, from, 0, aldata_size);
        if (err && err != UBI_IO_BITFLIPS) {
                ubi_warn("error %d while reading data from PEB %d",
                         err, from);
+               err = MOVE_SOURCE_RD_ERR;
                goto out_unlock_buf;
        }
 
        /*
-        * Now we have got to calculate how much data we have to to copy. In
+        * Now we have got to calculate how much data we have to copy. In
         * case of a static volume it is fairly easy - the VID header contains
         * the data size. In case of a dynamic volume it is more difficult - we
         * have to read the contents, cut 0xFF bytes from the end and copy only
@@ -1056,14 +1076,14 @@ int ubi_eba_copy_leb(struct ubi_device *ubi, int from, int to,
         */
        if (vid_hdr->vol_type == UBI_VID_DYNAMIC)
                aldata_size = data_size =
-                       ubi_calc_data_len(ubi, ubi->peb_buf1, data_size);
+                       ubi_calc_data_len(ubi, ubi->peb_buf, data_size);
 
        cond_resched();
-       crc = crc32(UBI_CRC32_INIT, ubi->peb_buf1, data_size);
+       crc = crc32(UBI_CRC32_INIT, ubi->peb_buf, data_size);
        cond_resched();
 
        /*
-        * It may turn out to me that the whole @from physical eraseblock
+        * It may turn out to be that the whole @from physical eraseblock
         * contains only 0xFF bytes. Then we have to only write the VID header
         * and do not write any data. This also means we should not set
         * @vid_hdr->copy_flag, @vid_hdr->data_size, and @vid_hdr->data_crc.
@@ -1073,28 +1093,37 @@ int ubi_eba_copy_leb(struct ubi_device *ubi, int from, int to,
                vid_hdr->data_size = cpu_to_be32(data_size);
                vid_hdr->data_crc = cpu_to_be32(crc);
        }
-       vid_hdr->sqnum = cpu_to_be64(next_sqnum(ubi));
+       vid_hdr->sqnum = cpu_to_be64(ubi_next_sqnum(ubi));
 
        err = ubi_io_write_vid_hdr(ubi, to, vid_hdr);
-       if (err)
+       if (err) {
+               if (err == -EIO)
+                       err = MOVE_TARGET_WR_ERR;
                goto out_unlock_buf;
+       }
 
        cond_resched();
 
        /* Read the VID header back and check if it was written correctly */
        err = ubi_io_read_vid_hdr(ubi, to, vid_hdr, 1);
        if (err) {
-               if (err != UBI_IO_BITFLIPS)
-                       ubi_warn("cannot read VID header back from PEB %d", to);
-               else
-                       err = 1;
+               if (err != UBI_IO_BITFLIPS) {
+                       ubi_warn("error %d while reading VID header back from PEB %d",
+                                err, to);
+                       if (is_error_sane(err))
+                               err = MOVE_TARGET_RD_ERR;
+               } else
+                       err = MOVE_TARGET_BITFLIPS;
                goto out_unlock_buf;
        }
 
        if (data_size > 0) {
-               err = ubi_io_write_data(ubi, ubi->peb_buf1, to, 0, aldata_size);
-               if (err)
+               err = ubi_io_write_data(ubi, ubi->peb_buf, to, 0, aldata_size);
+               if (err) {
+                       if (err == -EIO)
+                               err = MOVE_TARGET_WR_ERR;
                        goto out_unlock_buf;
+               }
 
                cond_resched();
 
@@ -1102,28 +1131,33 @@ int ubi_eba_copy_leb(struct ubi_device *ubi, int from, int to,
                 * We've written the data and are going to read it back to make
                 * sure it was written correctly.
                 */
-
-               err = ubi_io_read_data(ubi, ubi->peb_buf2, to, 0, aldata_size);
+               memset(ubi->peb_buf, 0xFF, aldata_size);
+               err = ubi_io_read_data(ubi, ubi->peb_buf, to, 0, aldata_size);
                if (err) {
-                       if (err != UBI_IO_BITFLIPS)
-                               ubi_warn("cannot read data back from PEB %d",
-                                        to);
-                       else
-                               err = 1;
+                       if (err != UBI_IO_BITFLIPS) {
+                               ubi_warn("error %d while reading data back from PEB %d",
+                                        err, to);
+                               if (is_error_sane(err))
+                                       err = MOVE_TARGET_RD_ERR;
+                       } else
+                               err = MOVE_TARGET_BITFLIPS;
                        goto out_unlock_buf;
                }
 
                cond_resched();
 
-               if (memcmp(ubi->peb_buf1, ubi->peb_buf2, aldata_size)) {
-                       ubi_warn("read data back from PEB %d - it is different",
+               if (crc != crc32(UBI_CRC32_INIT, ubi->peb_buf, data_size)) {
+                       ubi_warn("read data back from PEB %d and it is different",
                                 to);
+                       err = -EINVAL;
                        goto out_unlock_buf;
                }
        }
 
        ubi_assert(vol->eba_tbl[lnum] == from);
+       down_read(&ubi->fm_sem);
        vol->eba_tbl[lnum] = to;
+       up_read(&ubi->fm_sem);
 
 out_unlock_buf:
        mutex_unlock(&ubi->buf_mutex);
@@ -1133,28 +1167,165 @@ out_unlock_leb:
 }
 
 /**
- * ubi_eba_init_scan - initialize the EBA unit using scanning information.
+ * print_rsvd_warning - warn about not having enough reserved PEBs.
  * @ubi: UBI device description object
- * @si: scanning information
+ *
+ * This is a helper function for 'ubi_eba_init()' which is called when UBI
+ * cannot reserve enough PEBs for bad block handling. This function makes a
+ * decision whether we have to print a warning or not. The algorithm is as
+ * follows:
+ *   o if this is a new UBI image, then just print the warning
+ *   o if this is an UBI image which has already been used for some time, print
+ *     a warning only if we can reserve less than 10% of the expected amount of
+ *     the reserved PEB.
+ *
+ * The idea is that when UBI is used, PEBs become bad, and the reserved pool
+ * of PEBs becomes smaller, which is normal and we do not want to scare users
+ * with a warning every time they attach the MTD device. This was an issue
+ * reported by real users.
+ */
+static void print_rsvd_warning(struct ubi_device *ubi,
+                              struct ubi_attach_info *ai)
+{
+       /*
+        * The 1 << 18 (256KiB) number is picked randomly, just a reasonably
+        * large number to distinguish between newly flashed and used images.
+        */
+       if (ai->max_sqnum > (1 << 18)) {
+               int min = ubi->beb_rsvd_level / 10;
+
+               if (!min)
+                       min = 1;
+               if (ubi->beb_rsvd_pebs > min)
+                       return;
+       }
+
+       ubi_warn("cannot reserve enough PEBs for bad PEB handling, reserved %d, need %d",
+                ubi->beb_rsvd_pebs, ubi->beb_rsvd_level);
+       if (ubi->corr_peb_count)
+               ubi_warn("%d PEBs are corrupted and not used",
+                        ubi->corr_peb_count);
+}
+
+/**
+ * self_check_eba - run a self check on the EBA table constructed by fastmap.
+ * @ubi: UBI device description object
+ * @ai_fastmap: UBI attach info object created by fastmap
+ * @ai_scan: UBI attach info object created by scanning
+ *
+ * Returns < 0 in case of an internal error, 0 otherwise.
+ * If a bad EBA table entry was found it will be printed out and
+ * ubi_assert() triggers.
+ */
+int self_check_eba(struct ubi_device *ubi, struct ubi_attach_info *ai_fastmap,
+                  struct ubi_attach_info *ai_scan)
+{
+       int i, j, num_volumes, ret = 0;
+       int **scan_eba, **fm_eba;
+       struct ubi_ainf_volume *av;
+       struct ubi_volume *vol;
+       struct ubi_ainf_peb *aeb;
+       struct rb_node *rb;
+
+       num_volumes = ubi->vtbl_slots + UBI_INT_VOL_COUNT;
+
+       scan_eba = kmalloc(sizeof(*scan_eba) * num_volumes, GFP_KERNEL);
+       if (!scan_eba)
+               return -ENOMEM;
+
+       fm_eba = kmalloc(sizeof(*fm_eba) * num_volumes, GFP_KERNEL);
+       if (!fm_eba) {
+               kfree(scan_eba);
+               return -ENOMEM;
+       }
+
+       for (i = 0; i < num_volumes; i++) {
+               vol = ubi->volumes[i];
+               if (!vol)
+                       continue;
+
+               scan_eba[i] = kmalloc(vol->reserved_pebs * sizeof(**scan_eba),
+                                     GFP_KERNEL);
+               if (!scan_eba[i]) {
+                       ret = -ENOMEM;
+                       goto out_free;
+               }
+
+               fm_eba[i] = kmalloc(vol->reserved_pebs * sizeof(**fm_eba),
+                                   GFP_KERNEL);
+               if (!fm_eba[i]) {
+                       ret = -ENOMEM;
+                       goto out_free;
+               }
+
+               for (j = 0; j < vol->reserved_pebs; j++)
+                       scan_eba[i][j] = fm_eba[i][j] = UBI_LEB_UNMAPPED;
+
+               av = ubi_find_av(ai_scan, idx2vol_id(ubi, i));
+               if (!av)
+                       continue;
+
+               ubi_rb_for_each_entry(rb, aeb, &av->root, u.rb)
+                       scan_eba[i][aeb->lnum] = aeb->pnum;
+
+               av = ubi_find_av(ai_fastmap, idx2vol_id(ubi, i));
+               if (!av)
+                       continue;
+
+               ubi_rb_for_each_entry(rb, aeb, &av->root, u.rb)
+                       fm_eba[i][aeb->lnum] = aeb->pnum;
+
+               for (j = 0; j < vol->reserved_pebs; j++) {
+                       if (scan_eba[i][j] != fm_eba[i][j]) {
+                               if (scan_eba[i][j] == UBI_LEB_UNMAPPED ||
+                                       fm_eba[i][j] == UBI_LEB_UNMAPPED)
+                                       continue;
+
+                               ubi_err("LEB:%i:%i is PEB:%i instead of %i!",
+                                       vol->vol_id, i, fm_eba[i][j],
+                                       scan_eba[i][j]);
+                               ubi_assert(0);
+                       }
+               }
+       }
+
+out_free:
+       for (i = 0; i < num_volumes; i++) {
+               if (!ubi->volumes[i])
+                       continue;
+
+               kfree(scan_eba[i]);
+               kfree(fm_eba[i]);
+       }
+
+       kfree(scan_eba);
+       kfree(fm_eba);
+       return ret;
+}
+
+/**
+ * ubi_eba_init - initialize the EBA sub-system using attaching information.
+ * @ubi: UBI device description object
+ * @ai: attaching information
  *
  * This function returns zero in case of success and a negative error code in
  * case of failure.
  */
-int ubi_eba_init_scan(struct ubi_device *ubi, struct ubi_scan_info *si)
+int ubi_eba_init(struct ubi_device *ubi, struct ubi_attach_info *ai)
 {
        int i, j, err, num_volumes;
-       struct ubi_scan_volume *sv;
+       struct ubi_ainf_volume *av;
        struct ubi_volume *vol;
-       struct ubi_scan_leb *seb;
+       struct ubi_ainf_peb *aeb;
        struct rb_node *rb;
 
-       dbg_eba("initialize EBA unit");
+       dbg_eba("initialize EBA sub-system");
 
        spin_lock_init(&ubi->ltree_lock);
        mutex_init(&ubi->alc_mutex);
        ubi->ltree = RB_ROOT;
 
-       ubi->global_sqnum = si->max_sqnum + 1;
+       ubi->global_sqnum = ai->max_sqnum + 1;
        num_volumes = ubi->vtbl_slots + UBI_INT_VOL_COUNT;
 
        for (i = 0; i < num_volumes; i++) {
@@ -1174,24 +1345,27 @@ int ubi_eba_init_scan(struct ubi_device *ubi, struct ubi_scan_info *si)
                for (j = 0; j < vol->reserved_pebs; j++)
                        vol->eba_tbl[j] = UBI_LEB_UNMAPPED;
 
-               sv = ubi_scan_find_sv(si, idx2vol_id(ubi, i));
-               if (!sv)
+               av = ubi_find_av(ai, idx2vol_id(ubi, i));
+               if (!av)
                        continue;
 
-               ubi_rb_for_each_entry(rb, seb, &sv->root, u.rb) {
-                       if (seb->lnum >= vol->reserved_pebs)
+               ubi_rb_for_each_entry(rb, aeb, &av->root, u.rb) {
+                       if (aeb->lnum >= vol->reserved_pebs)
                                /*
                                 * This may happen in case of an unclean reboot
                                 * during re-size.
                                 */
-                               ubi_scan_move_to_list(sv, seb, &si->erase);
-                       vol->eba_tbl[seb->lnum] = seb->pnum;
+                               ubi_move_aeb_to_list(av, aeb, &ai->erase);
+                       vol->eba_tbl[aeb->lnum] = aeb->pnum;
                }
        }
 
        if (ubi->avail_pebs < EBA_RESERVED_PEBS) {
                ubi_err("no enough physical eraseblocks (%d, need %d)",
                        ubi->avail_pebs, EBA_RESERVED_PEBS);
+               if (ubi->corr_peb_count)
+                       ubi_err("%d PEBs are corrupted and not used",
+                               ubi->corr_peb_count);
                err = -ENOSPC;
                goto out_free;
        }
@@ -1204,9 +1378,7 @@ int ubi_eba_init_scan(struct ubi_device *ubi, struct ubi_scan_info *si)
                if (ubi->avail_pebs < ubi->beb_rsvd_level) {
                        /* No enough free physical eraseblocks */
                        ubi->beb_rsvd_pebs = ubi->avail_pebs;
-                       ubi_warn("cannot reserve enough PEBs for bad PEB "
-                                "handling, reserved %d, need %d",
-                                ubi->beb_rsvd_pebs, ubi->beb_rsvd_level);
+                       print_rsvd_warning(ubi, ai);
                } else
                        ubi->beb_rsvd_pebs = ubi->beb_rsvd_level;
 
@@ -1214,7 +1386,7 @@ int ubi_eba_init_scan(struct ubi_device *ubi, struct ubi_scan_info *si)
                ubi->rsvd_pebs  += ubi->beb_rsvd_pebs;
        }
 
-       dbg_eba("EBA unit is initialized");
+       dbg_eba("EBA sub-system is initialized");
        return 0;
 
 out_free:
@@ -1222,23 +1394,7 @@ out_free:
                if (!ubi->volumes[i])
                        continue;
                kfree(ubi->volumes[i]->eba_tbl);
+               ubi->volumes[i]->eba_tbl = NULL;
        }
        return err;
 }
-
-/**
- * ubi_eba_close - close EBA unit.
- * @ubi: UBI device description object
- */
-void ubi_eba_close(const struct ubi_device *ubi)
-{
-       int i, num_volumes = ubi->vtbl_slots + UBI_INT_VOL_COUNT;
-
-       dbg_eba("close EBA unit");
-
-       for (i = 0; i < num_volumes; i++) {
-               if (!ubi->volumes[i])
-                       continue;
-               kfree(ubi->volumes[i]->eba_tbl);
-       }
-}
diff --git a/drivers/mtd/ubi/fastmap.c b/drivers/mtd/ubi/fastmap.c
new file mode 100644 (file)
index 0000000..787522f
--- /dev/null
@@ -0,0 +1,1584 @@
+/*
+ * Copyright (c) 2012 Linutronix GmbH
+ * Author: Richard Weinberger <richard@nod.at>
+ *
+ * SPDX-License-Identifier:    GPL-2.0+
+ *
+ */
+
+#define __UBOOT__
+#ifndef __UBOOT__
+#include <linux/crc32.h>
+#else
+#include <div64.h>
+#include <malloc.h>
+#include <ubi_uboot.h>
+#endif
+
+#include <linux/compat.h>
+#include <linux/math64.h>
+#include "ubi.h"
+
+/**
+ * ubi_calc_fm_size - calculates the fastmap size in bytes for an UBI device.
+ * @ubi: UBI device description object
+ */
+size_t ubi_calc_fm_size(struct ubi_device *ubi)
+{
+       size_t size;
+
+       size = sizeof(struct ubi_fm_hdr) + \
+               sizeof(struct ubi_fm_scan_pool) + \
+               sizeof(struct ubi_fm_scan_pool) + \
+               (ubi->peb_count * sizeof(struct ubi_fm_ec)) + \
+               (sizeof(struct ubi_fm_eba) + \
+               (ubi->peb_count * sizeof(__be32))) + \
+               sizeof(struct ubi_fm_volhdr) * UBI_MAX_VOLUMES;
+       return roundup(size, ubi->leb_size);
+}
+
+
+/**
+ * new_fm_vhdr - allocate a new volume header for fastmap usage.
+ * @ubi: UBI device description object
+ * @vol_id: the VID of the new header
+ *
+ * Returns a new struct ubi_vid_hdr on success.
+ * NULL indicates out of memory.
+ */
+static struct ubi_vid_hdr *new_fm_vhdr(struct ubi_device *ubi, int vol_id)
+{
+       struct ubi_vid_hdr *new;
+
+       new = ubi_zalloc_vid_hdr(ubi, GFP_KERNEL);
+       if (!new)
+               goto out;
+
+       new->vol_type = UBI_VID_DYNAMIC;
+       new->vol_id = cpu_to_be32(vol_id);
+
+       /* UBI implementations without fastmap support have to delete the
+        * fastmap.
+        */
+       new->compat = UBI_COMPAT_DELETE;
+
+out:
+       return new;
+}
+
+/**
+ * add_aeb - create and add a attach erase block to a given list.
+ * @ai: UBI attach info object
+ * @list: the target list
+ * @pnum: PEB number of the new attach erase block
+ * @ec: erease counter of the new LEB
+ * @scrub: scrub this PEB after attaching
+ *
+ * Returns 0 on success, < 0 indicates an internal error.
+ */
+static int add_aeb(struct ubi_attach_info *ai, struct list_head *list,
+                  int pnum, int ec, int scrub)
+{
+       struct ubi_ainf_peb *aeb;
+
+       aeb = kmem_cache_alloc(ai->aeb_slab_cache, GFP_KERNEL);
+       if (!aeb)
+               return -ENOMEM;
+
+       aeb->pnum = pnum;
+       aeb->ec = ec;
+       aeb->lnum = -1;
+       aeb->scrub = scrub;
+       aeb->copy_flag = aeb->sqnum = 0;
+
+       ai->ec_sum += aeb->ec;
+       ai->ec_count++;
+
+       if (ai->max_ec < aeb->ec)
+               ai->max_ec = aeb->ec;
+
+       if (ai->min_ec > aeb->ec)
+               ai->min_ec = aeb->ec;
+
+       list_add_tail(&aeb->u.list, list);
+
+       return 0;
+}
+
+/**
+ * add_vol - create and add a new volume to ubi_attach_info.
+ * @ai: ubi_attach_info object
+ * @vol_id: VID of the new volume
+ * @used_ebs: number of used EBS
+ * @data_pad: data padding value of the new volume
+ * @vol_type: volume type
+ * @last_eb_bytes: number of bytes in the last LEB
+ *
+ * Returns the new struct ubi_ainf_volume on success.
+ * NULL indicates an error.
+ */
+static struct ubi_ainf_volume *add_vol(struct ubi_attach_info *ai, int vol_id,
+                                      int used_ebs, int data_pad, u8 vol_type,
+                                      int last_eb_bytes)
+{
+       struct ubi_ainf_volume *av;
+       struct rb_node **p = &ai->volumes.rb_node, *parent = NULL;
+
+       while (*p) {
+               parent = *p;
+               av = rb_entry(parent, struct ubi_ainf_volume, rb);
+
+               if (vol_id > av->vol_id)
+                       p = &(*p)->rb_left;
+               else if (vol_id > av->vol_id)
+                       p = &(*p)->rb_right;
+       }
+
+       av = kmalloc(sizeof(struct ubi_ainf_volume), GFP_KERNEL);
+       if (!av)
+               goto out;
+
+       av->highest_lnum = av->leb_count = 0;
+       av->vol_id = vol_id;
+       av->used_ebs = used_ebs;
+       av->data_pad = data_pad;
+       av->last_data_size = last_eb_bytes;
+       av->compat = 0;
+       av->vol_type = vol_type;
+       av->root = RB_ROOT;
+
+       dbg_bld("found volume (ID %i)", vol_id);
+
+       rb_link_node(&av->rb, parent, p);
+       rb_insert_color(&av->rb, &ai->volumes);
+
+out:
+       return av;
+}
+
+/**
+ * assign_aeb_to_av - assigns a SEB to a given ainf_volume and removes it
+ * from it's original list.
+ * @ai: ubi_attach_info object
+ * @aeb: the to be assigned SEB
+ * @av: target scan volume
+ */
+static void assign_aeb_to_av(struct ubi_attach_info *ai,
+                            struct ubi_ainf_peb *aeb,
+                            struct ubi_ainf_volume *av)
+{
+       struct ubi_ainf_peb *tmp_aeb;
+       struct rb_node **p = &ai->volumes.rb_node, *parent = NULL;
+
+       p = &av->root.rb_node;
+       while (*p) {
+               parent = *p;
+
+               tmp_aeb = rb_entry(parent, struct ubi_ainf_peb, u.rb);
+               if (aeb->lnum != tmp_aeb->lnum) {
+                       if (aeb->lnum < tmp_aeb->lnum)
+                               p = &(*p)->rb_left;
+                       else
+                               p = &(*p)->rb_right;
+
+                       continue;
+               } else
+                       break;
+       }
+
+       list_del(&aeb->u.list);
+       av->leb_count++;
+
+       rb_link_node(&aeb->u.rb, parent, p);
+       rb_insert_color(&aeb->u.rb, &av->root);
+}
+
+/**
+ * update_vol - inserts or updates a LEB which was found a pool.
+ * @ubi: the UBI device object
+ * @ai: attach info object
+ * @av: the volume this LEB belongs to
+ * @new_vh: the volume header derived from new_aeb
+ * @new_aeb: the AEB to be examined
+ *
+ * Returns 0 on success, < 0 indicates an internal error.
+ */
+static int update_vol(struct ubi_device *ubi, struct ubi_attach_info *ai,
+                     struct ubi_ainf_volume *av, struct ubi_vid_hdr *new_vh,
+                     struct ubi_ainf_peb *new_aeb)
+{
+       struct rb_node **p = &av->root.rb_node, *parent = NULL;
+       struct ubi_ainf_peb *aeb, *victim;
+       int cmp_res;
+
+       while (*p) {
+               parent = *p;
+               aeb = rb_entry(parent, struct ubi_ainf_peb, u.rb);
+
+               if (be32_to_cpu(new_vh->lnum) != aeb->lnum) {
+                       if (be32_to_cpu(new_vh->lnum) < aeb->lnum)
+                               p = &(*p)->rb_left;
+                       else
+                               p = &(*p)->rb_right;
+
+                       continue;
+               }
+
+               /* This case can happen if the fastmap gets written
+                * because of a volume change (creation, deletion, ..).
+                * Then a PEB can be within the persistent EBA and the pool.
+                */
+               if (aeb->pnum == new_aeb->pnum) {
+                       ubi_assert(aeb->lnum == new_aeb->lnum);
+                       kmem_cache_free(ai->aeb_slab_cache, new_aeb);
+
+                       return 0;
+               }
+
+               cmp_res = ubi_compare_lebs(ubi, aeb, new_aeb->pnum, new_vh);
+               if (cmp_res < 0)
+                       return cmp_res;
+
+               /* new_aeb is newer */
+               if (cmp_res & 1) {
+                       victim = kmem_cache_alloc(ai->aeb_slab_cache,
+                               GFP_KERNEL);
+                       if (!victim)
+                               return -ENOMEM;
+
+                       victim->ec = aeb->ec;
+                       victim->pnum = aeb->pnum;
+                       list_add_tail(&victim->u.list, &ai->erase);
+
+                       if (av->highest_lnum == be32_to_cpu(new_vh->lnum))
+                               av->last_data_size = \
+                                       be32_to_cpu(new_vh->data_size);
+
+                       dbg_bld("vol %i: AEB %i's PEB %i is the newer",
+                               av->vol_id, aeb->lnum, new_aeb->pnum);
+
+                       aeb->ec = new_aeb->ec;
+                       aeb->pnum = new_aeb->pnum;
+                       aeb->copy_flag = new_vh->copy_flag;
+                       aeb->scrub = new_aeb->scrub;
+                       kmem_cache_free(ai->aeb_slab_cache, new_aeb);
+
+               /* new_aeb is older */
+               } else {
+                       dbg_bld("vol %i: AEB %i's PEB %i is old, dropping it",
+                               av->vol_id, aeb->lnum, new_aeb->pnum);
+                       list_add_tail(&new_aeb->u.list, &ai->erase);
+               }
+
+               return 0;
+       }
+       /* This LEB is new, let's add it to the volume */
+
+       if (av->highest_lnum <= be32_to_cpu(new_vh->lnum)) {
+               av->highest_lnum = be32_to_cpu(new_vh->lnum);
+               av->last_data_size = be32_to_cpu(new_vh->data_size);
+       }
+
+       if (av->vol_type == UBI_STATIC_VOLUME)
+               av->used_ebs = be32_to_cpu(new_vh->used_ebs);
+
+       av->leb_count++;
+
+       rb_link_node(&new_aeb->u.rb, parent, p);
+       rb_insert_color(&new_aeb->u.rb, &av->root);
+
+       return 0;
+}
+
+/**
+ * process_pool_aeb - we found a non-empty PEB in a pool.
+ * @ubi: UBI device object
+ * @ai: attach info object
+ * @new_vh: the volume header derived from new_aeb
+ * @new_aeb: the AEB to be examined
+ *
+ * Returns 0 on success, < 0 indicates an internal error.
+ */
+static int process_pool_aeb(struct ubi_device *ubi, struct ubi_attach_info *ai,
+                           struct ubi_vid_hdr *new_vh,
+                           struct ubi_ainf_peb *new_aeb)
+{
+       struct ubi_ainf_volume *av, *tmp_av = NULL;
+       struct rb_node **p = &ai->volumes.rb_node, *parent = NULL;
+       int found = 0;
+
+       if (be32_to_cpu(new_vh->vol_id) == UBI_FM_SB_VOLUME_ID ||
+               be32_to_cpu(new_vh->vol_id) == UBI_FM_DATA_VOLUME_ID) {
+               kmem_cache_free(ai->aeb_slab_cache, new_aeb);
+
+               return 0;
+       }
+
+       /* Find the volume this SEB belongs to */
+       while (*p) {
+               parent = *p;
+               tmp_av = rb_entry(parent, struct ubi_ainf_volume, rb);
+
+               if (be32_to_cpu(new_vh->vol_id) > tmp_av->vol_id)
+                       p = &(*p)->rb_left;
+               else if (be32_to_cpu(new_vh->vol_id) < tmp_av->vol_id)
+                       p = &(*p)->rb_right;
+               else {
+                       found = 1;
+                       break;
+               }
+       }
+
+       if (found)
+               av = tmp_av;
+       else {
+               ubi_err("orphaned volume in fastmap pool!");
+               return UBI_BAD_FASTMAP;
+       }
+
+       ubi_assert(be32_to_cpu(new_vh->vol_id) == av->vol_id);
+
+       return update_vol(ubi, ai, av, new_vh, new_aeb);
+}
+
+/**
+ * unmap_peb - unmap a PEB.
+ * If fastmap detects a free PEB in the pool it has to check whether
+ * this PEB has been unmapped after writing the fastmap.
+ *
+ * @ai: UBI attach info object
+ * @pnum: The PEB to be unmapped
+ */
+static void unmap_peb(struct ubi_attach_info *ai, int pnum)
+{
+       struct ubi_ainf_volume *av;
+       struct rb_node *node, *node2;
+       struct ubi_ainf_peb *aeb;
+
+       for (node = rb_first(&ai->volumes); node; node = rb_next(node)) {
+               av = rb_entry(node, struct ubi_ainf_volume, rb);
+
+               for (node2 = rb_first(&av->root); node2;
+                    node2 = rb_next(node2)) {
+                       aeb = rb_entry(node2, struct ubi_ainf_peb, u.rb);
+                       if (aeb->pnum == pnum) {
+                               rb_erase(&aeb->u.rb, &av->root);
+                               kmem_cache_free(ai->aeb_slab_cache, aeb);
+                               return;
+                       }
+               }
+       }
+}
+
+/**
+ * scan_pool - scans a pool for changed (no longer empty PEBs).
+ * @ubi: UBI device object
+ * @ai: attach info object
+ * @pebs: an array of all PEB numbers in the to be scanned pool
+ * @pool_size: size of the pool (number of entries in @pebs)
+ * @max_sqnum: pointer to the maximal sequence number
+ * @eba_orphans: list of PEBs which need to be scanned
+ * @free: list of PEBs which are most likely free (and go into @ai->free)
+ *
+ * Returns 0 on success, if the pool is unusable UBI_BAD_FASTMAP is returned.
+ * < 0 indicates an internal error.
+ */
+#ifndef __UBOOT__
+static int scan_pool(struct ubi_device *ubi, struct ubi_attach_info *ai,
+                    int *pebs, int pool_size, unsigned long long *max_sqnum,
+                    struct list_head *eba_orphans, struct list_head *freef)
+#else
+static int scan_pool(struct ubi_device *ubi, struct ubi_attach_info *ai,
+                    __be32 *pebs, int pool_size, unsigned long long *max_sqnum,
+                    struct list_head *eba_orphans, struct list_head *freef)
+#endif
+{
+       struct ubi_vid_hdr *vh;
+       struct ubi_ec_hdr *ech;
+       struct ubi_ainf_peb *new_aeb, *tmp_aeb;
+       int i, pnum, err, found_orphan, ret = 0;
+
+       ech = kzalloc(ubi->ec_hdr_alsize, GFP_KERNEL);
+       if (!ech)
+               return -ENOMEM;
+
+       vh = ubi_zalloc_vid_hdr(ubi, GFP_KERNEL);
+       if (!vh) {
+               kfree(ech);
+               return -ENOMEM;
+       }
+
+       dbg_bld("scanning fastmap pool: size = %i", pool_size);
+
+       /*
+        * Now scan all PEBs in the pool to find changes which have been made
+        * after the creation of the fastmap
+        */
+       for (i = 0; i < pool_size; i++) {
+               int scrub = 0;
+               int image_seq;
+
+               pnum = be32_to_cpu(pebs[i]);
+
+               if (ubi_io_is_bad(ubi, pnum)) {
+                       ubi_err("bad PEB in fastmap pool!");
+                       ret = UBI_BAD_FASTMAP;
+                       goto out;
+               }
+
+               err = ubi_io_read_ec_hdr(ubi, pnum, ech, 0);
+               if (err && err != UBI_IO_BITFLIPS) {
+                       ubi_err("unable to read EC header! PEB:%i err:%i",
+                               pnum, err);
+                       ret = err > 0 ? UBI_BAD_FASTMAP : err;
+                       goto out;
+               } else if (ret == UBI_IO_BITFLIPS)
+                       scrub = 1;
+
+               /*
+                * Older UBI implementations have image_seq set to zero, so
+                * we shouldn't fail if image_seq == 0.
+                */
+               image_seq = be32_to_cpu(ech->image_seq);
+
+               if (image_seq && (image_seq != ubi->image_seq)) {
+                       ubi_err("bad image seq: 0x%x, expected: 0x%x",
+                               be32_to_cpu(ech->image_seq), ubi->image_seq);
+                       ret = UBI_BAD_FASTMAP;
+                       goto out;
+               }
+
+               err = ubi_io_read_vid_hdr(ubi, pnum, vh, 0);
+               if (err == UBI_IO_FF || err == UBI_IO_FF_BITFLIPS) {
+                       unsigned long long ec = be64_to_cpu(ech->ec);
+                       unmap_peb(ai, pnum);
+                       dbg_bld("Adding PEB to free: %i", pnum);
+                       if (err == UBI_IO_FF_BITFLIPS)
+                               add_aeb(ai, freef, pnum, ec, 1);
+                       else
+                               add_aeb(ai, freef, pnum, ec, 0);
+                       continue;
+               } else if (err == 0 || err == UBI_IO_BITFLIPS) {
+                       dbg_bld("Found non empty PEB:%i in pool", pnum);
+
+                       if (err == UBI_IO_BITFLIPS)
+                               scrub = 1;
+
+                       found_orphan = 0;
+                       list_for_each_entry(tmp_aeb, eba_orphans, u.list) {
+                               if (tmp_aeb->pnum == pnum) {
+                                       found_orphan = 1;
+                                       break;
+                               }
+                       }
+                       if (found_orphan) {
+                               list_del(&tmp_aeb->u.list);
+                               kmem_cache_free(ai->aeb_slab_cache, tmp_aeb);
+                       }
+
+                       new_aeb = kmem_cache_alloc(ai->aeb_slab_cache,
+                                                  GFP_KERNEL);
+                       if (!new_aeb) {
+                               ret = -ENOMEM;
+                               goto out;
+                       }
+
+                       new_aeb->ec = be64_to_cpu(ech->ec);
+                       new_aeb->pnum = pnum;
+                       new_aeb->lnum = be32_to_cpu(vh->lnum);
+                       new_aeb->sqnum = be64_to_cpu(vh->sqnum);
+                       new_aeb->copy_flag = vh->copy_flag;
+                       new_aeb->scrub = scrub;
+
+                       if (*max_sqnum < new_aeb->sqnum)
+                               *max_sqnum = new_aeb->sqnum;
+
+                       err = process_pool_aeb(ubi, ai, vh, new_aeb);
+                       if (err) {
+                               ret = err > 0 ? UBI_BAD_FASTMAP : err;
+                               goto out;
+                       }
+               } else {
+                       /* We are paranoid and fall back to scanning mode */
+                       ubi_err("fastmap pool PEBs contains damaged PEBs!");
+                       ret = err > 0 ? UBI_BAD_FASTMAP : err;
+                       goto out;
+               }
+
+       }
+
+out:
+       ubi_free_vid_hdr(ubi, vh);
+       kfree(ech);
+       return ret;
+}
+
+/**
+ * count_fastmap_pebs - Counts the PEBs found by fastmap.
+ * @ai: The UBI attach info object
+ */
+static int count_fastmap_pebs(struct ubi_attach_info *ai)
+{
+       struct ubi_ainf_peb *aeb;
+       struct ubi_ainf_volume *av;
+       struct rb_node *rb1, *rb2;
+       int n = 0;
+
+       list_for_each_entry(aeb, &ai->erase, u.list)
+               n++;
+
+       list_for_each_entry(aeb, &ai->free, u.list)
+               n++;
+
+        ubi_rb_for_each_entry(rb1, av, &ai->volumes, rb)
+               ubi_rb_for_each_entry(rb2, aeb, &av->root, u.rb)
+                       n++;
+
+       return n;
+}
+
+/**
+ * ubi_attach_fastmap - creates ubi_attach_info from a fastmap.
+ * @ubi: UBI device object
+ * @ai: UBI attach info object
+ * @fm: the fastmap to be attached
+ *
+ * Returns 0 on success, UBI_BAD_FASTMAP if the found fastmap was unusable.
+ * < 0 indicates an internal error.
+ */
+static int ubi_attach_fastmap(struct ubi_device *ubi,
+                             struct ubi_attach_info *ai,
+                             struct ubi_fastmap_layout *fm)
+{
+       struct list_head used, eba_orphans, freef;
+       struct ubi_ainf_volume *av;
+       struct ubi_ainf_peb *aeb, *tmp_aeb, *_tmp_aeb;
+       struct ubi_ec_hdr *ech;
+       struct ubi_fm_sb *fmsb;
+       struct ubi_fm_hdr *fmhdr;
+       struct ubi_fm_scan_pool *fmpl1, *fmpl2;
+       struct ubi_fm_ec *fmec;
+       struct ubi_fm_volhdr *fmvhdr;
+       struct ubi_fm_eba *fm_eba;
+       int ret, i, j, pool_size, wl_pool_size;
+       size_t fm_pos = 0, fm_size = ubi->fm_size;
+       unsigned long long max_sqnum = 0;
+       void *fm_raw = ubi->fm_buf;
+
+       INIT_LIST_HEAD(&used);
+       INIT_LIST_HEAD(&freef);
+       INIT_LIST_HEAD(&eba_orphans);
+       INIT_LIST_HEAD(&ai->corr);
+       INIT_LIST_HEAD(&ai->free);
+       INIT_LIST_HEAD(&ai->erase);
+       INIT_LIST_HEAD(&ai->alien);
+       ai->volumes = RB_ROOT;
+       ai->min_ec = UBI_MAX_ERASECOUNTER;
+
+       ai->aeb_slab_cache = kmem_cache_create("ubi_ainf_peb_slab",
+                                              sizeof(struct ubi_ainf_peb),
+                                              0, 0, NULL);
+       if (!ai->aeb_slab_cache) {
+               ret = -ENOMEM;
+               goto fail;
+       }
+
+       fmsb = (struct ubi_fm_sb *)(fm_raw);
+       ai->max_sqnum = fmsb->sqnum;
+       fm_pos += sizeof(struct ubi_fm_sb);
+       if (fm_pos >= fm_size)
+               goto fail_bad;
+
+       fmhdr = (struct ubi_fm_hdr *)(fm_raw + fm_pos);
+       fm_pos += sizeof(*fmhdr);
+       if (fm_pos >= fm_size)
+               goto fail_bad;
+
+       if (be32_to_cpu(fmhdr->magic) != UBI_FM_HDR_MAGIC) {
+               ubi_err("bad fastmap header magic: 0x%x, expected: 0x%x",
+                       be32_to_cpu(fmhdr->magic), UBI_FM_HDR_MAGIC);
+               goto fail_bad;
+       }
+
+       fmpl1 = (struct ubi_fm_scan_pool *)(fm_raw + fm_pos);
+       fm_pos += sizeof(*fmpl1);
+       if (fm_pos >= fm_size)
+               goto fail_bad;
+       if (be32_to_cpu(fmpl1->magic) != UBI_FM_POOL_MAGIC) {
+               ubi_err("bad fastmap pool magic: 0x%x, expected: 0x%x",
+                       be32_to_cpu(fmpl1->magic), UBI_FM_POOL_MAGIC);
+               goto fail_bad;
+       }
+
+       fmpl2 = (struct ubi_fm_scan_pool *)(fm_raw + fm_pos);
+       fm_pos += sizeof(*fmpl2);
+       if (fm_pos >= fm_size)
+               goto fail_bad;
+       if (be32_to_cpu(fmpl2->magic) != UBI_FM_POOL_MAGIC) {
+               ubi_err("bad fastmap pool magic: 0x%x, expected: 0x%x",
+                       be32_to_cpu(fmpl2->magic), UBI_FM_POOL_MAGIC);
+               goto fail_bad;
+       }
+
+       pool_size = be16_to_cpu(fmpl1->size);
+       wl_pool_size = be16_to_cpu(fmpl2->size);
+       fm->max_pool_size = be16_to_cpu(fmpl1->max_size);
+       fm->max_wl_pool_size = be16_to_cpu(fmpl2->max_size);
+
+       if (pool_size > UBI_FM_MAX_POOL_SIZE || pool_size < 0) {
+               ubi_err("bad pool size: %i", pool_size);
+               goto fail_bad;
+       }
+
+       if (wl_pool_size > UBI_FM_MAX_POOL_SIZE || wl_pool_size < 0) {
+               ubi_err("bad WL pool size: %i", wl_pool_size);
+               goto fail_bad;
+       }
+
+
+       if (fm->max_pool_size > UBI_FM_MAX_POOL_SIZE ||
+           fm->max_pool_size < 0) {
+               ubi_err("bad maximal pool size: %i", fm->max_pool_size);
+               goto fail_bad;
+       }
+
+       if (fm->max_wl_pool_size > UBI_FM_MAX_POOL_SIZE ||
+           fm->max_wl_pool_size < 0) {
+               ubi_err("bad maximal WL pool size: %i", fm->max_wl_pool_size);
+               goto fail_bad;
+       }
+
+       /* read EC values from free list */
+       for (i = 0; i < be32_to_cpu(fmhdr->free_peb_count); i++) {
+               fmec = (struct ubi_fm_ec *)(fm_raw + fm_pos);
+               fm_pos += sizeof(*fmec);
+               if (fm_pos >= fm_size)
+                       goto fail_bad;
+
+               add_aeb(ai, &ai->free, be32_to_cpu(fmec->pnum),
+                       be32_to_cpu(fmec->ec), 0);
+       }
+
+       /* read EC values from used list */
+       for (i = 0; i < be32_to_cpu(fmhdr->used_peb_count); i++) {
+               fmec = (struct ubi_fm_ec *)(fm_raw + fm_pos);
+               fm_pos += sizeof(*fmec);
+               if (fm_pos >= fm_size)
+                       goto fail_bad;
+
+               add_aeb(ai, &used, be32_to_cpu(fmec->pnum),
+                       be32_to_cpu(fmec->ec), 0);
+       }
+
+       /* read EC values from scrub list */
+       for (i = 0; i < be32_to_cpu(fmhdr->scrub_peb_count); i++) {
+               fmec = (struct ubi_fm_ec *)(fm_raw + fm_pos);
+               fm_pos += sizeof(*fmec);
+               if (fm_pos >= fm_size)
+                       goto fail_bad;
+
+               add_aeb(ai, &used, be32_to_cpu(fmec->pnum),
+                       be32_to_cpu(fmec->ec), 1);
+       }
+
+       /* read EC values from erase list */
+       for (i = 0; i < be32_to_cpu(fmhdr->erase_peb_count); i++) {
+               fmec = (struct ubi_fm_ec *)(fm_raw + fm_pos);
+               fm_pos += sizeof(*fmec);
+               if (fm_pos >= fm_size)
+                       goto fail_bad;
+
+               add_aeb(ai, &ai->erase, be32_to_cpu(fmec->pnum),
+                       be32_to_cpu(fmec->ec), 1);
+       }
+
+       ai->mean_ec = div_u64(ai->ec_sum, ai->ec_count);
+       ai->bad_peb_count = be32_to_cpu(fmhdr->bad_peb_count);
+
+       /* Iterate over all volumes and read their EBA table */
+       for (i = 0; i < be32_to_cpu(fmhdr->vol_count); i++) {
+               fmvhdr = (struct ubi_fm_volhdr *)(fm_raw + fm_pos);
+               fm_pos += sizeof(*fmvhdr);
+               if (fm_pos >= fm_size)
+                       goto fail_bad;
+
+               if (be32_to_cpu(fmvhdr->magic) != UBI_FM_VHDR_MAGIC) {
+                       ubi_err("bad fastmap vol header magic: 0x%x, " \
+                               "expected: 0x%x",
+                               be32_to_cpu(fmvhdr->magic), UBI_FM_VHDR_MAGIC);
+                       goto fail_bad;
+               }
+
+               av = add_vol(ai, be32_to_cpu(fmvhdr->vol_id),
+                            be32_to_cpu(fmvhdr->used_ebs),
+                            be32_to_cpu(fmvhdr->data_pad),
+                            fmvhdr->vol_type,
+                            be32_to_cpu(fmvhdr->last_eb_bytes));
+
+               if (!av)
+                       goto fail_bad;
+
+               ai->vols_found++;
+               if (ai->highest_vol_id < be32_to_cpu(fmvhdr->vol_id))
+                       ai->highest_vol_id = be32_to_cpu(fmvhdr->vol_id);
+
+               fm_eba = (struct ubi_fm_eba *)(fm_raw + fm_pos);
+               fm_pos += sizeof(*fm_eba);
+               fm_pos += (sizeof(__be32) * be32_to_cpu(fm_eba->reserved_pebs));
+               if (fm_pos >= fm_size)
+                       goto fail_bad;
+
+               if (be32_to_cpu(fm_eba->magic) != UBI_FM_EBA_MAGIC) {
+                       ubi_err("bad fastmap EBA header magic: 0x%x, " \
+                               "expected: 0x%x",
+                               be32_to_cpu(fm_eba->magic), UBI_FM_EBA_MAGIC);
+                       goto fail_bad;
+               }
+
+               for (j = 0; j < be32_to_cpu(fm_eba->reserved_pebs); j++) {
+                       int pnum = be32_to_cpu(fm_eba->pnum[j]);
+
+                       if ((int)be32_to_cpu(fm_eba->pnum[j]) < 0)
+                               continue;
+
+                       aeb = NULL;
+                       list_for_each_entry(tmp_aeb, &used, u.list) {
+                               if (tmp_aeb->pnum == pnum) {
+                                       aeb = tmp_aeb;
+                                       break;
+                               }
+                       }
+
+                       /* This can happen if a PEB is already in an EBA known
+                        * by this fastmap but the PEB itself is not in the used
+                        * list.
+                        * In this case the PEB can be within the fastmap pool
+                        * or while writing the fastmap it was in the protection
+                        * queue.
+                        */
+                       if (!aeb) {
+                               aeb = kmem_cache_alloc(ai->aeb_slab_cache,
+                                                      GFP_KERNEL);
+                               if (!aeb) {
+                                       ret = -ENOMEM;
+
+                                       goto fail;
+                               }
+
+                               aeb->lnum = j;
+                               aeb->pnum = be32_to_cpu(fm_eba->pnum[j]);
+                               aeb->ec = -1;
+                               aeb->scrub = aeb->copy_flag = aeb->sqnum = 0;
+                               list_add_tail(&aeb->u.list, &eba_orphans);
+                               continue;
+                       }
+
+                       aeb->lnum = j;
+
+                       if (av->highest_lnum <= aeb->lnum)
+                               av->highest_lnum = aeb->lnum;
+
+                       assign_aeb_to_av(ai, aeb, av);
+
+                       dbg_bld("inserting PEB:%i (LEB %i) to vol %i",
+                               aeb->pnum, aeb->lnum, av->vol_id);
+               }
+
+               ech = kzalloc(ubi->ec_hdr_alsize, GFP_KERNEL);
+               if (!ech) {
+                       ret = -ENOMEM;
+                       goto fail;
+               }
+
+               list_for_each_entry_safe(tmp_aeb, _tmp_aeb, &eba_orphans,
+                                        u.list) {
+                       int err;
+
+                       if (ubi_io_is_bad(ubi, tmp_aeb->pnum)) {
+                               ubi_err("bad PEB in fastmap EBA orphan list");
+                               ret = UBI_BAD_FASTMAP;
+                               kfree(ech);
+                               goto fail;
+                       }
+
+                       err = ubi_io_read_ec_hdr(ubi, tmp_aeb->pnum, ech, 0);
+                       if (err && err != UBI_IO_BITFLIPS) {
+                               ubi_err("unable to read EC header! PEB:%i " \
+                                       "err:%i", tmp_aeb->pnum, err);
+                               ret = err > 0 ? UBI_BAD_FASTMAP : err;
+                               kfree(ech);
+
+                               goto fail;
+                       } else if (err == UBI_IO_BITFLIPS)
+                               tmp_aeb->scrub = 1;
+
+                       tmp_aeb->ec = be64_to_cpu(ech->ec);
+                       assign_aeb_to_av(ai, tmp_aeb, av);
+               }
+
+               kfree(ech);
+       }
+
+       ret = scan_pool(ubi, ai, fmpl1->pebs, pool_size, &max_sqnum,
+                       &eba_orphans, &freef);
+       if (ret)
+               goto fail;
+
+       ret = scan_pool(ubi, ai, fmpl2->pebs, wl_pool_size, &max_sqnum,
+                       &eba_orphans, &freef);
+       if (ret)
+               goto fail;
+
+       if (max_sqnum > ai->max_sqnum)
+               ai->max_sqnum = max_sqnum;
+
+       list_for_each_entry_safe(tmp_aeb, _tmp_aeb, &freef, u.list)
+               list_move_tail(&tmp_aeb->u.list, &ai->free);
+
+       ubi_assert(list_empty(&used));
+       ubi_assert(list_empty(&eba_orphans));
+       ubi_assert(list_empty(&freef));
+
+       /*
+        * If fastmap is leaking PEBs (must not happen), raise a
+        * fat warning and fall back to scanning mode.
+        * We do this here because in ubi_wl_init() it's too late
+        * and we cannot fall back to scanning.
+        */
+#ifndef __UBOOT__
+       if (WARN_ON(count_fastmap_pebs(ai) != ubi->peb_count -
+                   ai->bad_peb_count - fm->used_blocks))
+               goto fail_bad;
+#else
+       if (count_fastmap_pebs(ai) != ubi->peb_count -
+                   ai->bad_peb_count - fm->used_blocks) {
+               WARN_ON(1);
+               goto fail_bad;
+       }
+#endif
+
+       return 0;
+
+fail_bad:
+       ret = UBI_BAD_FASTMAP;
+fail:
+       list_for_each_entry_safe(tmp_aeb, _tmp_aeb, &used, u.list) {
+               list_del(&tmp_aeb->u.list);
+               kmem_cache_free(ai->aeb_slab_cache, tmp_aeb);
+       }
+       list_for_each_entry_safe(tmp_aeb, _tmp_aeb, &eba_orphans, u.list) {
+               list_del(&tmp_aeb->u.list);
+               kmem_cache_free(ai->aeb_slab_cache, tmp_aeb);
+       }
+       list_for_each_entry_safe(tmp_aeb, _tmp_aeb, &freef, u.list) {
+               list_del(&tmp_aeb->u.list);
+               kmem_cache_free(ai->aeb_slab_cache, tmp_aeb);
+       }
+
+       return ret;
+}
+
+/**
+ * ubi_scan_fastmap - scan the fastmap.
+ * @ubi: UBI device object
+ * @ai: UBI attach info to be filled
+ * @fm_anchor: The fastmap starts at this PEB
+ *
+ * Returns 0 on success, UBI_NO_FASTMAP if no fastmap was found,
+ * UBI_BAD_FASTMAP if one was found but is not usable.
+ * < 0 indicates an internal error.
+ */
+int ubi_scan_fastmap(struct ubi_device *ubi, struct ubi_attach_info *ai,
+                    int fm_anchor)
+{
+       struct ubi_fm_sb *fmsb, *fmsb2;
+       struct ubi_vid_hdr *vh;
+       struct ubi_ec_hdr *ech;
+       struct ubi_fastmap_layout *fm;
+       int i, used_blocks, pnum, ret = 0;
+       size_t fm_size;
+       __be32 crc, tmp_crc;
+       unsigned long long sqnum = 0;
+
+       mutex_lock(&ubi->fm_mutex);
+       memset(ubi->fm_buf, 0, ubi->fm_size);
+
+       fmsb = kmalloc(sizeof(*fmsb), GFP_KERNEL);
+       if (!fmsb) {
+               ret = -ENOMEM;
+               goto out;
+       }
+
+       fm = kzalloc(sizeof(*fm), GFP_KERNEL);
+       if (!fm) {
+               ret = -ENOMEM;
+               kfree(fmsb);
+               goto out;
+       }
+
+       ret = ubi_io_read(ubi, fmsb, fm_anchor, ubi->leb_start, sizeof(*fmsb));
+       if (ret && ret != UBI_IO_BITFLIPS)
+               goto free_fm_sb;
+       else if (ret == UBI_IO_BITFLIPS)
+               fm->to_be_tortured[0] = 1;
+
+       if (be32_to_cpu(fmsb->magic) != UBI_FM_SB_MAGIC) {
+               ubi_err("bad super block magic: 0x%x, expected: 0x%x",
+                       be32_to_cpu(fmsb->magic), UBI_FM_SB_MAGIC);
+               ret = UBI_BAD_FASTMAP;
+               goto free_fm_sb;
+       }
+
+       if (fmsb->version != UBI_FM_FMT_VERSION) {
+               ubi_err("bad fastmap version: %i, expected: %i",
+                       fmsb->version, UBI_FM_FMT_VERSION);
+               ret = UBI_BAD_FASTMAP;
+               goto free_fm_sb;
+       }
+
+       used_blocks = be32_to_cpu(fmsb->used_blocks);
+       if (used_blocks > UBI_FM_MAX_BLOCKS || used_blocks < 1) {
+               ubi_err("number of fastmap blocks is invalid: %i", used_blocks);
+               ret = UBI_BAD_FASTMAP;
+               goto free_fm_sb;
+       }
+
+       fm_size = ubi->leb_size * used_blocks;
+       if (fm_size != ubi->fm_size) {
+               ubi_err("bad fastmap size: %zi, expected: %zi", fm_size,
+                       ubi->fm_size);
+               ret = UBI_BAD_FASTMAP;
+               goto free_fm_sb;
+       }
+
+       ech = kzalloc(ubi->ec_hdr_alsize, GFP_KERNEL);
+       if (!ech) {
+               ret = -ENOMEM;
+               goto free_fm_sb;
+       }
+
+       vh = ubi_zalloc_vid_hdr(ubi, GFP_KERNEL);
+       if (!vh) {
+               ret = -ENOMEM;
+               goto free_hdr;
+       }
+
+       for (i = 0; i < used_blocks; i++) {
+               int image_seq;
+
+               pnum = be32_to_cpu(fmsb->block_loc[i]);
+
+               if (ubi_io_is_bad(ubi, pnum)) {
+                       ret = UBI_BAD_FASTMAP;
+                       goto free_hdr;
+               }
+
+               ret = ubi_io_read_ec_hdr(ubi, pnum, ech, 0);
+               if (ret && ret != UBI_IO_BITFLIPS) {
+                       ubi_err("unable to read fastmap block# %i EC (PEB: %i)",
+                               i, pnum);
+                       if (ret > 0)
+                               ret = UBI_BAD_FASTMAP;
+                       goto free_hdr;
+               } else if (ret == UBI_IO_BITFLIPS)
+                       fm->to_be_tortured[i] = 1;
+
+               image_seq = be32_to_cpu(ech->image_seq);
+               if (!ubi->image_seq)
+                       ubi->image_seq = image_seq;
+
+               /*
+                * Older UBI implementations have image_seq set to zero, so
+                * we shouldn't fail if image_seq == 0.
+                */
+               if (image_seq && (image_seq != ubi->image_seq)) {
+                       ubi_err("wrong image seq:%d instead of %d",
+                               be32_to_cpu(ech->image_seq), ubi->image_seq);
+                       ret = UBI_BAD_FASTMAP;
+                       goto free_hdr;
+               }
+
+               ret = ubi_io_read_vid_hdr(ubi, pnum, vh, 0);
+               if (ret && ret != UBI_IO_BITFLIPS) {
+                       ubi_err("unable to read fastmap block# %i (PEB: %i)",
+                               i, pnum);
+                       goto free_hdr;
+               }
+
+               if (i == 0) {
+                       if (be32_to_cpu(vh->vol_id) != UBI_FM_SB_VOLUME_ID) {
+                               ubi_err("bad fastmap anchor vol_id: 0x%x," \
+                                       " expected: 0x%x",
+                                       be32_to_cpu(vh->vol_id),
+                                       UBI_FM_SB_VOLUME_ID);
+                               ret = UBI_BAD_FASTMAP;
+                               goto free_hdr;
+                       }
+               } else {
+                       if (be32_to_cpu(vh->vol_id) != UBI_FM_DATA_VOLUME_ID) {
+                               ubi_err("bad fastmap data vol_id: 0x%x," \
+                                       " expected: 0x%x",
+                                       be32_to_cpu(vh->vol_id),
+                                       UBI_FM_DATA_VOLUME_ID);
+                               ret = UBI_BAD_FASTMAP;
+                               goto free_hdr;
+                       }
+               }
+
+               if (sqnum < be64_to_cpu(vh->sqnum))
+                       sqnum = be64_to_cpu(vh->sqnum);
+
+               ret = ubi_io_read(ubi, ubi->fm_buf + (ubi->leb_size * i), pnum,
+                                 ubi->leb_start, ubi->leb_size);
+               if (ret && ret != UBI_IO_BITFLIPS) {
+                       ubi_err("unable to read fastmap block# %i (PEB: %i, " \
+                               "err: %i)", i, pnum, ret);
+                       goto free_hdr;
+               }
+       }
+
+       kfree(fmsb);
+       fmsb = NULL;
+
+       fmsb2 = (struct ubi_fm_sb *)(ubi->fm_buf);
+       tmp_crc = be32_to_cpu(fmsb2->data_crc);
+       fmsb2->data_crc = 0;
+       crc = crc32(UBI_CRC32_INIT, ubi->fm_buf, fm_size);
+       if (crc != tmp_crc) {
+               ubi_err("fastmap data CRC is invalid");
+               ubi_err("CRC should be: 0x%x, calc: 0x%x", tmp_crc, crc);
+               ret = UBI_BAD_FASTMAP;
+               goto free_hdr;
+       }
+
+       fmsb2->sqnum = sqnum;
+
+       fm->used_blocks = used_blocks;
+
+       ret = ubi_attach_fastmap(ubi, ai, fm);
+       if (ret) {
+               if (ret > 0)
+                       ret = UBI_BAD_FASTMAP;
+               goto free_hdr;
+       }
+
+       for (i = 0; i < used_blocks; i++) {
+               struct ubi_wl_entry *e;
+
+               e = kmem_cache_alloc(ubi_wl_entry_slab, GFP_KERNEL);
+               if (!e) {
+                       while (i--)
+                               kfree(fm->e[i]);
+
+                       ret = -ENOMEM;
+                       goto free_hdr;
+               }
+
+               e->pnum = be32_to_cpu(fmsb2->block_loc[i]);
+               e->ec = be32_to_cpu(fmsb2->block_ec[i]);
+               fm->e[i] = e;
+       }
+
+       ubi->fm = fm;
+       ubi->fm_pool.max_size = ubi->fm->max_pool_size;
+       ubi->fm_wl_pool.max_size = ubi->fm->max_wl_pool_size;
+       ubi_msg("attached by fastmap");
+       ubi_msg("fastmap pool size: %d", ubi->fm_pool.max_size);
+       ubi_msg("fastmap WL pool size: %d", ubi->fm_wl_pool.max_size);
+       ubi->fm_disabled = 0;
+
+       ubi_free_vid_hdr(ubi, vh);
+       kfree(ech);
+out:
+       mutex_unlock(&ubi->fm_mutex);
+       if (ret == UBI_BAD_FASTMAP)
+               ubi_err("Attach by fastmap failed, doing a full scan!");
+       return ret;
+
+free_hdr:
+       ubi_free_vid_hdr(ubi, vh);
+       kfree(ech);
+free_fm_sb:
+       kfree(fmsb);
+       kfree(fm);
+       goto out;
+}
+
+/**
+ * ubi_write_fastmap - writes a fastmap.
+ * @ubi: UBI device object
+ * @new_fm: the to be written fastmap
+ *
+ * Returns 0 on success, < 0 indicates an internal error.
+ */
+static int ubi_write_fastmap(struct ubi_device *ubi,
+                            struct ubi_fastmap_layout *new_fm)
+{
+       size_t fm_pos = 0;
+       void *fm_raw;
+       struct ubi_fm_sb *fmsb;
+       struct ubi_fm_hdr *fmh;
+       struct ubi_fm_scan_pool *fmpl1, *fmpl2;
+       struct ubi_fm_ec *fec;
+       struct ubi_fm_volhdr *fvh;
+       struct ubi_fm_eba *feba;
+       struct rb_node *node;
+       struct ubi_wl_entry *wl_e;
+       struct ubi_volume *vol;
+       struct ubi_vid_hdr *avhdr, *dvhdr;
+       struct ubi_work *ubi_wrk;
+       int ret, i, j, free_peb_count, used_peb_count, vol_count;
+       int scrub_peb_count, erase_peb_count;
+
+       fm_raw = ubi->fm_buf;
+       memset(ubi->fm_buf, 0, ubi->fm_size);
+
+       avhdr = new_fm_vhdr(ubi, UBI_FM_SB_VOLUME_ID);
+       if (!avhdr) {
+               ret = -ENOMEM;
+               goto out;
+       }
+
+       dvhdr = new_fm_vhdr(ubi, UBI_FM_DATA_VOLUME_ID);
+       if (!dvhdr) {
+               ret = -ENOMEM;
+               goto out_kfree;
+       }
+
+       spin_lock(&ubi->volumes_lock);
+       spin_lock(&ubi->wl_lock);
+
+       fmsb = (struct ubi_fm_sb *)fm_raw;
+       fm_pos += sizeof(*fmsb);
+       ubi_assert(fm_pos <= ubi->fm_size);
+
+       fmh = (struct ubi_fm_hdr *)(fm_raw + fm_pos);
+       fm_pos += sizeof(*fmh);
+       ubi_assert(fm_pos <= ubi->fm_size);
+
+       fmsb->magic = cpu_to_be32(UBI_FM_SB_MAGIC);
+       fmsb->version = UBI_FM_FMT_VERSION;
+       fmsb->used_blocks = cpu_to_be32(new_fm->used_blocks);
+       /* the max sqnum will be filled in while *reading* the fastmap */
+       fmsb->sqnum = 0;
+
+       fmh->magic = cpu_to_be32(UBI_FM_HDR_MAGIC);
+       free_peb_count = 0;
+       used_peb_count = 0;
+       scrub_peb_count = 0;
+       erase_peb_count = 0;
+       vol_count = 0;
+
+       fmpl1 = (struct ubi_fm_scan_pool *)(fm_raw + fm_pos);
+       fm_pos += sizeof(*fmpl1);
+       fmpl1->magic = cpu_to_be32(UBI_FM_POOL_MAGIC);
+       fmpl1->size = cpu_to_be16(ubi->fm_pool.size);
+       fmpl1->max_size = cpu_to_be16(ubi->fm_pool.max_size);
+
+       for (i = 0; i < ubi->fm_pool.size; i++)
+               fmpl1->pebs[i] = cpu_to_be32(ubi->fm_pool.pebs[i]);
+
+       fmpl2 = (struct ubi_fm_scan_pool *)(fm_raw + fm_pos);
+       fm_pos += sizeof(*fmpl2);
+       fmpl2->magic = cpu_to_be32(UBI_FM_POOL_MAGIC);
+       fmpl2->size = cpu_to_be16(ubi->fm_wl_pool.size);
+       fmpl2->max_size = cpu_to_be16(ubi->fm_wl_pool.max_size);
+
+       for (i = 0; i < ubi->fm_wl_pool.size; i++)
+               fmpl2->pebs[i] = cpu_to_be32(ubi->fm_wl_pool.pebs[i]);
+
+       for (node = rb_first(&ubi->free); node; node = rb_next(node)) {
+               wl_e = rb_entry(node, struct ubi_wl_entry, u.rb);
+               fec = (struct ubi_fm_ec *)(fm_raw + fm_pos);
+
+               fec->pnum = cpu_to_be32(wl_e->pnum);
+               fec->ec = cpu_to_be32(wl_e->ec);
+
+               free_peb_count++;
+               fm_pos += sizeof(*fec);
+               ubi_assert(fm_pos <= ubi->fm_size);
+       }
+       fmh->free_peb_count = cpu_to_be32(free_peb_count);
+
+       for (node = rb_first(&ubi->used); node; node = rb_next(node)) {
+               wl_e = rb_entry(node, struct ubi_wl_entry, u.rb);
+               fec = (struct ubi_fm_ec *)(fm_raw + fm_pos);
+
+               fec->pnum = cpu_to_be32(wl_e->pnum);
+               fec->ec = cpu_to_be32(wl_e->ec);
+
+               used_peb_count++;
+               fm_pos += sizeof(*fec);
+               ubi_assert(fm_pos <= ubi->fm_size);
+       }
+       fmh->used_peb_count = cpu_to_be32(used_peb_count);
+
+       for (node = rb_first(&ubi->scrub); node; node = rb_next(node)) {
+               wl_e = rb_entry(node, struct ubi_wl_entry, u.rb);
+               fec = (struct ubi_fm_ec *)(fm_raw + fm_pos);
+
+               fec->pnum = cpu_to_be32(wl_e->pnum);
+               fec->ec = cpu_to_be32(wl_e->ec);
+
+               scrub_peb_count++;
+               fm_pos += sizeof(*fec);
+               ubi_assert(fm_pos <= ubi->fm_size);
+       }
+       fmh->scrub_peb_count = cpu_to_be32(scrub_peb_count);
+
+
+       list_for_each_entry(ubi_wrk, &ubi->works, list) {
+               if (ubi_is_erase_work(ubi_wrk)) {
+                       wl_e = ubi_wrk->e;
+                       ubi_assert(wl_e);
+
+                       fec = (struct ubi_fm_ec *)(fm_raw + fm_pos);
+
+                       fec->pnum = cpu_to_be32(wl_e->pnum);
+                       fec->ec = cpu_to_be32(wl_e->ec);
+
+                       erase_peb_count++;
+                       fm_pos += sizeof(*fec);
+                       ubi_assert(fm_pos <= ubi->fm_size);
+               }
+       }
+       fmh->erase_peb_count = cpu_to_be32(erase_peb_count);
+
+       for (i = 0; i < UBI_MAX_VOLUMES + UBI_INT_VOL_COUNT; i++) {
+               vol = ubi->volumes[i];
+
+               if (!vol)
+                       continue;
+
+               vol_count++;
+
+               fvh = (struct ubi_fm_volhdr *)(fm_raw + fm_pos);
+               fm_pos += sizeof(*fvh);
+               ubi_assert(fm_pos <= ubi->fm_size);
+
+               fvh->magic = cpu_to_be32(UBI_FM_VHDR_MAGIC);
+               fvh->vol_id = cpu_to_be32(vol->vol_id);
+               fvh->vol_type = vol->vol_type;
+               fvh->used_ebs = cpu_to_be32(vol->used_ebs);
+               fvh->data_pad = cpu_to_be32(vol->data_pad);
+               fvh->last_eb_bytes = cpu_to_be32(vol->last_eb_bytes);
+
+               ubi_assert(vol->vol_type == UBI_DYNAMIC_VOLUME ||
+                       vol->vol_type == UBI_STATIC_VOLUME);
+
+               feba = (struct ubi_fm_eba *)(fm_raw + fm_pos);
+               fm_pos += sizeof(*feba) + (sizeof(__be32) * vol->reserved_pebs);
+               ubi_assert(fm_pos <= ubi->fm_size);
+
+               for (j = 0; j < vol->reserved_pebs; j++)
+                       feba->pnum[j] = cpu_to_be32(vol->eba_tbl[j]);
+
+               feba->reserved_pebs = cpu_to_be32(j);
+               feba->magic = cpu_to_be32(UBI_FM_EBA_MAGIC);
+       }
+       fmh->vol_count = cpu_to_be32(vol_count);
+       fmh->bad_peb_count = cpu_to_be32(ubi->bad_peb_count);
+
+       avhdr->sqnum = cpu_to_be64(ubi_next_sqnum(ubi));
+       avhdr->lnum = 0;
+
+       spin_unlock(&ubi->wl_lock);
+       spin_unlock(&ubi->volumes_lock);
+
+       dbg_bld("writing fastmap SB to PEB %i", new_fm->e[0]->pnum);
+       ret = ubi_io_write_vid_hdr(ubi, new_fm->e[0]->pnum, avhdr);
+       if (ret) {
+               ubi_err("unable to write vid_hdr to fastmap SB!");
+               goto out_kfree;
+       }
+
+       for (i = 0; i < new_fm->used_blocks; i++) {
+               fmsb->block_loc[i] = cpu_to_be32(new_fm->e[i]->pnum);
+               fmsb->block_ec[i] = cpu_to_be32(new_fm->e[i]->ec);
+       }
+
+       fmsb->data_crc = 0;
+       fmsb->data_crc = cpu_to_be32(crc32(UBI_CRC32_INIT, fm_raw,
+                                          ubi->fm_size));
+
+       for (i = 1; i < new_fm->used_blocks; i++) {
+               dvhdr->sqnum = cpu_to_be64(ubi_next_sqnum(ubi));
+               dvhdr->lnum = cpu_to_be32(i);
+               dbg_bld("writing fastmap data to PEB %i sqnum %llu",
+                       new_fm->e[i]->pnum, be64_to_cpu(dvhdr->sqnum));
+               ret = ubi_io_write_vid_hdr(ubi, new_fm->e[i]->pnum, dvhdr);
+               if (ret) {
+                       ubi_err("unable to write vid_hdr to PEB %i!",
+                               new_fm->e[i]->pnum);
+                       goto out_kfree;
+               }
+       }
+
+       for (i = 0; i < new_fm->used_blocks; i++) {
+               ret = ubi_io_write(ubi, fm_raw + (i * ubi->leb_size),
+                       new_fm->e[i]->pnum, ubi->leb_start, ubi->leb_size);
+               if (ret) {
+                       ubi_err("unable to write fastmap to PEB %i!",
+                               new_fm->e[i]->pnum);
+                       goto out_kfree;
+               }
+       }
+
+       ubi_assert(new_fm);
+       ubi->fm = new_fm;
+
+       dbg_bld("fastmap written!");
+
+out_kfree:
+       ubi_free_vid_hdr(ubi, avhdr);
+       ubi_free_vid_hdr(ubi, dvhdr);
+out:
+       return ret;
+}
+
+/**
+ * erase_block - Manually erase a PEB.
+ * @ubi: UBI device object
+ * @pnum: PEB to be erased
+ *
+ * Returns the new EC value on success, < 0 indicates an internal error.
+ */
+static int erase_block(struct ubi_device *ubi, int pnum)
+{
+       int ret;
+       struct ubi_ec_hdr *ec_hdr;
+       long long ec;
+
+       ec_hdr = kzalloc(ubi->ec_hdr_alsize, GFP_KERNEL);
+       if (!ec_hdr)
+               return -ENOMEM;
+
+       ret = ubi_io_read_ec_hdr(ubi, pnum, ec_hdr, 0);
+       if (ret < 0)
+               goto out;
+       else if (ret && ret != UBI_IO_BITFLIPS) {
+               ret = -EINVAL;
+               goto out;
+       }
+
+       ret = ubi_io_sync_erase(ubi, pnum, 0);
+       if (ret < 0)
+               goto out;
+
+       ec = be64_to_cpu(ec_hdr->ec);
+       ec += ret;
+       if (ec > UBI_MAX_ERASECOUNTER) {
+               ret = -EINVAL;
+               goto out;
+       }
+
+       ec_hdr->ec = cpu_to_be64(ec);
+       ret = ubi_io_write_ec_hdr(ubi, pnum, ec_hdr);
+       if (ret < 0)
+               goto out;
+
+       ret = ec;
+out:
+       kfree(ec_hdr);
+       return ret;
+}
+
+/**
+ * invalidate_fastmap - destroys a fastmap.
+ * @ubi: UBI device object
+ * @fm: the fastmap to be destroyed
+ *
+ * Returns 0 on success, < 0 indicates an internal error.
+ */
+static int invalidate_fastmap(struct ubi_device *ubi,
+                             struct ubi_fastmap_layout *fm)
+{
+       int ret;
+       struct ubi_vid_hdr *vh;
+
+       ret = erase_block(ubi, fm->e[0]->pnum);
+       if (ret < 0)
+               return ret;
+
+       vh = new_fm_vhdr(ubi, UBI_FM_SB_VOLUME_ID);
+       if (!vh)
+               return -ENOMEM;
+
+       /* deleting the current fastmap SB is not enough, an old SB may exist,
+        * so create a (corrupted) SB such that fastmap will find it and fall
+        * back to scanning mode in any case */
+       vh->sqnum = cpu_to_be64(ubi_next_sqnum(ubi));
+       ret = ubi_io_write_vid_hdr(ubi, fm->e[0]->pnum, vh);
+
+       return ret;
+}
+
+/**
+ * ubi_update_fastmap - will be called by UBI if a volume changes or
+ * a fastmap pool becomes full.
+ * @ubi: UBI device object
+ *
+ * Returns 0 on success, < 0 indicates an internal error.
+ */
+int ubi_update_fastmap(struct ubi_device *ubi)
+{
+       int ret, i;
+       struct ubi_fastmap_layout *new_fm, *old_fm;
+       struct ubi_wl_entry *tmp_e;
+
+       mutex_lock(&ubi->fm_mutex);
+
+       ubi_refill_pools(ubi);
+
+       if (ubi->ro_mode || ubi->fm_disabled) {
+               mutex_unlock(&ubi->fm_mutex);
+               return 0;
+       }
+
+       ret = ubi_ensure_anchor_pebs(ubi);
+       if (ret) {
+               mutex_unlock(&ubi->fm_mutex);
+               return ret;
+       }
+
+       new_fm = kzalloc(sizeof(*new_fm), GFP_KERNEL);
+       if (!new_fm) {
+               mutex_unlock(&ubi->fm_mutex);
+               return -ENOMEM;
+       }
+
+       new_fm->used_blocks = ubi->fm_size / ubi->leb_size;
+
+       for (i = 0; i < new_fm->used_blocks; i++) {
+               new_fm->e[i] = kmem_cache_alloc(ubi_wl_entry_slab, GFP_KERNEL);
+               if (!new_fm->e[i]) {
+                       while (i--)
+                               kfree(new_fm->e[i]);
+
+                       kfree(new_fm);
+                       mutex_unlock(&ubi->fm_mutex);
+                       return -ENOMEM;
+               }
+       }
+
+       old_fm = ubi->fm;
+       ubi->fm = NULL;
+
+       if (new_fm->used_blocks > UBI_FM_MAX_BLOCKS) {
+               ubi_err("fastmap too large");
+               ret = -ENOSPC;
+               goto err;
+       }
+
+       for (i = 1; i < new_fm->used_blocks; i++) {
+               spin_lock(&ubi->wl_lock);
+               tmp_e = ubi_wl_get_fm_peb(ubi, 0);
+               spin_unlock(&ubi->wl_lock);
+
+               if (!tmp_e && !old_fm) {
+                       int j;
+                       ubi_err("could not get any free erase block");
+
+                       for (j = 1; j < i; j++)
+                               ubi_wl_put_fm_peb(ubi, new_fm->e[j], j, 0);
+
+                       ret = -ENOSPC;
+                       goto err;
+               } else if (!tmp_e && old_fm) {
+                       ret = erase_block(ubi, old_fm->e[i]->pnum);
+                       if (ret < 0) {
+                               int j;
+
+                               for (j = 1; j < i; j++)
+                                       ubi_wl_put_fm_peb(ubi, new_fm->e[j],
+                                                         j, 0);
+
+                               ubi_err("could not erase old fastmap PEB");
+                               goto err;
+                       }
+
+                       new_fm->e[i]->pnum = old_fm->e[i]->pnum;
+                       new_fm->e[i]->ec = old_fm->e[i]->ec;
+               } else {
+                       new_fm->e[i]->pnum = tmp_e->pnum;
+                       new_fm->e[i]->ec = tmp_e->ec;
+
+                       if (old_fm)
+                               ubi_wl_put_fm_peb(ubi, old_fm->e[i], i,
+                                                 old_fm->to_be_tortured[i]);
+               }
+       }
+
+       spin_lock(&ubi->wl_lock);
+       tmp_e = ubi_wl_get_fm_peb(ubi, 1);
+       spin_unlock(&ubi->wl_lock);
+
+       if (old_fm) {
+               /* no fresh anchor PEB was found, reuse the old one */
+               if (!tmp_e) {
+                       ret = erase_block(ubi, old_fm->e[0]->pnum);
+                       if (ret < 0) {
+                               int i;
+                               ubi_err("could not erase old anchor PEB");
+
+                               for (i = 1; i < new_fm->used_blocks; i++)
+                                       ubi_wl_put_fm_peb(ubi, new_fm->e[i],
+                                                         i, 0);
+                               goto err;
+                       }
+
+                       new_fm->e[0]->pnum = old_fm->e[0]->pnum;
+                       new_fm->e[0]->ec = ret;
+               } else {
+                       /* we've got a new anchor PEB, return the old one */
+                       ubi_wl_put_fm_peb(ubi, old_fm->e[0], 0,
+                                         old_fm->to_be_tortured[0]);
+
+                       new_fm->e[0]->pnum = tmp_e->pnum;
+                       new_fm->e[0]->ec = tmp_e->ec;
+               }
+       } else {
+               if (!tmp_e) {
+                       int i;
+                       ubi_err("could not find any anchor PEB");
+
+                       for (i = 1; i < new_fm->used_blocks; i++)
+                               ubi_wl_put_fm_peb(ubi, new_fm->e[i], i, 0);
+
+                       ret = -ENOSPC;
+                       goto err;
+               }
+
+               new_fm->e[0]->pnum = tmp_e->pnum;
+               new_fm->e[0]->ec = tmp_e->ec;
+       }
+
+       down_write(&ubi->work_sem);
+       down_write(&ubi->fm_sem);
+       ret = ubi_write_fastmap(ubi, new_fm);
+       up_write(&ubi->fm_sem);
+       up_write(&ubi->work_sem);
+
+       if (ret)
+               goto err;
+
+out_unlock:
+       mutex_unlock(&ubi->fm_mutex);
+       kfree(old_fm);
+       return ret;
+
+err:
+       kfree(new_fm);
+
+       ubi_warn("Unable to write new fastmap, err=%i", ret);
+
+       ret = 0;
+       if (old_fm) {
+               ret = invalidate_fastmap(ubi, old_fm);
+               if (ret < 0)
+                       ubi_err("Unable to invalidiate current fastmap!");
+               else if (ret)
+                       ret = 0;
+       }
+       goto out_unlock;
+}
index 960befc6dda262b9ed22a76cdbc3fc855d3067f3..41d7eb763885ae3c527aa4b89c43614c3730e65b 100644 (file)
@@ -1,22 +1,21 @@
 /*
  * Copyright (c) International Business Machines Corp., 2006
  * Copyright (c) Nokia Corporation, 2006, 2007
- *
  * SPDX-License-Identifier:    GPL-2.0+
  *
  * Author: Artem Bityutskiy (Битюцкий Артём)
  */
 
 /*
- * UBI input/output unit.
+ * UBI input/output sub-system.
  *
- * This unit provides a uniform way to work with all kinds of the underlying
- * MTD devices. It also implements handy functions for reading and writing UBI
- * headers.
+ * This sub-system provides a uniform way to work with all kinds of the
+ * underlying MTD devices. It also implements handy functions for reading and
+ * writing UBI headers.
  *
  * We are trying to have a paranoid mindset and not to trust to what we read
- * from the flash media in order to be more secure and robust. So this unit
- * validates every single header it reads from the flash media.
+ * from the flash media in order to be more secure and robust. So this
+ * sub-system validates every single header it reads from the flash media.
  *
  * Some words about how the eraseblock headers are stored.
  *
@@ -52,9 +51,9 @@
  * device, e.g., make @ubi->min_io_size = 512 in the example above?
  *
  * A: because when writing a sub-page, MTD still writes a full 2K page but the
- * bytes which are no relevant to the sub-page are 0xFF. So, basically, writing
- * 4x512 sub-pages is 4 times slower then writing one 2KiB NAND page. Thus, we
- * prefer to use sub-pages only for EV and VID headers.
+ * bytes which are not relevant to the sub-page are 0xFF. So, basically,
+ * writing 4x512 sub-pages is 4 times slower than writing one 2KiB NAND page.
+ * Thus, we prefer to use sub-pages only for EC and VID headers.
  *
  * As it was noted above, the VID header may start at a non-aligned offset.
  * For example, in case of a 2KiB page NAND flash with a 512 bytes sub-page,
  * 512-byte chunks, we have to allocate one more buffer and copy our VID header
  * to offset 448 of this buffer.
  *
- * The I/O unit does the following trick in order to avoid this extra copy.
- * It always allocates a @ubi->vid_hdr_alsize bytes buffer for the VID header
- * and returns a pointer to offset @ubi->vid_hdr_shift of this buffer. When the
- * VID header is being written out, it shifts the VID header pointer back and
- * writes the whole sub-page.
+ * The I/O sub-system does the following trick in order to avoid this extra
+ * copy. It always allocates a @ubi->vid_hdr_alsize bytes buffer for the VID
+ * header and returns a pointer to offset @ubi->vid_hdr_shift of this buffer.
+ * When the VID header is being written out, it shifts the VID header pointer
+ * back and writes the whole sub-page.
  */
 
-#ifdef UBI_LINUX
+#define __UBOOT__
+#ifndef __UBOOT__
 #include <linux/crc32.h>
 #include <linux/err.h>
+#include <linux/slab.h>
+#else
+#include <ubi_uboot.h>
 #endif
 
-#include <ubi_uboot.h>
 #include "ubi.h"
 
-#ifdef CONFIG_MTD_UBI_DEBUG_PARANOID
-static int paranoid_check_not_bad(const struct ubi_device *ubi, int pnum);
-static int paranoid_check_peb_ec_hdr(const struct ubi_device *ubi, int pnum);
-static int paranoid_check_ec_hdr(const struct ubi_device *ubi, int pnum,
-                                const struct ubi_ec_hdr *ec_hdr);
-static int paranoid_check_peb_vid_hdr(const struct ubi_device *ubi, int pnum);
-static int paranoid_check_vid_hdr(const struct ubi_device *ubi, int pnum,
-                                 const struct ubi_vid_hdr *vid_hdr);
-static int paranoid_check_all_ff(struct ubi_device *ubi, int pnum, int offset,
-                                int len);
-#else
-#define paranoid_check_not_bad(ubi, pnum) 0
-#define paranoid_check_peb_ec_hdr(ubi, pnum)  0
-#define paranoid_check_ec_hdr(ubi, pnum, ec_hdr)  0
-#define paranoid_check_peb_vid_hdr(ubi, pnum) 0
-#define paranoid_check_vid_hdr(ubi, pnum, vid_hdr) 0
-#define paranoid_check_all_ff(ubi, pnum, offset, len) 0
-#endif
+static int self_check_not_bad(const struct ubi_device *ubi, int pnum);
+static int self_check_peb_ec_hdr(const struct ubi_device *ubi, int pnum);
+static int self_check_ec_hdr(const struct ubi_device *ubi, int pnum,
+                            const struct ubi_ec_hdr *ec_hdr);
+static int self_check_peb_vid_hdr(const struct ubi_device *ubi, int pnum);
+static int self_check_vid_hdr(const struct ubi_device *ubi, int pnum,
+                             const struct ubi_vid_hdr *vid_hdr);
+static int self_check_write(struct ubi_device *ubi, const void *buf, int pnum,
+                           int offset, int len);
 
 /**
  * ubi_io_read - read data from a physical eraseblock.
@@ -136,51 +129,77 @@ int ubi_io_read(const struct ubi_device *ubi, void *buf, int pnum, int offset,
        ubi_assert(offset >= 0 && offset + len <= ubi->peb_size);
        ubi_assert(len > 0);
 
-       err = paranoid_check_not_bad(ubi, pnum);
+       err = self_check_not_bad(ubi, pnum);
        if (err)
-               return err > 0 ? -EINVAL : err;
+               return err;
+
+       /*
+        * Deliberately corrupt the buffer to improve robustness. Indeed, if we
+        * do not do this, the following may happen:
+        * 1. The buffer contains data from previous operation, e.g., read from
+        *    another PEB previously. The data looks like expected, e.g., if we
+        *    just do not read anything and return - the caller would not
+        *    notice this. E.g., if we are reading a VID header, the buffer may
+        *    contain a valid VID header from another PEB.
+        * 2. The driver is buggy and returns us success or -EBADMSG or
+        *    -EUCLEAN, but it does not actually put any data to the buffer.
+        *
+        * This may confuse UBI or upper layers - they may think the buffer
+        * contains valid data while in fact it is just old data. This is
+        * especially possible because UBI (and UBIFS) relies on CRC, and
+        * treats data as correct even in case of ECC errors if the CRC is
+        * correct.
+        *
+        * Try to prevent this situation by changing the first byte of the
+        * buffer.
+        */
+       *((uint8_t *)buf) ^= 0xFF;
 
        addr = (loff_t)pnum * ubi->peb_size + offset;
 retry:
        err = mtd_read(ubi->mtd, addr, len, &read, buf);
        if (err) {
-               if (err == -EUCLEAN) {
+               const char *errstr = mtd_is_eccerr(err) ? " (ECC error)" : "";
+
+               if (mtd_is_bitflip(err)) {
                        /*
                         * -EUCLEAN is reported if there was a bit-flip which
                         * was corrected, so this is harmless.
+                        *
+                        * We do not report about it here unless debugging is
+                        * enabled. A corresponding message will be printed
+                        * later, when it is has been scrubbed.
                         */
                        ubi_msg("fixable bit-flip detected at PEB %d", pnum);
                        ubi_assert(len == read);
                        return UBI_IO_BITFLIPS;
                }
 
-               if (read != len && retries++ < UBI_IO_RETRIES) {
-                       dbg_io("error %d while reading %d bytes from PEB %d:%d, "
-                              "read only %zd bytes, retry",
-                              err, len, pnum, offset, read);
+               if (retries++ < UBI_IO_RETRIES) {
+                       ubi_warn("error %d%s while reading %d bytes from PEB %d:%d, read only %zd bytes, retry",
+                                err, errstr, len, pnum, offset, read);
                        yield();
                        goto retry;
                }
 
-               ubi_err("error %d while reading %d bytes from PEB %d:%d, "
-                       "read %zd bytes", err, len, pnum, offset, read);
-               ubi_dbg_dump_stack();
+               ubi_err("error %d%s while reading %d bytes from PEB %d:%d, read %zd bytes",
+                       err, errstr, len, pnum, offset, read);
+               dump_stack();
 
                /*
                 * The driver should never return -EBADMSG if it failed to read
                 * all the requested data. But some buggy drivers might do
                 * this, so we change it to -EIO.
                 */
-               if (read != len && err == -EBADMSG) {
+               if (read != len && mtd_is_eccerr(err)) {
                        ubi_assert(0);
-                       printk("%s[%d] not here\n", __func__, __LINE__);
-/*                     err = -EIO; */
+                       err = -EIO;
                }
        } else {
                ubi_assert(len == read);
 
-               if (ubi_dbg_is_bitflip()) {
-                       dbg_msg("bit-flip (emulated)");
+               if (ubi_dbg_is_bitflip(ubi)) {
+                       dbg_gen("bit-flip (emulated)");
                        err = UBI_IO_BITFLIPS;
                }
        }
@@ -224,46 +243,60 @@ int ubi_io_write(struct ubi_device *ubi, const void *buf, int pnum, int offset,
                return -EROFS;
        }
 
-       /* The below has to be compiled out if paranoid checks are disabled */
-
-       err = paranoid_check_not_bad(ubi, pnum);
+       err = self_check_not_bad(ubi, pnum);
        if (err)
-               return err > 0 ? -EINVAL : err;
+               return err;
 
        /* The area we are writing to has to contain all 0xFF bytes */
-       err = paranoid_check_all_ff(ubi, pnum, offset, len);
+       err = ubi_self_check_all_ff(ubi, pnum, offset, len);
        if (err)
-               return err > 0 ? -EINVAL : err;
+               return err;
 
        if (offset >= ubi->leb_start) {
                /*
                 * We write to the data area of the physical eraseblock. Make
                 * sure it has valid EC and VID headers.
                 */
-               err = paranoid_check_peb_ec_hdr(ubi, pnum);
+               err = self_check_peb_ec_hdr(ubi, pnum);
                if (err)
-                       return err > 0 ? -EINVAL : err;
-               err = paranoid_check_peb_vid_hdr(ubi, pnum);
+                       return err;
+               err = self_check_peb_vid_hdr(ubi, pnum);
                if (err)
-                       return err > 0 ? -EINVAL : err;
+                       return err;
        }
 
-       if (ubi_dbg_is_write_failure()) {
-               dbg_err("cannot write %d bytes to PEB %d:%d "
-                       "(emulated)", len, pnum, offset);
-               ubi_dbg_dump_stack();
+       if (ubi_dbg_is_write_failure(ubi)) {
+               ubi_err("cannot write %d bytes to PEB %d:%d (emulated)",
+                       len, pnum, offset);
+               dump_stack();
                return -EIO;
        }
 
        addr = (loff_t)pnum * ubi->peb_size + offset;
        err = mtd_write(ubi->mtd, addr, len, &written, buf);
        if (err) {
-               ubi_err("error %d while writing %d bytes to PEB %d:%d, written"
-                       " %zd bytes", err, len, pnum, offset, written);
-               ubi_dbg_dump_stack();
+               ubi_err("error %d while writing %d bytes to PEB %d:%d, written %zd bytes",
+                       err, len, pnum, offset, written);
+               dump_stack();
+               ubi_dump_flash(ubi, pnum, offset, len);
        } else
                ubi_assert(written == len);
 
+       if (!err) {
+               err = self_check_write(ubi, buf, pnum, offset, len);
+               if (err)
+                       return err;
+
+               /*
+                * Since we always write sequentially, the rest of the PEB has
+                * to contain only 0xFF bytes.
+                */
+               offset += len;
+               len = ubi->peb_size - offset;
+               if (len)
+                       err = ubi_self_check_all_ff(ubi, pnum, offset, len);
+       }
+
        return err;
 }
 
@@ -295,6 +328,12 @@ static int do_sync_erase(struct ubi_device *ubi, int pnum)
        wait_queue_head_t wq;
 
        dbg_io("erase PEB %d", pnum);
+       ubi_assert(pnum >= 0 && pnum < ubi->peb_count);
+
+       if (ubi->ro_mode) {
+               ubi_err("read-only mode");
+               return -EROFS;
+       }
 
 retry:
        init_waitqueue_head(&wq);
@@ -309,13 +348,13 @@ retry:
        err = mtd_erase(ubi->mtd, &ei);
        if (err) {
                if (retries++ < UBI_IO_RETRIES) {
-                       dbg_io("error %d while erasing PEB %d, retry",
-                              err, pnum);
+                       ubi_warn("error %d while erasing PEB %d, retry",
+                                err, pnum);
                        yield();
                        goto retry;
                }
                ubi_err("cannot erase PEB %d, error %d", pnum, err);
-               ubi_dbg_dump_stack();
+               dump_stack();
                return err;
        }
 
@@ -328,46 +367,27 @@ retry:
 
        if (ei.state == MTD_ERASE_FAILED) {
                if (retries++ < UBI_IO_RETRIES) {
-                       dbg_io("error while erasing PEB %d, retry", pnum);
+                       ubi_warn("error while erasing PEB %d, retry", pnum);
                        yield();
                        goto retry;
                }
                ubi_err("cannot erase PEB %d", pnum);
-               ubi_dbg_dump_stack();
+               dump_stack();
                return -EIO;
        }
 
-       err = paranoid_check_all_ff(ubi, pnum, 0, ubi->peb_size);
+       err = ubi_self_check_all_ff(ubi, pnum, 0, ubi->peb_size);
        if (err)
-               return err > 0 ? -EINVAL : err;
+               return err;
 
-       if (ubi_dbg_is_erase_failure() && !err) {
-               dbg_err("cannot erase PEB %d (emulated)", pnum);
+       if (ubi_dbg_is_erase_failure(ubi)) {
+               ubi_err("cannot erase PEB %d (emulated)", pnum);
                return -EIO;
        }
 
        return 0;
 }
 
-/**
- * check_pattern - check if buffer contains only a certain byte pattern.
- * @buf: buffer to check
- * @patt: the pattern to check
- * @size: buffer size in bytes
- *
- * This function returns %1 in there are only @patt bytes in @buf, and %0 if
- * something else was also found.
- */
-static int check_pattern(const void *buf, uint8_t patt, int size)
-{
-       int i;
-
-       for (i = 0; i < size; i++)
-               if (((const uint8_t *)buf)[i] != patt)
-                       return 0;
-       return 1;
-}
-
 /* Patterns to write to a physical eraseblock when torturing it */
 static uint8_t patterns[] = {0xa5, 0x5a, 0x0};
 
@@ -384,6 +404,7 @@ static int torture_peb(struct ubi_device *ubi, int pnum)
 {
        int err, i, patt_count;
 
+       ubi_msg("run torture test for PEB %d", pnum);
        patt_count = ARRAY_SIZE(patterns);
        ubi_assert(patt_count > 0);
 
@@ -394,11 +415,11 @@ static int torture_peb(struct ubi_device *ubi, int pnum)
                        goto out;
 
                /* Make sure the PEB contains only 0xFF bytes */
-               err = ubi_io_read(ubi, ubi->peb_buf1, pnum, 0, ubi->peb_size);
+               err = ubi_io_read(ubi, ubi->peb_buf, pnum, 0, ubi->peb_size);
                if (err)
                        goto out;
 
-               err = check_pattern(ubi->peb_buf1, 0xFF, ubi->peb_size);
+               err = ubi_check_pattern(ubi->peb_buf, 0xFF, ubi->peb_size);
                if (err == 0) {
                        ubi_err("erased PEB %d, but a non-0xFF byte found",
                                pnum);
@@ -407,17 +428,18 @@ static int torture_peb(struct ubi_device *ubi, int pnum)
                }
 
                /* Write a pattern and check it */
-               memset(ubi->peb_buf1, patterns[i], ubi->peb_size);
-               err = ubi_io_write(ubi, ubi->peb_buf1, pnum, 0, ubi->peb_size);
+               memset(ubi->peb_buf, patterns[i], ubi->peb_size);
+               err = ubi_io_write(ubi, ubi->peb_buf, pnum, 0, ubi->peb_size);
                if (err)
                        goto out;
 
-               memset(ubi->peb_buf1, ~patterns[i], ubi->peb_size);
-               err = ubi_io_read(ubi, ubi->peb_buf1, pnum, 0, ubi->peb_size);
+               memset(ubi->peb_buf, ~patterns[i], ubi->peb_size);
+               err = ubi_io_read(ubi, ubi->peb_buf, pnum, 0, ubi->peb_size);
                if (err)
                        goto out;
 
-               err = check_pattern(ubi->peb_buf1, patterns[i], ubi->peb_size);
+               err = ubi_check_pattern(ubi->peb_buf, patterns[i],
+                                       ubi->peb_size);
                if (err == 0) {
                        ubi_err("pattern %x checking failed for PEB %d",
                                patterns[i], pnum);
@@ -427,10 +449,11 @@ static int torture_peb(struct ubi_device *ubi, int pnum)
        }
 
        err = patt_count;
+       ubi_msg("PEB %d passed torture test, do not mark it as bad", pnum);
 
 out:
        mutex_unlock(&ubi->buf_mutex);
-       if (err == UBI_IO_BITFLIPS || err == -EBADMSG) {
+       if (err == UBI_IO_BITFLIPS || mtd_is_eccerr(err)) {
                /*
                 * If a bit-flip or data integrity error was detected, the test
                 * has not passed because it happened on a freshly erased
@@ -443,6 +466,80 @@ out:
        return err;
 }
 
+/**
+ * nor_erase_prepare - prepare a NOR flash PEB for erasure.
+ * @ubi: UBI device description object
+ * @pnum: physical eraseblock number to prepare
+ *
+ * NOR flash, or at least some of them, have peculiar embedded PEB erasure
+ * algorithm: the PEB is first filled with zeroes, then it is erased. And
+ * filling with zeroes starts from the end of the PEB. This was observed with
+ * Spansion S29GL512N NOR flash.
+ *
+ * This means that in case of a power cut we may end up with intact data at the
+ * beginning of the PEB, and all zeroes at the end of PEB. In other words, the
+ * EC and VID headers are OK, but a large chunk of data at the end of PEB is
+ * zeroed. This makes UBI mistakenly treat this PEB as used and associate it
+ * with an LEB, which leads to subsequent failures (e.g., UBIFS fails).
+ *
+ * This function is called before erasing NOR PEBs and it zeroes out EC and VID
+ * magic numbers in order to invalidate them and prevent the failures. Returns
+ * zero in case of success and a negative error code in case of failure.
+ */
+static int nor_erase_prepare(struct ubi_device *ubi, int pnum)
+{
+       int err;
+       size_t written;
+       loff_t addr;
+       uint32_t data = 0;
+       struct ubi_ec_hdr ec_hdr;
+
+       /*
+        * Note, we cannot generally define VID header buffers on stack,
+        * because of the way we deal with these buffers (see the header
+        * comment in this file). But we know this is a NOR-specific piece of
+        * code, so we can do this. But yes, this is error-prone and we should
+        * (pre-)allocate VID header buffer instead.
+        */
+       struct ubi_vid_hdr vid_hdr;
+
+       /*
+        * If VID or EC is valid, we have to corrupt them before erasing.
+        * It is important to first invalidate the EC header, and then the VID
+        * header. Otherwise a power cut may lead to valid EC header and
+        * invalid VID header, in which case UBI will treat this PEB as
+        * corrupted and will try to preserve it, and print scary warnings.
+        */
+       addr = (loff_t)pnum * ubi->peb_size;
+       err = ubi_io_read_ec_hdr(ubi, pnum, &ec_hdr, 0);
+       if (err != UBI_IO_BAD_HDR_EBADMSG && err != UBI_IO_BAD_HDR &&
+           err != UBI_IO_FF){
+               err = mtd_write(ubi->mtd, addr, 4, &written, (void *)&data);
+               if(err)
+                       goto error;
+       }
+
+       err = ubi_io_read_vid_hdr(ubi, pnum, &vid_hdr, 0);
+       if (err != UBI_IO_BAD_HDR_EBADMSG && err != UBI_IO_BAD_HDR &&
+           err != UBI_IO_FF){
+               addr += ubi->vid_hdr_aloffset;
+               err = mtd_write(ubi->mtd, addr, 4, &written, (void *)&data);
+               if (err)
+                       goto error;
+       }
+       return 0;
+
+error:
+       /*
+        * The PEB contains a valid VID or EC header, but we cannot invalidate
+        * it. Supposedly the flash media or the driver is screwed up, so
+        * return an error.
+        */
+       ubi_err("cannot invalidate PEB %d, write returned %d", pnum, err);
+       ubi_dump_flash(ubi, pnum, 0, ubi->peb_size);
+       return -EIO;
+}
+
 /**
  * ubi_io_sync_erase - synchronously erase a physical eraseblock.
  * @ubi: UBI device description object
@@ -452,7 +549,7 @@ out:
  * This function synchronously erases physical eraseblock @pnum. If @torture
  * flag is not zero, the physical eraseblock is checked by means of writing
  * different patterns to it and reading them back. If the torturing is enabled,
- * the physical eraseblock is erased more then once.
+ * the physical eraseblock is erased more than once.
  *
  * This function returns the number of erasures made in case of success, %-EIO
  * if the erasure failed or the torturing test failed, and other negative error
@@ -465,15 +562,21 @@ int ubi_io_sync_erase(struct ubi_device *ubi, int pnum, int torture)
 
        ubi_assert(pnum >= 0 && pnum < ubi->peb_count);
 
-       err = paranoid_check_not_bad(ubi, pnum);
+       err = self_check_not_bad(ubi, pnum);
        if (err != 0)
-               return err > 0 ? -EINVAL : err;
+               return err;
 
        if (ubi->ro_mode) {
                ubi_err("read-only mode");
                return -EROFS;
        }
 
+       if (ubi->nor_flash) {
+               err = nor_erase_prepare(ubi, pnum);
+               if (err)
+                       return err;
+       }
+
        if (torture) {
                ret = torture_peb(ubi, pnum);
                if (ret < 0)
@@ -564,8 +667,7 @@ static int validate_ec_hdr(const struct ubi_device *ubi,
        leb_start = be32_to_cpu(ec_hdr->data_offset);
 
        if (ec_hdr->version != UBI_VERSION) {
-               ubi_err("node with incompatible UBI version found: "
-                       "this UBI version is %d, image version is %d",
+               ubi_err("node with incompatible UBI version found: this UBI version is %d, image version is %d",
                        UBI_VERSION, (int)ec_hdr->version);
                goto bad;
        }
@@ -591,8 +693,8 @@ static int validate_ec_hdr(const struct ubi_device *ubi,
 
 bad:
        ubi_err("bad EC header");
-       ubi_dbg_dump_ec_hdr(ec_hdr);
-       ubi_dbg_dump_stack();
+       ubi_dump_ec_hdr(ec_hdr);
+       dump_stack();
        return 1;
 }
 
@@ -612,67 +714,58 @@ bad:
  * o %UBI_IO_BITFLIPS if the CRC is correct, but bit-flips were detected
  *   and corrected by the flash driver; this is harmless but may indicate that
  *   this eraseblock may become bad soon (but may be not);
- * o %UBI_IO_BAD_EC_HDR if the erase counter header is corrupted (a CRC error);
- * o %UBI_IO_PEB_EMPTY if the physical eraseblock is empty;
+ * o %UBI_IO_BAD_HDR if the erase counter header is corrupted (a CRC error);
+ * o %UBI_IO_BAD_HDR_EBADMSG is the same as %UBI_IO_BAD_HDR, but there also was
+ *   a data integrity error (uncorrectable ECC error in case of NAND);
+ * o %UBI_IO_FF if only 0xFF bytes were read (the PEB is supposedly empty)
  * o a negative error code in case of failure.
  */
 int ubi_io_read_ec_hdr(struct ubi_device *ubi, int pnum,
                       struct ubi_ec_hdr *ec_hdr, int verbose)
 {
-       int err, read_err = 0;
+       int err, read_err;
        uint32_t crc, magic, hdr_crc;
 
        dbg_io("read EC header from PEB %d", pnum);
        ubi_assert(pnum >= 0 && pnum < ubi->peb_count);
-       if (UBI_IO_DEBUG)
-               verbose = 1;
 
-       err = ubi_io_read(ubi, ec_hdr, pnum, 0, UBI_EC_HDR_SIZE);
-       if (err) {
-               if (err != UBI_IO_BITFLIPS && err != -EBADMSG)
-                       return err;
+       read_err = ubi_io_read(ubi, ec_hdr, pnum, 0, UBI_EC_HDR_SIZE);
+       if (read_err) {
+               if (read_err != UBI_IO_BITFLIPS && !mtd_is_eccerr(read_err))
+                       return read_err;
 
                /*
                 * We read all the data, but either a correctable bit-flip
-                * occurred, or MTD reported about some data integrity error,
-                * like an ECC error in case of NAND. The former is harmless,
-                * the later may mean that the read data is corrupted. But we
-                * have a CRC check-sum and we will detect this. If the EC
-                * header is still OK, we just report this as there was a
-                * bit-flip.
+                * occurred, or MTD reported a data integrity error
+                * (uncorrectable ECC error in case of NAND). The former is
+                * harmless, the later may mean that the read data is
+                * corrupted. But we have a CRC check-sum and we will detect
+                * this. If the EC header is still OK, we just report this as
+                * there was a bit-flip, to force scrubbing.
                 */
-               read_err = err;
        }
 
        magic = be32_to_cpu(ec_hdr->magic);
        if (magic != UBI_EC_HDR_MAGIC) {
+               if (mtd_is_eccerr(read_err))
+                       return UBI_IO_BAD_HDR_EBADMSG;
+
                /*
                 * The magic field is wrong. Let's check if we have read all
                 * 0xFF. If yes, this physical eraseblock is assumed to be
                 * empty.
-                *
-                * But if there was a read error, we do not test it for all
-                * 0xFFs. Even if it does contain all 0xFFs, this error
-                * indicates that something is still wrong with this physical
-                * eraseblock and we anyway cannot treat it as empty.
                 */
-               if (read_err != -EBADMSG &&
-                   check_pattern(ec_hdr, 0xFF, UBI_EC_HDR_SIZE)) {
+               if (ubi_check_pattern(ec_hdr, 0xFF, UBI_EC_HDR_SIZE)) {
                        /* The physical eraseblock is supposedly empty */
-
-                       /*
-                        * The below is just a paranoid check, it has to be
-                        * compiled out if paranoid checks are disabled.
-                        */
-                       err = paranoid_check_all_ff(ubi, pnum, 0,
-                                                   ubi->peb_size);
-                       if (err)
-                               return err > 0 ? UBI_IO_BAD_EC_HDR : err;
-
                        if (verbose)
-                               ubi_warn("no EC header found at PEB %d, "
-                                        "only 0xFF bytes", pnum);
-                       return UBI_IO_PEB_EMPTY;
+                               ubi_warn("no EC header found at PEB %d, only 0xFF bytes",
+                                        pnum);
+                       dbg_bld("no EC header found at PEB %d, only 0xFF bytes",
+                               pnum);
+                       if (!read_err)
+                               return UBI_IO_FF;
+                       else
+                               return UBI_IO_FF_BITFLIPS;
                }
 
                /*
@@ -680,11 +773,13 @@ int ubi_io_read_ec_hdr(struct ubi_device *ubi, int pnum,
                 * 0xFF bytes. Report that the header is corrupted.
                 */
                if (verbose) {
-                       ubi_warn("bad magic number at PEB %d: %08x instead of "
-                                "%08x", pnum, magic, UBI_EC_HDR_MAGIC);
-                       ubi_dbg_dump_ec_hdr(ec_hdr);
+                       ubi_warn("bad magic number at PEB %d: %08x instead of %08x",
+                                pnum, magic, UBI_EC_HDR_MAGIC);
+                       ubi_dump_ec_hdr(ec_hdr);
                }
-               return UBI_IO_BAD_EC_HDR;
+               dbg_bld("bad magic number at PEB %d: %08x instead of %08x",
+                       pnum, magic, UBI_EC_HDR_MAGIC);
+               return UBI_IO_BAD_HDR;
        }
 
        crc = crc32(UBI_CRC32_INIT, ec_hdr, UBI_EC_HDR_SIZE_CRC);
@@ -692,11 +787,17 @@ int ubi_io_read_ec_hdr(struct ubi_device *ubi, int pnum,
 
        if (hdr_crc != crc) {
                if (verbose) {
-                       ubi_warn("bad EC header CRC at PEB %d, calculated %#08x,"
-                                " read %#08x", pnum, crc, hdr_crc);
-                       ubi_dbg_dump_ec_hdr(ec_hdr);
+                       ubi_warn("bad EC header CRC at PEB %d, calculated %#08x, read %#08x",
+                                pnum, crc, hdr_crc);
+                       ubi_dump_ec_hdr(ec_hdr);
                }
-               return UBI_IO_BAD_EC_HDR;
+               dbg_bld("bad EC header CRC at PEB %d, calculated %#08x, read %#08x",
+                       pnum, crc, hdr_crc);
+
+               if (!read_err)
+                       return UBI_IO_BAD_HDR;
+               else
+                       return UBI_IO_BAD_HDR_EBADMSG;
        }
 
        /* And of course validate what has just been read from the media */
@@ -706,6 +807,10 @@ int ubi_io_read_ec_hdr(struct ubi_device *ubi, int pnum,
                return -EINVAL;
        }
 
+       /*
+        * If there was %-EBADMSG, but the header CRC is still OK, report about
+        * a bit-flip to force scrubbing on this PEB.
+        */
        return read_err ? UBI_IO_BITFLIPS : 0;
 }
 
@@ -737,12 +842,13 @@ int ubi_io_write_ec_hdr(struct ubi_device *ubi, int pnum,
        ec_hdr->version = UBI_VERSION;
        ec_hdr->vid_hdr_offset = cpu_to_be32(ubi->vid_hdr_offset);
        ec_hdr->data_offset = cpu_to_be32(ubi->leb_start);
+       ec_hdr->image_seq = cpu_to_be32(ubi->image_seq);
        crc = crc32(UBI_CRC32_INIT, ec_hdr, UBI_EC_HDR_SIZE_CRC);
        ec_hdr->hdr_crc = cpu_to_be32(crc);
 
-       err = paranoid_check_ec_hdr(ubi, pnum, ec_hdr);
+       err = self_check_ec_hdr(ubi, pnum, ec_hdr);
        if (err)
-               return -EINVAL;
+               return err;
 
        err = ubi_io_write(ubi, ec_hdr, pnum, 0, ubi->ec_hdr_alsize);
        return err;
@@ -771,40 +877,40 @@ static int validate_vid_hdr(const struct ubi_device *ubi,
        int usable_leb_size = ubi->leb_size - data_pad;
 
        if (copy_flag != 0 && copy_flag != 1) {
-               dbg_err("bad copy_flag");
+               ubi_err("bad copy_flag");
                goto bad;
        }
 
        if (vol_id < 0 || lnum < 0 || data_size < 0 || used_ebs < 0 ||
            data_pad < 0) {
-               dbg_err("negative values");
+               ubi_err("negative values");
                goto bad;
        }
 
        if (vol_id >= UBI_MAX_VOLUMES && vol_id < UBI_INTERNAL_VOL_START) {
-               dbg_err("bad vol_id");
+               ubi_err("bad vol_id");
                goto bad;
        }
 
        if (vol_id < UBI_INTERNAL_VOL_START && compat != 0) {
-               dbg_err("bad compat");
+               ubi_err("bad compat");
                goto bad;
        }
 
        if (vol_id >= UBI_INTERNAL_VOL_START && compat != UBI_COMPAT_DELETE &&
            compat != UBI_COMPAT_RO && compat != UBI_COMPAT_PRESERVE &&
            compat != UBI_COMPAT_REJECT) {
-               dbg_err("bad compat");
+               ubi_err("bad compat");
                goto bad;
        }
 
        if (vol_type != UBI_VID_DYNAMIC && vol_type != UBI_VID_STATIC) {
-               dbg_err("bad vol_type");
+               ubi_err("bad vol_type");
                goto bad;
        }
 
        if (data_pad >= ubi->leb_size / 2) {
-               dbg_err("bad data_pad");
+               ubi_err("bad data_pad");
                goto bad;
        }
 
@@ -816,45 +922,45 @@ static int validate_vid_hdr(const struct ubi_device *ubi,
                 * mapped logical eraseblocks.
                 */
                if (used_ebs == 0) {
-                       dbg_err("zero used_ebs");
+                       ubi_err("zero used_ebs");
                        goto bad;
                }
                if (data_size == 0) {
-                       dbg_err("zero data_size");
+                       ubi_err("zero data_size");
                        goto bad;
                }
                if (lnum < used_ebs - 1) {
                        if (data_size != usable_leb_size) {
-                               dbg_err("bad data_size");
+                               ubi_err("bad data_size");
                                goto bad;
                        }
                } else if (lnum == used_ebs - 1) {
                        if (data_size == 0) {
-                               dbg_err("bad data_size at last LEB");
+                               ubi_err("bad data_size at last LEB");
                                goto bad;
                        }
                } else {
-                       dbg_err("too high lnum");
+                       ubi_err("too high lnum");
                        goto bad;
                }
        } else {
                if (copy_flag == 0) {
                        if (data_crc != 0) {
-                               dbg_err("non-zero data CRC");
+                               ubi_err("non-zero data CRC");
                                goto bad;
                        }
                        if (data_size != 0) {
-                               dbg_err("non-zero data_size");
+                               ubi_err("non-zero data_size");
                                goto bad;
                        }
                } else {
                        if (data_size == 0) {
-                               dbg_err("zero data_size of copy");
+                               ubi_err("zero data_size of copy");
                                goto bad;
                        }
                }
                if (used_ebs != 0) {
-                       dbg_err("bad used_ebs");
+                       ubi_err("bad used_ebs");
                        goto bad;
                }
        }
@@ -863,8 +969,8 @@ static int validate_vid_hdr(const struct ubi_device *ubi,
 
 bad:
        ubi_err("bad VID header");
-       ubi_dbg_dump_vid_hdr(vid_hdr);
-       ubi_dbg_dump_stack();
+       ubi_dump_vid_hdr(vid_hdr);
+       dump_stack();
        return 1;
 }
 
@@ -878,88 +984,53 @@ bad:
  *
  * This function reads the volume identifier header from physical eraseblock
  * @pnum and stores it in @vid_hdr. It also checks CRC checksum of the read
- * volume identifier header. The following codes may be returned:
+ * volume identifier header. The error codes are the same as in
+ * 'ubi_io_read_ec_hdr()'.
  *
- * o %0 if the CRC checksum is correct and the header was successfully read;
- * o %UBI_IO_BITFLIPS if the CRC is correct, but bit-flips were detected
- *   and corrected by the flash driver; this is harmless but may indicate that
- *   this eraseblock may become bad soon;
- * o %UBI_IO_BAD_VID_HRD if the volume identifier header is corrupted (a CRC
- *   error detected);
- * o %UBI_IO_PEB_FREE if the physical eraseblock is free (i.e., there is no VID
- *   header there);
- * o a negative error code in case of failure.
+ * Note, the implementation of this function is also very similar to
+ * 'ubi_io_read_ec_hdr()', so refer commentaries in 'ubi_io_read_ec_hdr()'.
  */
 int ubi_io_read_vid_hdr(struct ubi_device *ubi, int pnum,
                        struct ubi_vid_hdr *vid_hdr, int verbose)
 {
-       int err, read_err = 0;
+       int err, read_err;
        uint32_t crc, magic, hdr_crc;
        void *p;
 
        dbg_io("read VID header from PEB %d", pnum);
        ubi_assert(pnum >= 0 &&  pnum < ubi->peb_count);
-       if (UBI_IO_DEBUG)
-               verbose = 1;
 
        p = (char *)vid_hdr - ubi->vid_hdr_shift;
-       err = ubi_io_read(ubi, p, pnum, ubi->vid_hdr_aloffset,
+       read_err = ubi_io_read(ubi, p, pnum, ubi->vid_hdr_aloffset,
                          ubi->vid_hdr_alsize);
-       if (err) {
-               if (err != UBI_IO_BITFLIPS && err != -EBADMSG)
-                       return err;
-
-               /*
-                * We read all the data, but either a correctable bit-flip
-                * occurred, or MTD reported about some data integrity error,
-                * like an ECC error in case of NAND. The former is harmless,
-                * the later may mean the read data is corrupted. But we have a
-                * CRC check-sum and we will identify this. If the VID header is
-                * still OK, we just report this as there was a bit-flip.
-                */
-               read_err = err;
-       }
+       if (read_err && read_err != UBI_IO_BITFLIPS && !mtd_is_eccerr(read_err))
+               return read_err;
 
        magic = be32_to_cpu(vid_hdr->magic);
        if (magic != UBI_VID_HDR_MAGIC) {
-               /*
-                * If we have read all 0xFF bytes, the VID header probably does
-                * not exist and the physical eraseblock is assumed to be free.
-                *
-                * But if there was a read error, we do not test the data for
-                * 0xFFs. Even if it does contain all 0xFFs, this error
-                * indicates that something is still wrong with this physical
-                * eraseblock and it cannot be regarded as free.
-                */
-               if (read_err != -EBADMSG &&
-                   check_pattern(vid_hdr, 0xFF, UBI_VID_HDR_SIZE)) {
-                       /* The physical eraseblock is supposedly free */
-
-                       /*
-                        * The below is just a paranoid check, it has to be
-                        * compiled out if paranoid checks are disabled.
-                        */
-                       err = paranoid_check_all_ff(ubi, pnum, ubi->leb_start,
-                                                   ubi->leb_size);
-                       if (err)
-                               return err > 0 ? UBI_IO_BAD_VID_HDR : err;
+               if (mtd_is_eccerr(read_err))
+                       return UBI_IO_BAD_HDR_EBADMSG;
 
+               if (ubi_check_pattern(vid_hdr, 0xFF, UBI_VID_HDR_SIZE)) {
                        if (verbose)
-                               ubi_warn("no VID header found at PEB %d, "
-                                        "only 0xFF bytes", pnum);
-                       return UBI_IO_PEB_FREE;
+                               ubi_warn("no VID header found at PEB %d, only 0xFF bytes",
+                                        pnum);
+                       dbg_bld("no VID header found at PEB %d, only 0xFF bytes",
+                               pnum);
+                       if (!read_err)
+                               return UBI_IO_FF;
+                       else
+                               return UBI_IO_FF_BITFLIPS;
                }
 
-               /*
-                * This is not a valid VID header, and these are not 0xFF
-                * bytes. Report that the header is corrupted.
-                */
                if (verbose) {
-                       ubi_warn("bad magic number at PEB %d: %08x instead of "
-                                "%08x", pnum, magic, UBI_VID_HDR_MAGIC);
-                       ubi_dbg_dump_vid_hdr(vid_hdr);
+                       ubi_warn("bad magic number at PEB %d: %08x instead of %08x",
+                                pnum, magic, UBI_VID_HDR_MAGIC);
+                       ubi_dump_vid_hdr(vid_hdr);
                }
-               return UBI_IO_BAD_VID_HDR;
+               dbg_bld("bad magic number at PEB %d: %08x instead of %08x",
+                       pnum, magic, UBI_VID_HDR_MAGIC);
+               return UBI_IO_BAD_HDR;
        }
 
        crc = crc32(UBI_CRC32_INIT, vid_hdr, UBI_VID_HDR_SIZE_CRC);
@@ -967,14 +1038,18 @@ int ubi_io_read_vid_hdr(struct ubi_device *ubi, int pnum,
 
        if (hdr_crc != crc) {
                if (verbose) {
-                       ubi_warn("bad CRC at PEB %d, calculated %#08x, "
-                                "read %#08x", pnum, crc, hdr_crc);
-                       ubi_dbg_dump_vid_hdr(vid_hdr);
+                       ubi_warn("bad CRC at PEB %d, calculated %#08x, read %#08x",
+                                pnum, crc, hdr_crc);
+                       ubi_dump_vid_hdr(vid_hdr);
                }
-               return UBI_IO_BAD_VID_HDR;
+               dbg_bld("bad CRC at PEB %d, calculated %#08x, read %#08x",
+                       pnum, crc, hdr_crc);
+               if (!read_err)
+                       return UBI_IO_BAD_HDR;
+               else
+                       return UBI_IO_BAD_HDR_EBADMSG;
        }
 
-       /* Validate the VID header that we have just read */
        err = validate_vid_hdr(ubi, vid_hdr);
        if (err) {
                ubi_err("validation failed for PEB %d", pnum);
@@ -1009,18 +1084,18 @@ int ubi_io_write_vid_hdr(struct ubi_device *ubi, int pnum,
        dbg_io("write VID header to PEB %d", pnum);
        ubi_assert(pnum >= 0 &&  pnum < ubi->peb_count);
 
-       err = paranoid_check_peb_ec_hdr(ubi, pnum);
+       err = self_check_peb_ec_hdr(ubi, pnum);
        if (err)
-               return err > 0 ? -EINVAL: err;
+               return err;
 
        vid_hdr->magic = cpu_to_be32(UBI_VID_HDR_MAGIC);
        vid_hdr->version = UBI_VERSION;
        crc = crc32(UBI_CRC32_INIT, vid_hdr, UBI_VID_HDR_SIZE_CRC);
        vid_hdr->hdr_crc = cpu_to_be32(crc);
 
-       err = paranoid_check_vid_hdr(ubi, pnum, vid_hdr);
+       err = self_check_vid_hdr(ubi, pnum, vid_hdr);
        if (err)
-               return -EINVAL;
+               return err;
 
        p = (char *)vid_hdr - ubi->vid_hdr_shift;
        err = ubi_io_write(ubi, p, pnum, ubi->vid_hdr_aloffset,
@@ -1028,44 +1103,48 @@ int ubi_io_write_vid_hdr(struct ubi_device *ubi, int pnum,
        return err;
 }
 
-#ifdef CONFIG_MTD_UBI_DEBUG_PARANOID
-
 /**
- * paranoid_check_not_bad - ensure that a physical eraseblock is not bad.
+ * self_check_not_bad - ensure that a physical eraseblock is not bad.
  * @ubi: UBI device description object
  * @pnum: physical eraseblock number to check
  *
- * This function returns zero if the physical eraseblock is good, a positive
- * number if it is bad and a negative error code if an error occurred.
+ * This function returns zero if the physical eraseblock is good, %-EINVAL if
+ * it is bad and a negative error code if an error occurred.
  */
-static int paranoid_check_not_bad(const struct ubi_device *ubi, int pnum)
+static int self_check_not_bad(const struct ubi_device *ubi, int pnum)
 {
        int err;
 
+       if (!ubi_dbg_chk_io(ubi))
+               return 0;
+
        err = ubi_io_is_bad(ubi, pnum);
        if (!err)
                return err;
 
-       ubi_err("paranoid check failed for PEB %d", pnum);
-       ubi_dbg_dump_stack();
-       return err;
+       ubi_err("self-check failed for PEB %d", pnum);
+       dump_stack();
+       return err > 0 ? -EINVAL : err;
 }
 
 /**
- * paranoid_check_ec_hdr - check if an erase counter header is all right.
+ * self_check_ec_hdr - check if an erase counter header is all right.
  * @ubi: UBI device description object
  * @pnum: physical eraseblock number the erase counter header belongs to
  * @ec_hdr: the erase counter header to check
  *
  * This function returns zero if the erase counter header contains valid
- * values, and %1 if not.
+ * values, and %-EINVAL if not.
  */
-static int paranoid_check_ec_hdr(const struct ubi_device *ubi, int pnum,
-                                const struct ubi_ec_hdr *ec_hdr)
+static int self_check_ec_hdr(const struct ubi_device *ubi, int pnum,
+                            const struct ubi_ec_hdr *ec_hdr)
 {
        int err;
        uint32_t magic;
 
+       if (!ubi_dbg_chk_io(ubi))
+               return 0;
+
        magic = be32_to_cpu(ec_hdr->magic);
        if (magic != UBI_EC_HDR_MAGIC) {
                ubi_err("bad magic %#08x, must be %#08x",
@@ -1075,53 +1154,55 @@ static int paranoid_check_ec_hdr(const struct ubi_device *ubi, int pnum,
 
        err = validate_ec_hdr(ubi, ec_hdr);
        if (err) {
-               ubi_err("paranoid check failed for PEB %d", pnum);
+               ubi_err("self-check failed for PEB %d", pnum);
                goto fail;
        }
 
        return 0;
 
 fail:
-       ubi_dbg_dump_ec_hdr(ec_hdr);
-       ubi_dbg_dump_stack();
-       return 1;
+       ubi_dump_ec_hdr(ec_hdr);
+       dump_stack();
+       return -EINVAL;
 }
 
 /**
- * paranoid_check_peb_ec_hdr - check that the erase counter header of a
- * physical eraseblock is in-place and is all right.
+ * self_check_peb_ec_hdr - check erase counter header.
  * @ubi: UBI device description object
  * @pnum: the physical eraseblock number to check
  *
- * This function returns zero if the erase counter header is all right, %1 if
- * not, and a negative error code if an error occurred.
+ * This function returns zero if the erase counter header is all right and and
+ * a negative error code if not or if an error occurred.
  */
-static int paranoid_check_peb_ec_hdr(const struct ubi_device *ubi, int pnum)
+static int self_check_peb_ec_hdr(const struct ubi_device *ubi, int pnum)
 {
        int err;
        uint32_t crc, hdr_crc;
        struct ubi_ec_hdr *ec_hdr;
 
+       if (!ubi_dbg_chk_io(ubi))
+               return 0;
+
        ec_hdr = kzalloc(ubi->ec_hdr_alsize, GFP_NOFS);
        if (!ec_hdr)
                return -ENOMEM;
 
        err = ubi_io_read(ubi, ec_hdr, pnum, 0, UBI_EC_HDR_SIZE);
-       if (err && err != UBI_IO_BITFLIPS && err != -EBADMSG)
+       if (err && err != UBI_IO_BITFLIPS && !mtd_is_eccerr(err))
                goto exit;
 
        crc = crc32(UBI_CRC32_INIT, ec_hdr, UBI_EC_HDR_SIZE_CRC);
        hdr_crc = be32_to_cpu(ec_hdr->hdr_crc);
        if (hdr_crc != crc) {
                ubi_err("bad CRC, calculated %#08x, read %#08x", crc, hdr_crc);
-               ubi_err("paranoid check failed for PEB %d", pnum);
-               ubi_dbg_dump_ec_hdr(ec_hdr);
-               ubi_dbg_dump_stack();
-               err = 1;
+               ubi_err("self-check failed for PEB %d", pnum);
+               ubi_dump_ec_hdr(ec_hdr);
+               dump_stack();
+               err = -EINVAL;
                goto exit;
        }
 
-       err = paranoid_check_ec_hdr(ubi, pnum, ec_hdr);
+       err = self_check_ec_hdr(ubi, pnum, ec_hdr);
 
 exit:
        kfree(ec_hdr);
@@ -1129,20 +1210,23 @@ exit:
 }
 
 /**
- * paranoid_check_vid_hdr - check that a volume identifier header is all right.
+ * self_check_vid_hdr - check that a volume identifier header is all right.
  * @ubi: UBI device description object
  * @pnum: physical eraseblock number the volume identifier header belongs to
  * @vid_hdr: the volume identifier header to check
  *
  * This function returns zero if the volume identifier header is all right, and
- * %1 if not.
+ * %-EINVAL if not.
  */
-static int paranoid_check_vid_hdr(const struct ubi_device *ubi, int pnum,
-                                 const struct ubi_vid_hdr *vid_hdr)
+static int self_check_vid_hdr(const struct ubi_device *ubi, int pnum,
+                             const struct ubi_vid_hdr *vid_hdr)
 {
        int err;
        uint32_t magic;
 
+       if (!ubi_dbg_chk_io(ubi))
+               return 0;
+
        magic = be32_to_cpu(vid_hdr->magic);
        if (magic != UBI_VID_HDR_MAGIC) {
                ubi_err("bad VID header magic %#08x at PEB %d, must be %#08x",
@@ -1152,36 +1236,38 @@ static int paranoid_check_vid_hdr(const struct ubi_device *ubi, int pnum,
 
        err = validate_vid_hdr(ubi, vid_hdr);
        if (err) {
-               ubi_err("paranoid check failed for PEB %d", pnum);
+               ubi_err("self-check failed for PEB %d", pnum);
                goto fail;
        }
 
        return err;
 
 fail:
-       ubi_err("paranoid check failed for PEB %d", pnum);
-       ubi_dbg_dump_vid_hdr(vid_hdr);
-       ubi_dbg_dump_stack();
-       return 1;
+       ubi_err("self-check failed for PEB %d", pnum);
+       ubi_dump_vid_hdr(vid_hdr);
+       dump_stack();
+       return -EINVAL;
 
 }
 
 /**
- * paranoid_check_peb_vid_hdr - check that the volume identifier header of a
- * physical eraseblock is in-place and is all right.
+ * self_check_peb_vid_hdr - check volume identifier header.
  * @ubi: UBI device description object
  * @pnum: the physical eraseblock number to check
  *
  * This function returns zero if the volume identifier header is all right,
- * %1 if not, and a negative error code if an error occurred.
+ * and a negative error code if not or if an error occurred.
  */
-static int paranoid_check_peb_vid_hdr(const struct ubi_device *ubi, int pnum)
+static int self_check_peb_vid_hdr(const struct ubi_device *ubi, int pnum)
 {
        int err;
        uint32_t crc, hdr_crc;
        struct ubi_vid_hdr *vid_hdr;
        void *p;
 
+       if (!ubi_dbg_chk_io(ubi))
+               return 0;
+
        vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_NOFS);
        if (!vid_hdr)
                return -ENOMEM;
@@ -1189,22 +1275,22 @@ static int paranoid_check_peb_vid_hdr(const struct ubi_device *ubi, int pnum)
        p = (char *)vid_hdr - ubi->vid_hdr_shift;
        err = ubi_io_read(ubi, p, pnum, ubi->vid_hdr_aloffset,
                          ubi->vid_hdr_alsize);
-       if (err && err != UBI_IO_BITFLIPS && err != -EBADMSG)
+       if (err && err != UBI_IO_BITFLIPS && !mtd_is_eccerr(err))
                goto exit;
 
        crc = crc32(UBI_CRC32_INIT, vid_hdr, UBI_EC_HDR_SIZE_CRC);
        hdr_crc = be32_to_cpu(vid_hdr->hdr_crc);
        if (hdr_crc != crc) {
-               ubi_err("bad VID header CRC at PEB %d, calculated %#08x, "
-                       "read %#08x", pnum, crc, hdr_crc);
-               ubi_err("paranoid check failed for PEB %d", pnum);
-               ubi_dbg_dump_vid_hdr(vid_hdr);
-               ubi_dbg_dump_stack();
-               err = 1;
+               ubi_err("bad VID header CRC at PEB %d, calculated %#08x, read %#08x",
+                       pnum, crc, hdr_crc);
+               ubi_err("self-check failed for PEB %d", pnum);
+               ubi_dump_vid_hdr(vid_hdr);
+               dump_stack();
+               err = -EINVAL;
                goto exit;
        }
 
-       err = paranoid_check_vid_hdr(ubi, pnum, vid_hdr);
+       err = self_check_vid_hdr(ubi, pnum, vid_hdr);
 
 exit:
        ubi_free_vid_hdr(ubi, vid_hdr);
@@ -1212,51 +1298,123 @@ exit:
 }
 
 /**
- * paranoid_check_all_ff - check that a region of flash is empty.
+ * self_check_write - make sure write succeeded.
+ * @ubi: UBI device description object
+ * @buf: buffer with data which were written
+ * @pnum: physical eraseblock number the data were written to
+ * @offset: offset within the physical eraseblock the data were written to
+ * @len: how many bytes were written
+ *
+ * This functions reads data which were recently written and compares it with
+ * the original data buffer - the data have to match. Returns zero if the data
+ * match and a negative error code if not or in case of failure.
+ */
+static int self_check_write(struct ubi_device *ubi, const void *buf, int pnum,
+                           int offset, int len)
+{
+       int err, i;
+       size_t read;
+       void *buf1;
+       loff_t addr = (loff_t)pnum * ubi->peb_size + offset;
+
+       if (!ubi_dbg_chk_io(ubi))
+               return 0;
+
+       buf1 = __vmalloc(len, GFP_NOFS, PAGE_KERNEL);
+       if (!buf1) {
+               ubi_err("cannot allocate memory to check writes");
+               return 0;
+       }
+
+       err = mtd_read(ubi->mtd, addr, len, &read, buf1);
+       if (err && !mtd_is_bitflip(err))
+               goto out_free;
+
+       for (i = 0; i < len; i++) {
+               uint8_t c = ((uint8_t *)buf)[i];
+               uint8_t c1 = ((uint8_t *)buf1)[i];
+#if !defined(CONFIG_UBI_SILENCE_MSG)
+               int dump_len = max_t(int, 128, len - i);
+#endif
+
+               if (c == c1)
+                       continue;
+
+               ubi_err("self-check failed for PEB %d:%d, len %d",
+                       pnum, offset, len);
+               ubi_msg("data differ at position %d", i);
+               ubi_msg("hex dump of the original buffer from %d to %d",
+                       i, i + dump_len);
+               print_hex_dump(KERN_DEBUG, "", DUMP_PREFIX_OFFSET, 32, 1,
+                              buf + i, dump_len, 1);
+               ubi_msg("hex dump of the read buffer from %d to %d",
+                       i, i + dump_len);
+               print_hex_dump(KERN_DEBUG, "", DUMP_PREFIX_OFFSET, 32, 1,
+                              buf1 + i, dump_len, 1);
+               dump_stack();
+               err = -EINVAL;
+               goto out_free;
+       }
+
+       vfree(buf1);
+       return 0;
+
+out_free:
+       vfree(buf1);
+       return err;
+}
+
+/**
+ * ubi_self_check_all_ff - check that a region of flash is empty.
  * @ubi: UBI device description object
  * @pnum: the physical eraseblock number to check
  * @offset: the starting offset within the physical eraseblock to check
  * @len: the length of the region to check
  *
  * This function returns zero if only 0xFF bytes are present at offset
- * @offset of the physical eraseblock @pnum, %1 if not, and a negative error
- * code if an error occurred.
+ * @offset of the physical eraseblock @pnum, and a negative error code if not
+ * or if an error occurred.
  */
-static int paranoid_check_all_ff(struct ubi_device *ubi, int pnum, int offset,
-                                int len)
+int ubi_self_check_all_ff(struct ubi_device *ubi, int pnum, int offset, int len)
 {
        size_t read;
        int err;
+       void *buf;
        loff_t addr = (loff_t)pnum * ubi->peb_size + offset;
 
-       mutex_lock(&ubi->dbg_buf_mutex);
-       err = mtd_read(ubi->mtd, addr, len, &read, ubi->dbg_peb_buf);
-       if (err && err != -EUCLEAN) {
-               ubi_err("error %d while reading %d bytes from PEB %d:%d, "
-                       "read %zd bytes", err, len, pnum, offset, read);
+       if (!ubi_dbg_chk_io(ubi))
+               return 0;
+
+       buf = __vmalloc(len, GFP_NOFS, PAGE_KERNEL);
+       if (!buf) {
+               ubi_err("cannot allocate memory to check for 0xFFs");
+               return 0;
+       }
+
+       err = mtd_read(ubi->mtd, addr, len, &read, buf);
+       if (err && !mtd_is_bitflip(err)) {
+               ubi_err("error %d while reading %d bytes from PEB %d:%d, read %zd bytes",
+                       err, len, pnum, offset, read);
                goto error;
        }
 
-       err = check_pattern(ubi->dbg_peb_buf, 0xFF, len);
+       err = ubi_check_pattern(buf, 0xFF, len);
        if (err == 0) {
-               ubi_err("flash region at PEB %d:%d, length %d does not "
-                       "contain all 0xFF bytes", pnum, offset, len);
+               ubi_err("flash region at PEB %d:%d, length %d does not contain all 0xFF bytes",
+                       pnum, offset, len);
                goto fail;
        }
-       mutex_unlock(&ubi->dbg_buf_mutex);
 
+       vfree(buf);
        return 0;
 
 fail:
-       ubi_err("paranoid check failed for PEB %d", pnum);
-       dbg_msg("hex dump of the %d-%d region", offset, offset + len);
-       print_hex_dump(KERN_DEBUG, "", DUMP_PREFIX_OFFSET, 32, 1,
-                      ubi->dbg_peb_buf, len, 1);
-       err = 1;
+       ubi_err("self-check failed for PEB %d", pnum);
+       ubi_msg("hex dump of the %d-%d region", offset, offset + len);
+       print_hex_dump(KERN_DEBUG, "", DUMP_PREFIX_OFFSET, 32, 1, buf, len, 1);
+       err = -EINVAL;
 error:
-       ubi_dbg_dump_stack();
-       mutex_unlock(&ubi->dbg_buf_mutex);
+       dump_stack();
+       vfree(buf);
        return err;
 }
-
-#endif /* CONFIG_MTD_UBI_DEBUG_PARANOID */
index 63c56c998e8d984daec5514db69a6faa69575511..0183c93b0b040560f43c7275844ccadf67b57e5a 100644 (file)
@@ -8,15 +8,42 @@
 
 /* This file mostly implements UBI kernel API functions */
 
-#ifdef UBI_LINUX
+#define __UBOOT__
+#ifndef __UBOOT__
 #include <linux/module.h>
-#include <linux/err.h>
+#include <linux/slab.h>
+#include <linux/namei.h>
+#include <linux/fs.h>
 #include <asm/div64.h>
+#else
+#include <ubi_uboot.h>
 #endif
+#include <linux/err.h>
 
-#include <ubi_uboot.h>
 #include "ubi.h"
 
+/**
+ * ubi_do_get_device_info - get information about UBI device.
+ * @ubi: UBI device description object
+ * @di: the information is stored here
+ *
+ * This function is the same as 'ubi_get_device_info()', but it assumes the UBI
+ * device is locked and cannot disappear.
+ */
+void ubi_do_get_device_info(struct ubi_device *ubi, struct ubi_device_info *di)
+{
+       di->ubi_num = ubi->ubi_num;
+       di->leb_size = ubi->leb_size;
+       di->leb_start = ubi->leb_start;
+       di->min_io_size = ubi->min_io_size;
+       di->max_write_size = ubi->max_write_size;
+       di->ro_mode = ubi->ro_mode;
+#ifndef __UBOOT__
+       di->cdev = ubi->cdev.dev;
+#endif
+}
+EXPORT_SYMBOL_GPL(ubi_do_get_device_info);
+
 /**
  * ubi_get_device_info - get information about UBI device.
  * @ubi_num: UBI device number
@@ -31,33 +58,24 @@ int ubi_get_device_info(int ubi_num, struct ubi_device_info *di)
 
        if (ubi_num < 0 || ubi_num >= UBI_MAX_DEVICES)
                return -EINVAL;
-
        ubi = ubi_get_device(ubi_num);
        if (!ubi)
                return -ENODEV;
-
-       di->ubi_num = ubi->ubi_num;
-       di->leb_size = ubi->leb_size;
-       di->min_io_size = ubi->min_io_size;
-       di->ro_mode = ubi->ro_mode;
-       di->cdev = ubi->cdev.dev;
-
+       ubi_do_get_device_info(ubi, di);
        ubi_put_device(ubi);
        return 0;
 }
 EXPORT_SYMBOL_GPL(ubi_get_device_info);
 
 /**
- * ubi_get_volume_info - get information about UBI volume.
- * @desc: volume descriptor
+ * ubi_do_get_volume_info - get information about UBI volume.
+ * @ubi: UBI device description object
+ * @vol: volume description object
  * @vi: the information is stored here
  */
-void ubi_get_volume_info(struct ubi_volume_desc *desc,
-                        struct ubi_volume_info *vi)
+void ubi_do_get_volume_info(struct ubi_device *ubi, struct ubi_volume *vol,
+                           struct ubi_volume_info *vi)
 {
-       const struct ubi_volume *vol = desc->vol;
-       const struct ubi_device *ubi = vol->ubi;
-
        vi->vol_id = vol->vol_id;
        vi->ubi_num = ubi->ubi_num;
        vi->size = vol->reserved_pebs;
@@ -71,6 +89,17 @@ void ubi_get_volume_info(struct ubi_volume_desc *desc,
        vi->name = vol->name;
        vi->cdev = vol->cdev.dev;
 }
+
+/**
+ * ubi_get_volume_info - get information about UBI volume.
+ * @desc: volume descriptor
+ * @vi: the information is stored here
+ */
+void ubi_get_volume_info(struct ubi_volume_desc *desc,
+                        struct ubi_volume_info *vi)
+{
+       ubi_do_get_volume_info(desc->vol->ubi, desc->vol, vi);
+}
 EXPORT_SYMBOL_GPL(ubi_get_volume_info);
 
 /**
@@ -98,7 +127,7 @@ struct ubi_volume_desc *ubi_open_volume(int ubi_num, int vol_id, int mode)
        struct ubi_device *ubi;
        struct ubi_volume *vol;
 
-       dbg_msg("open device %d volume %d, mode %d", ubi_num, vol_id, mode);
+       dbg_gen("open device %d, volume %d, mode %d", ubi_num, vol_id, mode);
 
        if (ubi_num < 0 || ubi_num >= UBI_MAX_DEVICES)
                return ERR_PTR(-EINVAL);
@@ -188,6 +217,8 @@ out_free:
        kfree(desc);
 out_put_ubi:
        ubi_put_device(ubi);
+       ubi_err("cannot open device %d, volume %d, error %d",
+               ubi_num, vol_id, err);
        return ERR_PTR(err);
 }
 EXPORT_SYMBOL_GPL(ubi_open_volume);
@@ -207,7 +238,7 @@ struct ubi_volume_desc *ubi_open_volume_nm(int ubi_num, const char *name,
        struct ubi_device *ubi;
        struct ubi_volume_desc *ret;
 
-       dbg_msg("open volume %s, mode %d", name, mode);
+       dbg_gen("open device %d, volume %s, mode %d", ubi_num, name, mode);
 
        if (!name)
                return ERR_PTR(-EINVAL);
@@ -249,6 +280,45 @@ struct ubi_volume_desc *ubi_open_volume_nm(int ubi_num, const char *name,
 }
 EXPORT_SYMBOL_GPL(ubi_open_volume_nm);
 
+#ifndef __UBOOT__
+/**
+ * ubi_open_volume_path - open UBI volume by its character device node path.
+ * @pathname: volume character device node path
+ * @mode: open mode
+ *
+ * This function is similar to 'ubi_open_volume()', but opens a volume the path
+ * to its character device node.
+ */
+struct ubi_volume_desc *ubi_open_volume_path(const char *pathname, int mode)
+{
+       int error, ubi_num, vol_id, mod;
+       struct inode *inode;
+       struct path path;
+
+       dbg_gen("open volume %s, mode %d", pathname, mode);
+
+       if (!pathname || !*pathname)
+               return ERR_PTR(-EINVAL);
+
+       error = kern_path(pathname, LOOKUP_FOLLOW, &path);
+       if (error)
+               return ERR_PTR(error);
+
+       inode = path.dentry->d_inode;
+       mod = inode->i_mode;
+       ubi_num = ubi_major2num(imajor(inode));
+       vol_id = iminor(inode) - 1;
+       path_put(&path);
+
+       if (!S_ISCHR(mod))
+               return ERR_PTR(-EINVAL);
+       if (vol_id >= 0 && ubi_num >= 0)
+               return ubi_open_volume(ubi_num, vol_id, mode);
+       return ERR_PTR(-ENODEV);
+}
+EXPORT_SYMBOL_GPL(ubi_open_volume_path);
+#endif
+
 /**
  * ubi_close_volume - close UBI volume.
  * @desc: volume descriptor
@@ -258,7 +328,8 @@ void ubi_close_volume(struct ubi_volume_desc *desc)
        struct ubi_volume *vol = desc->vol;
        struct ubi_device *ubi = vol->ubi;
 
-       dbg_msg("close volume %d, mode %d", vol->vol_id, desc->mode);
+       dbg_gen("close device %d, volume %d, mode %d",
+               ubi->ubi_num, vol->vol_id, desc->mode);
 
        spin_lock(&ubi->volumes_lock);
        switch (desc->mode) {
@@ -315,7 +386,7 @@ int ubi_leb_read(struct ubi_volume_desc *desc, int lnum, char *buf, int offset,
        struct ubi_device *ubi = vol->ubi;
        int err, vol_id = vol->vol_id;
 
-       dbg_msg("read %d bytes from LEB %d:%d:%d", len, vol_id, lnum, offset);
+       dbg_gen("read %d bytes from LEB %d:%d:%d", len, vol_id, lnum, offset);
 
        if (vol_id < 0 || vol_id >= ubi->vtbl_slots || lnum < 0 ||
            lnum >= vol->used_ebs || offset < 0 || len < 0 ||
@@ -353,11 +424,9 @@ EXPORT_SYMBOL_GPL(ubi_leb_read);
  * @buf: data to write
  * @offset: offset within the logical eraseblock where to write
  * @len: how many bytes to write
- * @dtype: expected data type
  *
  * This function writes @len bytes of data from @buf to offset @offset of
- * logical eraseblock @lnum. The @dtype argument describes expected lifetime of
- * the data.
+ * logical eraseblock @lnum.
  *
  * This function takes care of physical eraseblock write failures. If write to
  * the physical eraseblock write operation fails, the logical eraseblock is
@@ -374,13 +443,13 @@ EXPORT_SYMBOL_GPL(ubi_leb_read);
  * returns immediately with %-EBADF code.
  */
 int ubi_leb_write(struct ubi_volume_desc *desc, int lnum, const void *buf,
-                 int offset, int len, int dtype)
+                 int offset, int len)
 {
        struct ubi_volume *vol = desc->vol;
        struct ubi_device *ubi = vol->ubi;
        int vol_id = vol->vol_id;
 
-       dbg_msg("write %d bytes to LEB %d:%d:%d", len, vol_id, lnum, offset);
+       dbg_gen("write %d bytes to LEB %d:%d:%d", len, vol_id, lnum, offset);
 
        if (vol_id < 0 || vol_id >= ubi->vtbl_slots)
                return -EINVAL;
@@ -393,17 +462,13 @@ int ubi_leb_write(struct ubi_volume_desc *desc, int lnum, const void *buf,
            offset & (ubi->min_io_size - 1) || len & (ubi->min_io_size - 1))
                return -EINVAL;
 
-       if (dtype != UBI_LONGTERM && dtype != UBI_SHORTTERM &&
-           dtype != UBI_UNKNOWN)
-               return -EINVAL;
-
        if (vol->upd_marker)
                return -EBADF;
 
        if (len == 0)
                return 0;
 
-       return ubi_eba_write_leb(ubi, vol, lnum, buf, offset, len, dtype);
+       return ubi_eba_write_leb(ubi, vol, lnum, buf, offset, len);
 }
 EXPORT_SYMBOL_GPL(ubi_leb_write);
 
@@ -413,24 +478,23 @@ EXPORT_SYMBOL_GPL(ubi_leb_write);
  * @lnum: logical eraseblock number to change
  * @buf: data to write
  * @len: how many bytes to write
- * @dtype: expected data type
  *
  * This function changes the contents of a logical eraseblock atomically. @buf
  * has to contain new logical eraseblock data, and @len - the length of the
- * data, which has to be aligned. The length may be shorter then the logical
+ * data, which has to be aligned. The length may be shorter than the logical
  * eraseblock size, ant the logical eraseblock may be appended to more times
  * later on. This function guarantees that in case of an unclean reboot the old
  * contents is preserved. Returns zero in case of success and a negative error
  * code in case of failure.
  */
 int ubi_leb_change(struct ubi_volume_desc *desc, int lnum, const void *buf,
-                  int len, int dtype)
+                  int len)
 {
        struct ubi_volume *vol = desc->vol;
        struct ubi_device *ubi = vol->ubi;
        int vol_id = vol->vol_id;
 
-       dbg_msg("atomically write %d bytes to LEB %d:%d", len, vol_id, lnum);
+       dbg_gen("atomically write %d bytes to LEB %d:%d", len, vol_id, lnum);
 
        if (vol_id < 0 || vol_id >= ubi->vtbl_slots)
                return -EINVAL;
@@ -442,17 +506,13 @@ int ubi_leb_change(struct ubi_volume_desc *desc, int lnum, const void *buf,
            len > vol->usable_leb_size || len & (ubi->min_io_size - 1))
                return -EINVAL;
 
-       if (dtype != UBI_LONGTERM && dtype != UBI_SHORTTERM &&
-           dtype != UBI_UNKNOWN)
-               return -EINVAL;
-
        if (vol->upd_marker)
                return -EBADF;
 
        if (len == 0)
                return 0;
 
-       return ubi_eba_atomic_leb_change(ubi, vol, lnum, buf, len, dtype);
+       return ubi_eba_atomic_leb_change(ubi, vol, lnum, buf, len);
 }
 EXPORT_SYMBOL_GPL(ubi_leb_change);
 
@@ -474,7 +534,7 @@ int ubi_leb_erase(struct ubi_volume_desc *desc, int lnum)
        struct ubi_device *ubi = vol->ubi;
        int err;
 
-       dbg_msg("erase LEB %d:%d", vol->vol_id, lnum);
+       dbg_gen("erase LEB %d:%d", vol->vol_id, lnum);
 
        if (desc->mode == UBI_READONLY || vol->vol_type == UBI_STATIC_VOLUME)
                return -EROFS;
@@ -489,7 +549,7 @@ int ubi_leb_erase(struct ubi_volume_desc *desc, int lnum)
        if (err)
                return err;
 
-       return ubi_wl_flush(ubi);
+       return ubi_wl_flush(ubi, vol->vol_id, lnum);
 }
 EXPORT_SYMBOL_GPL(ubi_leb_erase);
 
@@ -500,7 +560,7 @@ EXPORT_SYMBOL_GPL(ubi_leb_erase);
  *
  * This function un-maps logical eraseblock @lnum and schedules the
  * corresponding physical eraseblock for erasure, so that it will eventually be
- * physically erased in background. This operation is much faster then the
+ * physically erased in background. This operation is much faster than the
  * erase operation.
  *
  * Unlike erase, the un-map operation does not guarantee that the logical
@@ -519,7 +579,7 @@ EXPORT_SYMBOL_GPL(ubi_leb_erase);
  *
  * The main and obvious use-case of this function is when the contents of a
  * logical eraseblock has to be re-written. Then it is much more efficient to
- * first un-map it, then write new data, rather then first erase it, then write
+ * first un-map it, then write new data, rather than first erase it, then write
  * new data. Note, once new data has been written to the logical eraseblock,
  * UBI guarantees that the old contents has gone forever. In other words, if an
  * unclean reboot happens after the logical eraseblock has been un-mapped and
@@ -534,7 +594,7 @@ int ubi_leb_unmap(struct ubi_volume_desc *desc, int lnum)
        struct ubi_volume *vol = desc->vol;
        struct ubi_device *ubi = vol->ubi;
 
-       dbg_msg("unmap LEB %d:%d", vol->vol_id, lnum);
+       dbg_gen("unmap LEB %d:%d", vol->vol_id, lnum);
 
        if (desc->mode == UBI_READONLY || vol->vol_type == UBI_STATIC_VOLUME)
                return -EROFS;
@@ -550,13 +610,12 @@ int ubi_leb_unmap(struct ubi_volume_desc *desc, int lnum)
 EXPORT_SYMBOL_GPL(ubi_leb_unmap);
 
 /**
- * ubi_leb_map - map logical erasblock to a physical eraseblock.
+ * ubi_leb_map - map logical eraseblock to a physical eraseblock.
  * @desc: volume descriptor
  * @lnum: logical eraseblock number
- * @dtype: expected data type
  *
  * This function maps an un-mapped logical eraseblock @lnum to a physical
- * eraseblock. This means, that after a successfull invocation of this
+ * eraseblock. This means, that after a successful invocation of this
  * function the logical eraseblock @lnum will be empty (contain only %0xFF
  * bytes) and be mapped to a physical eraseblock, even if an unclean reboot
  * happens.
@@ -566,12 +625,12 @@ EXPORT_SYMBOL_GPL(ubi_leb_unmap);
  * eraseblock is already mapped, and other negative error codes in case of
  * other failures.
  */
-int ubi_leb_map(struct ubi_volume_desc *desc, int lnum, int dtype)
+int ubi_leb_map(struct ubi_volume_desc *desc, int lnum)
 {
        struct ubi_volume *vol = desc->vol;
        struct ubi_device *ubi = vol->ubi;
 
-       dbg_msg("unmap LEB %d:%d", vol->vol_id, lnum);
+       dbg_gen("unmap LEB %d:%d", vol->vol_id, lnum);
 
        if (desc->mode == UBI_READONLY || vol->vol_type == UBI_STATIC_VOLUME)
                return -EROFS;
@@ -579,17 +638,13 @@ int ubi_leb_map(struct ubi_volume_desc *desc, int lnum, int dtype)
        if (lnum < 0 || lnum >= vol->reserved_pebs)
                return -EINVAL;
 
-       if (dtype != UBI_LONGTERM && dtype != UBI_SHORTTERM &&
-           dtype != UBI_UNKNOWN)
-               return -EINVAL;
-
        if (vol->upd_marker)
                return -EBADF;
 
        if (vol->eba_tbl[lnum] >= 0)
                return -EBADMSG;
 
-       return ubi_eba_write_leb(ubi, vol, lnum, NULL, 0, 0, dtype);
+       return ubi_eba_write_leb(ubi, vol, lnum, NULL, 0, 0);
 }
 EXPORT_SYMBOL_GPL(ubi_leb_map);
 
@@ -613,7 +668,7 @@ int ubi_is_mapped(struct ubi_volume_desc *desc, int lnum)
 {
        struct ubi_volume *vol = desc->vol;
 
-       dbg_msg("test LEB %d:%d", vol->vol_id, lnum);
+       dbg_gen("test LEB %d:%d", vol->vol_id, lnum);
 
        if (lnum < 0 || lnum >= vol->reserved_pebs)
                return -EINVAL;
@@ -624,3 +679,110 @@ int ubi_is_mapped(struct ubi_volume_desc *desc, int lnum)
        return vol->eba_tbl[lnum] >= 0;
 }
 EXPORT_SYMBOL_GPL(ubi_is_mapped);
+
+/**
+ * ubi_sync - synchronize UBI device buffers.
+ * @ubi_num: UBI device to synchronize
+ *
+ * The underlying MTD device may cache data in hardware or in software. This
+ * function ensures the caches are flushed. Returns zero in case of success and
+ * a negative error code in case of failure.
+ */
+int ubi_sync(int ubi_num)
+{
+       struct ubi_device *ubi;
+
+       ubi = ubi_get_device(ubi_num);
+       if (!ubi)
+               return -ENODEV;
+
+       mtd_sync(ubi->mtd);
+       ubi_put_device(ubi);
+       return 0;
+}
+EXPORT_SYMBOL_GPL(ubi_sync);
+
+/**
+ * ubi_flush - flush UBI work queue.
+ * @ubi_num: UBI device to flush work queue
+ * @vol_id: volume id to flush for
+ * @lnum: logical eraseblock number to flush for
+ *
+ * This function executes all pending works for a particular volume id / logical
+ * eraseblock number pair. If either value is set to %UBI_ALL, then it acts as
+ * a wildcard for all of the corresponding volume numbers or logical
+ * eraseblock numbers. It returns zero in case of success and a negative error
+ * code in case of failure.
+ */
+int ubi_flush(int ubi_num, int vol_id, int lnum)
+{
+       struct ubi_device *ubi;
+       int err = 0;
+
+       ubi = ubi_get_device(ubi_num);
+       if (!ubi)
+               return -ENODEV;
+
+       err = ubi_wl_flush(ubi, vol_id, lnum);
+       ubi_put_device(ubi);
+       return err;
+}
+EXPORT_SYMBOL_GPL(ubi_flush);
+
+#ifndef __UBOOT__
+BLOCKING_NOTIFIER_HEAD(ubi_notifiers);
+
+/**
+ * ubi_register_volume_notifier - register a volume notifier.
+ * @nb: the notifier description object
+ * @ignore_existing: if non-zero, do not send "added" notification for all
+ *                   already existing volumes
+ *
+ * This function registers a volume notifier, which means that
+ * 'nb->notifier_call()' will be invoked when an UBI  volume is created,
+ * removed, re-sized, re-named, or updated. The first argument of the function
+ * is the notification type. The second argument is pointer to a
+ * &struct ubi_notification object which describes the notification event.
+ * Using UBI API from the volume notifier is prohibited.
+ *
+ * This function returns zero in case of success and a negative error code
+ * in case of failure.
+ */
+int ubi_register_volume_notifier(struct notifier_block *nb,
+                                int ignore_existing)
+{
+       int err;
+
+       err = blocking_notifier_chain_register(&ubi_notifiers, nb);
+       if (err != 0)
+               return err;
+       if (ignore_existing)
+               return 0;
+
+       /*
+        * We are going to walk all UBI devices and all volumes, and
+        * notify the user about existing volumes by the %UBI_VOLUME_ADDED
+        * event. We have to lock the @ubi_devices_mutex to make sure UBI
+        * devices do not disappear.
+        */
+       mutex_lock(&ubi_devices_mutex);
+       ubi_enumerate_volumes(nb);
+       mutex_unlock(&ubi_devices_mutex);
+
+       return err;
+}
+EXPORT_SYMBOL_GPL(ubi_register_volume_notifier);
+
+/**
+ * ubi_unregister_volume_notifier - unregister the volume notifier.
+ * @nb: the notifier description object
+ *
+ * This function unregisters volume notifier @nm and returns zero in case of
+ * success and a negative error code in case of failure.
+ */
+int ubi_unregister_volume_notifier(struct notifier_block *nb)
+{
+       return blocking_notifier_chain_unregister(&ubi_notifiers, nb);
+}
+EXPORT_SYMBOL_GPL(ubi_unregister_volume_notifier);
+#endif
index 5ff55b4f77fea96c0a7236d82e4ad4a24e183fcc..49530b7448f9df8d7653658682b1a5dafe53b7ce 100644 (file)
@@ -81,14 +81,62 @@ int ubi_check_volume(struct ubi_device *ubi, int vol_id)
 }
 
 /**
- * ubi_calculate_rsvd_pool - calculate how many PEBs must be reserved for bad
+ * ubi_update_reserved - update bad eraseblock handling accounting data.
+ * @ubi: UBI device description object
+ *
+ * This function calculates the gap between current number of PEBs reserved for
+ * bad eraseblock handling and the required level of PEBs that must be
+ * reserved, and if necessary, reserves more PEBs to fill that gap, according
+ * to availability. Should be called with ubi->volumes_lock held.
+ */
+void ubi_update_reserved(struct ubi_device *ubi)
+{
+       int need = ubi->beb_rsvd_level - ubi->beb_rsvd_pebs;
+
+       if (need <= 0 || ubi->avail_pebs == 0)
+               return;
+
+       need = min_t(int, need, ubi->avail_pebs);
+       ubi->avail_pebs -= need;
+       ubi->rsvd_pebs += need;
+       ubi->beb_rsvd_pebs += need;
+       ubi_msg("reserved more %d PEBs for bad PEB handling", need);
+}
+
+/**
+ * ubi_calculate_reserved - calculate how many PEBs must be reserved for bad
  * eraseblock handling.
  * @ubi: UBI device description object
  */
 void ubi_calculate_reserved(struct ubi_device *ubi)
 {
-       ubi->beb_rsvd_level = ubi->good_peb_count/100;
-       ubi->beb_rsvd_level *= CONFIG_MTD_UBI_BEB_RESERVE;
-       if (ubi->beb_rsvd_level < MIN_RESEVED_PEBS)
-               ubi->beb_rsvd_level = MIN_RESEVED_PEBS;
+       /*
+        * Calculate the actual number of PEBs currently needed to be reserved
+        * for future bad eraseblock handling.
+        */
+       ubi->beb_rsvd_level = ubi->bad_peb_limit - ubi->bad_peb_count;
+       if (ubi->beb_rsvd_level < 0) {
+               ubi->beb_rsvd_level = 0;
+               ubi_warn("number of bad PEBs (%d) is above the expected limit (%d), not reserving any PEBs for bad PEB handling, will use available PEBs (if any)",
+                        ubi->bad_peb_count, ubi->bad_peb_limit);
+       }
+}
+
+/**
+ * ubi_check_pattern - check if buffer contains only a certain byte pattern.
+ * @buf: buffer to check
+ * @patt: the pattern to check
+ * @size: buffer size in bytes
+ *
+ * This function returns %1 in there are only @patt bytes in @buf, and %0 if
+ * something else was also found.
+ */
+int ubi_check_pattern(const void *buf, uint8_t patt, int size)
+{
+       int i;
+
+       for (i = 0; i < size; i++)
+               if (((const uint8_t *)buf)[i] != patt)
+                       return 0;
+       return 1;
 }
diff --git a/drivers/mtd/ubi/scan.c b/drivers/mtd/ubi/scan.c
deleted file mode 100644 (file)
index a6d0fbc..0000000
+++ /dev/null
@@ -1,1348 +0,0 @@
-/*
- * Copyright (c) International Business Machines Corp., 2006
- *
- * SPDX-License-Identifier:    GPL-2.0+
- *
- * Author: Artem Bityutskiy (Битюцкий Артём)
- */
-
-/*
- * UBI scanning unit.
- *
- * This unit is responsible for scanning the flash media, checking UBI
- * headers and providing complete information about the UBI flash image.
- *
- * The scanning information is represented by a &struct ubi_scan_info' object.
- * Information about found volumes is represented by &struct ubi_scan_volume
- * objects which are kept in volume RB-tree with root at the @volumes field.
- * The RB-tree is indexed by the volume ID.
- *
- * Found logical eraseblocks are represented by &struct ubi_scan_leb objects.
- * These objects are kept in per-volume RB-trees with the root at the
- * corresponding &struct ubi_scan_volume object. To put it differently, we keep
- * an RB-tree of per-volume objects and each of these objects is the root of
- * RB-tree of per-eraseblock objects.
- *
- * Corrupted physical eraseblocks are put to the @corr list, free physical
- * eraseblocks are put to the @free list and the physical eraseblock to be
- * erased are put to the @erase list.
- */
-
-#ifdef UBI_LINUX
-#include <linux/err.h>
-#include <linux/crc32.h>
-#include <asm/div64.h>
-#endif
-
-#include <ubi_uboot.h>
-#include "ubi.h"
-
-#ifdef CONFIG_MTD_UBI_DEBUG_PARANOID
-static int paranoid_check_si(struct ubi_device *ubi, struct ubi_scan_info *si);
-#else
-#define paranoid_check_si(ubi, si) 0
-#endif
-
-/* Temporary variables used during scanning */
-static struct ubi_ec_hdr *ech;
-static struct ubi_vid_hdr *vidh;
-
-/**
- * add_to_list - add physical eraseblock to a list.
- * @si: scanning information
- * @pnum: physical eraseblock number to add
- * @ec: erase counter of the physical eraseblock
- * @list: the list to add to
- *
- * This function adds physical eraseblock @pnum to free, erase, corrupted or
- * alien lists. Returns zero in case of success and a negative error code in
- * case of failure.
- */
-static int add_to_list(struct ubi_scan_info *si, int pnum, int ec,
-                      struct list_head *list)
-{
-       struct ubi_scan_leb *seb;
-
-       if (list == &si->free)
-               dbg_bld("add to free: PEB %d, EC %d", pnum, ec);
-       else if (list == &si->erase)
-               dbg_bld("add to erase: PEB %d, EC %d", pnum, ec);
-       else if (list == &si->corr)
-               dbg_bld("add to corrupted: PEB %d, EC %d", pnum, ec);
-       else if (list == &si->alien)
-               dbg_bld("add to alien: PEB %d, EC %d", pnum, ec);
-       else
-               BUG();
-
-       seb = kmalloc(sizeof(struct ubi_scan_leb), GFP_KERNEL);
-       if (!seb)
-               return -ENOMEM;
-
-       seb->pnum = pnum;
-       seb->ec = ec;
-       list_add_tail(&seb->u.list, list);
-       return 0;
-}
-
-/**
- * validate_vid_hdr - check that volume identifier header is correct and
- * consistent.
- * @vid_hdr: the volume identifier header to check
- * @sv: information about the volume this logical eraseblock belongs to
- * @pnum: physical eraseblock number the VID header came from
- *
- * This function checks that data stored in @vid_hdr is consistent. Returns
- * non-zero if an inconsistency was found and zero if not.
- *
- * Note, UBI does sanity check of everything it reads from the flash media.
- * Most of the checks are done in the I/O unit. Here we check that the
- * information in the VID header is consistent to the information in other VID
- * headers of the same volume.
- */
-static int validate_vid_hdr(const struct ubi_vid_hdr *vid_hdr,
-                           const struct ubi_scan_volume *sv, int pnum)
-{
-       int vol_type = vid_hdr->vol_type;
-       int vol_id = be32_to_cpu(vid_hdr->vol_id);
-       int used_ebs = be32_to_cpu(vid_hdr->used_ebs);
-       int data_pad = be32_to_cpu(vid_hdr->data_pad);
-
-       if (sv->leb_count != 0) {
-               int sv_vol_type;
-
-               /*
-                * This is not the first logical eraseblock belonging to this
-                * volume. Ensure that the data in its VID header is consistent
-                * to the data in previous logical eraseblock headers.
-                */
-
-               if (vol_id != sv->vol_id) {
-                       dbg_err("inconsistent vol_id");
-                       goto bad;
-               }
-
-               if (sv->vol_type == UBI_STATIC_VOLUME)
-                       sv_vol_type = UBI_VID_STATIC;
-               else
-                       sv_vol_type = UBI_VID_DYNAMIC;
-
-               if (vol_type != sv_vol_type) {
-                       dbg_err("inconsistent vol_type");
-                       goto bad;
-               }
-
-               if (used_ebs != sv->used_ebs) {
-                       dbg_err("inconsistent used_ebs");
-                       goto bad;
-               }
-
-               if (data_pad != sv->data_pad) {
-                       dbg_err("inconsistent data_pad");
-                       goto bad;
-               }
-       }
-
-       return 0;
-
-bad:
-       ubi_err("inconsistent VID header at PEB %d", pnum);
-       ubi_dbg_dump_vid_hdr(vid_hdr);
-       ubi_dbg_dump_sv(sv);
-       return -EINVAL;
-}
-
-/**
- * add_volume - add volume to the scanning information.
- * @si: scanning information
- * @vol_id: ID of the volume to add
- * @pnum: physical eraseblock number
- * @vid_hdr: volume identifier header
- *
- * If the volume corresponding to the @vid_hdr logical eraseblock is already
- * present in the scanning information, this function does nothing. Otherwise
- * it adds corresponding volume to the scanning information. Returns a pointer
- * to the scanning volume object in case of success and a negative error code
- * in case of failure.
- */
-static struct ubi_scan_volume *add_volume(struct ubi_scan_info *si, int vol_id,
-                                         int pnum,
-                                         const struct ubi_vid_hdr *vid_hdr)
-{
-       struct ubi_scan_volume *sv;
-       struct rb_node **p = &si->volumes.rb_node, *parent = NULL;
-
-       ubi_assert(vol_id == be32_to_cpu(vid_hdr->vol_id));
-
-       /* Walk the volume RB-tree to look if this volume is already present */
-       while (*p) {
-               parent = *p;
-               sv = rb_entry(parent, struct ubi_scan_volume, rb);
-
-               if (vol_id == sv->vol_id)
-                       return sv;
-
-               if (vol_id > sv->vol_id)
-                       p = &(*p)->rb_left;
-               else
-                       p = &(*p)->rb_right;
-       }
-
-       /* The volume is absent - add it */
-       sv = kmalloc(sizeof(struct ubi_scan_volume), GFP_KERNEL);
-       if (!sv)
-               return ERR_PTR(-ENOMEM);
-
-       sv->highest_lnum = sv->leb_count = 0;
-       sv->vol_id = vol_id;
-       sv->root = RB_ROOT;
-       sv->used_ebs = be32_to_cpu(vid_hdr->used_ebs);
-       sv->data_pad = be32_to_cpu(vid_hdr->data_pad);
-       sv->compat = vid_hdr->compat;
-       sv->vol_type = vid_hdr->vol_type == UBI_VID_DYNAMIC ? UBI_DYNAMIC_VOLUME
-                                                           : UBI_STATIC_VOLUME;
-       if (vol_id > si->highest_vol_id)
-               si->highest_vol_id = vol_id;
-
-       rb_link_node(&sv->rb, parent, p);
-       rb_insert_color(&sv->rb, &si->volumes);
-       si->vols_found += 1;
-       dbg_bld("added volume %d", vol_id);
-       return sv;
-}
-
-/**
- * compare_lebs - find out which logical eraseblock is newer.
- * @ubi: UBI device description object
- * @seb: first logical eraseblock to compare
- * @pnum: physical eraseblock number of the second logical eraseblock to
- * compare
- * @vid_hdr: volume identifier header of the second logical eraseblock
- *
- * This function compares 2 copies of a LEB and informs which one is newer. In
- * case of success this function returns a positive value, in case of failure, a
- * negative error code is returned. The success return codes use the following
- * bits:
- *     o bit 0 is cleared: the first PEB (described by @seb) is newer then the
- *       second PEB (described by @pnum and @vid_hdr);
- *     o bit 0 is set: the second PEB is newer;
- *     o bit 1 is cleared: no bit-flips were detected in the newer LEB;
- *     o bit 1 is set: bit-flips were detected in the newer LEB;
- *     o bit 2 is cleared: the older LEB is not corrupted;
- *     o bit 2 is set: the older LEB is corrupted.
- */
-static int compare_lebs(struct ubi_device *ubi, const struct ubi_scan_leb *seb,
-                       int pnum, const struct ubi_vid_hdr *vid_hdr)
-{
-       void *buf;
-       int len, err, second_is_newer, bitflips = 0, corrupted = 0;
-       uint32_t data_crc, crc;
-       struct ubi_vid_hdr *vh = NULL;
-       unsigned long long sqnum2 = be64_to_cpu(vid_hdr->sqnum);
-
-       if (seb->sqnum == 0 && sqnum2 == 0) {
-               long long abs, v1 = seb->leb_ver, v2 = be32_to_cpu(vid_hdr->leb_ver);
-
-               /*
-                * UBI constantly increases the logical eraseblock version
-                * number and it can overflow. Thus, we have to bear in mind
-                * that versions that are close to %0xFFFFFFFF are less then
-                * versions that are close to %0.
-                *
-                * The UBI WL unit guarantees that the number of pending tasks
-                * is not greater then %0x7FFFFFFF. So, if the difference
-                * between any two versions is greater or equivalent to
-                * %0x7FFFFFFF, there was an overflow and the logical
-                * eraseblock with lower version is actually newer then the one
-                * with higher version.
-                *
-                * FIXME: but this is anyway obsolete and will be removed at
-                * some point.
-                */
-               dbg_bld("using old crappy leb_ver stuff");
-
-               if (v1 == v2) {
-                       ubi_err("PEB %d and PEB %d have the same version %lld",
-                               seb->pnum, pnum, v1);
-                       return -EINVAL;
-               }
-
-               abs = v1 - v2;
-               if (abs < 0)
-                       abs = -abs;
-
-               if (abs < 0x7FFFFFFF)
-                       /* Non-overflow situation */
-                       second_is_newer = (v2 > v1);
-               else
-                       second_is_newer = (v2 < v1);
-       } else
-               /* Obviously the LEB with lower sequence counter is older */
-               second_is_newer = sqnum2 > seb->sqnum;
-
-       /*
-        * Now we know which copy is newer. If the copy flag of the PEB with
-        * newer version is not set, then we just return, otherwise we have to
-        * check data CRC. For the second PEB we already have the VID header,
-        * for the first one - we'll need to re-read it from flash.
-        *
-        * FIXME: this may be optimized so that we wouldn't read twice.
-        */
-
-       if (second_is_newer) {
-               if (!vid_hdr->copy_flag) {
-                       /* It is not a copy, so it is newer */
-                       dbg_bld("second PEB %d is newer, copy_flag is unset",
-                               pnum);
-                       return 1;
-               }
-       } else {
-               pnum = seb->pnum;
-
-               vh = ubi_zalloc_vid_hdr(ubi, GFP_KERNEL);
-               if (!vh)
-                       return -ENOMEM;
-
-               err = ubi_io_read_vid_hdr(ubi, pnum, vh, 0);
-               if (err) {
-                       if (err == UBI_IO_BITFLIPS)
-                               bitflips = 1;
-                       else {
-                               dbg_err("VID of PEB %d header is bad, but it "
-                                       "was OK earlier", pnum);
-                               if (err > 0)
-                                       err = -EIO;
-
-                               goto out_free_vidh;
-                       }
-               }
-
-               if (!vh->copy_flag) {
-                       /* It is not a copy, so it is newer */
-                       dbg_bld("first PEB %d is newer, copy_flag is unset",
-                               pnum);
-                       err = bitflips << 1;
-                       goto out_free_vidh;
-               }
-
-               vid_hdr = vh;
-       }
-
-       /* Read the data of the copy and check the CRC */
-
-       len = be32_to_cpu(vid_hdr->data_size);
-       buf = vmalloc(len);
-       if (!buf) {
-               err = -ENOMEM;
-               goto out_free_vidh;
-       }
-
-       err = ubi_io_read_data(ubi, buf, pnum, 0, len);
-       if (err && err != UBI_IO_BITFLIPS)
-               goto out_free_buf;
-
-       data_crc = be32_to_cpu(vid_hdr->data_crc);
-       crc = crc32(UBI_CRC32_INIT, buf, len);
-       if (crc != data_crc) {
-               dbg_bld("PEB %d CRC error: calculated %#08x, must be %#08x",
-                       pnum, crc, data_crc);
-               corrupted = 1;
-               bitflips = 0;
-               second_is_newer = !second_is_newer;
-       } else {
-               dbg_bld("PEB %d CRC is OK", pnum);
-               bitflips = !!err;
-       }
-
-       vfree(buf);
-       ubi_free_vid_hdr(ubi, vh);
-
-       if (second_is_newer)
-               dbg_bld("second PEB %d is newer, copy_flag is set", pnum);
-       else
-               dbg_bld("first PEB %d is newer, copy_flag is set", pnum);
-
-       return second_is_newer | (bitflips << 1) | (corrupted << 2);
-
-out_free_buf:
-       vfree(buf);
-out_free_vidh:
-       ubi_free_vid_hdr(ubi, vh);
-       return err;
-}
-
-/**
- * ubi_scan_add_used - add information about a physical eraseblock to the
- * scanning information.
- * @ubi: UBI device description object
- * @si: scanning information
- * @pnum: the physical eraseblock number
- * @ec: erase counter
- * @vid_hdr: the volume identifier header
- * @bitflips: if bit-flips were detected when this physical eraseblock was read
- *
- * This function adds information about a used physical eraseblock to the
- * 'used' tree of the corresponding volume. The function is rather complex
- * because it has to handle cases when this is not the first physical
- * eraseblock belonging to the same logical eraseblock, and the newer one has
- * to be picked, while the older one has to be dropped. This function returns
- * zero in case of success and a negative error code in case of failure.
- */
-int ubi_scan_add_used(struct ubi_device *ubi, struct ubi_scan_info *si,
-                     int pnum, int ec, const struct ubi_vid_hdr *vid_hdr,
-                     int bitflips)
-{
-       int err, vol_id, lnum;
-       uint32_t leb_ver;
-       unsigned long long sqnum;
-       struct ubi_scan_volume *sv;
-       struct ubi_scan_leb *seb;
-       struct rb_node **p, *parent = NULL;
-
-       vol_id = be32_to_cpu(vid_hdr->vol_id);
-       lnum = be32_to_cpu(vid_hdr->lnum);
-       sqnum = be64_to_cpu(vid_hdr->sqnum);
-       leb_ver = be32_to_cpu(vid_hdr->leb_ver);
-
-       dbg_bld("PEB %d, LEB %d:%d, EC %d, sqnum %llu, ver %u, bitflips %d",
-               pnum, vol_id, lnum, ec, sqnum, leb_ver, bitflips);
-
-       sv = add_volume(si, vol_id, pnum, vid_hdr);
-       if (IS_ERR(sv) < 0)
-               return PTR_ERR(sv);
-
-       if (si->max_sqnum < sqnum)
-               si->max_sqnum = sqnum;
-
-       /*
-        * Walk the RB-tree of logical eraseblocks of volume @vol_id to look
-        * if this is the first instance of this logical eraseblock or not.
-        */
-       p = &sv->root.rb_node;
-       while (*p) {
-               int cmp_res;
-
-               parent = *p;
-               seb = rb_entry(parent, struct ubi_scan_leb, u.rb);
-               if (lnum != seb->lnum) {
-                       if (lnum < seb->lnum)
-                               p = &(*p)->rb_left;
-                       else
-                               p = &(*p)->rb_right;
-                       continue;
-               }
-
-               /*
-                * There is already a physical eraseblock describing the same
-                * logical eraseblock present.
-                */
-
-               dbg_bld("this LEB already exists: PEB %d, sqnum %llu, "
-                       "LEB ver %u, EC %d", seb->pnum, seb->sqnum,
-                       seb->leb_ver, seb->ec);
-
-               /*
-                * Make sure that the logical eraseblocks have different
-                * versions. Otherwise the image is bad.
-                */
-               if (seb->leb_ver == leb_ver && leb_ver != 0) {
-                       ubi_err("two LEBs with same version %u", leb_ver);
-                       ubi_dbg_dump_seb(seb, 0);
-                       ubi_dbg_dump_vid_hdr(vid_hdr);
-                       return -EINVAL;
-               }
-
-               /*
-                * Make sure that the logical eraseblocks have different
-                * sequence numbers. Otherwise the image is bad.
-                *
-                * FIXME: remove 'sqnum != 0' check when leb_ver is removed.
-                */
-               if (seb->sqnum == sqnum && sqnum != 0) {
-                       ubi_err("two LEBs with same sequence number %llu",
-                               sqnum);
-                       ubi_dbg_dump_seb(seb, 0);
-                       ubi_dbg_dump_vid_hdr(vid_hdr);
-                       return -EINVAL;
-               }
-
-               /*
-                * Now we have to drop the older one and preserve the newer
-                * one.
-                */
-               cmp_res = compare_lebs(ubi, seb, pnum, vid_hdr);
-               if (cmp_res < 0)
-                       return cmp_res;
-
-               if (cmp_res & 1) {
-                       /*
-                        * This logical eraseblock is newer then the one
-                        * found earlier.
-                        */
-                       err = validate_vid_hdr(vid_hdr, sv, pnum);
-                       if (err)
-                               return err;
-
-                       if (cmp_res & 4)
-                               err = add_to_list(si, seb->pnum, seb->ec,
-                                                 &si->corr);
-                       else
-                               err = add_to_list(si, seb->pnum, seb->ec,
-                                                 &si->erase);
-                       if (err)
-                               return err;
-
-                       seb->ec = ec;
-                       seb->pnum = pnum;
-                       seb->scrub = ((cmp_res & 2) || bitflips);
-                       seb->sqnum = sqnum;
-                       seb->leb_ver = leb_ver;
-
-                       if (sv->highest_lnum == lnum)
-                               sv->last_data_size =
-                                       be32_to_cpu(vid_hdr->data_size);
-
-                       return 0;
-               } else {
-                       /*
-                        * This logical eraseblock is older then the one found
-                        * previously.
-                        */
-                       if (cmp_res & 4)
-                               return add_to_list(si, pnum, ec, &si->corr);
-                       else
-                               return add_to_list(si, pnum, ec, &si->erase);
-               }
-       }
-
-       /*
-        * We've met this logical eraseblock for the first time, add it to the
-        * scanning information.
-        */
-
-       err = validate_vid_hdr(vid_hdr, sv, pnum);
-       if (err)
-               return err;
-
-       seb = kmalloc(sizeof(struct ubi_scan_leb), GFP_KERNEL);
-       if (!seb)
-               return -ENOMEM;
-
-       seb->ec = ec;
-       seb->pnum = pnum;
-       seb->lnum = lnum;
-       seb->sqnum = sqnum;
-       seb->scrub = bitflips;
-       seb->leb_ver = leb_ver;
-
-       if (sv->highest_lnum <= lnum) {
-               sv->highest_lnum = lnum;
-               sv->last_data_size = be32_to_cpu(vid_hdr->data_size);
-       }
-
-       sv->leb_count += 1;
-       rb_link_node(&seb->u.rb, parent, p);
-       rb_insert_color(&seb->u.rb, &sv->root);
-       return 0;
-}
-
-/**
- * ubi_scan_find_sv - find information about a particular volume in the
- * scanning information.
- * @si: scanning information
- * @vol_id: the requested volume ID
- *
- * This function returns a pointer to the volume description or %NULL if there
- * are no data about this volume in the scanning information.
- */
-struct ubi_scan_volume *ubi_scan_find_sv(const struct ubi_scan_info *si,
-                                        int vol_id)
-{
-       struct ubi_scan_volume *sv;
-       struct rb_node *p = si->volumes.rb_node;
-
-       while (p) {
-               sv = rb_entry(p, struct ubi_scan_volume, rb);
-
-               if (vol_id == sv->vol_id)
-                       return sv;
-
-               if (vol_id > sv->vol_id)
-                       p = p->rb_left;
-               else
-                       p = p->rb_right;
-       }
-
-       return NULL;
-}
-
-/**
- * ubi_scan_find_seb - find information about a particular logical
- * eraseblock in the volume scanning information.
- * @sv: a pointer to the volume scanning information
- * @lnum: the requested logical eraseblock
- *
- * This function returns a pointer to the scanning logical eraseblock or %NULL
- * if there are no data about it in the scanning volume information.
- */
-struct ubi_scan_leb *ubi_scan_find_seb(const struct ubi_scan_volume *sv,
-                                      int lnum)
-{
-       struct ubi_scan_leb *seb;
-       struct rb_node *p = sv->root.rb_node;
-
-       while (p) {
-               seb = rb_entry(p, struct ubi_scan_leb, u.rb);
-
-               if (lnum == seb->lnum)
-                       return seb;
-
-               if (lnum > seb->lnum)
-                       p = p->rb_left;
-               else
-                       p = p->rb_right;
-       }
-
-       return NULL;
-}
-
-/**
- * ubi_scan_rm_volume - delete scanning information about a volume.
- * @si: scanning information
- * @sv: the volume scanning information to delete
- */
-void ubi_scan_rm_volume(struct ubi_scan_info *si, struct ubi_scan_volume *sv)
-{
-       struct rb_node *rb;
-       struct ubi_scan_leb *seb;
-
-       dbg_bld("remove scanning information about volume %d", sv->vol_id);
-
-       while ((rb = rb_first(&sv->root))) {
-               seb = rb_entry(rb, struct ubi_scan_leb, u.rb);
-               rb_erase(&seb->u.rb, &sv->root);
-               list_add_tail(&seb->u.list, &si->erase);
-       }
-
-       rb_erase(&sv->rb, &si->volumes);
-       kfree(sv);
-       si->vols_found -= 1;
-}
-
-/**
- * ubi_scan_erase_peb - erase a physical eraseblock.
- * @ubi: UBI device description object
- * @si: scanning information
- * @pnum: physical eraseblock number to erase;
- * @ec: erase counter value to write (%UBI_SCAN_UNKNOWN_EC if it is unknown)
- *
- * This function erases physical eraseblock 'pnum', and writes the erase
- * counter header to it. This function should only be used on UBI device
- * initialization stages, when the EBA unit had not been yet initialized. This
- * function returns zero in case of success and a negative error code in case
- * of failure.
- */
-int ubi_scan_erase_peb(struct ubi_device *ubi, const struct ubi_scan_info *si,
-                      int pnum, int ec)
-{
-       int err;
-       struct ubi_ec_hdr *ec_hdr;
-
-       if ((long long)ec >= UBI_MAX_ERASECOUNTER) {
-               /*
-                * Erase counter overflow. Upgrade UBI and use 64-bit
-                * erase counters internally.
-                */
-               ubi_err("erase counter overflow at PEB %d, EC %d", pnum, ec);
-               return -EINVAL;
-       }
-
-       ec_hdr = kzalloc(ubi->ec_hdr_alsize, GFP_KERNEL);
-       if (!ec_hdr)
-               return -ENOMEM;
-
-       ec_hdr->ec = cpu_to_be64(ec);
-
-       err = ubi_io_sync_erase(ubi, pnum, 0);
-       if (err < 0)
-               goto out_free;
-
-       err = ubi_io_write_ec_hdr(ubi, pnum, ec_hdr);
-
-out_free:
-       kfree(ec_hdr);
-       return err;
-}
-
-/**
- * ubi_scan_get_free_peb - get a free physical eraseblock.
- * @ubi: UBI device description object
- * @si: scanning information
- *
- * This function returns a free physical eraseblock. It is supposed to be
- * called on the UBI initialization stages when the wear-leveling unit is not
- * initialized yet. This function picks a physical eraseblocks from one of the
- * lists, writes the EC header if it is needed, and removes it from the list.
- *
- * This function returns scanning physical eraseblock information in case of
- * success and an error code in case of failure.
- */
-struct ubi_scan_leb *ubi_scan_get_free_peb(struct ubi_device *ubi,
-                                          struct ubi_scan_info *si)
-{
-       int err = 0, i;
-       struct ubi_scan_leb *seb;
-
-       if (!list_empty(&si->free)) {
-               seb = list_entry(si->free.next, struct ubi_scan_leb, u.list);
-               list_del(&seb->u.list);
-               dbg_bld("return free PEB %d, EC %d", seb->pnum, seb->ec);
-               return seb;
-       }
-
-       for (i = 0; i < 2; i++) {
-               struct list_head *head;
-               struct ubi_scan_leb *tmp_seb;
-
-               if (i == 0)
-                       head = &si->erase;
-               else
-                       head = &si->corr;
-
-               /*
-                * We try to erase the first physical eraseblock from the @head
-                * list and pick it if we succeed, or try to erase the
-                * next one if not. And so forth. We don't want to take care
-                * about bad eraseblocks here - they'll be handled later.
-                */
-               list_for_each_entry_safe(seb, tmp_seb, head, u.list) {
-                       if (seb->ec == UBI_SCAN_UNKNOWN_EC)
-                               seb->ec = si->mean_ec;
-
-                       err = ubi_scan_erase_peb(ubi, si, seb->pnum, seb->ec+1);
-                       if (err)
-                               continue;
-
-                       seb->ec += 1;
-                       list_del(&seb->u.list);
-                       dbg_bld("return PEB %d, EC %d", seb->pnum, seb->ec);
-                       return seb;
-               }
-       }
-
-       ubi_err("no eraseblocks found");
-       return ERR_PTR(-ENOSPC);
-}
-
-/**
- * process_eb - read UBI headers, check them and add corresponding data
- * to the scanning information.
- * @ubi: UBI device description object
- * @si: scanning information
- * @pnum: the physical eraseblock number
- *
- * This function returns a zero if the physical eraseblock was successfully
- * handled and a negative error code in case of failure.
- */
-static int process_eb(struct ubi_device *ubi, struct ubi_scan_info *si, int pnum)
-{
-       long long uninitialized_var(ec);
-       int err, bitflips = 0, vol_id, ec_corr = 0;
-
-       dbg_bld("scan PEB %d", pnum);
-
-       /* Skip bad physical eraseblocks */
-       err = ubi_io_is_bad(ubi, pnum);
-       if (err < 0)
-               return err;
-       else if (err) {
-               /*
-                * FIXME: this is actually duty of the I/O unit to initialize
-                * this, but MTD does not provide enough information.
-                */
-               si->bad_peb_count += 1;
-               return 0;
-       }
-
-       err = ubi_io_read_ec_hdr(ubi, pnum, ech, 0);
-       if (err < 0)
-               return err;
-       else if (err == UBI_IO_BITFLIPS)
-               bitflips = 1;
-       else if (err == UBI_IO_PEB_EMPTY)
-               return add_to_list(si, pnum, UBI_SCAN_UNKNOWN_EC, &si->erase);
-       else if (err == UBI_IO_BAD_EC_HDR) {
-               /*
-                * We have to also look at the VID header, possibly it is not
-                * corrupted. Set %bitflips flag in order to make this PEB be
-                * moved and EC be re-created.
-                */
-               ec_corr = 1;
-               ec = UBI_SCAN_UNKNOWN_EC;
-               bitflips = 1;
-       }
-
-       si->is_empty = 0;
-
-       if (!ec_corr) {
-               /* Make sure UBI version is OK */
-               if (ech->version != UBI_VERSION) {
-                       ubi_err("this UBI version is %d, image version is %d",
-                               UBI_VERSION, (int)ech->version);
-                       return -EINVAL;
-               }
-
-               ec = be64_to_cpu(ech->ec);
-               if (ec > UBI_MAX_ERASECOUNTER) {
-                       /*
-                        * Erase counter overflow. The EC headers have 64 bits
-                        * reserved, but we anyway make use of only 31 bit
-                        * values, as this seems to be enough for any existing
-                        * flash. Upgrade UBI and use 64-bit erase counters
-                        * internally.
-                        */
-                       ubi_err("erase counter overflow, max is %d",
-                               UBI_MAX_ERASECOUNTER);
-                       ubi_dbg_dump_ec_hdr(ech);
-                       return -EINVAL;
-               }
-       }
-
-       /* OK, we've done with the EC header, let's look at the VID header */
-
-       err = ubi_io_read_vid_hdr(ubi, pnum, vidh, 0);
-       if (err < 0)
-               return err;
-       else if (err == UBI_IO_BITFLIPS)
-               bitflips = 1;
-       else if (err == UBI_IO_BAD_VID_HDR ||
-                (err == UBI_IO_PEB_FREE && ec_corr)) {
-               /* VID header is corrupted */
-               err = add_to_list(si, pnum, ec, &si->corr);
-               if (err)
-                       return err;
-               goto adjust_mean_ec;
-       } else if (err == UBI_IO_PEB_FREE) {
-               /* No VID header - the physical eraseblock is free */
-               err = add_to_list(si, pnum, ec, &si->free);
-               if (err)
-                       return err;
-               goto adjust_mean_ec;
-       }
-
-       vol_id = be32_to_cpu(vidh->vol_id);
-       if (vol_id > UBI_MAX_VOLUMES && vol_id != UBI_LAYOUT_VOLUME_ID) {
-               int lnum = be32_to_cpu(vidh->lnum);
-
-               /* Unsupported internal volume */
-               switch (vidh->compat) {
-               case UBI_COMPAT_DELETE:
-                       ubi_msg("\"delete\" compatible internal volume %d:%d"
-                               " found, remove it", vol_id, lnum);
-                       err = add_to_list(si, pnum, ec, &si->corr);
-                       if (err)
-                               return err;
-                       break;
-
-               case UBI_COMPAT_RO:
-                       ubi_msg("read-only compatible internal volume %d:%d"
-                               " found, switch to read-only mode",
-                               vol_id, lnum);
-                       ubi->ro_mode = 1;
-                       break;
-
-               case UBI_COMPAT_PRESERVE:
-                       ubi_msg("\"preserve\" compatible internal volume %d:%d"
-                               " found", vol_id, lnum);
-                       err = add_to_list(si, pnum, ec, &si->alien);
-                       if (err)
-                               return err;
-                       si->alien_peb_count += 1;
-                       return 0;
-
-               case UBI_COMPAT_REJECT:
-                       ubi_err("incompatible internal volume %d:%d found",
-                               vol_id, lnum);
-                       return -EINVAL;
-               }
-       }
-
-       /* Both UBI headers seem to be fine */
-       err = ubi_scan_add_used(ubi, si, pnum, ec, vidh, bitflips);
-       if (err)
-               return err;
-
-adjust_mean_ec:
-       if (!ec_corr) {
-               si->ec_sum += ec;
-               si->ec_count += 1;
-               if (ec > si->max_ec)
-                       si->max_ec = ec;
-               if (ec < si->min_ec)
-                       si->min_ec = ec;
-       }
-
-       return 0;
-}
-
-/**
- * ubi_scan - scan an MTD device.
- * @ubi: UBI device description object
- *
- * This function does full scanning of an MTD device and returns complete
- * information about it. In case of failure, an error code is returned.
- */
-struct ubi_scan_info *ubi_scan(struct ubi_device *ubi)
-{
-       int err, pnum;
-       struct rb_node *rb1, *rb2;
-       struct ubi_scan_volume *sv;
-       struct ubi_scan_leb *seb;
-       struct ubi_scan_info *si;
-
-       si = kzalloc(sizeof(struct ubi_scan_info), GFP_KERNEL);
-       if (!si)
-               return ERR_PTR(-ENOMEM);
-
-       INIT_LIST_HEAD(&si->corr);
-       INIT_LIST_HEAD(&si->free);
-       INIT_LIST_HEAD(&si->erase);
-       INIT_LIST_HEAD(&si->alien);
-       si->volumes = RB_ROOT;
-       si->is_empty = 1;
-
-       err = -ENOMEM;
-       ech = kzalloc(ubi->ec_hdr_alsize, GFP_KERNEL);
-       if (!ech)
-               goto out_si;
-
-       vidh = ubi_zalloc_vid_hdr(ubi, GFP_KERNEL);
-       if (!vidh)
-               goto out_ech;
-
-       for (pnum = 0; pnum < ubi->peb_count; pnum++) {
-               cond_resched();
-
-               dbg_msg("process PEB %d", pnum);
-               err = process_eb(ubi, si, pnum);
-               if (err < 0)
-                       goto out_vidh;
-       }
-
-       dbg_msg("scanning is finished");
-
-       /* Calculate mean erase counter */
-       if (si->ec_count) {
-               do_div(si->ec_sum, si->ec_count);
-               si->mean_ec = si->ec_sum;
-       }
-
-       if (si->is_empty)
-               ubi_msg("empty MTD device detected");
-
-       /*
-        * In case of unknown erase counter we use the mean erase counter
-        * value.
-        */
-       ubi_rb_for_each_entry(rb1, sv, &si->volumes, rb) {
-               ubi_rb_for_each_entry(rb2, seb, &sv->root, u.rb)
-                       if (seb->ec == UBI_SCAN_UNKNOWN_EC)
-                               seb->ec = si->mean_ec;
-       }
-
-       list_for_each_entry(seb, &si->free, u.list) {
-               if (seb->ec == UBI_SCAN_UNKNOWN_EC)
-                       seb->ec = si->mean_ec;
-       }
-
-       list_for_each_entry(seb, &si->corr, u.list)
-               if (seb->ec == UBI_SCAN_UNKNOWN_EC)
-                       seb->ec = si->mean_ec;
-
-       list_for_each_entry(seb, &si->erase, u.list)
-               if (seb->ec == UBI_SCAN_UNKNOWN_EC)
-                       seb->ec = si->mean_ec;
-
-       err = paranoid_check_si(ubi, si);
-       if (err) {
-               if (err > 0)
-                       err = -EINVAL;
-               goto out_vidh;
-       }
-
-       ubi_free_vid_hdr(ubi, vidh);
-       kfree(ech);
-
-       return si;
-
-out_vidh:
-       ubi_free_vid_hdr(ubi, vidh);
-out_ech:
-       kfree(ech);
-out_si:
-       ubi_scan_destroy_si(si);
-       return ERR_PTR(err);
-}
-
-/**
- * destroy_sv - free the scanning volume information
- * @sv: scanning volume information
- *
- * This function destroys the volume RB-tree (@sv->root) and the scanning
- * volume information.
- */
-static void destroy_sv(struct ubi_scan_volume *sv)
-{
-       struct ubi_scan_leb *seb;
-       struct rb_node *this = sv->root.rb_node;
-
-       while (this) {
-               if (this->rb_left)
-                       this = this->rb_left;
-               else if (this->rb_right)
-                       this = this->rb_right;
-               else {
-                       seb = rb_entry(this, struct ubi_scan_leb, u.rb);
-                       this = rb_parent(this);
-                       if (this) {
-                               if (this->rb_left == &seb->u.rb)
-                                       this->rb_left = NULL;
-                               else
-                                       this->rb_right = NULL;
-                       }
-
-                       kfree(seb);
-               }
-       }
-       kfree(sv);
-}
-
-/**
- * ubi_scan_destroy_si - destroy scanning information.
- * @si: scanning information
- */
-void ubi_scan_destroy_si(struct ubi_scan_info *si)
-{
-       struct ubi_scan_leb *seb, *seb_tmp;
-       struct ubi_scan_volume *sv;
-       struct rb_node *rb;
-
-       list_for_each_entry_safe(seb, seb_tmp, &si->alien, u.list) {
-               list_del(&seb->u.list);
-               kfree(seb);
-       }
-       list_for_each_entry_safe(seb, seb_tmp, &si->erase, u.list) {
-               list_del(&seb->u.list);
-               kfree(seb);
-       }
-       list_for_each_entry_safe(seb, seb_tmp, &si->corr, u.list) {
-               list_del(&seb->u.list);
-               kfree(seb);
-       }
-       list_for_each_entry_safe(seb, seb_tmp, &si->free, u.list) {
-               list_del(&seb->u.list);
-               kfree(seb);
-       }
-
-       /* Destroy the volume RB-tree */
-       rb = si->volumes.rb_node;
-       while (rb) {
-               if (rb->rb_left)
-                       rb = rb->rb_left;
-               else if (rb->rb_right)
-                       rb = rb->rb_right;
-               else {
-                       sv = rb_entry(rb, struct ubi_scan_volume, rb);
-
-                       rb = rb_parent(rb);
-                       if (rb) {
-                               if (rb->rb_left == &sv->rb)
-                                       rb->rb_left = NULL;
-                               else
-                                       rb->rb_right = NULL;
-                       }
-
-                       destroy_sv(sv);
-               }
-       }
-
-       kfree(si);
-}
-
-#ifdef CONFIG_MTD_UBI_DEBUG_PARANOID
-
-/**
- * paranoid_check_si - check if the scanning information is correct and
- * consistent.
- * @ubi: UBI device description object
- * @si: scanning information
- *
- * This function returns zero if the scanning information is all right, %1 if
- * not and a negative error code if an error occurred.
- */
-static int paranoid_check_si(struct ubi_device *ubi, struct ubi_scan_info *si)
-{
-       int pnum, err, vols_found = 0;
-       struct rb_node *rb1, *rb2;
-       struct ubi_scan_volume *sv;
-       struct ubi_scan_leb *seb, *last_seb;
-       uint8_t *buf;
-
-       /*
-        * At first, check that scanning information is OK.
-        */
-       ubi_rb_for_each_entry(rb1, sv, &si->volumes, rb) {
-               int leb_count = 0;
-
-               cond_resched();
-
-               vols_found += 1;
-
-               if (si->is_empty) {
-                       ubi_err("bad is_empty flag");
-                       goto bad_sv;
-               }
-
-               if (sv->vol_id < 0 || sv->highest_lnum < 0 ||
-                   sv->leb_count < 0 || sv->vol_type < 0 || sv->used_ebs < 0 ||
-                   sv->data_pad < 0 || sv->last_data_size < 0) {
-                       ubi_err("negative values");
-                       goto bad_sv;
-               }
-
-               if (sv->vol_id >= UBI_MAX_VOLUMES &&
-                   sv->vol_id < UBI_INTERNAL_VOL_START) {
-                       ubi_err("bad vol_id");
-                       goto bad_sv;
-               }
-
-               if (sv->vol_id > si->highest_vol_id) {
-                       ubi_err("highest_vol_id is %d, but vol_id %d is there",
-                               si->highest_vol_id, sv->vol_id);
-                       goto out;
-               }
-
-               if (sv->vol_type != UBI_DYNAMIC_VOLUME &&
-                   sv->vol_type != UBI_STATIC_VOLUME) {
-                       ubi_err("bad vol_type");
-                       goto bad_sv;
-               }
-
-               if (sv->data_pad > ubi->leb_size / 2) {
-                       ubi_err("bad data_pad");
-                       goto bad_sv;
-               }
-
-               last_seb = NULL;
-               ubi_rb_for_each_entry(rb2, seb, &sv->root, u.rb) {
-                       cond_resched();
-
-                       last_seb = seb;
-                       leb_count += 1;
-
-                       if (seb->pnum < 0 || seb->ec < 0) {
-                               ubi_err("negative values");
-                               goto bad_seb;
-                       }
-
-                       if (seb->ec < si->min_ec) {
-                               ubi_err("bad si->min_ec (%d), %d found",
-                                       si->min_ec, seb->ec);
-                               goto bad_seb;
-                       }
-
-                       if (seb->ec > si->max_ec) {
-                               ubi_err("bad si->max_ec (%d), %d found",
-                                       si->max_ec, seb->ec);
-                               goto bad_seb;
-                       }
-
-                       if (seb->pnum >= ubi->peb_count) {
-                               ubi_err("too high PEB number %d, total PEBs %d",
-                                       seb->pnum, ubi->peb_count);
-                               goto bad_seb;
-                       }
-
-                       if (sv->vol_type == UBI_STATIC_VOLUME) {
-                               if (seb->lnum >= sv->used_ebs) {
-                                       ubi_err("bad lnum or used_ebs");
-                                       goto bad_seb;
-                               }
-                       } else {
-                               if (sv->used_ebs != 0) {
-                                       ubi_err("non-zero used_ebs");
-                                       goto bad_seb;
-                               }
-                       }
-
-                       if (seb->lnum > sv->highest_lnum) {
-                               ubi_err("incorrect highest_lnum or lnum");
-                               goto bad_seb;
-                       }
-               }
-
-               if (sv->leb_count != leb_count) {
-                       ubi_err("bad leb_count, %d objects in the tree",
-                               leb_count);
-                       goto bad_sv;
-               }
-
-               if (!last_seb)
-                       continue;
-
-               seb = last_seb;
-
-               if (seb->lnum != sv->highest_lnum) {
-                       ubi_err("bad highest_lnum");
-                       goto bad_seb;
-               }
-       }
-
-       if (vols_found != si->vols_found) {
-               ubi_err("bad si->vols_found %d, should be %d",
-                       si->vols_found, vols_found);
-               goto out;
-       }
-
-       /* Check that scanning information is correct */
-       ubi_rb_for_each_entry(rb1, sv, &si->volumes, rb) {
-               last_seb = NULL;
-               ubi_rb_for_each_entry(rb2, seb, &sv->root, u.rb) {
-                       int vol_type;
-
-                       cond_resched();
-
-                       last_seb = seb;
-
-                       err = ubi_io_read_vid_hdr(ubi, seb->pnum, vidh, 1);
-                       if (err && err != UBI_IO_BITFLIPS) {
-                               ubi_err("VID header is not OK (%d)", err);
-                               if (err > 0)
-                                       err = -EIO;
-                               return err;
-                       }
-
-                       vol_type = vidh->vol_type == UBI_VID_DYNAMIC ?
-                                  UBI_DYNAMIC_VOLUME : UBI_STATIC_VOLUME;
-                       if (sv->vol_type != vol_type) {
-                               ubi_err("bad vol_type");
-                               goto bad_vid_hdr;
-                       }
-
-                       if (seb->sqnum != be64_to_cpu(vidh->sqnum)) {
-                               ubi_err("bad sqnum %llu", seb->sqnum);
-                               goto bad_vid_hdr;
-                       }
-
-                       if (sv->vol_id != be32_to_cpu(vidh->vol_id)) {
-                               ubi_err("bad vol_id %d", sv->vol_id);
-                               goto bad_vid_hdr;
-                       }
-
-                       if (sv->compat != vidh->compat) {
-                               ubi_err("bad compat %d", vidh->compat);
-                               goto bad_vid_hdr;
-                       }
-
-                       if (seb->lnum != be32_to_cpu(vidh->lnum)) {
-                               ubi_err("bad lnum %d", seb->lnum);
-                               goto bad_vid_hdr;
-                       }
-
-                       if (sv->used_ebs != be32_to_cpu(vidh->used_ebs)) {
-                               ubi_err("bad used_ebs %d", sv->used_ebs);
-                               goto bad_vid_hdr;
-                       }
-
-                       if (sv->data_pad != be32_to_cpu(vidh->data_pad)) {
-                               ubi_err("bad data_pad %d", sv->data_pad);
-                               goto bad_vid_hdr;
-                       }
-
-                       if (seb->leb_ver != be32_to_cpu(vidh->leb_ver)) {
-                               ubi_err("bad leb_ver %u", seb->leb_ver);
-                               goto bad_vid_hdr;
-                       }
-               }
-
-               if (!last_seb)
-                       continue;
-
-               if (sv->highest_lnum != be32_to_cpu(vidh->lnum)) {
-                       ubi_err("bad highest_lnum %d", sv->highest_lnum);
-                       goto bad_vid_hdr;
-               }
-
-               if (sv->last_data_size != be32_to_cpu(vidh->data_size)) {
-                       ubi_err("bad last_data_size %d", sv->last_data_size);
-                       goto bad_vid_hdr;
-               }
-       }
-
-       /*
-        * Make sure that all the physical eraseblocks are in one of the lists
-        * or trees.
-        */
-       buf = kzalloc(ubi->peb_count, GFP_KERNEL);
-       if (!buf)
-               return -ENOMEM;
-
-       for (pnum = 0; pnum < ubi->peb_count; pnum++) {
-               err = ubi_io_is_bad(ubi, pnum);
-               if (err < 0) {
-                       kfree(buf);
-                       return err;
-               }
-               else if (err)
-                       buf[pnum] = 1;
-       }
-
-       ubi_rb_for_each_entry(rb1, sv, &si->volumes, rb)
-               ubi_rb_for_each_entry(rb2, seb, &sv->root, u.rb)
-                       buf[seb->pnum] = 1;
-
-       list_for_each_entry(seb, &si->free, u.list)
-               buf[seb->pnum] = 1;
-
-       list_for_each_entry(seb, &si->corr, u.list)
-               buf[seb->pnum] = 1;
-
-       list_for_each_entry(seb, &si->erase, u.list)
-               buf[seb->pnum] = 1;
-
-       list_for_each_entry(seb, &si->alien, u.list)
-               buf[seb->pnum] = 1;
-
-       err = 0;
-       for (pnum = 0; pnum < ubi->peb_count; pnum++)
-               if (!buf[pnum]) {
-                       ubi_err("PEB %d is not referred", pnum);
-                       err = 1;
-               }
-
-       kfree(buf);
-       if (err)
-               goto out;
-       return 0;
-
-bad_seb:
-       ubi_err("bad scanning information about LEB %d", seb->lnum);
-       ubi_dbg_dump_seb(seb, 0);
-       ubi_dbg_dump_sv(sv);
-       goto out;
-
-bad_sv:
-       ubi_err("bad scanning information about volume %d", sv->vol_id);
-       ubi_dbg_dump_sv(sv);
-       goto out;
-
-bad_vid_hdr:
-       ubi_err("bad scanning information about volume %d", sv->vol_id);
-       ubi_dbg_dump_sv(sv);
-       ubi_dbg_dump_vid_hdr(vidh);
-
-out:
-       ubi_dbg_dump_stack();
-       return 1;
-}
-
-#endif /* CONFIG_MTD_UBI_DEBUG_PARANOID */
diff --git a/drivers/mtd/ubi/scan.h b/drivers/mtd/ubi/scan.h
deleted file mode 100644 (file)
index 252b1f1..0000000
+++ /dev/null
@@ -1,153 +0,0 @@
-/*
- * Copyright (c) International Business Machines Corp., 2006
- *
- * SPDX-License-Identifier:    GPL-2.0+
- *
- * Author: Artem Bityutskiy (Битюцкий Артём)
- */
-
-#ifndef __UBI_SCAN_H__
-#define __UBI_SCAN_H__
-
-/* The erase counter value for this physical eraseblock is unknown */
-#define UBI_SCAN_UNKNOWN_EC (-1)
-
-/**
- * struct ubi_scan_leb - scanning information about a physical eraseblock.
- * @ec: erase counter (%UBI_SCAN_UNKNOWN_EC if it is unknown)
- * @pnum: physical eraseblock number
- * @lnum: logical eraseblock number
- * @scrub: if this physical eraseblock needs scrubbing
- * @sqnum: sequence number
- * @u: unions RB-tree or @list links
- * @u.rb: link in the per-volume RB-tree of &struct ubi_scan_leb objects
- * @u.list: link in one of the eraseblock lists
- * @leb_ver: logical eraseblock version (obsolete)
- *
- * One object of this type is allocated for each physical eraseblock during
- * scanning.
- */
-struct ubi_scan_leb {
-       int ec;
-       int pnum;
-       int lnum;
-       int scrub;
-       unsigned long long sqnum;
-       union {
-               struct rb_node rb;
-               struct list_head list;
-       } u;
-       uint32_t leb_ver;
-};
-
-/**
- * struct ubi_scan_volume - scanning information about a volume.
- * @vol_id: volume ID
- * @highest_lnum: highest logical eraseblock number in this volume
- * @leb_count: number of logical eraseblocks in this volume
- * @vol_type: volume type
- * @used_ebs: number of used logical eraseblocks in this volume (only for
- * static volumes)
- * @last_data_size: amount of data in the last logical eraseblock of this
- * volume (always equivalent to the usable logical eraseblock size in case of
- * dynamic volumes)
- * @data_pad: how many bytes at the end of logical eraseblocks of this volume
- * are not used (due to volume alignment)
- * @compat: compatibility flags of this volume
- * @rb: link in the volume RB-tree
- * @root: root of the RB-tree containing all the eraseblock belonging to this
- * volume (&struct ubi_scan_leb objects)
- *
- * One object of this type is allocated for each volume during scanning.
- */
-struct ubi_scan_volume {
-       int vol_id;
-       int highest_lnum;
-       int leb_count;
-       int vol_type;
-       int used_ebs;
-       int last_data_size;
-       int data_pad;
-       int compat;
-       struct rb_node rb;
-       struct rb_root root;
-};
-
-/**
- * struct ubi_scan_info - UBI scanning information.
- * @volumes: root of the volume RB-tree
- * @corr: list of corrupted physical eraseblocks
- * @free: list of free physical eraseblocks
- * @erase: list of physical eraseblocks which have to be erased
- * @alien: list of physical eraseblocks which should not be used by UBI (e.g.,
- * @bad_peb_count: count of bad physical eraseblocks
- * those belonging to "preserve"-compatible internal volumes)
- * @vols_found: number of volumes found during scanning
- * @highest_vol_id: highest volume ID
- * @alien_peb_count: count of physical eraseblocks in the @alien list
- * @is_empty: flag indicating whether the MTD device is empty or not
- * @min_ec: lowest erase counter value
- * @max_ec: highest erase counter value
- * @max_sqnum: highest sequence number value
- * @mean_ec: mean erase counter value
- * @ec_sum: a temporary variable used when calculating @mean_ec
- * @ec_count: a temporary variable used when calculating @mean_ec
- *
- * This data structure contains the result of scanning and may be used by other
- * UBI units to build final UBI data structures, further error-recovery and so
- * on.
- */
-struct ubi_scan_info {
-       struct rb_root volumes;
-       struct list_head corr;
-       struct list_head free;
-       struct list_head erase;
-       struct list_head alien;
-       int bad_peb_count;
-       int vols_found;
-       int highest_vol_id;
-       int alien_peb_count;
-       int is_empty;
-       int min_ec;
-       int max_ec;
-       unsigned long long max_sqnum;
-       int mean_ec;
-       uint64_t ec_sum;
-       int ec_count;
-};
-
-struct ubi_device;
-struct ubi_vid_hdr;
-
-/*
- * ubi_scan_move_to_list - move a physical eraseblock from the volume tree to a
- * list.
- *
- * @sv: volume scanning information
- * @seb: scanning eraseblock infprmation
- * @list: the list to move to
- */
-static inline void ubi_scan_move_to_list(struct ubi_scan_volume *sv,
-                                        struct ubi_scan_leb *seb,
-                                        struct list_head *list)
-{
-               rb_erase(&seb->u.rb, &sv->root);
-               list_add_tail(&seb->u.list, list);
-}
-
-int ubi_scan_add_used(struct ubi_device *ubi, struct ubi_scan_info *si,
-                     int pnum, int ec, const struct ubi_vid_hdr *vid_hdr,
-                     int bitflips);
-struct ubi_scan_volume *ubi_scan_find_sv(const struct ubi_scan_info *si,
-                                        int vol_id);
-struct ubi_scan_leb *ubi_scan_find_seb(const struct ubi_scan_volume *sv,
-                                      int lnum);
-void ubi_scan_rm_volume(struct ubi_scan_info *si, struct ubi_scan_volume *sv);
-struct ubi_scan_leb *ubi_scan_get_free_peb(struct ubi_device *ubi,
-                                          struct ubi_scan_info *si);
-int ubi_scan_erase_peb(struct ubi_device *ubi, const struct ubi_scan_info *si,
-                      int pnum, int ec);
-struct ubi_scan_info *ubi_scan(struct ubi_device *ubi);
-void ubi_scan_destroy_si(struct ubi_scan_info *si);
-
-#endif /* !__UBI_SCAN_H__ */
index 9012326d6101d761b509881796578141eb24c241..2809805c2c0856617c7a14a94671993512b5eee6 100644 (file)
@@ -86,10 +86,11 @@ enum {
  * Compatibility constants used by internal volumes.
  *
  * @UBI_COMPAT_DELETE: delete this internal volume before anything is written
- * to the flash
+ *                     to the flash
  * @UBI_COMPAT_RO: attach this device in read-only mode
  * @UBI_COMPAT_PRESERVE: preserve this internal volume - do not touch its
- * physical eraseblocks, don't allow the wear-leveling unit to move them
+ *                       physical eraseblocks, don't allow the wear-leveling
+ *                       sub-system to move them
  * @UBI_COMPAT_REJECT: reject this UBI image
  */
 enum {
@@ -111,18 +112,19 @@ enum {
  * struct ubi_ec_hdr - UBI erase counter header.
  * @magic: erase counter header magic number (%UBI_EC_HDR_MAGIC)
  * @version: version of UBI implementation which is supposed to accept this
- * UBI image
+ *           UBI image
  * @padding1: reserved for future, zeroes
  * @ec: the erase counter
  * @vid_hdr_offset: where the VID header starts
  * @data_offset: where the user data start
+ * @image_seq: image sequence number
  * @padding2: reserved for future, zeroes
  * @hdr_crc: erase counter header CRC checksum
  *
  * The erase counter header takes 64 bytes and has a plenty of unused space for
  * future usage. The unused fields are zeroed. The @version field is used to
  * indicate the version of UBI implementation which is supposed to be able to
- * work with this UBI image. If @version is greater then the current UBI
+ * work with this UBI image. If @version is greater than the current UBI
  * version, the image is rejected. This may be useful in future if something
  * is changed radically. This field is duplicated in the volume identifier
  * header.
@@ -131,6 +133,14 @@ enum {
  * volume identifier header and user data, relative to the beginning of the
  * physical eraseblock. These values have to be the same for all physical
  * eraseblocks.
+ *
+ * The @image_seq field is used to validate a UBI image that has been prepared
+ * for a UBI device. The @image_seq value can be any value, but it must be the
+ * same on all eraseblocks. UBI will ensure that all new erase counter headers
+ * also contain this value, and will check the value when attaching the flash.
+ * One way to make use of @image_seq is to increase its value by one every time
+ * an image is flashed over an existing image, then, if the flashing does not
+ * complete, UBI will detect the error when attaching the media.
  */
 struct ubi_ec_hdr {
        __be32  magic;
@@ -139,32 +149,32 @@ struct ubi_ec_hdr {
        __be64  ec; /* Warning: the current limit is 31-bit anyway! */
        __be32  vid_hdr_offset;
        __be32  data_offset;
-       __u8    padding2[36];
+       __be32  image_seq;
+       __u8    padding2[32];
        __be32  hdr_crc;
-} __attribute__ ((packed));
+} __packed;
 
 /**
  * struct ubi_vid_hdr - on-flash UBI volume identifier header.
  * @magic: volume identifier header magic number (%UBI_VID_HDR_MAGIC)
  * @version: UBI implementation version which is supposed to accept this UBI
- * image (%UBI_VERSION)
+ *           image (%UBI_VERSION)
  * @vol_type: volume type (%UBI_VID_DYNAMIC or %UBI_VID_STATIC)
  * @copy_flag: if this logical eraseblock was copied from another physical
- * eraseblock (for wear-leveling reasons)
+ *             eraseblock (for wear-leveling reasons)
  * @compat: compatibility of this volume (%0, %UBI_COMPAT_DELETE,
- * %UBI_COMPAT_IGNORE, %UBI_COMPAT_PRESERVE, or %UBI_COMPAT_REJECT)
+ *          %UBI_COMPAT_IGNORE, %UBI_COMPAT_PRESERVE, or %UBI_COMPAT_REJECT)
  * @vol_id: ID of this volume
  * @lnum: logical eraseblock number
- * @leb_ver: version of this logical eraseblock (IMPORTANT: obsolete, to be
- * removed, kept only for not breaking older UBI users)
+ * @padding1: reserved for future, zeroes
  * @data_size: how many bytes of data this logical eraseblock contains
  * @used_ebs: total number of used logical eraseblocks in this volume
  * @data_pad: how many bytes at the end of this physical eraseblock are not
- * used
+ *            used
  * @data_crc: CRC checksum of the data stored in this logical eraseblock
- * @padding1: reserved for future, zeroes
- * @sqnum: sequence number
  * @padding2: reserved for future, zeroes
+ * @sqnum: sequence number
+ * @padding3: reserved for future, zeroes
  * @hdr_crc: volume identifier header CRC checksum
  *
  * The @sqnum is the value of the global sequence counter at the time when this
@@ -175,7 +185,7 @@ struct ubi_ec_hdr {
  * (sequence number) is used to distinguish between older and newer versions of
  * logical eraseblocks.
  *
- * There are 2 situations when there may be more then one physical eraseblock
+ * There are 2 situations when there may be more than one physical eraseblock
  * corresponding to the same logical eraseblock, i.e., having the same @vol_id
  * and @lnum values in the volume identifier header. Suppose we have a logical
  * eraseblock L and it is mapped to the physical eraseblock P.
@@ -212,10 +222,6 @@ struct ubi_ec_hdr {
  * checksum is correct, this physical eraseblock is selected (P1). Otherwise
  * the older one (P) is selected.
  *
- * Note, there is an obsolete @leb_ver field which was used instead of @sqnum
- * in the past. But it is not used anymore and we keep it in order to be able
- * to deal with old UBI images. It will be removed at some point.
- *
  * There are 2 sorts of volumes in UBI: user volumes and internal volumes.
  * Internal volumes are not seen from outside and are used for various internal
  * UBI purposes. In this implementation there is only one internal volume - the
@@ -236,9 +242,9 @@ struct ubi_ec_hdr {
  * The @data_crc field contains the CRC checksum of the contents of the logical
  * eraseblock if this is a static volume. In case of dynamic volumes, it does
  * not contain the CRC checksum as a rule. The only exception is when the
- * data of the physical eraseblock was moved by the wear-leveling unit, then
- * the wear-leveling unit calculates the data CRC and stores it in the
- * @data_crc field. And of course, the @copy_flag is %in this case.
+ * data of the physical eraseblock was moved by the wear-leveling sub-system,
+ * then the wear-leveling sub-system calculates the data CRC and stores it in
+ * the @data_crc field. And of course, the @copy_flag is %in this case.
  *
  * The @data_size field is used only for static volumes because UBI has to know
  * how many bytes of data are stored in this eraseblock. For dynamic volumes,
@@ -265,23 +271,23 @@ struct ubi_vid_hdr {
        __u8    compat;
        __be32  vol_id;
        __be32  lnum;
-       __be32  leb_ver; /* obsolete, to be removed, don't use */
+       __u8    padding1[4];
        __be32  data_size;
        __be32  used_ebs;
        __be32  data_pad;
        __be32  data_crc;
-       __u8    padding1[4];
+       __u8    padding2[4];
        __be64  sqnum;
-       __u8    padding2[12];
+       __u8    padding3[12];
        __be32  hdr_crc;
-} __attribute__ ((packed));
+} __packed;
 
 /* Internal UBI volumes count */
 #define UBI_INT_VOL_COUNT 1
 
 /*
- * Starting ID of internal volumes. There is reserved room for 4096 internal
- * volumes.
+ * Starting ID of internal volumes: 0x7fffefff.
+ * There is reserved room for 4096 internal volumes.
  */
 #define UBI_INTERNAL_VOL_START (0x7FFFFFFF - 4096)
 
@@ -351,10 +357,151 @@ struct ubi_vtbl_record {
        __u8    vol_type;
        __u8    upd_marker;
        __be16  name_len;
+#ifndef __UBOOT__
        __u8    name[UBI_VOL_NAME_MAX+1];
+#else
+       char    name[UBI_VOL_NAME_MAX+1];
+#endif
        __u8    flags;
        __u8    padding[23];
        __be32  crc;
-} __attribute__ ((packed));
+} __packed;
+
+/* UBI fastmap on-flash data structures */
+
+#define UBI_FM_SB_VOLUME_ID    (UBI_LAYOUT_VOLUME_ID + 1)
+#define UBI_FM_DATA_VOLUME_ID  (UBI_LAYOUT_VOLUME_ID + 2)
 
+/* fastmap on-flash data structure format version */
+#define UBI_FM_FMT_VERSION     1
+
+#define UBI_FM_SB_MAGIC                0x7B11D69F
+#define UBI_FM_HDR_MAGIC       0xD4B82EF7
+#define UBI_FM_VHDR_MAGIC      0xFA370ED1
+#define UBI_FM_POOL_MAGIC      0x67AF4D08
+#define UBI_FM_EBA_MAGIC       0xf0c040a8
+
+/* A fastmap supber block can be located between PEB 0 and
+ * UBI_FM_MAX_START */
+#define UBI_FM_MAX_START       64
+
+/* A fastmap can use up to UBI_FM_MAX_BLOCKS PEBs */
+#define UBI_FM_MAX_BLOCKS      32
+
+/* 5% of the total number of PEBs have to be scanned while attaching
+ * from a fastmap.
+ * But the size of this pool is limited to be between UBI_FM_MIN_POOL_SIZE and
+ * UBI_FM_MAX_POOL_SIZE */
+#define UBI_FM_MIN_POOL_SIZE   8
+#define UBI_FM_MAX_POOL_SIZE   256
+
+#define UBI_FM_WL_POOL_SIZE    25
+
+/**
+ * struct ubi_fm_sb - UBI fastmap super block
+ * @magic: fastmap super block magic number (%UBI_FM_SB_MAGIC)
+ * @version: format version of this fastmap
+ * @data_crc: CRC over the fastmap data
+ * @used_blocks: number of PEBs used by this fastmap
+ * @block_loc: an array containing the location of all PEBs of the fastmap
+ * @block_ec: the erase counter of each used PEB
+ * @sqnum: highest sequence number value at the time while taking the fastmap
+ *
+ */
+struct ubi_fm_sb {
+       __be32 magic;
+       __u8 version;
+       __u8 padding1[3];
+       __be32 data_crc;
+       __be32 used_blocks;
+       __be32 block_loc[UBI_FM_MAX_BLOCKS];
+       __be32 block_ec[UBI_FM_MAX_BLOCKS];
+       __be64 sqnum;
+       __u8 padding2[32];
+} __packed;
+
+/**
+ * struct ubi_fm_hdr - header of the fastmap data set
+ * @magic: fastmap header magic number (%UBI_FM_HDR_MAGIC)
+ * @free_peb_count: number of free PEBs known by this fastmap
+ * @used_peb_count: number of used PEBs known by this fastmap
+ * @scrub_peb_count: number of to be scrubbed PEBs known by this fastmap
+ * @bad_peb_count: number of bad PEBs known by this fastmap
+ * @erase_peb_count: number of bad PEBs which have to be erased
+ * @vol_count: number of UBI volumes known by this fastmap
+ */
+struct ubi_fm_hdr {
+       __be32 magic;
+       __be32 free_peb_count;
+       __be32 used_peb_count;
+       __be32 scrub_peb_count;
+       __be32 bad_peb_count;
+       __be32 erase_peb_count;
+       __be32 vol_count;
+       __u8 padding[4];
+} __packed;
+
+/* struct ubi_fm_hdr is followed by two struct ubi_fm_scan_pool */
+
+/**
+ * struct ubi_fm_scan_pool - Fastmap pool PEBs to be scanned while attaching
+ * @magic: pool magic numer (%UBI_FM_POOL_MAGIC)
+ * @size: current pool size
+ * @max_size: maximal pool size
+ * @pebs: an array containing the location of all PEBs in this pool
+ */
+struct ubi_fm_scan_pool {
+       __be32 magic;
+       __be16 size;
+       __be16 max_size;
+       __be32 pebs[UBI_FM_MAX_POOL_SIZE];
+       __be32 padding[4];
+} __packed;
+
+/* ubi_fm_scan_pool is followed by nfree+nused struct ubi_fm_ec records */
+
+/**
+ * struct ubi_fm_ec - stores the erase counter of a PEB
+ * @pnum: PEB number
+ * @ec: ec of this PEB
+ */
+struct ubi_fm_ec {
+       __be32 pnum;
+       __be32 ec;
+} __packed;
+
+/**
+ * struct ubi_fm_volhdr - Fastmap volume header
+ * it identifies the start of an eba table
+ * @magic: Fastmap volume header magic number (%UBI_FM_VHDR_MAGIC)
+ * @vol_id: volume id of the fastmapped volume
+ * @vol_type: type of the fastmapped volume
+ * @data_pad: data_pad value of the fastmapped volume
+ * @used_ebs: number of used LEBs within this volume
+ * @last_eb_bytes: number of bytes used in the last LEB
+ */
+struct ubi_fm_volhdr {
+       __be32 magic;
+       __be32 vol_id;
+       __u8 vol_type;
+       __u8 padding1[3];
+       __be32 data_pad;
+       __be32 used_ebs;
+       __be32 last_eb_bytes;
+       __u8 padding2[8];
+} __packed;
+
+/* struct ubi_fm_volhdr is followed by one struct ubi_fm_eba records */
+
+/**
+ * struct ubi_fm_eba - denotes an association beween a PEB and LEB
+ * @magic: EBA table magic number
+ * @reserved_pebs: number of table entries
+ * @pnum: PEB number of LEB (LEB is the index)
+ */
+struct ubi_fm_eba {
+       __be32 magic;
+       __be32 reserved_pebs;
+       __be32 pnum[0];
+} __packed;
 #endif /* !__UBI_MEDIA_H__ */
index f4f71655ed194d0c49e5a2106847cf48f9771d43..1c395730710bd0d5a92fde156f3b7a20db2e8b75 100644 (file)
@@ -10,7 +10,8 @@
 #ifndef __UBI_UBI_H__
 #define __UBI_UBI_H__
 
-#ifdef UBI_LINUX
+#define __UBOOT__
+#ifndef __UBOOT__
 #include <linux/init.h>
 #include <linux/types.h>
 #include <linux/list.h>
 #include <linux/fs.h>
 #include <linux/cdev.h>
 #include <linux/device.h>
+#include <linux/slab.h>
 #include <linux/string.h>
 #include <linux/vmalloc.h>
-#include <linux/mtd/mtd.h>
-#include <linux/mtd/ubi.h>
+#include <linux/notifier.h>
+#include <asm/pgtable.h>
+#else
+#include <ubi_uboot.h>
 #endif
-
-#include <linux/types.h>
-#include <linux/list.h>
-#include <linux/rbtree.h>
-#include <linux/string.h>
 #include <linux/mtd/mtd.h>
 #include <linux/mtd/ubi.h>
-
 #include "ubi-media.h"
-#include "scan.h"
-#include "debug.h"
+#include <mtd/ubi-user.h>
 
 /* Maximum number of supported UBI devices */
 #define UBI_MAX_DEVICES 32
 #else
 #define ubi_msg(fmt, ...) printk(KERN_NOTICE "UBI: " fmt "\n", ##__VA_ARGS__)
 #endif
+
 /* UBI warning messages */
-#define ubi_warn(fmt, ...) printk(KERN_WARNING "UBI warning: %s: " fmt "\n", \
-                                 __func__, ##__VA_ARGS__)
+#define ubi_warn(fmt, ...) pr_warn("UBI warning: %s: " fmt "\n",  \
+                                  __func__, ##__VA_ARGS__)
 /* UBI error messages */
-#define ubi_err(fmt, ...) printk(KERN_ERR "UBI error: %s: " fmt "\n", \
+#define ubi_err(fmt, ...) pr_err("UBI error: %s: " fmt "\n",      \
                                 __func__, ##__VA_ARGS__)
 
-/* Lowest number PEBs reserved for bad PEB handling */
-#define MIN_RESEVED_PEBS 2
-
 /* Background thread name pattern */
 #define UBI_BGT_NAME_PATTERN "ubi_bgt%dd"
 
-/* This marker in the EBA table means that the LEB is um-mapped */
+/*
+ * This marker in the EBA table means that the LEB is um-mapped.
+ * NOTE! It has to have the same value as %UBI_ALL.
+ */
 #define UBI_LEB_UNMAPPED -1
 
 /*
 #define UBI_IO_RETRIES 3
 
 /*
- * Error codes returned by the I/O unit.
- *
- * UBI_IO_PEB_EMPTY: the physical eraseblock is empty, i.e. it contains only
- * 0xFF bytes
- * UBI_IO_PEB_FREE: the physical eraseblock is free, i.e. it contains only a
- * valid erase counter header, and the rest are %0xFF bytes
- * UBI_IO_BAD_EC_HDR: the erase counter header is corrupted (bad magic or CRC)
- * UBI_IO_BAD_VID_HDR: the volume identifier header is corrupted (bad magic or
- * CRC)
+ * Length of the protection queue. The length is effectively equivalent to the
+ * number of (global) erase cycles PEBs are protected from the wear-leveling
+ * worker.
+ */
+#define UBI_PROT_QUEUE_LEN 10
+
+/* The volume ID/LEB number/erase counter is unknown */
+#define UBI_UNKNOWN -1
+
+/*
+ * The UBI debugfs directory name pattern and maximum name length (3 for "ubi"
+ * + 2 for the number plus 1 for the trailing zero byte.
+ */
+#define UBI_DFS_DIR_NAME "ubi%d"
+#define UBI_DFS_DIR_LEN  (3 + 2 + 1)
+
+/*
+ * Error codes returned by the I/O sub-system.
+ *
+ * UBI_IO_FF: the read region of flash contains only 0xFFs
+ * UBI_IO_FF_BITFLIPS: the same as %UBI_IO_FF, but also also there was a data
+ *                     integrity error reported by the MTD driver
+ *                     (uncorrectable ECC error in case of NAND)
+ * UBI_IO_BAD_HDR: the EC or VID header is corrupted (bad magic or CRC)
+ * UBI_IO_BAD_HDR_EBADMSG: the same as %UBI_IO_BAD_HDR, but also there was a
+ *                         data integrity error reported by the MTD driver
+ *                         (uncorrectable ECC error in case of NAND)
  * UBI_IO_BITFLIPS: bit-flips were detected and corrected
+ *
+ * Note, it is probably better to have bit-flip and ebadmsg as flags which can
+ * be or'ed with other error code. But this is a big change because there are
+ * may callers, so it does not worth the risk of introducing a bug
+ */
+enum {
+       UBI_IO_FF = 1,
+       UBI_IO_FF_BITFLIPS,
+       UBI_IO_BAD_HDR,
+       UBI_IO_BAD_HDR_EBADMSG,
+       UBI_IO_BITFLIPS,
+};
+
+/*
+ * Return codes of the 'ubi_eba_copy_leb()' function.
+ *
+ * MOVE_CANCEL_RACE: canceled because the volume is being deleted, the source
+ *                   PEB was put meanwhile, or there is I/O on the source PEB
+ * MOVE_SOURCE_RD_ERR: canceled because there was a read error from the source
+ *                     PEB
+ * MOVE_TARGET_RD_ERR: canceled because there was a read error from the target
+ *                     PEB
+ * MOVE_TARGET_WR_ERR: canceled because there was a write error to the target
+ *                     PEB
+ * MOVE_TARGET_BITFLIPS: canceled because a bit-flip was detected in the
+ *                       target PEB
+ * MOVE_RETRY: retry scrubbing the PEB
  */
 enum {
-       UBI_IO_PEB_EMPTY = 1,
-       UBI_IO_PEB_FREE,
-       UBI_IO_BAD_EC_HDR,
-       UBI_IO_BAD_VID_HDR,
-       UBI_IO_BITFLIPS
+       MOVE_CANCEL_RACE = 1,
+       MOVE_SOURCE_RD_ERR,
+       MOVE_TARGET_RD_ERR,
+       MOVE_TARGET_WR_ERR,
+       MOVE_TARGET_BITFLIPS,
+       MOVE_RETRY,
+};
+
+/*
+ * Return codes of the fastmap sub-system
+ *
+ * UBI_NO_FASTMAP: No fastmap super block was found
+ * UBI_BAD_FASTMAP: A fastmap was found but it's unusable
+ */
+enum {
+       UBI_NO_FASTMAP = 1,
+       UBI_BAD_FASTMAP,
 };
 
 /**
  * struct ubi_wl_entry - wear-leveling entry.
- * @rb: link in the corresponding RB-tree
+ * @u.rb: link in the corresponding (free/used) RB-tree
+ * @u.list: link in the protection queue
  * @ec: erase counter
  * @pnum: physical eraseblock number
  *
- * This data structure is used in the WL unit. Each physical eraseblock has a
- * corresponding &struct wl_entry object which may be kept in different
- * RB-trees. See WL unit for details.
+ * This data structure is used in the WL sub-system. Each physical eraseblock
+ * has a corresponding &struct wl_entry object which may be kept in different
+ * RB-trees. See WL sub-system for details.
  */
 struct ubi_wl_entry {
-       struct rb_node rb;
+       union {
+               struct rb_node rb;
+               struct list_head list;
+       } u;
        int ec;
        int pnum;
 };
@@ -119,10 +178,10 @@ struct ubi_wl_entry {
  * @mutex: read/write mutex to implement read/write access serialization to
  *         the (@vol_id, @lnum) logical eraseblock
  *
- * This data structure is used in the EBA unit to implement per-LEB locking.
- * When a logical eraseblock is being locked - corresponding
+ * This data structure is used in the EBA sub-system to implement per-LEB
+ * locking. When a logical eraseblock is being locked - corresponding
  * &struct ubi_ltree_entry object is inserted to the lock tree (@ubi->ltree).
- * See EBA unit for details.
+ * See EBA sub-system for details.
  */
 struct ubi_ltree_entry {
        struct rb_node rb;
@@ -132,8 +191,64 @@ struct ubi_ltree_entry {
        struct rw_semaphore mutex;
 };
 
+/**
+ * struct ubi_rename_entry - volume re-name description data structure.
+ * @new_name_len: new volume name length
+ * @new_name: new volume name
+ * @remove: if not zero, this volume should be removed, not re-named
+ * @desc: descriptor of the volume
+ * @list: links re-name entries into a list
+ *
+ * This data structure is utilized in the multiple volume re-name code. Namely,
+ * UBI first creates a list of &struct ubi_rename_entry objects from the
+ * &struct ubi_rnvol_req request object, and then utilizes this list to do all
+ * the job.
+ */
+struct ubi_rename_entry {
+       int new_name_len;
+       char new_name[UBI_VOL_NAME_MAX + 1];
+       int remove;
+       struct ubi_volume_desc *desc;
+       struct list_head list;
+};
+
 struct ubi_volume_desc;
 
+/**
+ * struct ubi_fastmap_layout - in-memory fastmap data structure.
+ * @e: PEBs used by the current fastmap
+ * @to_be_tortured: if non-zero tortured this PEB
+ * @used_blocks: number of used PEBs
+ * @max_pool_size: maximal size of the user pool
+ * @max_wl_pool_size: maximal size of the pool used by the WL sub-system
+ */
+struct ubi_fastmap_layout {
+       struct ubi_wl_entry *e[UBI_FM_MAX_BLOCKS];
+       int to_be_tortured[UBI_FM_MAX_BLOCKS];
+       int used_blocks;
+       int max_pool_size;
+       int max_wl_pool_size;
+};
+
+/**
+ * struct ubi_fm_pool - in-memory fastmap pool
+ * @pebs: PEBs in this pool
+ * @used: number of used PEBs
+ * @size: total number of PEBs in this pool
+ * @max_size: maximal size of the pool
+ *
+ * A pool gets filled with up to max_size.
+ * If all PEBs within the pool are used a new fastmap will be written
+ * to the flash and the pool gets refilled with empty PEBs.
+ *
+ */
+struct ubi_fm_pool {
+       int pebs[UBI_FM_MAX_POOL_SIZE];
+       int used;
+       int size;
+       int max_size;
+};
+
 /**
  * struct ubi_volume - UBI volume description data structure.
  * @dev: device object to make use of the the Linux device model
@@ -160,8 +275,6 @@ struct ubi_volume_desc;
  * @upd_ebs: how many eraseblocks are expected to be updated
  * @ch_lnum: LEB number which is being changing by the atomic LEB change
  *           operation
- * @ch_dtype: data persistency type which is being changing by the atomic LEB
- *            change operation
  * @upd_bytes: how many bytes are expected to be received for volume update or
  *             atomic LEB change
  * @upd_received: how many bytes were already received for volume update or
@@ -175,10 +288,7 @@ struct ubi_volume_desc;
  * @upd_marker: %1 if the update marker is set for this volume
  * @updating: %1 if the volume is being updated
  * @changing_leb: %1 if the atomic LEB change ioctl command is in progress
- *
- * @gluebi_desc: gluebi UBI volume descriptor
- * @gluebi_refcount: reference count of the gluebi MTD device
- * @gluebi_mtd: MTD device description object of the gluebi MTD device
+ * @direct_writes: %1 if direct writes are enabled for this volume
  *
  * The @corrupted field indicates that the volume's contents is corrupted.
  * Since UBI protects only static volumes, this field is not relevant to
@@ -202,16 +312,19 @@ struct ubi_volume {
        int vol_type;
        int usable_leb_size;
        int used_ebs;
+#ifndef __UBOOT__
        int last_eb_bytes;
+#else
+       u32 last_eb_bytes;
+#endif
        long long used_bytes;
        int alignment;
        int data_pad;
        int name_len;
-       char name[UBI_VOL_NAME_MAX+1];
+       char name[UBI_VOL_NAME_MAX + 1];
 
        int upd_ebs;
        int ch_lnum;
-       int ch_dtype;
        long long upd_bytes;
        long long upd_received;
        void *upd_buf;
@@ -222,22 +335,11 @@ struct ubi_volume {
        unsigned int upd_marker:1;
        unsigned int updating:1;
        unsigned int changing_leb:1;
-
-#ifdef CONFIG_MTD_UBI_GLUEBI
-       /*
-        * Gluebi-related stuff may be compiled out.
-        * TODO: this should not be built into UBI but should be a separate
-        * ubimtd driver which works on top of UBI and emulates MTD devices.
-        */
-       struct ubi_volume_desc *gluebi_desc;
-       int gluebi_refcount;
-       struct mtd_info gluebi_mtd;
-#endif
+       unsigned int direct_writes:1;
 };
 
 /**
- * struct ubi_volume_desc - descriptor of the UBI volume returned when it is
- * opened.
+ * struct ubi_volume_desc - UBI volume descriptor returned when it is opened.
  * @vol: reference to the corresponding volume description object
  * @mode: open mode (%UBI_READONLY, %UBI_READWRITE, or %UBI_EXCLUSIVE)
  */
@@ -248,6 +350,37 @@ struct ubi_volume_desc {
 
 struct ubi_wl_entry;
 
+/**
+ * struct ubi_debug_info - debugging information for an UBI device.
+ *
+ * @chk_gen: if UBI general extra checks are enabled
+ * @chk_io: if UBI I/O extra checks are enabled
+ * @disable_bgt: disable the background task for testing purposes
+ * @emulate_bitflips: emulate bit-flips for testing purposes
+ * @emulate_io_failures: emulate write/erase failures for testing purposes
+ * @dfs_dir_name: name of debugfs directory containing files of this UBI device
+ * @dfs_dir: direntry object of the UBI device debugfs directory
+ * @dfs_chk_gen: debugfs knob to enable UBI general extra checks
+ * @dfs_chk_io: debugfs knob to enable UBI I/O extra checks
+ * @dfs_disable_bgt: debugfs knob to disable the background task
+ * @dfs_emulate_bitflips: debugfs knob to emulate bit-flips
+ * @dfs_emulate_io_failures: debugfs knob to emulate write/erase failures
+ */
+struct ubi_debug_info {
+       unsigned int chk_gen:1;
+       unsigned int chk_io:1;
+       unsigned int disable_bgt:1;
+       unsigned int emulate_bitflips:1;
+       unsigned int emulate_io_failures:1;
+       char dfs_dir_name[UBI_DFS_DIR_LEN + 1];
+       struct dentry *dfs_dir;
+       struct dentry *dfs_chk_gen;
+       struct dentry *dfs_chk_io;
+       struct dentry *dfs_disable_bgt;
+       struct dentry *dfs_emulate_bitflips;
+       struct dentry *dfs_emulate_io_failures;
+};
+
 /**
  * struct ubi_device - UBI device description structure
  * @dev: UBI device object to use the the Linux device model
@@ -261,6 +394,7 @@ struct ubi_wl_entry;
  *                @vol->readers, @vol->writers, @vol->exclusive,
  *                @vol->ref_count, @vol->mapping and @vol->eba_tbl.
  * @ref_count: count of references on the UBI device
+ * @image_seq: image sequence number recorded on EC headers
  *
  * @rsvd_pebs: count of reserved physical eraseblocks
  * @avail_pebs: count of available physical eraseblocks
@@ -269,12 +403,13 @@ struct ubi_wl_entry;
  * @beb_rsvd_level: normal level of PEBs reserved for bad PEB handling
  *
  * @autoresize_vol_id: ID of the volume which has to be auto-resized at the end
- *                     of UBI ititializetion
+ *                     of UBI initialization
  * @vtbl_slots: how many slots are available in the volume table
  * @vtbl_size: size of the volume table in bytes
  * @vtbl: in-RAM volume table copy
- * @volumes_mutex: protects on-flash volume table and serializes volume
- *                 changes, like creation, deletion, update, resize
+ * @device_mutex: protects on-flash volume table and serializes volume
+ *                creation, deletion, update, re-size, re-name and set
+ *                property
  *
  * @max_ec: current highest erase counter value
  * @mean_ec: current mean erase counter value
@@ -284,20 +419,33 @@ struct ubi_wl_entry;
  * @ltree: the lock tree
  * @alc_mutex: serializes "atomic LEB change" operations
  *
+ * @fm_disabled: non-zero if fastmap is disabled (default)
+ * @fm: in-memory data structure of the currently used fastmap
+ * @fm_pool: in-memory data structure of the fastmap pool
+ * @fm_wl_pool: in-memory data structure of the fastmap pool used by the WL
+ *             sub-system
+ * @fm_mutex: serializes ubi_update_fastmap() and protects @fm_buf
+ * @fm_buf: vmalloc()'d buffer which holds the raw fastmap
+ * @fm_size: fastmap size in bytes
+ * @fm_sem: allows ubi_update_fastmap() to block EBA table changes
+ * @fm_work: fastmap work queue
+ *
  * @used: RB-tree of used physical eraseblocks
+ * @erroneous: RB-tree of erroneous used physical eraseblocks
  * @free: RB-tree of free physical eraseblocks
+ * @free_count: Contains the number of elements in @free
  * @scrub: RB-tree of physical eraseblocks which need scrubbing
- * @prot: protection trees
- * @prot.pnum: protection tree indexed by physical eraseblock numbers
- * @prot.aec: protection tree indexed by absolute erase counter value
- * @wl_lock: protects the @used, @free, @prot, @lookuptbl, @abs_ec, @move_from,
- *           @move_to, @move_to_put @erase_pending, @wl_scheduled, and @works
- *           fields
+ * @pq: protection queue (contain physical eraseblocks which are temporarily
+ *      protected from the wear-leveling worker)
+ * @pq_head: protection queue head
+ * @wl_lock: protects the @used, @free, @pq, @pq_head, @lookuptbl, @move_from,
+ *          @move_to, @move_to_put @erase_pending, @wl_scheduled, @works,
+ *          @erroneous, and @erroneous_peb_count fields
  * @move_mutex: serializes eraseblock moves
+ * @work_sem: synchronizes the WL worker with use tasks
  * @wl_scheduled: non-zero if the wear-leveling was scheduled
  * @lookuptbl: a table to quickly find a &struct ubi_wl_entry object for any
  *             physical eraseblock
- * @abs_ec: absolute erase counter
  * @move_from: physical eraseblock from where the data is being moved
  * @move_to: physical eraseblock where the data is being moved to
  * @move_to_put: if the "to" PEB was put
@@ -310,30 +458,38 @@ struct ubi_wl_entry;
  * @flash_size: underlying MTD device size (in bytes)
  * @peb_count: count of physical eraseblocks on the MTD device
  * @peb_size: physical eraseblock size
+ * @bad_peb_limit: top limit of expected bad physical eraseblocks
  * @bad_peb_count: count of bad physical eraseblocks
  * @good_peb_count: count of good physical eraseblocks
+ * @corr_peb_count: count of corrupted physical eraseblocks (preserved and not
+ *                  used by UBI)
+ * @erroneous_peb_count: count of erroneous physical eraseblocks in @erroneous
+ * @max_erroneous: maximum allowed amount of erroneous physical eraseblocks
  * @min_io_size: minimal input/output unit size of the underlying MTD device
  * @hdrs_min_io_size: minimal I/O unit size used for VID and EC headers
  * @ro_mode: if the UBI device is in read-only mode
  * @leb_size: logical eraseblock size
  * @leb_start: starting offset of logical eraseblocks within physical
- * eraseblocks
+ *             eraseblocks
  * @ec_hdr_alsize: size of the EC header aligned to @hdrs_min_io_size
  * @vid_hdr_alsize: size of the VID header aligned to @hdrs_min_io_size
  * @vid_hdr_offset: starting offset of the volume identifier header (might be
- * unaligned)
+ *                  unaligned)
  * @vid_hdr_aloffset: starting offset of the VID header aligned to
  * @hdrs_min_io_size
  * @vid_hdr_shift: contains @vid_hdr_offset - @vid_hdr_aloffset
  * @bad_allowed: whether the MTD device admits of bad physical eraseblocks or
  *               not
+ * @nor_flash: non-zero if working on top of NOR flash
+ * @max_write_size: maximum amount of bytes the underlying flash can write at a
+ *                  time (MTD write buffer size)
  * @mtd: MTD device descriptor
  *
- * @peb_buf1: a buffer of PEB size used for different purposes
- * @peb_buf2: another buffer of PEB size used for different purposes
- * @buf_mutex: proptects @peb_buf1 and @peb_buf2
- * @dbg_peb_buf: buffer of PEB size used for debugging
- * @dbg_buf_mutex: proptects @dbg_peb_buf
+ * @peb_buf: a buffer of PEB size used for different purposes
+ * @buf_mutex: protects @peb_buf
+ * @ckvol_mutex: serializes static volume checking when opening
+ *
+ * @dbg: debugging information for this UBI device
  */
 struct ubi_device {
        struct cdev cdev;
@@ -344,42 +500,56 @@ struct ubi_device {
        struct ubi_volume *volumes[UBI_MAX_VOLUMES+UBI_INT_VOL_COUNT];
        spinlock_t volumes_lock;
        int ref_count;
+       int image_seq;
 
        int rsvd_pebs;
        int avail_pebs;
        int beb_rsvd_pebs;
        int beb_rsvd_level;
+       int bad_peb_limit;
 
        int autoresize_vol_id;
        int vtbl_slots;
        int vtbl_size;
        struct ubi_vtbl_record *vtbl;
-       struct mutex volumes_mutex;
+       struct mutex device_mutex;
 
        int max_ec;
-       /* TODO: mean_ec is not updated run-time, fix */
+       /* Note, mean_ec is not updated run-time - should be fixed */
        int mean_ec;
 
-       /* EBA unit's stuff */
+       /* EBA sub-system's stuff */
        unsigned long long global_sqnum;
        spinlock_t ltree_lock;
        struct rb_root ltree;
        struct mutex alc_mutex;
 
-       /* Wear-leveling unit's stuff */
+       /* Fastmap stuff */
+       int fm_disabled;
+       struct ubi_fastmap_layout *fm;
+       struct ubi_fm_pool fm_pool;
+       struct ubi_fm_pool fm_wl_pool;
+       struct rw_semaphore fm_sem;
+       struct mutex fm_mutex;
+       void *fm_buf;
+       size_t fm_size;
+#ifndef __UBOOT__
+       struct work_struct fm_work;
+#endif
+
+       /* Wear-leveling sub-system's stuff */
        struct rb_root used;
+       struct rb_root erroneous;
        struct rb_root free;
+       int free_count;
        struct rb_root scrub;
-       struct {
-               struct rb_root pnum;
-               struct rb_root aec;
-       } prot;
+       struct list_head pq[UBI_PROT_QUEUE_LEN];
+       int pq_head;
        spinlock_t wl_lock;
        struct mutex move_mutex;
        struct rw_semaphore work_sem;
        int wl_scheduled;
        struct ubi_wl_entry **lookuptbl;
-       unsigned long long abs_ec;
        struct ubi_wl_entry *move_from;
        struct ubi_wl_entry *move_to;
        int move_to_put;
@@ -389,12 +559,15 @@ struct ubi_device {
        int thread_enabled;
        char bgt_name[sizeof(UBI_BGT_NAME_PATTERN)+2];
 
-       /* I/O unit's stuff */
+       /* I/O sub-system's stuff */
        long long flash_size;
        int peb_count;
        int peb_size;
        int bad_peb_count;
        int good_peb_count;
+       int corr_peb_count;
+       int erroneous_peb_count;
+       int max_erroneous;
        int min_io_size;
        int hdrs_min_io_size;
        int ro_mode;
@@ -405,35 +578,195 @@ struct ubi_device {
        int vid_hdr_offset;
        int vid_hdr_aloffset;
        int vid_hdr_shift;
-       int bad_allowed;
+       unsigned int bad_allowed:1;
+       unsigned int nor_flash:1;
+       int max_write_size;
        struct mtd_info *mtd;
 
-       void *peb_buf1;
-       void *peb_buf2;
+       void *peb_buf;
        struct mutex buf_mutex;
        struct mutex ckvol_mutex;
-#ifdef CONFIG_MTD_UBI_DEBUG
-       void *dbg_peb_buf;
-       struct mutex dbg_buf_mutex;
-#endif
+
+       struct ubi_debug_info dbg;
 };
 
+/**
+ * struct ubi_ainf_peb - attach information about a physical eraseblock.
+ * @ec: erase counter (%UBI_UNKNOWN if it is unknown)
+ * @pnum: physical eraseblock number
+ * @vol_id: ID of the volume this LEB belongs to
+ * @lnum: logical eraseblock number
+ * @scrub: if this physical eraseblock needs scrubbing
+ * @copy_flag: this LEB is a copy (@copy_flag is set in VID header of this LEB)
+ * @sqnum: sequence number
+ * @u: unions RB-tree or @list links
+ * @u.rb: link in the per-volume RB-tree of &struct ubi_ainf_peb objects
+ * @u.list: link in one of the eraseblock lists
+ *
+ * One object of this type is allocated for each physical eraseblock when
+ * attaching an MTD device. Note, if this PEB does not belong to any LEB /
+ * volume, the @vol_id and @lnum fields are initialized to %UBI_UNKNOWN.
+ */
+struct ubi_ainf_peb {
+       int ec;
+       int pnum;
+       int vol_id;
+       int lnum;
+       unsigned int scrub:1;
+       unsigned int copy_flag:1;
+       unsigned long long sqnum;
+       union {
+               struct rb_node rb;
+               struct list_head list;
+       } u;
+};
+
+/**
+ * struct ubi_ainf_volume - attaching information about a volume.
+ * @vol_id: volume ID
+ * @highest_lnum: highest logical eraseblock number in this volume
+ * @leb_count: number of logical eraseblocks in this volume
+ * @vol_type: volume type
+ * @used_ebs: number of used logical eraseblocks in this volume (only for
+ *            static volumes)
+ * @last_data_size: amount of data in the last logical eraseblock of this
+ *                  volume (always equivalent to the usable logical eraseblock
+ *                  size in case of dynamic volumes)
+ * @data_pad: how many bytes at the end of logical eraseblocks of this volume
+ *            are not used (due to volume alignment)
+ * @compat: compatibility flags of this volume
+ * @rb: link in the volume RB-tree
+ * @root: root of the RB-tree containing all the eraseblock belonging to this
+ *        volume (&struct ubi_ainf_peb objects)
+ *
+ * One object of this type is allocated for each volume when attaching an MTD
+ * device.
+ */
+struct ubi_ainf_volume {
+       int vol_id;
+       int highest_lnum;
+       int leb_count;
+       int vol_type;
+       int used_ebs;
+       int last_data_size;
+       int data_pad;
+       int compat;
+       struct rb_node rb;
+       struct rb_root root;
+};
+
+/**
+ * struct ubi_attach_info - MTD device attaching information.
+ * @volumes: root of the volume RB-tree
+ * @corr: list of corrupted physical eraseblocks
+ * @free: list of free physical eraseblocks
+ * @erase: list of physical eraseblocks which have to be erased
+ * @alien: list of physical eraseblocks which should not be used by UBI (e.g.,
+ *         those belonging to "preserve"-compatible internal volumes)
+ * @corr_peb_count: count of PEBs in the @corr list
+ * @empty_peb_count: count of PEBs which are presumably empty (contain only
+ *                   0xFF bytes)
+ * @alien_peb_count: count of PEBs in the @alien list
+ * @bad_peb_count: count of bad physical eraseblocks
+ * @maybe_bad_peb_count: count of bad physical eraseblocks which are not marked
+ *                       as bad yet, but which look like bad
+ * @vols_found: number of volumes found
+ * @highest_vol_id: highest volume ID
+ * @is_empty: flag indicating whether the MTD device is empty or not
+ * @min_ec: lowest erase counter value
+ * @max_ec: highest erase counter value
+ * @max_sqnum: highest sequence number value
+ * @mean_ec: mean erase counter value
+ * @ec_sum: a temporary variable used when calculating @mean_ec
+ * @ec_count: a temporary variable used when calculating @mean_ec
+ * @aeb_slab_cache: slab cache for &struct ubi_ainf_peb objects
+ *
+ * This data structure contains the result of attaching an MTD device and may
+ * be used by other UBI sub-systems to build final UBI data structures, further
+ * error-recovery and so on.
+ */
+struct ubi_attach_info {
+       struct rb_root volumes;
+       struct list_head corr;
+       struct list_head free;
+       struct list_head erase;
+       struct list_head alien;
+       int corr_peb_count;
+       int empty_peb_count;
+       int alien_peb_count;
+       int bad_peb_count;
+       int maybe_bad_peb_count;
+       int vols_found;
+       int highest_vol_id;
+       int is_empty;
+       int min_ec;
+       int max_ec;
+       unsigned long long max_sqnum;
+       int mean_ec;
+       uint64_t ec_sum;
+       int ec_count;
+       struct kmem_cache *aeb_slab_cache;
+};
+
+/**
+ * struct ubi_work - UBI work description data structure.
+ * @list: a link in the list of pending works
+ * @func: worker function
+ * @e: physical eraseblock to erase
+ * @vol_id: the volume ID on which this erasure is being performed
+ * @lnum: the logical eraseblock number
+ * @torture: if the physical eraseblock has to be tortured
+ * @anchor: produce a anchor PEB to by used by fastmap
+ *
+ * The @func pointer points to the worker function. If the @cancel argument is
+ * not zero, the worker has to free the resources and exit immediately. The
+ * worker has to return zero in case of success and a negative error code in
+ * case of failure.
+ */
+struct ubi_work {
+       struct list_head list;
+       int (*func)(struct ubi_device *ubi, struct ubi_work *wrk, int cancel);
+       /* The below fields are only relevant to erasure works */
+       struct ubi_wl_entry *e;
+       int vol_id;
+       int lnum;
+       int torture;
+       int anchor;
+};
+
+#include "debug.h"
+
 extern struct kmem_cache *ubi_wl_entry_slab;
-extern struct file_operations ubi_ctrl_cdev_operations;
-extern struct file_operations ubi_cdev_operations;
-extern struct file_operations ubi_vol_cdev_operations;
+extern const struct file_operations ubi_ctrl_cdev_operations;
+extern const struct file_operations ubi_cdev_operations;
+extern const struct file_operations ubi_vol_cdev_operations;
 extern struct class *ubi_class;
 extern struct mutex ubi_devices_mutex;
+extern struct blocking_notifier_head ubi_notifiers;
+
+/* attach.c */
+int ubi_add_to_av(struct ubi_device *ubi, struct ubi_attach_info *ai, int pnum,
+                 int ec, const struct ubi_vid_hdr *vid_hdr, int bitflips);
+struct ubi_ainf_volume *ubi_find_av(const struct ubi_attach_info *ai,
+                                   int vol_id);
+void ubi_remove_av(struct ubi_attach_info *ai, struct ubi_ainf_volume *av);
+struct ubi_ainf_peb *ubi_early_get_peb(struct ubi_device *ubi,
+                                      struct ubi_attach_info *ai);
+int ubi_attach(struct ubi_device *ubi, int force_scan);
+void ubi_destroy_ai(struct ubi_attach_info *ai);
 
 /* vtbl.c */
 int ubi_change_vtbl_record(struct ubi_device *ubi, int idx,
                           struct ubi_vtbl_record *vtbl_rec);
-int ubi_read_volume_table(struct ubi_device *ubi, struct ubi_scan_info *si);
+int ubi_vtbl_rename_volumes(struct ubi_device *ubi,
+                           struct list_head *rename_list);
+int ubi_read_volume_table(struct ubi_device *ubi, struct ubi_attach_info *ai);
 
 /* vmt.c */
 int ubi_create_volume(struct ubi_device *ubi, struct ubi_mkvol_req *req);
-int ubi_remove_volume(struct ubi_volume_desc *desc);
+int ubi_remove_volume(struct ubi_volume_desc *desc, int no_vtbl);
 int ubi_resize_volume(struct ubi_volume_desc *desc, int reserved_pebs);
+int ubi_rename_volumes(struct ubi_device *ubi, struct list_head *rename_list);
 int ubi_add_volume(struct ubi_device *ubi, struct ubi_volume *vol);
 void ubi_free_volume(struct ubi_device *ubi, struct ubi_volume *vol);
 
@@ -448,9 +781,12 @@ int ubi_more_leb_change_data(struct ubi_device *ubi, struct ubi_volume *vol,
                             const void __user *buf, int count);
 
 /* misc.c */
-int ubi_calc_data_len(const struct ubi_device *ubi, const void *buf, int length);
+int ubi_calc_data_len(const struct ubi_device *ubi, const void *buf,
+                     int length);
 int ubi_check_volume(struct ubi_device *ubi, int vol_id);
+void ubi_update_reserved(struct ubi_device *ubi);
 void ubi_calculate_reserved(struct ubi_device *ubi);
+int ubi_check_pattern(const void *buf, uint8_t patt, int size);
 
 /* gluebi.c */
 #ifdef CONFIG_MTD_UBI_GLUEBI
@@ -474,25 +810,33 @@ int ubi_eba_unmap_leb(struct ubi_device *ubi, struct ubi_volume *vol,
 int ubi_eba_read_leb(struct ubi_device *ubi, struct ubi_volume *vol, int lnum,
                     void *buf, int offset, int len, int check);
 int ubi_eba_write_leb(struct ubi_device *ubi, struct ubi_volume *vol, int lnum,
-                     const void *buf, int offset, int len, int dtype);
+                     const void *buf, int offset, int len);
 int ubi_eba_write_leb_st(struct ubi_device *ubi, struct ubi_volume *vol,
-                        int lnum, const void *buf, int len, int dtype,
-                        int used_ebs);
+                        int lnum, const void *buf, int len, int used_ebs);
 int ubi_eba_atomic_leb_change(struct ubi_device *ubi, struct ubi_volume *vol,
-                             int lnum, const void *buf, int len, int dtype);
+                             int lnum, const void *buf, int len);
 int ubi_eba_copy_leb(struct ubi_device *ubi, int from, int to,
                     struct ubi_vid_hdr *vid_hdr);
-int ubi_eba_init_scan(struct ubi_device *ubi, struct ubi_scan_info *si);
-void ubi_eba_close(const struct ubi_device *ubi);
+int ubi_eba_init(struct ubi_device *ubi, struct ubi_attach_info *ai);
+unsigned long long ubi_next_sqnum(struct ubi_device *ubi);
+int self_check_eba(struct ubi_device *ubi, struct ubi_attach_info *ai_fastmap,
+                  struct ubi_attach_info *ai_scan);
 
 /* wl.c */
-int ubi_wl_get_peb(struct ubi_device *ubi, int dtype);
-int ubi_wl_put_peb(struct ubi_device *ubi, int pnum, int torture);
-int ubi_wl_flush(struct ubi_device *ubi);
+int ubi_wl_get_peb(struct ubi_device *ubi);
+int ubi_wl_put_peb(struct ubi_device *ubi, int vol_id, int lnum,
+                  int pnum, int torture);
+int ubi_wl_flush(struct ubi_device *ubi, int vol_id, int lnum);
 int ubi_wl_scrub_peb(struct ubi_device *ubi, int pnum);
-int ubi_wl_init_scan(struct ubi_device *ubi, struct ubi_scan_info *si);
+int ubi_wl_init(struct ubi_device *ubi, struct ubi_attach_info *ai);
 void ubi_wl_close(struct ubi_device *ubi);
 int ubi_thread(void *u);
+struct ubi_wl_entry *ubi_wl_get_fm_peb(struct ubi_device *ubi, int anchor);
+int ubi_wl_put_fm_peb(struct ubi_device *ubi, struct ubi_wl_entry *used_e,
+                     int lnum, int torture);
+int ubi_is_erase_work(struct ubi_work *wrk);
+void ubi_refill_pools(struct ubi_device *ubi);
+int ubi_ensure_anchor_pebs(struct ubi_device *ubi);
 
 /* io.c */
 int ubi_io_read(const struct ubi_device *ubi, void *buf, int pnum, int offset,
@@ -512,16 +856,37 @@ int ubi_io_write_vid_hdr(struct ubi_device *ubi, int pnum,
                         struct ubi_vid_hdr *vid_hdr);
 
 /* build.c */
-int ubi_attach_mtd_dev(struct mtd_info *mtd, int ubi_num, int vid_hdr_offset);
+int ubi_attach_mtd_dev(struct mtd_info *mtd, int ubi_num,
+                      int vid_hdr_offset, int max_beb_per1024);
 int ubi_detach_mtd_dev(int ubi_num, int anyway);
 struct ubi_device *ubi_get_device(int ubi_num);
 void ubi_put_device(struct ubi_device *ubi);
 struct ubi_device *ubi_get_by_major(int major);
 int ubi_major2num(int major);
+int ubi_volume_notify(struct ubi_device *ubi, struct ubi_volume *vol,
+                     int ntype);
+int ubi_notify_all(struct ubi_device *ubi, int ntype,
+                  struct notifier_block *nb);
+int ubi_enumerate_volumes(struct notifier_block *nb);
+void ubi_free_internal_volumes(struct ubi_device *ubi);
+
+/* kapi.c */
+void ubi_do_get_device_info(struct ubi_device *ubi, struct ubi_device_info *di);
+void ubi_do_get_volume_info(struct ubi_device *ubi, struct ubi_volume *vol,
+                           struct ubi_volume_info *vi);
+/* scan.c */
+int ubi_compare_lebs(struct ubi_device *ubi, const struct ubi_ainf_peb *aeb,
+                     int pnum, const struct ubi_vid_hdr *vid_hdr);
+
+/* fastmap.c */
+size_t ubi_calc_fm_size(struct ubi_device *ubi);
+int ubi_update_fastmap(struct ubi_device *ubi);
+int ubi_scan_fastmap(struct ubi_device *ubi, struct ubi_attach_info *ai,
+                    int fm_anchor);
 
 /*
  * ubi_rb_for_each_entry - walk an RB-tree.
- * @rb: a pointer to type 'struct rb_node' to to use as a loop counter
+ * @rb: a pointer to type 'struct rb_node' to use as a loop counter
  * @pos: a pointer to RB-tree entry type to use as a loop counter
  * @root: RB-tree's root
  * @member: the name of the 'struct rb_node' within the RB-tree entry
@@ -530,7 +895,23 @@ int ubi_major2num(int major);
        for (rb = rb_first(root),                                            \
             pos = (rb ? container_of(rb, typeof(*pos), member) : NULL);     \
             rb;                                                             \
-            rb = rb_next(rb), pos = container_of(rb, typeof(*pos), member))
+            rb = rb_next(rb),                                               \
+            pos = (rb ? container_of(rb, typeof(*pos), member) : NULL))
+
+/*
+ * ubi_move_aeb_to_list - move a PEB from the volume tree to a list.
+ *
+ * @av: volume attaching information
+ * @aeb: attaching eraseblock information
+ * @list: the list to move to
+ */
+static inline void ubi_move_aeb_to_list(struct ubi_ainf_volume *av,
+                                        struct ubi_ainf_peb *aeb,
+                                        struct list_head *list)
+{
+               rb_erase(&aeb->u.rb, &av->root);
+               list_add_tail(&aeb->u.list, list);
+}
 
 /**
  * ubi_zalloc_vid_hdr - allocate a volume identifier header object.
@@ -606,6 +987,7 @@ static inline void ubi_ro_mode(struct ubi_device *ubi)
        if (!ubi->ro_mode) {
                ubi->ro_mode = 1;
                ubi_warn("switch to read-only mode");
+               dump_stack();
        }
 }
 
index e597f82b874dcf3926feb7b44a46a550d8d46bce..220c120515ac508f58eb9e1459cb8aed4d3eeb0e 100644 (file)
  * transaction with a roll-back capability.
  */
 
-#ifdef UBI_LINUX
-#include <linux/err.h>
-#include <asm/uaccess.h>
-#include <asm/div64.h>
+#define __UBOOT__
+#ifndef __UBOOT__
+#include <linux/uaccess.h>
+#else
+#include <div64.h>
+#include <ubi_uboot.h>
 #endif
+#include <linux/err.h>
+#include <linux/math64.h>
 
-#include <ubi_uboot.h>
 #include "ubi.h"
 
 /**
@@ -48,22 +51,21 @@ static int set_update_marker(struct ubi_device *ubi, struct ubi_volume *vol)
        int err;
        struct ubi_vtbl_record vtbl_rec;
 
-       dbg_msg("set update marker for volume %d", vol->vol_id);
+       dbg_gen("set update marker for volume %d", vol->vol_id);
 
        if (vol->upd_marker) {
                ubi_assert(ubi->vtbl[vol->vol_id].upd_marker);
-               dbg_msg("already set");
+               dbg_gen("already set");
                return 0;
        }
 
-       memcpy(&vtbl_rec, &ubi->vtbl[vol->vol_id],
-              sizeof(struct ubi_vtbl_record));
+       vtbl_rec = ubi->vtbl[vol->vol_id];
        vtbl_rec.upd_marker = 1;
 
-       mutex_lock(&ubi->volumes_mutex);
+       mutex_lock(&ubi->device_mutex);
        err = ubi_change_vtbl_record(ubi, vol->vol_id, &vtbl_rec);
-       mutex_unlock(&ubi->volumes_mutex);
        vol->upd_marker = 1;
+       mutex_unlock(&ubi->device_mutex);
        return err;
 }
 
@@ -81,31 +83,29 @@ static int clear_update_marker(struct ubi_device *ubi, struct ubi_volume *vol,
                               long long bytes)
 {
        int err;
-       uint64_t tmp;
        struct ubi_vtbl_record vtbl_rec;
 
-       dbg_msg("clear update marker for volume %d", vol->vol_id);
+       dbg_gen("clear update marker for volume %d", vol->vol_id);
 
-       memcpy(&vtbl_rec, &ubi->vtbl[vol->vol_id],
-              sizeof(struct ubi_vtbl_record));
+       vtbl_rec = ubi->vtbl[vol->vol_id];
        ubi_assert(vol->upd_marker && vtbl_rec.upd_marker);
        vtbl_rec.upd_marker = 0;
 
        if (vol->vol_type == UBI_STATIC_VOLUME) {
                vol->corrupted = 0;
-               vol->used_bytes = tmp = bytes;
-               vol->last_eb_bytes = do_div(tmp, vol->usable_leb_size);
-               vol->used_ebs = tmp;
+               vol->used_bytes = bytes;
+               vol->used_ebs = div_u64_rem(bytes, vol->usable_leb_size,
+                                           &vol->last_eb_bytes);
                if (vol->last_eb_bytes)
                        vol->used_ebs += 1;
                else
                        vol->last_eb_bytes = vol->usable_leb_size;
        }
 
-       mutex_lock(&ubi->volumes_mutex);
+       mutex_lock(&ubi->device_mutex);
        err = ubi_change_vtbl_record(ubi, vol->vol_id, &vtbl_rec);
-       mutex_unlock(&ubi->volumes_mutex);
        vol->upd_marker = 0;
+       mutex_unlock(&ubi->device_mutex);
        return err;
 }
 
@@ -123,9 +123,8 @@ int ubi_start_update(struct ubi_device *ubi, struct ubi_volume *vol,
                     long long bytes)
 {
        int i, err;
-       uint64_t tmp;
 
-       dbg_msg("start update of volume %d, %llu bytes", vol->vol_id, bytes);
+       dbg_gen("start update of volume %d, %llu bytes", vol->vol_id, bytes);
        ubi_assert(!vol->updating && !vol->changing_leb);
        vol->updating = 1;
 
@@ -141,21 +140,23 @@ int ubi_start_update(struct ubi_device *ubi, struct ubi_volume *vol,
        }
 
        if (bytes == 0) {
+               err = ubi_wl_flush(ubi, UBI_ALL, UBI_ALL);
+               if (err)
+                       return err;
+
                err = clear_update_marker(ubi, vol, 0);
                if (err)
                        return err;
-               err = ubi_wl_flush(ubi);
-               if (!err)
-                       vol->updating = 0;
+               vol->updating = 0;
+               return 0;
        }
 
        vol->upd_buf = vmalloc(ubi->leb_size);
        if (!vol->upd_buf)
                return -ENOMEM;
 
-       tmp = bytes;
-       vol->upd_ebs = !!do_div(tmp, vol->usable_leb_size);
-       vol->upd_ebs += tmp;
+       vol->upd_ebs = div_u64(bytes + vol->usable_leb_size - 1,
+                              vol->usable_leb_size);
        vol->upd_bytes = bytes;
        vol->upd_received = 0;
        return 0;
@@ -175,17 +176,15 @@ int ubi_start_leb_change(struct ubi_device *ubi, struct ubi_volume *vol,
 {
        ubi_assert(!vol->updating && !vol->changing_leb);
 
-       dbg_msg("start changing LEB %d:%d, %u bytes",
+       dbg_gen("start changing LEB %d:%d, %u bytes",
                vol->vol_id, req->lnum, req->bytes);
        if (req->bytes == 0)
-               return ubi_eba_atomic_leb_change(ubi, vol, req->lnum, NULL, 0,
-                                                req->dtype);
+               return ubi_eba_atomic_leb_change(ubi, vol, req->lnum, NULL, 0);
 
        vol->upd_bytes = req->bytes;
        vol->upd_received = 0;
        vol->changing_leb = 1;
        vol->ch_lnum = req->lnum;
-       vol->ch_dtype = req->dtype;
 
        vol->upd_buf = vmalloc(req->bytes);
        if (!vol->upd_buf)
@@ -234,11 +233,11 @@ static int write_leb(struct ubi_device *ubi, struct ubi_volume *vol, int lnum,
                memset(buf + len, 0xFF, l - len);
                len = ubi_calc_data_len(ubi, buf, l);
                if (len == 0) {
-                       dbg_msg("all %d bytes contain 0xFF - skip", len);
+                       dbg_gen("all %d bytes contain 0xFF - skip", len);
                        return 0;
                }
 
-               err = ubi_eba_write_leb(ubi, vol, lnum, buf, 0, len, UBI_UNKNOWN);
+               err = ubi_eba_write_leb(ubi, vol, lnum, buf, 0, len);
        } else {
                /*
                 * When writing static volume, and this is the last logical
@@ -250,8 +249,7 @@ static int write_leb(struct ubi_device *ubi, struct ubi_volume *vol, int lnum,
                 * contain zeros, not random trash.
                 */
                memset(buf + len, 0, vol->usable_leb_size - len);
-               err = ubi_eba_write_leb_st(ubi, vol, lnum, buf, len,
-                                          UBI_UNKNOWN, used_ebs);
+               err = ubi_eba_write_leb_st(ubi, vol, lnum, buf, len, used_ebs);
        }
 
        return err;
@@ -259,6 +257,7 @@ static int write_leb(struct ubi_device *ubi, struct ubi_volume *vol, int lnum,
 
 /**
  * ubi_more_update_data - write more update data.
+ * @ubi: UBI device description object
  * @vol: volume description object
  * @buf: write data (user-space memory buffer)
  * @count: how much bytes to write
@@ -272,19 +271,20 @@ static int write_leb(struct ubi_device *ubi, struct ubi_volume *vol, int lnum,
 int ubi_more_update_data(struct ubi_device *ubi, struct ubi_volume *vol,
                         const void __user *buf, int count)
 {
-       uint64_t tmp;
+#ifndef __UBOOT__
        int lnum, offs, err = 0, len, to_write = count;
+#else
+       int lnum, err = 0, len, to_write = count;
+       u32 offs;
+#endif
 
-       dbg_msg("write %d of %lld bytes, %lld already passed",
+       dbg_gen("write %d of %lld bytes, %lld already passed",
                count, vol->upd_bytes, vol->upd_received);
 
        if (ubi->ro_mode)
                return -EROFS;
 
-       tmp = vol->upd_received;
-       offs = do_div(tmp, vol->usable_leb_size);
-       lnum = tmp;
-
+       lnum = div_u64_rem(vol->upd_received,  vol->usable_leb_size, &offs);
        if (vol->upd_received + count > vol->upd_bytes)
                to_write = count = vol->upd_bytes - vol->upd_received;
 
@@ -359,16 +359,16 @@ int ubi_more_update_data(struct ubi_device *ubi, struct ubi_volume *vol,
 
        ubi_assert(vol->upd_received <= vol->upd_bytes);
        if (vol->upd_received == vol->upd_bytes) {
+               err = ubi_wl_flush(ubi, UBI_ALL, UBI_ALL);
+               if (err)
+                       return err;
                /* The update is finished, clear the update marker */
                err = clear_update_marker(ubi, vol, vol->upd_bytes);
                if (err)
                        return err;
-               err = ubi_wl_flush(ubi);
-               if (err == 0) {
-                       vol->updating = 0;
-                       err = to_write;
-                       vfree(vol->upd_buf);
-               }
+               vol->updating = 0;
+               err = to_write;
+               vfree(vol->upd_buf);
        }
 
        return err;
@@ -376,6 +376,7 @@ int ubi_more_update_data(struct ubi_device *ubi, struct ubi_volume *vol,
 
 /**
  * ubi_more_leb_change_data - accept more data for atomic LEB change.
+ * @ubi: UBI device description object
  * @vol: volume description object
  * @buf: write data (user-space memory buffer)
  * @count: how much bytes to write
@@ -392,7 +393,7 @@ int ubi_more_leb_change_data(struct ubi_device *ubi, struct ubi_volume *vol,
 {
        int err;
 
-       dbg_msg("write %d of %lld bytes, %lld already passed",
+       dbg_gen("write %d of %lld bytes, %lld already passed",
                count, vol->upd_bytes, vol->upd_received);
 
        if (ubi->ro_mode)
@@ -410,10 +411,11 @@ int ubi_more_leb_change_data(struct ubi_device *ubi, struct ubi_volume *vol,
        if (vol->upd_received == vol->upd_bytes) {
                int len = ALIGN((int)vol->upd_bytes, ubi->min_io_size);
 
-               memset(vol->upd_buf + vol->upd_bytes, 0xFF, len - vol->upd_bytes);
+               memset(vol->upd_buf + vol->upd_bytes, 0xFF,
+                      len - vol->upd_bytes);
                len = ubi_calc_data_len(ubi, vol->upd_buf, len);
                err = ubi_eba_atomic_leb_change(ubi, vol, vol->ch_lnum,
-                                               vol->upd_buf, len, UBI_UNKNOWN);
+                                               vol->upd_buf, len);
                if (err)
                        return err;
        }
index c4e894b43a70fc3c9270efa215313653b4afe70f..d9665a446a3dbc1d5174af3dcbc0c88ba79a9170 100644 (file)
  * resizing.
  */
 
-#ifdef UBI_LINUX
+#define __UBOOT__
+#ifndef __UBOOT__
 #include <linux/err.h>
-#include <asm/div64.h>
+#include <linux/slab.h>
+#include <linux/export.h>
+#else
+#include <div64.h>
+#include <ubi_uboot.h>
 #endif
+#include <linux/math64.h>
 
-#include <ubi_uboot.h>
 #include "ubi.h"
 
-#ifdef CONFIG_MTD_UBI_DEBUG_PARANOID
-static void paranoid_check_volumes(struct ubi_device *ubi);
-#else
-#define paranoid_check_volumes(ubi)
-#endif
+static int self_check_volumes(struct ubi_device *ubi);
 
-#ifdef UBI_LINUX
+#ifndef __UBOOT__
 static ssize_t vol_attribute_show(struct device *dev,
                                  struct device_attribute *attr, char *buf);
 
@@ -121,10 +122,11 @@ static void vol_release(struct device *dev)
 {
        struct ubi_volume *vol = container_of(dev, struct ubi_volume, dev);
 
+       kfree(vol->eba_tbl);
        kfree(vol);
 }
 
-#ifdef UBI_LINUX
+#ifndef __UBOOT__
 /**
  * volume_sysfs_init - initialize sysfs for new volume.
  * @ubi: UBI device description object
@@ -193,14 +195,13 @@ static void volume_sysfs_close(struct ubi_volume *vol)
  * %UBI_VOL_NUM_AUTO, this function automatically assign ID to the new volume
  * and saves it in @req->vol_id. Returns zero in case of success and a negative
  * error code in case of failure. Note, the caller has to have the
- * @ubi->volumes_mutex locked.
+ * @ubi->device_mutex locked.
  */
 int ubi_create_volume(struct ubi_device *ubi, struct ubi_mkvol_req *req)
 {
-       int i, err, vol_id = req->vol_id, dont_free = 0;
+       int i, err, vol_id = req->vol_id, do_free = 1;
        struct ubi_volume *vol;
        struct ubi_vtbl_record vtbl_rec;
-       uint64_t bytes;
        dev_t dev;
 
        if (ubi->ro_mode)
@@ -213,7 +214,7 @@ int ubi_create_volume(struct ubi_device *ubi, struct ubi_mkvol_req *req)
        spin_lock(&ubi->volumes_lock);
        if (vol_id == UBI_VOL_NUM_AUTO) {
                /* Find unused volume ID */
-               dbg_msg("search for vacant volume ID");
+               dbg_gen("search for vacant volume ID");
                for (i = 0; i < ubi->vtbl_slots; i++)
                        if (!ubi->volumes[i]) {
                                vol_id = i;
@@ -221,21 +222,21 @@ int ubi_create_volume(struct ubi_device *ubi, struct ubi_mkvol_req *req)
                        }
 
                if (vol_id == UBI_VOL_NUM_AUTO) {
-                       dbg_err("out of volume IDs");
+                       ubi_err("out of volume IDs");
                        err = -ENFILE;
                        goto out_unlock;
                }
                req->vol_id = vol_id;
        }
 
-       dbg_msg("volume ID %d, %llu bytes, type %d, name %s",
-               vol_id, (unsigned long long)req->bytes,
+       dbg_gen("create device %d, volume %d, %llu bytes, type %d, name %s",
+               ubi->ubi_num, vol_id, (unsigned long long)req->bytes,
                (int)req->vol_type, req->name);
 
        /* Ensure that this volume does not exist */
        err = -EEXIST;
        if (ubi->volumes[vol_id]) {
-               dbg_err("volume %d already exists", vol_id);
+               ubi_err("volume %d already exists", vol_id);
                goto out_unlock;
        }
 
@@ -244,20 +245,21 @@ int ubi_create_volume(struct ubi_device *ubi, struct ubi_mkvol_req *req)
                if (ubi->volumes[i] &&
                    ubi->volumes[i]->name_len == req->name_len &&
                    !strcmp(ubi->volumes[i]->name, req->name)) {
-                       dbg_err("volume \"%s\" exists (ID %d)", req->name, i);
+                       ubi_err("volume \"%s\" exists (ID %d)", req->name, i);
                        goto out_unlock;
                }
 
        /* Calculate how many eraseblocks are requested */
        vol->usable_leb_size = ubi->leb_size - ubi->leb_size % req->alignment;
-       bytes = req->bytes;
-       if (do_div(bytes, vol->usable_leb_size))
-               vol->reserved_pebs = 1;
-       vol->reserved_pebs += bytes;
+       vol->reserved_pebs += div_u64(req->bytes + vol->usable_leb_size - 1,
+                                     vol->usable_leb_size);
 
        /* Reserve physical eraseblocks */
        if (vol->reserved_pebs > ubi->avail_pebs) {
-               dbg_err("not enough PEBs, only %d available", ubi->avail_pebs);
+               ubi_err("not enough PEBs, only %d available", ubi->avail_pebs);
+               if (ubi->corr_peb_count)
+                       ubi_err("%d PEBs are corrupted and not used",
+                               ubi->corr_peb_count);
                err = -ENOSPC;
                goto out_unlock;
        }
@@ -270,14 +272,14 @@ int ubi_create_volume(struct ubi_device *ubi, struct ubi_mkvol_req *req)
        vol->data_pad  = ubi->leb_size % vol->alignment;
        vol->vol_type  = req->vol_type;
        vol->name_len  = req->name_len;
-       memcpy(vol->name, req->name, vol->name_len + 1);
+       memcpy(vol->name, req->name, vol->name_len);
        vol->ubi = ubi;
 
        /*
         * Finish all pending erases because there may be some LEBs belonging
         * to the same volume ID.
         */
-       err = ubi_wl_flush(ubi);
+       err = ubi_wl_flush(ubi, vol_id, UBI_ALL);
        if (err)
                goto out_acc;
 
@@ -296,10 +298,10 @@ int ubi_create_volume(struct ubi_device *ubi, struct ubi_mkvol_req *req)
                vol->used_bytes =
                        (long long)vol->used_ebs * vol->usable_leb_size;
        } else {
-               bytes = vol->used_bytes;
-               vol->last_eb_bytes = do_div(bytes, vol->usable_leb_size);
-               vol->used_ebs = bytes;
-               if (vol->last_eb_bytes)
+               vol->used_ebs = div_u64_rem(vol->used_bytes,
+                                           vol->usable_leb_size,
+                                           &vol->last_eb_bytes);
+               if (vol->last_eb_bytes != 0)
                        vol->used_ebs += 1;
                else
                        vol->last_eb_bytes = vol->usable_leb_size;
@@ -315,20 +317,16 @@ int ubi_create_volume(struct ubi_device *ubi, struct ubi_mkvol_req *req)
                goto out_mapping;
        }
 
-       err = ubi_create_gluebi(ubi, vol);
-       if (err)
-               goto out_cdev;
-
        vol->dev.release = vol_release;
        vol->dev.parent = &ubi->dev;
        vol->dev.devt = dev;
        vol->dev.class = ubi_class;
 
-       sprintf(&vol->dev.bus_id[0], "%s_%d", ubi->ubi_name, vol->vol_id);
+       dev_set_name(&vol->dev, "%s_%d", ubi->ubi_name, vol->vol_id);
        err = device_register(&vol->dev);
        if (err) {
                ubi_err("cannot register device");
-               goto out_gluebi;
+               goto out_cdev;
        }
 
        err = volume_sysfs_init(ubi, vol);
@@ -345,7 +343,7 @@ int ubi_create_volume(struct ubi_device *ubi, struct ubi_mkvol_req *req)
                vtbl_rec.vol_type = UBI_VID_DYNAMIC;
        else
                vtbl_rec.vol_type = UBI_VID_STATIC;
-       memcpy(vtbl_rec.name, vol->name, vol->name_len + 1);
+       memcpy(vtbl_rec.name, vol->name, vol->name_len);
 
        err = ubi_change_vtbl_record(ubi, vol_id, &vtbl_rec);
        if (err)
@@ -356,39 +354,37 @@ int ubi_create_volume(struct ubi_device *ubi, struct ubi_mkvol_req *req)
        ubi->vol_count += 1;
        spin_unlock(&ubi->volumes_lock);
 
-       paranoid_check_volumes(ubi);
-       return 0;
+       ubi_volume_notify(ubi, vol, UBI_VOLUME_ADDED);
+       self_check_volumes(ubi);
+       return err;
 
 out_sysfs:
        /*
-        * We have registered our device, we should not free the volume*
+        * We have registered our device, we should not free the volume
         * description object in this function in case of an error - it is
         * freed by the release function.
         *
         * Get device reference to prevent the release function from being
         * called just after sysfs has been closed.
         */
-       dont_free = 1;
+       do_free = 0;
        get_device(&vol->dev);
        volume_sysfs_close(vol);
-out_gluebi:
-       if (ubi_destroy_gluebi(vol))
-               dbg_err("cannot destroy gluebi for volume %d:%d",
-                       ubi->ubi_num, vol_id);
 out_cdev:
        cdev_del(&vol->cdev);
 out_mapping:
-       kfree(vol->eba_tbl);
+       if (do_free)
+               kfree(vol->eba_tbl);
 out_acc:
        spin_lock(&ubi->volumes_lock);
        ubi->rsvd_pebs -= vol->reserved_pebs;
        ubi->avail_pebs += vol->reserved_pebs;
 out_unlock:
        spin_unlock(&ubi->volumes_lock);
-       if (dont_free)
-               put_device(&vol->dev);
-       else
+       if (do_free)
                kfree(vol);
+       else
+               put_device(&vol->dev);
        ubi_err("cannot create volume %d, error %d", vol_id, err);
        return err;
 }
@@ -396,19 +392,20 @@ out_unlock:
 /**
  * ubi_remove_volume - remove volume.
  * @desc: volume descriptor
+ * @no_vtbl: do not change volume table if not zero
  *
  * This function removes volume described by @desc. The volume has to be opened
  * in "exclusive" mode. Returns zero in case of success and a negative error
- * code in case of failure. The caller has to have the @ubi->volumes_mutex
+ * code in case of failure. The caller has to have the @ubi->device_mutex
  * locked.
  */
-int ubi_remove_volume(struct ubi_volume_desc *desc)
+int ubi_remove_volume(struct ubi_volume_desc *desc, int no_vtbl)
 {
        struct ubi_volume *vol = desc->vol;
        struct ubi_device *ubi = vol->ubi;
        int i, err, vol_id = vol->vol_id, reserved_pebs = vol->reserved_pebs;
 
-       dbg_msg("remove UBI volume %d", vol_id);
+       dbg_gen("remove device %d, volume %d", ubi->ubi_num, vol_id);
        ubi_assert(desc->mode == UBI_EXCLUSIVE);
        ubi_assert(vol == ubi->volumes[vol_id]);
 
@@ -427,13 +424,11 @@ int ubi_remove_volume(struct ubi_volume_desc *desc)
        ubi->volumes[vol_id] = NULL;
        spin_unlock(&ubi->volumes_lock);
 
-       err = ubi_destroy_gluebi(vol);
-       if (err)
-               goto out_err;
-
-       err = ubi_change_vtbl_record(ubi, vol_id, NULL);
-       if (err)
-               goto out_err;
+       if (!no_vtbl) {
+               err = ubi_change_vtbl_record(ubi, vol_id, NULL);
+               if (err)
+                       goto out_err;
+       }
 
        for (i = 0; i < vol->reserved_pebs; i++) {
                err = ubi_eba_unmap_leb(ubi, vol, i);
@@ -441,28 +436,21 @@ int ubi_remove_volume(struct ubi_volume_desc *desc)
                        goto out_err;
        }
 
-       kfree(vol->eba_tbl);
-       vol->eba_tbl = NULL;
        cdev_del(&vol->cdev);
        volume_sysfs_close(vol);
 
        spin_lock(&ubi->volumes_lock);
        ubi->rsvd_pebs -= reserved_pebs;
        ubi->avail_pebs += reserved_pebs;
-       i = ubi->beb_rsvd_level - ubi->beb_rsvd_pebs;
-       if (i > 0) {
-               i = ubi->avail_pebs >= i ? i : ubi->avail_pebs;
-               ubi->avail_pebs -= i;
-               ubi->rsvd_pebs += i;
-               ubi->beb_rsvd_pebs += i;
-               if (i > 0)
-                       ubi_msg("reserve more %d PEBs", i);
-       }
+       ubi_update_reserved(ubi);
        ubi->vol_count -= 1;
        spin_unlock(&ubi->volumes_lock);
 
-       paranoid_check_volumes(ubi);
-       return 0;
+       ubi_volume_notify(ubi, vol, UBI_VOLUME_REMOVED);
+       if (!no_vtbl)
+               self_check_volumes(ubi);
+
+       return err;
 
 out_err:
        ubi_err("cannot remove volume %d, error %d", vol_id, err);
@@ -480,7 +468,7 @@ out_unlock:
  *
  * This function re-sizes the volume and returns zero in case of success, and a
  * negative error code in case of failure. The caller has to have the
- * @ubi->volumes_mutex locked.
+ * @ubi->device_mutex locked.
  */
 int ubi_resize_volume(struct ubi_volume_desc *desc, int reserved_pebs)
 {
@@ -493,12 +481,12 @@ int ubi_resize_volume(struct ubi_volume_desc *desc, int reserved_pebs)
        if (ubi->ro_mode)
                return -EROFS;
 
-       dbg_msg("re-size volume %d to from %d to %d PEBs",
-               vol_id, vol->reserved_pebs, reserved_pebs);
+       dbg_gen("re-size device %d, volume %d to from %d to %d PEBs",
+               ubi->ubi_num, vol_id, vol->reserved_pebs, reserved_pebs);
 
        if (vol->vol_type == UBI_STATIC_VOLUME &&
            reserved_pebs < vol->used_ebs) {
-               dbg_err("too small size %d, %d LEBs contain data",
+               ubi_err("too small size %d, %d LEBs contain data",
                        reserved_pebs, vol->used_ebs);
                return -EINVAL;
        }
@@ -527,8 +515,11 @@ int ubi_resize_volume(struct ubi_volume_desc *desc, int reserved_pebs)
        if (pebs > 0) {
                spin_lock(&ubi->volumes_lock);
                if (pebs > ubi->avail_pebs) {
-                       dbg_err("not enough PEBs: requested %d, available %d",
+                       ubi_err("not enough PEBs: requested %d, available %d",
                                pebs, ubi->avail_pebs);
+                       if (ubi->corr_peb_count)
+                               ubi_err("%d PEBs are corrupted and not used",
+                                       ubi->corr_peb_count);
                        spin_unlock(&ubi->volumes_lock);
                        err = -ENOSPC;
                        goto out_free;
@@ -543,7 +534,7 @@ int ubi_resize_volume(struct ubi_volume_desc *desc, int reserved_pebs)
        }
 
        /* Change volume table record */
-       memcpy(&vtbl_rec, &ubi->vtbl[vol_id], sizeof(struct ubi_vtbl_record));
+       vtbl_rec = ubi->vtbl[vol_id];
        vtbl_rec.reserved_pebs = cpu_to_be32(reserved_pebs);
        err = ubi_change_vtbl_record(ubi, vol_id, &vtbl_rec);
        if (err)
@@ -558,15 +549,7 @@ int ubi_resize_volume(struct ubi_volume_desc *desc, int reserved_pebs)
                spin_lock(&ubi->volumes_lock);
                ubi->rsvd_pebs += pebs;
                ubi->avail_pebs -= pebs;
-               pebs = ubi->beb_rsvd_level - ubi->beb_rsvd_pebs;
-               if (pebs > 0) {
-                       pebs = ubi->avail_pebs >= pebs ? pebs : ubi->avail_pebs;
-                       ubi->avail_pebs -= pebs;
-                       ubi->rsvd_pebs += pebs;
-                       ubi->beb_rsvd_pebs += pebs;
-                       if (pebs > 0)
-                               ubi_msg("reserve more %d PEBs", pebs);
-               }
+               ubi_update_reserved(ubi);
                for (i = 0; i < reserved_pebs; i++)
                        new_mapping[i] = vol->eba_tbl[i];
                kfree(vol->eba_tbl);
@@ -582,8 +565,9 @@ int ubi_resize_volume(struct ubi_volume_desc *desc, int reserved_pebs)
                        (long long)vol->used_ebs * vol->usable_leb_size;
        }
 
-       paranoid_check_volumes(ubi);
-       return 0;
+       ubi_volume_notify(ubi, vol, UBI_VOLUME_RESIZED);
+       self_check_volumes(ubi);
+       return err;
 
 out_acc:
        if (pebs > 0) {
@@ -597,6 +581,45 @@ out_free:
        return err;
 }
 
+/**
+ * ubi_rename_volumes - re-name UBI volumes.
+ * @ubi: UBI device description object
+ * @rename_list: list of &struct ubi_rename_entry objects
+ *
+ * This function re-names or removes volumes specified in the re-name list.
+ * Returns zero in case of success and a negative error code in case of
+ * failure.
+ */
+int ubi_rename_volumes(struct ubi_device *ubi, struct list_head *rename_list)
+{
+       int err;
+       struct ubi_rename_entry *re;
+
+       err = ubi_vtbl_rename_volumes(ubi, rename_list);
+       if (err)
+               return err;
+
+       list_for_each_entry(re, rename_list, list) {
+               if (re->remove) {
+                       err = ubi_remove_volume(re->desc, 1);
+                       if (err)
+                               break;
+               } else {
+                       struct ubi_volume *vol = re->desc->vol;
+
+                       spin_lock(&ubi->volumes_lock);
+                       vol->name_len = re->new_name_len;
+                       memcpy(vol->name, re->new_name, re->new_name_len + 1);
+                       spin_unlock(&ubi->volumes_lock);
+                       ubi_volume_notify(ubi, vol, UBI_VOLUME_RENAMED);
+               }
+       }
+
+       if (!err)
+               self_check_volumes(ubi);
+       return err;
+}
+
 /**
  * ubi_add_volume - add volume.
  * @ubi: UBI device description object
@@ -611,8 +634,7 @@ int ubi_add_volume(struct ubi_device *ubi, struct ubi_volume *vol)
        int err, vol_id = vol->vol_id;
        dev_t dev;
 
-       dbg_msg("add volume %d", vol_id);
-       ubi_dbg_dump_vol_info(vol);
+       dbg_gen("add volume %d", vol_id);
 
        /* Register character device for the volume */
        cdev_init(&vol->cdev, &ubi_vol_cdev_operations);
@@ -625,32 +647,25 @@ int ubi_add_volume(struct ubi_device *ubi, struct ubi_volume *vol)
                return err;
        }
 
-       err = ubi_create_gluebi(ubi, vol);
-       if (err)
-               goto out_cdev;
-
        vol->dev.release = vol_release;
        vol->dev.parent = &ubi->dev;
        vol->dev.devt = dev;
        vol->dev.class = ubi_class;
-       sprintf(&vol->dev.bus_id[0], "%s_%d", ubi->ubi_name, vol->vol_id);
+       dev_set_name(&vol->dev, "%s_%d", ubi->ubi_name, vol->vol_id);
        err = device_register(&vol->dev);
        if (err)
-               goto out_gluebi;
+               goto out_cdev;
 
        err = volume_sysfs_init(ubi, vol);
        if (err) {
                cdev_del(&vol->cdev);
-               err = ubi_destroy_gluebi(vol);
                volume_sysfs_close(vol);
                return err;
        }
 
-       paranoid_check_volumes(ubi);
-       return 0;
+       self_check_volumes(ubi);
+       return err;
 
-out_gluebi:
-       err = ubi_destroy_gluebi(vol);
 out_cdev:
        cdev_del(&vol->cdev);
        return err;
@@ -666,22 +681,21 @@ out_cdev:
  */
 void ubi_free_volume(struct ubi_device *ubi, struct ubi_volume *vol)
 {
-       dbg_msg("free volume %d", vol->vol_id);
+       dbg_gen("free volume %d", vol->vol_id);
 
        ubi->volumes[vol->vol_id] = NULL;
-       ubi_destroy_gluebi(vol);
        cdev_del(&vol->cdev);
        volume_sysfs_close(vol);
 }
 
-#ifdef CONFIG_MTD_UBI_DEBUG_PARANOID
-
 /**
- * paranoid_check_volume - check volume information.
+ * self_check_volume - check volume information.
  * @ubi: UBI device description object
  * @vol_id: volume ID
+ *
+ * Returns zero if volume is all right and a a negative error code if not.
  */
-static void paranoid_check_volume(struct ubi_device *ubi, int vol_id)
+static int self_check_volume(struct ubi_device *ubi, int vol_id)
 {
        int idx = vol_id2idx(ubi, vol_id);
        int reserved_pebs, alignment, data_pad, vol_type, name_len, upd_marker;
@@ -699,16 +713,7 @@ static void paranoid_check_volume(struct ubi_device *ubi, int vol_id)
                        goto fail;
                }
                spin_unlock(&ubi->volumes_lock);
-               return;
-       }
-
-       if (vol->exclusive) {
-               /*
-                * The volume may be being created at the moment, do not check
-                * it (e.g., it may be in the middle of ubi_create_volume().
-                */
-               spin_unlock(&ubi->volumes_lock);
-               return;
+               return 0;
        }
 
        if (vol->reserved_pebs < 0 || vol->alignment < 0 || vol->data_pad < 0 ||
@@ -740,7 +745,7 @@ static void paranoid_check_volume(struct ubi_device *ubi, int vol_id)
        }
 
        if (vol->upd_marker && vol->corrupted) {
-               dbg_err("update marker and corrupted simultaneously");
+               ubi_err("update marker and corrupted simultaneously");
                goto fail;
        }
 
@@ -760,11 +765,6 @@ static void paranoid_check_volume(struct ubi_device *ubi, int vol_id)
                goto fail;
        }
 
-       if (!vol->name) {
-               ubi_err("NULL volume name");
-               goto fail;
-       }
-
        n = strnlen(vol->name, vol->name_len + 1);
        if (n != vol->name_len) {
                ubi_err("bad name_len %lld", n);
@@ -818,31 +818,42 @@ static void paranoid_check_volume(struct ubi_device *ubi, int vol_id)
 
        if (alignment != vol->alignment || data_pad != vol->data_pad ||
            upd_marker != vol->upd_marker || vol_type != vol->vol_type ||
-           name_len!= vol->name_len || strncmp(name, vol->name, name_len)) {
+           name_len != vol->name_len || strncmp(name, vol->name, name_len)) {
                ubi_err("volume info is different");
                goto fail;
        }
 
        spin_unlock(&ubi->volumes_lock);
-       return;
+       return 0;
 
 fail:
-       ubi_err("paranoid check failed for volume %d", vol_id);
-       ubi_dbg_dump_vol_info(vol);
-       ubi_dbg_dump_vtbl_record(&ubi->vtbl[vol_id], vol_id);
+       ubi_err("self-check failed for volume %d", vol_id);
+       if (vol)
+               ubi_dump_vol_info(vol);
+       ubi_dump_vtbl_record(&ubi->vtbl[vol_id], vol_id);
+       dump_stack();
        spin_unlock(&ubi->volumes_lock);
-       BUG();
+       return -EINVAL;
 }
 
 /**
- * paranoid_check_volumes - check information about all volumes.
+ * self_check_volumes - check information about all volumes.
  * @ubi: UBI device description object
+ *
+ * Returns zero if volumes are all right and a a negative error code if not.
  */
-static void paranoid_check_volumes(struct ubi_device *ubi)
+static int self_check_volumes(struct ubi_device *ubi)
 {
-       int i;
+       int i, err = 0;
 
-       for (i = 0; i < ubi->vtbl_slots; i++)
-               paranoid_check_volume(ubi, i);
+       if (!ubi_dbg_chk_gen(ubi))
+               return 0;
+
+       for (i = 0; i < ubi->vtbl_slots; i++) {
+               err = self_check_volume(ubi, i);
+               if (err)
+                       break;
+       }
+
+       return err;
 }
-#endif
index 3fbb4a0a9d40b2884b77962d21d05d22f7be799c..e6c8f5bbe0e00f54d32b180a95f1a9f28b7633b7 100644 (file)
  * LEB 1. This scheme guarantees recoverability from unclean reboots.
  *
  * In this UBI implementation the on-flash volume table does not contain any
- * information about how many data static volumes contain. This information may
- * be found from the scanning data.
+ * information about how much data static volumes contain.
  *
  * But it would still be beneficial to store this information in the volume
  * table. For example, suppose we have a static volume X, and all its physical
  * eraseblocks became bad for some reasons. Suppose we are attaching the
- * corresponding MTD device, the scanning has found no logical eraseblocks
+ * corresponding MTD device, for some reason we find no logical eraseblocks
  * corresponding to the volume X. According to the volume table volume X does
  * exist. So we don't know whether it is just empty or all its physical
- * eraseblocks went bad. So we cannot alarm the user about this corruption.
+ * eraseblocks went bad. So we cannot alarm the user properly.
  *
  * The volume table also stores so-called "update marker", which is used for
  * volume updates. Before updating the volume, the update marker is set, and
  * damaged.
  */
 
-#ifdef UBI_LINUX
+#define __UBOOT__
+#ifndef __UBOOT__
 #include <linux/crc32.h>
 #include <linux/err.h>
+#include <linux/slab.h>
 #include <asm/div64.h>
+#else
+#include <ubi_uboot.h>
 #endif
 
-#include <ubi_uboot.h>
+#include <linux/err.h>
 #include "ubi.h"
 
-#ifdef CONFIG_MTD_UBI_DEBUG_PARANOID
-static void paranoid_vtbl_check(const struct ubi_device *ubi);
-#else
-#define paranoid_vtbl_check(ubi)
-#endif
+static void self_vtbl_check(const struct ubi_device *ubi);
 
 /* Empty volume table record */
 static struct ubi_vtbl_record empty_vtbl_record;
@@ -97,18 +96,68 @@ int ubi_change_vtbl_record(struct ubi_device *ubi, int idx,
                        return err;
 
                err = ubi_eba_write_leb(ubi, layout_vol, i, ubi->vtbl, 0,
-                                       ubi->vtbl_size, UBI_LONGTERM);
+                                       ubi->vtbl_size);
+               if (err)
+                       return err;
+       }
+
+       self_vtbl_check(ubi);
+       return 0;
+}
+
+/**
+ * ubi_vtbl_rename_volumes - rename UBI volumes in the volume table.
+ * @ubi: UBI device description object
+ * @rename_list: list of &struct ubi_rename_entry objects
+ *
+ * This function re-names multiple volumes specified in @req in the volume
+ * table. Returns zero in case of success and a negative error code in case of
+ * failure.
+ */
+int ubi_vtbl_rename_volumes(struct ubi_device *ubi,
+                           struct list_head *rename_list)
+{
+       int i, err;
+       struct ubi_rename_entry *re;
+       struct ubi_volume *layout_vol;
+
+       list_for_each_entry(re, rename_list, list) {
+               uint32_t crc;
+               struct ubi_volume *vol = re->desc->vol;
+               struct ubi_vtbl_record *vtbl_rec = &ubi->vtbl[vol->vol_id];
+
+               if (re->remove) {
+                       memcpy(vtbl_rec, &empty_vtbl_record,
+                              sizeof(struct ubi_vtbl_record));
+                       continue;
+               }
+
+               vtbl_rec->name_len = cpu_to_be16(re->new_name_len);
+               memcpy(vtbl_rec->name, re->new_name, re->new_name_len);
+               memset(vtbl_rec->name + re->new_name_len, 0,
+                      UBI_VOL_NAME_MAX + 1 - re->new_name_len);
+               crc = crc32(UBI_CRC32_INIT, vtbl_rec,
+                           UBI_VTBL_RECORD_SIZE_CRC);
+               vtbl_rec->crc = cpu_to_be32(crc);
+       }
+
+       layout_vol = ubi->volumes[vol_id2idx(ubi, UBI_LAYOUT_VOLUME_ID)];
+       for (i = 0; i < UBI_LAYOUT_VOLUME_EBS; i++) {
+               err = ubi_eba_unmap_leb(ubi, layout_vol, i);
+               if (err)
+                       return err;
+
+               err = ubi_eba_write_leb(ubi, layout_vol, i, ubi->vtbl, 0,
+                                       ubi->vtbl_size);
                if (err)
                        return err;
        }
 
-       paranoid_vtbl_check(ubi);
        return 0;
 }
 
 /**
- * vtbl_check - check if volume table is not corrupted and contains sensible
- *              data.
+ * vtbl_check - check if volume table is not corrupted and sensible.
  * @ubi: UBI device description object
  * @vtbl: volume table
  *
@@ -132,13 +181,13 @@ static int vtbl_check(const struct ubi_device *ubi,
                upd_marker = vtbl[i].upd_marker;
                vol_type = vtbl[i].vol_type;
                name_len = be16_to_cpu(vtbl[i].name_len);
-               name = (const char *) &vtbl[i].name[0];
+               name = &vtbl[i].name[0];
 
                crc = crc32(UBI_CRC32_INIT, &vtbl[i], UBI_VTBL_RECORD_SIZE_CRC);
                if (be32_to_cpu(vtbl[i].crc) != crc) {
                        ubi_err("bad CRC at record %u: %#08x, not %#08x",
                                 i, crc, be32_to_cpu(vtbl[i].crc));
-                       ubi_dbg_dump_vtbl_record(&vtbl[i], i);
+                       ubi_dump_vtbl_record(&vtbl[i], i);
                        return 1;
                }
 
@@ -170,7 +219,7 @@ static int vtbl_check(const struct ubi_device *ubi,
 
                n = ubi->leb_size % alignment;
                if (data_pad != n) {
-                       dbg_err("bad data_pad, has to be %d", n);
+                       ubi_err("bad data_pad, has to be %d", n);
                        err = 6;
                        goto bad;
                }
@@ -186,8 +235,8 @@ static int vtbl_check(const struct ubi_device *ubi,
                }
 
                if (reserved_pebs > ubi->good_peb_count) {
-                       dbg_err("too large reserved_pebs, good PEBs %d",
-                               ubi->good_peb_count);
+                       ubi_err("too large reserved_pebs %d, good PEBs %d",
+                               reserved_pebs, ubi->good_peb_count);
                        err = 9;
                        goto bad;
                }
@@ -215,11 +264,15 @@ static int vtbl_check(const struct ubi_device *ubi,
                        int len2 = be16_to_cpu(vtbl[n].name_len);
 
                        if (len1 > 0 && len1 == len2 &&
-                           !strncmp((char *)vtbl[i].name, (char *)vtbl[n].name, len1)) {
-                               ubi_err("volumes %d and %d have the same name"
-                                       " \"%s\"", i, n, vtbl[i].name);
-                               ubi_dbg_dump_vtbl_record(&vtbl[i], i);
-                               ubi_dbg_dump_vtbl_record(&vtbl[n], n);
+#ifndef __UBOOT__
+                           !strncmp(vtbl[i].name, vtbl[n].name, len1)) {
+#else
+                           !strncmp((char *)vtbl[i].name, vtbl[n].name, len1)) {
+#endif
+                               ubi_err("volumes %d and %d have the same name \"%s\"",
+                                       i, n, vtbl[i].name);
+                               ubi_dump_vtbl_record(&vtbl[i], i);
+                               ubi_dump_vtbl_record(&vtbl[n], n);
                                return -EINVAL;
                        }
                }
@@ -229,76 +282,64 @@ static int vtbl_check(const struct ubi_device *ubi,
 
 bad:
        ubi_err("volume table check failed: record %d, error %d", i, err);
-       ubi_dbg_dump_vtbl_record(&vtbl[i], i);
+       ubi_dump_vtbl_record(&vtbl[i], i);
        return -EINVAL;
 }
 
 /**
  * create_vtbl - create a copy of volume table.
  * @ubi: UBI device description object
- * @si: scanning information
+ * @ai: attaching information
  * @copy: number of the volume table copy
  * @vtbl: contents of the volume table
  *
  * This function returns zero in case of success and a negative error code in
  * case of failure.
  */
-static int create_vtbl(struct ubi_device *ubi, struct ubi_scan_info *si,
+static int create_vtbl(struct ubi_device *ubi, struct ubi_attach_info *ai,
                       int copy, void *vtbl)
 {
        int err, tries = 0;
-       static struct ubi_vid_hdr *vid_hdr;
-       struct ubi_scan_volume *sv;
-       struct ubi_scan_leb *new_seb, *old_seb = NULL;
+       struct ubi_vid_hdr *vid_hdr;
+       struct ubi_ainf_peb *new_aeb;
 
-       ubi_msg("create volume table (copy #%d)", copy + 1);
+       dbg_gen("create volume table (copy #%d)", copy + 1);
 
        vid_hdr = ubi_zalloc_vid_hdr(ubi, GFP_KERNEL);
        if (!vid_hdr)
                return -ENOMEM;
 
-       /*
-        * Check if there is a logical eraseblock which would have to contain
-        * this volume table copy was found during scanning. It has to be wiped
-        * out.
-        */
-       sv = ubi_scan_find_sv(si, UBI_LAYOUT_VOLUME_ID);
-       if (sv)
-               old_seb = ubi_scan_find_seb(sv, copy);
-
 retry:
-       new_seb = ubi_scan_get_free_peb(ubi, si);
-       if (IS_ERR(new_seb)) {
-               err = PTR_ERR(new_seb);
+       new_aeb = ubi_early_get_peb(ubi, ai);
+       if (IS_ERR(new_aeb)) {
+               err = PTR_ERR(new_aeb);
                goto out_free;
        }
 
-       vid_hdr->vol_type = UBI_VID_DYNAMIC;
+       vid_hdr->vol_type = UBI_LAYOUT_VOLUME_TYPE;
        vid_hdr->vol_id = cpu_to_be32(UBI_LAYOUT_VOLUME_ID);
        vid_hdr->compat = UBI_LAYOUT_VOLUME_COMPAT;
        vid_hdr->data_size = vid_hdr->used_ebs =
                             vid_hdr->data_pad = cpu_to_be32(0);
        vid_hdr->lnum = cpu_to_be32(copy);
-       vid_hdr->sqnum = cpu_to_be64(++si->max_sqnum);
-       vid_hdr->leb_ver = cpu_to_be32(old_seb ? old_seb->leb_ver + 1: 0);
+       vid_hdr->sqnum = cpu_to_be64(++ai->max_sqnum);
 
        /* The EC header is already there, write the VID header */
-       err = ubi_io_write_vid_hdr(ubi, new_seb->pnum, vid_hdr);
+       err = ubi_io_write_vid_hdr(ubi, new_aeb->pnum, vid_hdr);
        if (err)
                goto write_error;
 
        /* Write the layout volume contents */
-       err = ubi_io_write_data(ubi, vtbl, new_seb->pnum, 0, ubi->vtbl_size);
+       err = ubi_io_write_data(ubi, vtbl, new_aeb->pnum, 0, ubi->vtbl_size);
        if (err)
                goto write_error;
 
        /*
-        * And add it to the scanning information. Don't delete the old
-        * @old_seb as it will be deleted and freed in 'ubi_scan_add_used()'.
+        * And add it to the attaching information. Don't delete the old version
+        * of this LEB as it will be deleted and freed in 'ubi_add_to_av()'.
         */
-       err = ubi_scan_add_used(ubi, si, new_seb->pnum, new_seb->ec,
-                               vid_hdr, 0);
-       kfree(new_seb);
+       err = ubi_add_to_av(ubi, ai, new_aeb->pnum, new_aeb->ec, vid_hdr, 0);
+       kmem_cache_free(ai->aeb_slab_cache, new_aeb);
        ubi_free_vid_hdr(ubi, vid_hdr);
        return err;
 
@@ -308,10 +349,10 @@ write_error:
                 * Probably this physical eraseblock went bad, try to pick
                 * another one.
                 */
-               list_add_tail(&new_seb->u.list, &si->corr);
+               list_add(&new_aeb->u.list, &ai->erase);
                goto retry;
        }
-       kfree(new_seb);
+       kmem_cache_free(ai->aeb_slab_cache, new_aeb);
 out_free:
        ubi_free_vid_hdr(ubi, vid_hdr);
        return err;
@@ -321,20 +362,20 @@ out_free:
 /**
  * process_lvol - process the layout volume.
  * @ubi: UBI device description object
- * @si: scanning information
- * @sv: layout volume scanning information
+ * @ai: attaching information
+ * @av: layout volume attaching information
  *
  * This function is responsible for reading the layout volume, ensuring it is
  * not corrupted, and recovering from corruptions if needed. Returns volume
  * table in case of success and a negative error code in case of failure.
  */
 static struct ubi_vtbl_record *process_lvol(struct ubi_device *ubi,
-                                           struct ubi_scan_info *si,
-                                           struct ubi_scan_volume *sv)
+                                           struct ubi_attach_info *ai,
+                                           struct ubi_ainf_volume *av)
 {
        int err;
        struct rb_node *rb;
-       struct ubi_scan_leb *seb;
+       struct ubi_ainf_peb *aeb;
        struct ubi_vtbl_record *leb[UBI_LAYOUT_VOLUME_EBS] = { NULL, NULL };
        int leb_corrupted[UBI_LAYOUT_VOLUME_EBS] = {1, 1};
 
@@ -356,25 +397,24 @@ static struct ubi_vtbl_record *process_lvol(struct ubi_device *ubi,
         * 0 contains more recent information.
         *
         * So the plan is to first check LEB 0. Then
-        * a. if LEB 0 is OK, it must be containing the most resent data; then
+        * a. if LEB 0 is OK, it must be containing the most recent data; then
         *    we compare it with LEB 1, and if they are different, we copy LEB
         *    0 to LEB 1;
         * b. if LEB 0 is corrupted, but LEB 1 has to be OK, and we copy LEB 1
         *    to LEB 0.
         */
 
-       dbg_msg("check layout volume");
+       dbg_gen("check layout volume");
 
        /* Read both LEB 0 and LEB 1 into memory */
-       ubi_rb_for_each_entry(rb, seb, &sv->root, u.rb) {
-               leb[seb->lnum] = vmalloc(ubi->vtbl_size);
-               if (!leb[seb->lnum]) {
+       ubi_rb_for_each_entry(rb, aeb, &av->root, u.rb) {
+               leb[aeb->lnum] = vzalloc(ubi->vtbl_size);
+               if (!leb[aeb->lnum]) {
                        err = -ENOMEM;
                        goto out_free;
                }
-               memset(leb[seb->lnum], 0, ubi->vtbl_size);
 
-               err = ubi_io_read_data(ubi, leb[seb->lnum], seb->pnum, 0,
+               err = ubi_io_read_data(ubi, leb[aeb->lnum], aeb->pnum, 0,
                                       ubi->vtbl_size);
                if (err == UBI_IO_BITFLIPS || mtd_is_eccerr(err))
                        /*
@@ -382,12 +422,12 @@ static struct ubi_vtbl_record *process_lvol(struct ubi_device *ubi,
                         * uncorrectable ECC error, but we have our own CRC and
                         * the data will be checked later. If the data is OK,
                         * the PEB will be scrubbed (because we set
-                        * seb->scrub). If the data is not OK, the contents of
+                        * aeb->scrub). If the data is not OK, the contents of
                         * the PEB will be recovered from the second copy, and
-                        * seb->scrub will be cleared in
-                        * 'ubi_scan_add_used()'.
+                        * aeb->scrub will be cleared in
+                        * 'ubi_add_to_av()'.
                         */
-                       seb->scrub = 1;
+                       aeb->scrub = 1;
                else if (err)
                        goto out_free;
        }
@@ -402,10 +442,11 @@ static struct ubi_vtbl_record *process_lvol(struct ubi_device *ubi,
        if (!leb_corrupted[0]) {
                /* LEB 0 is OK */
                if (leb[1])
-                       leb_corrupted[1] = memcmp(leb[0], leb[1], ubi->vtbl_size);
+                       leb_corrupted[1] = memcmp(leb[0], leb[1],
+                                                 ubi->vtbl_size);
                if (leb_corrupted[1]) {
                        ubi_warn("volume table copy #2 is corrupted");
-                       err = create_vtbl(ubi, si, 1, leb[0]);
+                       err = create_vtbl(ubi, ai, 1, leb[0]);
                        if (err)
                                goto out_free;
                        ubi_msg("volume table was restored");
@@ -428,7 +469,7 @@ static struct ubi_vtbl_record *process_lvol(struct ubi_device *ubi,
                }
 
                ubi_warn("volume table copy #1 is corrupted");
-               err = create_vtbl(ubi, si, 0, leb[1]);
+               err = create_vtbl(ubi, ai, 0, leb[1]);
                if (err)
                        goto out_free;
                ubi_msg("volume table was restored");
@@ -446,21 +487,20 @@ out_free:
 /**
  * create_empty_lvol - create empty layout volume.
  * @ubi: UBI device description object
- * @si: scanning information
+ * @ai: attaching information
  *
  * This function returns volume table contents in case of success and a
  * negative error code in case of failure.
  */
 static struct ubi_vtbl_record *create_empty_lvol(struct ubi_device *ubi,
-                                                struct ubi_scan_info *si)
+                                                struct ubi_attach_info *ai)
 {
        int i;
        struct ubi_vtbl_record *vtbl;
 
-       vtbl = vmalloc(ubi->vtbl_size);
+       vtbl = vzalloc(ubi->vtbl_size);
        if (!vtbl)
                return ERR_PTR(-ENOMEM);
-       memset(vtbl, 0, ubi->vtbl_size);
 
        for (i = 0; i < ubi->vtbl_slots; i++)
                memcpy(&vtbl[i], &empty_vtbl_record, UBI_VTBL_RECORD_SIZE);
@@ -468,7 +508,7 @@ static struct ubi_vtbl_record *create_empty_lvol(struct ubi_device *ubi,
        for (i = 0; i < UBI_LAYOUT_VOLUME_EBS; i++) {
                int err;
 
-               err = create_vtbl(ubi, si, i, vtbl);
+               err = create_vtbl(ubi, ai, i, vtbl);
                if (err) {
                        vfree(vtbl);
                        return ERR_PTR(err);
@@ -481,18 +521,19 @@ static struct ubi_vtbl_record *create_empty_lvol(struct ubi_device *ubi,
 /**
  * init_volumes - initialize volume information for existing volumes.
  * @ubi: UBI device description object
- * @si: scanning information
+ * @ai: scanning information
  * @vtbl: volume table
  *
  * This function allocates volume description objects for existing volumes.
  * Returns zero in case of success and a negative error code in case of
  * failure.
  */
-static int init_volumes(struct ubi_device *ubi, const struct ubi_scan_info *si,
+static int init_volumes(struct ubi_device *ubi,
+                       const struct ubi_attach_info *ai,
                        const struct ubi_vtbl_record *vtbl)
 {
        int i, reserved_pebs = 0;
-       struct ubi_scan_volume *sv;
+       struct ubi_ainf_volume *av;
        struct ubi_volume *vol;
 
        for (i = 0; i < ubi->vtbl_slots; i++) {
@@ -520,8 +561,8 @@ static int init_volumes(struct ubi_device *ubi, const struct ubi_scan_info *si,
                if (vtbl[i].flags & UBI_VTBL_AUTORESIZE_FLG) {
                        /* Auto re-size flag may be set only for one volume */
                        if (ubi->autoresize_vol_id != -1) {
-                               ubi_err("more then one auto-resize volume (%d "
-                                       "and %d)", ubi->autoresize_vol_id, i);
+                               ubi_err("more than one auto-resize volume (%d and %d)",
+                                       ubi->autoresize_vol_id, i);
                                kfree(vol);
                                return -EINVAL;
                        }
@@ -548,8 +589,8 @@ static int init_volumes(struct ubi_device *ubi, const struct ubi_scan_info *si,
                }
 
                /* Static volumes only */
-               sv = ubi_scan_find_sv(si, i);
-               if (!sv) {
+               av = ubi_find_av(ai, i);
+               if (!av) {
                        /*
                         * No eraseblocks belonging to this volume found. We
                         * don't actually know whether this static volume is
@@ -561,22 +602,22 @@ static int init_volumes(struct ubi_device *ubi, const struct ubi_scan_info *si,
                        continue;
                }
 
-               if (sv->leb_count != sv->used_ebs) {
+               if (av->leb_count != av->used_ebs) {
                        /*
                         * We found a static volume which misses several
                         * eraseblocks. Treat it as corrupted.
                         */
                        ubi_warn("static volume %d misses %d LEBs - corrupted",
-                                sv->vol_id, sv->used_ebs - sv->leb_count);
+                                av->vol_id, av->used_ebs - av->leb_count);
                        vol->corrupted = 1;
                        continue;
                }
 
-               vol->used_ebs = sv->used_ebs;
+               vol->used_ebs = av->used_ebs;
                vol->used_bytes =
                        (long long)(vol->used_ebs - 1) * vol->usable_leb_size;
-               vol->used_bytes += sv->last_data_size;
-               vol->last_eb_bytes = sv->last_data_size;
+               vol->used_bytes += av->last_data_size;
+               vol->last_eb_bytes = av->last_data_size;
        }
 
        /* And add the layout volume */
@@ -585,7 +626,7 @@ static int init_volumes(struct ubi_device *ubi, const struct ubi_scan_info *si,
                return -ENOMEM;
 
        vol->reserved_pebs = UBI_LAYOUT_VOLUME_EBS;
-       vol->alignment = 1;
+       vol->alignment = UBI_LAYOUT_VOLUME_ALIGN;
        vol->vol_type = UBI_DYNAMIC_VOLUME;
        vol->name_len = sizeof(UBI_LAYOUT_VOLUME_NAME) - 1;
        memcpy(vol->name, UBI_LAYOUT_VOLUME_NAME, vol->name_len + 1);
@@ -603,9 +644,13 @@ static int init_volumes(struct ubi_device *ubi, const struct ubi_scan_info *si,
        ubi->vol_count += 1;
        vol->ubi = ubi;
 
-       if (reserved_pebs > ubi->avail_pebs)
+       if (reserved_pebs > ubi->avail_pebs) {
                ubi_err("not enough PEBs, required %d, available %d",
                        reserved_pebs, ubi->avail_pebs);
+               if (ubi->corr_peb_count)
+                       ubi_err("%d PEBs are corrupted and not used",
+                               ubi->corr_peb_count);
+       }
        ubi->rsvd_pebs += reserved_pebs;
        ubi->avail_pebs -= reserved_pebs;
 
@@ -613,105 +658,104 @@ static int init_volumes(struct ubi_device *ubi, const struct ubi_scan_info *si,
 }
 
 /**
- * check_sv - check volume scanning information.
+ * check_av - check volume attaching information.
  * @vol: UBI volume description object
- * @sv: volume scanning information
+ * @av: volume attaching information
  *
- * This function returns zero if the volume scanning information is consistent
+ * This function returns zero if the volume attaching information is consistent
  * to the data read from the volume tabla, and %-EINVAL if not.
  */
-static int check_sv(const struct ubi_volume *vol,
-                   const struct ubi_scan_volume *sv)
+static int check_av(const struct ubi_volume *vol,
+                   const struct ubi_ainf_volume *av)
 {
        int err;
 
-       if (sv->highest_lnum >= vol->reserved_pebs) {
+       if (av->highest_lnum >= vol->reserved_pebs) {
                err = 1;
                goto bad;
        }
-       if (sv->leb_count > vol->reserved_pebs) {
+       if (av->leb_count > vol->reserved_pebs) {
                err = 2;
                goto bad;
        }
-       if (sv->vol_type != vol->vol_type) {
+       if (av->vol_type != vol->vol_type) {
                err = 3;
                goto bad;
        }
-       if (sv->used_ebs > vol->reserved_pebs) {
+       if (av->used_ebs > vol->reserved_pebs) {
                err = 4;
                goto bad;
        }
-       if (sv->data_pad != vol->data_pad) {
+       if (av->data_pad != vol->data_pad) {
                err = 5;
                goto bad;
        }
        return 0;
 
 bad:
-       ubi_err("bad scanning information, error %d", err);
-       ubi_dbg_dump_sv(sv);
-       ubi_dbg_dump_vol_info(vol);
+       ubi_err("bad attaching information, error %d", err);
+       ubi_dump_av(av);
+       ubi_dump_vol_info(vol);
        return -EINVAL;
 }
 
 /**
- * check_scanning_info - check that scanning information.
+ * check_attaching_info - check that attaching information.
  * @ubi: UBI device description object
- * @si: scanning information
+ * @ai: attaching information
  *
  * Even though we protect on-flash data by CRC checksums, we still don't trust
- * the media. This function ensures that scanning information is consistent to
- * the information read from the volume table. Returns zero if the scanning
+ * the media. This function ensures that attaching information is consistent to
+ * the information read from the volume table. Returns zero if the attaching
  * information is OK and %-EINVAL if it is not.
  */
-static int check_scanning_info(const struct ubi_device *ubi,
-                              struct ubi_scan_info *si)
+static int check_attaching_info(const struct ubi_device *ubi,
+                              struct ubi_attach_info *ai)
 {
        int err, i;
-       struct ubi_scan_volume *sv;
+       struct ubi_ainf_volume *av;
        struct ubi_volume *vol;
 
-       if (si->vols_found > UBI_INT_VOL_COUNT + ubi->vtbl_slots) {
-               ubi_err("scanning found %d volumes, maximum is %d + %d",
-                       si->vols_found, UBI_INT_VOL_COUNT, ubi->vtbl_slots);
+       if (ai->vols_found > UBI_INT_VOL_COUNT + ubi->vtbl_slots) {
+               ubi_err("found %d volumes while attaching, maximum is %d + %d",
+                       ai->vols_found, UBI_INT_VOL_COUNT, ubi->vtbl_slots);
                return -EINVAL;
        }
 
-       if (si->highest_vol_id >= ubi->vtbl_slots + UBI_INT_VOL_COUNT &&
-           si->highest_vol_id < UBI_INTERNAL_VOL_START) {
-               ubi_err("too large volume ID %d found by scanning",
-                       si->highest_vol_id);
+       if (ai->highest_vol_id >= ubi->vtbl_slots + UBI_INT_VOL_COUNT &&
+           ai->highest_vol_id < UBI_INTERNAL_VOL_START) {
+               ubi_err("too large volume ID %d found", ai->highest_vol_id);
                return -EINVAL;
        }
 
        for (i = 0; i < ubi->vtbl_slots + UBI_INT_VOL_COUNT; i++) {
                cond_resched();
 
-               sv = ubi_scan_find_sv(si, i);
+               av = ubi_find_av(ai, i);
                vol = ubi->volumes[i];
                if (!vol) {
-                       if (sv)
-                               ubi_scan_rm_volume(si, sv);
+                       if (av)
+                               ubi_remove_av(ai, av);
                        continue;
                }
 
                if (vol->reserved_pebs == 0) {
                        ubi_assert(i < ubi->vtbl_slots);
 
-                       if (!sv)
+                       if (!av)
                                continue;
 
                        /*
-                        * During scanning we found a volume which does not
+                        * During attaching we found a volume which does not
                         * exist according to the information in the volume
                         * table. This must have happened due to an unclean
                         * reboot while the volume was being removed. Discard
                         * these eraseblocks.
                         */
-                       ubi_msg("finish volume %d removal", sv->vol_id);
-                       ubi_scan_rm_volume(si, sv);
-               } else if (sv) {
-                       err = check_sv(vol, sv);
+                       ubi_msg("finish volume %d removal", av->vol_id);
+                       ubi_remove_av(ai, av);
+               } else if (av) {
+                       err = check_av(vol, av);
                        if (err)
                                return err;
                }
@@ -721,19 +765,18 @@ static int check_scanning_info(const struct ubi_device *ubi,
 }
 
 /**
- * ubi_read_volume_table - read volume table.
- * information.
+ * ubi_read_volume_table - read the volume table.
  * @ubi: UBI device description object
- * @si: scanning information
+ * @ai: attaching information
  *
  * This function reads volume table, checks it, recover from errors if needed,
  * or creates it if needed. Returns zero in case of success and a negative
  * error code in case of failure.
  */
-int ubi_read_volume_table(struct ubi_device *ubi, struct ubi_scan_info *si)
+int ubi_read_volume_table(struct ubi_device *ubi, struct ubi_attach_info *ai)
 {
        int i, err;
-       struct ubi_scan_volume *sv;
+       struct ubi_ainf_volume *av;
 
        empty_vtbl_record.crc = cpu_to_be32(0xf116c36b);
 
@@ -748,8 +791,8 @@ int ubi_read_volume_table(struct ubi_device *ubi, struct ubi_scan_info *si)
        ubi->vtbl_size = ubi->vtbl_slots * UBI_VTBL_RECORD_SIZE;
        ubi->vtbl_size = ALIGN(ubi->vtbl_size, ubi->min_io_size);
 
-       sv = ubi_scan_find_sv(si, UBI_LAYOUT_VOLUME_ID);
-       if (!sv) {
+       av = ubi_find_av(ai, UBI_LAYOUT_VOLUME_ID);
+       if (!av) {
                /*
                 * No logical eraseblocks belonging to the layout volume were
                 * found. This could mean that the flash is just empty. In
@@ -758,8 +801,8 @@ int ubi_read_volume_table(struct ubi_device *ubi, struct ubi_scan_info *si)
                 * But if flash is not empty this must be a corruption or the
                 * MTD device just contains garbage.
                 */
-               if (si->is_empty) {
-                       ubi->vtbl = create_empty_lvol(ubi, si);
+               if (ai->is_empty) {
+                       ubi->vtbl = create_empty_lvol(ubi, ai);
                        if (IS_ERR(ubi->vtbl))
                                return PTR_ERR(ubi->vtbl);
                } else {
@@ -767,33 +810,33 @@ int ubi_read_volume_table(struct ubi_device *ubi, struct ubi_scan_info *si)
                        return -EINVAL;
                }
        } else {
-               if (sv->leb_count > UBI_LAYOUT_VOLUME_EBS) {
+               if (av->leb_count > UBI_LAYOUT_VOLUME_EBS) {
                        /* This must not happen with proper UBI images */
-                       dbg_err("too many LEBs (%d) in layout volume",
-                               sv->leb_count);
+                       ubi_err("too many LEBs (%d) in layout volume",
+                               av->leb_count);
                        return -EINVAL;
                }
 
-               ubi->vtbl = process_lvol(ubi, si, sv);
+               ubi->vtbl = process_lvol(ubi, ai, av);
                if (IS_ERR(ubi->vtbl))
                        return PTR_ERR(ubi->vtbl);
        }
 
-       ubi->avail_pebs = ubi->good_peb_count;
+       ubi->avail_pebs = ubi->good_peb_count - ubi->corr_peb_count;
 
        /*
         * The layout volume is OK, initialize the corresponding in-RAM data
         * structures.
         */
-       err = init_volumes(ubi, si, ubi->vtbl);
+       err = init_volumes(ubi, ai, ubi->vtbl);
        if (err)
                goto out_free;
 
        /*
-        * Get sure that the scanning information is consistent to the
+        * Make sure that the attaching information is consistent to the
         * information stored in the volume table.
         */
-       err = check_scanning_info(ubi, si);
+       err = check_attaching_info(ubi, ai);
        if (err)
                goto out_free;
 
@@ -801,26 +844,24 @@ int ubi_read_volume_table(struct ubi_device *ubi, struct ubi_scan_info *si)
 
 out_free:
        vfree(ubi->vtbl);
-       for (i = 0; i < ubi->vtbl_slots + UBI_INT_VOL_COUNT; i++)
-               if (ubi->volumes[i]) {
-                       kfree(ubi->volumes[i]);
-                       ubi->volumes[i] = NULL;
-               }
+       for (i = 0; i < ubi->vtbl_slots + UBI_INT_VOL_COUNT; i++) {
+               kfree(ubi->volumes[i]);
+               ubi->volumes[i] = NULL;
+       }
        return err;
 }
 
-#ifdef CONFIG_MTD_UBI_DEBUG_PARANOID
-
 /**
- * paranoid_vtbl_check - check volume table.
+ * self_vtbl_check - check volume table.
  * @ubi: UBI device description object
  */
-static void paranoid_vtbl_check(const struct ubi_device *ubi)
+static void self_vtbl_check(const struct ubi_device *ubi)
 {
+       if (!ubi_dbg_chk_gen(ubi))
+               return;
+
        if (vtbl_check(ubi, ubi->vtbl)) {
-               ubi_err("paranoid check failed");
+               ubi_err("self-check failed");
                BUG();
        }
 }
-
-#endif /* CONFIG_MTD_UBI_DEBUG_PARANOID */
index 1eaa88b36fef21ee15fb28e2b7f1aae423a9752a..2987ffc093338ab04a54f10efa8819dc2d0e00c5 100644 (file)
  */
 
 /*
- * UBI wear-leveling unit.
+ * UBI wear-leveling sub-system.
  *
- * This unit is responsible for wear-leveling. It works in terms of physical
- * eraseblocks and erase counters and knows nothing about logical eraseblocks,
- * volumes, etc. From this unit's perspective all physical eraseblocks are of
- * two types - used and free. Used physical eraseblocks are those that were
- * "get" by the 'ubi_wl_get_peb()' function, and free physical eraseblocks are
- * those that were put by the 'ubi_wl_put_peb()' function.
+ * This sub-system is responsible for wear-leveling. It works in terms of
+ * physical eraseblocks and erase counters and knows nothing about logical
+ * eraseblocks, volumes, etc. From this sub-system's perspective all physical
+ * eraseblocks are of two types - used and free. Used physical eraseblocks are
+ * those that were "get" by the 'ubi_wl_get_peb()' function, and free physical
+ * eraseblocks are those that were put by the 'ubi_wl_put_peb()' function.
  *
  * Physical eraseblocks returned by 'ubi_wl_get_peb()' have only erase counter
- * header. The rest of the physical eraseblock contains only 0xFF bytes.
+ * header. The rest of the physical eraseblock contains only %0xFF bytes.
  *
- * When physical eraseblocks are returned to the WL unit by means of the
+ * When physical eraseblocks are returned to the WL sub-system by means of the
  * 'ubi_wl_put_peb()' function, they are scheduled for erasure. The erasure is
  * done asynchronously in context of the per-UBI device background thread,
- * which is also managed by the WL unit.
+ * which is also managed by the WL sub-system.
  *
  * The wear-leveling is ensured by means of moving the contents of used
  * physical eraseblocks with low erase counter to free physical eraseblocks
  * with high erase counter.
  *
- * The 'ubi_wl_get_peb()' function accepts data type hints which help to pick
- * an "optimal" physical eraseblock. For example, when it is known that the
- * physical eraseblock will be "put" soon because it contains short-term data,
- * the WL unit may pick a free physical eraseblock with low erase counter, and
- * so forth.
+ * If the WL sub-system fails to erase a physical eraseblock, it marks it as
+ * bad.
  *
- * If the WL unit fails to erase a physical eraseblock, it marks it as bad.
+ * This sub-system is also responsible for scrubbing. If a bit-flip is detected
+ * in a physical eraseblock, it has to be moved. Technically this is the same
+ * as moving it for wear-leveling reasons.
  *
- * This unit is also responsible for scrubbing. If a bit-flip is detected in a
- * physical eraseblock, it has to be moved. Technically this is the same as
- * moving it for wear-leveling reasons.
+ * As it was said, for the UBI sub-system all physical eraseblocks are either
+ * "free" or "used". Free eraseblock are kept in the @wl->free RB-tree, while
+ * used eraseblocks are kept in @wl->used, @wl->erroneous, or @wl->scrub
+ * RB-trees, as well as (temporarily) in the @wl->pq queue.
  *
- * As it was said, for the UBI unit all physical eraseblocks are either "free"
- * or "used". Free eraseblock are kept in the @wl->free RB-tree, while used
- * eraseblocks are kept in a set of different RB-trees: @wl->used,
- * @wl->prot.pnum, @wl->prot.aec, and @wl->scrub.
+ * When the WL sub-system returns a physical eraseblock, the physical
+ * eraseblock is protected from being moved for some "time". For this reason,
+ * the physical eraseblock is not directly moved from the @wl->free tree to the
+ * @wl->used tree. There is a protection queue in between where this
+ * physical eraseblock is temporarily stored (@wl->pq).
+ *
+ * All this protection stuff is needed because:
+ *  o we don't want to move physical eraseblocks just after we have given them
+ *    to the user; instead, we first want to let users fill them up with data;
+ *
+ *  o there is a chance that the user will put the physical eraseblock very
+ *    soon, so it makes sense not to move it for some time, but wait.
+ *
+ * Physical eraseblocks stay protected only for limited time. But the "time" is
+ * measured in erase cycles in this case. This is implemented with help of the
+ * protection queue. Eraseblocks are put to the tail of this queue when they
+ * are returned by the 'ubi_wl_get_peb()', and eraseblocks are removed from the
+ * head of the queue on each erase operation (for any eraseblock). So the
+ * length of the queue defines how may (global) erase cycles PEBs are protected.
+ *
+ * To put it differently, each physical eraseblock has 2 main states: free and
+ * used. The former state corresponds to the @wl->free tree. The latter state
+ * is split up on several sub-states:
+ * o the WL movement is allowed (@wl->used tree);
+ * o the WL movement is disallowed (@wl->erroneous) because the PEB is
+ *   erroneous - e.g., there was a read error;
+ * o the WL movement is temporarily prohibited (@wl->pq queue);
+ * o scrubbing is needed (@wl->scrub tree).
+ *
+ * Depending on the sub-state, wear-leveling entries of the used physical
+ * eraseblocks may be kept in one of those structures.
  *
  * Note, in this implementation, we keep a small in-RAM object for each physical
  * eraseblock. This is surely not a scalable solution. But it appears to be good
  * enough for moderately large flashes and it is simple. In future, one may
- * re-work this unit and make it more scalable.
+ * re-work this sub-system and make it more scalable.
  *
- * At the moment this unit does not utilize the sequence number, which was
- * introduced relatively recently. But it would be wise to do this because the
- * sequence number of a logical eraseblock characterizes how old is it. For
+ * At the moment this sub-system does not utilize the sequence number, which
+ * was introduced relatively recently. But it would be wise to do this because
+ * the sequence number of a logical eraseblock characterizes how old is it. For
  * example, when we move a PEB with low erase counter, and we need to pick the
  * target PEB, we pick a PEB with the highest EC if our PEB is "old" and we
  * pick target PEB with an average EC if our PEB is not very "old". This is a
- * room for future re-works of the WL unit.
- *
- * FIXME: looks too complex, should be simplified (later).
+ * room for future re-works of the WL sub-system.
  */
 
-#ifdef UBI_LINUX
+#define __UBOOT__
+#ifndef __UBOOT__
 #include <linux/slab.h>
 #include <linux/crc32.h>
 #include <linux/freezer.h>
 #include <linux/kthread.h>
+#else
+#include <ubi_uboot.h>
 #endif
 
-#include <ubi_uboot.h>
 #include "ubi.h"
 
 /* Number of physical eraseblocks reserved for wear-leveling purposes */
 #define WL_RESERVED_PEBS 1
 
-/*
- * How many erase cycles are short term, unknown, and long term physical
- * eraseblocks protected.
- */
-#define ST_PROTECTION 16
-#define U_PROTECTION  10
-#define LT_PROTECTION 4
-
 /*
  * Maximum difference between two erase counters. If this threshold is
- * exceeded, the WL unit starts moving data from used physical eraseblocks with
- * low erase counter to free physical eraseblocks with high erase counter.
+ * exceeded, the WL sub-system starts moving data from used physical
+ * eraseblocks with low erase counter to free physical eraseblocks with high
+ * erase counter.
  */
 #define UBI_WL_THRESHOLD CONFIG_MTD_UBI_WL_THRESHOLD
 
 /*
- * When a physical eraseblock is moved, the WL unit has to pick the target
+ * When a physical eraseblock is moved, the WL sub-system has to pick the target
  * physical eraseblock to move to. The simplest way would be just to pick the
  * one with the highest erase counter. But in certain workloads this could lead
  * to an unlimited wear of one or few physical eraseblock. Indeed, imagine a
  * situation when the picked physical eraseblock is constantly erased after the
  * data is written to it. So, we have a constant which limits the highest erase
- * counter of the free physical eraseblock to pick. Namely, the WL unit does
- * not pick eraseblocks with erase counter greater then the lowest erase
+ * counter of the free physical eraseblock to pick. Namely, the WL sub-system
+ * does not pick eraseblocks with erase counter greater than the lowest erase
  * counter plus %WL_FREE_MAX_DIFF.
  */
 #define WL_FREE_MAX_DIFF (2*UBI_WL_THRESHOLD)
  */
 #define WL_MAX_FAILURES 32
 
+static int self_check_ec(struct ubi_device *ubi, int pnum, int ec);
+static int self_check_in_wl_tree(const struct ubi_device *ubi,
+                                struct ubi_wl_entry *e, struct rb_root *root);
+static int self_check_in_pq(const struct ubi_device *ubi,
+                           struct ubi_wl_entry *e);
+
+#ifdef CONFIG_MTD_UBI_FASTMAP
+#ifndef __UBOOT__
 /**
- * struct ubi_wl_prot_entry - PEB protection entry.
- * @rb_pnum: link in the @wl->prot.pnum RB-tree
- * @rb_aec: link in the @wl->prot.aec RB-tree
- * @abs_ec: the absolute erase counter value when the protection ends
- * @e: the wear-leveling entry of the physical eraseblock under protection
- *
- * When the WL unit returns a physical eraseblock, the physical eraseblock is
- * protected from being moved for some "time". For this reason, the physical
- * eraseblock is not directly moved from the @wl->free tree to the @wl->used
- * tree. There is one more tree in between where this physical eraseblock is
- * temporarily stored (@wl->prot).
- *
- * All this protection stuff is needed because:
- *  o we don't want to move physical eraseblocks just after we have given them
- *    to the user; instead, we first want to let users fill them up with data;
- *
- *  o there is a chance that the user will put the physical eraseblock very
- *    soon, so it makes sense not to move it for some time, but wait; this is
- *    especially important in case of "short term" physical eraseblocks.
- *
- * Physical eraseblocks stay protected only for limited time. But the "time" is
- * measured in erase cycles in this case. This is implemented with help of the
- * absolute erase counter (@wl->abs_ec). When it reaches certain value, the
- * physical eraseblocks are moved from the protection trees (@wl->prot.*) to
- * the @wl->used tree.
- *
- * Protected physical eraseblocks are searched by physical eraseblock number
- * (when they are put) and by the absolute erase counter (to check if it is
- * time to move them to the @wl->used tree). So there are actually 2 RB-trees
- * storing the protected physical eraseblocks: @wl->prot.pnum and
- * @wl->prot.aec. They are referred to as the "protection" trees. The
- * first one is indexed by the physical eraseblock number. The second one is
- * indexed by the absolute erase counter. Both trees store
- * &struct ubi_wl_prot_entry objects.
- *
- * Each physical eraseblock has 2 main states: free and used. The former state
- * corresponds to the @wl->free tree. The latter state is split up on several
- * sub-states:
- * o the WL movement is allowed (@wl->used tree);
- * o the WL movement is temporarily prohibited (@wl->prot.pnum and
- * @wl->prot.aec trees);
- * o scrubbing is needed (@wl->scrub tree).
- *
- * Depending on the sub-state, wear-leveling entries of the used physical
- * eraseblocks may be kept in one of those trees.
+ * update_fastmap_work_fn - calls ubi_update_fastmap from a work queue
+ * @wrk: the work description object
  */
-struct ubi_wl_prot_entry {
-       struct rb_node rb_pnum;
-       struct rb_node rb_aec;
-       unsigned long long abs_ec;
-       struct ubi_wl_entry *e;
-};
+static void update_fastmap_work_fn(struct work_struct *wrk)
+{
+       struct ubi_device *ubi = container_of(wrk, struct ubi_device, fm_work);
+       ubi_update_fastmap(ubi);
+}
+#endif
 
 /**
- * struct ubi_work - UBI work description data structure.
- * @list: a link in the list of pending works
- * @func: worker function
- * @priv: private data of the worker function
- *
- * @e: physical eraseblock to erase
- * @torture: if the physical eraseblock has to be tortured
- *
- * The @func pointer points to the worker function. If the @cancel argument is
- * not zero, the worker has to free the resources and exit immediately. The
- * worker has to return zero in case of success and a negative error code in
- * case of failure.
+ *  ubi_ubi_is_fm_block - returns 1 if a PEB is currently used in a fastmap.
+ *  @ubi: UBI device description object
+ *  @pnum: the to be checked PEB
  */
-struct ubi_work {
-       struct list_head list;
-       int (*func)(struct ubi_device *ubi, struct ubi_work *wrk, int cancel);
-       /* The below fields are only relevant to erasure works */
-       struct ubi_wl_entry *e;
-       int torture;
-};
+static int ubi_is_fm_block(struct ubi_device *ubi, int pnum)
+{
+       int i;
+
+       if (!ubi->fm)
+               return 0;
+
+       for (i = 0; i < ubi->fm->used_blocks; i++)
+               if (ubi->fm->e[i]->pnum == pnum)
+                       return 1;
 
-#ifdef CONFIG_MTD_UBI_DEBUG_PARANOID
-static int paranoid_check_ec(struct ubi_device *ubi, int pnum, int ec);
-static int paranoid_check_in_wl_tree(struct ubi_wl_entry *e,
-                                    struct rb_root *root);
+       return 0;
+}
 #else
-#define paranoid_check_ec(ubi, pnum, ec) 0
-#define paranoid_check_in_wl_tree(e, root)
+static int ubi_is_fm_block(struct ubi_device *ubi, int pnum)
+{
+       return 0;
+}
 #endif
 
 /**
@@ -210,7 +189,7 @@ static void wl_tree_add(struct ubi_wl_entry *e, struct rb_root *root)
                struct ubi_wl_entry *e1;
 
                parent = *p;
-               e1 = rb_entry(parent, struct ubi_wl_entry, rb);
+               e1 = rb_entry(parent, struct ubi_wl_entry, u.rb);
 
                if (e->ec < e1->ec)
                        p = &(*p)->rb_left;
@@ -225,8 +204,8 @@ static void wl_tree_add(struct ubi_wl_entry *e, struct rb_root *root)
                }
        }
 
-       rb_link_node(&e->rb, parent, p);
-       rb_insert_color(&e->rb, root);
+       rb_link_node(&e->u.rb, parent, p);
+       rb_insert_color(&e->u.rb, root);
 }
 
 /**
@@ -289,18 +268,16 @@ static int produce_free_peb(struct ubi_device *ubi)
 {
        int err;
 
-       spin_lock(&ubi->wl_lock);
        while (!ubi->free.rb_node) {
                spin_unlock(&ubi->wl_lock);
 
                dbg_wl("do one work synchronously");
                err = do_work(ubi);
-               if (err)
-                       return err;
 
                spin_lock(&ubi->wl_lock);
+               if (err)
+                       return err;
        }
-       spin_unlock(&ubi->wl_lock);
 
        return 0;
 }
@@ -321,7 +298,7 @@ static int in_wl_tree(struct ubi_wl_entry *e, struct rb_root *root)
        while (p) {
                struct ubi_wl_entry *e1;
 
-               e1 = rb_entry(p, struct ubi_wl_entry, rb);
+               e1 = rb_entry(p, struct ubi_wl_entry, u.rb);
 
                if (e->pnum == e1->pnum) {
                        ubi_assert(e == e1);
@@ -345,223 +322,401 @@ static int in_wl_tree(struct ubi_wl_entry *e, struct rb_root *root)
 }
 
 /**
- * prot_tree_add - add physical eraseblock to protection trees.
+ * prot_queue_add - add physical eraseblock to the protection queue.
  * @ubi: UBI device description object
  * @e: the physical eraseblock to add
- * @pe: protection entry object to use
- * @abs_ec: absolute erase counter value when this physical eraseblock has
- * to be removed from the protection trees.
  *
- * @wl->lock has to be locked.
+ * This function adds @e to the tail of the protection queue @ubi->pq, where
+ * @e will stay for %UBI_PROT_QUEUE_LEN erase operations and will be
+ * temporarily protected from the wear-leveling worker. Note, @wl->lock has to
+ * be locked.
  */
-static void prot_tree_add(struct ubi_device *ubi, struct ubi_wl_entry *e,
-                         struct ubi_wl_prot_entry *pe, int abs_ec)
+static void prot_queue_add(struct ubi_device *ubi, struct ubi_wl_entry *e)
 {
-       struct rb_node **p, *parent = NULL;
-       struct ubi_wl_prot_entry *pe1;
+       int pq_tail = ubi->pq_head - 1;
 
-       pe->e = e;
-       pe->abs_ec = ubi->abs_ec + abs_ec;
-
-       p = &ubi->prot.pnum.rb_node;
-       while (*p) {
-               parent = *p;
-               pe1 = rb_entry(parent, struct ubi_wl_prot_entry, rb_pnum);
-
-               if (e->pnum < pe1->e->pnum)
-                       p = &(*p)->rb_left;
-               else
-                       p = &(*p)->rb_right;
-       }
-       rb_link_node(&pe->rb_pnum, parent, p);
-       rb_insert_color(&pe->rb_pnum, &ubi->prot.pnum);
-
-       p = &ubi->prot.aec.rb_node;
-       parent = NULL;
-       while (*p) {
-               parent = *p;
-               pe1 = rb_entry(parent, struct ubi_wl_prot_entry, rb_aec);
-
-               if (pe->abs_ec < pe1->abs_ec)
-                       p = &(*p)->rb_left;
-               else
-                       p = &(*p)->rb_right;
-       }
-       rb_link_node(&pe->rb_aec, parent, p);
-       rb_insert_color(&pe->rb_aec, &ubi->prot.aec);
+       if (pq_tail < 0)
+               pq_tail = UBI_PROT_QUEUE_LEN - 1;
+       ubi_assert(pq_tail >= 0 && pq_tail < UBI_PROT_QUEUE_LEN);
+       list_add_tail(&e->u.list, &ubi->pq[pq_tail]);
+       dbg_wl("added PEB %d EC %d to the protection queue", e->pnum, e->ec);
 }
 
 /**
  * find_wl_entry - find wear-leveling entry closest to certain erase counter.
+ * @ubi: UBI device description object
  * @root: the RB-tree where to look for
- * @max: highest possible erase counter
+ * @diff: maximum possible difference from the smallest erase counter
  *
  * This function looks for a wear leveling entry with erase counter closest to
- * @max and less then @max.
+ * min + @diff, where min is the smallest erase counter.
  */
-static struct ubi_wl_entry *find_wl_entry(struct rb_root *root, int max)
+static struct ubi_wl_entry *find_wl_entry(struct ubi_device *ubi,
+                                         struct rb_root *root, int diff)
 {
        struct rb_node *p;
-       struct ubi_wl_entry *e;
+       struct ubi_wl_entry *e, *prev_e = NULL;
+       int max;
 
-       e = rb_entry(rb_first(root), struct ubi_wl_entry, rb);
-       max += e->ec;
+       e = rb_entry(rb_first(root), struct ubi_wl_entry, u.rb);
+       max = e->ec + diff;
 
        p = root->rb_node;
        while (p) {
                struct ubi_wl_entry *e1;
 
-               e1 = rb_entry(p, struct ubi_wl_entry, rb);
+               e1 = rb_entry(p, struct ubi_wl_entry, u.rb);
                if (e1->ec >= max)
                        p = p->rb_left;
                else {
                        p = p->rb_right;
+                       prev_e = e;
                        e = e1;
                }
        }
 
+       /* If no fastmap has been written and this WL entry can be used
+        * as anchor PEB, hold it back and return the second best WL entry
+        * such that fastmap can use the anchor PEB later. */
+       if (prev_e && !ubi->fm_disabled &&
+           !ubi->fm && e->pnum < UBI_FM_MAX_START)
+               return prev_e;
+
        return e;
 }
 
 /**
- * ubi_wl_get_peb - get a physical eraseblock.
+ * find_mean_wl_entry - find wear-leveling entry with medium erase counter.
  * @ubi: UBI device description object
- * @dtype: type of data which will be stored in this physical eraseblock
+ * @root: the RB-tree where to look for
  *
- * This function returns a physical eraseblock in case of success and a
- * negative error code in case of failure. Might sleep.
+ * This function looks for a wear leveling entry with medium erase counter,
+ * but not greater or equivalent than the lowest erase counter plus
+ * %WL_FREE_MAX_DIFF/2.
  */
-int ubi_wl_get_peb(struct ubi_device *ubi, int dtype)
+static struct ubi_wl_entry *find_mean_wl_entry(struct ubi_device *ubi,
+                                              struct rb_root *root)
 {
-       int err, protect, medium_ec;
        struct ubi_wl_entry *e, *first, *last;
-       struct ubi_wl_prot_entry *pe;
 
-       ubi_assert(dtype == UBI_LONGTERM || dtype == UBI_SHORTTERM ||
-                  dtype == UBI_UNKNOWN);
+       first = rb_entry(rb_first(root), struct ubi_wl_entry, u.rb);
+       last = rb_entry(rb_last(root), struct ubi_wl_entry, u.rb);
 
-       pe = kmalloc(sizeof(struct ubi_wl_prot_entry), GFP_NOFS);
-       if (!pe)
-               return -ENOMEM;
+       if (last->ec - first->ec < WL_FREE_MAX_DIFF) {
+               e = rb_entry(root->rb_node, struct ubi_wl_entry, u.rb);
+
+#ifdef CONFIG_MTD_UBI_FASTMAP
+               /* If no fastmap has been written and this WL entry can be used
+                * as anchor PEB, hold it back and return the second best
+                * WL entry such that fastmap can use the anchor PEB later. */
+               if (e && !ubi->fm_disabled && !ubi->fm &&
+                   e->pnum < UBI_FM_MAX_START)
+                       e = rb_entry(rb_next(root->rb_node),
+                                    struct ubi_wl_entry, u.rb);
+#endif
+       } else
+               e = find_wl_entry(ubi, root, WL_FREE_MAX_DIFF/2);
+
+       return e;
+}
+
+#ifdef CONFIG_MTD_UBI_FASTMAP
+/**
+ * find_anchor_wl_entry - find wear-leveling entry to used as anchor PEB.
+ * @root: the RB-tree where to look for
+ */
+static struct ubi_wl_entry *find_anchor_wl_entry(struct rb_root *root)
+{
+       struct rb_node *p;
+       struct ubi_wl_entry *e, *victim = NULL;
+       int max_ec = UBI_MAX_ERASECOUNTER;
+
+       ubi_rb_for_each_entry(p, e, root, u.rb) {
+               if (e->pnum < UBI_FM_MAX_START && e->ec < max_ec) {
+                       victim = e;
+                       max_ec = e->ec;
+               }
+       }
+
+       return victim;
+}
+
+static int anchor_pebs_avalible(struct rb_root *root)
+{
+       struct rb_node *p;
+       struct ubi_wl_entry *e;
+
+       ubi_rb_for_each_entry(p, e, root, u.rb)
+               if (e->pnum < UBI_FM_MAX_START)
+                       return 1;
+
+       return 0;
+}
+
+/**
+ * ubi_wl_get_fm_peb - find a physical erase block with a given maximal number.
+ * @ubi: UBI device description object
+ * @anchor: This PEB will be used as anchor PEB by fastmap
+ *
+ * The function returns a physical erase block with a given maximal number
+ * and removes it from the wl subsystem.
+ * Must be called with wl_lock held!
+ */
+struct ubi_wl_entry *ubi_wl_get_fm_peb(struct ubi_device *ubi, int anchor)
+{
+       struct ubi_wl_entry *e = NULL;
+
+       if (!ubi->free.rb_node || (ubi->free_count - ubi->beb_rsvd_pebs < 1))
+               goto out;
+
+       if (anchor)
+               e = find_anchor_wl_entry(&ubi->free);
+       else
+               e = find_mean_wl_entry(ubi, &ubi->free);
+
+       if (!e)
+               goto out;
+
+       self_check_in_wl_tree(ubi, e, &ubi->free);
+
+       /* remove it from the free list,
+        * the wl subsystem does no longer know this erase block */
+       rb_erase(&e->u.rb, &ubi->free);
+       ubi->free_count--;
+out:
+       return e;
+}
+#endif
+
+/**
+ * __wl_get_peb - get a physical eraseblock.
+ * @ubi: UBI device description object
+ *
+ * This function returns a physical eraseblock in case of success and a
+ * negative error code in case of failure.
+ */
+static int __wl_get_peb(struct ubi_device *ubi)
+{
+       int err;
+       struct ubi_wl_entry *e;
 
 retry:
-       spin_lock(&ubi->wl_lock);
        if (!ubi->free.rb_node) {
                if (ubi->works_count == 0) {
-                       ubi_assert(list_empty(&ubi->works));
                        ubi_err("no free eraseblocks");
-                       spin_unlock(&ubi->wl_lock);
-                       kfree(pe);
+                       ubi_assert(list_empty(&ubi->works));
                        return -ENOSPC;
                }
-               spin_unlock(&ubi->wl_lock);
 
                err = produce_free_peb(ubi);
-               if (err < 0) {
-                       kfree(pe);
+               if (err < 0)
                        return err;
-               }
                goto retry;
        }
 
-       switch (dtype) {
-               case UBI_LONGTERM:
-                       /*
-                        * For long term data we pick a physical eraseblock
-                        * with high erase counter. But the highest erase
-                        * counter we can pick is bounded by the the lowest
-                        * erase counter plus %WL_FREE_MAX_DIFF.
-                        */
-                       e = find_wl_entry(&ubi->free, WL_FREE_MAX_DIFF);
-                       protect = LT_PROTECTION;
-                       break;
-               case UBI_UNKNOWN:
-                       /*
-                        * For unknown data we pick a physical eraseblock with
-                        * medium erase counter. But we by no means can pick a
-                        * physical eraseblock with erase counter greater or
-                        * equivalent than the lowest erase counter plus
-                        * %WL_FREE_MAX_DIFF.
-                        */
-                       first = rb_entry(rb_first(&ubi->free),
-                                        struct ubi_wl_entry, rb);
-                       last = rb_entry(rb_last(&ubi->free),
-                                       struct ubi_wl_entry, rb);
-
-                       if (last->ec - first->ec < WL_FREE_MAX_DIFF)
-                               e = rb_entry(ubi->free.rb_node,
-                                               struct ubi_wl_entry, rb);
-                       else {
-                               medium_ec = (first->ec + WL_FREE_MAX_DIFF)/2;
-                               e = find_wl_entry(&ubi->free, medium_ec);
-                       }
-                       protect = U_PROTECTION;
-                       break;
-               case UBI_SHORTTERM:
-                       /*
-                        * For short term data we pick a physical eraseblock
-                        * with the lowest erase counter as we expect it will
-                        * be erased soon.
-                        */
-                       e = rb_entry(rb_first(&ubi->free),
-                                    struct ubi_wl_entry, rb);
-                       protect = ST_PROTECTION;
-                       break;
-               default:
-                       protect = 0;
-                       e = NULL;
-                       BUG();
+       e = find_mean_wl_entry(ubi, &ubi->free);
+       if (!e) {
+               ubi_err("no free eraseblocks");
+               return -ENOSPC;
        }
 
+       self_check_in_wl_tree(ubi, e, &ubi->free);
+
        /*
-        * Move the physical eraseblock to the protection trees where it will
+        * Move the physical eraseblock to the protection queue where it will
         * be protected from being moved for some time.
         */
-       paranoid_check_in_wl_tree(e, &ubi->free);
-       rb_erase(&e->rb, &ubi->free);
-       prot_tree_add(ubi, e, pe, protect);
+       rb_erase(&e->u.rb, &ubi->free);
+       ubi->free_count--;
+       dbg_wl("PEB %d EC %d", e->pnum, e->ec);
+#ifndef CONFIG_MTD_UBI_FASTMAP
+       /* We have to enqueue e only if fastmap is disabled,
+        * is fastmap enabled prot_queue_add() will be called by
+        * ubi_wl_get_peb() after removing e from the pool. */
+       prot_queue_add(ubi, e);
+#endif
+       return e->pnum;
+}
 
-       dbg_wl("PEB %d EC %d, protection %d", e->pnum, e->ec, protect);
-       spin_unlock(&ubi->wl_lock);
+#ifdef CONFIG_MTD_UBI_FASTMAP
+/**
+ * return_unused_pool_pebs - returns unused PEB to the free tree.
+ * @ubi: UBI device description object
+ * @pool: fastmap pool description object
+ */
+static void return_unused_pool_pebs(struct ubi_device *ubi,
+                                   struct ubi_fm_pool *pool)
+{
+       int i;
+       struct ubi_wl_entry *e;
 
-       return e->pnum;
+       for (i = pool->used; i < pool->size; i++) {
+               e = ubi->lookuptbl[pool->pebs[i]];
+               wl_tree_add(e, &ubi->free);
+               ubi->free_count++;
+       }
 }
 
 /**
- * prot_tree_del - remove a physical eraseblock from the protection trees
+ * refill_wl_pool - refills all the fastmap pool used by the
+ * WL sub-system.
  * @ubi: UBI device description object
- * @pnum: the physical eraseblock to remove
+ */
+static void refill_wl_pool(struct ubi_device *ubi)
+{
+       struct ubi_wl_entry *e;
+       struct ubi_fm_pool *pool = &ubi->fm_wl_pool;
+
+       return_unused_pool_pebs(ubi, pool);
+
+       for (pool->size = 0; pool->size < pool->max_size; pool->size++) {
+               if (!ubi->free.rb_node ||
+                  (ubi->free_count - ubi->beb_rsvd_pebs < 5))
+                       break;
+
+               e = find_wl_entry(ubi, &ubi->free, WL_FREE_MAX_DIFF);
+               self_check_in_wl_tree(ubi, e, &ubi->free);
+               rb_erase(&e->u.rb, &ubi->free);
+               ubi->free_count--;
+
+               pool->pebs[pool->size] = e->pnum;
+       }
+       pool->used = 0;
+}
+
+/**
+ * refill_wl_user_pool - refills all the fastmap pool used by ubi_wl_get_peb.
+ * @ubi: UBI device description object
+ */
+static void refill_wl_user_pool(struct ubi_device *ubi)
+{
+       struct ubi_fm_pool *pool = &ubi->fm_pool;
+
+       return_unused_pool_pebs(ubi, pool);
+
+       for (pool->size = 0; pool->size < pool->max_size; pool->size++) {
+               pool->pebs[pool->size] = __wl_get_peb(ubi);
+               if (pool->pebs[pool->size] < 0)
+                       break;
+       }
+       pool->used = 0;
+}
+
+/**
+ * ubi_refill_pools - refills all fastmap PEB pools.
+ * @ubi: UBI device description object
+ */
+void ubi_refill_pools(struct ubi_device *ubi)
+{
+       spin_lock(&ubi->wl_lock);
+       refill_wl_pool(ubi);
+       refill_wl_user_pool(ubi);
+       spin_unlock(&ubi->wl_lock);
+}
+
+/* ubi_wl_get_peb - works exaclty like __wl_get_peb but keeps track of
+ * the fastmap pool.
+ */
+int ubi_wl_get_peb(struct ubi_device *ubi)
+{
+       int ret;
+       struct ubi_fm_pool *pool = &ubi->fm_pool;
+       struct ubi_fm_pool *wl_pool = &ubi->fm_wl_pool;
+
+       if (!pool->size || !wl_pool->size || pool->used == pool->size ||
+           wl_pool->used == wl_pool->size)
+               ubi_update_fastmap(ubi);
+
+       /* we got not a single free PEB */
+       if (!pool->size)
+               ret = -ENOSPC;
+       else {
+               spin_lock(&ubi->wl_lock);
+               ret = pool->pebs[pool->used++];
+               prot_queue_add(ubi, ubi->lookuptbl[ret]);
+               spin_unlock(&ubi->wl_lock);
+       }
+
+       return ret;
+}
+
+/* get_peb_for_wl - returns a PEB to be used internally by the WL sub-system.
  *
- * This function returns PEB @pnum from the protection trees and returns zero
- * in case of success and %-ENODEV if the PEB was not found in the protection
- * trees.
+ * @ubi: UBI device description object
  */
-static int prot_tree_del(struct ubi_device *ubi, int pnum)
+static struct ubi_wl_entry *get_peb_for_wl(struct ubi_device *ubi)
 {
-       struct rb_node *p;
-       struct ubi_wl_prot_entry *pe = NULL;
+       struct ubi_fm_pool *pool = &ubi->fm_wl_pool;
+       int pnum;
+
+       if (pool->used == pool->size || !pool->size) {
+               /* We cannot update the fastmap here because this
+                * function is called in atomic context.
+                * Let's fail here and refill/update it as soon as possible. */
+#ifndef __UBOOT__
+               schedule_work(&ubi->fm_work);
+#else
+               /* In U-Boot we must call this directly */
+               ubi_update_fastmap(ubi);
+#endif
+               return NULL;
+       } else {
+               pnum = pool->pebs[pool->used++];
+               return ubi->lookuptbl[pnum];
+       }
+}
+#else
+static struct ubi_wl_entry *get_peb_for_wl(struct ubi_device *ubi)
+{
+       struct ubi_wl_entry *e;
 
-       p = ubi->prot.pnum.rb_node;
-       while (p) {
+       e = find_wl_entry(ubi, &ubi->free, WL_FREE_MAX_DIFF);
+       self_check_in_wl_tree(ubi, e, &ubi->free);
+       rb_erase(&e->u.rb, &ubi->free);
 
-               pe = rb_entry(p, struct ubi_wl_prot_entry, rb_pnum);
+       return e;
+}
 
-               if (pnum == pe->e->pnum)
-                       goto found;
+int ubi_wl_get_peb(struct ubi_device *ubi)
+{
+       int peb, err;
 
-               if (pnum < pe->e->pnum)
-                       p = p->rb_left;
-               else
-                       p = p->rb_right;
+       spin_lock(&ubi->wl_lock);
+       peb = __wl_get_peb(ubi);
+       spin_unlock(&ubi->wl_lock);
+
+       err = ubi_self_check_all_ff(ubi, peb, ubi->vid_hdr_aloffset,
+                                   ubi->peb_size - ubi->vid_hdr_aloffset);
+       if (err) {
+               ubi_err("new PEB %d does not contain all 0xFF bytes", peb);
+               return err;
        }
 
-       return -ENODEV;
+       return peb;
+}
+#endif
+
+/**
+ * prot_queue_del - remove a physical eraseblock from the protection queue.
+ * @ubi: UBI device description object
+ * @pnum: the physical eraseblock to remove
+ *
+ * This function deletes PEB @pnum from the protection queue and returns zero
+ * in case of success and %-ENODEV if the PEB was not found.
+ */
+static int prot_queue_del(struct ubi_device *ubi, int pnum)
+{
+       struct ubi_wl_entry *e;
 
-found:
-       ubi_assert(pe->e->pnum == pnum);
-       rb_erase(&pe->rb_aec, &ubi->prot.aec);
-       rb_erase(&pe->rb_pnum, &ubi->prot.pnum);
-       kfree(pe);
+       e = ubi->lookuptbl[pnum];
+       if (!e)
+               return -ENODEV;
+
+       if (self_check_in_pq(ubi, e))
+               return -ENODEV;
+
+       list_del(&e->u.list);
+       dbg_wl("deleted PEB %d from the protection queue", e->pnum);
        return 0;
 }
 
@@ -574,7 +729,8 @@ found:
  * This function returns zero in case of success and a negative error code in
  * case of failure.
  */
-static int sync_erase(struct ubi_device *ubi, struct ubi_wl_entry *e, int torture)
+static int sync_erase(struct ubi_device *ubi, struct ubi_wl_entry *e,
+                     int torture)
 {
        int err;
        struct ubi_ec_hdr *ec_hdr;
@@ -582,8 +738,8 @@ static int sync_erase(struct ubi_device *ubi, struct ubi_wl_entry *e, int tortur
 
        dbg_wl("erase PEB %d, old EC %llu", e->pnum, ec);
 
-       err = paranoid_check_ec(ubi, e->pnum, e->ec);
-       if (err > 0)
+       err = self_check_ec(ubi, e->pnum, e->ec);
+       if (err)
                return -EINVAL;
 
        ec_hdr = kzalloc(ubi->ec_hdr_alsize, GFP_NOFS);
@@ -626,105 +782,213 @@ out_free:
 }
 
 /**
- * check_protection_over - check if it is time to stop protecting some
- * physical eraseblocks.
+ * serve_prot_queue - check if it is time to stop protecting PEBs.
  * @ubi: UBI device description object
  *
- * This function is called after each erase operation, when the absolute erase
- * counter is incremented, to check if some physical eraseblock  have not to be
- * protected any longer. These physical eraseblocks are moved from the
- * protection trees to the used tree.
+ * This function is called after each erase operation and removes PEBs from the
+ * tail of the protection queue. These PEBs have been protected for long enough
+ * and should be moved to the used tree.
  */
-static void check_protection_over(struct ubi_device *ubi)
+static void serve_prot_queue(struct ubi_device *ubi)
 {
-       struct ubi_wl_prot_entry *pe;
+       struct ubi_wl_entry *e, *tmp;
+       int count;
 
        /*
         * There may be several protected physical eraseblock to remove,
         * process them all.
         */
-       while (1) {
-               spin_lock(&ubi->wl_lock);
-               if (!ubi->prot.aec.rb_node) {
-                       spin_unlock(&ubi->wl_lock);
-                       break;
-               }
-
-               pe = rb_entry(rb_first(&ubi->prot.aec),
-                             struct ubi_wl_prot_entry, rb_aec);
+repeat:
+       count = 0;
+       spin_lock(&ubi->wl_lock);
+       list_for_each_entry_safe(e, tmp, &ubi->pq[ubi->pq_head], u.list) {
+               dbg_wl("PEB %d EC %d protection over, move to used tree",
+                       e->pnum, e->ec);
 
-               if (pe->abs_ec > ubi->abs_ec) {
+               list_del(&e->u.list);
+               wl_tree_add(e, &ubi->used);
+               if (count++ > 32) {
+                       /*
+                        * Let's be nice and avoid holding the spinlock for
+                        * too long.
+                        */
                        spin_unlock(&ubi->wl_lock);
-                       break;
+                       cond_resched();
+                       goto repeat;
                }
-
-               dbg_wl("PEB %d protection over, abs_ec %llu, PEB abs_ec %llu",
-                      pe->e->pnum, ubi->abs_ec, pe->abs_ec);
-               rb_erase(&pe->rb_aec, &ubi->prot.aec);
-               rb_erase(&pe->rb_pnum, &ubi->prot.pnum);
-               wl_tree_add(pe->e, &ubi->used);
-               spin_unlock(&ubi->wl_lock);
-
-               kfree(pe);
-               cond_resched();
        }
+
+       ubi->pq_head += 1;
+       if (ubi->pq_head == UBI_PROT_QUEUE_LEN)
+               ubi->pq_head = 0;
+       ubi_assert(ubi->pq_head >= 0 && ubi->pq_head < UBI_PROT_QUEUE_LEN);
+       spin_unlock(&ubi->wl_lock);
 }
 
 /**
- * schedule_ubi_work - schedule a work.
+ * __schedule_ubi_work - schedule a work.
  * @ubi: UBI device description object
  * @wrk: the work to schedule
  *
- * This function enqueues a work defined by @wrk to the tail of the pending
- * works list.
+ * This function adds a work defined by @wrk to the tail of the pending works
+ * list. Can only be used of ubi->work_sem is already held in read mode!
  */
-static void schedule_ubi_work(struct ubi_device *ubi, struct ubi_work *wrk)
+static void __schedule_ubi_work(struct ubi_device *ubi, struct ubi_work *wrk)
 {
        spin_lock(&ubi->wl_lock);
        list_add_tail(&wrk->list, &ubi->works);
        ubi_assert(ubi->works_count >= 0);
        ubi->works_count += 1;
-
+#ifndef __UBOOT__
+       if (ubi->thread_enabled && !ubi_dbg_is_bgt_disabled(ubi))
+               wake_up_process(ubi->bgt_thread);
+#else
        /*
         * U-Boot special: We have no bgt_thread in U-Boot!
         * So just call do_work() here directly.
         */
        do_work(ubi);
-
+#endif
        spin_unlock(&ubi->wl_lock);
 }
 
-static int erase_worker(struct ubi_device *ubi, struct ubi_work *wl_wrk,
-                       int cancel);
-
 /**
- * schedule_erase - schedule an erase work.
+ * schedule_ubi_work - schedule a work.
+ * @ubi: UBI device description object
+ * @wrk: the work to schedule
+ *
+ * This function adds a work defined by @wrk to the tail of the pending works
+ * list.
+ */
+static void schedule_ubi_work(struct ubi_device *ubi, struct ubi_work *wrk)
+{
+       down_read(&ubi->work_sem);
+       __schedule_ubi_work(ubi, wrk);
+       up_read(&ubi->work_sem);
+}
+
+static int erase_worker(struct ubi_device *ubi, struct ubi_work *wl_wrk,
+                       int cancel);
+
+#ifdef CONFIG_MTD_UBI_FASTMAP
+/**
+ * ubi_is_erase_work - checks whether a work is erase work.
+ * @wrk: The work object to be checked
+ */
+int ubi_is_erase_work(struct ubi_work *wrk)
+{
+       return wrk->func == erase_worker;
+}
+#endif
+
+/**
+ * schedule_erase - schedule an erase work.
+ * @ubi: UBI device description object
+ * @e: the WL entry of the physical eraseblock to erase
+ * @vol_id: the volume ID that last used this PEB
+ * @lnum: the last used logical eraseblock number for the PEB
+ * @torture: if the physical eraseblock has to be tortured
+ *
+ * This function returns zero in case of success and a %-ENOMEM in case of
+ * failure.
+ */
+static int schedule_erase(struct ubi_device *ubi, struct ubi_wl_entry *e,
+                         int vol_id, int lnum, int torture)
+{
+       struct ubi_work *wl_wrk;
+
+       ubi_assert(e);
+       ubi_assert(!ubi_is_fm_block(ubi, e->pnum));
+
+       dbg_wl("schedule erasure of PEB %d, EC %d, torture %d",
+              e->pnum, e->ec, torture);
+
+       wl_wrk = kmalloc(sizeof(struct ubi_work), GFP_NOFS);
+       if (!wl_wrk)
+               return -ENOMEM;
+
+       wl_wrk->func = &erase_worker;
+       wl_wrk->e = e;
+       wl_wrk->vol_id = vol_id;
+       wl_wrk->lnum = lnum;
+       wl_wrk->torture = torture;
+
+       schedule_ubi_work(ubi, wl_wrk);
+       return 0;
+}
+
+/**
+ * do_sync_erase - run the erase worker synchronously.
  * @ubi: UBI device description object
  * @e: the WL entry of the physical eraseblock to erase
+ * @vol_id: the volume ID that last used this PEB
+ * @lnum: the last used logical eraseblock number for the PEB
  * @torture: if the physical eraseblock has to be tortured
  *
- * This function returns zero in case of success and a %-ENOMEM in case of
- * failure.
  */
-static int schedule_erase(struct ubi_device *ubi, struct ubi_wl_entry *e,
-                         int torture)
+static int do_sync_erase(struct ubi_device *ubi, struct ubi_wl_entry *e,
+                        int vol_id, int lnum, int torture)
 {
        struct ubi_work *wl_wrk;
 
-       dbg_wl("schedule erasure of PEB %d, EC %d, torture %d",
-              e->pnum, e->ec, torture);
+       dbg_wl("sync erase of PEB %i", e->pnum);
 
        wl_wrk = kmalloc(sizeof(struct ubi_work), GFP_NOFS);
        if (!wl_wrk)
                return -ENOMEM;
 
-       wl_wrk->func = &erase_worker;
        wl_wrk->e = e;
+       wl_wrk->vol_id = vol_id;
+       wl_wrk->lnum = lnum;
        wl_wrk->torture = torture;
 
-       schedule_ubi_work(ubi, wl_wrk);
-       return 0;
+       return erase_worker(ubi, wl_wrk, 0);
+}
+
+#ifdef CONFIG_MTD_UBI_FASTMAP
+/**
+ * ubi_wl_put_fm_peb - returns a PEB used in a fastmap to the wear-leveling
+ * sub-system.
+ * see: ubi_wl_put_peb()
+ *
+ * @ubi: UBI device description object
+ * @fm_e: physical eraseblock to return
+ * @lnum: the last used logical eraseblock number for the PEB
+ * @torture: if this physical eraseblock has to be tortured
+ */
+int ubi_wl_put_fm_peb(struct ubi_device *ubi, struct ubi_wl_entry *fm_e,
+                     int lnum, int torture)
+{
+       struct ubi_wl_entry *e;
+       int vol_id, pnum = fm_e->pnum;
+
+       dbg_wl("PEB %d", pnum);
+
+       ubi_assert(pnum >= 0);
+       ubi_assert(pnum < ubi->peb_count);
+
+       spin_lock(&ubi->wl_lock);
+       e = ubi->lookuptbl[pnum];
+
+       /* This can happen if we recovered from a fastmap the very
+        * first time and writing now a new one. In this case the wl system
+        * has never seen any PEB used by the original fastmap.
+        */
+       if (!e) {
+               e = fm_e;
+               ubi_assert(e->ec >= 0);
+               ubi->lookuptbl[pnum] = e;
+       } else {
+               e->ec = fm_e->ec;
+               kfree(fm_e);
+       }
+
+       spin_unlock(&ubi->wl_lock);
+
+       vol_id = lnum ? UBI_FM_DATA_VOLUME_ID : UBI_FM_SB_VOLUME_ID;
+       return schedule_erase(ubi, e, vol_id, lnum, torture);
 }
+#endif
 
 /**
  * wear_leveling_worker - wear-leveling worker function.
@@ -739,13 +1003,15 @@ static int schedule_erase(struct ubi_device *ubi, struct ubi_wl_entry *e,
 static int wear_leveling_worker(struct ubi_device *ubi, struct ubi_work *wrk,
                                int cancel)
 {
-       int err, put = 0, scrubbing = 0, protect = 0;
-       struct ubi_wl_prot_entry *uninitialized_var(pe);
+       int err, scrubbing = 0, torture = 0, protect = 0, erroneous = 0;
+       int vol_id = -1, uninitialized_var(lnum);
+#ifdef CONFIG_MTD_UBI_FASTMAP
+       int anchor = wrk->anchor;
+#endif
        struct ubi_wl_entry *e1, *e2;
        struct ubi_vid_hdr *vid_hdr;
 
        kfree(wrk);
-
        if (cancel)
                return 0;
 
@@ -775,36 +1041,61 @@ static int wear_leveling_worker(struct ubi_device *ubi, struct ubi_work *wrk,
                goto out_cancel;
        }
 
+#ifdef CONFIG_MTD_UBI_FASTMAP
+       /* Check whether we need to produce an anchor PEB */
+       if (!anchor)
+               anchor = !anchor_pebs_avalible(&ubi->free);
+
+       if (anchor) {
+               e1 = find_anchor_wl_entry(&ubi->used);
+               if (!e1)
+                       goto out_cancel;
+               e2 = get_peb_for_wl(ubi);
+               if (!e2)
+                       goto out_cancel;
+
+               self_check_in_wl_tree(ubi, e1, &ubi->used);
+               rb_erase(&e1->u.rb, &ubi->used);
+               dbg_wl("anchor-move PEB %d to PEB %d", e1->pnum, e2->pnum);
+       } else if (!ubi->scrub.rb_node) {
+#else
        if (!ubi->scrub.rb_node) {
+#endif
                /*
                 * Now pick the least worn-out used physical eraseblock and a
                 * highly worn-out free physical eraseblock. If the erase
                 * counters differ much enough, start wear-leveling.
                 */
-               e1 = rb_entry(rb_first(&ubi->used), struct ubi_wl_entry, rb);
-               e2 = find_wl_entry(&ubi->free, WL_FREE_MAX_DIFF);
+               e1 = rb_entry(rb_first(&ubi->used), struct ubi_wl_entry, u.rb);
+               e2 = get_peb_for_wl(ubi);
+               if (!e2)
+                       goto out_cancel;
 
                if (!(e2->ec - e1->ec >= UBI_WL_THRESHOLD)) {
                        dbg_wl("no WL needed: min used EC %d, max free EC %d",
                               e1->ec, e2->ec);
+
+                       /* Give the unused PEB back */
+                       wl_tree_add(e2, &ubi->free);
                        goto out_cancel;
                }
-               paranoid_check_in_wl_tree(e1, &ubi->used);
-               rb_erase(&e1->rb, &ubi->used);
+               self_check_in_wl_tree(ubi, e1, &ubi->used);
+               rb_erase(&e1->u.rb, &ubi->used);
                dbg_wl("move PEB %d EC %d to PEB %d EC %d",
                       e1->pnum, e1->ec, e2->pnum, e2->ec);
        } else {
                /* Perform scrubbing */
                scrubbing = 1;
-               e1 = rb_entry(rb_first(&ubi->scrub), struct ubi_wl_entry, rb);
-               e2 = find_wl_entry(&ubi->free, WL_FREE_MAX_DIFF);
-               paranoid_check_in_wl_tree(e1, &ubi->scrub);
-               rb_erase(&e1->rb, &ubi->scrub);
+               e1 = rb_entry(rb_first(&ubi->scrub), struct ubi_wl_entry, u.rb);
+               e2 = get_peb_for_wl(ubi);
+               if (!e2)
+                       goto out_cancel;
+
+               self_check_in_wl_tree(ubi, e1, &ubi->scrub);
+               rb_erase(&e1->u.rb, &ubi->scrub);
                dbg_wl("scrub PEB %d to PEB %d", e1->pnum, e2->pnum);
        }
 
-       paranoid_check_in_wl_tree(e2, &ubi->free);
-       rb_erase(&e2->rb, &ubi->free);
        ubi->move_from = e1;
        ubi->move_to = e2;
        spin_unlock(&ubi->wl_lock);
@@ -822,81 +1113,127 @@ static int wear_leveling_worker(struct ubi_device *ubi, struct ubi_work *wrk,
 
        err = ubi_io_read_vid_hdr(ubi, e1->pnum, vid_hdr, 0);
        if (err && err != UBI_IO_BITFLIPS) {
-               if (err == UBI_IO_PEB_FREE) {
+               if (err == UBI_IO_FF) {
                        /*
                         * We are trying to move PEB without a VID header. UBI
                         * always write VID headers shortly after the PEB was
-                        * given, so we have a situation when it did not have
-                        * chance to write it down because it was preempted.
-                        * Just re-schedule the work, so that next time it will
-                        * likely have the VID header in place.
+                        * given, so we have a situation when it has not yet
+                        * had a chance to write it, because it was preempted.
+                        * So add this PEB to the protection queue so far,
+                        * because presumably more data will be written there
+                        * (including the missing VID header), and then we'll
+                        * move it.
                         */
                        dbg_wl("PEB %d has no VID header", e1->pnum);
+                       protect = 1;
+                       goto out_not_moved;
+               } else if (err == UBI_IO_FF_BITFLIPS) {
+                       /*
+                        * The same situation as %UBI_IO_FF, but bit-flips were
+                        * detected. It is better to schedule this PEB for
+                        * scrubbing.
+                        */
+                       dbg_wl("PEB %d has no VID header but has bit-flips",
+                              e1->pnum);
+                       scrubbing = 1;
                        goto out_not_moved;
                }
 
                ubi_err("error %d while reading VID header from PEB %d",
                        err, e1->pnum);
-               if (err > 0)
-                       err = -EIO;
                goto out_error;
        }
 
+       vol_id = be32_to_cpu(vid_hdr->vol_id);
+       lnum = be32_to_cpu(vid_hdr->lnum);
+
        err = ubi_eba_copy_leb(ubi, e1->pnum, e2->pnum, vid_hdr);
        if (err) {
-
-               if (err < 0)
-                       goto out_error;
-               if (err == 1)
+               if (err == MOVE_CANCEL_RACE) {
+                       /*
+                        * The LEB has not been moved because the volume is
+                        * being deleted or the PEB has been put meanwhile. We
+                        * should prevent this PEB from being selected for
+                        * wear-leveling movement again, so put it to the
+                        * protection queue.
+                        */
+                       protect = 1;
+                       goto out_not_moved;
+               }
+               if (err == MOVE_RETRY) {
+                       scrubbing = 1;
+                       goto out_not_moved;
+               }
+               if (err == MOVE_TARGET_BITFLIPS || err == MOVE_TARGET_WR_ERR ||
+                   err == MOVE_TARGET_RD_ERR) {
+                       /*
+                        * Target PEB had bit-flips or write error - torture it.
+                        */
+                       torture = 1;
                        goto out_not_moved;
+               }
 
-               /*
-                * For some reason the LEB was not moved - it might be because
-                * the volume is being deleted. We should prevent this PEB from
-                * being selected for wear-levelling movement for some "time",
-                * so put it to the protection tree.
-                */
+               if (err == MOVE_SOURCE_RD_ERR) {
+                       /*
+                        * An error happened while reading the source PEB. Do
+                        * not switch to R/O mode in this case, and give the
+                        * upper layers a possibility to recover from this,
+                        * e.g. by unmapping corresponding LEB. Instead, just
+                        * put this PEB to the @ubi->erroneous list to prevent
+                        * UBI from trying to move it over and over again.
+                        */
+                       if (ubi->erroneous_peb_count > ubi->max_erroneous) {
+                               ubi_err("too many erroneous eraseblocks (%d)",
+                                       ubi->erroneous_peb_count);
+                               goto out_error;
+                       }
+                       erroneous = 1;
+                       goto out_not_moved;
+               }
 
-               dbg_wl("cancelled moving PEB %d", e1->pnum);
-               pe = kmalloc(sizeof(struct ubi_wl_prot_entry), GFP_NOFS);
-               if (!pe) {
-                       err = -ENOMEM;
+               if (err < 0)
                        goto out_error;
-               }
 
-               protect = 1;
+               ubi_assert(0);
        }
 
+       /* The PEB has been successfully moved */
+       if (scrubbing)
+               ubi_msg("scrubbed PEB %d (LEB %d:%d), data moved to PEB %d",
+                       e1->pnum, vol_id, lnum, e2->pnum);
        ubi_free_vid_hdr(ubi, vid_hdr);
+
        spin_lock(&ubi->wl_lock);
-       if (protect)
-               prot_tree_add(ubi, e1, pe, protect);
-       if (!ubi->move_to_put)
+       if (!ubi->move_to_put) {
                wl_tree_add(e2, &ubi->used);
-       else
-               put = 1;
+               e2 = NULL;
+       }
        ubi->move_from = ubi->move_to = NULL;
        ubi->move_to_put = ubi->wl_scheduled = 0;
        spin_unlock(&ubi->wl_lock);
 
-       if (put) {
+       err = do_sync_erase(ubi, e1, vol_id, lnum, 0);
+       if (err) {
+               kmem_cache_free(ubi_wl_entry_slab, e1);
+               if (e2)
+                       kmem_cache_free(ubi_wl_entry_slab, e2);
+               goto out_ro;
+       }
+
+       if (e2) {
                /*
                 * Well, the target PEB was put meanwhile, schedule it for
                 * erasure.
                 */
-               dbg_wl("PEB %d was put meanwhile, erase", e2->pnum);
-               err = schedule_erase(ubi, e2, 0);
-               if (err)
-                       goto out_error;
-       }
-
-       if (!protect) {
-               err = schedule_erase(ubi, e1, 0);
-               if (err)
-                       goto out_error;
+               dbg_wl("PEB %d (LEB %d:%d) was put meanwhile, erase",
+                      e2->pnum, vol_id, lnum);
+               err = do_sync_erase(ubi, e2, vol_id, lnum, 0);
+               if (err) {
+                       kmem_cache_free(ubi_wl_entry_slab, e2);
+                       goto out_ro;
+               }
        }
 
-
        dbg_wl("done");
        mutex_unlock(&ubi->move_mutex);
        return 0;
@@ -904,42 +1241,60 @@ static int wear_leveling_worker(struct ubi_device *ubi, struct ubi_work *wrk,
        /*
         * For some reasons the LEB was not moved, might be an error, might be
         * something else. @e1 was not changed, so return it back. @e2 might
-        * be changed, schedule it for erasure.
+        * have been changed, schedule it for erasure.
         */
 out_not_moved:
-       ubi_free_vid_hdr(ubi, vid_hdr);
+       if (vol_id != -1)
+               dbg_wl("cancel moving PEB %d (LEB %d:%d) to PEB %d (%d)",
+                      e1->pnum, vol_id, lnum, e2->pnum, err);
+       else
+               dbg_wl("cancel moving PEB %d to PEB %d (%d)",
+                      e1->pnum, e2->pnum, err);
        spin_lock(&ubi->wl_lock);
-       if (scrubbing)
+       if (protect)
+               prot_queue_add(ubi, e1);
+       else if (erroneous) {
+               wl_tree_add(e1, &ubi->erroneous);
+               ubi->erroneous_peb_count += 1;
+       } else if (scrubbing)
                wl_tree_add(e1, &ubi->scrub);
        else
                wl_tree_add(e1, &ubi->used);
+       ubi_assert(!ubi->move_to_put);
        ubi->move_from = ubi->move_to = NULL;
-       ubi->move_to_put = ubi->wl_scheduled = 0;
+       ubi->wl_scheduled = 0;
        spin_unlock(&ubi->wl_lock);
 
-       err = schedule_erase(ubi, e2, 0);
-       if (err)
-               goto out_error;
-
+       ubi_free_vid_hdr(ubi, vid_hdr);
+       err = do_sync_erase(ubi, e2, vol_id, lnum, torture);
+       if (err) {
+               kmem_cache_free(ubi_wl_entry_slab, e2);
+               goto out_ro;
+       }
        mutex_unlock(&ubi->move_mutex);
        return 0;
 
 out_error:
-       ubi_err("error %d while moving PEB %d to PEB %d",
-               err, e1->pnum, e2->pnum);
-
-       ubi_free_vid_hdr(ubi, vid_hdr);
+       if (vol_id != -1)
+               ubi_err("error %d while moving PEB %d to PEB %d",
+                       err, e1->pnum, e2->pnum);
+       else
+               ubi_err("error %d while moving PEB %d (LEB %d:%d) to PEB %d",
+                       err, e1->pnum, vol_id, lnum, e2->pnum);
        spin_lock(&ubi->wl_lock);
        ubi->move_from = ubi->move_to = NULL;
        ubi->move_to_put = ubi->wl_scheduled = 0;
        spin_unlock(&ubi->wl_lock);
 
+       ubi_free_vid_hdr(ubi, vid_hdr);
        kmem_cache_free(ubi_wl_entry_slab, e1);
        kmem_cache_free(ubi_wl_entry_slab, e2);
-       ubi_ro_mode(ubi);
 
+out_ro:
+       ubi_ro_mode(ubi);
        mutex_unlock(&ubi->move_mutex);
-       return err;
+       ubi_assert(err != 0);
+       return err < 0 ? err : -EIO;
 
 out_cancel:
        ubi->wl_scheduled = 0;
@@ -952,12 +1307,13 @@ out_cancel:
 /**
  * ensure_wear_leveling - schedule wear-leveling if it is needed.
  * @ubi: UBI device description object
+ * @nested: set to non-zero if this function is called from UBI worker
  *
  * This function checks if it is time to start wear-leveling and schedules it
  * if yes. This function returns zero in case of success and a negative error
  * code in case of failure.
  */
-static int ensure_wear_leveling(struct ubi_device *ubi)
+static int ensure_wear_leveling(struct ubi_device *ubi, int nested)
 {
        int err = 0;
        struct ubi_wl_entry *e1;
@@ -981,11 +1337,11 @@ static int ensure_wear_leveling(struct ubi_device *ubi)
                /*
                 * We schedule wear-leveling only if the difference between the
                 * lowest erase counter of used physical eraseblocks and a high
-                * erase counter of free physical eraseblocks is greater then
+                * erase counter of free physical eraseblocks is greater than
                 * %UBI_WL_THRESHOLD.
                 */
-               e1 = rb_entry(rb_first(&ubi->used), struct ubi_wl_entry, rb);
-               e2 = find_wl_entry(&ubi->free, WL_FREE_MAX_DIFF);
+               e1 = rb_entry(rb_first(&ubi->used), struct ubi_wl_entry, u.rb);
+               e2 = find_wl_entry(ubi, &ubi->free, WL_FREE_MAX_DIFF);
 
                if (!(e2->ec - e1->ec >= UBI_WL_THRESHOLD))
                        goto out_unlock;
@@ -1002,8 +1358,12 @@ static int ensure_wear_leveling(struct ubi_device *ubi)
                goto out_cancel;
        }
 
+       wrk->anchor = 0;
        wrk->func = &wear_leveling_worker;
-       schedule_ubi_work(ubi, wrk);
+       if (nested)
+               __schedule_ubi_work(ubi, wrk);
+       else
+               schedule_ubi_work(ubi, wrk);
        return err;
 
 out_cancel:
@@ -1014,6 +1374,38 @@ out_unlock:
        return err;
 }
 
+#ifdef CONFIG_MTD_UBI_FASTMAP
+/**
+ * ubi_ensure_anchor_pebs - schedule wear-leveling to produce an anchor PEB.
+ * @ubi: UBI device description object
+ */
+int ubi_ensure_anchor_pebs(struct ubi_device *ubi)
+{
+       struct ubi_work *wrk;
+
+       spin_lock(&ubi->wl_lock);
+       if (ubi->wl_scheduled) {
+               spin_unlock(&ubi->wl_lock);
+               return 0;
+       }
+       ubi->wl_scheduled = 1;
+       spin_unlock(&ubi->wl_lock);
+
+       wrk = kmalloc(sizeof(struct ubi_work), GFP_NOFS);
+       if (!wrk) {
+               spin_lock(&ubi->wl_lock);
+               ubi->wl_scheduled = 0;
+               spin_unlock(&ubi->wl_lock);
+               return -ENOMEM;
+       }
+
+       wrk->anchor = 1;
+       wrk->func = &wear_leveling_worker;
+       schedule_ubi_work(ubi, wrk);
+       return 0;
+}
+#endif
+
 /**
  * erase_worker - physical eraseblock erase worker function.
  * @ubi: UBI device description object
@@ -1029,7 +1421,10 @@ static int erase_worker(struct ubi_device *ubi, struct ubi_work *wl_wrk,
                        int cancel)
 {
        struct ubi_wl_entry *e = wl_wrk->e;
-       int pnum = e->pnum, err, need;
+       int pnum = e->pnum;
+       int vol_id = wl_wrk->vol_id;
+       int lnum = wl_wrk->lnum;
+       int err, available_consumed = 0;
 
        if (cancel) {
                dbg_wl("cancel erasure of PEB %d EC %d", pnum, e->ec);
@@ -1038,7 +1433,10 @@ static int erase_worker(struct ubi_device *ubi, struct ubi_work *wl_wrk,
                return 0;
        }
 
-       dbg_wl("erase PEB %d EC %d", pnum, e->ec);
+       dbg_wl("erase PEB %d EC %d LEB %d:%d",
+              pnum, e->ec, wl_wrk->vol_id, wl_wrk->lnum);
+
+       ubi_assert(!ubi_is_fm_block(ubi, e->pnum));
 
        err = sync_erase(ubi, e, wl_wrk->torture);
        if (!err) {
@@ -1046,44 +1444,45 @@ static int erase_worker(struct ubi_device *ubi, struct ubi_work *wl_wrk,
                kfree(wl_wrk);
 
                spin_lock(&ubi->wl_lock);
-               ubi->abs_ec += 1;
                wl_tree_add(e, &ubi->free);
+               ubi->free_count++;
                spin_unlock(&ubi->wl_lock);
 
                /*
-                * One more erase operation has happened, take care about protected
-                * physical eraseblocks.
+                * One more erase operation has happened, take care about
+                * protected physical eraseblocks.
                 */
-               check_protection_over(ubi);
+               serve_prot_queue(ubi);
 
                /* And take care about wear-leveling */
-               err = ensure_wear_leveling(ubi);
+               err = ensure_wear_leveling(ubi, 1);
                return err;
        }
 
        ubi_err("failed to erase PEB %d, error %d", pnum, err);
        kfree(wl_wrk);
-       kmem_cache_free(ubi_wl_entry_slab, e);
 
        if (err == -EINTR || err == -ENOMEM || err == -EAGAIN ||
            err == -EBUSY) {
                int err1;
 
                /* Re-schedule the LEB for erasure */
-               err1 = schedule_erase(ubi, e, 0);
+               err1 = schedule_erase(ubi, e, vol_id, lnum, 0);
                if (err1) {
                        err = err1;
                        goto out_ro;
                }
                return err;
-       } else if (err != -EIO) {
+       }
+
+       kmem_cache_free(ubi_wl_entry_slab, e);
+       if (err != -EIO)
                /*
                 * If this is not %-EIO, we have no idea what to do. Scheduling
                 * this physical eraseblock for erasure again would cause
-                * errors again and again. Well, lets switch to RO mode.
+                * errors again and again. Well, lets switch to R/O mode.
                 */
                goto out_ro;
-       }
 
        /* It is %-EIO, the PEB went bad */
 
@@ -1093,48 +1492,62 @@ static int erase_worker(struct ubi_device *ubi, struct ubi_work *wl_wrk,
        }
 
        spin_lock(&ubi->volumes_lock);
-       need = ubi->beb_rsvd_level - ubi->beb_rsvd_pebs + 1;
-       if (need > 0) {
-               need = ubi->avail_pebs >= need ? need : ubi->avail_pebs;
-               ubi->avail_pebs -= need;
-               ubi->rsvd_pebs += need;
-               ubi->beb_rsvd_pebs += need;
-               if (need > 0)
-                       ubi_msg("reserve more %d PEBs", need);
-       }
-
        if (ubi->beb_rsvd_pebs == 0) {
-               spin_unlock(&ubi->volumes_lock);
-               ubi_err("no reserved physical eraseblocks");
-               goto out_ro;
+               if (ubi->avail_pebs == 0) {
+                       spin_unlock(&ubi->volumes_lock);
+                       ubi_err("no reserved/available physical eraseblocks");
+                       goto out_ro;
+               }
+               ubi->avail_pebs -= 1;
+               available_consumed = 1;
        }
-
        spin_unlock(&ubi->volumes_lock);
-       ubi_msg("mark PEB %d as bad", pnum);
 
+       ubi_msg("mark PEB %d as bad", pnum);
        err = ubi_io_mark_bad(ubi, pnum);
        if (err)
                goto out_ro;
 
        spin_lock(&ubi->volumes_lock);
-       ubi->beb_rsvd_pebs -= 1;
+       if (ubi->beb_rsvd_pebs > 0) {
+               if (available_consumed) {
+                       /*
+                        * The amount of reserved PEBs increased since we last
+                        * checked.
+                        */
+                       ubi->avail_pebs += 1;
+                       available_consumed = 0;
+               }
+               ubi->beb_rsvd_pebs -= 1;
+       }
        ubi->bad_peb_count += 1;
        ubi->good_peb_count -= 1;
        ubi_calculate_reserved(ubi);
-       if (ubi->beb_rsvd_pebs == 0)
-               ubi_warn("last PEB from the reserved pool was used");
+       if (available_consumed)
+               ubi_warn("no PEBs in the reserved pool, used an available PEB");
+       else if (ubi->beb_rsvd_pebs)
+               ubi_msg("%d PEBs left in the reserve", ubi->beb_rsvd_pebs);
+       else
+               ubi_warn("last PEB from the reserve was used");
        spin_unlock(&ubi->volumes_lock);
 
        return err;
 
 out_ro:
+       if (available_consumed) {
+               spin_lock(&ubi->volumes_lock);
+               ubi->avail_pebs += 1;
+               spin_unlock(&ubi->volumes_lock);
+       }
        ubi_ro_mode(ubi);
        return err;
 }
 
 /**
- * ubi_wl_put_peb - return a physical eraseblock to the wear-leveling unit.
+ * ubi_wl_put_peb - return a PEB to the wear-leveling sub-system.
  * @ubi: UBI device description object
+ * @vol_id: the volume ID that last used this PEB
+ * @lnum: the last used logical eraseblock number for the PEB
  * @pnum: physical eraseblock to return
  * @torture: if this physical eraseblock has to be tortured
  *
@@ -1143,7 +1556,8 @@ out_ro:
  * occurred to this @pnum and it has to be tested. This function returns zero
  * in case of success, and a negative error code in case of failure.
  */
-int ubi_wl_put_peb(struct ubi_device *ubi, int pnum, int torture)
+int ubi_wl_put_peb(struct ubi_device *ubi, int vol_id, int lnum,
+                  int pnum, int torture)
 {
        int err;
        struct ubi_wl_entry *e;
@@ -1172,11 +1586,11 @@ retry:
                /*
                 * User is putting the physical eraseblock which was selected
                 * as the target the data is moved to. It may happen if the EBA
-                * unit already re-mapped the LEB in 'ubi_eba_copy_leb()' but
-                * the WL unit has not put the PEB to the "used" tree yet, but
-                * it is about to do this. So we just set a flag which will
-                * tell the WL worker that the PEB is not needed anymore and
-                * should be scheduled for erasure.
+                * sub-system already re-mapped the LEB in 'ubi_eba_copy_leb()'
+                * but the WL sub-system has not put the PEB to the "used" tree
+                * yet, but it is about to do this. So we just set a flag which
+                * will tell the WL worker that the PEB is not needed anymore
+                * and should be scheduled for erasure.
                 */
                dbg_wl("PEB %d is the target of data moving", pnum);
                ubi_assert(!ubi->move_to_put);
@@ -1185,13 +1599,20 @@ retry:
                return 0;
        } else {
                if (in_wl_tree(e, &ubi->used)) {
-                       paranoid_check_in_wl_tree(e, &ubi->used);
-                       rb_erase(&e->rb, &ubi->used);
+                       self_check_in_wl_tree(ubi, e, &ubi->used);
+                       rb_erase(&e->u.rb, &ubi->used);
                } else if (in_wl_tree(e, &ubi->scrub)) {
-                       paranoid_check_in_wl_tree(e, &ubi->scrub);
-                       rb_erase(&e->rb, &ubi->scrub);
+                       self_check_in_wl_tree(ubi, e, &ubi->scrub);
+                       rb_erase(&e->u.rb, &ubi->scrub);
+               } else if (in_wl_tree(e, &ubi->erroneous)) {
+                       self_check_in_wl_tree(ubi, e, &ubi->erroneous);
+                       rb_erase(&e->u.rb, &ubi->erroneous);
+                       ubi->erroneous_peb_count -= 1;
+                       ubi_assert(ubi->erroneous_peb_count >= 0);
+                       /* Erroneous PEBs should be tortured */
+                       torture = 1;
                } else {
-                       err = prot_tree_del(ubi, e->pnum);
+                       err = prot_queue_del(ubi, e->pnum);
                        if (err) {
                                ubi_err("PEB %d not found", pnum);
                                ubi_ro_mode(ubi);
@@ -1202,7 +1623,7 @@ retry:
        }
        spin_unlock(&ubi->wl_lock);
 
-       err = schedule_erase(ubi, e, torture);
+       err = schedule_erase(ubi, e, vol_id, lnum, torture);
        if (err) {
                spin_lock(&ubi->wl_lock);
                wl_tree_add(e, &ubi->used);
@@ -1231,7 +1652,8 @@ int ubi_wl_scrub_peb(struct ubi_device *ubi, int pnum)
 retry:
        spin_lock(&ubi->wl_lock);
        e = ubi->lookuptbl[pnum];
-       if (e == ubi->move_from || in_wl_tree(e, &ubi->scrub)) {
+       if (e == ubi->move_from || in_wl_tree(e, &ubi->scrub) ||
+                                  in_wl_tree(e, &ubi->erroneous)) {
                spin_unlock(&ubi->wl_lock);
                return 0;
        }
@@ -1250,12 +1672,12 @@ retry:
        }
 
        if (in_wl_tree(e, &ubi->used)) {
-               paranoid_check_in_wl_tree(e, &ubi->used);
-               rb_erase(&e->rb, &ubi->used);
+               self_check_in_wl_tree(ubi, e, &ubi->used);
+               rb_erase(&e->u.rb, &ubi->used);
        } else {
                int err;
 
-               err = prot_tree_del(ubi, e->pnum);
+               err = prot_queue_del(ubi, e->pnum);
                if (err) {
                        ubi_err("PEB %d not found", pnum);
                        ubi_ro_mode(ubi);
@@ -1271,29 +1693,60 @@ retry:
         * Technically scrubbing is the same as wear-leveling, so it is done
         * by the WL worker.
         */
-       return ensure_wear_leveling(ubi);
+       return ensure_wear_leveling(ubi, 0);
 }
 
 /**
  * ubi_wl_flush - flush all pending works.
  * @ubi: UBI device description object
+ * @vol_id: the volume id to flush for
+ * @lnum: the logical eraseblock number to flush for
  *
- * This function returns zero in case of success and a negative error code in
- * case of failure.
+ * This function executes all pending works for a particular volume id /
+ * logical eraseblock number pair. If either value is set to %UBI_ALL, then it
+ * acts as a wildcard for all of the corresponding volume numbers or logical
+ * eraseblock numbers. It returns zero in case of success and a negative error
+ * code in case of failure.
  */
-int ubi_wl_flush(struct ubi_device *ubi)
+int ubi_wl_flush(struct ubi_device *ubi, int vol_id, int lnum)
 {
-       int err;
+       int err = 0;
+       int found = 1;
 
        /*
-        * Erase while the pending works queue is not empty, but not more then
+        * Erase while the pending works queue is not empty, but not more than
         * the number of currently pending works.
         */
-       dbg_wl("flush (%d pending works)", ubi->works_count);
-       while (ubi->works_count) {
-               err = do_work(ubi);
-               if (err)
-                       return err;
+       dbg_wl("flush pending work for LEB %d:%d (%d pending works)",
+              vol_id, lnum, ubi->works_count);
+
+       while (found) {
+               struct ubi_work *wrk;
+               found = 0;
+
+               down_read(&ubi->work_sem);
+               spin_lock(&ubi->wl_lock);
+               list_for_each_entry(wrk, &ubi->works, list) {
+                       if ((vol_id == UBI_ALL || wrk->vol_id == vol_id) &&
+                           (lnum == UBI_ALL || wrk->lnum == lnum)) {
+                               list_del(&wrk->list);
+                               ubi->works_count -= 1;
+                               ubi_assert(ubi->works_count >= 0);
+                               spin_unlock(&ubi->wl_lock);
+
+                               err = wrk->func(ubi, wrk, 0);
+                               if (err) {
+                                       up_read(&ubi->work_sem);
+                                       return err;
+                               }
+
+                               spin_lock(&ubi->wl_lock);
+                               found = 1;
+                               break;
+                       }
+               }
+               spin_unlock(&ubi->wl_lock);
+               up_read(&ubi->work_sem);
        }
 
        /*
@@ -1303,18 +1756,7 @@ int ubi_wl_flush(struct ubi_device *ubi)
        down_write(&ubi->work_sem);
        up_write(&ubi->work_sem);
 
-       /*
-        * And in case last was the WL worker and it cancelled the LEB
-        * movement, flush again.
-        */
-       while (ubi->works_count) {
-               dbg_wl("flush more (%d pending works)", ubi->works_count);
-               err = do_work(ubi);
-               if (err)
-                       return err;
-       }
-
-       return 0;
+       return err;
 }
 
 /**
@@ -1333,11 +1775,11 @@ static void tree_destroy(struct rb_root *root)
                else if (rb->rb_right)
                        rb = rb->rb_right;
                else {
-                       e = rb_entry(rb, struct ubi_wl_entry, rb);
+                       e = rb_entry(rb, struct ubi_wl_entry, u.rb);
 
                        rb = rb_parent(rb);
                        if (rb) {
-                               if (rb->rb_left == &e->rb)
+                               if (rb->rb_left == &e->u.rb)
                                        rb->rb_left = NULL;
                                else
                                        rb->rb_right = NULL;
@@ -1372,7 +1814,7 @@ int ubi_thread(void *u)
 
                spin_lock(&ubi->wl_lock);
                if (list_empty(&ubi->works) || ubi->ro_mode ||
-                              !ubi->thread_enabled) {
+                   !ubi->thread_enabled || ubi_dbg_is_bgt_disabled(ubi)) {
                        set_current_state(TASK_INTERRUPTIBLE);
                        spin_unlock(&ubi->wl_lock);
                        schedule();
@@ -1392,7 +1834,8 @@ int ubi_thread(void *u)
                                ubi_msg("%s: %d consecutive failures",
                                        ubi->bgt_name, WL_MAX_FAILURES);
                                ubi_ro_mode(ubi);
-                               break;
+                               ubi->thread_enabled = 0;
+                               continue;
                        }
                } else
                        failures = 0;
@@ -1422,30 +1865,32 @@ static void cancel_pending(struct ubi_device *ubi)
 }
 
 /**
- * ubi_wl_init_scan - initialize the wear-leveling unit using scanning
- * information.
+ * ubi_wl_init - initialize the WL sub-system using attaching information.
  * @ubi: UBI device description object
- * @si: scanning information
+ * @ai: attaching information
  *
  * This function returns zero in case of success, and a negative error code in
  * case of failure.
  */
-int ubi_wl_init_scan(struct ubi_device *ubi, struct ubi_scan_info *si)
+int ubi_wl_init(struct ubi_device *ubi, struct ubi_attach_info *ai)
 {
-       int err;
+       int err, i, reserved_pebs, found_pebs = 0;
        struct rb_node *rb1, *rb2;
-       struct ubi_scan_volume *sv;
-       struct ubi_scan_leb *seb, *tmp;
+       struct ubi_ainf_volume *av;
+       struct ubi_ainf_peb *aeb, *tmp;
        struct ubi_wl_entry *e;
 
-
-       ubi->used = ubi->free = ubi->scrub = RB_ROOT;
-       ubi->prot.pnum = ubi->prot.aec = RB_ROOT;
+       ubi->used = ubi->erroneous = ubi->free = ubi->scrub = RB_ROOT;
        spin_lock_init(&ubi->wl_lock);
        mutex_init(&ubi->move_mutex);
        init_rwsem(&ubi->work_sem);
-       ubi->max_ec = si->max_ec;
+       ubi->max_ec = ai->max_ec;
        INIT_LIST_HEAD(&ubi->works);
+#ifndef __UBOOT__
+#ifdef CONFIG_MTD_UBI_FASTMAP
+       INIT_WORK(&ubi->fm_work, update_fastmap_work_fn);
+#endif
+#endif
 
        sprintf(ubi->bgt_name, UBI_BGT_NAME_PATTERN, ubi->ubi_num);
 
@@ -1454,64 +1899,63 @@ int ubi_wl_init_scan(struct ubi_device *ubi, struct ubi_scan_info *si)
        if (!ubi->lookuptbl)
                return err;
 
-       list_for_each_entry_safe(seb, tmp, &si->erase, u.list) {
+       for (i = 0; i < UBI_PROT_QUEUE_LEN; i++)
+               INIT_LIST_HEAD(&ubi->pq[i]);
+       ubi->pq_head = 0;
+
+       list_for_each_entry_safe(aeb, tmp, &ai->erase, u.list) {
                cond_resched();
 
                e = kmem_cache_alloc(ubi_wl_entry_slab, GFP_KERNEL);
                if (!e)
                        goto out_free;
 
-               e->pnum = seb->pnum;
-               e->ec = seb->ec;
+               e->pnum = aeb->pnum;
+               e->ec = aeb->ec;
+               ubi_assert(!ubi_is_fm_block(ubi, e->pnum));
                ubi->lookuptbl[e->pnum] = e;
-               if (schedule_erase(ubi, e, 0)) {
+               if (schedule_erase(ubi, e, aeb->vol_id, aeb->lnum, 0)) {
                        kmem_cache_free(ubi_wl_entry_slab, e);
                        goto out_free;
                }
+
+               found_pebs++;
        }
 
-       list_for_each_entry(seb, &si->free, u.list) {
+       ubi->free_count = 0;
+       list_for_each_entry(aeb, &ai->free, u.list) {
                cond_resched();
 
                e = kmem_cache_alloc(ubi_wl_entry_slab, GFP_KERNEL);
                if (!e)
                        goto out_free;
 
-               e->pnum = seb->pnum;
-               e->ec = seb->ec;
+               e->pnum = aeb->pnum;
+               e->ec = aeb->ec;
                ubi_assert(e->ec >= 0);
-               wl_tree_add(e, &ubi->free);
-               ubi->lookuptbl[e->pnum] = e;
-       }
-
-       list_for_each_entry(seb, &si->corr, u.list) {
-               cond_resched();
+               ubi_assert(!ubi_is_fm_block(ubi, e->pnum));
 
-               e = kmem_cache_alloc(ubi_wl_entry_slab, GFP_KERNEL);
-               if (!e)
-                       goto out_free;
+               wl_tree_add(e, &ubi->free);
+               ubi->free_count++;
 
-               e->pnum = seb->pnum;
-               e->ec = seb->ec;
                ubi->lookuptbl[e->pnum] = e;
-               if (schedule_erase(ubi, e, 0)) {
-                       kmem_cache_free(ubi_wl_entry_slab, e);
-                       goto out_free;
-               }
+
+               found_pebs++;
        }
 
-       ubi_rb_for_each_entry(rb1, sv, &si->volumes, rb) {
-               ubi_rb_for_each_entry(rb2, seb, &sv->root, u.rb) {
+       ubi_rb_for_each_entry(rb1, av, &ai->volumes, rb) {
+               ubi_rb_for_each_entry(rb2, aeb, &av->root, u.rb) {
                        cond_resched();
 
                        e = kmem_cache_alloc(ubi_wl_entry_slab, GFP_KERNEL);
                        if (!e)
                                goto out_free;
 
-                       e->pnum = seb->pnum;
-                       e->ec = seb->ec;
+                       e->pnum = aeb->pnum;
+                       e->ec = aeb->ec;
                        ubi->lookuptbl[e->pnum] = e;
-                       if (!seb->scrub) {
+
+                       if (!aeb->scrub) {
                                dbg_wl("add PEB %d EC %d to the used tree",
                                       e->pnum, e->ec);
                                wl_tree_add(e, &ubi->used);
@@ -1520,20 +1964,38 @@ int ubi_wl_init_scan(struct ubi_device *ubi, struct ubi_scan_info *si)
                                       e->pnum, e->ec);
                                wl_tree_add(e, &ubi->scrub);
                        }
+
+                       found_pebs++;
                }
        }
 
-       if (ubi->avail_pebs < WL_RESERVED_PEBS) {
+       dbg_wl("found %i PEBs", found_pebs);
+
+       if (ubi->fm)
+               ubi_assert(ubi->good_peb_count == \
+                          found_pebs + ubi->fm->used_blocks);
+       else
+               ubi_assert(ubi->good_peb_count == found_pebs);
+
+       reserved_pebs = WL_RESERVED_PEBS;
+#ifdef CONFIG_MTD_UBI_FASTMAP
+       /* Reserve enough LEBs to store two fastmaps. */
+       reserved_pebs += (ubi->fm_size / ubi->leb_size) * 2;
+#endif
+
+       if (ubi->avail_pebs < reserved_pebs) {
                ubi_err("no enough physical eraseblocks (%d, need %d)",
-                       ubi->avail_pebs, WL_RESERVED_PEBS);
-               err = -ENOSPC;
+                       ubi->avail_pebs, reserved_pebs);
+               if (ubi->corr_peb_count)
+                       ubi_err("%d PEBs are corrupted and not used",
+                               ubi->corr_peb_count);
                goto out_free;
        }
-       ubi->avail_pebs -= WL_RESERVED_PEBS;
-       ubi->rsvd_pebs += WL_RESERVED_PEBS;
+       ubi->avail_pebs -= reserved_pebs;
+       ubi->rsvd_pebs += reserved_pebs;
 
        /* Schedule wear-leveling if needed */
-       err = ensure_wear_leveling(ubi);
+       err = ensure_wear_leveling(ubi, 0);
        if (err)
                goto out_free;
 
@@ -1549,72 +2011,57 @@ out_free:
 }
 
 /**
- * protection_trees_destroy - destroy the protection RB-trees.
+ * protection_queue_destroy - destroy the protection queue.
  * @ubi: UBI device description object
  */
-static void protection_trees_destroy(struct ubi_device *ubi)
+static void protection_queue_destroy(struct ubi_device *ubi)
 {
-       struct rb_node *rb;
-       struct ubi_wl_prot_entry *pe;
-
-       rb = ubi->prot.aec.rb_node;
-       while (rb) {
-               if (rb->rb_left)
-                       rb = rb->rb_left;
-               else if (rb->rb_right)
-                       rb = rb->rb_right;
-               else {
-                       pe = rb_entry(rb, struct ubi_wl_prot_entry, rb_aec);
-
-                       rb = rb_parent(rb);
-                       if (rb) {
-                               if (rb->rb_left == &pe->rb_aec)
-                                       rb->rb_left = NULL;
-                               else
-                                       rb->rb_right = NULL;
-                       }
+       int i;
+       struct ubi_wl_entry *e, *tmp;
 
-                       kmem_cache_free(ubi_wl_entry_slab, pe->e);
-                       kfree(pe);
+       for (i = 0; i < UBI_PROT_QUEUE_LEN; ++i) {
+               list_for_each_entry_safe(e, tmp, &ubi->pq[i], u.list) {
+                       list_del(&e->u.list);
+                       kmem_cache_free(ubi_wl_entry_slab, e);
                }
        }
 }
 
 /**
- * ubi_wl_close - close the wear-leveling unit.
+ * ubi_wl_close - close the wear-leveling sub-system.
  * @ubi: UBI device description object
  */
 void ubi_wl_close(struct ubi_device *ubi)
 {
-       dbg_wl("close the UBI wear-leveling unit");
-
+       dbg_wl("close the WL sub-system");
        cancel_pending(ubi);
-       protection_trees_destroy(ubi);
+       protection_queue_destroy(ubi);
        tree_destroy(&ubi->used);
+       tree_destroy(&ubi->erroneous);
        tree_destroy(&ubi->free);
        tree_destroy(&ubi->scrub);
        kfree(ubi->lookuptbl);
 }
 
-#ifdef CONFIG_MTD_UBI_DEBUG_PARANOID
-
 /**
- * paranoid_check_ec - make sure that the erase counter of a physical eraseblock
- * is correct.
+ * self_check_ec - make sure that the erase counter of a PEB is correct.
  * @ubi: UBI device description object
  * @pnum: the physical eraseblock number to check
  * @ec: the erase counter to check
  *
  * This function returns zero if the erase counter of physical eraseblock @pnum
- * is equivalent to @ec, %1 if not, and a negative error code if an error
+ * is equivalent to @ec, and a negative error code if not or if an error
  * occurred.
  */
-static int paranoid_check_ec(struct ubi_device *ubi, int pnum, int ec)
+static int self_check_ec(struct ubi_device *ubi, int pnum, int ec)
 {
        int err;
        long long read_ec;
        struct ubi_ec_hdr *ec_hdr;
 
+       if (!ubi_dbg_chk_gen(ubi))
+               return 0;
+
        ec_hdr = kzalloc(ubi->ec_hdr_alsize, GFP_NOFS);
        if (!ec_hdr)
                return -ENOMEM;
@@ -1627,10 +2074,10 @@ static int paranoid_check_ec(struct ubi_device *ubi, int pnum, int ec)
        }
 
        read_ec = be64_to_cpu(ec_hdr->ec);
-       if (ec != read_ec) {
-               ubi_err("paranoid check failed for PEB %d", pnum);
+       if (ec != read_ec && read_ec - ec > 1) {
+               ubi_err("self-check failed for PEB %d", pnum);
                ubi_err("read EC is %lld, should be %d", read_ec, ec);
-               ubi_dbg_dump_stack();
+               dump_stack();
                err = 1;
        } else
                err = 0;
@@ -1641,24 +2088,53 @@ out_free:
 }
 
 /**
- * paranoid_check_in_wl_tree - make sure that a wear-leveling entry is present
- * in a WL RB-tree.
+ * self_check_in_wl_tree - check that wear-leveling entry is in WL RB-tree.
+ * @ubi: UBI device description object
  * @e: the wear-leveling entry to check
  * @root: the root of the tree
  *
- * This function returns zero if @e is in the @root RB-tree and %1 if it
+ * This function returns zero if @e is in the @root RB-tree and %-EINVAL if it
  * is not.
  */
-static int paranoid_check_in_wl_tree(struct ubi_wl_entry *e,
-                                    struct rb_root *root)
+static int self_check_in_wl_tree(const struct ubi_device *ubi,
+                                struct ubi_wl_entry *e, struct rb_root *root)
 {
+       if (!ubi_dbg_chk_gen(ubi))
+               return 0;
+
        if (in_wl_tree(e, root))
                return 0;
 
-       ubi_err("paranoid check failed for PEB %d, EC %d, RB-tree %p ",
+       ubi_err("self-check failed for PEB %d, EC %d, RB-tree %p ",
                e->pnum, e->ec, root);
-       ubi_dbg_dump_stack();
-       return 1;
+       dump_stack();
+       return -EINVAL;
 }
 
-#endif /* CONFIG_MTD_UBI_DEBUG_PARANOID */
+/**
+ * self_check_in_pq - check if wear-leveling entry is in the protection
+ *                        queue.
+ * @ubi: UBI device description object
+ * @e: the wear-leveling entry to check
+ *
+ * This function returns zero if @e is in @ubi->pq and %-EINVAL if it is not.
+ */
+static int self_check_in_pq(const struct ubi_device *ubi,
+                           struct ubi_wl_entry *e)
+{
+       struct ubi_wl_entry *p;
+       int i;
+
+       if (!ubi_dbg_chk_gen(ubi))
+               return 0;
+
+       for (i = 0; i < UBI_PROT_QUEUE_LEN; ++i)
+               list_for_each_entry(p, &ubi->pq[i], u.list)
+                       if (p == e)
+                               return 0;
+
+       ubi_err("self-check failed for PEB %d, EC %d, Protect queue",
+               e->pnum, e->ec);
+       dump_stack();
+       return -EINVAL;
+}
index 85377ea2a614adad1912eac7e226cc253aa2f866..9ed40177cb993ccb4d137c9bef58eb81f8a497e9 100644 (file)
@@ -3,18 +3,7 @@
  *
  * Copyright (C) 2006-2008 Nokia Corporation.
  *
- * This program is free software; you can redistribute it and/or modify it
- * under the terms of the GNU General Public License version 2 as published by
- * the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
- * more details.
- *
- * You should have received a copy of the GNU General Public License along with
- * this program; if not, write to the Free Software Foundation, Inc., 51
- * Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
+ * SPDX-License-Identifier:    GPL-2.0+
  *
  * Authors: Adrian Hunter
  *          Artem Bityutskiy (Битюцкий Артём)
  */
 
 #include "ubifs.h"
+#define __UBOOT__
+#ifndef __UBOOT__
+#include <linux/writeback.h>
+#else
+#include <linux/err.h>
+#endif
 #include <linux/math64.h>
 
+/*
+ * When pessimistic budget calculations say that there is no enough space,
+ * UBIFS starts writing back dirty inodes and pages, doing garbage collection,
+ * or committing. The below constant defines maximum number of times UBIFS
+ * repeats the operations.
+ */
+#define MAX_MKSPC_RETRIES 3
+
+/*
+ * The below constant defines amount of dirty pages which should be written
+ * back at when trying to shrink the liability.
+ */
+#define NR_TO_WRITE 16
+
+#ifndef __UBOOT__
+/**
+ * shrink_liability - write-back some dirty pages/inodes.
+ * @c: UBIFS file-system description object
+ * @nr_to_write: how many dirty pages to write-back
+ *
+ * This function shrinks UBIFS liability by means of writing back some amount
+ * of dirty inodes and their pages.
+ *
+ * Note, this function synchronizes even VFS inodes which are locked
+ * (@i_mutex) by the caller of the budgeting function, because write-back does
+ * not touch @i_mutex.
+ */
+static void shrink_liability(struct ubifs_info *c, int nr_to_write)
+{
+       down_read(&c->vfs_sb->s_umount);
+       writeback_inodes_sb(c->vfs_sb, WB_REASON_FS_FREE_SPACE);
+       up_read(&c->vfs_sb->s_umount);
+}
+
+/**
+ * run_gc - run garbage collector.
+ * @c: UBIFS file-system description object
+ *
+ * This function runs garbage collector to make some more free space. Returns
+ * zero if a free LEB has been produced, %-EAGAIN if commit is required, and a
+ * negative error code in case of failure.
+ */
+static int run_gc(struct ubifs_info *c)
+{
+       int err, lnum;
+
+       /* Make some free space by garbage-collecting dirty space */
+       down_read(&c->commit_sem);
+       lnum = ubifs_garbage_collect(c, 1);
+       up_read(&c->commit_sem);
+       if (lnum < 0)
+               return lnum;
+
+       /* GC freed one LEB, return it to lprops */
+       dbg_budg("GC freed LEB %d", lnum);
+       err = ubifs_return_leb(c, lnum);
+       if (err)
+               return err;
+       return 0;
+}
+
 /**
- * ubifs_calc_min_idx_lebs - calculate amount of eraseblocks for the index.
+ * get_liability - calculate current liability.
  * @c: UBIFS file-system description object
  *
- * This function calculates and returns the number of eraseblocks which should
- * be kept for index usage.
+ * This function calculates and returns current UBIFS liability, i.e. the
+ * amount of bytes UBIFS has "promised" to write to the media.
+ */
+static long long get_liability(struct ubifs_info *c)
+{
+       long long liab;
+
+       spin_lock(&c->space_lock);
+       liab = c->bi.idx_growth + c->bi.data_growth + c->bi.dd_growth;
+       spin_unlock(&c->space_lock);
+       return liab;
+}
+
+/**
+ * make_free_space - make more free space on the file-system.
+ * @c: UBIFS file-system description object
+ *
+ * This function is called when an operation cannot be budgeted because there
+ * is supposedly no free space. But in most cases there is some free space:
+ *   o budgeting is pessimistic, so it always budgets more than it is actually
+ *     needed, so shrinking the liability is one way to make free space - the
+ *     cached data will take less space then it was budgeted for;
+ *   o GC may turn some dark space into free space (budgeting treats dark space
+ *     as not available);
+ *   o commit may free some LEB, i.e., turn freeable LEBs into free LEBs.
+ *
+ * So this function tries to do the above. Returns %-EAGAIN if some free space
+ * was presumably made and the caller has to re-try budgeting the operation.
+ * Returns %-ENOSPC if it couldn't do more free space, and other negative error
+ * codes on failures.
+ */
+static int make_free_space(struct ubifs_info *c)
+{
+       int err, retries = 0;
+       long long liab1, liab2;
+
+       do {
+               liab1 = get_liability(c);
+               /*
+                * We probably have some dirty pages or inodes (liability), try
+                * to write them back.
+                */
+               dbg_budg("liability %lld, run write-back", liab1);
+               shrink_liability(c, NR_TO_WRITE);
+
+               liab2 = get_liability(c);
+               if (liab2 < liab1)
+                       return -EAGAIN;
+
+               dbg_budg("new liability %lld (not shrunk)", liab2);
+
+               /* Liability did not shrink again, try GC */
+               dbg_budg("Run GC");
+               err = run_gc(c);
+               if (!err)
+                       return -EAGAIN;
+
+               if (err != -EAGAIN && err != -ENOSPC)
+                       /* Some real error happened */
+                       return err;
+
+               dbg_budg("Run commit (retries %d)", retries);
+               err = ubifs_run_commit(c);
+               if (err)
+                       return err;
+       } while (retries++ < MAX_MKSPC_RETRIES);
+
+       return -ENOSPC;
+}
+#endif
+
+/**
+ * ubifs_calc_min_idx_lebs - calculate amount of LEBs for the index.
+ * @c: UBIFS file-system description object
+ *
+ * This function calculates and returns the number of LEBs which should be kept
+ * for index usage.
  */
 int ubifs_calc_min_idx_lebs(struct ubifs_info *c)
 {
-       int idx_lebs, eff_leb_size = c->leb_size - c->max_idx_node_sz;
+       int idx_lebs;
        long long idx_size;
 
-       idx_size = c->old_idx_sz + c->budg_idx_growth + c->budg_uncommitted_idx;
-
+       idx_size = c->bi.old_idx_sz + c->bi.idx_growth + c->bi.uncommitted_idx;
        /* And make sure we have thrice the index size of space reserved */
-       idx_size = idx_size + (idx_size << 1);
-
+       idx_size += idx_size << 1;
        /*
         * We do not maintain 'old_idx_size' as 'old_idx_lebs'/'old_idx_bytes'
         * pair, nor similarly the two variables for the new index size, so we
         * have to do this costly 64-bit division on fast-path.
         */
-       idx_size += eff_leb_size - 1;
-       idx_lebs = div_u64(idx_size, eff_leb_size);
+       idx_lebs = div_u64(idx_size + c->idx_leb_size - 1, c->idx_leb_size);
        /*
         * The index head is not available for the in-the-gaps method, so add an
         * extra LEB to compensate.
@@ -67,6 +195,424 @@ int ubifs_calc_min_idx_lebs(struct ubifs_info *c)
        return idx_lebs;
 }
 
+#ifndef __UBOOT__
+/**
+ * ubifs_calc_available - calculate available FS space.
+ * @c: UBIFS file-system description object
+ * @min_idx_lebs: minimum number of LEBs reserved for the index
+ *
+ * This function calculates and returns amount of FS space available for use.
+ */
+long long ubifs_calc_available(const struct ubifs_info *c, int min_idx_lebs)
+{
+       int subtract_lebs;
+       long long available;
+
+       available = c->main_bytes - c->lst.total_used;
+
+       /*
+        * Now 'available' contains theoretically available flash space
+        * assuming there is no index, so we have to subtract the space which
+        * is reserved for the index.
+        */
+       subtract_lebs = min_idx_lebs;
+
+       /* Take into account that GC reserves one LEB for its own needs */
+       subtract_lebs += 1;
+
+       /*
+        * The GC journal head LEB is not really accessible. And since
+        * different write types go to different heads, we may count only on
+        * one head's space.
+        */
+       subtract_lebs += c->jhead_cnt - 1;
+
+       /* We also reserve one LEB for deletions, which bypass budgeting */
+       subtract_lebs += 1;
+
+       available -= (long long)subtract_lebs * c->leb_size;
+
+       /* Subtract the dead space which is not available for use */
+       available -= c->lst.total_dead;
+
+       /*
+        * Subtract dark space, which might or might not be usable - it depends
+        * on the data which we have on the media and which will be written. If
+        * this is a lot of uncompressed or not-compressible data, the dark
+        * space cannot be used.
+        */
+       available -= c->lst.total_dark;
+
+       /*
+        * However, there is more dark space. The index may be bigger than
+        * @min_idx_lebs. Those extra LEBs are assumed to be available, but
+        * their dark space is not included in total_dark, so it is subtracted
+        * here.
+        */
+       if (c->lst.idx_lebs > min_idx_lebs) {
+               subtract_lebs = c->lst.idx_lebs - min_idx_lebs;
+               available -= subtract_lebs * c->dark_wm;
+       }
+
+       /* The calculations are rough and may end up with a negative number */
+       return available > 0 ? available : 0;
+}
+
+/**
+ * can_use_rp - check whether the user is allowed to use reserved pool.
+ * @c: UBIFS file-system description object
+ *
+ * UBIFS has so-called "reserved pool" which is flash space reserved
+ * for the superuser and for uses whose UID/GID is recorded in UBIFS superblock.
+ * This function checks whether current user is allowed to use reserved pool.
+ * Returns %1  current user is allowed to use reserved pool and %0 otherwise.
+ */
+static int can_use_rp(struct ubifs_info *c)
+{
+       if (uid_eq(current_fsuid(), c->rp_uid) || capable(CAP_SYS_RESOURCE) ||
+           (!gid_eq(c->rp_gid, GLOBAL_ROOT_GID) && in_group_p(c->rp_gid)))
+               return 1;
+       return 0;
+}
+
+/**
+ * do_budget_space - reserve flash space for index and data growth.
+ * @c: UBIFS file-system description object
+ *
+ * This function makes sure UBIFS has enough free LEBs for index growth and
+ * data.
+ *
+ * When budgeting index space, UBIFS reserves thrice as many LEBs as the index
+ * would take if it was consolidated and written to the flash. This guarantees
+ * that the "in-the-gaps" commit method always succeeds and UBIFS will always
+ * be able to commit dirty index. So this function basically adds amount of
+ * budgeted index space to the size of the current index, multiplies this by 3,
+ * and makes sure this does not exceed the amount of free LEBs.
+ *
+ * Notes about @c->bi.min_idx_lebs and @c->lst.idx_lebs variables:
+ * o @c->lst.idx_lebs is the number of LEBs the index currently uses. It might
+ *    be large, because UBIFS does not do any index consolidation as long as
+ *    there is free space. IOW, the index may take a lot of LEBs, but the LEBs
+ *    will contain a lot of dirt.
+ * o @c->bi.min_idx_lebs is the number of LEBS the index presumably takes. IOW,
+ *    the index may be consolidated to take up to @c->bi.min_idx_lebs LEBs.
+ *
+ * This function returns zero in case of success, and %-ENOSPC in case of
+ * failure.
+ */
+static int do_budget_space(struct ubifs_info *c)
+{
+       long long outstanding, available;
+       int lebs, rsvd_idx_lebs, min_idx_lebs;
+
+       /* First budget index space */
+       min_idx_lebs = ubifs_calc_min_idx_lebs(c);
+
+       /* Now 'min_idx_lebs' contains number of LEBs to reserve */
+       if (min_idx_lebs > c->lst.idx_lebs)
+               rsvd_idx_lebs = min_idx_lebs - c->lst.idx_lebs;
+       else
+               rsvd_idx_lebs = 0;
+
+       /*
+        * The number of LEBs that are available to be used by the index is:
+        *
+        *    @c->lst.empty_lebs + @c->freeable_cnt + @c->idx_gc_cnt -
+        *    @c->lst.taken_empty_lebs
+        *
+        * @c->lst.empty_lebs are available because they are empty.
+        * @c->freeable_cnt are available because they contain only free and
+        * dirty space, @c->idx_gc_cnt are available because they are index
+        * LEBs that have been garbage collected and are awaiting the commit
+        * before they can be used. And the in-the-gaps method will grab these
+        * if it needs them. @c->lst.taken_empty_lebs are empty LEBs that have
+        * already been allocated for some purpose.
+        *
+        * Note, @c->idx_gc_cnt is included to both @c->lst.empty_lebs (because
+        * these LEBs are empty) and to @c->lst.taken_empty_lebs (because they
+        * are taken until after the commit).
+        *
+        * Note, @c->lst.taken_empty_lebs may temporarily be higher by one
+        * because of the way we serialize LEB allocations and budgeting. See a
+        * comment in 'ubifs_find_free_space()'.
+        */
+       lebs = c->lst.empty_lebs + c->freeable_cnt + c->idx_gc_cnt -
+              c->lst.taken_empty_lebs;
+       if (unlikely(rsvd_idx_lebs > lebs)) {
+               dbg_budg("out of indexing space: min_idx_lebs %d (old %d), rsvd_idx_lebs %d",
+                        min_idx_lebs, c->bi.min_idx_lebs, rsvd_idx_lebs);
+               return -ENOSPC;
+       }
+
+       available = ubifs_calc_available(c, min_idx_lebs);
+       outstanding = c->bi.data_growth + c->bi.dd_growth;
+
+       if (unlikely(available < outstanding)) {
+               dbg_budg("out of data space: available %lld, outstanding %lld",
+                        available, outstanding);
+               return -ENOSPC;
+       }
+
+       if (available - outstanding <= c->rp_size && !can_use_rp(c))
+               return -ENOSPC;
+
+       c->bi.min_idx_lebs = min_idx_lebs;
+       return 0;
+}
+
+/**
+ * calc_idx_growth - calculate approximate index growth from budgeting request.
+ * @c: UBIFS file-system description object
+ * @req: budgeting request
+ *
+ * For now we assume each new node adds one znode. But this is rather poor
+ * approximation, though.
+ */
+static int calc_idx_growth(const struct ubifs_info *c,
+                          const struct ubifs_budget_req *req)
+{
+       int znodes;
+
+       znodes = req->new_ino + (req->new_page << UBIFS_BLOCKS_PER_PAGE_SHIFT) +
+                req->new_dent;
+       return znodes * c->max_idx_node_sz;
+}
+
+/**
+ * calc_data_growth - calculate approximate amount of new data from budgeting
+ * request.
+ * @c: UBIFS file-system description object
+ * @req: budgeting request
+ */
+static int calc_data_growth(const struct ubifs_info *c,
+                           const struct ubifs_budget_req *req)
+{
+       int data_growth;
+
+       data_growth = req->new_ino  ? c->bi.inode_budget : 0;
+       if (req->new_page)
+               data_growth += c->bi.page_budget;
+       if (req->new_dent)
+               data_growth += c->bi.dent_budget;
+       data_growth += req->new_ino_d;
+       return data_growth;
+}
+
+/**
+ * calc_dd_growth - calculate approximate amount of data which makes other data
+ * dirty from budgeting request.
+ * @c: UBIFS file-system description object
+ * @req: budgeting request
+ */
+static int calc_dd_growth(const struct ubifs_info *c,
+                         const struct ubifs_budget_req *req)
+{
+       int dd_growth;
+
+       dd_growth = req->dirtied_page ? c->bi.page_budget : 0;
+
+       if (req->dirtied_ino)
+               dd_growth += c->bi.inode_budget << (req->dirtied_ino - 1);
+       if (req->mod_dent)
+               dd_growth += c->bi.dent_budget;
+       dd_growth += req->dirtied_ino_d;
+       return dd_growth;
+}
+
+/**
+ * ubifs_budget_space - ensure there is enough space to complete an operation.
+ * @c: UBIFS file-system description object
+ * @req: budget request
+ *
+ * This function allocates budget for an operation. It uses pessimistic
+ * approximation of how much flash space the operation needs. The goal of this
+ * function is to make sure UBIFS always has flash space to flush all dirty
+ * pages, dirty inodes, and dirty znodes (liability). This function may force
+ * commit, garbage-collection or write-back. Returns zero in case of success,
+ * %-ENOSPC if there is no free space and other negative error codes in case of
+ * failures.
+ */
+int ubifs_budget_space(struct ubifs_info *c, struct ubifs_budget_req *req)
+{
+       int uninitialized_var(cmt_retries), uninitialized_var(wb_retries);
+       int err, idx_growth, data_growth, dd_growth, retried = 0;
+
+       ubifs_assert(req->new_page <= 1);
+       ubifs_assert(req->dirtied_page <= 1);
+       ubifs_assert(req->new_dent <= 1);
+       ubifs_assert(req->mod_dent <= 1);
+       ubifs_assert(req->new_ino <= 1);
+       ubifs_assert(req->new_ino_d <= UBIFS_MAX_INO_DATA);
+       ubifs_assert(req->dirtied_ino <= 4);
+       ubifs_assert(req->dirtied_ino_d <= UBIFS_MAX_INO_DATA * 4);
+       ubifs_assert(!(req->new_ino_d & 7));
+       ubifs_assert(!(req->dirtied_ino_d & 7));
+
+       data_growth = calc_data_growth(c, req);
+       dd_growth = calc_dd_growth(c, req);
+       if (!data_growth && !dd_growth)
+               return 0;
+       idx_growth = calc_idx_growth(c, req);
+
+again:
+       spin_lock(&c->space_lock);
+       ubifs_assert(c->bi.idx_growth >= 0);
+       ubifs_assert(c->bi.data_growth >= 0);
+       ubifs_assert(c->bi.dd_growth >= 0);
+
+       if (unlikely(c->bi.nospace) && (c->bi.nospace_rp || !can_use_rp(c))) {
+               dbg_budg("no space");
+               spin_unlock(&c->space_lock);
+               return -ENOSPC;
+       }
+
+       c->bi.idx_growth += idx_growth;
+       c->bi.data_growth += data_growth;
+       c->bi.dd_growth += dd_growth;
+
+       err = do_budget_space(c);
+       if (likely(!err)) {
+               req->idx_growth = idx_growth;
+               req->data_growth = data_growth;
+               req->dd_growth = dd_growth;
+               spin_unlock(&c->space_lock);
+               return 0;
+       }
+
+       /* Restore the old values */
+       c->bi.idx_growth -= idx_growth;
+       c->bi.data_growth -= data_growth;
+       c->bi.dd_growth -= dd_growth;
+       spin_unlock(&c->space_lock);
+
+       if (req->fast) {
+               dbg_budg("no space for fast budgeting");
+               return err;
+       }
+
+       err = make_free_space(c);
+       cond_resched();
+       if (err == -EAGAIN) {
+               dbg_budg("try again");
+               goto again;
+       } else if (err == -ENOSPC) {
+               if (!retried) {
+                       retried = 1;
+                       dbg_budg("-ENOSPC, but anyway try once again");
+                       goto again;
+               }
+               dbg_budg("FS is full, -ENOSPC");
+               c->bi.nospace = 1;
+               if (can_use_rp(c) || c->rp_size == 0)
+                       c->bi.nospace_rp = 1;
+               smp_wmb();
+       } else
+               ubifs_err("cannot budget space, error %d", err);
+       return err;
+}
+
+/**
+ * ubifs_release_budget - release budgeted free space.
+ * @c: UBIFS file-system description object
+ * @req: budget request
+ *
+ * This function releases the space budgeted by 'ubifs_budget_space()'. Note,
+ * since the index changes (which were budgeted for in @req->idx_growth) will
+ * only be written to the media on commit, this function moves the index budget
+ * from @c->bi.idx_growth to @c->bi.uncommitted_idx. The latter will be zeroed
+ * by the commit operation.
+ */
+void ubifs_release_budget(struct ubifs_info *c, struct ubifs_budget_req *req)
+{
+       ubifs_assert(req->new_page <= 1);
+       ubifs_assert(req->dirtied_page <= 1);
+       ubifs_assert(req->new_dent <= 1);
+       ubifs_assert(req->mod_dent <= 1);
+       ubifs_assert(req->new_ino <= 1);
+       ubifs_assert(req->new_ino_d <= UBIFS_MAX_INO_DATA);
+       ubifs_assert(req->dirtied_ino <= 4);
+       ubifs_assert(req->dirtied_ino_d <= UBIFS_MAX_INO_DATA * 4);
+       ubifs_assert(!(req->new_ino_d & 7));
+       ubifs_assert(!(req->dirtied_ino_d & 7));
+       if (!req->recalculate) {
+               ubifs_assert(req->idx_growth >= 0);
+               ubifs_assert(req->data_growth >= 0);
+               ubifs_assert(req->dd_growth >= 0);
+       }
+
+       if (req->recalculate) {
+               req->data_growth = calc_data_growth(c, req);
+               req->dd_growth = calc_dd_growth(c, req);
+               req->idx_growth = calc_idx_growth(c, req);
+       }
+
+       if (!req->data_growth && !req->dd_growth)
+               return;
+
+       c->bi.nospace = c->bi.nospace_rp = 0;
+       smp_wmb();
+
+       spin_lock(&c->space_lock);
+       c->bi.idx_growth -= req->idx_growth;
+       c->bi.uncommitted_idx += req->idx_growth;
+       c->bi.data_growth -= req->data_growth;
+       c->bi.dd_growth -= req->dd_growth;
+       c->bi.min_idx_lebs = ubifs_calc_min_idx_lebs(c);
+
+       ubifs_assert(c->bi.idx_growth >= 0);
+       ubifs_assert(c->bi.data_growth >= 0);
+       ubifs_assert(c->bi.dd_growth >= 0);
+       ubifs_assert(c->bi.min_idx_lebs < c->main_lebs);
+       ubifs_assert(!(c->bi.idx_growth & 7));
+       ubifs_assert(!(c->bi.data_growth & 7));
+       ubifs_assert(!(c->bi.dd_growth & 7));
+       spin_unlock(&c->space_lock);
+}
+
+/**
+ * ubifs_convert_page_budget - convert budget of a new page.
+ * @c: UBIFS file-system description object
+ *
+ * This function converts budget which was allocated for a new page of data to
+ * the budget of changing an existing page of data. The latter is smaller than
+ * the former, so this function only does simple re-calculation and does not
+ * involve any write-back.
+ */
+void ubifs_convert_page_budget(struct ubifs_info *c)
+{
+       spin_lock(&c->space_lock);
+       /* Release the index growth reservation */
+       c->bi.idx_growth -= c->max_idx_node_sz << UBIFS_BLOCKS_PER_PAGE_SHIFT;
+       /* Release the data growth reservation */
+       c->bi.data_growth -= c->bi.page_budget;
+       /* Increase the dirty data growth reservation instead */
+       c->bi.dd_growth += c->bi.page_budget;
+       /* And re-calculate the indexing space reservation */
+       c->bi.min_idx_lebs = ubifs_calc_min_idx_lebs(c);
+       spin_unlock(&c->space_lock);
+}
+
+/**
+ * ubifs_release_dirty_inode_budget - release dirty inode budget.
+ * @c: UBIFS file-system description object
+ * @ui: UBIFS inode to release the budget for
+ *
+ * This function releases budget corresponding to a dirty inode. It is usually
+ * called when after the inode has been written to the media and marked as
+ * clean. It also causes the "no space" flags to be cleared.
+ */
+void ubifs_release_dirty_inode_budget(struct ubifs_info *c,
+                                     struct ubifs_inode *ui)
+{
+       struct ubifs_budget_req req;
+
+       memset(&req, 0, sizeof(struct ubifs_budget_req));
+       /* The "no space" flags will be cleared because dd_growth is > 0 */
+       req.dd_growth = c->bi.inode_budget + ALIGN(ui->data_len, 8);
+       ubifs_release_budget(c, &req);
+}
+#endif
+
 /**
  * ubifs_reported_space - calculate reported free space.
  * @c: the UBIFS file-system description object
@@ -111,3 +657,75 @@ long long ubifs_reported_space(const struct ubifs_info *c, long long free)
        free *= factor;
        return div_u64(free, divisor);
 }
+
+#ifndef __UBOOT__
+/**
+ * ubifs_get_free_space_nolock - return amount of free space.
+ * @c: UBIFS file-system description object
+ *
+ * This function calculates amount of free space to report to user-space.
+ *
+ * Because UBIFS may introduce substantial overhead (the index, node headers,
+ * alignment, wastage at the end of LEBs, etc), it cannot report real amount of
+ * free flash space it has (well, because not all dirty space is reclaimable,
+ * UBIFS does not actually know the real amount). If UBIFS did so, it would
+ * bread user expectations about what free space is. Users seem to accustomed
+ * to assume that if the file-system reports N bytes of free space, they would
+ * be able to fit a file of N bytes to the FS. This almost works for
+ * traditional file-systems, because they have way less overhead than UBIFS.
+ * So, to keep users happy, UBIFS tries to take the overhead into account.
+ */
+long long ubifs_get_free_space_nolock(struct ubifs_info *c)
+{
+       int rsvd_idx_lebs, lebs;
+       long long available, outstanding, free;
+
+       ubifs_assert(c->bi.min_idx_lebs == ubifs_calc_min_idx_lebs(c));
+       outstanding = c->bi.data_growth + c->bi.dd_growth;
+       available = ubifs_calc_available(c, c->bi.min_idx_lebs);
+
+       /*
+        * When reporting free space to user-space, UBIFS guarantees that it is
+        * possible to write a file of free space size. This means that for
+        * empty LEBs we may use more precise calculations than
+        * 'ubifs_calc_available()' is using. Namely, we know that in empty
+        * LEBs we would waste only @c->leb_overhead bytes, not @c->dark_wm.
+        * Thus, amend the available space.
+        *
+        * Note, the calculations below are similar to what we have in
+        * 'do_budget_space()', so refer there for comments.
+        */
+       if (c->bi.min_idx_lebs > c->lst.idx_lebs)
+               rsvd_idx_lebs = c->bi.min_idx_lebs - c->lst.idx_lebs;
+       else
+               rsvd_idx_lebs = 0;
+       lebs = c->lst.empty_lebs + c->freeable_cnt + c->idx_gc_cnt -
+              c->lst.taken_empty_lebs;
+       lebs -= rsvd_idx_lebs;
+       available += lebs * (c->dark_wm - c->leb_overhead);
+
+       if (available > outstanding)
+               free = ubifs_reported_space(c, available - outstanding);
+       else
+               free = 0;
+       return free;
+}
+
+/**
+ * ubifs_get_free_space - return amount of free space.
+ * @c: UBIFS file-system description object
+ *
+ * This function calculates and returns amount of free space to report to
+ * user-space.
+ */
+long long ubifs_get_free_space(struct ubifs_info *c)
+{
+       long long free;
+
+       spin_lock(&c->space_lock);
+       free = ubifs_get_free_space_nolock(c);
+       spin_unlock(&c->space_lock);
+
+       return free;
+}
+#endif
index 6afb8835a58e480e92f977bda8725dfd77154a77..2f50a554bbe8dd5591410697c36be789a40cdf7c 100644 (file)
@@ -3,18 +3,7 @@
  *
  * Copyright (C) 2006-2008 Nokia Corporation
  *
- * This program is free software; you can redistribute it and/or modify it
- * under the terms of the GNU General Public License version 2 as published by
- * the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
- * more details.
- *
- * You should have received a copy of the GNU General Public License along with
- * this program; if not, write to the Free Software Foundation, Inc., 51
- * Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
+ * SPDX-License-Identifier:    GPL-2.0+
  *
  * Authors: Artem Bityutskiy (Битюцкий Артём)
  *          Adrian Hunter
  * various local functions of those subsystems.
  */
 
-#define UBIFS_DBG_PRESERVE_UBI
-
+#define __UBOOT__
+#ifndef __UBOOT__
+#include <linux/module.h>
+#include <linux/debugfs.h>
+#include <linux/math64.h>
+#include <linux/uaccess.h>
+#include <linux/random.h>
+#else
+#include <linux/compat.h>
+#include <linux/err.h>
+#endif
 #include "ubifs.h"
 
-#ifdef CONFIG_UBIFS_FS_DEBUG
-
-DEFINE_SPINLOCK(dbg_lock);
+#ifndef __UBOOT__
+static DEFINE_SPINLOCK(dbg_lock);
+#endif
 
-static char dbg_key_buf0[128];
-static char dbg_key_buf1[128];
-
-unsigned int ubifs_msg_flags = UBIFS_MSG_FLAGS_DEFAULT;
-unsigned int ubifs_chk_flags = UBIFS_CHK_FLAGS_DEFAULT;
-unsigned int ubifs_tst_flags;
-
-module_param_named(debug_msgs, ubifs_msg_flags, uint, S_IRUGO | S_IWUSR);
-module_param_named(debug_chks, ubifs_chk_flags, uint, S_IRUGO | S_IWUSR);
-module_param_named(debug_tsts, ubifs_tst_flags, uint, S_IRUGO | S_IWUSR);
+static const char *get_key_fmt(int fmt)
+{
+       switch (fmt) {
+       case UBIFS_SIMPLE_KEY_FMT:
+               return "simple";
+       default:
+               return "unknown/invalid format";
+       }
+}
 
-MODULE_PARM_DESC(debug_msgs, "Debug message type flags");
-MODULE_PARM_DESC(debug_chks, "Debug check flags");
-MODULE_PARM_DESC(debug_tsts, "Debug special test flags");
+static const char *get_key_hash(int hash)
+{
+       switch (hash) {
+       case UBIFS_KEY_HASH_R5:
+               return "R5";
+       case UBIFS_KEY_HASH_TEST:
+               return "test";
+       default:
+               return "unknown/invalid name hash";
+       }
+}
 
 static const char *get_key_type(int type)
 {
@@ -68,8 +73,32 @@ static const char *get_key_type(int type)
        }
 }
 
-static void sprintf_key(const struct ubifs_info *c, const union ubifs_key *key,
-                       char *buffer)
+#ifndef __UBOOT__
+static const char *get_dent_type(int type)
+{
+       switch (type) {
+       case UBIFS_ITYPE_REG:
+               return "file";
+       case UBIFS_ITYPE_DIR:
+               return "dir";
+       case UBIFS_ITYPE_LNK:
+               return "symlink";
+       case UBIFS_ITYPE_BLK:
+               return "blkdev";
+       case UBIFS_ITYPE_CHR:
+               return "char dev";
+       case UBIFS_ITYPE_FIFO:
+               return "fifo";
+       case UBIFS_ITYPE_SOCK:
+               return "socket";
+       default:
+               return "unknown/invalid type";
+       }
+}
+#endif
+
+const char *dbg_snprintf_key(const struct ubifs_info *c,
+                            const union ubifs_key *key, char *buffer, int len)
 {
        char *p = buffer;
        int type = key_type(c, key);
@@ -77,80 +106,3038 @@ static void sprintf_key(const struct ubifs_info *c, const union ubifs_key *key,
        if (c->key_fmt == UBIFS_SIMPLE_KEY_FMT) {
                switch (type) {
                case UBIFS_INO_KEY:
-                       sprintf(p, "(%lu, %s)", (unsigned long)key_inum(c, key),
-                              get_key_type(type));
+                       len -= snprintf(p, len, "(%lu, %s)",
+                                       (unsigned long)key_inum(c, key),
+                                       get_key_type(type));
                        break;
                case UBIFS_DENT_KEY:
                case UBIFS_XENT_KEY:
-                       sprintf(p, "(%lu, %s, %#08x)",
-                               (unsigned long)key_inum(c, key),
-                               get_key_type(type), key_hash(c, key));
+                       len -= snprintf(p, len, "(%lu, %s, %#08x)",
+                                       (unsigned long)key_inum(c, key),
+                                       get_key_type(type), key_hash(c, key));
                        break;
                case UBIFS_DATA_KEY:
-                       sprintf(p, "(%lu, %s, %u)",
-                               (unsigned long)key_inum(c, key),
-                               get_key_type(type), key_block(c, key));
+                       len -= snprintf(p, len, "(%lu, %s, %u)",
+                                       (unsigned long)key_inum(c, key),
+                                       get_key_type(type), key_block(c, key));
                        break;
                case UBIFS_TRUN_KEY:
-                       sprintf(p, "(%lu, %s)",
-                               (unsigned long)key_inum(c, key),
-                               get_key_type(type));
+                       len -= snprintf(p, len, "(%lu, %s)",
+                                       (unsigned long)key_inum(c, key),
+                                       get_key_type(type));
                        break;
                default:
-                       sprintf(p, "(bad key type: %#08x, %#08x)",
-                               key->u32[0], key->u32[1]);
+                       len -= snprintf(p, len, "(bad key type: %#08x, %#08x)",
+                                       key->u32[0], key->u32[1]);
                }
        } else
-               sprintf(p, "bad key format %d", c->key_fmt);
+               len -= snprintf(p, len, "bad key format %d", c->key_fmt);
+       ubifs_assert(len > 0);
+       return p;
+}
+
+const char *dbg_ntype(int type)
+{
+       switch (type) {
+       case UBIFS_PAD_NODE:
+               return "padding node";
+       case UBIFS_SB_NODE:
+               return "superblock node";
+       case UBIFS_MST_NODE:
+               return "master node";
+       case UBIFS_REF_NODE:
+               return "reference node";
+       case UBIFS_INO_NODE:
+               return "inode node";
+       case UBIFS_DENT_NODE:
+               return "direntry node";
+       case UBIFS_XENT_NODE:
+               return "xentry node";
+       case UBIFS_DATA_NODE:
+               return "data node";
+       case UBIFS_TRUN_NODE:
+               return "truncate node";
+       case UBIFS_IDX_NODE:
+               return "indexing node";
+       case UBIFS_CS_NODE:
+               return "commit start node";
+       case UBIFS_ORPH_NODE:
+               return "orphan node";
+       default:
+               return "unknown node";
+       }
+}
+
+static const char *dbg_gtype(int type)
+{
+       switch (type) {
+       case UBIFS_NO_NODE_GROUP:
+               return "no node group";
+       case UBIFS_IN_NODE_GROUP:
+               return "in node group";
+       case UBIFS_LAST_OF_NODE_GROUP:
+               return "last of node group";
+       default:
+               return "unknown";
+       }
+}
+
+const char *dbg_cstate(int cmt_state)
+{
+       switch (cmt_state) {
+       case COMMIT_RESTING:
+               return "commit resting";
+       case COMMIT_BACKGROUND:
+               return "background commit requested";
+       case COMMIT_REQUIRED:
+               return "commit required";
+       case COMMIT_RUNNING_BACKGROUND:
+               return "BACKGROUND commit running";
+       case COMMIT_RUNNING_REQUIRED:
+               return "commit running and required";
+       case COMMIT_BROKEN:
+               return "broken commit";
+       default:
+               return "unknown commit state";
+       }
+}
+
+const char *dbg_jhead(int jhead)
+{
+       switch (jhead) {
+       case GCHD:
+               return "0 (GC)";
+       case BASEHD:
+               return "1 (base)";
+       case DATAHD:
+               return "2 (data)";
+       default:
+               return "unknown journal head";
+       }
+}
+
+static void dump_ch(const struct ubifs_ch *ch)
+{
+       pr_err("\tmagic          %#x\n", le32_to_cpu(ch->magic));
+       pr_err("\tcrc            %#x\n", le32_to_cpu(ch->crc));
+       pr_err("\tnode_type      %d (%s)\n", ch->node_type,
+              dbg_ntype(ch->node_type));
+       pr_err("\tgroup_type     %d (%s)\n", ch->group_type,
+              dbg_gtype(ch->group_type));
+       pr_err("\tsqnum          %llu\n",
+              (unsigned long long)le64_to_cpu(ch->sqnum));
+       pr_err("\tlen            %u\n", le32_to_cpu(ch->len));
+}
+
+void ubifs_dump_inode(struct ubifs_info *c, const struct inode *inode)
+{
+#ifndef __UBOOT__
+       const struct ubifs_inode *ui = ubifs_inode(inode);
+       struct qstr nm = { .name = NULL };
+       union ubifs_key key;
+       struct ubifs_dent_node *dent, *pdent = NULL;
+       int count = 2;
+
+       pr_err("Dump in-memory inode:");
+       pr_err("\tinode          %lu\n", inode->i_ino);
+       pr_err("\tsize           %llu\n",
+              (unsigned long long)i_size_read(inode));
+       pr_err("\tnlink          %u\n", inode->i_nlink);
+       pr_err("\tuid            %u\n", (unsigned int)i_uid_read(inode));
+       pr_err("\tgid            %u\n", (unsigned int)i_gid_read(inode));
+       pr_err("\tatime          %u.%u\n",
+              (unsigned int)inode->i_atime.tv_sec,
+              (unsigned int)inode->i_atime.tv_nsec);
+       pr_err("\tmtime          %u.%u\n",
+              (unsigned int)inode->i_mtime.tv_sec,
+              (unsigned int)inode->i_mtime.tv_nsec);
+       pr_err("\tctime          %u.%u\n",
+              (unsigned int)inode->i_ctime.tv_sec,
+              (unsigned int)inode->i_ctime.tv_nsec);
+       pr_err("\tcreat_sqnum    %llu\n", ui->creat_sqnum);
+       pr_err("\txattr_size     %u\n", ui->xattr_size);
+       pr_err("\txattr_cnt      %u\n", ui->xattr_cnt);
+       pr_err("\txattr_names    %u\n", ui->xattr_names);
+       pr_err("\tdirty          %u\n", ui->dirty);
+       pr_err("\txattr          %u\n", ui->xattr);
+       pr_err("\tbulk_read      %u\n", ui->xattr);
+       pr_err("\tsynced_i_size  %llu\n",
+              (unsigned long long)ui->synced_i_size);
+       pr_err("\tui_size        %llu\n",
+              (unsigned long long)ui->ui_size);
+       pr_err("\tflags          %d\n", ui->flags);
+       pr_err("\tcompr_type     %d\n", ui->compr_type);
+       pr_err("\tlast_page_read %lu\n", ui->last_page_read);
+       pr_err("\tread_in_a_row  %lu\n", ui->read_in_a_row);
+       pr_err("\tdata_len       %d\n", ui->data_len);
+
+       if (!S_ISDIR(inode->i_mode))
+               return;
+
+       pr_err("List of directory entries:\n");
+       ubifs_assert(!mutex_is_locked(&c->tnc_mutex));
+
+       lowest_dent_key(c, &key, inode->i_ino);
+       while (1) {
+               dent = ubifs_tnc_next_ent(c, &key, &nm);
+               if (IS_ERR(dent)) {
+                       if (PTR_ERR(dent) != -ENOENT)
+                               pr_err("error %ld\n", PTR_ERR(dent));
+                       break;
+               }
+
+               pr_err("\t%d: %s (%s)\n",
+                      count++, dent->name, get_dent_type(dent->type));
+
+               nm.name = dent->name;
+               nm.len = le16_to_cpu(dent->nlen);
+               kfree(pdent);
+               pdent = dent;
+               key_read(c, &dent->key, &key);
+       }
+       kfree(pdent);
+#endif
+}
+
+void ubifs_dump_node(const struct ubifs_info *c, const void *node)
+{
+       int i, n;
+       union ubifs_key key;
+       const struct ubifs_ch *ch = node;
+       char key_buf[DBG_KEY_BUF_LEN];
+
+       /* If the magic is incorrect, just hexdump the first bytes */
+       if (le32_to_cpu(ch->magic) != UBIFS_NODE_MAGIC) {
+               pr_err("Not a node, first %zu bytes:", UBIFS_CH_SZ);
+               print_hex_dump(KERN_ERR, "", DUMP_PREFIX_OFFSET, 32, 1,
+                              (void *)node, UBIFS_CH_SZ, 1);
+               return;
+       }
+
+       spin_lock(&dbg_lock);
+       dump_ch(node);
+
+       switch (ch->node_type) {
+       case UBIFS_PAD_NODE:
+       {
+               const struct ubifs_pad_node *pad = node;
+
+               pr_err("\tpad_len        %u\n", le32_to_cpu(pad->pad_len));
+               break;
+       }
+       case UBIFS_SB_NODE:
+       {
+               const struct ubifs_sb_node *sup = node;
+               unsigned int sup_flags = le32_to_cpu(sup->flags);
+
+               pr_err("\tkey_hash       %d (%s)\n",
+                      (int)sup->key_hash, get_key_hash(sup->key_hash));
+               pr_err("\tkey_fmt        %d (%s)\n",
+                      (int)sup->key_fmt, get_key_fmt(sup->key_fmt));
+               pr_err("\tflags          %#x\n", sup_flags);
+               pr_err("\t  big_lpt      %u\n",
+                      !!(sup_flags & UBIFS_FLG_BIGLPT));
+               pr_err("\t  space_fixup  %u\n",
+                      !!(sup_flags & UBIFS_FLG_SPACE_FIXUP));
+               pr_err("\tmin_io_size    %u\n", le32_to_cpu(sup->min_io_size));
+               pr_err("\tleb_size       %u\n", le32_to_cpu(sup->leb_size));
+               pr_err("\tleb_cnt        %u\n", le32_to_cpu(sup->leb_cnt));
+               pr_err("\tmax_leb_cnt    %u\n", le32_to_cpu(sup->max_leb_cnt));
+               pr_err("\tmax_bud_bytes  %llu\n",
+                      (unsigned long long)le64_to_cpu(sup->max_bud_bytes));
+               pr_err("\tlog_lebs       %u\n", le32_to_cpu(sup->log_lebs));
+               pr_err("\tlpt_lebs       %u\n", le32_to_cpu(sup->lpt_lebs));
+               pr_err("\torph_lebs      %u\n", le32_to_cpu(sup->orph_lebs));
+               pr_err("\tjhead_cnt      %u\n", le32_to_cpu(sup->jhead_cnt));
+               pr_err("\tfanout         %u\n", le32_to_cpu(sup->fanout));
+               pr_err("\tlsave_cnt      %u\n", le32_to_cpu(sup->lsave_cnt));
+               pr_err("\tdefault_compr  %u\n",
+                      (int)le16_to_cpu(sup->default_compr));
+               pr_err("\trp_size        %llu\n",
+                      (unsigned long long)le64_to_cpu(sup->rp_size));
+               pr_err("\trp_uid         %u\n", le32_to_cpu(sup->rp_uid));
+               pr_err("\trp_gid         %u\n", le32_to_cpu(sup->rp_gid));
+               pr_err("\tfmt_version    %u\n", le32_to_cpu(sup->fmt_version));
+               pr_err("\ttime_gran      %u\n", le32_to_cpu(sup->time_gran));
+               pr_err("\tUUID           %pUB\n", sup->uuid);
+               break;
+       }
+       case UBIFS_MST_NODE:
+       {
+               const struct ubifs_mst_node *mst = node;
+
+               pr_err("\thighest_inum   %llu\n",
+                      (unsigned long long)le64_to_cpu(mst->highest_inum));
+               pr_err("\tcommit number  %llu\n",
+                      (unsigned long long)le64_to_cpu(mst->cmt_no));
+               pr_err("\tflags          %#x\n", le32_to_cpu(mst->flags));
+               pr_err("\tlog_lnum       %u\n", le32_to_cpu(mst->log_lnum));
+               pr_err("\troot_lnum      %u\n", le32_to_cpu(mst->root_lnum));
+               pr_err("\troot_offs      %u\n", le32_to_cpu(mst->root_offs));
+               pr_err("\troot_len       %u\n", le32_to_cpu(mst->root_len));
+               pr_err("\tgc_lnum        %u\n", le32_to_cpu(mst->gc_lnum));
+               pr_err("\tihead_lnum     %u\n", le32_to_cpu(mst->ihead_lnum));
+               pr_err("\tihead_offs     %u\n", le32_to_cpu(mst->ihead_offs));
+               pr_err("\tindex_size     %llu\n",
+                      (unsigned long long)le64_to_cpu(mst->index_size));
+               pr_err("\tlpt_lnum       %u\n", le32_to_cpu(mst->lpt_lnum));
+               pr_err("\tlpt_offs       %u\n", le32_to_cpu(mst->lpt_offs));
+               pr_err("\tnhead_lnum     %u\n", le32_to_cpu(mst->nhead_lnum));
+               pr_err("\tnhead_offs     %u\n", le32_to_cpu(mst->nhead_offs));
+               pr_err("\tltab_lnum      %u\n", le32_to_cpu(mst->ltab_lnum));
+               pr_err("\tltab_offs      %u\n", le32_to_cpu(mst->ltab_offs));
+               pr_err("\tlsave_lnum     %u\n", le32_to_cpu(mst->lsave_lnum));
+               pr_err("\tlsave_offs     %u\n", le32_to_cpu(mst->lsave_offs));
+               pr_err("\tlscan_lnum     %u\n", le32_to_cpu(mst->lscan_lnum));
+               pr_err("\tleb_cnt        %u\n", le32_to_cpu(mst->leb_cnt));
+               pr_err("\tempty_lebs     %u\n", le32_to_cpu(mst->empty_lebs));
+               pr_err("\tidx_lebs       %u\n", le32_to_cpu(mst->idx_lebs));
+               pr_err("\ttotal_free     %llu\n",
+                      (unsigned long long)le64_to_cpu(mst->total_free));
+               pr_err("\ttotal_dirty    %llu\n",
+                      (unsigned long long)le64_to_cpu(mst->total_dirty));
+               pr_err("\ttotal_used     %llu\n",
+                      (unsigned long long)le64_to_cpu(mst->total_used));
+               pr_err("\ttotal_dead     %llu\n",
+                      (unsigned long long)le64_to_cpu(mst->total_dead));
+               pr_err("\ttotal_dark     %llu\n",
+                      (unsigned long long)le64_to_cpu(mst->total_dark));
+               break;
+       }
+       case UBIFS_REF_NODE:
+       {
+               const struct ubifs_ref_node *ref = node;
+
+               pr_err("\tlnum           %u\n", le32_to_cpu(ref->lnum));
+               pr_err("\toffs           %u\n", le32_to_cpu(ref->offs));
+               pr_err("\tjhead          %u\n", le32_to_cpu(ref->jhead));
+               break;
+       }
+       case UBIFS_INO_NODE:
+       {
+               const struct ubifs_ino_node *ino = node;
+
+               key_read(c, &ino->key, &key);
+               pr_err("\tkey            %s\n",
+                      dbg_snprintf_key(c, &key, key_buf, DBG_KEY_BUF_LEN));
+               pr_err("\tcreat_sqnum    %llu\n",
+                      (unsigned long long)le64_to_cpu(ino->creat_sqnum));
+               pr_err("\tsize           %llu\n",
+                      (unsigned long long)le64_to_cpu(ino->size));
+               pr_err("\tnlink          %u\n", le32_to_cpu(ino->nlink));
+               pr_err("\tatime          %lld.%u\n",
+                      (long long)le64_to_cpu(ino->atime_sec),
+                      le32_to_cpu(ino->atime_nsec));
+               pr_err("\tmtime          %lld.%u\n",
+                      (long long)le64_to_cpu(ino->mtime_sec),
+                      le32_to_cpu(ino->mtime_nsec));
+               pr_err("\tctime          %lld.%u\n",
+                      (long long)le64_to_cpu(ino->ctime_sec),
+                      le32_to_cpu(ino->ctime_nsec));
+               pr_err("\tuid            %u\n", le32_to_cpu(ino->uid));
+               pr_err("\tgid            %u\n", le32_to_cpu(ino->gid));
+               pr_err("\tmode           %u\n", le32_to_cpu(ino->mode));
+               pr_err("\tflags          %#x\n", le32_to_cpu(ino->flags));
+               pr_err("\txattr_cnt      %u\n", le32_to_cpu(ino->xattr_cnt));
+               pr_err("\txattr_size     %u\n", le32_to_cpu(ino->xattr_size));
+               pr_err("\txattr_names    %u\n", le32_to_cpu(ino->xattr_names));
+               pr_err("\tcompr_type     %#x\n",
+                      (int)le16_to_cpu(ino->compr_type));
+               pr_err("\tdata len       %u\n", le32_to_cpu(ino->data_len));
+               break;
+       }
+       case UBIFS_DENT_NODE:
+       case UBIFS_XENT_NODE:
+       {
+               const struct ubifs_dent_node *dent = node;
+               int nlen = le16_to_cpu(dent->nlen);
+
+               key_read(c, &dent->key, &key);
+               pr_err("\tkey            %s\n",
+                      dbg_snprintf_key(c, &key, key_buf, DBG_KEY_BUF_LEN));
+               pr_err("\tinum           %llu\n",
+                      (unsigned long long)le64_to_cpu(dent->inum));
+               pr_err("\ttype           %d\n", (int)dent->type);
+               pr_err("\tnlen           %d\n", nlen);
+               pr_err("\tname           ");
+
+               if (nlen > UBIFS_MAX_NLEN)
+                       pr_err("(bad name length, not printing, bad or corrupted node)");
+               else {
+                       for (i = 0; i < nlen && dent->name[i]; i++)
+                               pr_cont("%c", dent->name[i]);
+               }
+               pr_cont("\n");
+
+               break;
+       }
+       case UBIFS_DATA_NODE:
+       {
+               const struct ubifs_data_node *dn = node;
+               int dlen = le32_to_cpu(ch->len) - UBIFS_DATA_NODE_SZ;
+
+               key_read(c, &dn->key, &key);
+               pr_err("\tkey            %s\n",
+                      dbg_snprintf_key(c, &key, key_buf, DBG_KEY_BUF_LEN));
+               pr_err("\tsize           %u\n", le32_to_cpu(dn->size));
+               pr_err("\tcompr_typ      %d\n",
+                      (int)le16_to_cpu(dn->compr_type));
+               pr_err("\tdata size      %d\n", dlen);
+               pr_err("\tdata:\n");
+               print_hex_dump(KERN_ERR, "\t", DUMP_PREFIX_OFFSET, 32, 1,
+                              (void *)&dn->data, dlen, 0);
+               break;
+       }
+       case UBIFS_TRUN_NODE:
+       {
+               const struct ubifs_trun_node *trun = node;
+
+               pr_err("\tinum           %u\n", le32_to_cpu(trun->inum));
+               pr_err("\told_size       %llu\n",
+                      (unsigned long long)le64_to_cpu(trun->old_size));
+               pr_err("\tnew_size       %llu\n",
+                      (unsigned long long)le64_to_cpu(trun->new_size));
+               break;
+       }
+       case UBIFS_IDX_NODE:
+       {
+               const struct ubifs_idx_node *idx = node;
+
+               n = le16_to_cpu(idx->child_cnt);
+               pr_err("\tchild_cnt      %d\n", n);
+               pr_err("\tlevel          %d\n", (int)le16_to_cpu(idx->level));
+               pr_err("\tBranches:\n");
+
+               for (i = 0; i < n && i < c->fanout - 1; i++) {
+                       const struct ubifs_branch *br;
+
+                       br = ubifs_idx_branch(c, idx, i);
+                       key_read(c, &br->key, &key);
+                       pr_err("\t%d: LEB %d:%d len %d key %s\n",
+                              i, le32_to_cpu(br->lnum), le32_to_cpu(br->offs),
+                              le32_to_cpu(br->len),
+                              dbg_snprintf_key(c, &key, key_buf,
+                                               DBG_KEY_BUF_LEN));
+               }
+               break;
+       }
+       case UBIFS_CS_NODE:
+               break;
+       case UBIFS_ORPH_NODE:
+       {
+               const struct ubifs_orph_node *orph = node;
+
+               pr_err("\tcommit number  %llu\n",
+                      (unsigned long long)
+                               le64_to_cpu(orph->cmt_no) & LLONG_MAX);
+               pr_err("\tlast node flag %llu\n",
+                      (unsigned long long)(le64_to_cpu(orph->cmt_no)) >> 63);
+               n = (le32_to_cpu(ch->len) - UBIFS_ORPH_NODE_SZ) >> 3;
+               pr_err("\t%d orphan inode numbers:\n", n);
+               for (i = 0; i < n; i++)
+                       pr_err("\t  ino %llu\n",
+                              (unsigned long long)le64_to_cpu(orph->inos[i]));
+               break;
+       }
+       default:
+               pr_err("node type %d was not recognized\n",
+                      (int)ch->node_type);
+       }
+       spin_unlock(&dbg_lock);
+}
+
+void ubifs_dump_budget_req(const struct ubifs_budget_req *req)
+{
+       spin_lock(&dbg_lock);
+       pr_err("Budgeting request: new_ino %d, dirtied_ino %d\n",
+              req->new_ino, req->dirtied_ino);
+       pr_err("\tnew_ino_d   %d, dirtied_ino_d %d\n",
+              req->new_ino_d, req->dirtied_ino_d);
+       pr_err("\tnew_page    %d, dirtied_page %d\n",
+              req->new_page, req->dirtied_page);
+       pr_err("\tnew_dent    %d, mod_dent     %d\n",
+              req->new_dent, req->mod_dent);
+       pr_err("\tidx_growth  %d\n", req->idx_growth);
+       pr_err("\tdata_growth %d dd_growth     %d\n",
+              req->data_growth, req->dd_growth);
+       spin_unlock(&dbg_lock);
+}
+
+void ubifs_dump_lstats(const struct ubifs_lp_stats *lst)
+{
+       spin_lock(&dbg_lock);
+       pr_err("(pid %d) Lprops statistics: empty_lebs %d, idx_lebs  %d\n",
+              current->pid, lst->empty_lebs, lst->idx_lebs);
+       pr_err("\ttaken_empty_lebs %d, total_free %lld, total_dirty %lld\n",
+              lst->taken_empty_lebs, lst->total_free, lst->total_dirty);
+       pr_err("\ttotal_used %lld, total_dark %lld, total_dead %lld\n",
+              lst->total_used, lst->total_dark, lst->total_dead);
+       spin_unlock(&dbg_lock);
+}
+
+#ifndef __UBOOT__
+void ubifs_dump_budg(struct ubifs_info *c, const struct ubifs_budg_info *bi)
+{
+       int i;
+       struct rb_node *rb;
+       struct ubifs_bud *bud;
+       struct ubifs_gced_idx_leb *idx_gc;
+       long long available, outstanding, free;
+
+       spin_lock(&c->space_lock);
+       spin_lock(&dbg_lock);
+       pr_err("(pid %d) Budgeting info: data budget sum %lld, total budget sum %lld\n",
+              current->pid, bi->data_growth + bi->dd_growth,
+              bi->data_growth + bi->dd_growth + bi->idx_growth);
+       pr_err("\tbudg_data_growth %lld, budg_dd_growth %lld, budg_idx_growth %lld\n",
+              bi->data_growth, bi->dd_growth, bi->idx_growth);
+       pr_err("\tmin_idx_lebs %d, old_idx_sz %llu, uncommitted_idx %lld\n",
+              bi->min_idx_lebs, bi->old_idx_sz, bi->uncommitted_idx);
+       pr_err("\tpage_budget %d, inode_budget %d, dent_budget %d\n",
+              bi->page_budget, bi->inode_budget, bi->dent_budget);
+       pr_err("\tnospace %u, nospace_rp %u\n", bi->nospace, bi->nospace_rp);
+       pr_err("\tdark_wm %d, dead_wm %d, max_idx_node_sz %d\n",
+              c->dark_wm, c->dead_wm, c->max_idx_node_sz);
+
+       if (bi != &c->bi)
+               /*
+                * If we are dumping saved budgeting data, do not print
+                * additional information which is about the current state, not
+                * the old one which corresponded to the saved budgeting data.
+                */
+               goto out_unlock;
+
+       pr_err("\tfreeable_cnt %d, calc_idx_sz %lld, idx_gc_cnt %d\n",
+              c->freeable_cnt, c->calc_idx_sz, c->idx_gc_cnt);
+       pr_err("\tdirty_pg_cnt %ld, dirty_zn_cnt %ld, clean_zn_cnt %ld\n",
+              atomic_long_read(&c->dirty_pg_cnt),
+              atomic_long_read(&c->dirty_zn_cnt),
+              atomic_long_read(&c->clean_zn_cnt));
+       pr_err("\tgc_lnum %d, ihead_lnum %d\n", c->gc_lnum, c->ihead_lnum);
+
+       /* If we are in R/O mode, journal heads do not exist */
+       if (c->jheads)
+               for (i = 0; i < c->jhead_cnt; i++)
+                       pr_err("\tjhead %s\t LEB %d\n",
+                              dbg_jhead(c->jheads[i].wbuf.jhead),
+                              c->jheads[i].wbuf.lnum);
+       for (rb = rb_first(&c->buds); rb; rb = rb_next(rb)) {
+               bud = rb_entry(rb, struct ubifs_bud, rb);
+               pr_err("\tbud LEB %d\n", bud->lnum);
+       }
+       list_for_each_entry(bud, &c->old_buds, list)
+               pr_err("\told bud LEB %d\n", bud->lnum);
+       list_for_each_entry(idx_gc, &c->idx_gc, list)
+               pr_err("\tGC'ed idx LEB %d unmap %d\n",
+                      idx_gc->lnum, idx_gc->unmap);
+       pr_err("\tcommit state %d\n", c->cmt_state);
+
+       /* Print budgeting predictions */
+       available = ubifs_calc_available(c, c->bi.min_idx_lebs);
+       outstanding = c->bi.data_growth + c->bi.dd_growth;
+       free = ubifs_get_free_space_nolock(c);
+       pr_err("Budgeting predictions:\n");
+       pr_err("\tavailable: %lld, outstanding %lld, free %lld\n",
+              available, outstanding, free);
+out_unlock:
+       spin_unlock(&dbg_lock);
+       spin_unlock(&c->space_lock);
+}
+#else
+void ubifs_dump_budg(struct ubifs_info *c, const struct ubifs_budg_info *bi)
+{
+}
+#endif
+
+void ubifs_dump_lprop(const struct ubifs_info *c, const struct ubifs_lprops *lp)
+{
+       int i, spc, dark = 0, dead = 0;
+       struct rb_node *rb;
+       struct ubifs_bud *bud;
+
+       spc = lp->free + lp->dirty;
+       if (spc < c->dead_wm)
+               dead = spc;
+       else
+               dark = ubifs_calc_dark(c, spc);
+
+       if (lp->flags & LPROPS_INDEX)
+               pr_err("LEB %-7d free %-8d dirty %-8d used %-8d free + dirty %-8d flags %#x (",
+                      lp->lnum, lp->free, lp->dirty, c->leb_size - spc, spc,
+                      lp->flags);
+       else
+               pr_err("LEB %-7d free %-8d dirty %-8d used %-8d free + dirty %-8d dark %-4d dead %-4d nodes fit %-3d flags %#-4x (",
+                      lp->lnum, lp->free, lp->dirty, c->leb_size - spc, spc,
+                      dark, dead, (int)(spc / UBIFS_MAX_NODE_SZ), lp->flags);
+
+       if (lp->flags & LPROPS_TAKEN) {
+               if (lp->flags & LPROPS_INDEX)
+                       pr_cont("index, taken");
+               else
+                       pr_cont("taken");
+       } else {
+               const char *s;
+
+               if (lp->flags & LPROPS_INDEX) {
+                       switch (lp->flags & LPROPS_CAT_MASK) {
+                       case LPROPS_DIRTY_IDX:
+                               s = "dirty index";
+                               break;
+                       case LPROPS_FRDI_IDX:
+                               s = "freeable index";
+                               break;
+                       default:
+                               s = "index";
+                       }
+               } else {
+                       switch (lp->flags & LPROPS_CAT_MASK) {
+                       case LPROPS_UNCAT:
+                               s = "not categorized";
+                               break;
+                       case LPROPS_DIRTY:
+                               s = "dirty";
+                               break;
+                       case LPROPS_FREE:
+                               s = "free";
+                               break;
+                       case LPROPS_EMPTY:
+                               s = "empty";
+                               break;
+                       case LPROPS_FREEABLE:
+                               s = "freeable";
+                               break;
+                       default:
+                               s = NULL;
+                               break;
+                       }
+               }
+               pr_cont("%s", s);
+       }
+
+       for (rb = rb_first((struct rb_root *)&c->buds); rb; rb = rb_next(rb)) {
+               bud = rb_entry(rb, struct ubifs_bud, rb);
+               if (bud->lnum == lp->lnum) {
+                       int head = 0;
+                       for (i = 0; i < c->jhead_cnt; i++) {
+                               /*
+                                * Note, if we are in R/O mode or in the middle
+                                * of mounting/re-mounting, the write-buffers do
+                                * not exist.
+                                */
+                               if (c->jheads &&
+                                   lp->lnum == c->jheads[i].wbuf.lnum) {
+                                       pr_cont(", jhead %s", dbg_jhead(i));
+                                       head = 1;
+                               }
+                       }
+                       if (!head)
+                               pr_cont(", bud of jhead %s",
+                                      dbg_jhead(bud->jhead));
+               }
+       }
+       if (lp->lnum == c->gc_lnum)
+               pr_cont(", GC LEB");
+       pr_cont(")\n");
+}
+
+void ubifs_dump_lprops(struct ubifs_info *c)
+{
+       int lnum, err;
+       struct ubifs_lprops lp;
+       struct ubifs_lp_stats lst;
+
+       pr_err("(pid %d) start dumping LEB properties\n", current->pid);
+       ubifs_get_lp_stats(c, &lst);
+       ubifs_dump_lstats(&lst);
+
+       for (lnum = c->main_first; lnum < c->leb_cnt; lnum++) {
+               err = ubifs_read_one_lp(c, lnum, &lp);
+               if (err)
+                       ubifs_err("cannot read lprops for LEB %d", lnum);
+
+               ubifs_dump_lprop(c, &lp);
+       }
+       pr_err("(pid %d) finish dumping LEB properties\n", current->pid);
 }
 
-const char *dbg_key_str0(const struct ubifs_info *c, const union ubifs_key *key)
+void ubifs_dump_lpt_info(struct ubifs_info *c)
 {
-       /* dbg_lock must be held */
-       sprintf_key(c, key, dbg_key_buf0);
-       return dbg_key_buf0;
+       int i;
+
+       spin_lock(&dbg_lock);
+       pr_err("(pid %d) dumping LPT information\n", current->pid);
+       pr_err("\tlpt_sz:        %lld\n", c->lpt_sz);
+       pr_err("\tpnode_sz:      %d\n", c->pnode_sz);
+       pr_err("\tnnode_sz:      %d\n", c->nnode_sz);
+       pr_err("\tltab_sz:       %d\n", c->ltab_sz);
+       pr_err("\tlsave_sz:      %d\n", c->lsave_sz);
+       pr_err("\tbig_lpt:       %d\n", c->big_lpt);
+       pr_err("\tlpt_hght:      %d\n", c->lpt_hght);
+       pr_err("\tpnode_cnt:     %d\n", c->pnode_cnt);
+       pr_err("\tnnode_cnt:     %d\n", c->nnode_cnt);
+       pr_err("\tdirty_pn_cnt:  %d\n", c->dirty_pn_cnt);
+       pr_err("\tdirty_nn_cnt:  %d\n", c->dirty_nn_cnt);
+       pr_err("\tlsave_cnt:     %d\n", c->lsave_cnt);
+       pr_err("\tspace_bits:    %d\n", c->space_bits);
+       pr_err("\tlpt_lnum_bits: %d\n", c->lpt_lnum_bits);
+       pr_err("\tlpt_offs_bits: %d\n", c->lpt_offs_bits);
+       pr_err("\tlpt_spc_bits:  %d\n", c->lpt_spc_bits);
+       pr_err("\tpcnt_bits:     %d\n", c->pcnt_bits);
+       pr_err("\tlnum_bits:     %d\n", c->lnum_bits);
+       pr_err("\tLPT root is at %d:%d\n", c->lpt_lnum, c->lpt_offs);
+       pr_err("\tLPT head is at %d:%d\n",
+              c->nhead_lnum, c->nhead_offs);
+       pr_err("\tLPT ltab is at %d:%d\n", c->ltab_lnum, c->ltab_offs);
+       if (c->big_lpt)
+               pr_err("\tLPT lsave is at %d:%d\n",
+                      c->lsave_lnum, c->lsave_offs);
+       for (i = 0; i < c->lpt_lebs; i++)
+               pr_err("\tLPT LEB %d free %d dirty %d tgc %d cmt %d\n",
+                      i + c->lpt_first, c->ltab[i].free, c->ltab[i].dirty,
+                      c->ltab[i].tgc, c->ltab[i].cmt);
+       spin_unlock(&dbg_lock);
+}
+
+void ubifs_dump_sleb(const struct ubifs_info *c,
+                    const struct ubifs_scan_leb *sleb, int offs)
+{
+       struct ubifs_scan_node *snod;
+
+       pr_err("(pid %d) start dumping scanned data from LEB %d:%d\n",
+              current->pid, sleb->lnum, offs);
+
+       list_for_each_entry(snod, &sleb->nodes, list) {
+               cond_resched();
+               pr_err("Dumping node at LEB %d:%d len %d\n",
+                      sleb->lnum, snod->offs, snod->len);
+               ubifs_dump_node(c, snod->node);
+       }
+}
+
+void ubifs_dump_leb(const struct ubifs_info *c, int lnum)
+{
+       struct ubifs_scan_leb *sleb;
+       struct ubifs_scan_node *snod;
+       void *buf;
+
+       pr_err("(pid %d) start dumping LEB %d\n", current->pid, lnum);
+
+       buf = __vmalloc(c->leb_size, GFP_NOFS, PAGE_KERNEL);
+       if (!buf) {
+               ubifs_err("cannot allocate memory for dumping LEB %d", lnum);
+               return;
+       }
+
+       sleb = ubifs_scan(c, lnum, 0, buf, 0);
+       if (IS_ERR(sleb)) {
+               ubifs_err("scan error %d", (int)PTR_ERR(sleb));
+               goto out;
+       }
+
+       pr_err("LEB %d has %d nodes ending at %d\n", lnum,
+              sleb->nodes_cnt, sleb->endpt);
+
+       list_for_each_entry(snod, &sleb->nodes, list) {
+               cond_resched();
+               pr_err("Dumping node at LEB %d:%d len %d\n", lnum,
+                      snod->offs, snod->len);
+               ubifs_dump_node(c, snod->node);
+       }
+
+       pr_err("(pid %d) finish dumping LEB %d\n", current->pid, lnum);
+       ubifs_scan_destroy(sleb);
+
+out:
+       vfree(buf);
+       return;
+}
+
+void ubifs_dump_znode(const struct ubifs_info *c,
+                     const struct ubifs_znode *znode)
+{
+       int n;
+       const struct ubifs_zbranch *zbr;
+       char key_buf[DBG_KEY_BUF_LEN];
+
+       spin_lock(&dbg_lock);
+       if (znode->parent)
+               zbr = &znode->parent->zbranch[znode->iip];
+       else
+               zbr = &c->zroot;
+
+       pr_err("znode %p, LEB %d:%d len %d parent %p iip %d level %d child_cnt %d flags %lx\n",
+              znode, zbr->lnum, zbr->offs, zbr->len, znode->parent, znode->iip,
+              znode->level, znode->child_cnt, znode->flags);
+
+       if (znode->child_cnt <= 0 || znode->child_cnt > c->fanout) {
+               spin_unlock(&dbg_lock);
+               return;
+       }
+
+       pr_err("zbranches:\n");
+       for (n = 0; n < znode->child_cnt; n++) {
+               zbr = &znode->zbranch[n];
+               if (znode->level > 0)
+                       pr_err("\t%d: znode %p LEB %d:%d len %d key %s\n",
+                              n, zbr->znode, zbr->lnum, zbr->offs, zbr->len,
+                              dbg_snprintf_key(c, &zbr->key, key_buf,
+                                               DBG_KEY_BUF_LEN));
+               else
+                       pr_err("\t%d: LNC %p LEB %d:%d len %d key %s\n",
+                              n, zbr->znode, zbr->lnum, zbr->offs, zbr->len,
+                              dbg_snprintf_key(c, &zbr->key, key_buf,
+                                               DBG_KEY_BUF_LEN));
+       }
+       spin_unlock(&dbg_lock);
+}
+
+void ubifs_dump_heap(struct ubifs_info *c, struct ubifs_lpt_heap *heap, int cat)
+{
+       int i;
+
+       pr_err("(pid %d) start dumping heap cat %d (%d elements)\n",
+              current->pid, cat, heap->cnt);
+       for (i = 0; i < heap->cnt; i++) {
+               struct ubifs_lprops *lprops = heap->arr[i];
+
+               pr_err("\t%d. LEB %d hpos %d free %d dirty %d flags %d\n",
+                      i, lprops->lnum, lprops->hpos, lprops->free,
+                      lprops->dirty, lprops->flags);
+       }
+       pr_err("(pid %d) finish dumping heap\n", current->pid);
+}
+
+void ubifs_dump_pnode(struct ubifs_info *c, struct ubifs_pnode *pnode,
+                     struct ubifs_nnode *parent, int iip)
+{
+       int i;
+
+       pr_err("(pid %d) dumping pnode:\n", current->pid);
+       pr_err("\taddress %zx parent %zx cnext %zx\n",
+              (size_t)pnode, (size_t)parent, (size_t)pnode->cnext);
+       pr_err("\tflags %lu iip %d level %d num %d\n",
+              pnode->flags, iip, pnode->level, pnode->num);
+       for (i = 0; i < UBIFS_LPT_FANOUT; i++) {
+               struct ubifs_lprops *lp = &pnode->lprops[i];
+
+               pr_err("\t%d: free %d dirty %d flags %d lnum %d\n",
+                      i, lp->free, lp->dirty, lp->flags, lp->lnum);
+       }
+}
+
+void ubifs_dump_tnc(struct ubifs_info *c)
+{
+       struct ubifs_znode *znode;
+       int level;
+
+       pr_err("\n");
+       pr_err("(pid %d) start dumping TNC tree\n", current->pid);
+       znode = ubifs_tnc_levelorder_next(c->zroot.znode, NULL);
+       level = znode->level;
+       pr_err("== Level %d ==\n", level);
+       while (znode) {
+               if (level != znode->level) {
+                       level = znode->level;
+                       pr_err("== Level %d ==\n", level);
+               }
+               ubifs_dump_znode(c, znode);
+               znode = ubifs_tnc_levelorder_next(c->zroot.znode, znode);
+       }
+       pr_err("(pid %d) finish dumping TNC tree\n", current->pid);
 }
 
-const char *dbg_key_str1(const struct ubifs_info *c, const union ubifs_key *key)
+static int dump_znode(struct ubifs_info *c, struct ubifs_znode *znode,
+                     void *priv)
 {
-       /* dbg_lock must be held */
-       sprintf_key(c, key, dbg_key_buf1);
-       return dbg_key_buf1;
+       ubifs_dump_znode(c, znode);
+       return 0;
 }
 
 /**
- * ubifs_debugging_init - initialize UBIFS debugging.
+ * ubifs_dump_index - dump the on-flash index.
  * @c: UBIFS file-system description object
  *
- * This function initializes debugging-related data for the file system.
- * Returns zero in case of success and a negative error code in case of
- * failure.
+ * This function dumps whole UBIFS indexing B-tree, unlike 'ubifs_dump_tnc()'
+ * which dumps only in-memory znodes and does not read znodes which from flash.
  */
-int ubifs_debugging_init(struct ubifs_info *c)
+void ubifs_dump_index(struct ubifs_info *c)
 {
-       c->dbg = kzalloc(sizeof(struct ubifs_debug_info), GFP_KERNEL);
-       if (!c->dbg)
-               return -ENOMEM;
+       dbg_walk_index(c, NULL, dump_znode, NULL);
+}
+
+#ifndef __UBOOT__
+/**
+ * dbg_save_space_info - save information about flash space.
+ * @c: UBIFS file-system description object
+ *
+ * This function saves information about UBIFS free space, dirty space, etc, in
+ * order to check it later.
+ */
+void dbg_save_space_info(struct ubifs_info *c)
+{
+       struct ubifs_debug_info *d = c->dbg;
+       int freeable_cnt;
+
+       spin_lock(&c->space_lock);
+       memcpy(&d->saved_lst, &c->lst, sizeof(struct ubifs_lp_stats));
+       memcpy(&d->saved_bi, &c->bi, sizeof(struct ubifs_budg_info));
+       d->saved_idx_gc_cnt = c->idx_gc_cnt;
+
+       /*
+        * We use a dirty hack here and zero out @c->freeable_cnt, because it
+        * affects the free space calculations, and UBIFS might not know about
+        * all freeable eraseblocks. Indeed, we know about freeable eraseblocks
+        * only when we read their lprops, and we do this only lazily, upon the
+        * need. So at any given point of time @c->freeable_cnt might be not
+        * exactly accurate.
+        *
+        * Just one example about the issue we hit when we did not zero
+        * @c->freeable_cnt.
+        * 1. The file-system is mounted R/O, c->freeable_cnt is %0. We save the
+        *    amount of free space in @d->saved_free
+        * 2. We re-mount R/W, which makes UBIFS to read the "lsave"
+        *    information from flash, where we cache LEBs from various
+        *    categories ('ubifs_remount_fs()' -> 'ubifs_lpt_init()'
+        *    -> 'lpt_init_wr()' -> 'read_lsave()' -> 'ubifs_lpt_lookup()'
+        *    -> 'ubifs_get_pnode()' -> 'update_cats()'
+        *    -> 'ubifs_add_to_cat()').
+        * 3. Lsave contains a freeable eraseblock, and @c->freeable_cnt
+        *    becomes %1.
+        * 4. We calculate the amount of free space when the re-mount is
+        *    finished in 'dbg_check_space_info()' and it does not match
+        *    @d->saved_free.
+        */
+       freeable_cnt = c->freeable_cnt;
+       c->freeable_cnt = 0;
+       d->saved_free = ubifs_get_free_space_nolock(c);
+       c->freeable_cnt = freeable_cnt;
+       spin_unlock(&c->space_lock);
+}
+
+/**
+ * dbg_check_space_info - check flash space information.
+ * @c: UBIFS file-system description object
+ *
+ * This function compares current flash space information with the information
+ * which was saved when the 'dbg_save_space_info()' function was called.
+ * Returns zero if the information has not changed, and %-EINVAL it it has
+ * changed.
+ */
+int dbg_check_space_info(struct ubifs_info *c)
+{
+       struct ubifs_debug_info *d = c->dbg;
+       struct ubifs_lp_stats lst;
+       long long free;
+       int freeable_cnt;
 
-       c->dbg->buf = vmalloc(c->leb_size);
-       if (!c->dbg->buf)
+       spin_lock(&c->space_lock);
+       freeable_cnt = c->freeable_cnt;
+       c->freeable_cnt = 0;
+       free = ubifs_get_free_space_nolock(c);
+       c->freeable_cnt = freeable_cnt;
+       spin_unlock(&c->space_lock);
+
+       if (free != d->saved_free) {
+               ubifs_err("free space changed from %lld to %lld",
+                         d->saved_free, free);
                goto out;
+       }
 
        return 0;
 
 out:
-       kfree(c->dbg);
-       return -ENOMEM;
+       ubifs_msg("saved lprops statistics dump");
+       ubifs_dump_lstats(&d->saved_lst);
+       ubifs_msg("saved budgeting info dump");
+       ubifs_dump_budg(c, &d->saved_bi);
+       ubifs_msg("saved idx_gc_cnt %d", d->saved_idx_gc_cnt);
+       ubifs_msg("current lprops statistics dump");
+       ubifs_get_lp_stats(c, &lst);
+       ubifs_dump_lstats(&lst);
+       ubifs_msg("current budgeting info dump");
+       ubifs_dump_budg(c, &c->bi);
+       dump_stack();
+       return -EINVAL;
 }
 
 /**
- * ubifs_debugging_exit - free debugging data.
+ * dbg_check_synced_i_size - check synchronized inode size.
  * @c: UBIFS file-system description object
+ * @inode: inode to check
+ *
+ * If inode is clean, synchronized inode size has to be equivalent to current
+ * inode size. This function has to be called only for locked inodes (@i_mutex
+ * has to be locked). Returns %0 if synchronized inode size if correct, and
+ * %-EINVAL if not.
  */
-void ubifs_debugging_exit(struct ubifs_info *c)
+int dbg_check_synced_i_size(const struct ubifs_info *c, struct inode *inode)
 {
-       vfree(c->dbg->buf);
-       kfree(c->dbg);
+       int err = 0;
+       struct ubifs_inode *ui = ubifs_inode(inode);
+
+       if (!dbg_is_chk_gen(c))
+               return 0;
+       if (!S_ISREG(inode->i_mode))
+               return 0;
+
+       mutex_lock(&ui->ui_mutex);
+       spin_lock(&ui->ui_lock);
+       if (ui->ui_size != ui->synced_i_size && !ui->dirty) {
+               ubifs_err("ui_size is %lld, synced_i_size is %lld, but inode is clean",
+                         ui->ui_size, ui->synced_i_size);
+               ubifs_err("i_ino %lu, i_mode %#x, i_size %lld", inode->i_ino,
+                         inode->i_mode, i_size_read(inode));
+               dump_stack();
+               err = -EINVAL;
+       }
+       spin_unlock(&ui->ui_lock);
+       mutex_unlock(&ui->ui_mutex);
+       return err;
+}
+
+/*
+ * dbg_check_dir - check directory inode size and link count.
+ * @c: UBIFS file-system description object
+ * @dir: the directory to calculate size for
+ * @size: the result is returned here
+ *
+ * This function makes sure that directory size and link count are correct.
+ * Returns zero in case of success and a negative error code in case of
+ * failure.
+ *
+ * Note, it is good idea to make sure the @dir->i_mutex is locked before
+ * calling this function.
+ */
+int dbg_check_dir(struct ubifs_info *c, const struct inode *dir)
+{
+       unsigned int nlink = 2;
+       union ubifs_key key;
+       struct ubifs_dent_node *dent, *pdent = NULL;
+       struct qstr nm = { .name = NULL };
+       loff_t size = UBIFS_INO_NODE_SZ;
+
+       if (!dbg_is_chk_gen(c))
+               return 0;
+
+       if (!S_ISDIR(dir->i_mode))
+               return 0;
+
+       lowest_dent_key(c, &key, dir->i_ino);
+       while (1) {
+               int err;
+
+               dent = ubifs_tnc_next_ent(c, &key, &nm);
+               if (IS_ERR(dent)) {
+                       err = PTR_ERR(dent);
+                       if (err == -ENOENT)
+                               break;
+                       return err;
+               }
+
+               nm.name = dent->name;
+               nm.len = le16_to_cpu(dent->nlen);
+               size += CALC_DENT_SIZE(nm.len);
+               if (dent->type == UBIFS_ITYPE_DIR)
+                       nlink += 1;
+               kfree(pdent);
+               pdent = dent;
+               key_read(c, &dent->key, &key);
+       }
+       kfree(pdent);
+
+       if (i_size_read(dir) != size) {
+               ubifs_err("directory inode %lu has size %llu, but calculated size is %llu",
+                         dir->i_ino, (unsigned long long)i_size_read(dir),
+                         (unsigned long long)size);
+               ubifs_dump_inode(c, dir);
+               dump_stack();
+               return -EINVAL;
+       }
+       if (dir->i_nlink != nlink) {
+               ubifs_err("directory inode %lu has nlink %u, but calculated nlink is %u",
+                         dir->i_ino, dir->i_nlink, nlink);
+               ubifs_dump_inode(c, dir);
+               dump_stack();
+               return -EINVAL;
+       }
+
+       return 0;
+}
+
+/**
+ * dbg_check_key_order - make sure that colliding keys are properly ordered.
+ * @c: UBIFS file-system description object
+ * @zbr1: first zbranch
+ * @zbr2: following zbranch
+ *
+ * In UBIFS indexing B-tree colliding keys has to be sorted in binary order of
+ * names of the direntries/xentries which are referred by the keys. This
+ * function reads direntries/xentries referred by @zbr1 and @zbr2 and makes
+ * sure the name of direntry/xentry referred by @zbr1 is less than
+ * direntry/xentry referred by @zbr2. Returns zero if this is true, %1 if not,
+ * and a negative error code in case of failure.
+ */
+static int dbg_check_key_order(struct ubifs_info *c, struct ubifs_zbranch *zbr1,
+                              struct ubifs_zbranch *zbr2)
+{
+       int err, nlen1, nlen2, cmp;
+       struct ubifs_dent_node *dent1, *dent2;
+       union ubifs_key key;
+       char key_buf[DBG_KEY_BUF_LEN];
+
+       ubifs_assert(!keys_cmp(c, &zbr1->key, &zbr2->key));
+       dent1 = kmalloc(UBIFS_MAX_DENT_NODE_SZ, GFP_NOFS);
+       if (!dent1)
+               return -ENOMEM;
+       dent2 = kmalloc(UBIFS_MAX_DENT_NODE_SZ, GFP_NOFS);
+       if (!dent2) {
+               err = -ENOMEM;
+               goto out_free;
+       }
+
+       err = ubifs_tnc_read_node(c, zbr1, dent1);
+       if (err)
+               goto out_free;
+       err = ubifs_validate_entry(c, dent1);
+       if (err)
+               goto out_free;
+
+       err = ubifs_tnc_read_node(c, zbr2, dent2);
+       if (err)
+               goto out_free;
+       err = ubifs_validate_entry(c, dent2);
+       if (err)
+               goto out_free;
+
+       /* Make sure node keys are the same as in zbranch */
+       err = 1;
+       key_read(c, &dent1->key, &key);
+       if (keys_cmp(c, &zbr1->key, &key)) {
+               ubifs_err("1st entry at %d:%d has key %s", zbr1->lnum,
+                         zbr1->offs, dbg_snprintf_key(c, &key, key_buf,
+                                                      DBG_KEY_BUF_LEN));
+               ubifs_err("but it should have key %s according to tnc",
+                         dbg_snprintf_key(c, &zbr1->key, key_buf,
+                                          DBG_KEY_BUF_LEN));
+               ubifs_dump_node(c, dent1);
+               goto out_free;
+       }
+
+       key_read(c, &dent2->key, &key);
+       if (keys_cmp(c, &zbr2->key, &key)) {
+               ubifs_err("2nd entry at %d:%d has key %s", zbr1->lnum,
+                         zbr1->offs, dbg_snprintf_key(c, &key, key_buf,
+                                                      DBG_KEY_BUF_LEN));
+               ubifs_err("but it should have key %s according to tnc",
+                         dbg_snprintf_key(c, &zbr2->key, key_buf,
+                                          DBG_KEY_BUF_LEN));
+               ubifs_dump_node(c, dent2);
+               goto out_free;
+       }
+
+       nlen1 = le16_to_cpu(dent1->nlen);
+       nlen2 = le16_to_cpu(dent2->nlen);
+
+       cmp = memcmp(dent1->name, dent2->name, min_t(int, nlen1, nlen2));
+       if (cmp < 0 || (cmp == 0 && nlen1 < nlen2)) {
+               err = 0;
+               goto out_free;
+       }
+       if (cmp == 0 && nlen1 == nlen2)
+               ubifs_err("2 xent/dent nodes with the same name");
+       else
+               ubifs_err("bad order of colliding key %s",
+                         dbg_snprintf_key(c, &key, key_buf, DBG_KEY_BUF_LEN));
+
+       ubifs_msg("first node at %d:%d\n", zbr1->lnum, zbr1->offs);
+       ubifs_dump_node(c, dent1);
+       ubifs_msg("second node at %d:%d\n", zbr2->lnum, zbr2->offs);
+       ubifs_dump_node(c, dent2);
+
+out_free:
+       kfree(dent2);
+       kfree(dent1);
+       return err;
 }
 
-#endif /* CONFIG_UBIFS_FS_DEBUG */
+/**
+ * dbg_check_znode - check if znode is all right.
+ * @c: UBIFS file-system description object
+ * @zbr: zbranch which points to this znode
+ *
+ * This function makes sure that znode referred to by @zbr is all right.
+ * Returns zero if it is, and %-EINVAL if it is not.
+ */
+static int dbg_check_znode(struct ubifs_info *c, struct ubifs_zbranch *zbr)
+{
+       struct ubifs_znode *znode = zbr->znode;
+       struct ubifs_znode *zp = znode->parent;
+       int n, err, cmp;
+
+       if (znode->child_cnt <= 0 || znode->child_cnt > c->fanout) {
+               err = 1;
+               goto out;
+       }
+       if (znode->level < 0) {
+               err = 2;
+               goto out;
+       }
+       if (znode->iip < 0 || znode->iip >= c->fanout) {
+               err = 3;
+               goto out;
+       }
+
+       if (zbr->len == 0)
+               /* Only dirty zbranch may have no on-flash nodes */
+               if (!ubifs_zn_dirty(znode)) {
+                       err = 4;
+                       goto out;
+               }
+
+       if (ubifs_zn_dirty(znode)) {
+               /*
+                * If znode is dirty, its parent has to be dirty as well. The
+                * order of the operation is important, so we have to have
+                * memory barriers.
+                */
+               smp_mb();
+               if (zp && !ubifs_zn_dirty(zp)) {
+                       /*
+                        * The dirty flag is atomic and is cleared outside the
+                        * TNC mutex, so znode's dirty flag may now have
+                        * been cleared. The child is always cleared before the
+                        * parent, so we just need to check again.
+                        */
+                       smp_mb();
+                       if (ubifs_zn_dirty(znode)) {
+                               err = 5;
+                               goto out;
+                       }
+               }
+       }
+
+       if (zp) {
+               const union ubifs_key *min, *max;
+
+               if (znode->level != zp->level - 1) {
+                       err = 6;
+                       goto out;
+               }
+
+               /* Make sure the 'parent' pointer in our znode is correct */
+               err = ubifs_search_zbranch(c, zp, &zbr->key, &n);
+               if (!err) {
+                       /* This zbranch does not exist in the parent */
+                       err = 7;
+                       goto out;
+               }
+
+               if (znode->iip >= zp->child_cnt) {
+                       err = 8;
+                       goto out;
+               }
+
+               if (znode->iip != n) {
+                       /* This may happen only in case of collisions */
+                       if (keys_cmp(c, &zp->zbranch[n].key,
+                                    &zp->zbranch[znode->iip].key)) {
+                               err = 9;
+                               goto out;
+                       }
+                       n = znode->iip;
+               }
+
+               /*
+                * Make sure that the first key in our znode is greater than or
+                * equal to the key in the pointing zbranch.
+                */
+               min = &zbr->key;
+               cmp = keys_cmp(c, min, &znode->zbranch[0].key);
+               if (cmp == 1) {
+                       err = 10;
+                       goto out;
+               }
+
+               if (n + 1 < zp->child_cnt) {
+                       max = &zp->zbranch[n + 1].key;
+
+                       /*
+                        * Make sure the last key in our znode is less or
+                        * equivalent than the key in the zbranch which goes
+                        * after our pointing zbranch.
+                        */
+                       cmp = keys_cmp(c, max,
+                               &znode->zbranch[znode->child_cnt - 1].key);
+                       if (cmp == -1) {
+                               err = 11;
+                               goto out;
+                       }
+               }
+       } else {
+               /* This may only be root znode */
+               if (zbr != &c->zroot) {
+                       err = 12;
+                       goto out;
+               }
+       }
+
+       /*
+        * Make sure that next key is greater or equivalent then the previous
+        * one.
+        */
+       for (n = 1; n < znode->child_cnt; n++) {
+               cmp = keys_cmp(c, &znode->zbranch[n - 1].key,
+                              &znode->zbranch[n].key);
+               if (cmp > 0) {
+                       err = 13;
+                       goto out;
+               }
+               if (cmp == 0) {
+                       /* This can only be keys with colliding hash */
+                       if (!is_hash_key(c, &znode->zbranch[n].key)) {
+                               err = 14;
+                               goto out;
+                       }
+
+                       if (znode->level != 0 || c->replaying)
+                               continue;
+
+                       /*
+                        * Colliding keys should follow binary order of
+                        * corresponding xentry/dentry names.
+                        */
+                       err = dbg_check_key_order(c, &znode->zbranch[n - 1],
+                                                 &znode->zbranch[n]);
+                       if (err < 0)
+                               return err;
+                       if (err) {
+                               err = 15;
+                               goto out;
+                       }
+               }
+       }
+
+       for (n = 0; n < znode->child_cnt; n++) {
+               if (!znode->zbranch[n].znode &&
+                   (znode->zbranch[n].lnum == 0 ||
+                    znode->zbranch[n].len == 0)) {
+                       err = 16;
+                       goto out;
+               }
+
+               if (znode->zbranch[n].lnum != 0 &&
+                   znode->zbranch[n].len == 0) {
+                       err = 17;
+                       goto out;
+               }
+
+               if (znode->zbranch[n].lnum == 0 &&
+                   znode->zbranch[n].len != 0) {
+                       err = 18;
+                       goto out;
+               }
+
+               if (znode->zbranch[n].lnum == 0 &&
+                   znode->zbranch[n].offs != 0) {
+                       err = 19;
+                       goto out;
+               }
+
+               if (znode->level != 0 && znode->zbranch[n].znode)
+                       if (znode->zbranch[n].znode->parent != znode) {
+                               err = 20;
+                               goto out;
+                       }
+       }
+
+       return 0;
+
+out:
+       ubifs_err("failed, error %d", err);
+       ubifs_msg("dump of the znode");
+       ubifs_dump_znode(c, znode);
+       if (zp) {
+               ubifs_msg("dump of the parent znode");
+               ubifs_dump_znode(c, zp);
+       }
+       dump_stack();
+       return -EINVAL;
+}
+#else
+
+int dbg_check_dir(struct ubifs_info *c, const struct inode *dir)
+{
+       return 0;
+}
+
+void dbg_debugfs_exit_fs(struct ubifs_info *c)
+{
+       return;
+}
+
+int ubifs_debugging_init(struct ubifs_info *c)
+{
+       return 0;
+}
+void ubifs_debugging_exit(struct ubifs_info *c)
+{
+}
+int dbg_check_filesystem(struct ubifs_info *c)
+{
+       return 0;
+}
+int dbg_debugfs_init_fs(struct ubifs_info *c)
+{
+       return 0;
+}
+#endif
+
+#ifndef __UBOOT__
+/**
+ * dbg_check_tnc - check TNC tree.
+ * @c: UBIFS file-system description object
+ * @extra: do extra checks that are possible at start commit
+ *
+ * This function traverses whole TNC tree and checks every znode. Returns zero
+ * if everything is all right and %-EINVAL if something is wrong with TNC.
+ */
+int dbg_check_tnc(struct ubifs_info *c, int extra)
+{
+       struct ubifs_znode *znode;
+       long clean_cnt = 0, dirty_cnt = 0;
+       int err, last;
+
+       if (!dbg_is_chk_index(c))
+               return 0;
+
+       ubifs_assert(mutex_is_locked(&c->tnc_mutex));
+       if (!c->zroot.znode)
+               return 0;
+
+       znode = ubifs_tnc_postorder_first(c->zroot.znode);
+       while (1) {
+               struct ubifs_znode *prev;
+               struct ubifs_zbranch *zbr;
+
+               if (!znode->parent)
+                       zbr = &c->zroot;
+               else
+                       zbr = &znode->parent->zbranch[znode->iip];
+
+               err = dbg_check_znode(c, zbr);
+               if (err)
+                       return err;
+
+               if (extra) {
+                       if (ubifs_zn_dirty(znode))
+                               dirty_cnt += 1;
+                       else
+                               clean_cnt += 1;
+               }
+
+               prev = znode;
+               znode = ubifs_tnc_postorder_next(znode);
+               if (!znode)
+                       break;
+
+               /*
+                * If the last key of this znode is equivalent to the first key
+                * of the next znode (collision), then check order of the keys.
+                */
+               last = prev->child_cnt - 1;
+               if (prev->level == 0 && znode->level == 0 && !c->replaying &&
+                   !keys_cmp(c, &prev->zbranch[last].key,
+                             &znode->zbranch[0].key)) {
+                       err = dbg_check_key_order(c, &prev->zbranch[last],
+                                                 &znode->zbranch[0]);
+                       if (err < 0)
+                               return err;
+                       if (err) {
+                               ubifs_msg("first znode");
+                               ubifs_dump_znode(c, prev);
+                               ubifs_msg("second znode");
+                               ubifs_dump_znode(c, znode);
+                               return -EINVAL;
+                       }
+               }
+       }
+
+       if (extra) {
+               if (clean_cnt != atomic_long_read(&c->clean_zn_cnt)) {
+                       ubifs_err("incorrect clean_zn_cnt %ld, calculated %ld",
+                                 atomic_long_read(&c->clean_zn_cnt),
+                                 clean_cnt);
+                       return -EINVAL;
+               }
+               if (dirty_cnt != atomic_long_read(&c->dirty_zn_cnt)) {
+                       ubifs_err("incorrect dirty_zn_cnt %ld, calculated %ld",
+                                 atomic_long_read(&c->dirty_zn_cnt),
+                                 dirty_cnt);
+                       return -EINVAL;
+               }
+       }
+
+       return 0;
+}
+#else
+int dbg_check_tnc(struct ubifs_info *c, int extra)
+{
+       return 0;
+}
+#endif
+
+/**
+ * dbg_walk_index - walk the on-flash index.
+ * @c: UBIFS file-system description object
+ * @leaf_cb: called for each leaf node
+ * @znode_cb: called for each indexing node
+ * @priv: private data which is passed to callbacks
+ *
+ * This function walks the UBIFS index and calls the @leaf_cb for each leaf
+ * node and @znode_cb for each indexing node. Returns zero in case of success
+ * and a negative error code in case of failure.
+ *
+ * It would be better if this function removed every znode it pulled to into
+ * the TNC, so that the behavior more closely matched the non-debugging
+ * behavior.
+ */
+int dbg_walk_index(struct ubifs_info *c, dbg_leaf_callback leaf_cb,
+                  dbg_znode_callback znode_cb, void *priv)
+{
+       int err;
+       struct ubifs_zbranch *zbr;
+       struct ubifs_znode *znode, *child;
+
+       mutex_lock(&c->tnc_mutex);
+       /* If the root indexing node is not in TNC - pull it */
+       if (!c->zroot.znode) {
+               c->zroot.znode = ubifs_load_znode(c, &c->zroot, NULL, 0);
+               if (IS_ERR(c->zroot.znode)) {
+                       err = PTR_ERR(c->zroot.znode);
+                       c->zroot.znode = NULL;
+                       goto out_unlock;
+               }
+       }
+
+       /*
+        * We are going to traverse the indexing tree in the postorder manner.
+        * Go down and find the leftmost indexing node where we are going to
+        * start from.
+        */
+       znode = c->zroot.znode;
+       while (znode->level > 0) {
+               zbr = &znode->zbranch[0];
+               child = zbr->znode;
+               if (!child) {
+                       child = ubifs_load_znode(c, zbr, znode, 0);
+                       if (IS_ERR(child)) {
+                               err = PTR_ERR(child);
+                               goto out_unlock;
+                       }
+                       zbr->znode = child;
+               }
+
+               znode = child;
+       }
+
+       /* Iterate over all indexing nodes */
+       while (1) {
+               int idx;
+
+               cond_resched();
+
+               if (znode_cb) {
+                       err = znode_cb(c, znode, priv);
+                       if (err) {
+                               ubifs_err("znode checking function returned error %d",
+                                         err);
+                               ubifs_dump_znode(c, znode);
+                               goto out_dump;
+                       }
+               }
+               if (leaf_cb && znode->level == 0) {
+                       for (idx = 0; idx < znode->child_cnt; idx++) {
+                               zbr = &znode->zbranch[idx];
+                               err = leaf_cb(c, zbr, priv);
+                               if (err) {
+                                       ubifs_err("leaf checking function returned error %d, for leaf at LEB %d:%d",
+                                                 err, zbr->lnum, zbr->offs);
+                                       goto out_dump;
+                               }
+                       }
+               }
+
+               if (!znode->parent)
+                       break;
+
+               idx = znode->iip + 1;
+               znode = znode->parent;
+               if (idx < znode->child_cnt) {
+                       /* Switch to the next index in the parent */
+                       zbr = &znode->zbranch[idx];
+                       child = zbr->znode;
+                       if (!child) {
+                               child = ubifs_load_znode(c, zbr, znode, idx);
+                               if (IS_ERR(child)) {
+                                       err = PTR_ERR(child);
+                                       goto out_unlock;
+                               }
+                               zbr->znode = child;
+                       }
+                       znode = child;
+               } else
+                       /*
+                        * This is the last child, switch to the parent and
+                        * continue.
+                        */
+                       continue;
+
+               /* Go to the lowest leftmost znode in the new sub-tree */
+               while (znode->level > 0) {
+                       zbr = &znode->zbranch[0];
+                       child = zbr->znode;
+                       if (!child) {
+                               child = ubifs_load_znode(c, zbr, znode, 0);
+                               if (IS_ERR(child)) {
+                                       err = PTR_ERR(child);
+                                       goto out_unlock;
+                               }
+                               zbr->znode = child;
+                       }
+                       znode = child;
+               }
+       }
+
+       mutex_unlock(&c->tnc_mutex);
+       return 0;
+
+out_dump:
+       if (znode->parent)
+               zbr = &znode->parent->zbranch[znode->iip];
+       else
+               zbr = &c->zroot;
+       ubifs_msg("dump of znode at LEB %d:%d", zbr->lnum, zbr->offs);
+       ubifs_dump_znode(c, znode);
+out_unlock:
+       mutex_unlock(&c->tnc_mutex);
+       return err;
+}
+
+/**
+ * add_size - add znode size to partially calculated index size.
+ * @c: UBIFS file-system description object
+ * @znode: znode to add size for
+ * @priv: partially calculated index size
+ *
+ * This is a helper function for 'dbg_check_idx_size()' which is called for
+ * every indexing node and adds its size to the 'long long' variable pointed to
+ * by @priv.
+ */
+static int add_size(struct ubifs_info *c, struct ubifs_znode *znode, void *priv)
+{
+       long long *idx_size = priv;
+       int add;
+
+       add = ubifs_idx_node_sz(c, znode->child_cnt);
+       add = ALIGN(add, 8);
+       *idx_size += add;
+       return 0;
+}
+
+/**
+ * dbg_check_idx_size - check index size.
+ * @c: UBIFS file-system description object
+ * @idx_size: size to check
+ *
+ * This function walks the UBIFS index, calculates its size and checks that the
+ * size is equivalent to @idx_size. Returns zero in case of success and a
+ * negative error code in case of failure.
+ */
+int dbg_check_idx_size(struct ubifs_info *c, long long idx_size)
+{
+       int err;
+       long long calc = 0;
+
+       if (!dbg_is_chk_index(c))
+               return 0;
+
+       err = dbg_walk_index(c, NULL, add_size, &calc);
+       if (err) {
+               ubifs_err("error %d while walking the index", err);
+               return err;
+       }
+
+       if (calc != idx_size) {
+               ubifs_err("index size check failed: calculated size is %lld, should be %lld",
+                         calc, idx_size);
+               dump_stack();
+               return -EINVAL;
+       }
+
+       return 0;
+}
+
+#ifndef __UBOOT__
+/**
+ * struct fsck_inode - information about an inode used when checking the file-system.
+ * @rb: link in the RB-tree of inodes
+ * @inum: inode number
+ * @mode: inode type, permissions, etc
+ * @nlink: inode link count
+ * @xattr_cnt: count of extended attributes
+ * @references: how many directory/xattr entries refer this inode (calculated
+ *              while walking the index)
+ * @calc_cnt: for directory inode count of child directories
+ * @size: inode size (read from on-flash inode)
+ * @xattr_sz: summary size of all extended attributes (read from on-flash
+ *            inode)
+ * @calc_sz: for directories calculated directory size
+ * @calc_xcnt: count of extended attributes
+ * @calc_xsz: calculated summary size of all extended attributes
+ * @xattr_nms: sum of lengths of all extended attribute names belonging to this
+ *             inode (read from on-flash inode)
+ * @calc_xnms: calculated sum of lengths of all extended attribute names
+ */
+struct fsck_inode {
+       struct rb_node rb;
+       ino_t inum;
+       umode_t mode;
+       unsigned int nlink;
+       unsigned int xattr_cnt;
+       int references;
+       int calc_cnt;
+       long long size;
+       unsigned int xattr_sz;
+       long long calc_sz;
+       long long calc_xcnt;
+       long long calc_xsz;
+       unsigned int xattr_nms;
+       long long calc_xnms;
+};
+
+/**
+ * struct fsck_data - private FS checking information.
+ * @inodes: RB-tree of all inodes (contains @struct fsck_inode objects)
+ */
+struct fsck_data {
+       struct rb_root inodes;
+};
+
+/**
+ * add_inode - add inode information to RB-tree of inodes.
+ * @c: UBIFS file-system description object
+ * @fsckd: FS checking information
+ * @ino: raw UBIFS inode to add
+ *
+ * This is a helper function for 'check_leaf()' which adds information about
+ * inode @ino to the RB-tree of inodes. Returns inode information pointer in
+ * case of success and a negative error code in case of failure.
+ */
+static struct fsck_inode *add_inode(struct ubifs_info *c,
+                                   struct fsck_data *fsckd,
+                                   struct ubifs_ino_node *ino)
+{
+       struct rb_node **p, *parent = NULL;
+       struct fsck_inode *fscki;
+       ino_t inum = key_inum_flash(c, &ino->key);
+       struct inode *inode;
+       struct ubifs_inode *ui;
+
+       p = &fsckd->inodes.rb_node;
+       while (*p) {
+               parent = *p;
+               fscki = rb_entry(parent, struct fsck_inode, rb);
+               if (inum < fscki->inum)
+                       p = &(*p)->rb_left;
+               else if (inum > fscki->inum)
+                       p = &(*p)->rb_right;
+               else
+                       return fscki;
+       }
+
+       if (inum > c->highest_inum) {
+               ubifs_err("too high inode number, max. is %lu",
+                         (unsigned long)c->highest_inum);
+               return ERR_PTR(-EINVAL);
+       }
+
+       fscki = kzalloc(sizeof(struct fsck_inode), GFP_NOFS);
+       if (!fscki)
+               return ERR_PTR(-ENOMEM);
+
+       inode = ilookup(c->vfs_sb, inum);
+
+       fscki->inum = inum;
+       /*
+        * If the inode is present in the VFS inode cache, use it instead of
+        * the on-flash inode which might be out-of-date. E.g., the size might
+        * be out-of-date. If we do not do this, the following may happen, for
+        * example:
+        *   1. A power cut happens
+        *   2. We mount the file-system R/O, the replay process fixes up the
+        *      inode size in the VFS cache, but on on-flash.
+        *   3. 'check_leaf()' fails because it hits a data node beyond inode
+        *      size.
+        */
+       if (!inode) {
+               fscki->nlink = le32_to_cpu(ino->nlink);
+               fscki->size = le64_to_cpu(ino->size);
+               fscki->xattr_cnt = le32_to_cpu(ino->xattr_cnt);
+               fscki->xattr_sz = le32_to_cpu(ino->xattr_size);
+               fscki->xattr_nms = le32_to_cpu(ino->xattr_names);
+               fscki->mode = le32_to_cpu(ino->mode);
+       } else {
+               ui = ubifs_inode(inode);
+               fscki->nlink = inode->i_nlink;
+               fscki->size = inode->i_size;
+               fscki->xattr_cnt = ui->xattr_cnt;
+               fscki->xattr_sz = ui->xattr_size;
+               fscki->xattr_nms = ui->xattr_names;
+               fscki->mode = inode->i_mode;
+               iput(inode);
+       }
+
+       if (S_ISDIR(fscki->mode)) {
+               fscki->calc_sz = UBIFS_INO_NODE_SZ;
+               fscki->calc_cnt = 2;
+       }
+
+       rb_link_node(&fscki->rb, parent, p);
+       rb_insert_color(&fscki->rb, &fsckd->inodes);
+
+       return fscki;
+}
+
+/**
+ * search_inode - search inode in the RB-tree of inodes.
+ * @fsckd: FS checking information
+ * @inum: inode number to search
+ *
+ * This is a helper function for 'check_leaf()' which searches inode @inum in
+ * the RB-tree of inodes and returns an inode information pointer or %NULL if
+ * the inode was not found.
+ */
+static struct fsck_inode *search_inode(struct fsck_data *fsckd, ino_t inum)
+{
+       struct rb_node *p;
+       struct fsck_inode *fscki;
+
+       p = fsckd->inodes.rb_node;
+       while (p) {
+               fscki = rb_entry(p, struct fsck_inode, rb);
+               if (inum < fscki->inum)
+                       p = p->rb_left;
+               else if (inum > fscki->inum)
+                       p = p->rb_right;
+               else
+                       return fscki;
+       }
+       return NULL;
+}
+
+/**
+ * read_add_inode - read inode node and add it to RB-tree of inodes.
+ * @c: UBIFS file-system description object
+ * @fsckd: FS checking information
+ * @inum: inode number to read
+ *
+ * This is a helper function for 'check_leaf()' which finds inode node @inum in
+ * the index, reads it, and adds it to the RB-tree of inodes. Returns inode
+ * information pointer in case of success and a negative error code in case of
+ * failure.
+ */
+static struct fsck_inode *read_add_inode(struct ubifs_info *c,
+                                        struct fsck_data *fsckd, ino_t inum)
+{
+       int n, err;
+       union ubifs_key key;
+       struct ubifs_znode *znode;
+       struct ubifs_zbranch *zbr;
+       struct ubifs_ino_node *ino;
+       struct fsck_inode *fscki;
+
+       fscki = search_inode(fsckd, inum);
+       if (fscki)
+               return fscki;
+
+       ino_key_init(c, &key, inum);
+       err = ubifs_lookup_level0(c, &key, &znode, &n);
+       if (!err) {
+               ubifs_err("inode %lu not found in index", (unsigned long)inum);
+               return ERR_PTR(-ENOENT);
+       } else if (err < 0) {
+               ubifs_err("error %d while looking up inode %lu",
+                         err, (unsigned long)inum);
+               return ERR_PTR(err);
+       }
+
+       zbr = &znode->zbranch[n];
+       if (zbr->len < UBIFS_INO_NODE_SZ) {
+               ubifs_err("bad node %lu node length %d",
+                         (unsigned long)inum, zbr->len);
+               return ERR_PTR(-EINVAL);
+       }
+
+       ino = kmalloc(zbr->len, GFP_NOFS);
+       if (!ino)
+               return ERR_PTR(-ENOMEM);
+
+       err = ubifs_tnc_read_node(c, zbr, ino);
+       if (err) {
+               ubifs_err("cannot read inode node at LEB %d:%d, error %d",
+                         zbr->lnum, zbr->offs, err);
+               kfree(ino);
+               return ERR_PTR(err);
+       }
+
+       fscki = add_inode(c, fsckd, ino);
+       kfree(ino);
+       if (IS_ERR(fscki)) {
+               ubifs_err("error %ld while adding inode %lu node",
+                         PTR_ERR(fscki), (unsigned long)inum);
+               return fscki;
+       }
+
+       return fscki;
+}
+
+/**
+ * check_leaf - check leaf node.
+ * @c: UBIFS file-system description object
+ * @zbr: zbranch of the leaf node to check
+ * @priv: FS checking information
+ *
+ * This is a helper function for 'dbg_check_filesystem()' which is called for
+ * every single leaf node while walking the indexing tree. It checks that the
+ * leaf node referred from the indexing tree exists, has correct CRC, and does
+ * some other basic validation. This function is also responsible for building
+ * an RB-tree of inodes - it adds all inodes into the RB-tree. It also
+ * calculates reference count, size, etc for each inode in order to later
+ * compare them to the information stored inside the inodes and detect possible
+ * inconsistencies. Returns zero in case of success and a negative error code
+ * in case of failure.
+ */
+static int check_leaf(struct ubifs_info *c, struct ubifs_zbranch *zbr,
+                     void *priv)
+{
+       ino_t inum;
+       void *node;
+       struct ubifs_ch *ch;
+       int err, type = key_type(c, &zbr->key);
+       struct fsck_inode *fscki;
+
+       if (zbr->len < UBIFS_CH_SZ) {
+               ubifs_err("bad leaf length %d (LEB %d:%d)",
+                         zbr->len, zbr->lnum, zbr->offs);
+               return -EINVAL;
+       }
+
+       node = kmalloc(zbr->len, GFP_NOFS);
+       if (!node)
+               return -ENOMEM;
+
+       err = ubifs_tnc_read_node(c, zbr, node);
+       if (err) {
+               ubifs_err("cannot read leaf node at LEB %d:%d, error %d",
+                         zbr->lnum, zbr->offs, err);
+               goto out_free;
+       }
+
+       /* If this is an inode node, add it to RB-tree of inodes */
+       if (type == UBIFS_INO_KEY) {
+               fscki = add_inode(c, priv, node);
+               if (IS_ERR(fscki)) {
+                       err = PTR_ERR(fscki);
+                       ubifs_err("error %d while adding inode node", err);
+                       goto out_dump;
+               }
+               goto out;
+       }
+
+       if (type != UBIFS_DENT_KEY && type != UBIFS_XENT_KEY &&
+           type != UBIFS_DATA_KEY) {
+               ubifs_err("unexpected node type %d at LEB %d:%d",
+                         type, zbr->lnum, zbr->offs);
+               err = -EINVAL;
+               goto out_free;
+       }
+
+       ch = node;
+       if (le64_to_cpu(ch->sqnum) > c->max_sqnum) {
+               ubifs_err("too high sequence number, max. is %llu",
+                         c->max_sqnum);
+               err = -EINVAL;
+               goto out_dump;
+       }
+
+       if (type == UBIFS_DATA_KEY) {
+               long long blk_offs;
+               struct ubifs_data_node *dn = node;
+
+               /*
+                * Search the inode node this data node belongs to and insert
+                * it to the RB-tree of inodes.
+                */
+               inum = key_inum_flash(c, &dn->key);
+               fscki = read_add_inode(c, priv, inum);
+               if (IS_ERR(fscki)) {
+                       err = PTR_ERR(fscki);
+                       ubifs_err("error %d while processing data node and trying to find inode node %lu",
+                                 err, (unsigned long)inum);
+                       goto out_dump;
+               }
+
+               /* Make sure the data node is within inode size */
+               blk_offs = key_block_flash(c, &dn->key);
+               blk_offs <<= UBIFS_BLOCK_SHIFT;
+               blk_offs += le32_to_cpu(dn->size);
+               if (blk_offs > fscki->size) {
+                       ubifs_err("data node at LEB %d:%d is not within inode size %lld",
+                                 zbr->lnum, zbr->offs, fscki->size);
+                       err = -EINVAL;
+                       goto out_dump;
+               }
+       } else {
+               int nlen;
+               struct ubifs_dent_node *dent = node;
+               struct fsck_inode *fscki1;
+
+               err = ubifs_validate_entry(c, dent);
+               if (err)
+                       goto out_dump;
+
+               /*
+                * Search the inode node this entry refers to and the parent
+                * inode node and insert them to the RB-tree of inodes.
+                */
+               inum = le64_to_cpu(dent->inum);
+               fscki = read_add_inode(c, priv, inum);
+               if (IS_ERR(fscki)) {
+                       err = PTR_ERR(fscki);
+                       ubifs_err("error %d while processing entry node and trying to find inode node %lu",
+                                 err, (unsigned long)inum);
+                       goto out_dump;
+               }
+
+               /* Count how many direntries or xentries refers this inode */
+               fscki->references += 1;
+
+               inum = key_inum_flash(c, &dent->key);
+               fscki1 = read_add_inode(c, priv, inum);
+               if (IS_ERR(fscki1)) {
+                       err = PTR_ERR(fscki1);
+                       ubifs_err("error %d while processing entry node and trying to find parent inode node %lu",
+                                 err, (unsigned long)inum);
+                       goto out_dump;
+               }
+
+               nlen = le16_to_cpu(dent->nlen);
+               if (type == UBIFS_XENT_KEY) {
+                       fscki1->calc_xcnt += 1;
+                       fscki1->calc_xsz += CALC_DENT_SIZE(nlen);
+                       fscki1->calc_xsz += CALC_XATTR_BYTES(fscki->size);
+                       fscki1->calc_xnms += nlen;
+               } else {
+                       fscki1->calc_sz += CALC_DENT_SIZE(nlen);
+                       if (dent->type == UBIFS_ITYPE_DIR)
+                               fscki1->calc_cnt += 1;
+               }
+       }
+
+out:
+       kfree(node);
+       return 0;
+
+out_dump:
+       ubifs_msg("dump of node at LEB %d:%d", zbr->lnum, zbr->offs);
+       ubifs_dump_node(c, node);
+out_free:
+       kfree(node);
+       return err;
+}
+
+/**
+ * free_inodes - free RB-tree of inodes.
+ * @fsckd: FS checking information
+ */
+static void free_inodes(struct fsck_data *fsckd)
+{
+       struct fsck_inode *fscki, *n;
+
+       rbtree_postorder_for_each_entry_safe(fscki, n, &fsckd->inodes, rb)
+               kfree(fscki);
+}
+
+/**
+ * check_inodes - checks all inodes.
+ * @c: UBIFS file-system description object
+ * @fsckd: FS checking information
+ *
+ * This is a helper function for 'dbg_check_filesystem()' which walks the
+ * RB-tree of inodes after the index scan has been finished, and checks that
+ * inode nlink, size, etc are correct. Returns zero if inodes are fine,
+ * %-EINVAL if not, and a negative error code in case of failure.
+ */
+static int check_inodes(struct ubifs_info *c, struct fsck_data *fsckd)
+{
+       int n, err;
+       union ubifs_key key;
+       struct ubifs_znode *znode;
+       struct ubifs_zbranch *zbr;
+       struct ubifs_ino_node *ino;
+       struct fsck_inode *fscki;
+       struct rb_node *this = rb_first(&fsckd->inodes);
+
+       while (this) {
+               fscki = rb_entry(this, struct fsck_inode, rb);
+               this = rb_next(this);
+
+               if (S_ISDIR(fscki->mode)) {
+                       /*
+                        * Directories have to have exactly one reference (they
+                        * cannot have hardlinks), although root inode is an
+                        * exception.
+                        */
+                       if (fscki->inum != UBIFS_ROOT_INO &&
+                           fscki->references != 1) {
+                               ubifs_err("directory inode %lu has %d direntries which refer it, but should be 1",
+                                         (unsigned long)fscki->inum,
+                                         fscki->references);
+                               goto out_dump;
+                       }
+                       if (fscki->inum == UBIFS_ROOT_INO &&
+                           fscki->references != 0) {
+                               ubifs_err("root inode %lu has non-zero (%d) direntries which refer it",
+                                         (unsigned long)fscki->inum,
+                                         fscki->references);
+                               goto out_dump;
+                       }
+                       if (fscki->calc_sz != fscki->size) {
+                               ubifs_err("directory inode %lu size is %lld, but calculated size is %lld",
+                                         (unsigned long)fscki->inum,
+                                         fscki->size, fscki->calc_sz);
+                               goto out_dump;
+                       }
+                       if (fscki->calc_cnt != fscki->nlink) {
+                               ubifs_err("directory inode %lu nlink is %d, but calculated nlink is %d",
+                                         (unsigned long)fscki->inum,
+                                         fscki->nlink, fscki->calc_cnt);
+                               goto out_dump;
+                       }
+               } else {
+                       if (fscki->references != fscki->nlink) {
+                               ubifs_err("inode %lu nlink is %d, but calculated nlink is %d",
+                                         (unsigned long)fscki->inum,
+                                         fscki->nlink, fscki->references);
+                               goto out_dump;
+                       }
+               }
+               if (fscki->xattr_sz != fscki->calc_xsz) {
+                       ubifs_err("inode %lu has xattr size %u, but calculated size is %lld",
+                                 (unsigned long)fscki->inum, fscki->xattr_sz,
+                                 fscki->calc_xsz);
+                       goto out_dump;
+               }
+               if (fscki->xattr_cnt != fscki->calc_xcnt) {
+                       ubifs_err("inode %lu has %u xattrs, but calculated count is %lld",
+                                 (unsigned long)fscki->inum,
+                                 fscki->xattr_cnt, fscki->calc_xcnt);
+                       goto out_dump;
+               }
+               if (fscki->xattr_nms != fscki->calc_xnms) {
+                       ubifs_err("inode %lu has xattr names' size %u, but calculated names' size is %lld",
+                                 (unsigned long)fscki->inum, fscki->xattr_nms,
+                                 fscki->calc_xnms);
+                       goto out_dump;
+               }
+       }
+
+       return 0;
+
+out_dump:
+       /* Read the bad inode and dump it */
+       ino_key_init(c, &key, fscki->inum);
+       err = ubifs_lookup_level0(c, &key, &znode, &n);
+       if (!err) {
+               ubifs_err("inode %lu not found in index",
+                         (unsigned long)fscki->inum);
+               return -ENOENT;
+       } else if (err < 0) {
+               ubifs_err("error %d while looking up inode %lu",
+                         err, (unsigned long)fscki->inum);
+               return err;
+       }
+
+       zbr = &znode->zbranch[n];
+       ino = kmalloc(zbr->len, GFP_NOFS);
+       if (!ino)
+               return -ENOMEM;
+
+       err = ubifs_tnc_read_node(c, zbr, ino);
+       if (err) {
+               ubifs_err("cannot read inode node at LEB %d:%d, error %d",
+                         zbr->lnum, zbr->offs, err);
+               kfree(ino);
+               return err;
+       }
+
+       ubifs_msg("dump of the inode %lu sitting in LEB %d:%d",
+                 (unsigned long)fscki->inum, zbr->lnum, zbr->offs);
+       ubifs_dump_node(c, ino);
+       kfree(ino);
+       return -EINVAL;
+}
+
+/**
+ * dbg_check_filesystem - check the file-system.
+ * @c: UBIFS file-system description object
+ *
+ * This function checks the file system, namely:
+ * o makes sure that all leaf nodes exist and their CRCs are correct;
+ * o makes sure inode nlink, size, xattr size/count are correct (for all
+ *   inodes).
+ *
+ * The function reads whole indexing tree and all nodes, so it is pretty
+ * heavy-weight. Returns zero if the file-system is consistent, %-EINVAL if
+ * not, and a negative error code in case of failure.
+ */
+int dbg_check_filesystem(struct ubifs_info *c)
+{
+       int err;
+       struct fsck_data fsckd;
+
+       if (!dbg_is_chk_fs(c))
+               return 0;
+
+       fsckd.inodes = RB_ROOT;
+       err = dbg_walk_index(c, check_leaf, NULL, &fsckd);
+       if (err)
+               goto out_free;
+
+       err = check_inodes(c, &fsckd);
+       if (err)
+               goto out_free;
+
+       free_inodes(&fsckd);
+       return 0;
+
+out_free:
+       ubifs_err("file-system check failed with error %d", err);
+       dump_stack();
+       free_inodes(&fsckd);
+       return err;
+}
+
+/**
+ * dbg_check_data_nodes_order - check that list of data nodes is sorted.
+ * @c: UBIFS file-system description object
+ * @head: the list of nodes ('struct ubifs_scan_node' objects)
+ *
+ * This function returns zero if the list of data nodes is sorted correctly,
+ * and %-EINVAL if not.
+ */
+int dbg_check_data_nodes_order(struct ubifs_info *c, struct list_head *head)
+{
+       struct list_head *cur;
+       struct ubifs_scan_node *sa, *sb;
+
+       if (!dbg_is_chk_gen(c))
+               return 0;
+
+       for (cur = head->next; cur->next != head; cur = cur->next) {
+               ino_t inuma, inumb;
+               uint32_t blka, blkb;
+
+               cond_resched();
+               sa = container_of(cur, struct ubifs_scan_node, list);
+               sb = container_of(cur->next, struct ubifs_scan_node, list);
+
+               if (sa->type != UBIFS_DATA_NODE) {
+                       ubifs_err("bad node type %d", sa->type);
+                       ubifs_dump_node(c, sa->node);
+                       return -EINVAL;
+               }
+               if (sb->type != UBIFS_DATA_NODE) {
+                       ubifs_err("bad node type %d", sb->type);
+                       ubifs_dump_node(c, sb->node);
+                       return -EINVAL;
+               }
+
+               inuma = key_inum(c, &sa->key);
+               inumb = key_inum(c, &sb->key);
+
+               if (inuma < inumb)
+                       continue;
+               if (inuma > inumb) {
+                       ubifs_err("larger inum %lu goes before inum %lu",
+                                 (unsigned long)inuma, (unsigned long)inumb);
+                       goto error_dump;
+               }
+
+               blka = key_block(c, &sa->key);
+               blkb = key_block(c, &sb->key);
+
+               if (blka > blkb) {
+                       ubifs_err("larger block %u goes before %u", blka, blkb);
+                       goto error_dump;
+               }
+               if (blka == blkb) {
+                       ubifs_err("two data nodes for the same block");
+                       goto error_dump;
+               }
+       }
+
+       return 0;
+
+error_dump:
+       ubifs_dump_node(c, sa->node);
+       ubifs_dump_node(c, sb->node);
+       return -EINVAL;
+}
+
+/**
+ * dbg_check_nondata_nodes_order - check that list of data nodes is sorted.
+ * @c: UBIFS file-system description object
+ * @head: the list of nodes ('struct ubifs_scan_node' objects)
+ *
+ * This function returns zero if the list of non-data nodes is sorted correctly,
+ * and %-EINVAL if not.
+ */
+int dbg_check_nondata_nodes_order(struct ubifs_info *c, struct list_head *head)
+{
+       struct list_head *cur;
+       struct ubifs_scan_node *sa, *sb;
+
+       if (!dbg_is_chk_gen(c))
+               return 0;
+
+       for (cur = head->next; cur->next != head; cur = cur->next) {
+               ino_t inuma, inumb;
+               uint32_t hasha, hashb;
+
+               cond_resched();
+               sa = container_of(cur, struct ubifs_scan_node, list);
+               sb = container_of(cur->next, struct ubifs_scan_node, list);
+
+               if (sa->type != UBIFS_INO_NODE && sa->type != UBIFS_DENT_NODE &&
+                   sa->type != UBIFS_XENT_NODE) {
+                       ubifs_err("bad node type %d", sa->type);
+                       ubifs_dump_node(c, sa->node);
+                       return -EINVAL;
+               }
+               if (sa->type != UBIFS_INO_NODE && sa->type != UBIFS_DENT_NODE &&
+                   sa->type != UBIFS_XENT_NODE) {
+                       ubifs_err("bad node type %d", sb->type);
+                       ubifs_dump_node(c, sb->node);
+                       return -EINVAL;
+               }
+
+               if (sa->type != UBIFS_INO_NODE && sb->type == UBIFS_INO_NODE) {
+                       ubifs_err("non-inode node goes before inode node");
+                       goto error_dump;
+               }
+
+               if (sa->type == UBIFS_INO_NODE && sb->type != UBIFS_INO_NODE)
+                       continue;
+
+               if (sa->type == UBIFS_INO_NODE && sb->type == UBIFS_INO_NODE) {
+                       /* Inode nodes are sorted in descending size order */
+                       if (sa->len < sb->len) {
+                               ubifs_err("smaller inode node goes first");
+                               goto error_dump;
+                       }
+                       continue;
+               }
+
+               /*
+                * This is either a dentry or xentry, which should be sorted in
+                * ascending (parent ino, hash) order.
+                */
+               inuma = key_inum(c, &sa->key);
+               inumb = key_inum(c, &sb->key);
+
+               if (inuma < inumb)
+                       continue;
+               if (inuma > inumb) {
+                       ubifs_err("larger inum %lu goes before inum %lu",
+                                 (unsigned long)inuma, (unsigned long)inumb);
+                       goto error_dump;
+               }
+
+               hasha = key_block(c, &sa->key);
+               hashb = key_block(c, &sb->key);
+
+               if (hasha > hashb) {
+                       ubifs_err("larger hash %u goes before %u",
+                                 hasha, hashb);
+                       goto error_dump;
+               }
+       }
+
+       return 0;
+
+error_dump:
+       ubifs_msg("dumping first node");
+       ubifs_dump_node(c, sa->node);
+       ubifs_msg("dumping second node");
+       ubifs_dump_node(c, sb->node);
+       return -EINVAL;
+       return 0;
+}
+
+static inline int chance(unsigned int n, unsigned int out_of)
+{
+       return !!((prandom_u32() % out_of) + 1 <= n);
+
+}
+
+static int power_cut_emulated(struct ubifs_info *c, int lnum, int write)
+{
+       struct ubifs_debug_info *d = c->dbg;
+
+       ubifs_assert(dbg_is_tst_rcvry(c));
+
+       if (!d->pc_cnt) {
+               /* First call - decide delay to the power cut */
+               if (chance(1, 2)) {
+                       unsigned long delay;
+
+                       if (chance(1, 2)) {
+                               d->pc_delay = 1;
+                               /* Fail withing 1 minute */
+                               delay = prandom_u32() % 60000;
+                               d->pc_timeout = jiffies;
+                               d->pc_timeout += msecs_to_jiffies(delay);
+                               ubifs_warn("failing after %lums", delay);
+                       } else {
+                               d->pc_delay = 2;
+                               delay = prandom_u32() % 10000;
+                               /* Fail within 10000 operations */
+                               d->pc_cnt_max = delay;
+                               ubifs_warn("failing after %lu calls", delay);
+                       }
+               }
+
+               d->pc_cnt += 1;
+       }
+
+       /* Determine if failure delay has expired */
+       if (d->pc_delay == 1 && time_before(jiffies, d->pc_timeout))
+                       return 0;
+       if (d->pc_delay == 2 && d->pc_cnt++ < d->pc_cnt_max)
+                       return 0;
+
+       if (lnum == UBIFS_SB_LNUM) {
+               if (write && chance(1, 2))
+                       return 0;
+               if (chance(19, 20))
+                       return 0;
+               ubifs_warn("failing in super block LEB %d", lnum);
+       } else if (lnum == UBIFS_MST_LNUM || lnum == UBIFS_MST_LNUM + 1) {
+               if (chance(19, 20))
+                       return 0;
+               ubifs_warn("failing in master LEB %d", lnum);
+       } else if (lnum >= UBIFS_LOG_LNUM && lnum <= c->log_last) {
+               if (write && chance(99, 100))
+                       return 0;
+               if (chance(399, 400))
+                       return 0;
+               ubifs_warn("failing in log LEB %d", lnum);
+       } else if (lnum >= c->lpt_first && lnum <= c->lpt_last) {
+               if (write && chance(7, 8))
+                       return 0;
+               if (chance(19, 20))
+                       return 0;
+               ubifs_warn("failing in LPT LEB %d", lnum);
+       } else if (lnum >= c->orph_first && lnum <= c->orph_last) {
+               if (write && chance(1, 2))
+                       return 0;
+               if (chance(9, 10))
+                       return 0;
+               ubifs_warn("failing in orphan LEB %d", lnum);
+       } else if (lnum == c->ihead_lnum) {
+               if (chance(99, 100))
+                       return 0;
+               ubifs_warn("failing in index head LEB %d", lnum);
+       } else if (c->jheads && lnum == c->jheads[GCHD].wbuf.lnum) {
+               if (chance(9, 10))
+                       return 0;
+               ubifs_warn("failing in GC head LEB %d", lnum);
+       } else if (write && !RB_EMPTY_ROOT(&c->buds) &&
+                  !ubifs_search_bud(c, lnum)) {
+               if (chance(19, 20))
+                       return 0;
+               ubifs_warn("failing in non-bud LEB %d", lnum);
+       } else if (c->cmt_state == COMMIT_RUNNING_BACKGROUND ||
+                  c->cmt_state == COMMIT_RUNNING_REQUIRED) {
+               if (chance(999, 1000))
+                       return 0;
+               ubifs_warn("failing in bud LEB %d commit running", lnum);
+       } else {
+               if (chance(9999, 10000))
+                       return 0;
+               ubifs_warn("failing in bud LEB %d commit not running", lnum);
+       }
+
+       d->pc_happened = 1;
+       ubifs_warn("========== Power cut emulated ==========");
+       dump_stack();
+       return 1;
+}
+
+static int corrupt_data(const struct ubifs_info *c, const void *buf,
+                       unsigned int len)
+{
+       unsigned int from, to, ffs = chance(1, 2);
+       unsigned char *p = (void *)buf;
+
+       from = prandom_u32() % len;
+       /* Corruption span max to end of write unit */
+       to = min(len, ALIGN(from + 1, c->max_write_size));
+
+       ubifs_warn("filled bytes %u-%u with %s", from, to - 1,
+                  ffs ? "0xFFs" : "random data");
+
+       if (ffs)
+               memset(p + from, 0xFF, to - from);
+       else
+               prandom_bytes(p + from, to - from);
+
+       return to;
+}
+
+int dbg_leb_write(struct ubifs_info *c, int lnum, const void *buf,
+                 int offs, int len)
+{
+       int err, failing;
+
+       if (c->dbg->pc_happened)
+               return -EROFS;
+
+       failing = power_cut_emulated(c, lnum, 1);
+       if (failing) {
+               len = corrupt_data(c, buf, len);
+               ubifs_warn("actually write %d bytes to LEB %d:%d (the buffer was corrupted)",
+                          len, lnum, offs);
+       }
+       err = ubi_leb_write(c->ubi, lnum, buf, offs, len);
+       if (err)
+               return err;
+       if (failing)
+               return -EROFS;
+       return 0;
+}
+
+int dbg_leb_change(struct ubifs_info *c, int lnum, const void *buf,
+                  int len)
+{
+       int err;
+
+       if (c->dbg->pc_happened)
+               return -EROFS;
+       if (power_cut_emulated(c, lnum, 1))
+               return -EROFS;
+       err = ubi_leb_change(c->ubi, lnum, buf, len);
+       if (err)
+               return err;
+       if (power_cut_emulated(c, lnum, 1))
+               return -EROFS;
+       return 0;
+}
+
+int dbg_leb_unmap(struct ubifs_info *c, int lnum)
+{
+       int err;
+
+       if (c->dbg->pc_happened)
+               return -EROFS;
+       if (power_cut_emulated(c, lnum, 0))
+               return -EROFS;
+       err = ubi_leb_unmap(c->ubi, lnum);
+       if (err)
+               return err;
+       if (power_cut_emulated(c, lnum, 0))
+               return -EROFS;
+       return 0;
+}
+
+int dbg_leb_map(struct ubifs_info *c, int lnum)
+{
+       int err;
+
+       if (c->dbg->pc_happened)
+               return -EROFS;
+       if (power_cut_emulated(c, lnum, 0))
+               return -EROFS;
+       err = ubi_leb_map(c->ubi, lnum);
+       if (err)
+               return err;
+       if (power_cut_emulated(c, lnum, 0))
+               return -EROFS;
+       return 0;
+}
+
+/*
+ * Root directory for UBIFS stuff in debugfs. Contains sub-directories which
+ * contain the stuff specific to particular file-system mounts.
+ */
+static struct dentry *dfs_rootdir;
+
+static int dfs_file_open(struct inode *inode, struct file *file)
+{
+       file->private_data = inode->i_private;
+       return nonseekable_open(inode, file);
+}
+
+/**
+ * provide_user_output - provide output to the user reading a debugfs file.
+ * @val: boolean value for the answer
+ * @u: the buffer to store the answer at
+ * @count: size of the buffer
+ * @ppos: position in the @u output buffer
+ *
+ * This is a simple helper function which stores @val boolean value in the user
+ * buffer when the user reads one of UBIFS debugfs files. Returns amount of
+ * bytes written to @u in case of success and a negative error code in case of
+ * failure.
+ */
+static int provide_user_output(int val, char __user *u, size_t count,
+                              loff_t *ppos)
+{
+       char buf[3];
+
+       if (val)
+               buf[0] = '1';
+       else
+               buf[0] = '0';
+       buf[1] = '\n';
+       buf[2] = 0x00;
+
+       return simple_read_from_buffer(u, count, ppos, buf, 2);
+}
+
+static ssize_t dfs_file_read(struct file *file, char __user *u, size_t count,
+                            loff_t *ppos)
+{
+       struct dentry *dent = file->f_path.dentry;
+       struct ubifs_info *c = file->private_data;
+       struct ubifs_debug_info *d = c->dbg;
+       int val;
+
+       if (dent == d->dfs_chk_gen)
+               val = d->chk_gen;
+       else if (dent == d->dfs_chk_index)
+               val = d->chk_index;
+       else if (dent == d->dfs_chk_orph)
+               val = d->chk_orph;
+       else if (dent == d->dfs_chk_lprops)
+               val = d->chk_lprops;
+       else if (dent == d->dfs_chk_fs)
+               val = d->chk_fs;
+       else if (dent == d->dfs_tst_rcvry)
+               val = d->tst_rcvry;
+       else if (dent == d->dfs_ro_error)
+               val = c->ro_error;
+       else
+               return -EINVAL;
+
+       return provide_user_output(val, u, count, ppos);
+}
+
+/**
+ * interpret_user_input - interpret user debugfs file input.
+ * @u: user-provided buffer with the input
+ * @count: buffer size
+ *
+ * This is a helper function which interpret user input to a boolean UBIFS
+ * debugfs file. Returns %0 or %1 in case of success and a negative error code
+ * in case of failure.
+ */
+static int interpret_user_input(const char __user *u, size_t count)
+{
+       size_t buf_size;
+       char buf[8];
+
+       buf_size = min_t(size_t, count, (sizeof(buf) - 1));
+       if (copy_from_user(buf, u, buf_size))
+               return -EFAULT;
+
+       if (buf[0] == '1')
+               return 1;
+       else if (buf[0] == '0')
+               return 0;
+
+       return -EINVAL;
+}
+
+static ssize_t dfs_file_write(struct file *file, const char __user *u,
+                             size_t count, loff_t *ppos)
+{
+       struct ubifs_info *c = file->private_data;
+       struct ubifs_debug_info *d = c->dbg;
+       struct dentry *dent = file->f_path.dentry;
+       int val;
+
+       /*
+        * TODO: this is racy - the file-system might have already been
+        * unmounted and we'd oops in this case. The plan is to fix it with
+        * help of 'iterate_supers_type()' which we should have in v3.0: when
+        * a debugfs opened, we rember FS's UUID in file->private_data. Then
+        * whenever we access the FS via a debugfs file, we iterate all UBIFS
+        * superblocks and fine the one with the same UUID, and take the
+        * locking right.
+        *
+        * The other way to go suggested by Al Viro is to create a separate
+        * 'ubifs-debug' file-system instead.
+        */
+       if (file->f_path.dentry == d->dfs_dump_lprops) {
+               ubifs_dump_lprops(c);
+               return count;
+       }
+       if (file->f_path.dentry == d->dfs_dump_budg) {
+               ubifs_dump_budg(c, &c->bi);
+               return count;
+       }
+       if (file->f_path.dentry == d->dfs_dump_tnc) {
+               mutex_lock(&c->tnc_mutex);
+               ubifs_dump_tnc(c);
+               mutex_unlock(&c->tnc_mutex);
+               return count;
+       }
+
+       val = interpret_user_input(u, count);
+       if (val < 0)
+               return val;
+
+       if (dent == d->dfs_chk_gen)
+               d->chk_gen = val;
+       else if (dent == d->dfs_chk_index)
+               d->chk_index = val;
+       else if (dent == d->dfs_chk_orph)
+               d->chk_orph = val;
+       else if (dent == d->dfs_chk_lprops)
+               d->chk_lprops = val;
+       else if (dent == d->dfs_chk_fs)
+               d->chk_fs = val;
+       else if (dent == d->dfs_tst_rcvry)
+               d->tst_rcvry = val;
+       else if (dent == d->dfs_ro_error)
+               c->ro_error = !!val;
+       else
+               return -EINVAL;
+
+       return count;
+}
+
+static const struct file_operations dfs_fops = {
+       .open = dfs_file_open,
+       .read = dfs_file_read,
+       .write = dfs_file_write,
+       .owner = THIS_MODULE,
+       .llseek = no_llseek,
+};
+
+/**
+ * dbg_debugfs_init_fs - initialize debugfs for UBIFS instance.
+ * @c: UBIFS file-system description object
+ *
+ * This function creates all debugfs files for this instance of UBIFS. Returns
+ * zero in case of success and a negative error code in case of failure.
+ *
+ * Note, the only reason we have not merged this function with the
+ * 'ubifs_debugging_init()' function is because it is better to initialize
+ * debugfs interfaces at the very end of the mount process, and remove them at
+ * the very beginning of the mount process.
+ */
+int dbg_debugfs_init_fs(struct ubifs_info *c)
+{
+       int err, n;
+       const char *fname;
+       struct dentry *dent;
+       struct ubifs_debug_info *d = c->dbg;
+
+       if (!IS_ENABLED(CONFIG_DEBUG_FS))
+               return 0;
+
+       n = snprintf(d->dfs_dir_name, UBIFS_DFS_DIR_LEN + 1, UBIFS_DFS_DIR_NAME,
+                    c->vi.ubi_num, c->vi.vol_id);
+       if (n == UBIFS_DFS_DIR_LEN) {
+               /* The array size is too small */
+               fname = UBIFS_DFS_DIR_NAME;
+               dent = ERR_PTR(-EINVAL);
+               goto out;
+       }
+
+       fname = d->dfs_dir_name;
+       dent = debugfs_create_dir(fname, dfs_rootdir);
+       if (IS_ERR_OR_NULL(dent))
+               goto out;
+       d->dfs_dir = dent;
+
+       fname = "dump_lprops";
+       dent = debugfs_create_file(fname, S_IWUSR, d->dfs_dir, c, &dfs_fops);
+       if (IS_ERR_OR_NULL(dent))
+               goto out_remove;
+       d->dfs_dump_lprops = dent;
+
+       fname = "dump_budg";
+       dent = debugfs_create_file(fname, S_IWUSR, d->dfs_dir, c, &dfs_fops);
+       if (IS_ERR_OR_NULL(dent))
+               goto out_remove;
+       d->dfs_dump_budg = dent;
+
+       fname = "dump_tnc";
+       dent = debugfs_create_file(fname, S_IWUSR, d->dfs_dir, c, &dfs_fops);
+       if (IS_ERR_OR_NULL(dent))
+               goto out_remove;
+       d->dfs_dump_tnc = dent;
+
+       fname = "chk_general";
+       dent = debugfs_create_file(fname, S_IRUSR | S_IWUSR, d->dfs_dir, c,
+                                  &dfs_fops);
+       if (IS_ERR_OR_NULL(dent))
+               goto out_remove;
+       d->dfs_chk_gen = dent;
+
+       fname = "chk_index";
+       dent = debugfs_create_file(fname, S_IRUSR | S_IWUSR, d->dfs_dir, c,
+                                  &dfs_fops);
+       if (IS_ERR_OR_NULL(dent))
+               goto out_remove;
+       d->dfs_chk_index = dent;
+
+       fname = "chk_orphans";
+       dent = debugfs_create_file(fname, S_IRUSR | S_IWUSR, d->dfs_dir, c,
+                                  &dfs_fops);
+       if (IS_ERR_OR_NULL(dent))
+               goto out_remove;
+       d->dfs_chk_orph = dent;
+
+       fname = "chk_lprops";
+       dent = debugfs_create_file(fname, S_IRUSR | S_IWUSR, d->dfs_dir, c,
+                                  &dfs_fops);
+       if (IS_ERR_OR_NULL(dent))
+               goto out_remove;
+       d->dfs_chk_lprops = dent;
+
+       fname = "chk_fs";
+       dent = debugfs_create_file(fname, S_IRUSR | S_IWUSR, d->dfs_dir, c,
+                                  &dfs_fops);
+       if (IS_ERR_OR_NULL(dent))
+               goto out_remove;
+       d->dfs_chk_fs = dent;
+
+       fname = "tst_recovery";
+       dent = debugfs_create_file(fname, S_IRUSR | S_IWUSR, d->dfs_dir, c,
+                                  &dfs_fops);
+       if (IS_ERR_OR_NULL(dent))
+               goto out_remove;
+       d->dfs_tst_rcvry = dent;
+
+       fname = "ro_error";
+       dent = debugfs_create_file(fname, S_IRUSR | S_IWUSR, d->dfs_dir, c,
+                                  &dfs_fops);
+       if (IS_ERR_OR_NULL(dent))
+               goto out_remove;
+       d->dfs_ro_error = dent;
+
+       return 0;
+
+out_remove:
+       debugfs_remove_recursive(d->dfs_dir);
+out:
+       err = dent ? PTR_ERR(dent) : -ENODEV;
+       ubifs_err("cannot create \"%s\" debugfs file or directory, error %d\n",
+                 fname, err);
+       return err;
+}
+
+/**
+ * dbg_debugfs_exit_fs - remove all debugfs files.
+ * @c: UBIFS file-system description object
+ */
+void dbg_debugfs_exit_fs(struct ubifs_info *c)
+{
+       if (IS_ENABLED(CONFIG_DEBUG_FS))
+               debugfs_remove_recursive(c->dbg->dfs_dir);
+}
+
+struct ubifs_global_debug_info ubifs_dbg;
+
+static struct dentry *dfs_chk_gen;
+static struct dentry *dfs_chk_index;
+static struct dentry *dfs_chk_orph;
+static struct dentry *dfs_chk_lprops;
+static struct dentry *dfs_chk_fs;
+static struct dentry *dfs_tst_rcvry;
+
+static ssize_t dfs_global_file_read(struct file *file, char __user *u,
+                                   size_t count, loff_t *ppos)
+{
+       struct dentry *dent = file->f_path.dentry;
+       int val;
+
+       if (dent == dfs_chk_gen)
+               val = ubifs_dbg.chk_gen;
+       else if (dent == dfs_chk_index)
+               val = ubifs_dbg.chk_index;
+       else if (dent == dfs_chk_orph)
+               val = ubifs_dbg.chk_orph;
+       else if (dent == dfs_chk_lprops)
+               val = ubifs_dbg.chk_lprops;
+       else if (dent == dfs_chk_fs)
+               val = ubifs_dbg.chk_fs;
+       else if (dent == dfs_tst_rcvry)
+               val = ubifs_dbg.tst_rcvry;
+       else
+               return -EINVAL;
+
+       return provide_user_output(val, u, count, ppos);
+}
+
+static ssize_t dfs_global_file_write(struct file *file, const char __user *u,
+                                    size_t count, loff_t *ppos)
+{
+       struct dentry *dent = file->f_path.dentry;
+       int val;
+
+       val = interpret_user_input(u, count);
+       if (val < 0)
+               return val;
+
+       if (dent == dfs_chk_gen)
+               ubifs_dbg.chk_gen = val;
+       else if (dent == dfs_chk_index)
+               ubifs_dbg.chk_index = val;
+       else if (dent == dfs_chk_orph)
+               ubifs_dbg.chk_orph = val;
+       else if (dent == dfs_chk_lprops)
+               ubifs_dbg.chk_lprops = val;
+       else if (dent == dfs_chk_fs)
+               ubifs_dbg.chk_fs = val;
+       else if (dent == dfs_tst_rcvry)
+               ubifs_dbg.tst_rcvry = val;
+       else
+               return -EINVAL;
+
+       return count;
+}
+
+static const struct file_operations dfs_global_fops = {
+       .read = dfs_global_file_read,
+       .write = dfs_global_file_write,
+       .owner = THIS_MODULE,
+       .llseek = no_llseek,
+};
+
+/**
+ * dbg_debugfs_init - initialize debugfs file-system.
+ *
+ * UBIFS uses debugfs file-system to expose various debugging knobs to
+ * user-space. This function creates "ubifs" directory in the debugfs
+ * file-system. Returns zero in case of success and a negative error code in
+ * case of failure.
+ */
+int dbg_debugfs_init(void)
+{
+       int err;
+       const char *fname;
+       struct dentry *dent;
+
+       if (!IS_ENABLED(CONFIG_DEBUG_FS))
+               return 0;
+
+       fname = "ubifs";
+       dent = debugfs_create_dir(fname, NULL);
+       if (IS_ERR_OR_NULL(dent))
+               goto out;
+       dfs_rootdir = dent;
+
+       fname = "chk_general";
+       dent = debugfs_create_file(fname, S_IRUSR | S_IWUSR, dfs_rootdir, NULL,
+                                  &dfs_global_fops);
+       if (IS_ERR_OR_NULL(dent))
+               goto out_remove;
+       dfs_chk_gen = dent;
+
+       fname = "chk_index";
+       dent = debugfs_create_file(fname, S_IRUSR | S_IWUSR, dfs_rootdir, NULL,
+                                  &dfs_global_fops);
+       if (IS_ERR_OR_NULL(dent))
+               goto out_remove;
+       dfs_chk_index = dent;
+
+       fname = "chk_orphans";
+       dent = debugfs_create_file(fname, S_IRUSR | S_IWUSR, dfs_rootdir, NULL,
+                                  &dfs_global_fops);
+       if (IS_ERR_OR_NULL(dent))
+               goto out_remove;
+       dfs_chk_orph = dent;
+
+       fname = "chk_lprops";
+       dent = debugfs_create_file(fname, S_IRUSR | S_IWUSR, dfs_rootdir, NULL,
+                                  &dfs_global_fops);
+       if (IS_ERR_OR_NULL(dent))
+               goto out_remove;
+       dfs_chk_lprops = dent;
+
+       fname = "chk_fs";
+       dent = debugfs_create_file(fname, S_IRUSR | S_IWUSR, dfs_rootdir, NULL,
+                                  &dfs_global_fops);
+       if (IS_ERR_OR_NULL(dent))
+               goto out_remove;
+       dfs_chk_fs = dent;
+
+       fname = "tst_recovery";
+       dent = debugfs_create_file(fname, S_IRUSR | S_IWUSR, dfs_rootdir, NULL,
+                                  &dfs_global_fops);
+       if (IS_ERR_OR_NULL(dent))
+               goto out_remove;
+       dfs_tst_rcvry = dent;
+
+       return 0;
+
+out_remove:
+       debugfs_remove_recursive(dfs_rootdir);
+out:
+       err = dent ? PTR_ERR(dent) : -ENODEV;
+       ubifs_err("cannot create \"%s\" debugfs file or directory, error %d\n",
+                 fname, err);
+       return err;
+}
+
+/**
+ * dbg_debugfs_exit - remove the "ubifs" directory from debugfs file-system.
+ */
+void dbg_debugfs_exit(void)
+{
+       if (IS_ENABLED(CONFIG_DEBUG_FS))
+               debugfs_remove_recursive(dfs_rootdir);
+}
+
+/**
+ * ubifs_debugging_init - initialize UBIFS debugging.
+ * @c: UBIFS file-system description object
+ *
+ * This function initializes debugging-related data for the file system.
+ * Returns zero in case of success and a negative error code in case of
+ * failure.
+ */
+int ubifs_debugging_init(struct ubifs_info *c)
+{
+       c->dbg = kzalloc(sizeof(struct ubifs_debug_info), GFP_KERNEL);
+       if (!c->dbg)
+               return -ENOMEM;
+
+       return 0;
+}
+
+/**
+ * ubifs_debugging_exit - free debugging data.
+ * @c: UBIFS file-system description object
+ */
+void ubifs_debugging_exit(struct ubifs_info *c)
+{
+       kfree(c->dbg);
+}
+#endif
index 62617b69270cac34bc0c4bd267eddd3b6fcb3dcd..6d325af8bcc296305815536f748aed790e697df4 100644 (file)
@@ -3,18 +3,7 @@
  *
  * Copyright (C) 2006-2008 Nokia Corporation.
  *
- * This program is free software; you can redistribute it and/or modify it
- * under the terms of the GNU General Public License version 2 as published by
- * the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
- * more details.
- *
- * You should have received a copy of the GNU General Public License along with
- * this program; if not, write to the Free Software Foundation, Inc., 51
- * Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
+ * SPDX-License-Identifier:    GPL-2.0+
  *
  * Authors: Artem Bityutskiy (Битюцкий Артём)
  *          Adrian Hunter
 #ifndef __UBIFS_DEBUG_H__
 #define __UBIFS_DEBUG_H__
 
-#ifdef CONFIG_UBIFS_FS_DEBUG
+#define __UBOOT__
+/* Checking helper functions */
+typedef int (*dbg_leaf_callback)(struct ubifs_info *c,
+                                struct ubifs_zbranch *zbr, void *priv);
+typedef int (*dbg_znode_callback)(struct ubifs_info *c,
+                                 struct ubifs_znode *znode, void *priv);
+
+/*
+ * The UBIFS debugfs directory name pattern and maximum name length (3 for "ubi"
+ * + 1 for "_" and plus 2x2 for 2 UBI numbers and 1 for the trailing zero byte.
+ */
+#define UBIFS_DFS_DIR_NAME "ubi%d_%d"
+#define UBIFS_DFS_DIR_LEN  (3 + 1 + 2*2 + 1)
 
 /**
  * ubifs_debug_info - per-FS debugging information.
- * @buf: a buffer of LEB size, used for various purposes
  * @old_zroot: old index root - used by 'dbg_check_old_index()'
  * @old_zroot_level: old index root level - used by 'dbg_check_old_index()'
  * @old_zroot_sqnum: old index root sqnum - used by 'dbg_check_old_index()'
- * @failure_mode: failure mode for recovery testing
- * @fail_delay: 0=>don't delay, 1=>delay a time, 2=>delay a number of calls
- * @fail_timeout: time in jiffies when delay of failure mode expires
- * @fail_cnt: current number of calls to failure mode I/O functions
- * @fail_cnt_max: number of calls by which to delay failure mode
+ *
+ * @pc_happened: non-zero if an emulated power cut happened
+ * @pc_delay: 0=>don't delay, 1=>delay a time, 2=>delay a number of calls
+ * @pc_timeout: time in jiffies when delay of failure mode expires
+ * @pc_cnt: current number of calls to failure mode I/O functions
+ * @pc_cnt_max: number of calls by which to delay failure mode
+ *
  * @chk_lpt_sz: used by LPT tree size checker
  * @chk_lpt_sz2: used by LPT tree size checker
  * @chk_lpt_wastage: used by LPT tree size checker
  * @new_ihead_offs: used by debugging to check @c->ihead_offs
  *
  * @saved_lst: saved lprops statistics (used by 'dbg_save_space_info()')
- * @saved_free: saved free space (used by 'dbg_save_space_info()')
+ * @saved_bi: saved budgeting information
+ * @saved_free: saved amount of free space
+ * @saved_idx_gc_cnt: saved value of @c->idx_gc_cnt
+ *
+ * @chk_gen: if general extra checks are enabled
+ * @chk_index: if index xtra checks are enabled
+ * @chk_orph: if orphans extra checks are enabled
+ * @chk_lprops: if lprops extra checks are enabled
+ * @chk_fs: if UBIFS contents extra checks are enabled
+ * @tst_rcvry: if UBIFS recovery testing mode enabled
  *
- * dfs_dir_name: name of debugfs directory containing this file-system's files
- * dfs_dir: direntry object of the file-system debugfs directory
- * dfs_dump_lprops: "dump lprops" debugfs knob
- * dfs_dump_budg: "dump budgeting information" debugfs knob
- * dfs_dump_tnc: "dump TNC" debugfs knob
+ * @dfs_dir_name: name of debugfs directory containing this file-system's files
+ * @dfs_dir: direntry object of the file-system debugfs directory
+ * @dfs_dump_lprops: "dump lprops" debugfs knob
+ * @dfs_dump_budg: "dump budgeting information" debugfs knob
+ * @dfs_dump_tnc: "dump TNC" debugfs knob
+ * @dfs_chk_gen: debugfs knob to enable UBIFS general extra checks
+ * @dfs_chk_index: debugfs knob to enable UBIFS index extra checks
+ * @dfs_chk_orph: debugfs knob to enable UBIFS orphans extra checks
+ * @dfs_chk_lprops: debugfs knob to enable UBIFS LEP properties extra checks
+ * @dfs_chk_fs: debugfs knob to enable UBIFS contents extra checks
+ * @dfs_tst_rcvry: debugfs knob to enable UBIFS recovery testing
+ * @dfs_ro_error: debugfs knob to switch UBIFS to R/O mode (different to
+ *                re-mounting to R/O mode because it does not flush any buffers
+ *                and UBIFS just starts returning -EROFS on all write
+ *               operations)
  */
 struct ubifs_debug_info {
-       void *buf;
        struct ubifs_zbranch old_zroot;
        int old_zroot_level;
        unsigned long long old_zroot_sqnum;
-       int failure_mode;
-       int fail_delay;
-       unsigned long fail_timeout;
-       unsigned int fail_cnt;
-       unsigned int fail_cnt_max;
+
+       int pc_happened;
+       int pc_delay;
+       unsigned long pc_timeout;
+       unsigned int pc_cnt;
+       unsigned int pc_cnt_max;
+
        long long chk_lpt_sz;
        long long chk_lpt_sz2;
        long long chk_lpt_wastage;
@@ -72,321 +94,285 @@ struct ubifs_debug_info {
        int new_ihead_offs;
 
        struct ubifs_lp_stats saved_lst;
+       struct ubifs_budg_info saved_bi;
        long long saved_free;
+       int saved_idx_gc_cnt;
+
+       unsigned int chk_gen:1;
+       unsigned int chk_index:1;
+       unsigned int chk_orph:1;
+       unsigned int chk_lprops:1;
+       unsigned int chk_fs:1;
+       unsigned int tst_rcvry:1;
 
-       char dfs_dir_name[100];
+       char dfs_dir_name[UBIFS_DFS_DIR_LEN + 1];
        struct dentry *dfs_dir;
        struct dentry *dfs_dump_lprops;
        struct dentry *dfs_dump_budg;
        struct dentry *dfs_dump_tnc;
+       struct dentry *dfs_chk_gen;
+       struct dentry *dfs_chk_index;
+       struct dentry *dfs_chk_orph;
+       struct dentry *dfs_chk_lprops;
+       struct dentry *dfs_chk_fs;
+       struct dentry *dfs_tst_rcvry;
+       struct dentry *dfs_ro_error;
 };
 
-#define UBIFS_DBG(op) op
+/**
+ * ubifs_global_debug_info - global (not per-FS) UBIFS debugging information.
+ *
+ * @chk_gen: if general extra checks are enabled
+ * @chk_index: if index xtra checks are enabled
+ * @chk_orph: if orphans extra checks are enabled
+ * @chk_lprops: if lprops extra checks are enabled
+ * @chk_fs: if UBIFS contents extra checks are enabled
+ * @tst_rcvry: if UBIFS recovery testing mode enabled
+ */
+struct ubifs_global_debug_info {
+       unsigned int chk_gen:1;
+       unsigned int chk_index:1;
+       unsigned int chk_orph:1;
+       unsigned int chk_lprops:1;
+       unsigned int chk_fs:1;
+       unsigned int tst_rcvry:1;
+};
 
+#ifndef __UBOOT__
 #define ubifs_assert(expr) do {                                                \
        if (unlikely(!(expr))) {                                               \
-               printk(KERN_CRIT "UBIFS assert failed in %s at %u (pid %d)\n", \
-                      __func__, __LINE__, 0);                      \
-               dbg_dump_stack();                                              \
+               pr_crit("UBIFS assert failed in %s at %u (pid %d)\n",          \
+                      __func__, __LINE__, current->pid);                      \
+               dump_stack();                                                  \
        }                                                                      \
 } while (0)
 
 #define ubifs_assert_cmt_locked(c) do {                                        \
        if (unlikely(down_write_trylock(&(c)->commit_sem))) {                  \
                up_write(&(c)->commit_sem);                                    \
-               printk(KERN_CRIT "commit lock is not locked!\n");              \
+               pr_crit("commit lock is not locked!\n");                       \
                ubifs_assert(0);                                               \
        }                                                                      \
 } while (0)
 
-#define dbg_dump_stack() do {                                                  \
-       if (!dbg_failure_mode)                                                 \
+#define ubifs_dbg_msg(type, fmt, ...) \
+       pr_debug("UBIFS DBG " type " (pid %d): " fmt "\n", current->pid,       \
+                ##__VA_ARGS__)
+
+#define DBG_KEY_BUF_LEN 48
+#define ubifs_dbg_msg_key(type, key, fmt, ...) do {                            \
+       char __tmp_key_buf[DBG_KEY_BUF_LEN];                                   \
+       pr_debug("UBIFS DBG " type " (pid %d): " fmt "%s\n", current->pid,     \
+                ##__VA_ARGS__,                                                \
+                dbg_snprintf_key(c, key, __tmp_key_buf, DBG_KEY_BUF_LEN));    \
+} while (0)
+#else
+#define ubifs_assert(expr) do {                                                \
+       if (unlikely(!(expr))) {                                               \
+               pr_crit("UBIFS assert failed in %s at %u\n",                   \
+                      __func__, __LINE__);                                    \
                dump_stack();                                                  \
+       }                                                                      \
 } while (0)
 
-/* Generic debugging messages */
-#define dbg_msg(fmt, ...) do {                                                 \
-       spin_lock(&dbg_lock);                                                  \
-       printk(KERN_DEBUG "UBIFS DBG (pid %d): %s: " fmt "\n", 0,   \
-              __func__, ##__VA_ARGS__);                                       \
-       spin_unlock(&dbg_lock);                                                \
+#define ubifs_assert_cmt_locked(c) do {                                        \
+       if (unlikely(down_write_trylock(&(c)->commit_sem))) {                  \
+               up_write(&(c)->commit_sem);                                    \
+               pr_crit("commit lock is not locked!\n");                       \
+               ubifs_assert(0);                                               \
+       }                                                                      \
 } while (0)
 
-#define dbg_do_msg(typ, fmt, ...) do {                                         \
-       if (ubifs_msg_flags & typ)                                             \
-               dbg_msg(fmt, ##__VA_ARGS__);                                   \
+#define ubifs_dbg_msg(type, fmt, ...) \
+       pr_debug("UBIFS DBG " type ": " fmt "\n",                              \
+                ##__VA_ARGS__)
+
+#define DBG_KEY_BUF_LEN 48
+#if defined CONFIG_MTD_DEBUG
+#define ubifs_dbg_msg_key(type, key, fmt, ...) do {                            \
+       char __tmp_key_buf[DBG_KEY_BUF_LEN];                                   \
+       pr_debug("UBIFS DBG " type ": " fmt "%s\n",                            \
+                ##__VA_ARGS__,                                                \
+                dbg_snprintf_key(c, key, __tmp_key_buf, DBG_KEY_BUF_LEN));    \
 } while (0)
-
-#define dbg_err(fmt, ...) do {                                                 \
-       spin_lock(&dbg_lock);                                                  \
-       ubifs_err(fmt, ##__VA_ARGS__);                                         \
-       spin_unlock(&dbg_lock);                                                \
+#else
+#define ubifs_dbg_msg_key(type, key, fmt, ...) do {                            \
+       pr_debug("UBIFS DBG\n");                                               \
 } while (0)
 
-const char *dbg_key_str0(const struct ubifs_info *c,
-                        const union ubifs_key *key);
-const char *dbg_key_str1(const struct ubifs_info *c,
-                        const union ubifs_key *key);
+#endif
 
-/*
- * DBGKEY macros require @dbg_lock to be held, which it is in the dbg message
- * macros.
- */
-#define DBGKEY(key)    dbg_key_str0(c, (key))
-#define DBGKEY1(key)   dbg_key_str1(c, (key))
+#endif
 
 /* General messages */
-#define dbg_gen(fmt, ...)   dbg_do_msg(UBIFS_MSG_GEN, fmt, ##__VA_ARGS__)
-
+#define dbg_gen(fmt, ...)   ubifs_dbg_msg("gen", fmt, ##__VA_ARGS__)
 /* Additional journal messages */
-#define dbg_jnl(fmt, ...)   dbg_do_msg(UBIFS_MSG_JNL, fmt, ##__VA_ARGS__)
-
+#define dbg_jnl(fmt, ...)   ubifs_dbg_msg("jnl", fmt, ##__VA_ARGS__)
+#define dbg_jnlk(key, fmt, ...) \
+       ubifs_dbg_msg_key("jnl", key, fmt, ##__VA_ARGS__)
 /* Additional TNC messages */
-#define dbg_tnc(fmt, ...)   dbg_do_msg(UBIFS_MSG_TNC, fmt, ##__VA_ARGS__)
-
+#define dbg_tnc(fmt, ...)   ubifs_dbg_msg("tnc", fmt, ##__VA_ARGS__)
+#define dbg_tnck(key, fmt, ...) \
+       ubifs_dbg_msg_key("tnc", key, fmt, ##__VA_ARGS__)
 /* Additional lprops messages */
-#define dbg_lp(fmt, ...)    dbg_do_msg(UBIFS_MSG_LP, fmt, ##__VA_ARGS__)
-
+#define dbg_lp(fmt, ...)    ubifs_dbg_msg("lp", fmt, ##__VA_ARGS__)
 /* Additional LEB find messages */
-#define dbg_find(fmt, ...)  dbg_do_msg(UBIFS_MSG_FIND, fmt, ##__VA_ARGS__)
-
+#define dbg_find(fmt, ...)  ubifs_dbg_msg("find", fmt, ##__VA_ARGS__)
 /* Additional mount messages */
-#define dbg_mnt(fmt, ...)   dbg_do_msg(UBIFS_MSG_MNT, fmt, ##__VA_ARGS__)
-
+#define dbg_mnt(fmt, ...)   ubifs_dbg_msg("mnt", fmt, ##__VA_ARGS__)
+#define dbg_mntk(key, fmt, ...) \
+       ubifs_dbg_msg_key("mnt", key, fmt, ##__VA_ARGS__)
 /* Additional I/O messages */
-#define dbg_io(fmt, ...)    dbg_do_msg(UBIFS_MSG_IO, fmt, ##__VA_ARGS__)
-
+#define dbg_io(fmt, ...)    ubifs_dbg_msg("io", fmt, ##__VA_ARGS__)
 /* Additional commit messages */
-#define dbg_cmt(fmt, ...)   dbg_do_msg(UBIFS_MSG_CMT, fmt, ##__VA_ARGS__)
-
+#define dbg_cmt(fmt, ...)   ubifs_dbg_msg("cmt", fmt, ##__VA_ARGS__)
 /* Additional budgeting messages */
-#define dbg_budg(fmt, ...)  dbg_do_msg(UBIFS_MSG_BUDG, fmt, ##__VA_ARGS__)
-
+#define dbg_budg(fmt, ...)  ubifs_dbg_msg("budg", fmt, ##__VA_ARGS__)
 /* Additional log messages */
-#define dbg_log(fmt, ...)   dbg_do_msg(UBIFS_MSG_LOG, fmt, ##__VA_ARGS__)
-
+#define dbg_log(fmt, ...)   ubifs_dbg_msg("log", fmt, ##__VA_ARGS__)
 /* Additional gc messages */
-#define dbg_gc(fmt, ...)    dbg_do_msg(UBIFS_MSG_GC, fmt, ##__VA_ARGS__)
-
+#define dbg_gc(fmt, ...)    ubifs_dbg_msg("gc", fmt, ##__VA_ARGS__)
 /* Additional scan messages */
-#define dbg_scan(fmt, ...)  dbg_do_msg(UBIFS_MSG_SCAN, fmt, ##__VA_ARGS__)
-
+#define dbg_scan(fmt, ...)  ubifs_dbg_msg("scan", fmt, ##__VA_ARGS__)
 /* Additional recovery messages */
-#define dbg_rcvry(fmt, ...) dbg_do_msg(UBIFS_MSG_RCVRY, fmt, ##__VA_ARGS__)
+#define dbg_rcvry(fmt, ...) ubifs_dbg_msg("rcvry", fmt, ##__VA_ARGS__)
+
+#ifndef __UBOOT__
+extern struct ubifs_global_debug_info ubifs_dbg;
+
+static inline int dbg_is_chk_gen(const struct ubifs_info *c)
+{
+       return !!(ubifs_dbg.chk_gen || c->dbg->chk_gen);
+}
+static inline int dbg_is_chk_index(const struct ubifs_info *c)
+{
+       return !!(ubifs_dbg.chk_index || c->dbg->chk_index);
+}
+static inline int dbg_is_chk_orph(const struct ubifs_info *c)
+{
+       return !!(ubifs_dbg.chk_orph || c->dbg->chk_orph);
+}
+static inline int dbg_is_chk_lprops(const struct ubifs_info *c)
+{
+       return !!(ubifs_dbg.chk_lprops || c->dbg->chk_lprops);
+}
+static inline int dbg_is_chk_fs(const struct ubifs_info *c)
+{
+       return !!(ubifs_dbg.chk_fs || c->dbg->chk_fs);
+}
+static inline int dbg_is_tst_rcvry(const struct ubifs_info *c)
+{
+       return !!(ubifs_dbg.tst_rcvry || c->dbg->tst_rcvry);
+}
+static inline int dbg_is_power_cut(const struct ubifs_info *c)
+{
+       return !!c->dbg->pc_happened;
+}
 
-/*
- * Debugging message type flags (must match msg_type_names in debug.c).
- *
- * UBIFS_MSG_GEN: general messages
- * UBIFS_MSG_JNL: journal messages
- * UBIFS_MSG_MNT: mount messages
- * UBIFS_MSG_CMT: commit messages
- * UBIFS_MSG_FIND: LEB find messages
- * UBIFS_MSG_BUDG: budgeting messages
- * UBIFS_MSG_GC: garbage collection messages
- * UBIFS_MSG_TNC: TNC messages
- * UBIFS_MSG_LP: lprops messages
- * UBIFS_MSG_IO: I/O messages
- * UBIFS_MSG_LOG: log messages
- * UBIFS_MSG_SCAN: scan messages
- * UBIFS_MSG_RCVRY: recovery messages
- */
-enum {
-       UBIFS_MSG_GEN   = 0x1,
-       UBIFS_MSG_JNL   = 0x2,
-       UBIFS_MSG_MNT   = 0x4,
-       UBIFS_MSG_CMT   = 0x8,
-       UBIFS_MSG_FIND  = 0x10,
-       UBIFS_MSG_BUDG  = 0x20,
-       UBIFS_MSG_GC    = 0x40,
-       UBIFS_MSG_TNC   = 0x80,
-       UBIFS_MSG_LP    = 0x100,
-       UBIFS_MSG_IO    = 0x200,
-       UBIFS_MSG_LOG   = 0x400,
-       UBIFS_MSG_SCAN  = 0x800,
-       UBIFS_MSG_RCVRY = 0x1000,
-};
-
-/* Debugging message type flags for each default debug message level */
-#define UBIFS_MSG_LVL_0 0
-#define UBIFS_MSG_LVL_1 0x1
-#define UBIFS_MSG_LVL_2 0x7f
-#define UBIFS_MSG_LVL_3 0xffff
-
-/*
- * Debugging check flags (must match chk_names in debug.c).
- *
- * UBIFS_CHK_GEN: general checks
- * UBIFS_CHK_TNC: check TNC
- * UBIFS_CHK_IDX_SZ: check index size
- * UBIFS_CHK_ORPH: check orphans
- * UBIFS_CHK_OLD_IDX: check the old index
- * UBIFS_CHK_LPROPS: check lprops
- * UBIFS_CHK_FS: check the file-system
- */
-enum {
-       UBIFS_CHK_GEN     = 0x1,
-       UBIFS_CHK_TNC     = 0x2,
-       UBIFS_CHK_IDX_SZ  = 0x4,
-       UBIFS_CHK_ORPH    = 0x8,
-       UBIFS_CHK_OLD_IDX = 0x10,
-       UBIFS_CHK_LPROPS  = 0x20,
-       UBIFS_CHK_FS      = 0x40,
-};
-
-/*
- * Special testing flags (must match tst_names in debug.c).
- *
- * UBIFS_TST_FORCE_IN_THE_GAPS: force the use of in-the-gaps method
- * UBIFS_TST_RCVRY: failure mode for recovery testing
- */
-enum {
-       UBIFS_TST_FORCE_IN_THE_GAPS = 0x2,
-       UBIFS_TST_RCVRY             = 0x4,
-};
-
-#if CONFIG_UBIFS_FS_DEBUG_MSG_LVL == 1
-#define UBIFS_MSG_FLAGS_DEFAULT UBIFS_MSG_LVL_1
-#elif CONFIG_UBIFS_FS_DEBUG_MSG_LVL == 2
-#define UBIFS_MSG_FLAGS_DEFAULT UBIFS_MSG_LVL_2
-#elif CONFIG_UBIFS_FS_DEBUG_MSG_LVL == 3
-#define UBIFS_MSG_FLAGS_DEFAULT UBIFS_MSG_LVL_3
-#else
-#define UBIFS_MSG_FLAGS_DEFAULT UBIFS_MSG_LVL_0
-#endif
-
-#ifdef CONFIG_UBIFS_FS_DEBUG_CHKS
-#define UBIFS_CHK_FLAGS_DEFAULT 0xffffffff
+int ubifs_debugging_init(struct ubifs_info *c);
+void ubifs_debugging_exit(struct ubifs_info *c);
 #else
-#define UBIFS_CHK_FLAGS_DEFAULT 0
-#endif
-
-#define dbg_ntype(type)                       ""
-#define dbg_cstate(cmt_state)                 ""
-#define dbg_get_key_dump(c, key)              ({})
-#define dbg_dump_inode(c, inode)              ({})
-#define dbg_dump_node(c, node)                ({})
-#define dbg_dump_budget_req(req)              ({})
-#define dbg_dump_lstats(lst)                  ({})
-#define dbg_dump_budg(c)                      ({})
-#define dbg_dump_lprop(c, lp)                 ({})
-#define dbg_dump_lprops(c)                    ({})
-#define dbg_dump_lpt_info(c)                  ({})
-#define dbg_dump_leb(c, lnum)                 ({})
-#define dbg_dump_znode(c, znode)              ({})
-#define dbg_dump_heap(c, heap, cat)           ({})
-#define dbg_dump_pnode(c, pnode, parent, iip) ({})
-#define dbg_dump_tnc(c)                       ({})
-#define dbg_dump_index(c)                     ({})
-
-#define dbg_walk_index(c, leaf_cb, znode_cb, priv) 0
-#define dbg_old_index_check_init(c, zroot)         0
-#define dbg_check_old_index(c, zroot)              0
-#define dbg_check_cats(c)                          0
-#define dbg_check_ltab(c)                          0
-#define dbg_chk_lpt_free_spc(c)                    0
-#define dbg_chk_lpt_sz(c, action, len)             0
-#define dbg_check_synced_i_size(inode)             0
-#define dbg_check_dir_size(c, dir)                 0
-#define dbg_check_tnc(c, x)                        0
-#define dbg_check_idx_size(c, idx_size)            0
-#define dbg_check_filesystem(c)                    0
-#define dbg_check_heap(c, heap, cat, add_pos)      ({})
-#define dbg_check_lprops(c)                        0
-#define dbg_check_lpt_nodes(c, cnode, row, col)    0
-#define dbg_force_in_the_gaps_enabled              0
-#define dbg_force_in_the_gaps()                    0
-#define dbg_failure_mode                           0
-#define dbg_failure_mode_registration(c)           ({})
-#define dbg_failure_mode_deregistration(c)         ({})
+static inline int dbg_is_chk_gen(const struct ubifs_info *c)
+{
+       return 0;
+}
+static inline int dbg_is_chk_index(const struct ubifs_info *c)
+{
+       return 0;
+}
+static inline int dbg_is_chk_orph(const struct ubifs_info *c)
+{
+       return 0;
+}
+static inline int dbg_is_chk_lprops(const struct ubifs_info *c)
+{
+       return 0;
+}
+static inline int dbg_is_chk_fs(const struct ubifs_info *c)
+{
+       return 0;
+}
+static inline int dbg_is_tst_rcvry(const struct ubifs_info *c)
+{
+       return 0;
+}
+static inline int dbg_is_power_cut(const struct ubifs_info *c)
+{
+       return 0;
+}
 
 int ubifs_debugging_init(struct ubifs_info *c);
 void ubifs_debugging_exit(struct ubifs_info *c);
 
-#else /* !CONFIG_UBIFS_FS_DEBUG */
-
-#define UBIFS_DBG(op)
-
-/* Use "if (0)" to make compiler check arguments even if debugging is off */
-#define ubifs_assert(expr)  do {                                               \
-       if (0 && (expr))                                                       \
-               printk(KERN_CRIT "UBIFS assert failed in %s at %u (pid %d)\n", \
-                      __func__, __LINE__, 0);                      \
-} while (0)
-
-#define dbg_err(fmt, ...)   do {                                               \
-       if (0)                                                                 \
-               ubifs_err(fmt, ##__VA_ARGS__);                                 \
-} while (0)
-
-#define dbg_msg(fmt, ...) do {                                                 \
-       if (0)                                                                 \
-               printk(KERN_DEBUG "UBIFS DBG (pid %d): %s: " fmt "\n",         \
-                      0, __func__, ##__VA_ARGS__);                 \
-} while (0)
-
-#define dbg_dump_stack()
-#define ubifs_assert_cmt_locked(c)
-
-#define dbg_gen(fmt, ...)   dbg_msg(fmt, ##__VA_ARGS__)
-#define dbg_jnl(fmt, ...)   dbg_msg(fmt, ##__VA_ARGS__)
-#define dbg_tnc(fmt, ...)   dbg_msg(fmt, ##__VA_ARGS__)
-#define dbg_lp(fmt, ...)    dbg_msg(fmt, ##__VA_ARGS__)
-#define dbg_find(fmt, ...)  dbg_msg(fmt, ##__VA_ARGS__)
-#define dbg_mnt(fmt, ...)   dbg_msg(fmt, ##__VA_ARGS__)
-#define dbg_io(fmt, ...)    dbg_msg(fmt, ##__VA_ARGS__)
-#define dbg_cmt(fmt, ...)   dbg_msg(fmt, ##__VA_ARGS__)
-#define dbg_budg(fmt, ...)  dbg_msg(fmt, ##__VA_ARGS__)
-#define dbg_log(fmt, ...)   dbg_msg(fmt, ##__VA_ARGS__)
-#define dbg_gc(fmt, ...)    dbg_msg(fmt, ##__VA_ARGS__)
-#define dbg_scan(fmt, ...)  dbg_msg(fmt, ##__VA_ARGS__)
-#define dbg_rcvry(fmt, ...) dbg_msg(fmt, ##__VA_ARGS__)
-
-#define DBGKEY(key)  ((char *)(key))
-#define DBGKEY1(key) ((char *)(key))
-
-#define ubifs_debugging_init(c)                0
-#define ubifs_debugging_exit(c)                ({})
-
-#define dbg_ntype(type)                       ""
-#define dbg_cstate(cmt_state)                 ""
-#define dbg_get_key_dump(c, key)              ({})
-#define dbg_dump_inode(c, inode)              ({})
-#define dbg_dump_node(c, node)                ({})
-#define dbg_dump_budget_req(req)              ({})
-#define dbg_dump_lstats(lst)                  ({})
-#define dbg_dump_budg(c)                      ({})
-#define dbg_dump_lprop(c, lp)                 ({})
-#define dbg_dump_lprops(c)                    ({})
-#define dbg_dump_lpt_info(c)                  ({})
-#define dbg_dump_leb(c, lnum)                 ({})
-#define dbg_dump_znode(c, znode)              ({})
-#define dbg_dump_heap(c, heap, cat)           ({})
-#define dbg_dump_pnode(c, pnode, parent, iip) ({})
-#define dbg_dump_tnc(c)                       ({})
-#define dbg_dump_index(c)                     ({})
-
-#define dbg_walk_index(c, leaf_cb, znode_cb, priv) 0
-#define dbg_old_index_check_init(c, zroot)         0
-#define dbg_check_old_index(c, zroot)              0
-#define dbg_check_cats(c)                          0
-#define dbg_check_ltab(c)                          0
-#define dbg_chk_lpt_free_spc(c)                    0
-#define dbg_chk_lpt_sz(c, action, len)             0
-#define dbg_check_synced_i_size(inode)             0
-#define dbg_check_dir_size(c, dir)                 0
-#define dbg_check_tnc(c, x)                        0
-#define dbg_check_idx_size(c, idx_size)            0
-#define dbg_check_filesystem(c)                    0
-#define dbg_check_heap(c, heap, cat, add_pos)      ({})
-#define dbg_check_lprops(c)                        0
-#define dbg_check_lpt_nodes(c, cnode, row, col)    0
-#define dbg_force_in_the_gaps_enabled              0
-#define dbg_force_in_the_gaps()                    0
-#define dbg_failure_mode                           0
-#define dbg_failure_mode_registration(c)           ({})
-#define dbg_failure_mode_deregistration(c)         ({})
+#endif
 
-#endif /* !CONFIG_UBIFS_FS_DEBUG */
+/* Dump functions */
+const char *dbg_ntype(int type);
+const char *dbg_cstate(int cmt_state);
+const char *dbg_jhead(int jhead);
+const char *dbg_get_key_dump(const struct ubifs_info *c,
+                            const union ubifs_key *key);
+const char *dbg_snprintf_key(const struct ubifs_info *c,
+                            const union ubifs_key *key, char *buffer, int len);
+void ubifs_dump_inode(struct ubifs_info *c, const struct inode *inode);
+void ubifs_dump_node(const struct ubifs_info *c, const void *node);
+void ubifs_dump_budget_req(const struct ubifs_budget_req *req);
+void ubifs_dump_lstats(const struct ubifs_lp_stats *lst);
+void ubifs_dump_budg(struct ubifs_info *c, const struct ubifs_budg_info *bi);
+void ubifs_dump_lprop(const struct ubifs_info *c,
+                     const struct ubifs_lprops *lp);
+void ubifs_dump_lprops(struct ubifs_info *c);
+void ubifs_dump_lpt_info(struct ubifs_info *c);
+void ubifs_dump_leb(const struct ubifs_info *c, int lnum);
+void ubifs_dump_sleb(const struct ubifs_info *c,
+                    const struct ubifs_scan_leb *sleb, int offs);
+void ubifs_dump_znode(const struct ubifs_info *c,
+                     const struct ubifs_znode *znode);
+void ubifs_dump_heap(struct ubifs_info *c, struct ubifs_lpt_heap *heap,
+                    int cat);
+void ubifs_dump_pnode(struct ubifs_info *c, struct ubifs_pnode *pnode,
+                     struct ubifs_nnode *parent, int iip);
+void ubifs_dump_tnc(struct ubifs_info *c);
+void ubifs_dump_index(struct ubifs_info *c);
+void ubifs_dump_lpt_lebs(const struct ubifs_info *c);
+
+int dbg_walk_index(struct ubifs_info *c, dbg_leaf_callback leaf_cb,
+                  dbg_znode_callback znode_cb, void *priv);
+
+/* Checking functions */
+void dbg_save_space_info(struct ubifs_info *c);
+int dbg_check_space_info(struct ubifs_info *c);
+int dbg_check_lprops(struct ubifs_info *c);
+int dbg_old_index_check_init(struct ubifs_info *c, struct ubifs_zbranch *zroot);
+int dbg_check_old_index(struct ubifs_info *c, struct ubifs_zbranch *zroot);
+int dbg_check_cats(struct ubifs_info *c);
+int dbg_check_ltab(struct ubifs_info *c);
+int dbg_chk_lpt_free_spc(struct ubifs_info *c);
+int dbg_chk_lpt_sz(struct ubifs_info *c, int action, int len);
+int dbg_check_synced_i_size(const struct ubifs_info *c, struct inode *inode);
+int dbg_check_dir(struct ubifs_info *c, const struct inode *dir);
+int dbg_check_tnc(struct ubifs_info *c, int extra);
+int dbg_check_idx_size(struct ubifs_info *c, long long idx_size);
+int dbg_check_filesystem(struct ubifs_info *c);
+void dbg_check_heap(struct ubifs_info *c, struct ubifs_lpt_heap *heap, int cat,
+                   int add_pos);
+int dbg_check_lpt_nodes(struct ubifs_info *c, struct ubifs_cnode *cnode,
+                       int row, int col);
+int dbg_check_inode_size(struct ubifs_info *c, const struct inode *inode,
+                        loff_t size);
+int dbg_check_data_nodes_order(struct ubifs_info *c, struct list_head *head);
+int dbg_check_nondata_nodes_order(struct ubifs_info *c, struct list_head *head);
+
+int dbg_leb_write(struct ubifs_info *c, int lnum, const void *buf, int offs,
+                 int len);
+int dbg_leb_change(struct ubifs_info *c, int lnum, const void *buf, int len);
+int dbg_leb_unmap(struct ubifs_info *c, int lnum);
+int dbg_leb_map(struct ubifs_info *c, int lnum);
+
+/* Debugfs-related stuff */
+int dbg_debugfs_init(void);
+void dbg_debugfs_exit(void);
+int dbg_debugfs_init_fs(struct ubifs_info *c);
+void dbg_debugfs_exit_fs(struct ubifs_info *c);
 
 #endif /* !__UBIFS_DEBUG_H__ */
index aae5c65eaeeedec16264b2b3e37b18914eba6969..f87341e1087a4b5c7ab0682ea36d4af89e20b9d7 100644 (file)
@@ -4,18 +4,7 @@
  * Copyright (C) 2006-2008 Nokia Corporation.
  * Copyright (C) 2006, 2007 University of Szeged, Hungary
  *
- * This program is free software; you can redistribute it and/or modify it
- * under the terms of the GNU General Public License version 2 as published by
- * the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
- * more details.
- *
- * You should have received a copy of the GNU General Public License along with
- * this program; if not, write to the Free Software Foundation, Inc., 51
- * Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
+ * SPDX-License-Identifier:    GPL-2.0+
  *
  * Authors: Artem Bityutskiy (Битюцкий Артём)
  *          Adrian Hunter
  * buffer is full or when it is not used for some time (by timer). This is
  * similar to the mechanism is used by JFFS2.
  *
+ * UBIFS distinguishes between minimum write size (@c->min_io_size) and maximum
+ * write size (@c->max_write_size). The latter is the maximum amount of bytes
+ * the underlying flash is able to program at a time, and writing in
+ * @c->max_write_size units should presumably be faster. Obviously,
+ * @c->min_io_size <= @c->max_write_size. Write-buffers are of
+ * @c->max_write_size bytes in size for maximum performance. However, when a
+ * write-buffer is flushed, only the portion of it (aligned to @c->min_io_size
+ * boundary) which contains data is written, not the whole write-buffer,
+ * because this is more space-efficient.
+ *
+ * This optimization adds few complications to the code. Indeed, on the one
+ * hand, we want to write in optimal @c->max_write_size bytes chunks, which
+ * also means aligning writes at the @c->max_write_size bytes offsets. On the
+ * other hand, we do not want to waste space when synchronizing the write
+ * buffer, so during synchronization we writes in smaller chunks. And this makes
+ * the next write offset to be not aligned to @c->max_write_size bytes. So the
+ * have to make sure that the write-buffer offset (@wbuf->offs) becomes aligned
+ * to @c->max_write_size bytes again. We do this by temporarily shrinking
+ * write-buffer size (@wbuf->size).
+ *
  * Write-buffers are defined by 'struct ubifs_wbuf' objects and protected by
  * mutexes defined inside these objects. Since sometimes upper-level code
  * has to lock the write-buffer (e.g. journal space reservation code), many
  * UBIFS uses padding when it pads to the next min. I/O unit. In this case it
  * uses padding nodes or padding bytes, if the padding node does not fit.
  *
- * All UBIFS nodes are protected by CRC checksums and UBIFS checks all nodes
- * every time they are read from the flash media.
+ * All UBIFS nodes are protected by CRC checksums and UBIFS checks CRC when
+ * they are read from the flash media.
  */
 
+#define __UBOOT__
+#ifndef __UBOOT__
+#include <linux/crc32.h>
+#include <linux/slab.h>
+#else
+#include <linux/compat.h>
+#include <linux/err.h>
+#endif
 #include "ubifs.h"
 
 /**
  */
 void ubifs_ro_mode(struct ubifs_info *c, int err)
 {
-       if (!c->ro_media) {
-               c->ro_media = 1;
+       if (!c->ro_error) {
+               c->ro_error = 1;
                c->no_chk_data_crc = 0;
+               c->vfs_sb->s_flags |= MS_RDONLY;
                ubifs_warn("switched to read-only mode, error %d", err);
-               dbg_dump_stack();
+               dump_stack();
+       }
+}
+
+/*
+ * Below are simple wrappers over UBI I/O functions which include some
+ * additional checks and UBIFS debugging stuff. See corresponding UBI function
+ * for more information.
+ */
+
+int ubifs_leb_read(const struct ubifs_info *c, int lnum, void *buf, int offs,
+                  int len, int even_ebadmsg)
+{
+       int err;
+
+       err = ubi_read(c->ubi, lnum, buf, offs, len);
+       /*
+        * In case of %-EBADMSG print the error message only if the
+        * @even_ebadmsg is true.
+        */
+       if (err && (err != -EBADMSG || even_ebadmsg)) {
+               ubifs_err("reading %d bytes from LEB %d:%d failed, error %d",
+                         len, lnum, offs, err);
+               dump_stack();
+       }
+       return err;
+}
+
+int ubifs_leb_write(struct ubifs_info *c, int lnum, const void *buf, int offs,
+                   int len)
+{
+       int err;
+
+       ubifs_assert(!c->ro_media && !c->ro_mount);
+       if (c->ro_error)
+               return -EROFS;
+       if (!dbg_is_tst_rcvry(c))
+               err = ubi_leb_write(c->ubi, lnum, buf, offs, len);
+       else
+               err = dbg_leb_write(c, lnum, buf, offs, len);
+       if (err) {
+               ubifs_err("writing %d bytes to LEB %d:%d failed, error %d",
+                         len, lnum, offs, err);
+               ubifs_ro_mode(c, err);
+               dump_stack();
+       }
+       return err;
+}
+
+int ubifs_leb_change(struct ubifs_info *c, int lnum, const void *buf, int len)
+{
+       int err;
+
+       ubifs_assert(!c->ro_media && !c->ro_mount);
+       if (c->ro_error)
+               return -EROFS;
+       if (!dbg_is_tst_rcvry(c))
+               err = ubi_leb_change(c->ubi, lnum, buf, len);
+       else
+               err = dbg_leb_change(c, lnum, buf, len);
+       if (err) {
+               ubifs_err("changing %d bytes in LEB %d failed, error %d",
+                         len, lnum, err);
+               ubifs_ro_mode(c, err);
+               dump_stack();
        }
+       return err;
+}
+
+int ubifs_leb_unmap(struct ubifs_info *c, int lnum)
+{
+       int err;
+
+       ubifs_assert(!c->ro_media && !c->ro_mount);
+       if (c->ro_error)
+               return -EROFS;
+       if (!dbg_is_tst_rcvry(c))
+               err = ubi_leb_unmap(c->ubi, lnum);
+       else
+               err = dbg_leb_unmap(c, lnum);
+       if (err) {
+               ubifs_err("unmap LEB %d failed, error %d", lnum, err);
+               ubifs_ro_mode(c, err);
+               dump_stack();
+       }
+       return err;
+}
+
+int ubifs_leb_map(struct ubifs_info *c, int lnum)
+{
+       int err;
+
+       ubifs_assert(!c->ro_media && !c->ro_mount);
+       if (c->ro_error)
+               return -EROFS;
+       if (!dbg_is_tst_rcvry(c))
+               err = ubi_leb_map(c->ubi, lnum);
+       else
+               err = dbg_leb_map(c, lnum);
+       if (err) {
+               ubifs_err("mapping LEB %d failed, error %d", lnum, err);
+               ubifs_ro_mode(c, err);
+               dump_stack();
+       }
+       return err;
+}
+
+int ubifs_is_mapped(const struct ubifs_info *c, int lnum)
+{
+       int err;
+
+       err = ubi_is_mapped(c->ubi, lnum);
+       if (err < 0) {
+               ubifs_err("ubi_is_mapped failed for LEB %d, error %d",
+                         lnum, err);
+               dump_stack();
+       }
+       return err;
 }
 
 /**
@@ -85,8 +219,12 @@ void ubifs_ro_mode(struct ubifs_info *c, int err)
  * This function may skip data nodes CRC checking if @c->no_chk_data_crc is
  * true, which is controlled by corresponding UBIFS mount option. However, if
  * @must_chk_crc is true, then @c->no_chk_data_crc is ignored and CRC is
- * checked. Similarly, if @c->always_chk_crc is true, @c->no_chk_data_crc is
- * ignored and CRC is checked.
+ * checked. Similarly, if @c->mounting or @c->remounting_rw is true (we are
+ * mounting or re-mounting to R/W mode), @c->no_chk_data_crc is ignored and CRC
+ * is checked. This is because during mounting or re-mounting from R/O mode to
+ * R/W mode we may read journal nodes (when replying the journal or doing the
+ * recovery) and the journal nodes may potentially be corrupted, so checking is
+ * required.
  *
  * This function returns zero in case of success and %-EUCLEAN in case of bad
  * CRC or magic.
@@ -128,8 +266,8 @@ int ubifs_check_node(const struct ubifs_info *c, const void *buf, int lnum,
                   node_len > c->ranges[type].max_len)
                goto out_len;
 
-       if (!must_chk_crc && type == UBIFS_DATA_NODE && !c->always_chk_crc &&
-            c->no_chk_data_crc)
+       if (!must_chk_crc && type == UBIFS_DATA_NODE && !c->mounting &&
+           !c->remounting_rw && c->no_chk_data_crc)
                return 0;
 
        crc = crc32(UBIFS_CRC32_INIT, buf + 8, node_len - 8);
@@ -150,8 +288,8 @@ out_len:
 out:
        if (!quiet) {
                ubifs_err("bad node at LEB %d:%d", lnum, offs);
-               dbg_dump_node(c, buf);
-               dbg_dump_stack();
+               ubifs_dump_node(c, buf);
+               dump_stack();
        }
        return err;
 }
@@ -256,6 +394,571 @@ void ubifs_prepare_node(struct ubifs_info *c, void *node, int len, int pad)
        }
 }
 
+/**
+ * ubifs_prep_grp_node - prepare node of a group to be written to flash.
+ * @c: UBIFS file-system description object
+ * @node: the node to pad
+ * @len: node length
+ * @last: indicates the last node of the group
+ *
+ * This function prepares node at @node to be written to the media - it
+ * calculates node CRC and fills the common header.
+ */
+void ubifs_prep_grp_node(struct ubifs_info *c, void *node, int len, int last)
+{
+       uint32_t crc;
+       struct ubifs_ch *ch = node;
+       unsigned long long sqnum = next_sqnum(c);
+
+       ubifs_assert(len >= UBIFS_CH_SZ);
+
+       ch->magic = cpu_to_le32(UBIFS_NODE_MAGIC);
+       ch->len = cpu_to_le32(len);
+       if (last)
+               ch->group_type = UBIFS_LAST_OF_NODE_GROUP;
+       else
+               ch->group_type = UBIFS_IN_NODE_GROUP;
+       ch->sqnum = cpu_to_le64(sqnum);
+       ch->padding[0] = ch->padding[1] = 0;
+       crc = crc32(UBIFS_CRC32_INIT, node + 8, len - 8);
+       ch->crc = cpu_to_le32(crc);
+}
+
+#ifndef __UBOOT__
+/**
+ * wbuf_timer_callback - write-buffer timer callback function.
+ * @data: timer data (write-buffer descriptor)
+ *
+ * This function is called when the write-buffer timer expires.
+ */
+static enum hrtimer_restart wbuf_timer_callback_nolock(struct hrtimer *timer)
+{
+       struct ubifs_wbuf *wbuf = container_of(timer, struct ubifs_wbuf, timer);
+
+       dbg_io("jhead %s", dbg_jhead(wbuf->jhead));
+       wbuf->need_sync = 1;
+       wbuf->c->need_wbuf_sync = 1;
+       ubifs_wake_up_bgt(wbuf->c);
+       return HRTIMER_NORESTART;
+}
+
+/**
+ * new_wbuf_timer - start new write-buffer timer.
+ * @wbuf: write-buffer descriptor
+ */
+static void new_wbuf_timer_nolock(struct ubifs_wbuf *wbuf)
+{
+       ubifs_assert(!hrtimer_active(&wbuf->timer));
+
+       if (wbuf->no_timer)
+               return;
+       dbg_io("set timer for jhead %s, %llu-%llu millisecs",
+              dbg_jhead(wbuf->jhead),
+              div_u64(ktime_to_ns(wbuf->softlimit), USEC_PER_SEC),
+              div_u64(ktime_to_ns(wbuf->softlimit) + wbuf->delta,
+                      USEC_PER_SEC));
+       hrtimer_start_range_ns(&wbuf->timer, wbuf->softlimit, wbuf->delta,
+                              HRTIMER_MODE_REL);
+}
+#endif
+
+/**
+ * cancel_wbuf_timer - cancel write-buffer timer.
+ * @wbuf: write-buffer descriptor
+ */
+static void cancel_wbuf_timer_nolock(struct ubifs_wbuf *wbuf)
+{
+       if (wbuf->no_timer)
+               return;
+       wbuf->need_sync = 0;
+#ifndef __UBOOT__
+       hrtimer_cancel(&wbuf->timer);
+#endif
+}
+
+/**
+ * ubifs_wbuf_sync_nolock - synchronize write-buffer.
+ * @wbuf: write-buffer to synchronize
+ *
+ * This function synchronizes write-buffer @buf and returns zero in case of
+ * success or a negative error code in case of failure.
+ *
+ * Note, although write-buffers are of @c->max_write_size, this function does
+ * not necessarily writes all @c->max_write_size bytes to the flash. Instead,
+ * if the write-buffer is only partially filled with data, only the used part
+ * of the write-buffer (aligned on @c->min_io_size boundary) is synchronized.
+ * This way we waste less space.
+ */
+int ubifs_wbuf_sync_nolock(struct ubifs_wbuf *wbuf)
+{
+       struct ubifs_info *c = wbuf->c;
+       int err, dirt, sync_len;
+
+       cancel_wbuf_timer_nolock(wbuf);
+       if (!wbuf->used || wbuf->lnum == -1)
+               /* Write-buffer is empty or not seeked */
+               return 0;
+
+       dbg_io("LEB %d:%d, %d bytes, jhead %s",
+              wbuf->lnum, wbuf->offs, wbuf->used, dbg_jhead(wbuf->jhead));
+       ubifs_assert(!(wbuf->avail & 7));
+       ubifs_assert(wbuf->offs + wbuf->size <= c->leb_size);
+       ubifs_assert(wbuf->size >= c->min_io_size);
+       ubifs_assert(wbuf->size <= c->max_write_size);
+       ubifs_assert(wbuf->size % c->min_io_size == 0);
+       ubifs_assert(!c->ro_media && !c->ro_mount);
+       if (c->leb_size - wbuf->offs >= c->max_write_size)
+               ubifs_assert(!((wbuf->offs + wbuf->size) % c->max_write_size));
+
+       if (c->ro_error)
+               return -EROFS;
+
+       /*
+        * Do not write whole write buffer but write only the minimum necessary
+        * amount of min. I/O units.
+        */
+       sync_len = ALIGN(wbuf->used, c->min_io_size);
+       dirt = sync_len - wbuf->used;
+       if (dirt)
+               ubifs_pad(c, wbuf->buf + wbuf->used, dirt);
+       err = ubifs_leb_write(c, wbuf->lnum, wbuf->buf, wbuf->offs, sync_len);
+       if (err)
+               return err;
+
+       spin_lock(&wbuf->lock);
+       wbuf->offs += sync_len;
+       /*
+        * Now @wbuf->offs is not necessarily aligned to @c->max_write_size.
+        * But our goal is to optimize writes and make sure we write in
+        * @c->max_write_size chunks and to @c->max_write_size-aligned offset.
+        * Thus, if @wbuf->offs is not aligned to @c->max_write_size now, make
+        * sure that @wbuf->offs + @wbuf->size is aligned to
+        * @c->max_write_size. This way we make sure that after next
+        * write-buffer flush we are again at the optimal offset (aligned to
+        * @c->max_write_size).
+        */
+       if (c->leb_size - wbuf->offs < c->max_write_size)
+               wbuf->size = c->leb_size - wbuf->offs;
+       else if (wbuf->offs & (c->max_write_size - 1))
+               wbuf->size = ALIGN(wbuf->offs, c->max_write_size) - wbuf->offs;
+       else
+               wbuf->size = c->max_write_size;
+       wbuf->avail = wbuf->size;
+       wbuf->used = 0;
+       wbuf->next_ino = 0;
+       spin_unlock(&wbuf->lock);
+
+       if (wbuf->sync_callback)
+               err = wbuf->sync_callback(c, wbuf->lnum,
+                                         c->leb_size - wbuf->offs, dirt);
+       return err;
+}
+
+/**
+ * ubifs_wbuf_seek_nolock - seek write-buffer.
+ * @wbuf: write-buffer
+ * @lnum: logical eraseblock number to seek to
+ * @offs: logical eraseblock offset to seek to
+ *
+ * This function targets the write-buffer to logical eraseblock @lnum:@offs.
+ * The write-buffer has to be empty. Returns zero in case of success and a
+ * negative error code in case of failure.
+ */
+int ubifs_wbuf_seek_nolock(struct ubifs_wbuf *wbuf, int lnum, int offs)
+{
+       const struct ubifs_info *c = wbuf->c;
+
+       dbg_io("LEB %d:%d, jhead %s", lnum, offs, dbg_jhead(wbuf->jhead));
+       ubifs_assert(lnum >= 0 && lnum < c->leb_cnt);
+       ubifs_assert(offs >= 0 && offs <= c->leb_size);
+       ubifs_assert(offs % c->min_io_size == 0 && !(offs & 7));
+       ubifs_assert(lnum != wbuf->lnum);
+       ubifs_assert(wbuf->used == 0);
+
+       spin_lock(&wbuf->lock);
+       wbuf->lnum = lnum;
+       wbuf->offs = offs;
+       if (c->leb_size - wbuf->offs < c->max_write_size)
+               wbuf->size = c->leb_size - wbuf->offs;
+       else if (wbuf->offs & (c->max_write_size - 1))
+               wbuf->size = ALIGN(wbuf->offs, c->max_write_size) - wbuf->offs;
+       else
+               wbuf->size = c->max_write_size;
+       wbuf->avail = wbuf->size;
+       wbuf->used = 0;
+       spin_unlock(&wbuf->lock);
+
+       return 0;
+}
+
+#ifndef __UBOOT__
+/**
+ * ubifs_bg_wbufs_sync - synchronize write-buffers.
+ * @c: UBIFS file-system description object
+ *
+ * This function is called by background thread to synchronize write-buffers.
+ * Returns zero in case of success and a negative error code in case of
+ * failure.
+ */
+int ubifs_bg_wbufs_sync(struct ubifs_info *c)
+{
+       int err, i;
+
+       ubifs_assert(!c->ro_media && !c->ro_mount);
+       if (!c->need_wbuf_sync)
+               return 0;
+       c->need_wbuf_sync = 0;
+
+       if (c->ro_error) {
+               err = -EROFS;
+               goto out_timers;
+       }
+
+       dbg_io("synchronize");
+       for (i = 0; i < c->jhead_cnt; i++) {
+               struct ubifs_wbuf *wbuf = &c->jheads[i].wbuf;
+
+               cond_resched();
+
+               /*
+                * If the mutex is locked then wbuf is being changed, so
+                * synchronization is not necessary.
+                */
+               if (mutex_is_locked(&wbuf->io_mutex))
+                       continue;
+
+               mutex_lock_nested(&wbuf->io_mutex, wbuf->jhead);
+               if (!wbuf->need_sync) {
+                       mutex_unlock(&wbuf->io_mutex);
+                       continue;
+               }
+
+               err = ubifs_wbuf_sync_nolock(wbuf);
+               mutex_unlock(&wbuf->io_mutex);
+               if (err) {
+                       ubifs_err("cannot sync write-buffer, error %d", err);
+                       ubifs_ro_mode(c, err);
+                       goto out_timers;
+               }
+       }
+
+       return 0;
+
+out_timers:
+       /* Cancel all timers to prevent repeated errors */
+       for (i = 0; i < c->jhead_cnt; i++) {
+               struct ubifs_wbuf *wbuf = &c->jheads[i].wbuf;
+
+               mutex_lock_nested(&wbuf->io_mutex, wbuf->jhead);
+               cancel_wbuf_timer_nolock(wbuf);
+               mutex_unlock(&wbuf->io_mutex);
+       }
+       return err;
+}
+
+/**
+ * ubifs_wbuf_write_nolock - write data to flash via write-buffer.
+ * @wbuf: write-buffer
+ * @buf: node to write
+ * @len: node length
+ *
+ * This function writes data to flash via write-buffer @wbuf. This means that
+ * the last piece of the node won't reach the flash media immediately if it
+ * does not take whole max. write unit (@c->max_write_size). Instead, the node
+ * will sit in RAM until the write-buffer is synchronized (e.g., by timer, or
+ * because more data are appended to the write-buffer).
+ *
+ * This function returns zero in case of success and a negative error code in
+ * case of failure. If the node cannot be written because there is no more
+ * space in this logical eraseblock, %-ENOSPC is returned.
+ */
+int ubifs_wbuf_write_nolock(struct ubifs_wbuf *wbuf, void *buf, int len)
+{
+       struct ubifs_info *c = wbuf->c;
+       int err, written, n, aligned_len = ALIGN(len, 8);
+
+       dbg_io("%d bytes (%s) to jhead %s wbuf at LEB %d:%d", len,
+              dbg_ntype(((struct ubifs_ch *)buf)->node_type),
+              dbg_jhead(wbuf->jhead), wbuf->lnum, wbuf->offs + wbuf->used);
+       ubifs_assert(len > 0 && wbuf->lnum >= 0 && wbuf->lnum < c->leb_cnt);
+       ubifs_assert(wbuf->offs >= 0 && wbuf->offs % c->min_io_size == 0);
+       ubifs_assert(!(wbuf->offs & 7) && wbuf->offs <= c->leb_size);
+       ubifs_assert(wbuf->avail > 0 && wbuf->avail <= wbuf->size);
+       ubifs_assert(wbuf->size >= c->min_io_size);
+       ubifs_assert(wbuf->size <= c->max_write_size);
+       ubifs_assert(wbuf->size % c->min_io_size == 0);
+       ubifs_assert(mutex_is_locked(&wbuf->io_mutex));
+       ubifs_assert(!c->ro_media && !c->ro_mount);
+       ubifs_assert(!c->space_fixup);
+       if (c->leb_size - wbuf->offs >= c->max_write_size)
+               ubifs_assert(!((wbuf->offs + wbuf->size) % c->max_write_size));
+
+       if (c->leb_size - wbuf->offs - wbuf->used < aligned_len) {
+               err = -ENOSPC;
+               goto out;
+       }
+
+       cancel_wbuf_timer_nolock(wbuf);
+
+       if (c->ro_error)
+               return -EROFS;
+
+       if (aligned_len <= wbuf->avail) {
+               /*
+                * The node is not very large and fits entirely within
+                * write-buffer.
+                */
+               memcpy(wbuf->buf + wbuf->used, buf, len);
+
+               if (aligned_len == wbuf->avail) {
+                       dbg_io("flush jhead %s wbuf to LEB %d:%d",
+                              dbg_jhead(wbuf->jhead), wbuf->lnum, wbuf->offs);
+                       err = ubifs_leb_write(c, wbuf->lnum, wbuf->buf,
+                                             wbuf->offs, wbuf->size);
+                       if (err)
+                               goto out;
+
+                       spin_lock(&wbuf->lock);
+                       wbuf->offs += wbuf->size;
+                       if (c->leb_size - wbuf->offs >= c->max_write_size)
+                               wbuf->size = c->max_write_size;
+                       else
+                               wbuf->size = c->leb_size - wbuf->offs;
+                       wbuf->avail = wbuf->size;
+                       wbuf->used = 0;
+                       wbuf->next_ino = 0;
+                       spin_unlock(&wbuf->lock);
+               } else {
+                       spin_lock(&wbuf->lock);
+                       wbuf->avail -= aligned_len;
+                       wbuf->used += aligned_len;
+                       spin_unlock(&wbuf->lock);
+               }
+
+               goto exit;
+       }
+
+       written = 0;
+
+       if (wbuf->used) {
+               /*
+                * The node is large enough and does not fit entirely within
+                * current available space. We have to fill and flush
+                * write-buffer and switch to the next max. write unit.
+                */
+               dbg_io("flush jhead %s wbuf to LEB %d:%d",
+                      dbg_jhead(wbuf->jhead), wbuf->lnum, wbuf->offs);
+               memcpy(wbuf->buf + wbuf->used, buf, wbuf->avail);
+               err = ubifs_leb_write(c, wbuf->lnum, wbuf->buf, wbuf->offs,
+                                     wbuf->size);
+               if (err)
+                       goto out;
+
+               wbuf->offs += wbuf->size;
+               len -= wbuf->avail;
+               aligned_len -= wbuf->avail;
+               written += wbuf->avail;
+       } else if (wbuf->offs & (c->max_write_size - 1)) {
+               /*
+                * The write-buffer offset is not aligned to
+                * @c->max_write_size and @wbuf->size is less than
+                * @c->max_write_size. Write @wbuf->size bytes to make sure the
+                * following writes are done in optimal @c->max_write_size
+                * chunks.
+                */
+               dbg_io("write %d bytes to LEB %d:%d",
+                      wbuf->size, wbuf->lnum, wbuf->offs);
+               err = ubifs_leb_write(c, wbuf->lnum, buf, wbuf->offs,
+                                     wbuf->size);
+               if (err)
+                       goto out;
+
+               wbuf->offs += wbuf->size;
+               len -= wbuf->size;
+               aligned_len -= wbuf->size;
+               written += wbuf->size;
+       }
+
+       /*
+        * The remaining data may take more whole max. write units, so write the
+        * remains multiple to max. write unit size directly to the flash media.
+        * We align node length to 8-byte boundary because we anyway flash wbuf
+        * if the remaining space is less than 8 bytes.
+        */
+       n = aligned_len >> c->max_write_shift;
+       if (n) {
+               n <<= c->max_write_shift;
+               dbg_io("write %d bytes to LEB %d:%d", n, wbuf->lnum,
+                      wbuf->offs);
+               err = ubifs_leb_write(c, wbuf->lnum, buf + written,
+                                     wbuf->offs, n);
+               if (err)
+                       goto out;
+               wbuf->offs += n;
+               aligned_len -= n;
+               len -= n;
+               written += n;
+       }
+
+       spin_lock(&wbuf->lock);
+       if (aligned_len)
+               /*
+                * And now we have what's left and what does not take whole
+                * max. write unit, so write it to the write-buffer and we are
+                * done.
+                */
+               memcpy(wbuf->buf, buf + written, len);
+
+       if (c->leb_size - wbuf->offs >= c->max_write_size)
+               wbuf->size = c->max_write_size;
+       else
+               wbuf->size = c->leb_size - wbuf->offs;
+       wbuf->avail = wbuf->size - aligned_len;
+       wbuf->used = aligned_len;
+       wbuf->next_ino = 0;
+       spin_unlock(&wbuf->lock);
+
+exit:
+       if (wbuf->sync_callback) {
+               int free = c->leb_size - wbuf->offs - wbuf->used;
+
+               err = wbuf->sync_callback(c, wbuf->lnum, free, 0);
+               if (err)
+                       goto out;
+       }
+
+       if (wbuf->used)
+               new_wbuf_timer_nolock(wbuf);
+
+       return 0;
+
+out:
+       ubifs_err("cannot write %d bytes to LEB %d:%d, error %d",
+                 len, wbuf->lnum, wbuf->offs, err);
+       ubifs_dump_node(c, buf);
+       dump_stack();
+       ubifs_dump_leb(c, wbuf->lnum);
+       return err;
+}
+
+/**
+ * ubifs_write_node - write node to the media.
+ * @c: UBIFS file-system description object
+ * @buf: the node to write
+ * @len: node length
+ * @lnum: logical eraseblock number
+ * @offs: offset within the logical eraseblock
+ *
+ * This function automatically fills node magic number, assigns sequence
+ * number, and calculates node CRC checksum. The length of the @buf buffer has
+ * to be aligned to the minimal I/O unit size. This function automatically
+ * appends padding node and padding bytes if needed. Returns zero in case of
+ * success and a negative error code in case of failure.
+ */
+int ubifs_write_node(struct ubifs_info *c, void *buf, int len, int lnum,
+                    int offs)
+{
+       int err, buf_len = ALIGN(len, c->min_io_size);
+
+       dbg_io("LEB %d:%d, %s, length %d (aligned %d)",
+              lnum, offs, dbg_ntype(((struct ubifs_ch *)buf)->node_type), len,
+              buf_len);
+       ubifs_assert(lnum >= 0 && lnum < c->leb_cnt && offs >= 0);
+       ubifs_assert(offs % c->min_io_size == 0 && offs < c->leb_size);
+       ubifs_assert(!c->ro_media && !c->ro_mount);
+       ubifs_assert(!c->space_fixup);
+
+       if (c->ro_error)
+               return -EROFS;
+
+       ubifs_prepare_node(c, buf, len, 1);
+       err = ubifs_leb_write(c, lnum, buf, offs, buf_len);
+       if (err)
+               ubifs_dump_node(c, buf);
+
+       return err;
+}
+#endif
+
+/**
+ * ubifs_read_node_wbuf - read node from the media or write-buffer.
+ * @wbuf: wbuf to check for un-written data
+ * @buf: buffer to read to
+ * @type: node type
+ * @len: node length
+ * @lnum: logical eraseblock number
+ * @offs: offset within the logical eraseblock
+ *
+ * This function reads a node of known type and length, checks it and stores
+ * in @buf. If the node partially or fully sits in the write-buffer, this
+ * function takes data from the buffer, otherwise it reads the flash media.
+ * Returns zero in case of success, %-EUCLEAN if CRC mismatched and a negative
+ * error code in case of failure.
+ */
+int ubifs_read_node_wbuf(struct ubifs_wbuf *wbuf, void *buf, int type, int len,
+                        int lnum, int offs)
+{
+       const struct ubifs_info *c = wbuf->c;
+       int err, rlen, overlap;
+       struct ubifs_ch *ch = buf;
+
+       dbg_io("LEB %d:%d, %s, length %d, jhead %s", lnum, offs,
+              dbg_ntype(type), len, dbg_jhead(wbuf->jhead));
+       ubifs_assert(wbuf && lnum >= 0 && lnum < c->leb_cnt && offs >= 0);
+       ubifs_assert(!(offs & 7) && offs < c->leb_size);
+       ubifs_assert(type >= 0 && type < UBIFS_NODE_TYPES_CNT);
+
+       spin_lock(&wbuf->lock);
+       overlap = (lnum == wbuf->lnum && offs + len > wbuf->offs);
+       if (!overlap) {
+               /* We may safely unlock the write-buffer and read the data */
+               spin_unlock(&wbuf->lock);
+               return ubifs_read_node(c, buf, type, len, lnum, offs);
+       }
+
+       /* Don't read under wbuf */
+       rlen = wbuf->offs - offs;
+       if (rlen < 0)
+               rlen = 0;
+
+       /* Copy the rest from the write-buffer */
+       memcpy(buf + rlen, wbuf->buf + offs + rlen - wbuf->offs, len - rlen);
+       spin_unlock(&wbuf->lock);
+
+       if (rlen > 0) {
+               /* Read everything that goes before write-buffer */
+               err = ubifs_leb_read(c, lnum, buf, offs, rlen, 0);
+               if (err && err != -EBADMSG)
+                       return err;
+       }
+
+       if (type != ch->node_type) {
+               ubifs_err("bad node type (%d but expected %d)",
+                         ch->node_type, type);
+               goto out;
+       }
+
+       err = ubifs_check_node(c, buf, lnum, offs, 0, 0);
+       if (err) {
+               ubifs_err("expected node type %d", type);
+               return err;
+       }
+
+       rlen = le32_to_cpu(ch->len);
+       if (rlen != len) {
+               ubifs_err("bad node length %d, expected %d", rlen, len);
+               goto out;
+       }
+
+       return 0;
+
+out:
+       ubifs_err("bad node at LEB %d:%d", lnum, offs);
+       ubifs_dump_node(c, buf);
+       dump_stack();
+       return -EINVAL;
+}
+
 /**
  * ubifs_read_node - read node.
  * @c: UBIFS file-system description object
@@ -281,12 +984,9 @@ int ubifs_read_node(const struct ubifs_info *c, void *buf, int type, int len,
        ubifs_assert(!(offs & 7) && offs < c->leb_size);
        ubifs_assert(type >= 0 && type < UBIFS_NODE_TYPES_CNT);
 
-       err = ubi_read(c->ubi, lnum, buf, offs, len);
-       if (err && err != -EBADMSG) {
-               ubifs_err("cannot read node %d from LEB %d:%d, error %d",
-                         type, lnum, offs, err);
+       err = ubifs_leb_read(c, lnum, buf, offs, len, 0);
+       if (err && err != -EBADMSG)
                return err;
-       }
 
        if (type != ch->node_type) {
                ubifs_err("bad node type (%d but expected %d)",
@@ -309,8 +1009,143 @@ int ubifs_read_node(const struct ubifs_info *c, void *buf, int type, int len,
        return 0;
 
 out:
-       ubifs_err("bad node at LEB %d:%d", lnum, offs);
-       dbg_dump_node(c, buf);
-       dbg_dump_stack();
+       ubifs_err("bad node at LEB %d:%d, LEB mapping status %d", lnum, offs,
+                 ubi_is_mapped(c->ubi, lnum));
+       ubifs_dump_node(c, buf);
+       dump_stack();
        return -EINVAL;
 }
+
+/**
+ * ubifs_wbuf_init - initialize write-buffer.
+ * @c: UBIFS file-system description object
+ * @wbuf: write-buffer to initialize
+ *
+ * This function initializes write-buffer. Returns zero in case of success
+ * %-ENOMEM in case of failure.
+ */
+int ubifs_wbuf_init(struct ubifs_info *c, struct ubifs_wbuf *wbuf)
+{
+       size_t size;
+
+       wbuf->buf = kmalloc(c->max_write_size, GFP_KERNEL);
+       if (!wbuf->buf)
+               return -ENOMEM;
+
+       size = (c->max_write_size / UBIFS_CH_SZ + 1) * sizeof(ino_t);
+       wbuf->inodes = kmalloc(size, GFP_KERNEL);
+       if (!wbuf->inodes) {
+               kfree(wbuf->buf);
+               wbuf->buf = NULL;
+               return -ENOMEM;
+       }
+
+       wbuf->used = 0;
+       wbuf->lnum = wbuf->offs = -1;
+       /*
+        * If the LEB starts at the max. write size aligned address, then
+        * write-buffer size has to be set to @c->max_write_size. Otherwise,
+        * set it to something smaller so that it ends at the closest max.
+        * write size boundary.
+        */
+       size = c->max_write_size - (c->leb_start % c->max_write_size);
+       wbuf->avail = wbuf->size = size;
+       wbuf->sync_callback = NULL;
+       mutex_init(&wbuf->io_mutex);
+       spin_lock_init(&wbuf->lock);
+       wbuf->c = c;
+       wbuf->next_ino = 0;
+
+#ifndef __UBOOT__
+       hrtimer_init(&wbuf->timer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
+       wbuf->timer.function = wbuf_timer_callback_nolock;
+       wbuf->softlimit = ktime_set(WBUF_TIMEOUT_SOFTLIMIT, 0);
+       wbuf->delta = WBUF_TIMEOUT_HARDLIMIT - WBUF_TIMEOUT_SOFTLIMIT;
+       wbuf->delta *= 1000000000ULL;
+       ubifs_assert(wbuf->delta <= ULONG_MAX);
+#endif
+       return 0;
+}
+
+/**
+ * ubifs_wbuf_add_ino_nolock - add an inode number into the wbuf inode array.
+ * @wbuf: the write-buffer where to add
+ * @inum: the inode number
+ *
+ * This function adds an inode number to the inode array of the write-buffer.
+ */
+void ubifs_wbuf_add_ino_nolock(struct ubifs_wbuf *wbuf, ino_t inum)
+{
+       if (!wbuf->buf)
+               /* NOR flash or something similar */
+               return;
+
+       spin_lock(&wbuf->lock);
+       if (wbuf->used)
+               wbuf->inodes[wbuf->next_ino++] = inum;
+       spin_unlock(&wbuf->lock);
+}
+
+/**
+ * wbuf_has_ino - returns if the wbuf contains data from the inode.
+ * @wbuf: the write-buffer
+ * @inum: the inode number
+ *
+ * This function returns with %1 if the write-buffer contains some data from the
+ * given inode otherwise it returns with %0.
+ */
+static int wbuf_has_ino(struct ubifs_wbuf *wbuf, ino_t inum)
+{
+       int i, ret = 0;
+
+       spin_lock(&wbuf->lock);
+       for (i = 0; i < wbuf->next_ino; i++)
+               if (inum == wbuf->inodes[i]) {
+                       ret = 1;
+                       break;
+               }
+       spin_unlock(&wbuf->lock);
+
+       return ret;
+}
+
+/**
+ * ubifs_sync_wbufs_by_inode - synchronize write-buffers for an inode.
+ * @c: UBIFS file-system description object
+ * @inode: inode to synchronize
+ *
+ * This function synchronizes write-buffers which contain nodes belonging to
+ * @inode. Returns zero in case of success and a negative error code in case of
+ * failure.
+ */
+int ubifs_sync_wbufs_by_inode(struct ubifs_info *c, struct inode *inode)
+{
+       int i, err = 0;
+
+       for (i = 0; i < c->jhead_cnt; i++) {
+               struct ubifs_wbuf *wbuf = &c->jheads[i].wbuf;
+
+               if (i == GCHD)
+                       /*
+                        * GC head is special, do not look at it. Even if the
+                        * head contains something related to this inode, it is
+                        * a _copy_ of corresponding on-flash node which sits
+                        * somewhere else.
+                        */
+                       continue;
+
+               if (!wbuf_has_ino(wbuf, inode->i_ino))
+                       continue;
+
+               mutex_lock_nested(&wbuf->io_mutex, wbuf->jhead);
+               if (wbuf_has_ino(wbuf, inode->i_ino))
+                       err = ubifs_wbuf_sync_nolock(wbuf);
+               mutex_unlock(&wbuf->io_mutex);
+
+               if (err) {
+                       ubifs_ro_mode(c, err);
+                       return err;
+               }
+       }
+       return 0;
+}
index efb3430a2581c4b723cafc6f846ed7b5841ed933..b5c4884e3032fa3aa30a54230d3c5f72b58598dc 100644 (file)
@@ -3,18 +3,7 @@
  *
  * Copyright (C) 2006-2008 Nokia Corporation.
  *
- * This program is free software; you can redistribute it and/or modify it
- * under the terms of the GNU General Public License version 2 as published by
- * the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
- * more details.
- *
- * You should have received a copy of the GNU General Public License along with
- * this program; if not, write to the Free Software Foundation, Inc., 51
- * Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
+ * SPDX-License-Identifier:    GPL-2.0+
  *
  * Authors: Artem Bityutskiy (Битюцкий Артём)
  *          Adrian Hunter
@@ -228,23 +217,6 @@ static inline void xent_key_init(const struct ubifs_info *c,
        key->u32[1] = hash | (UBIFS_XENT_KEY << UBIFS_S_KEY_HASH_BITS);
 }
 
-/**
- * xent_key_init_hash - initialize extended attribute entry key without
- *                      re-calculating hash function.
- * @c: UBIFS file-system description object
- * @key: key to initialize
- * @inum: host inode number
- * @hash: extended attribute entry name hash
- */
-static inline void xent_key_init_hash(const struct ubifs_info *c,
-                                     union ubifs_key *key, ino_t inum,
-                                     uint32_t hash)
-{
-       ubifs_assert(!(hash & ~UBIFS_S_KEY_HASH_MASK));
-       key->u32[0] = inum;
-       key->u32[1] = hash | (UBIFS_XENT_KEY << UBIFS_S_KEY_HASH_BITS);
-}
-
 /**
  * xent_key_init_flash - initialize on-flash extended attribute entry key.
  * @c: UBIFS file-system description object
@@ -295,22 +267,15 @@ static inline void data_key_init(const struct ubifs_info *c,
 }
 
 /**
- * data_key_init_flash - initialize on-flash data key.
+ * highest_data_key - get the highest possible data key for an inode.
  * @c: UBIFS file-system description object
- * @k: key to initialize
+ * @key: key to initialize
  * @inum: inode number
- * @block: block number
  */
-static inline void data_key_init_flash(const struct ubifs_info *c, void *k,
-                                      ino_t inum, unsigned int block)
+static inline void highest_data_key(const struct ubifs_info *c,
+                                  union ubifs_key *key, ino_t inum)
 {
-       union ubifs_key *key = k;
-
-       ubifs_assert(!(block & ~UBIFS_S_KEY_BLOCK_MASK));
-       key->j32[0] = cpu_to_le32(inum);
-       key->j32[1] = cpu_to_le32(block |
-                                 (UBIFS_DATA_KEY << UBIFS_S_KEY_BLOCK_BITS));
-       memset(k + 8, 0, UBIFS_MAX_KEY_LEN - 8);
+       data_key_init(c, key, inum, UBIFS_S_KEY_BLOCK_MASK);
 }
 
 /**
@@ -329,6 +294,20 @@ static inline void trun_key_init(const struct ubifs_info *c,
        key->u32[1] = UBIFS_TRUN_KEY << UBIFS_S_KEY_BLOCK_BITS;
 }
 
+/**
+ * invalid_key_init - initialize invalid node key.
+ * @c: UBIFS file-system description object
+ * @key: key to initialize
+ *
+ * This is a helper function which marks a @key object as invalid.
+ */
+static inline void invalid_key_init(const struct ubifs_info *c,
+                                   union ubifs_key *key)
+{
+       key->u32[0] = 0xDEADBEAF;
+       key->u32[1] = UBIFS_INVALID_KEY;
+}
+
 /**
  * key_type - get key type.
  * @c: UBIFS file-system description object
@@ -381,8 +360,8 @@ static inline ino_t key_inum_flash(const struct ubifs_info *c, const void *k)
  * @c: UBIFS file-system description object
  * @key: the key to get hash from
  */
-static inline int key_hash(const struct ubifs_info *c,
-                          const union ubifs_key *key)
+static inline uint32_t key_hash(const struct ubifs_info *c,
+                               const union ubifs_key *key)
 {
        return key->u32[1] & UBIFS_S_KEY_HASH_MASK;
 }
@@ -392,7 +371,7 @@ static inline int key_hash(const struct ubifs_info *c,
  * @c: UBIFS file-system description object
  * @k: the key to get hash from
  */
-static inline int key_hash_flash(const struct ubifs_info *c, const void *k)
+static inline uint32_t key_hash_flash(const struct ubifs_info *c, const void *k)
 {
        const union ubifs_key *key = k;
 
@@ -554,4 +533,5 @@ static inline unsigned long long key_max_inode_size(const struct ubifs_info *c)
                return 0;
        }
 }
+
 #endif /* !__UBIFS_KEY_H__ */
index 68a9bd98f8ca5a8a9576ad412a068968b53b7c62..ced04240c54efa1a443ff3bc7c3192eddbbc2d04 100644 (file)
@@ -3,18 +3,7 @@
  *
  * Copyright (C) 2006-2008 Nokia Corporation.
  *
- * This program is free software; you can redistribute it and/or modify it
- * under the terms of the GNU General Public License version 2 as published by
- * the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
- * more details.
- *
- * You should have received a copy of the GNU General Public License along with
- * this program; if not, write to the Free Software Foundation, Inc., 51
- * Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
+ * SPDX-License-Identifier:    GPL-2.0+
  *
  * Authors: Artem Bityutskiy (Битюцкий Артём)
  *          Adrian Hunter
  * journal.
  */
 
+#define __UBOOT__
+#ifdef __UBOOT__
+#include <linux/err.h>
+#endif
 #include "ubifs.h"
 
+static int dbg_check_bud_bytes(struct ubifs_info *c);
+
 /**
  * ubifs_search_bud - search bud LEB.
  * @c: UBIFS file-system description object
@@ -59,6 +54,57 @@ struct ubifs_bud *ubifs_search_bud(struct ubifs_info *c, int lnum)
        return NULL;
 }
 
+/**
+ * ubifs_get_wbuf - get the wbuf associated with a LEB, if there is one.
+ * @c: UBIFS file-system description object
+ * @lnum: logical eraseblock number to search
+ *
+ * This functions returns the wbuf for @lnum or %NULL if there is not one.
+ */
+struct ubifs_wbuf *ubifs_get_wbuf(struct ubifs_info *c, int lnum)
+{
+       struct rb_node *p;
+       struct ubifs_bud *bud;
+       int jhead;
+
+       if (!c->jheads)
+               return NULL;
+
+       spin_lock(&c->buds_lock);
+       p = c->buds.rb_node;
+       while (p) {
+               bud = rb_entry(p, struct ubifs_bud, rb);
+               if (lnum < bud->lnum)
+                       p = p->rb_left;
+               else if (lnum > bud->lnum)
+                       p = p->rb_right;
+               else {
+                       jhead = bud->jhead;
+                       spin_unlock(&c->buds_lock);
+                       return &c->jheads[jhead].wbuf;
+               }
+       }
+       spin_unlock(&c->buds_lock);
+       return NULL;
+}
+
+/**
+ * empty_log_bytes - calculate amount of empty space in the log.
+ * @c: UBIFS file-system description object
+ */
+static inline long long empty_log_bytes(const struct ubifs_info *c)
+{
+       long long h, t;
+
+       h = (long long)c->lhead_lnum * c->leb_size + c->lhead_offs;
+       t = (long long)c->ltail_lnum * c->leb_size;
+
+       if (h >= t)
+               return c->log_bytes - h + t;
+       else
+               return t - h;
+}
+
 /**
  * ubifs_add_bud - add bud LEB to the tree of buds and its journal head list.
  * @c: UBIFS file-system description object
@@ -88,7 +134,7 @@ void ubifs_add_bud(struct ubifs_info *c, struct ubifs_bud *bud)
                jhead = &c->jheads[bud->jhead];
                list_add_tail(&bud->list, &jhead->buds_list);
        } else
-               ubifs_assert(c->replaying && (c->vfs_sb->s_flags & MS_RDONLY));
+               ubifs_assert(c->replaying && c->ro_mount);
 
        /*
         * Note, although this is a new bud, we anyway account this space now,
@@ -98,7 +144,594 @@ void ubifs_add_bud(struct ubifs_info *c, struct ubifs_bud *bud)
         */
        c->bud_bytes += c->leb_size - bud->start;
 
-       dbg_log("LEB %d:%d, jhead %d, bud_bytes %lld", bud->lnum,
-               bud->start, bud->jhead, c->bud_bytes);
+       dbg_log("LEB %d:%d, jhead %s, bud_bytes %lld", bud->lnum,
+               bud->start, dbg_jhead(bud->jhead), c->bud_bytes);
+       spin_unlock(&c->buds_lock);
+}
+
+/**
+ * ubifs_add_bud_to_log - add a new bud to the log.
+ * @c: UBIFS file-system description object
+ * @jhead: journal head the bud belongs to
+ * @lnum: LEB number of the bud
+ * @offs: starting offset of the bud
+ *
+ * This function writes reference node for the new bud LEB @lnum it to the log,
+ * and adds it to the buds tress. It also makes sure that log size does not
+ * exceed the 'c->max_bud_bytes' limit. Returns zero in case of success,
+ * %-EAGAIN if commit is required, and a negative error codes in case of
+ * failure.
+ */
+int ubifs_add_bud_to_log(struct ubifs_info *c, int jhead, int lnum, int offs)
+{
+       int err;
+       struct ubifs_bud *bud;
+       struct ubifs_ref_node *ref;
+
+       bud = kmalloc(sizeof(struct ubifs_bud), GFP_NOFS);
+       if (!bud)
+               return -ENOMEM;
+       ref = kzalloc(c->ref_node_alsz, GFP_NOFS);
+       if (!ref) {
+               kfree(bud);
+               return -ENOMEM;
+       }
+
+       mutex_lock(&c->log_mutex);
+       ubifs_assert(!c->ro_media && !c->ro_mount);
+       if (c->ro_error) {
+               err = -EROFS;
+               goto out_unlock;
+       }
+
+       /* Make sure we have enough space in the log */
+       if (empty_log_bytes(c) - c->ref_node_alsz < c->min_log_bytes) {
+               dbg_log("not enough log space - %lld, required %d",
+                       empty_log_bytes(c), c->min_log_bytes);
+               ubifs_commit_required(c);
+               err = -EAGAIN;
+               goto out_unlock;
+       }
+
+       /*
+        * Make sure the amount of space in buds will not exceed the
+        * 'c->max_bud_bytes' limit, because we want to guarantee mount time
+        * limits.
+        *
+        * It is not necessary to hold @c->buds_lock when reading @c->bud_bytes
+        * because we are holding @c->log_mutex. All @c->bud_bytes take place
+        * when both @c->log_mutex and @c->bud_bytes are locked.
+        */
+       if (c->bud_bytes + c->leb_size - offs > c->max_bud_bytes) {
+               dbg_log("bud bytes %lld (%lld max), require commit",
+                       c->bud_bytes, c->max_bud_bytes);
+               ubifs_commit_required(c);
+               err = -EAGAIN;
+               goto out_unlock;
+       }
+
+       /*
+        * If the journal is full enough - start background commit. Note, it is
+        * OK to read 'c->cmt_state' without spinlock because integer reads
+        * are atomic in the kernel.
+        */
+       if (c->bud_bytes >= c->bg_bud_bytes &&
+           c->cmt_state == COMMIT_RESTING) {
+               dbg_log("bud bytes %lld (%lld max), initiate BG commit",
+                       c->bud_bytes, c->max_bud_bytes);
+               ubifs_request_bg_commit(c);
+       }
+
+       bud->lnum = lnum;
+       bud->start = offs;
+       bud->jhead = jhead;
+
+       ref->ch.node_type = UBIFS_REF_NODE;
+       ref->lnum = cpu_to_le32(bud->lnum);
+       ref->offs = cpu_to_le32(bud->start);
+       ref->jhead = cpu_to_le32(jhead);
+
+       if (c->lhead_offs > c->leb_size - c->ref_node_alsz) {
+               c->lhead_lnum = ubifs_next_log_lnum(c, c->lhead_lnum);
+               c->lhead_offs = 0;
+       }
+
+       if (c->lhead_offs == 0) {
+               /* Must ensure next log LEB has been unmapped */
+               err = ubifs_leb_unmap(c, c->lhead_lnum);
+               if (err)
+                       goto out_unlock;
+       }
+
+       if (bud->start == 0) {
+               /*
+                * Before writing the LEB reference which refers an empty LEB
+                * to the log, we have to make sure it is mapped, because
+                * otherwise we'd risk to refer an LEB with garbage in case of
+                * an unclean reboot, because the target LEB might have been
+                * unmapped, but not yet physically erased.
+                */
+               err = ubifs_leb_map(c, bud->lnum);
+               if (err)
+                       goto out_unlock;
+       }
+
+       dbg_log("write ref LEB %d:%d",
+               c->lhead_lnum, c->lhead_offs);
+       err = ubifs_write_node(c, ref, UBIFS_REF_NODE_SZ, c->lhead_lnum,
+                              c->lhead_offs);
+       if (err)
+               goto out_unlock;
+
+       c->lhead_offs += c->ref_node_alsz;
+
+       ubifs_add_bud(c, bud);
+
+       mutex_unlock(&c->log_mutex);
+       kfree(ref);
+       return 0;
+
+out_unlock:
+       mutex_unlock(&c->log_mutex);
+       kfree(ref);
+       kfree(bud);
+       return err;
+}
+
+/**
+ * remove_buds - remove used buds.
+ * @c: UBIFS file-system description object
+ *
+ * This function removes use buds from the buds tree. It does not remove the
+ * buds which are pointed to by journal heads.
+ */
+static void remove_buds(struct ubifs_info *c)
+{
+       struct rb_node *p;
+
+       ubifs_assert(list_empty(&c->old_buds));
+       c->cmt_bud_bytes = 0;
+       spin_lock(&c->buds_lock);
+       p = rb_first(&c->buds);
+       while (p) {
+               struct rb_node *p1 = p;
+               struct ubifs_bud *bud;
+               struct ubifs_wbuf *wbuf;
+
+               p = rb_next(p);
+               bud = rb_entry(p1, struct ubifs_bud, rb);
+               wbuf = &c->jheads[bud->jhead].wbuf;
+
+               if (wbuf->lnum == bud->lnum) {
+                       /*
+                        * Do not remove buds which are pointed to by journal
+                        * heads (non-closed buds).
+                        */
+                       c->cmt_bud_bytes += wbuf->offs - bud->start;
+                       dbg_log("preserve %d:%d, jhead %s, bud bytes %d, cmt_bud_bytes %lld",
+                               bud->lnum, bud->start, dbg_jhead(bud->jhead),
+                               wbuf->offs - bud->start, c->cmt_bud_bytes);
+                       bud->start = wbuf->offs;
+               } else {
+                       c->cmt_bud_bytes += c->leb_size - bud->start;
+                       dbg_log("remove %d:%d, jhead %s, bud bytes %d, cmt_bud_bytes %lld",
+                               bud->lnum, bud->start, dbg_jhead(bud->jhead),
+                               c->leb_size - bud->start, c->cmt_bud_bytes);
+                       rb_erase(p1, &c->buds);
+                       /*
+                        * If the commit does not finish, the recovery will need
+                        * to replay the journal, in which case the old buds
+                        * must be unchanged. Do not release them until post
+                        * commit i.e. do not allow them to be garbage
+                        * collected.
+                        */
+                       list_move(&bud->list, &c->old_buds);
+               }
+       }
+       spin_unlock(&c->buds_lock);
+}
+
+/**
+ * ubifs_log_start_commit - start commit.
+ * @c: UBIFS file-system description object
+ * @ltail_lnum: return new log tail LEB number
+ *
+ * The commit operation starts with writing "commit start" node to the log and
+ * reference nodes for all journal heads which will define new journal after
+ * the commit has been finished. The commit start and reference nodes are
+ * written in one go to the nearest empty log LEB (hence, when commit is
+ * finished UBIFS may safely unmap all the previous log LEBs). This function
+ * returns zero in case of success and a negative error code in case of
+ * failure.
+ */
+int ubifs_log_start_commit(struct ubifs_info *c, int *ltail_lnum)
+{
+       void *buf;
+       struct ubifs_cs_node *cs;
+       struct ubifs_ref_node *ref;
+       int err, i, max_len, len;
+
+       err = dbg_check_bud_bytes(c);
+       if (err)
+               return err;
+
+       max_len = UBIFS_CS_NODE_SZ + c->jhead_cnt * UBIFS_REF_NODE_SZ;
+       max_len = ALIGN(max_len, c->min_io_size);
+       buf = cs = kmalloc(max_len, GFP_NOFS);
+       if (!buf)
+               return -ENOMEM;
+
+       cs->ch.node_type = UBIFS_CS_NODE;
+       cs->cmt_no = cpu_to_le64(c->cmt_no);
+       ubifs_prepare_node(c, cs, UBIFS_CS_NODE_SZ, 0);
+
+       /*
+        * Note, we do not lock 'c->log_mutex' because this is the commit start
+        * phase and we are exclusively using the log. And we do not lock
+        * write-buffer because nobody can write to the file-system at this
+        * phase.
+        */
+
+       len = UBIFS_CS_NODE_SZ;
+       for (i = 0; i < c->jhead_cnt; i++) {
+               int lnum = c->jheads[i].wbuf.lnum;
+               int offs = c->jheads[i].wbuf.offs;
+
+               if (lnum == -1 || offs == c->leb_size)
+                       continue;
+
+               dbg_log("add ref to LEB %d:%d for jhead %s",
+                       lnum, offs, dbg_jhead(i));
+               ref = buf + len;
+               ref->ch.node_type = UBIFS_REF_NODE;
+               ref->lnum = cpu_to_le32(lnum);
+               ref->offs = cpu_to_le32(offs);
+               ref->jhead = cpu_to_le32(i);
+
+               ubifs_prepare_node(c, ref, UBIFS_REF_NODE_SZ, 0);
+               len += UBIFS_REF_NODE_SZ;
+       }
+
+       ubifs_pad(c, buf + len, ALIGN(len, c->min_io_size) - len);
+
+       /* Switch to the next log LEB */
+       if (c->lhead_offs) {
+               c->lhead_lnum = ubifs_next_log_lnum(c, c->lhead_lnum);
+               c->lhead_offs = 0;
+       }
+
+       if (c->lhead_offs == 0) {
+               /* Must ensure next LEB has been unmapped */
+               err = ubifs_leb_unmap(c, c->lhead_lnum);
+               if (err)
+                       goto out;
+       }
+
+       len = ALIGN(len, c->min_io_size);
+       dbg_log("writing commit start at LEB %d:0, len %d", c->lhead_lnum, len);
+       err = ubifs_leb_write(c, c->lhead_lnum, cs, 0, len);
+       if (err)
+               goto out;
+
+       *ltail_lnum = c->lhead_lnum;
+
+       c->lhead_offs += len;
+       if (c->lhead_offs == c->leb_size) {
+               c->lhead_lnum = ubifs_next_log_lnum(c, c->lhead_lnum);
+               c->lhead_offs = 0;
+       }
+
+       remove_buds(c);
+
+       /*
+        * We have started the commit and now users may use the rest of the log
+        * for new writes.
+        */
+       c->min_log_bytes = 0;
+
+out:
+       kfree(buf);
+       return err;
+}
+
+/**
+ * ubifs_log_end_commit - end commit.
+ * @c: UBIFS file-system description object
+ * @ltail_lnum: new log tail LEB number
+ *
+ * This function is called on when the commit operation was finished. It
+ * moves log tail to new position and unmaps LEBs which contain obsolete data.
+ * Returns zero in case of success and a negative error code in case of
+ * failure.
+ */
+int ubifs_log_end_commit(struct ubifs_info *c, int ltail_lnum)
+{
+       int err;
+
+       /*
+        * At this phase we have to lock 'c->log_mutex' because UBIFS allows FS
+        * writes during commit. Its only short "commit" start phase when
+        * writers are blocked.
+        */
+       mutex_lock(&c->log_mutex);
+
+       dbg_log("old tail was LEB %d:0, new tail is LEB %d:0",
+               c->ltail_lnum, ltail_lnum);
+
+       c->ltail_lnum = ltail_lnum;
+       /*
+        * The commit is finished and from now on it must be guaranteed that
+        * there is always enough space for the next commit.
+        */
+       c->min_log_bytes = c->leb_size;
+
+       spin_lock(&c->buds_lock);
+       c->bud_bytes -= c->cmt_bud_bytes;
+       spin_unlock(&c->buds_lock);
+
+       err = dbg_check_bud_bytes(c);
+
+       mutex_unlock(&c->log_mutex);
+       return err;
+}
+
+/**
+ * ubifs_log_post_commit - things to do after commit is completed.
+ * @c: UBIFS file-system description object
+ * @old_ltail_lnum: old log tail LEB number
+ *
+ * Release buds only after commit is completed, because they must be unchanged
+ * if recovery is needed.
+ *
+ * Unmap log LEBs only after commit is completed, because they may be needed for
+ * recovery.
+ *
+ * This function returns %0 on success and a negative error code on failure.
+ */
+int ubifs_log_post_commit(struct ubifs_info *c, int old_ltail_lnum)
+{
+       int lnum, err = 0;
+
+       while (!list_empty(&c->old_buds)) {
+               struct ubifs_bud *bud;
+
+               bud = list_entry(c->old_buds.next, struct ubifs_bud, list);
+               err = ubifs_return_leb(c, bud->lnum);
+               if (err)
+                       return err;
+               list_del(&bud->list);
+               kfree(bud);
+       }
+       mutex_lock(&c->log_mutex);
+       for (lnum = old_ltail_lnum; lnum != c->ltail_lnum;
+            lnum = ubifs_next_log_lnum(c, lnum)) {
+               dbg_log("unmap log LEB %d", lnum);
+               err = ubifs_leb_unmap(c, lnum);
+               if (err)
+                       goto out;
+       }
+out:
+       mutex_unlock(&c->log_mutex);
+       return err;
+}
+
+/**
+ * struct done_ref - references that have been done.
+ * @rb: rb-tree node
+ * @lnum: LEB number
+ */
+struct done_ref {
+       struct rb_node rb;
+       int lnum;
+};
+
+/**
+ * done_already - determine if a reference has been done already.
+ * @done_tree: rb-tree to store references that have been done
+ * @lnum: LEB number of reference
+ *
+ * This function returns %1 if the reference has been done, %0 if not, otherwise
+ * a negative error code is returned.
+ */
+static int done_already(struct rb_root *done_tree, int lnum)
+{
+       struct rb_node **p = &done_tree->rb_node, *parent = NULL;
+       struct done_ref *dr;
+
+       while (*p) {
+               parent = *p;
+               dr = rb_entry(parent, struct done_ref, rb);
+               if (lnum < dr->lnum)
+                       p = &(*p)->rb_left;
+               else if (lnum > dr->lnum)
+                       p = &(*p)->rb_right;
+               else
+                       return 1;
+       }
+
+       dr = kzalloc(sizeof(struct done_ref), GFP_NOFS);
+       if (!dr)
+               return -ENOMEM;
+
+       dr->lnum = lnum;
+
+       rb_link_node(&dr->rb, parent, p);
+       rb_insert_color(&dr->rb, done_tree);
+
+       return 0;
+}
+
+/**
+ * destroy_done_tree - destroy the done tree.
+ * @done_tree: done tree to destroy
+ */
+static void destroy_done_tree(struct rb_root *done_tree)
+{
+       struct done_ref *dr, *n;
+
+       rbtree_postorder_for_each_entry_safe(dr, n, done_tree, rb)
+               kfree(dr);
+}
+
+/**
+ * add_node - add a node to the consolidated log.
+ * @c: UBIFS file-system description object
+ * @buf: buffer to which to add
+ * @lnum: LEB number to which to write is passed and returned here
+ * @offs: offset to where to write is passed and returned here
+ * @node: node to add
+ *
+ * This function returns %0 on success and a negative error code on failure.
+ */
+static int add_node(struct ubifs_info *c, void *buf, int *lnum, int *offs,
+                   void *node)
+{
+       struct ubifs_ch *ch = node;
+       int len = le32_to_cpu(ch->len), remains = c->leb_size - *offs;
+
+       if (len > remains) {
+               int sz = ALIGN(*offs, c->min_io_size), err;
+
+               ubifs_pad(c, buf + *offs, sz - *offs);
+               err = ubifs_leb_change(c, *lnum, buf, sz);
+               if (err)
+                       return err;
+               *lnum = ubifs_next_log_lnum(c, *lnum);
+               *offs = 0;
+       }
+       memcpy(buf + *offs, node, len);
+       *offs += ALIGN(len, 8);
+       return 0;
+}
+
+/**
+ * ubifs_consolidate_log - consolidate the log.
+ * @c: UBIFS file-system description object
+ *
+ * Repeated failed commits could cause the log to be full, but at least 1 LEB is
+ * needed for commit. This function rewrites the reference nodes in the log
+ * omitting duplicates, and failed CS nodes, and leaving no gaps.
+ *
+ * This function returns %0 on success and a negative error code on failure.
+ */
+int ubifs_consolidate_log(struct ubifs_info *c)
+{
+       struct ubifs_scan_leb *sleb;
+       struct ubifs_scan_node *snod;
+       struct rb_root done_tree = RB_ROOT;
+       int lnum, err, first = 1, write_lnum, offs = 0;
+       void *buf;
+
+       dbg_rcvry("log tail LEB %d, log head LEB %d", c->ltail_lnum,
+                 c->lhead_lnum);
+       buf = vmalloc(c->leb_size);
+       if (!buf)
+               return -ENOMEM;
+       lnum = c->ltail_lnum;
+       write_lnum = lnum;
+       while (1) {
+               sleb = ubifs_scan(c, lnum, 0, c->sbuf, 0);
+               if (IS_ERR(sleb)) {
+                       err = PTR_ERR(sleb);
+                       goto out_free;
+               }
+               list_for_each_entry(snod, &sleb->nodes, list) {
+                       switch (snod->type) {
+                       case UBIFS_REF_NODE: {
+                               struct ubifs_ref_node *ref = snod->node;
+                               int ref_lnum = le32_to_cpu(ref->lnum);
+
+                               err = done_already(&done_tree, ref_lnum);
+                               if (err < 0)
+                                       goto out_scan;
+                               if (err != 1) {
+                                       err = add_node(c, buf, &write_lnum,
+                                                      &offs, snod->node);
+                                       if (err)
+                                               goto out_scan;
+                               }
+                               break;
+                       }
+                       case UBIFS_CS_NODE:
+                               if (!first)
+                                       break;
+                               err = add_node(c, buf, &write_lnum, &offs,
+                                              snod->node);
+                               if (err)
+                                       goto out_scan;
+                               first = 0;
+                               break;
+                       }
+               }
+               ubifs_scan_destroy(sleb);
+               if (lnum == c->lhead_lnum)
+                       break;
+               lnum = ubifs_next_log_lnum(c, lnum);
+       }
+       if (offs) {
+               int sz = ALIGN(offs, c->min_io_size);
+
+               ubifs_pad(c, buf + offs, sz - offs);
+               err = ubifs_leb_change(c, write_lnum, buf, sz);
+               if (err)
+                       goto out_free;
+               offs = ALIGN(offs, c->min_io_size);
+       }
+       destroy_done_tree(&done_tree);
+       vfree(buf);
+       if (write_lnum == c->lhead_lnum) {
+               ubifs_err("log is too full");
+               return -EINVAL;
+       }
+       /* Unmap remaining LEBs */
+       lnum = write_lnum;
+       do {
+               lnum = ubifs_next_log_lnum(c, lnum);
+               err = ubifs_leb_unmap(c, lnum);
+               if (err)
+                       return err;
+       } while (lnum != c->lhead_lnum);
+       c->lhead_lnum = write_lnum;
+       c->lhead_offs = offs;
+       dbg_rcvry("new log head at %d:%d", c->lhead_lnum, c->lhead_offs);
+       return 0;
+
+out_scan:
+       ubifs_scan_destroy(sleb);
+out_free:
+       destroy_done_tree(&done_tree);
+       vfree(buf);
+       return err;
+}
+
+/**
+ * dbg_check_bud_bytes - make sure bud bytes calculation are all right.
+ * @c: UBIFS file-system description object
+ *
+ * This function makes sure the amount of flash space used by closed buds
+ * ('c->bud_bytes' is correct). Returns zero in case of success and %-EINVAL in
+ * case of failure.
+ */
+static int dbg_check_bud_bytes(struct ubifs_info *c)
+{
+       int i, err = 0;
+       struct ubifs_bud *bud;
+       long long bud_bytes = 0;
+
+       if (!dbg_is_chk_gen(c))
+               return 0;
+
+       spin_lock(&c->buds_lock);
+       for (i = 0; i < c->jhead_cnt; i++)
+               list_for_each_entry(bud, &c->jheads[i].buds_list, list)
+                       bud_bytes += c->leb_size - bud->start;
+
+       if (c->bud_bytes != bud_bytes) {
+               ubifs_err("bad bud_bytes %lld, calculated %lld",
+                         c->bud_bytes, bud_bytes);
+               err = -EINVAL;
+       }
        spin_unlock(&c->buds_lock);
+
+       return err;
 }
index 8ce4949fcf4dc8d0de8ae5576fabe0f6ecea9139..fc6686bb08b14f52a7fef241d7811764552a6897 100644 (file)
@@ -3,18 +3,7 @@
  *
  * Copyright (C) 2006-2008 Nokia Corporation.
  *
- * This program is free software; you can redistribute it and/or modify it
- * under the terms of the GNU General Public License version 2 as published by
- * the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
- * more details.
- *
- * You should have received a copy of the GNU General Public License along with
- * this program; if not, write to the Free Software Foundation, Inc., 51
- * Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
+ * SPDX-License-Identifier:    GPL-2.0+
  *
  * Authors: Adrian Hunter
  *          Artem Bityutskiy (Битюцкий Артём)
  * an empty LEB for the journal, or a very dirty LEB for garbage collection.
  */
 
+#define __UBOOT__
+#ifdef __UBOOT__
+#include <linux/err.h>
+#endif
 #include "ubifs.h"
 
 /**
@@ -281,7 +274,7 @@ void ubifs_add_to_cat(struct ubifs_info *c, struct ubifs_lprops *lprops,
        case LPROPS_FREE:
                if (add_to_lpt_heap(c, lprops, cat))
                        break;
-               /* No more room on heap so make it uncategorized */
+               /* No more room on heap so make it un-categorized */
                cat = LPROPS_UNCAT;
                /* Fall through */
        case LPROPS_UNCAT:
@@ -300,8 +293,11 @@ void ubifs_add_to_cat(struct ubifs_info *c, struct ubifs_lprops *lprops,
        default:
                ubifs_assert(0);
        }
+
        lprops->flags &= ~LPROPS_CAT_MASK;
        lprops->flags |= cat;
+       c->in_a_category_cnt += 1;
+       ubifs_assert(c->in_a_category_cnt <= c->main_lebs);
 }
 
 /**
@@ -334,6 +330,9 @@ static void ubifs_remove_from_cat(struct ubifs_info *c,
        default:
                ubifs_assert(0);
        }
+
+       c->in_a_category_cnt -= 1;
+       ubifs_assert(c->in_a_category_cnt >= 0);
 }
 
 /**
@@ -375,8 +374,8 @@ void ubifs_replace_cat(struct ubifs_info *c, struct ubifs_lprops *old_lprops,
  * @lprops: LEB properties
  *
  * A LEB may have fallen off of the bottom of a heap, and ended up as
- * uncategorized even though it has enough space for us now. If that is the case
- * this function will put the LEB back onto a heap.
+ * un-categorized even though it has enough space for us now. If that is the
+ * case this function will put the LEB back onto a heap.
  */
 void ubifs_ensure_cat(struct ubifs_info *c, struct ubifs_lprops *lprops)
 {
@@ -436,10 +435,10 @@ int ubifs_categorize_lprops(const struct ubifs_info *c,
 /**
  * change_category - change LEB properties category.
  * @c: UBIFS file-system description object
- * @lprops: LEB properties to recategorize
+ * @lprops: LEB properties to re-categorize
  *
  * LEB properties are categorized to enable fast find operations. When the LEB
- * properties change they must be recategorized.
+ * properties change they must be re-categorized.
  */
 static void change_category(struct ubifs_info *c, struct ubifs_lprops *lprops)
 {
@@ -447,7 +446,7 @@ static void change_category(struct ubifs_info *c, struct ubifs_lprops *lprops)
        int new_cat = ubifs_categorize_lprops(c, lprops);
 
        if (old_cat == new_cat) {
-               struct ubifs_lpt_heap *heap = &c->lpt_heap[new_cat - 1];
+               struct ubifs_lpt_heap *heap;
 
                /* lprops on a heap now must be moved up or down */
                if (new_cat < 1 || new_cat > LPROPS_HEAP_CNT)
@@ -461,21 +460,18 @@ static void change_category(struct ubifs_info *c, struct ubifs_lprops *lprops)
 }
 
 /**
- * calc_dark - calculate LEB dark space size.
+ * ubifs_calc_dark - calculate LEB dark space size.
  * @c: the UBIFS file-system description object
  * @spc: amount of free and dirty space in the LEB
  *
- * This function calculates amount of dark space in an LEB which has @spc bytes
- * of free and dirty space. Returns the calculations result.
+ * This function calculates and returns amount of dark space in an LEB which
+ * has @spc bytes of free and dirty space.
  *
- * Dark space is the space which is not always usable - it depends on which
- * nodes are written in which order. E.g., if an LEB has only 512 free bytes,
- * it is dark space, because it cannot fit a large data node. So UBIFS cannot
- * count on this LEB and treat these 512 bytes as usable because it is not true
- * if, for example, only big chunks of uncompressible data will be written to
- * the FS.
+ * UBIFS is trying to account the space which might not be usable, and this
+ * space is called "dark space". For example, if an LEB has only %512 free
+ * bytes, it is dark space, because it cannot fit a large data node.
  */
-static int calc_dark(struct ubifs_info *c, int spc)
+int ubifs_calc_dark(const struct ubifs_info *c, int spc)
 {
        ubifs_assert(!(spc & 7));
 
@@ -507,7 +503,7 @@ static int is_lprops_dirty(struct ubifs_info *c, struct ubifs_lprops *lprops)
        pnode = (struct ubifs_pnode *)container_of(lprops - pos,
                                                   struct ubifs_pnode,
                                                   lprops[0]);
-       return !test_bit(COW_ZNODE, &pnode->flags) &&
+       return !test_bit(COW_CNODE, &pnode->flags) &&
               test_bit(DIRTY_CNODE, &pnode->flags);
 }
 
@@ -518,7 +514,7 @@ static int is_lprops_dirty(struct ubifs_info *c, struct ubifs_lprops *lprops)
  * @free: new free space amount
  * @dirty: new dirty space amount
  * @flags: new flags
- * @idx_gc_cnt: change to the count of idx_gc list
+ * @idx_gc_cnt: change to the count of @idx_gc list
  *
  * This function changes LEB properties (@free, @dirty or @flag). However, the
  * property which has the %LPROPS_NC value is not changed. Returns a pointer to
@@ -535,7 +531,7 @@ const struct ubifs_lprops *ubifs_change_lp(struct ubifs_info *c,
 {
        /*
         * This is the only function that is allowed to change lprops, so we
-        * discard the const qualifier.
+        * discard the "const" qualifier.
         */
        struct ubifs_lprops *lprops = (struct ubifs_lprops *)lp;
 
@@ -575,7 +571,7 @@ const struct ubifs_lprops *ubifs_change_lp(struct ubifs_info *c,
                if (old_spc < c->dead_wm)
                        c->lst.total_dead -= old_spc;
                else
-                       c->lst.total_dark -= calc_dark(c, old_spc);
+                       c->lst.total_dark -= ubifs_calc_dark(c, old_spc);
 
                c->lst.total_used -= c->leb_size - old_spc;
        }
@@ -616,7 +612,7 @@ const struct ubifs_lprops *ubifs_change_lp(struct ubifs_info *c,
                if (new_spc < c->dead_wm)
                        c->lst.total_dead += new_spc;
                else
-                       c->lst.total_dark += calc_dark(c, new_spc);
+                       c->lst.total_dark += ubifs_calc_dark(c, new_spc);
 
                c->lst.total_used += c->leb_size - new_spc;
        }
@@ -678,6 +674,9 @@ int ubifs_change_one_lp(struct ubifs_info *c, int lnum, int free, int dirty,
 
 out:
        ubifs_release_lprops(c);
+       if (err)
+               ubifs_err("cannot change properties of LEB %d, error %d",
+                         lnum, err);
        return err;
 }
 
@@ -714,6 +713,9 @@ int ubifs_update_one_lp(struct ubifs_info *c, int lnum, int free, int dirty,
 
 out:
        ubifs_release_lprops(c);
+       if (err)
+               ubifs_err("cannot update properties of LEB %d, error %d",
+                         lnum, err);
        return err;
 }
 
@@ -737,6 +739,8 @@ int ubifs_read_one_lp(struct ubifs_info *c, int lnum, struct ubifs_lprops *lp)
        lpp = ubifs_lpt_lookup(c, lnum);
        if (IS_ERR(lpp)) {
                err = PTR_ERR(lpp);
+               ubifs_err("cannot read properties of LEB %d, error %d",
+                         lnum, err);
                goto out;
        }
 
@@ -840,3 +844,471 @@ const struct ubifs_lprops *ubifs_fast_find_frdi_idx(struct ubifs_info *c)
        ubifs_assert(lprops->free + lprops->dirty == c->leb_size);
        return lprops;
 }
+
+/*
+ * Everything below is related to debugging.
+ */
+
+/**
+ * dbg_check_cats - check category heaps and lists.
+ * @c: UBIFS file-system description object
+ *
+ * This function returns %0 on success and a negative error code on failure.
+ */
+int dbg_check_cats(struct ubifs_info *c)
+{
+       struct ubifs_lprops *lprops;
+       struct list_head *pos;
+       int i, cat;
+
+       if (!dbg_is_chk_gen(c) && !dbg_is_chk_lprops(c))
+               return 0;
+
+       list_for_each_entry(lprops, &c->empty_list, list) {
+               if (lprops->free != c->leb_size) {
+                       ubifs_err("non-empty LEB %d on empty list (free %d dirty %d flags %d)",
+                                 lprops->lnum, lprops->free, lprops->dirty,
+                                 lprops->flags);
+                       return -EINVAL;
+               }
+               if (lprops->flags & LPROPS_TAKEN) {
+                       ubifs_err("taken LEB %d on empty list (free %d dirty %d flags %d)",
+                                 lprops->lnum, lprops->free, lprops->dirty,
+                                 lprops->flags);
+                       return -EINVAL;
+               }
+       }
+
+       i = 0;
+       list_for_each_entry(lprops, &c->freeable_list, list) {
+               if (lprops->free + lprops->dirty != c->leb_size) {
+                       ubifs_err("non-freeable LEB %d on freeable list (free %d dirty %d flags %d)",
+                                 lprops->lnum, lprops->free, lprops->dirty,
+                                 lprops->flags);
+                       return -EINVAL;
+               }
+               if (lprops->flags & LPROPS_TAKEN) {
+                       ubifs_err("taken LEB %d on freeable list (free %d dirty %d flags %d)",
+                                 lprops->lnum, lprops->free, lprops->dirty,
+                                 lprops->flags);
+                       return -EINVAL;
+               }
+               i += 1;
+       }
+       if (i != c->freeable_cnt) {
+               ubifs_err("freeable list count %d expected %d", i,
+                         c->freeable_cnt);
+               return -EINVAL;
+       }
+
+       i = 0;
+       list_for_each(pos, &c->idx_gc)
+               i += 1;
+       if (i != c->idx_gc_cnt) {
+               ubifs_err("idx_gc list count %d expected %d", i,
+                         c->idx_gc_cnt);
+               return -EINVAL;
+       }
+
+       list_for_each_entry(lprops, &c->frdi_idx_list, list) {
+               if (lprops->free + lprops->dirty != c->leb_size) {
+                       ubifs_err("non-freeable LEB %d on frdi_idx list (free %d dirty %d flags %d)",
+                                 lprops->lnum, lprops->free, lprops->dirty,
+                                 lprops->flags);
+                       return -EINVAL;
+               }
+               if (lprops->flags & LPROPS_TAKEN) {
+                       ubifs_err("taken LEB %d on frdi_idx list (free %d dirty %d flags %d)",
+                                 lprops->lnum, lprops->free, lprops->dirty,
+                                 lprops->flags);
+                       return -EINVAL;
+               }
+               if (!(lprops->flags & LPROPS_INDEX)) {
+                       ubifs_err("non-index LEB %d on frdi_idx list (free %d dirty %d flags %d)",
+                                 lprops->lnum, lprops->free, lprops->dirty,
+                                 lprops->flags);
+                       return -EINVAL;
+               }
+       }
+
+       for (cat = 1; cat <= LPROPS_HEAP_CNT; cat++) {
+               struct ubifs_lpt_heap *heap = &c->lpt_heap[cat - 1];
+
+               for (i = 0; i < heap->cnt; i++) {
+                       lprops = heap->arr[i];
+                       if (!lprops) {
+                               ubifs_err("null ptr in LPT heap cat %d", cat);
+                               return -EINVAL;
+                       }
+                       if (lprops->hpos != i) {
+                               ubifs_err("bad ptr in LPT heap cat %d", cat);
+                               return -EINVAL;
+                       }
+                       if (lprops->flags & LPROPS_TAKEN) {
+                               ubifs_err("taken LEB in LPT heap cat %d", cat);
+                               return -EINVAL;
+                       }
+               }
+       }
+
+       return 0;
+}
+
+void dbg_check_heap(struct ubifs_info *c, struct ubifs_lpt_heap *heap, int cat,
+                   int add_pos)
+{
+       int i = 0, j, err = 0;
+
+       if (!dbg_is_chk_gen(c) && !dbg_is_chk_lprops(c))
+               return;
+
+       for (i = 0; i < heap->cnt; i++) {
+               struct ubifs_lprops *lprops = heap->arr[i];
+               struct ubifs_lprops *lp;
+
+               if (i != add_pos)
+                       if ((lprops->flags & LPROPS_CAT_MASK) != cat) {
+                               err = 1;
+                               goto out;
+                       }
+               if (lprops->hpos != i) {
+                       err = 2;
+                       goto out;
+               }
+               lp = ubifs_lpt_lookup(c, lprops->lnum);
+               if (IS_ERR(lp)) {
+                       err = 3;
+                       goto out;
+               }
+               if (lprops != lp) {
+                       ubifs_err("lprops %zx lp %zx lprops->lnum %d lp->lnum %d",
+                                 (size_t)lprops, (size_t)lp, lprops->lnum,
+                                 lp->lnum);
+                       err = 4;
+                       goto out;
+               }
+               for (j = 0; j < i; j++) {
+                       lp = heap->arr[j];
+                       if (lp == lprops) {
+                               err = 5;
+                               goto out;
+                       }
+                       if (lp->lnum == lprops->lnum) {
+                               err = 6;
+                               goto out;
+                       }
+               }
+       }
+out:
+       if (err) {
+               ubifs_err("failed cat %d hpos %d err %d", cat, i, err);
+               dump_stack();
+               ubifs_dump_heap(c, heap, cat);
+       }
+}
+
+/**
+ * scan_check_cb - scan callback.
+ * @c: the UBIFS file-system description object
+ * @lp: LEB properties to scan
+ * @in_tree: whether the LEB properties are in main memory
+ * @lst: lprops statistics to update
+ *
+ * This function returns a code that indicates whether the scan should continue
+ * (%LPT_SCAN_CONTINUE), whether the LEB properties should be added to the tree
+ * in main memory (%LPT_SCAN_ADD), or whether the scan should stop
+ * (%LPT_SCAN_STOP).
+ */
+static int scan_check_cb(struct ubifs_info *c,
+                        const struct ubifs_lprops *lp, int in_tree,
+                        struct ubifs_lp_stats *lst)
+{
+       struct ubifs_scan_leb *sleb;
+       struct ubifs_scan_node *snod;
+       int cat, lnum = lp->lnum, is_idx = 0, used = 0, freef, dirty, ret;
+       void *buf = NULL;
+
+       cat = lp->flags & LPROPS_CAT_MASK;
+       if (cat != LPROPS_UNCAT) {
+               cat = ubifs_categorize_lprops(c, lp);
+               if (cat != (lp->flags & LPROPS_CAT_MASK)) {
+                       ubifs_err("bad LEB category %d expected %d",
+                                 (lp->flags & LPROPS_CAT_MASK), cat);
+                       return -EINVAL;
+               }
+       }
+
+       /* Check lp is on its category list (if it has one) */
+       if (in_tree) {
+               struct list_head *list = NULL;
+
+               switch (cat) {
+               case LPROPS_EMPTY:
+                       list = &c->empty_list;
+                       break;
+               case LPROPS_FREEABLE:
+                       list = &c->freeable_list;
+                       break;
+               case LPROPS_FRDI_IDX:
+                       list = &c->frdi_idx_list;
+                       break;
+               case LPROPS_UNCAT:
+                       list = &c->uncat_list;
+                       break;
+               }
+               if (list) {
+                       struct ubifs_lprops *lprops;
+                       int found = 0;
+
+                       list_for_each_entry(lprops, list, list) {
+                               if (lprops == lp) {
+                                       found = 1;
+                                       break;
+                               }
+                       }
+                       if (!found) {
+                               ubifs_err("bad LPT list (category %d)", cat);
+                               return -EINVAL;
+                       }
+               }
+       }
+
+       /* Check lp is on its category heap (if it has one) */
+       if (in_tree && cat > 0 && cat <= LPROPS_HEAP_CNT) {
+               struct ubifs_lpt_heap *heap = &c->lpt_heap[cat - 1];
+
+               if ((lp->hpos != -1 && heap->arr[lp->hpos]->lnum != lnum) ||
+                   lp != heap->arr[lp->hpos]) {
+                       ubifs_err("bad LPT heap (category %d)", cat);
+                       return -EINVAL;
+               }
+       }
+
+       buf = __vmalloc(c->leb_size, GFP_NOFS, PAGE_KERNEL);
+       if (!buf)
+               return -ENOMEM;
+
+       /*
+        * After an unclean unmount, empty and freeable LEBs
+        * may contain garbage - do not scan them.
+        */
+       if (lp->free == c->leb_size) {
+               lst->empty_lebs += 1;
+               lst->total_free += c->leb_size;
+               lst->total_dark += ubifs_calc_dark(c, c->leb_size);
+               return LPT_SCAN_CONTINUE;
+       }
+       if (lp->free + lp->dirty == c->leb_size &&
+           !(lp->flags & LPROPS_INDEX)) {
+               lst->total_free  += lp->free;
+               lst->total_dirty += lp->dirty;
+               lst->total_dark  +=  ubifs_calc_dark(c, c->leb_size);
+               return LPT_SCAN_CONTINUE;
+       }
+
+       sleb = ubifs_scan(c, lnum, 0, buf, 0);
+       if (IS_ERR(sleb)) {
+               ret = PTR_ERR(sleb);
+               if (ret == -EUCLEAN) {
+                       ubifs_dump_lprops(c);
+                       ubifs_dump_budg(c, &c->bi);
+               }
+               goto out;
+       }
+
+       is_idx = -1;
+       list_for_each_entry(snod, &sleb->nodes, list) {
+               int found, level = 0;
+
+               cond_resched();
+
+               if (is_idx == -1)
+                       is_idx = (snod->type == UBIFS_IDX_NODE) ? 1 : 0;
+
+               if (is_idx && snod->type != UBIFS_IDX_NODE) {
+                       ubifs_err("indexing node in data LEB %d:%d",
+                                 lnum, snod->offs);
+                       goto out_destroy;
+               }
+
+               if (snod->type == UBIFS_IDX_NODE) {
+                       struct ubifs_idx_node *idx = snod->node;
+
+                       key_read(c, ubifs_idx_key(c, idx), &snod->key);
+                       level = le16_to_cpu(idx->level);
+               }
+
+               found = ubifs_tnc_has_node(c, &snod->key, level, lnum,
+                                          snod->offs, is_idx);
+               if (found) {
+                       if (found < 0)
+                               goto out_destroy;
+                       used += ALIGN(snod->len, 8);
+               }
+       }
+
+       freef = c->leb_size - sleb->endpt;
+       dirty = sleb->endpt - used;
+
+       if (freef > c->leb_size || freef < 0 || dirty > c->leb_size ||
+           dirty < 0) {
+               ubifs_err("bad calculated accounting for LEB %d: free %d, dirty %d",
+                         lnum, freef, dirty);
+               goto out_destroy;
+       }
+
+       if (lp->free + lp->dirty == c->leb_size &&
+           freef + dirty == c->leb_size)
+               if ((is_idx && !(lp->flags & LPROPS_INDEX)) ||
+                   (!is_idx && freef == c->leb_size) ||
+                   lp->free == c->leb_size) {
+                       /*
+                        * Empty or freeable LEBs could contain index
+                        * nodes from an uncompleted commit due to an
+                        * unclean unmount. Or they could be empty for
+                        * the same reason. Or it may simply not have been
+                        * unmapped.
+                        */
+                       freef = lp->free;
+                       dirty = lp->dirty;
+                       is_idx = 0;
+                   }
+
+       if (is_idx && lp->free + lp->dirty == freef + dirty &&
+           lnum != c->ihead_lnum) {
+               /*
+                * After an unclean unmount, an index LEB could have a different
+                * amount of free space than the value recorded by lprops. That
+                * is because the in-the-gaps method may use free space or
+                * create free space (as a side-effect of using ubi_leb_change
+                * and not writing the whole LEB). The incorrect free space
+                * value is not a problem because the index is only ever
+                * allocated empty LEBs, so there will never be an attempt to
+                * write to the free space at the end of an index LEB - except
+                * by the in-the-gaps method for which it is not a problem.
+                */
+               freef = lp->free;
+               dirty = lp->dirty;
+       }
+
+       if (lp->free != freef || lp->dirty != dirty)
+               goto out_print;
+
+       if (is_idx && !(lp->flags & LPROPS_INDEX)) {
+               if (freef == c->leb_size)
+                       /* Free but not unmapped LEB, it's fine */
+                       is_idx = 0;
+               else {
+                       ubifs_err("indexing node without indexing flag");
+                       goto out_print;
+               }
+       }
+
+       if (!is_idx && (lp->flags & LPROPS_INDEX)) {
+               ubifs_err("data node with indexing flag");
+               goto out_print;
+       }
+
+       if (freef == c->leb_size)
+               lst->empty_lebs += 1;
+
+       if (is_idx)
+               lst->idx_lebs += 1;
+
+       if (!(lp->flags & LPROPS_INDEX))
+               lst->total_used += c->leb_size - freef - dirty;
+       lst->total_free += freef;
+       lst->total_dirty += dirty;
+
+       if (!(lp->flags & LPROPS_INDEX)) {
+               int spc = freef + dirty;
+
+               if (spc < c->dead_wm)
+                       lst->total_dead += spc;
+               else
+                       lst->total_dark += ubifs_calc_dark(c, spc);
+       }
+
+       ubifs_scan_destroy(sleb);
+       vfree(buf);
+       return LPT_SCAN_CONTINUE;
+
+out_print:
+       ubifs_err("bad accounting of LEB %d: free %d, dirty %d flags %#x, should be free %d, dirty %d",
+                 lnum, lp->free, lp->dirty, lp->flags, freef, dirty);
+       ubifs_dump_leb(c, lnum);
+out_destroy:
+       ubifs_scan_destroy(sleb);
+       ret = -EINVAL;
+out:
+       vfree(buf);
+       return ret;
+}
+
+/**
+ * dbg_check_lprops - check all LEB properties.
+ * @c: UBIFS file-system description object
+ *
+ * This function checks all LEB properties and makes sure they are all correct.
+ * It returns zero if everything is fine, %-EINVAL if there is an inconsistency
+ * and other negative error codes in case of other errors. This function is
+ * called while the file system is locked (because of commit start), so no
+ * additional locking is required. Note that locking the LPT mutex would cause
+ * a circular lock dependency with the TNC mutex.
+ */
+int dbg_check_lprops(struct ubifs_info *c)
+{
+       int i, err;
+       struct ubifs_lp_stats lst;
+
+       if (!dbg_is_chk_lprops(c))
+               return 0;
+
+       /*
+        * As we are going to scan the media, the write buffers have to be
+        * synchronized.
+        */
+       for (i = 0; i < c->jhead_cnt; i++) {
+               err = ubifs_wbuf_sync(&c->jheads[i].wbuf);
+               if (err)
+                       return err;
+       }
+
+       memset(&lst, 0, sizeof(struct ubifs_lp_stats));
+       err = ubifs_lpt_scan_nolock(c, c->main_first, c->leb_cnt - 1,
+                                   (ubifs_lpt_scan_callback)scan_check_cb,
+                                   &lst);
+       if (err && err != -ENOSPC)
+               goto out;
+
+       if (lst.empty_lebs != c->lst.empty_lebs ||
+           lst.idx_lebs != c->lst.idx_lebs ||
+           lst.total_free != c->lst.total_free ||
+           lst.total_dirty != c->lst.total_dirty ||
+           lst.total_used != c->lst.total_used) {
+               ubifs_err("bad overall accounting");
+               ubifs_err("calculated: empty_lebs %d, idx_lebs %d, total_free %lld, total_dirty %lld, total_used %lld",
+                         lst.empty_lebs, lst.idx_lebs, lst.total_free,
+                         lst.total_dirty, lst.total_used);
+               ubifs_err("read from lprops: empty_lebs %d, idx_lebs %d, total_free %lld, total_dirty %lld, total_used %lld",
+                         c->lst.empty_lebs, c->lst.idx_lebs, c->lst.total_free,
+                         c->lst.total_dirty, c->lst.total_used);
+               err = -EINVAL;
+               goto out;
+       }
+
+       if (lst.total_dead != c->lst.total_dead ||
+           lst.total_dark != c->lst.total_dark) {
+               ubifs_err("bad dead/dark space accounting");
+               ubifs_err("calculated: total_dead %lld, total_dark %lld",
+                         lst.total_dead, lst.total_dark);
+               ubifs_err("read from lprops: total_dead %lld, total_dark %lld",
+                         c->lst.total_dead, c->lst.total_dark);
+               err = -EINVAL;
+               goto out;
+       }
+
+       err = dbg_check_cats(c);
+out:
+       return err;
+}
index 1a50d4cc27902d07c1c80b2e07954b4f973a4578..c49d3b0687f5e11d395c476a089e1621e95b33a5 100644 (file)
@@ -3,18 +3,7 @@
  *
  * Copyright (C) 2006-2008 Nokia Corporation.
  *
- * This program is free software; you can redistribute it and/or modify it
- * under the terms of the GNU General Public License version 2 as published by
- * the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
- * more details.
- *
- * You should have received a copy of the GNU General Public License along with
- * this program; if not, write to the Free Software Foundation, Inc., 51
- * Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
+ * SPDX-License-Identifier:    GPL-2.0+
  *
  * Authors: Adrian Hunter
  *          Artem Bityutskiy (Битюцкий Артём)
  */
 
 #include "ubifs.h"
-#include "crc16.h"
+#define __UBOOT__
+#ifndef __UBOOT__
+#include <linux/crc16.h>
 #include <linux/math64.h>
+#include <linux/slab.h>
+#else
+#include <linux/compat.h>
+#include <linux/err.h>
+#include <ubi_uboot.h>
+#include "crc16.h"
+#endif
 
 /**
  * do_calc_lpt_geom - calculate sizes for the LPT area.
@@ -158,6 +156,119 @@ int ubifs_calc_lpt_geom(struct ubifs_info *c)
        return 0;
 }
 
+/**
+ * calc_dflt_lpt_geom - calculate default LPT geometry.
+ * @c: the UBIFS file-system description object
+ * @main_lebs: number of main area LEBs is passed and returned here
+ * @big_lpt: whether the LPT area is "big" is returned here
+ *
+ * The size of the LPT area depends on parameters that themselves are dependent
+ * on the size of the LPT area. This function, successively recalculates the LPT
+ * area geometry until the parameters and resultant geometry are consistent.
+ *
+ * This function returns %0 on success and a negative error code on failure.
+ */
+static int calc_dflt_lpt_geom(struct ubifs_info *c, int *main_lebs,
+                             int *big_lpt)
+{
+       int i, lebs_needed;
+       long long sz;
+
+       /* Start by assuming the minimum number of LPT LEBs */
+       c->lpt_lebs = UBIFS_MIN_LPT_LEBS;
+       c->main_lebs = *main_lebs - c->lpt_lebs;
+       if (c->main_lebs <= 0)
+               return -EINVAL;
+
+       /* And assume we will use the small LPT model */
+       c->big_lpt = 0;
+
+       /*
+        * Calculate the geometry based on assumptions above and then see if it
+        * makes sense
+        */
+       do_calc_lpt_geom(c);
+
+       /* Small LPT model must have lpt_sz < leb_size */
+       if (c->lpt_sz > c->leb_size) {
+               /* Nope, so try again using big LPT model */
+               c->big_lpt = 1;
+               do_calc_lpt_geom(c);
+       }
+
+       /* Now check there are enough LPT LEBs */
+       for (i = 0; i < 64 ; i++) {
+               sz = c->lpt_sz * 4; /* Allow 4 times the size */
+               lebs_needed = div_u64(sz + c->leb_size - 1, c->leb_size);
+               if (lebs_needed > c->lpt_lebs) {
+                       /* Not enough LPT LEBs so try again with more */
+                       c->lpt_lebs = lebs_needed;
+                       c->main_lebs = *main_lebs - c->lpt_lebs;
+                       if (c->main_lebs <= 0)
+                               return -EINVAL;
+                       do_calc_lpt_geom(c);
+                       continue;
+               }
+               if (c->ltab_sz > c->leb_size) {
+                       ubifs_err("LPT ltab too big");
+                       return -EINVAL;
+               }
+               *main_lebs = c->main_lebs;
+               *big_lpt = c->big_lpt;
+               return 0;
+       }
+       return -EINVAL;
+}
+
+/**
+ * pack_bits - pack bit fields end-to-end.
+ * @addr: address at which to pack (passed and next address returned)
+ * @pos: bit position at which to pack (passed and next position returned)
+ * @val: value to pack
+ * @nrbits: number of bits of value to pack (1-32)
+ */
+static void pack_bits(uint8_t **addr, int *pos, uint32_t val, int nrbits)
+{
+       uint8_t *p = *addr;
+       int b = *pos;
+
+       ubifs_assert(nrbits > 0);
+       ubifs_assert(nrbits <= 32);
+       ubifs_assert(*pos >= 0);
+       ubifs_assert(*pos < 8);
+       ubifs_assert((val >> nrbits) == 0 || nrbits == 32);
+       if (b) {
+               *p |= ((uint8_t)val) << b;
+               nrbits += b;
+               if (nrbits > 8) {
+                       *++p = (uint8_t)(val >>= (8 - b));
+                       if (nrbits > 16) {
+                               *++p = (uint8_t)(val >>= 8);
+                               if (nrbits > 24) {
+                                       *++p = (uint8_t)(val >>= 8);
+                                       if (nrbits > 32)
+                                               *++p = (uint8_t)(val >>= 8);
+                               }
+                       }
+               }
+       } else {
+               *p = (uint8_t)val;
+               if (nrbits > 8) {
+                       *++p = (uint8_t)(val >>= 8);
+                       if (nrbits > 16) {
+                               *++p = (uint8_t)(val >>= 8);
+                               if (nrbits > 24)
+                                       *++p = (uint8_t)(val >>= 8);
+                       }
+               }
+       }
+       b = nrbits & 7;
+       if (b == 0)
+               p++;
+       *addr = p;
+       *pos = b;
+}
+
 /**
  * ubifs_unpack_bits - unpack bit fields.
  * @addr: address at which to unpack (passed and next address returned)
@@ -227,6 +338,118 @@ uint32_t ubifs_unpack_bits(uint8_t **addr, int *pos, int nrbits)
        return val;
 }
 
+/**
+ * ubifs_pack_pnode - pack all the bit fields of a pnode.
+ * @c: UBIFS file-system description object
+ * @buf: buffer into which to pack
+ * @pnode: pnode to pack
+ */
+void ubifs_pack_pnode(struct ubifs_info *c, void *buf,
+                     struct ubifs_pnode *pnode)
+{
+       uint8_t *addr = buf + UBIFS_LPT_CRC_BYTES;
+       int i, pos = 0;
+       uint16_t crc;
+
+       pack_bits(&addr, &pos, UBIFS_LPT_PNODE, UBIFS_LPT_TYPE_BITS);
+       if (c->big_lpt)
+               pack_bits(&addr, &pos, pnode->num, c->pcnt_bits);
+       for (i = 0; i < UBIFS_LPT_FANOUT; i++) {
+               pack_bits(&addr, &pos, pnode->lprops[i].free >> 3,
+                         c->space_bits);
+               pack_bits(&addr, &pos, pnode->lprops[i].dirty >> 3,
+                         c->space_bits);
+               if (pnode->lprops[i].flags & LPROPS_INDEX)
+                       pack_bits(&addr, &pos, 1, 1);
+               else
+                       pack_bits(&addr, &pos, 0, 1);
+       }
+       crc = crc16(-1, buf + UBIFS_LPT_CRC_BYTES,
+                   c->pnode_sz - UBIFS_LPT_CRC_BYTES);
+       addr = buf;
+       pos = 0;
+       pack_bits(&addr, &pos, crc, UBIFS_LPT_CRC_BITS);
+}
+
+/**
+ * ubifs_pack_nnode - pack all the bit fields of a nnode.
+ * @c: UBIFS file-system description object
+ * @buf: buffer into which to pack
+ * @nnode: nnode to pack
+ */
+void ubifs_pack_nnode(struct ubifs_info *c, void *buf,
+                     struct ubifs_nnode *nnode)
+{
+       uint8_t *addr = buf + UBIFS_LPT_CRC_BYTES;
+       int i, pos = 0;
+       uint16_t crc;
+
+       pack_bits(&addr, &pos, UBIFS_LPT_NNODE, UBIFS_LPT_TYPE_BITS);
+       if (c->big_lpt)
+               pack_bits(&addr, &pos, nnode->num, c->pcnt_bits);
+       for (i = 0; i < UBIFS_LPT_FANOUT; i++) {
+               int lnum = nnode->nbranch[i].lnum;
+
+               if (lnum == 0)
+                       lnum = c->lpt_last + 1;
+               pack_bits(&addr, &pos, lnum - c->lpt_first, c->lpt_lnum_bits);
+               pack_bits(&addr, &pos, nnode->nbranch[i].offs,
+                         c->lpt_offs_bits);
+       }
+       crc = crc16(-1, buf + UBIFS_LPT_CRC_BYTES,
+                   c->nnode_sz - UBIFS_LPT_CRC_BYTES);
+       addr = buf;
+       pos = 0;
+       pack_bits(&addr, &pos, crc, UBIFS_LPT_CRC_BITS);
+}
+
+/**
+ * ubifs_pack_ltab - pack the LPT's own lprops table.
+ * @c: UBIFS file-system description object
+ * @buf: buffer into which to pack
+ * @ltab: LPT's own lprops table to pack
+ */
+void ubifs_pack_ltab(struct ubifs_info *c, void *buf,
+                    struct ubifs_lpt_lprops *ltab)
+{
+       uint8_t *addr = buf + UBIFS_LPT_CRC_BYTES;
+       int i, pos = 0;
+       uint16_t crc;
+
+       pack_bits(&addr, &pos, UBIFS_LPT_LTAB, UBIFS_LPT_TYPE_BITS);
+       for (i = 0; i < c->lpt_lebs; i++) {
+               pack_bits(&addr, &pos, ltab[i].free, c->lpt_spc_bits);
+               pack_bits(&addr, &pos, ltab[i].dirty, c->lpt_spc_bits);
+       }
+       crc = crc16(-1, buf + UBIFS_LPT_CRC_BYTES,
+                   c->ltab_sz - UBIFS_LPT_CRC_BYTES);
+       addr = buf;
+       pos = 0;
+       pack_bits(&addr, &pos, crc, UBIFS_LPT_CRC_BITS);
+}
+
+/**
+ * ubifs_pack_lsave - pack the LPT's save table.
+ * @c: UBIFS file-system description object
+ * @buf: buffer into which to pack
+ * @lsave: LPT's save table to pack
+ */
+void ubifs_pack_lsave(struct ubifs_info *c, void *buf, int *lsave)
+{
+       uint8_t *addr = buf + UBIFS_LPT_CRC_BYTES;
+       int i, pos = 0;
+       uint16_t crc;
+
+       pack_bits(&addr, &pos, UBIFS_LPT_LSAVE, UBIFS_LPT_TYPE_BITS);
+       for (i = 0; i < c->lsave_cnt; i++)
+               pack_bits(&addr, &pos, lsave[i], c->lnum_bits);
+       crc = crc16(-1, buf + UBIFS_LPT_CRC_BYTES,
+                   c->lsave_sz - UBIFS_LPT_CRC_BYTES);
+       addr = buf;
+       pos = 0;
+       pack_bits(&addr, &pos, crc, UBIFS_LPT_CRC_BITS);
+}
+
 /**
  * ubifs_add_lpt_dirt - add dirty space to LPT LEB properties.
  * @c: UBIFS file-system description object
@@ -243,6 +466,23 @@ void ubifs_add_lpt_dirt(struct ubifs_info *c, int lnum, int dirty)
        c->ltab[lnum - c->lpt_first].dirty += dirty;
 }
 
+/**
+ * set_ltab - set LPT LEB properties.
+ * @c: UBIFS file-system description object
+ * @lnum: LEB number
+ * @free: amount of free space
+ * @dirty: amount of dirty space
+ */
+static void set_ltab(struct ubifs_info *c, int lnum, int free, int dirty)
+{
+       dbg_lp("LEB %d free %d dirty %d to %d %d",
+              lnum, c->ltab[lnum - c->lpt_first].free,
+              c->ltab[lnum - c->lpt_first].dirty, free, dirty);
+       ubifs_assert(lnum >= c->lpt_first && lnum <= c->lpt_last);
+       c->ltab[lnum - c->lpt_first].free = free;
+       c->ltab[lnum - c->lpt_first].dirty = dirty;
+}
+
 /**
  * ubifs_add_nnode_dirt - add dirty space to LPT LEB properties.
  * @c: UBIFS file-system description object
@@ -275,6 +515,31 @@ static void add_pnode_dirt(struct ubifs_info *c, struct ubifs_pnode *pnode)
                           c->pnode_sz);
 }
 
+/**
+ * calc_nnode_num - calculate nnode number.
+ * @row: the row in the tree (root is zero)
+ * @col: the column in the row (leftmost is zero)
+ *
+ * The nnode number is a number that uniquely identifies a nnode and can be used
+ * easily to traverse the tree from the root to that nnode.
+ *
+ * This function calculates and returns the nnode number for the nnode at @row
+ * and @col.
+ */
+static int calc_nnode_num(int row, int col)
+{
+       int num, bits;
+
+       num = 1;
+       while (row--) {
+               bits = (col & (UBIFS_LPT_FANOUT - 1));
+               col >>= UBIFS_LPT_FANOUT_SHIFT;
+               num <<= UBIFS_LPT_FANOUT_SHIFT;
+               num |= bits;
+       }
+       return num;
+}
+
 /**
  * calc_nnode_num_from_parent - calculate nnode number.
  * @c: UBIFS file-system description object
@@ -327,6 +592,269 @@ static int calc_pnode_num_from_parent(const struct ubifs_info *c,
        return num;
 }
 
+/**
+ * ubifs_create_dflt_lpt - create default LPT.
+ * @c: UBIFS file-system description object
+ * @main_lebs: number of main area LEBs is passed and returned here
+ * @lpt_first: LEB number of first LPT LEB
+ * @lpt_lebs: number of LEBs for LPT is passed and returned here
+ * @big_lpt: use big LPT model is passed and returned here
+ *
+ * This function returns %0 on success and a negative error code on failure.
+ */
+int ubifs_create_dflt_lpt(struct ubifs_info *c, int *main_lebs, int lpt_first,
+                         int *lpt_lebs, int *big_lpt)
+{
+       int lnum, err = 0, node_sz, iopos, i, j, cnt, len, alen, row;
+       int blnum, boffs, bsz, bcnt;
+       struct ubifs_pnode *pnode = NULL;
+       struct ubifs_nnode *nnode = NULL;
+       void *buf = NULL, *p;
+       struct ubifs_lpt_lprops *ltab = NULL;
+       int *lsave = NULL;
+
+       err = calc_dflt_lpt_geom(c, main_lebs, big_lpt);
+       if (err)
+               return err;
+       *lpt_lebs = c->lpt_lebs;
+
+       /* Needed by 'ubifs_pack_nnode()' and 'set_ltab()' */
+       c->lpt_first = lpt_first;
+       /* Needed by 'set_ltab()' */
+       c->lpt_last = lpt_first + c->lpt_lebs - 1;
+       /* Needed by 'ubifs_pack_lsave()' */
+       c->main_first = c->leb_cnt - *main_lebs;
+
+       lsave = kmalloc(sizeof(int) * c->lsave_cnt, GFP_KERNEL);
+       pnode = kzalloc(sizeof(struct ubifs_pnode), GFP_KERNEL);
+       nnode = kzalloc(sizeof(struct ubifs_nnode), GFP_KERNEL);
+       buf = vmalloc(c->leb_size);
+       ltab = vmalloc(sizeof(struct ubifs_lpt_lprops) * c->lpt_lebs);
+       if (!pnode || !nnode || !buf || !ltab || !lsave) {
+               err = -ENOMEM;
+               goto out;
+       }
+
+       ubifs_assert(!c->ltab);
+       c->ltab = ltab; /* Needed by set_ltab */
+
+       /* Initialize LPT's own lprops */
+       for (i = 0; i < c->lpt_lebs; i++) {
+               ltab[i].free = c->leb_size;
+               ltab[i].dirty = 0;
+               ltab[i].tgc = 0;
+               ltab[i].cmt = 0;
+       }
+
+       lnum = lpt_first;
+       p = buf;
+       /* Number of leaf nodes (pnodes) */
+       cnt = c->pnode_cnt;
+
+       /*
+        * The first pnode contains the LEB properties for the LEBs that contain
+        * the root inode node and the root index node of the index tree.
+        */
+       node_sz = ALIGN(ubifs_idx_node_sz(c, 1), 8);
+       iopos = ALIGN(node_sz, c->min_io_size);
+       pnode->lprops[0].free = c->leb_size - iopos;
+       pnode->lprops[0].dirty = iopos - node_sz;
+       pnode->lprops[0].flags = LPROPS_INDEX;
+
+       node_sz = UBIFS_INO_NODE_SZ;
+       iopos = ALIGN(node_sz, c->min_io_size);
+       pnode->lprops[1].free = c->leb_size - iopos;
+       pnode->lprops[1].dirty = iopos - node_sz;
+
+       for (i = 2; i < UBIFS_LPT_FANOUT; i++)
+               pnode->lprops[i].free = c->leb_size;
+
+       /* Add first pnode */
+       ubifs_pack_pnode(c, p, pnode);
+       p += c->pnode_sz;
+       len = c->pnode_sz;
+       pnode->num += 1;
+
+       /* Reset pnode values for remaining pnodes */
+       pnode->lprops[0].free = c->leb_size;
+       pnode->lprops[0].dirty = 0;
+       pnode->lprops[0].flags = 0;
+
+       pnode->lprops[1].free = c->leb_size;
+       pnode->lprops[1].dirty = 0;
+
+       /*
+        * To calculate the internal node branches, we keep information about
+        * the level below.
+        */
+       blnum = lnum; /* LEB number of level below */
+       boffs = 0; /* Offset of level below */
+       bcnt = cnt; /* Number of nodes in level below */
+       bsz = c->pnode_sz; /* Size of nodes in level below */
+
+       /* Add all remaining pnodes */
+       for (i = 1; i < cnt; i++) {
+               if (len + c->pnode_sz > c->leb_size) {
+                       alen = ALIGN(len, c->min_io_size);
+                       set_ltab(c, lnum, c->leb_size - alen, alen - len);
+                       memset(p, 0xff, alen - len);
+                       err = ubifs_leb_change(c, lnum++, buf, alen);
+                       if (err)
+                               goto out;
+                       p = buf;
+                       len = 0;
+               }
+               ubifs_pack_pnode(c, p, pnode);
+               p += c->pnode_sz;
+               len += c->pnode_sz;
+               /*
+                * pnodes are simply numbered left to right starting at zero,
+                * which means the pnode number can be used easily to traverse
+                * down the tree to the corresponding pnode.
+                */
+               pnode->num += 1;
+       }
+
+       row = 0;
+       for (i = UBIFS_LPT_FANOUT; cnt > i; i <<= UBIFS_LPT_FANOUT_SHIFT)
+               row += 1;
+       /* Add all nnodes, one level at a time */
+       while (1) {
+               /* Number of internal nodes (nnodes) at next level */
+               cnt = DIV_ROUND_UP(cnt, UBIFS_LPT_FANOUT);
+               for (i = 0; i < cnt; i++) {
+                       if (len + c->nnode_sz > c->leb_size) {
+                               alen = ALIGN(len, c->min_io_size);
+                               set_ltab(c, lnum, c->leb_size - alen,
+                                           alen - len);
+                               memset(p, 0xff, alen - len);
+                               err = ubifs_leb_change(c, lnum++, buf, alen);
+                               if (err)
+                                       goto out;
+                               p = buf;
+                               len = 0;
+                       }
+                       /* Only 1 nnode at this level, so it is the root */
+                       if (cnt == 1) {
+                               c->lpt_lnum = lnum;
+                               c->lpt_offs = len;
+                       }
+                       /* Set branches to the level below */
+                       for (j = 0; j < UBIFS_LPT_FANOUT; j++) {
+                               if (bcnt) {
+                                       if (boffs + bsz > c->leb_size) {
+                                               blnum += 1;
+                                               boffs = 0;
+                                       }
+                                       nnode->nbranch[j].lnum = blnum;
+                                       nnode->nbranch[j].offs = boffs;
+                                       boffs += bsz;
+                                       bcnt--;
+                               } else {
+                                       nnode->nbranch[j].lnum = 0;
+                                       nnode->nbranch[j].offs = 0;
+                               }
+                       }
+                       nnode->num = calc_nnode_num(row, i);
+                       ubifs_pack_nnode(c, p, nnode);
+                       p += c->nnode_sz;
+                       len += c->nnode_sz;
+               }
+               /* Only 1 nnode at this level, so it is the root */
+               if (cnt == 1)
+                       break;
+               /* Update the information about the level below */
+               bcnt = cnt;
+               bsz = c->nnode_sz;
+               row -= 1;
+       }
+
+       if (*big_lpt) {
+               /* Need to add LPT's save table */
+               if (len + c->lsave_sz > c->leb_size) {
+                       alen = ALIGN(len, c->min_io_size);
+                       set_ltab(c, lnum, c->leb_size - alen, alen - len);
+                       memset(p, 0xff, alen - len);
+                       err = ubifs_leb_change(c, lnum++, buf, alen);
+                       if (err)
+                               goto out;
+                       p = buf;
+                       len = 0;
+               }
+
+               c->lsave_lnum = lnum;
+               c->lsave_offs = len;
+
+               for (i = 0; i < c->lsave_cnt && i < *main_lebs; i++)
+                       lsave[i] = c->main_first + i;
+               for (; i < c->lsave_cnt; i++)
+                       lsave[i] = c->main_first;
+
+               ubifs_pack_lsave(c, p, lsave);
+               p += c->lsave_sz;
+               len += c->lsave_sz;
+       }
+
+       /* Need to add LPT's own LEB properties table */
+       if (len + c->ltab_sz > c->leb_size) {
+               alen = ALIGN(len, c->min_io_size);
+               set_ltab(c, lnum, c->leb_size - alen, alen - len);
+               memset(p, 0xff, alen - len);
+               err = ubifs_leb_change(c, lnum++, buf, alen);
+               if (err)
+                       goto out;
+               p = buf;
+               len = 0;
+       }
+
+       c->ltab_lnum = lnum;
+       c->ltab_offs = len;
+
+       /* Update ltab before packing it */
+       len += c->ltab_sz;
+       alen = ALIGN(len, c->min_io_size);
+       set_ltab(c, lnum, c->leb_size - alen, alen - len);
+
+       ubifs_pack_ltab(c, p, ltab);
+       p += c->ltab_sz;
+
+       /* Write remaining buffer */
+       memset(p, 0xff, alen - len);
+       err = ubifs_leb_change(c, lnum, buf, alen);
+       if (err)
+               goto out;
+
+       c->nhead_lnum = lnum;
+       c->nhead_offs = ALIGN(len, c->min_io_size);
+
+       dbg_lp("space_bits %d", c->space_bits);
+       dbg_lp("lpt_lnum_bits %d", c->lpt_lnum_bits);
+       dbg_lp("lpt_offs_bits %d", c->lpt_offs_bits);
+       dbg_lp("lpt_spc_bits %d", c->lpt_spc_bits);
+       dbg_lp("pcnt_bits %d", c->pcnt_bits);
+       dbg_lp("lnum_bits %d", c->lnum_bits);
+       dbg_lp("pnode_sz %d", c->pnode_sz);
+       dbg_lp("nnode_sz %d", c->nnode_sz);
+       dbg_lp("ltab_sz %d", c->ltab_sz);
+       dbg_lp("lsave_sz %d", c->lsave_sz);
+       dbg_lp("lsave_cnt %d", c->lsave_cnt);
+       dbg_lp("lpt_hght %d", c->lpt_hght);
+       dbg_lp("big_lpt %d", c->big_lpt);
+       dbg_lp("LPT root is at %d:%d", c->lpt_lnum, c->lpt_offs);
+       dbg_lp("LPT head is at %d:%d", c->nhead_lnum, c->nhead_offs);
+       dbg_lp("LPT ltab is at %d:%d", c->ltab_lnum, c->ltab_offs);
+       if (c->big_lpt)
+               dbg_lp("LPT lsave is at %d:%d", c->lsave_lnum, c->lsave_offs);
+out:
+       c->ltab = NULL;
+       kfree(lsave);
+       vfree(ltab);
+       vfree(buf);
+       kfree(nnode);
+       kfree(pnode);
+       return err;
+}
+
 /**
  * update_cats - add LEB properties of a pnode to LEB category lists and heaps.
  * @c: UBIFS file-system description object
@@ -392,7 +920,7 @@ static int check_lpt_crc(void *buf, int len)
        if (crc != calc_crc) {
                ubifs_err("invalid crc in LPT node: crc %hx calc %hx", crc,
                          calc_crc);
-               dbg_dump_stack();
+               dump_stack();
                return -EINVAL;
        }
        return 0;
@@ -415,7 +943,7 @@ static int check_lpt_type(uint8_t **addr, int *pos, int type)
        if (node_type != type) {
                ubifs_err("invalid type (%d) in LPT node type %d", node_type,
                          type);
-               dbg_dump_stack();
+               dump_stack();
                return -EINVAL;
        }
        return 0;
@@ -524,6 +1052,34 @@ static int unpack_ltab(const struct ubifs_info *c, void *buf)
        return err;
 }
 
+#ifndef __UBOOT__
+/**
+ * unpack_lsave - unpack the LPT's save table.
+ * @c: UBIFS file-system description object
+ * @buf: buffer from which to unpack
+ *
+ * This function returns %0 on success and a negative error code on failure.
+ */
+static int unpack_lsave(const struct ubifs_info *c, void *buf)
+{
+       uint8_t *addr = buf + UBIFS_LPT_CRC_BYTES;
+       int i, pos = 0, err;
+
+       err = check_lpt_type(&addr, &pos, UBIFS_LPT_LSAVE);
+       if (err)
+               return err;
+       for (i = 0; i < c->lsave_cnt; i++) {
+               int lnum = ubifs_unpack_bits(&addr, &pos, c->lnum_bits);
+
+               if (lnum < c->main_first || lnum >= c->leb_cnt)
+                       return -EINVAL;
+               c->lsave[i] = lnum;
+       }
+       err = check_lpt_crc(buf, c->lsave_sz);
+       return err;
+}
+#endif
+
 /**
  * validate_nnode - validate a nnode.
  * @c: UBIFS file-system description object
@@ -662,7 +1218,7 @@ int ubifs_read_nnode(struct ubifs_info *c, struct ubifs_nnode *parent, int iip)
                if (c->big_lpt)
                        nnode->num = calc_nnode_num_from_parent(c, parent, iip);
        } else {
-               err = ubi_read(c->ubi, lnum, buf, offs, c->nnode_sz);
+               err = ubifs_leb_read(c, lnum, buf, offs, c->nnode_sz, 1);
                if (err)
                        goto out;
                err = ubifs_unpack_nnode(c, buf, nnode);
@@ -687,6 +1243,7 @@ int ubifs_read_nnode(struct ubifs_info *c, struct ubifs_nnode *parent, int iip)
 
 out:
        ubifs_err("error %d reading nnode at %d:%d", err, lnum, offs);
+       dump_stack();
        kfree(nnode);
        return err;
 }
@@ -710,10 +1267,9 @@ static int read_pnode(struct ubifs_info *c, struct ubifs_nnode *parent, int iip)
        lnum = branch->lnum;
        offs = branch->offs;
        pnode = kzalloc(sizeof(struct ubifs_pnode), GFP_NOFS);
-       if (!pnode) {
-               err = -ENOMEM;
-               goto out;
-       }
+       if (!pnode)
+               return -ENOMEM;
+
        if (lnum == 0) {
                /*
                 * This pnode was not written which just means that the LEB
@@ -731,7 +1287,7 @@ static int read_pnode(struct ubifs_info *c, struct ubifs_nnode *parent, int iip)
                        lprops->flags = ubifs_categorize_lprops(c, lprops);
                }
        } else {
-               err = ubi_read(c->ubi, lnum, buf, offs, c->pnode_sz);
+               err = ubifs_leb_read(c, lnum, buf, offs, c->pnode_sz, 1);
                if (err)
                        goto out;
                err = unpack_pnode(c, buf, pnode);
@@ -752,8 +1308,9 @@ static int read_pnode(struct ubifs_info *c, struct ubifs_nnode *parent, int iip)
 
 out:
        ubifs_err("error %d reading pnode at %d:%d", err, lnum, offs);
-       dbg_dump_pnode(c, pnode, parent, iip);
-       dbg_msg("calc num: %d", calc_pnode_num_from_parent(c, parent, iip));
+       ubifs_dump_pnode(c, pnode, parent, iip);
+       dump_stack();
+       ubifs_err("calc num: %d", calc_pnode_num_from_parent(c, parent, iip));
        kfree(pnode);
        return err;
 }
@@ -772,7 +1329,7 @@ static int read_ltab(struct ubifs_info *c)
        buf = vmalloc(c->ltab_sz);
        if (!buf)
                return -ENOMEM;
-       err = ubi_read(c->ubi, c->ltab_lnum, buf, c->ltab_offs, c->ltab_sz);
+       err = ubifs_leb_read(c, c->ltab_lnum, buf, c->ltab_offs, c->ltab_sz, 1);
        if (err)
                goto out;
        err = unpack_ltab(c, buf);
@@ -781,6 +1338,50 @@ out:
        return err;
 }
 
+#ifndef __UBOOT__
+/**
+ * read_lsave - read LPT's save table.
+ * @c: UBIFS file-system description object
+ *
+ * This function returns %0 on success and a negative error code on failure.
+ */
+static int read_lsave(struct ubifs_info *c)
+{
+       int err, i;
+       void *buf;
+
+       buf = vmalloc(c->lsave_sz);
+       if (!buf)
+               return -ENOMEM;
+       err = ubifs_leb_read(c, c->lsave_lnum, buf, c->lsave_offs,
+                            c->lsave_sz, 1);
+       if (err)
+               goto out;
+       err = unpack_lsave(c, buf);
+       if (err)
+               goto out;
+       for (i = 0; i < c->lsave_cnt; i++) {
+               int lnum = c->lsave[i];
+               struct ubifs_lprops *lprops;
+
+               /*
+                * Due to automatic resizing, the values in the lsave table
+                * could be beyond the volume size - just ignore them.
+                */
+               if (lnum >= c->leb_cnt)
+                       continue;
+               lprops = ubifs_lpt_lookup(c, lnum);
+               if (IS_ERR(lprops)) {
+                       err = PTR_ERR(lprops);
+                       goto out;
+               }
+       }
+out:
+       vfree(buf);
+       return err;
+}
+#endif
+
 /**
  * ubifs_get_nnode - get a nnode.
  * @c: UBIFS file-system description object
@@ -861,13 +1462,13 @@ struct ubifs_lprops *ubifs_lpt_lookup(struct ubifs_info *c, int lnum)
                shft -= UBIFS_LPT_FANOUT_SHIFT;
                nnode = ubifs_get_nnode(c, nnode, iip);
                if (IS_ERR(nnode))
-                       return ERR_PTR(PTR_ERR(nnode));
+                       return ERR_CAST(nnode);
        }
        iip = ((i >> shft) & (UBIFS_LPT_FANOUT - 1));
        shft -= UBIFS_LPT_FANOUT_SHIFT;
        pnode = ubifs_get_pnode(c, nnode, iip);
        if (IS_ERR(pnode))
-               return ERR_PTR(PTR_ERR(pnode));
+               return ERR_CAST(pnode);
        iip = (i & (UBIFS_LPT_FANOUT - 1));
        dbg_lp("LEB %d, free %d, dirty %d, flags %d", lnum,
               pnode->lprops[iip].free, pnode->lprops[iip].dirty,
@@ -990,7 +1591,7 @@ struct ubifs_lprops *ubifs_lpt_lookup_dirty(struct ubifs_info *c, int lnum)
        nnode = c->nroot;
        nnode = dirty_cow_nnode(c, nnode);
        if (IS_ERR(nnode))
-               return ERR_PTR(PTR_ERR(nnode));
+               return ERR_CAST(nnode);
        i = lnum - c->main_first;
        shft = c->lpt_hght * UBIFS_LPT_FANOUT_SHIFT;
        for (h = 1; h < c->lpt_hght; h++) {
@@ -998,19 +1599,19 @@ struct ubifs_lprops *ubifs_lpt_lookup_dirty(struct ubifs_info *c, int lnum)
                shft -= UBIFS_LPT_FANOUT_SHIFT;
                nnode = ubifs_get_nnode(c, nnode, iip);
                if (IS_ERR(nnode))
-                       return ERR_PTR(PTR_ERR(nnode));
+                       return ERR_CAST(nnode);
                nnode = dirty_cow_nnode(c, nnode);
                if (IS_ERR(nnode))
-                       return ERR_PTR(PTR_ERR(nnode));
+                       return ERR_CAST(nnode);
        }
        iip = ((i >> shft) & (UBIFS_LPT_FANOUT - 1));
        shft -= UBIFS_LPT_FANOUT_SHIFT;
        pnode = ubifs_get_pnode(c, nnode, iip);
        if (IS_ERR(pnode))
-               return ERR_PTR(PTR_ERR(pnode));
+               return ERR_CAST(pnode);
        pnode = dirty_cow_pnode(c, pnode);
        if (IS_ERR(pnode))
-               return ERR_PTR(PTR_ERR(pnode));
+               return ERR_CAST(pnode);
        iip = (i & (UBIFS_LPT_FANOUT - 1));
        dbg_lp("LEB %d, free %d, dirty %d, flags %d", lnum,
               pnode->lprops[iip].free, pnode->lprops[iip].dirty,
@@ -1079,6 +1680,47 @@ static int lpt_init_rd(struct ubifs_info *c)
        return 0;
 }
 
+#ifndef __UBOOT__
+/**
+ * lpt_init_wr - initialize the LPT for writing.
+ * @c: UBIFS file-system description object
+ *
+ * 'lpt_init_rd()' must have been called already.
+ *
+ * This function returns %0 on success and a negative error code on failure.
+ */
+static int lpt_init_wr(struct ubifs_info *c)
+{
+       int err, i;
+
+       c->ltab_cmt = vmalloc(sizeof(struct ubifs_lpt_lprops) * c->lpt_lebs);
+       if (!c->ltab_cmt)
+               return -ENOMEM;
+
+       c->lpt_buf = vmalloc(c->leb_size);
+       if (!c->lpt_buf)
+               return -ENOMEM;
+
+       if (c->big_lpt) {
+               c->lsave = kmalloc(sizeof(int) * c->lsave_cnt, GFP_NOFS);
+               if (!c->lsave)
+                       return -ENOMEM;
+               err = read_lsave(c);
+               if (err)
+                       return err;
+       }
+
+       for (i = 0; i < c->lpt_lebs; i++)
+               if (c->ltab[i].free == c->leb_size) {
+                       err = ubifs_leb_unmap(c, i + c->lpt_first);
+                       if (err)
+                               return err;
+               }
+
+       return 0;
+}
+#endif
+
 /**
  * ubifs_lpt_init - initialize the LPT.
  * @c: UBIFS file-system description object
@@ -1097,9 +1739,547 @@ int ubifs_lpt_init(struct ubifs_info *c, int rd, int wr)
 
        if (rd) {
                err = lpt_init_rd(c);
+               if (err)
+                       goto out_err;
+       }
+
+#ifndef __UBOOT__
+       if (wr) {
+               err = lpt_init_wr(c);
+               if (err)
+                       goto out_err;
+       }
+#endif
+
+       return 0;
+
+out_err:
+#ifndef __UBOOT__
+       if (wr)
+               ubifs_lpt_free(c, 1);
+#endif
+       if (rd)
+               ubifs_lpt_free(c, 0);
+       return err;
+}
+
+/**
+ * struct lpt_scan_node - somewhere to put nodes while we scan LPT.
+ * @nnode: where to keep a nnode
+ * @pnode: where to keep a pnode
+ * @cnode: where to keep a cnode
+ * @in_tree: is the node in the tree in memory
+ * @ptr.nnode: pointer to the nnode (if it is an nnode) which may be here or in
+ * the tree
+ * @ptr.pnode: ditto for pnode
+ * @ptr.cnode: ditto for cnode
+ */
+struct lpt_scan_node {
+       union {
+               struct ubifs_nnode nnode;
+               struct ubifs_pnode pnode;
+               struct ubifs_cnode cnode;
+       };
+       int in_tree;
+       union {
+               struct ubifs_nnode *nnode;
+               struct ubifs_pnode *pnode;
+               struct ubifs_cnode *cnode;
+       } ptr;
+};
+
+/**
+ * scan_get_nnode - for the scan, get a nnode from either the tree or flash.
+ * @c: the UBIFS file-system description object
+ * @path: where to put the nnode
+ * @parent: parent of the nnode
+ * @iip: index in parent of the nnode
+ *
+ * This function returns a pointer to the nnode on success or a negative error
+ * code on failure.
+ */
+static struct ubifs_nnode *scan_get_nnode(struct ubifs_info *c,
+                                         struct lpt_scan_node *path,
+                                         struct ubifs_nnode *parent, int iip)
+{
+       struct ubifs_nbranch *branch;
+       struct ubifs_nnode *nnode;
+       void *buf = c->lpt_nod_buf;
+       int err;
+
+       branch = &parent->nbranch[iip];
+       nnode = branch->nnode;
+       if (nnode) {
+               path->in_tree = 1;
+               path->ptr.nnode = nnode;
+               return nnode;
+       }
+       nnode = &path->nnode;
+       path->in_tree = 0;
+       path->ptr.nnode = nnode;
+       memset(nnode, 0, sizeof(struct ubifs_nnode));
+       if (branch->lnum == 0) {
+               /*
+                * This nnode was not written which just means that the LEB
+                * properties in the subtree below it describe empty LEBs. We
+                * make the nnode as though we had read it, which in fact means
+                * doing almost nothing.
+                */
+               if (c->big_lpt)
+                       nnode->num = calc_nnode_num_from_parent(c, parent, iip);
+       } else {
+               err = ubifs_leb_read(c, branch->lnum, buf, branch->offs,
+                                    c->nnode_sz, 1);
+               if (err)
+                       return ERR_PTR(err);
+               err = ubifs_unpack_nnode(c, buf, nnode);
+               if (err)
+                       return ERR_PTR(err);
+       }
+       err = validate_nnode(c, nnode, parent, iip);
+       if (err)
+               return ERR_PTR(err);
+       if (!c->big_lpt)
+               nnode->num = calc_nnode_num_from_parent(c, parent, iip);
+       nnode->level = parent->level - 1;
+       nnode->parent = parent;
+       nnode->iip = iip;
+       return nnode;
+}
+
+/**
+ * scan_get_pnode - for the scan, get a pnode from either the tree or flash.
+ * @c: the UBIFS file-system description object
+ * @path: where to put the pnode
+ * @parent: parent of the pnode
+ * @iip: index in parent of the pnode
+ *
+ * This function returns a pointer to the pnode on success or a negative error
+ * code on failure.
+ */
+static struct ubifs_pnode *scan_get_pnode(struct ubifs_info *c,
+                                         struct lpt_scan_node *path,
+                                         struct ubifs_nnode *parent, int iip)
+{
+       struct ubifs_nbranch *branch;
+       struct ubifs_pnode *pnode;
+       void *buf = c->lpt_nod_buf;
+       int err;
+
+       branch = &parent->nbranch[iip];
+       pnode = branch->pnode;
+       if (pnode) {
+               path->in_tree = 1;
+               path->ptr.pnode = pnode;
+               return pnode;
+       }
+       pnode = &path->pnode;
+       path->in_tree = 0;
+       path->ptr.pnode = pnode;
+       memset(pnode, 0, sizeof(struct ubifs_pnode));
+       if (branch->lnum == 0) {
+               /*
+                * This pnode was not written which just means that the LEB
+                * properties in it describe empty LEBs. We make the pnode as
+                * though we had read it.
+                */
+               int i;
+
+               if (c->big_lpt)
+                       pnode->num = calc_pnode_num_from_parent(c, parent, iip);
+               for (i = 0; i < UBIFS_LPT_FANOUT; i++) {
+                       struct ubifs_lprops * const lprops = &pnode->lprops[i];
+
+                       lprops->free = c->leb_size;
+                       lprops->flags = ubifs_categorize_lprops(c, lprops);
+               }
+       } else {
+               ubifs_assert(branch->lnum >= c->lpt_first &&
+                            branch->lnum <= c->lpt_last);
+               ubifs_assert(branch->offs >= 0 && branch->offs < c->leb_size);
+               err = ubifs_leb_read(c, branch->lnum, buf, branch->offs,
+                                    c->pnode_sz, 1);
+               if (err)
+                       return ERR_PTR(err);
+               err = unpack_pnode(c, buf, pnode);
+               if (err)
+                       return ERR_PTR(err);
+       }
+       err = validate_pnode(c, pnode, parent, iip);
+       if (err)
+               return ERR_PTR(err);
+       if (!c->big_lpt)
+               pnode->num = calc_pnode_num_from_parent(c, parent, iip);
+       pnode->parent = parent;
+       pnode->iip = iip;
+       set_pnode_lnum(c, pnode);
+       return pnode;
+}
+
+/**
+ * ubifs_lpt_scan_nolock - scan the LPT.
+ * @c: the UBIFS file-system description object
+ * @start_lnum: LEB number from which to start scanning
+ * @end_lnum: LEB number at which to stop scanning
+ * @scan_cb: callback function called for each lprops
+ * @data: data to be passed to the callback function
+ *
+ * This function returns %0 on success and a negative error code on failure.
+ */
+int ubifs_lpt_scan_nolock(struct ubifs_info *c, int start_lnum, int end_lnum,
+                         ubifs_lpt_scan_callback scan_cb, void *data)
+{
+       int err = 0, i, h, iip, shft;
+       struct ubifs_nnode *nnode;
+       struct ubifs_pnode *pnode;
+       struct lpt_scan_node *path;
+
+       if (start_lnum == -1) {
+               start_lnum = end_lnum + 1;
+               if (start_lnum >= c->leb_cnt)
+                       start_lnum = c->main_first;
+       }
+
+       ubifs_assert(start_lnum >= c->main_first && start_lnum < c->leb_cnt);
+       ubifs_assert(end_lnum >= c->main_first && end_lnum < c->leb_cnt);
+
+       if (!c->nroot) {
+               err = ubifs_read_nnode(c, NULL, 0);
                if (err)
                        return err;
        }
 
+       path = kmalloc(sizeof(struct lpt_scan_node) * (c->lpt_hght + 1),
+                      GFP_NOFS);
+       if (!path)
+               return -ENOMEM;
+
+       path[0].ptr.nnode = c->nroot;
+       path[0].in_tree = 1;
+again:
+       /* Descend to the pnode containing start_lnum */
+       nnode = c->nroot;
+       i = start_lnum - c->main_first;
+       shft = c->lpt_hght * UBIFS_LPT_FANOUT_SHIFT;
+       for (h = 1; h < c->lpt_hght; h++) {
+               iip = ((i >> shft) & (UBIFS_LPT_FANOUT - 1));
+               shft -= UBIFS_LPT_FANOUT_SHIFT;
+               nnode = scan_get_nnode(c, path + h, nnode, iip);
+               if (IS_ERR(nnode)) {
+                       err = PTR_ERR(nnode);
+                       goto out;
+               }
+       }
+       iip = ((i >> shft) & (UBIFS_LPT_FANOUT - 1));
+       shft -= UBIFS_LPT_FANOUT_SHIFT;
+       pnode = scan_get_pnode(c, path + h, nnode, iip);
+       if (IS_ERR(pnode)) {
+               err = PTR_ERR(pnode);
+               goto out;
+       }
+       iip = (i & (UBIFS_LPT_FANOUT - 1));
+
+       /* Loop for each lprops */
+       while (1) {
+               struct ubifs_lprops *lprops = &pnode->lprops[iip];
+               int ret, lnum = lprops->lnum;
+
+               ret = scan_cb(c, lprops, path[h].in_tree, data);
+               if (ret < 0) {
+                       err = ret;
+                       goto out;
+               }
+               if (ret & LPT_SCAN_ADD) {
+                       /* Add all the nodes in path to the tree in memory */
+                       for (h = 1; h < c->lpt_hght; h++) {
+                               const size_t sz = sizeof(struct ubifs_nnode);
+                               struct ubifs_nnode *parent;
+
+                               if (path[h].in_tree)
+                                       continue;
+                               nnode = kmemdup(&path[h].nnode, sz, GFP_NOFS);
+                               if (!nnode) {
+                                       err = -ENOMEM;
+                                       goto out;
+                               }
+                               parent = nnode->parent;
+                               parent->nbranch[nnode->iip].nnode = nnode;
+                               path[h].ptr.nnode = nnode;
+                               path[h].in_tree = 1;
+                               path[h + 1].cnode.parent = nnode;
+                       }
+                       if (path[h].in_tree)
+                               ubifs_ensure_cat(c, lprops);
+                       else {
+                               const size_t sz = sizeof(struct ubifs_pnode);
+                               struct ubifs_nnode *parent;
+
+                               pnode = kmemdup(&path[h].pnode, sz, GFP_NOFS);
+                               if (!pnode) {
+                                       err = -ENOMEM;
+                                       goto out;
+                               }
+                               parent = pnode->parent;
+                               parent->nbranch[pnode->iip].pnode = pnode;
+                               path[h].ptr.pnode = pnode;
+                               path[h].in_tree = 1;
+                               update_cats(c, pnode);
+                               c->pnodes_have += 1;
+                       }
+                       err = dbg_check_lpt_nodes(c, (struct ubifs_cnode *)
+                                                 c->nroot, 0, 0);
+                       if (err)
+                               goto out;
+                       err = dbg_check_cats(c);
+                       if (err)
+                               goto out;
+               }
+               if (ret & LPT_SCAN_STOP) {
+                       err = 0;
+                       break;
+               }
+               /* Get the next lprops */
+               if (lnum == end_lnum) {
+                       /*
+                        * We got to the end without finding what we were
+                        * looking for
+                        */
+                       err = -ENOSPC;
+                       goto out;
+               }
+               if (lnum + 1 >= c->leb_cnt) {
+                       /* Wrap-around to the beginning */
+                       start_lnum = c->main_first;
+                       goto again;
+               }
+               if (iip + 1 < UBIFS_LPT_FANOUT) {
+                       /* Next lprops is in the same pnode */
+                       iip += 1;
+                       continue;
+               }
+               /* We need to get the next pnode. Go up until we can go right */
+               iip = pnode->iip;
+               while (1) {
+                       h -= 1;
+                       ubifs_assert(h >= 0);
+                       nnode = path[h].ptr.nnode;
+                       if (iip + 1 < UBIFS_LPT_FANOUT)
+                               break;
+                       iip = nnode->iip;
+               }
+               /* Go right */
+               iip += 1;
+               /* Descend to the pnode */
+               h += 1;
+               for (; h < c->lpt_hght; h++) {
+                       nnode = scan_get_nnode(c, path + h, nnode, iip);
+                       if (IS_ERR(nnode)) {
+                               err = PTR_ERR(nnode);
+                               goto out;
+                       }
+                       iip = 0;
+               }
+               pnode = scan_get_pnode(c, path + h, nnode, iip);
+               if (IS_ERR(pnode)) {
+                       err = PTR_ERR(pnode);
+                       goto out;
+               }
+               iip = 0;
+       }
+out:
+       kfree(path);
+       return err;
+}
+
+/**
+ * dbg_chk_pnode - check a pnode.
+ * @c: the UBIFS file-system description object
+ * @pnode: pnode to check
+ * @col: pnode column
+ *
+ * This function returns %0 on success and a negative error code on failure.
+ */
+static int dbg_chk_pnode(struct ubifs_info *c, struct ubifs_pnode *pnode,
+                        int col)
+{
+       int i;
+
+       if (pnode->num != col) {
+               ubifs_err("pnode num %d expected %d parent num %d iip %d",
+                         pnode->num, col, pnode->parent->num, pnode->iip);
+               return -EINVAL;
+       }
+       for (i = 0; i < UBIFS_LPT_FANOUT; i++) {
+               struct ubifs_lprops *lp, *lprops = &pnode->lprops[i];
+               int lnum = (pnode->num << UBIFS_LPT_FANOUT_SHIFT) + i +
+                          c->main_first;
+               int found, cat = lprops->flags & LPROPS_CAT_MASK;
+               struct ubifs_lpt_heap *heap;
+               struct list_head *list = NULL;
+
+               if (lnum >= c->leb_cnt)
+                       continue;
+               if (lprops->lnum != lnum) {
+                       ubifs_err("bad LEB number %d expected %d",
+                                 lprops->lnum, lnum);
+                       return -EINVAL;
+               }
+               if (lprops->flags & LPROPS_TAKEN) {
+                       if (cat != LPROPS_UNCAT) {
+                               ubifs_err("LEB %d taken but not uncat %d",
+                                         lprops->lnum, cat);
+                               return -EINVAL;
+                       }
+                       continue;
+               }
+               if (lprops->flags & LPROPS_INDEX) {
+                       switch (cat) {
+                       case LPROPS_UNCAT:
+                       case LPROPS_DIRTY_IDX:
+                       case LPROPS_FRDI_IDX:
+                               break;
+                       default:
+                               ubifs_err("LEB %d index but cat %d",
+                                         lprops->lnum, cat);
+                               return -EINVAL;
+                       }
+               } else {
+                       switch (cat) {
+                       case LPROPS_UNCAT:
+                       case LPROPS_DIRTY:
+                       case LPROPS_FREE:
+                       case LPROPS_EMPTY:
+                       case LPROPS_FREEABLE:
+                               break;
+                       default:
+                               ubifs_err("LEB %d not index but cat %d",
+                                         lprops->lnum, cat);
+                               return -EINVAL;
+                       }
+               }
+               switch (cat) {
+               case LPROPS_UNCAT:
+                       list = &c->uncat_list;
+                       break;
+               case LPROPS_EMPTY:
+                       list = &c->empty_list;
+                       break;
+               case LPROPS_FREEABLE:
+                       list = &c->freeable_list;
+                       break;
+               case LPROPS_FRDI_IDX:
+                       list = &c->frdi_idx_list;
+                       break;
+               }
+               found = 0;
+               switch (cat) {
+               case LPROPS_DIRTY:
+               case LPROPS_DIRTY_IDX:
+               case LPROPS_FREE:
+                       heap = &c->lpt_heap[cat - 1];
+                       if (lprops->hpos < heap->cnt &&
+                           heap->arr[lprops->hpos] == lprops)
+                               found = 1;
+                       break;
+               case LPROPS_UNCAT:
+               case LPROPS_EMPTY:
+               case LPROPS_FREEABLE:
+               case LPROPS_FRDI_IDX:
+                       list_for_each_entry(lp, list, list)
+                               if (lprops == lp) {
+                                       found = 1;
+                                       break;
+                               }
+                       break;
+               }
+               if (!found) {
+                       ubifs_err("LEB %d cat %d not found in cat heap/list",
+                                 lprops->lnum, cat);
+                       return -EINVAL;
+               }
+               switch (cat) {
+               case LPROPS_EMPTY:
+                       if (lprops->free != c->leb_size) {
+                               ubifs_err("LEB %d cat %d free %d dirty %d",
+                                         lprops->lnum, cat, lprops->free,
+                                         lprops->dirty);
+                               return -EINVAL;
+                       }
+               case LPROPS_FREEABLE:
+               case LPROPS_FRDI_IDX:
+                       if (lprops->free + lprops->dirty != c->leb_size) {
+                               ubifs_err("LEB %d cat %d free %d dirty %d",
+                                         lprops->lnum, cat, lprops->free,
+                                         lprops->dirty);
+                               return -EINVAL;
+                       }
+               }
+       }
+       return 0;
+}
+
+/**
+ * dbg_check_lpt_nodes - check nnodes and pnodes.
+ * @c: the UBIFS file-system description object
+ * @cnode: next cnode (nnode or pnode) to check
+ * @row: row of cnode (root is zero)
+ * @col: column of cnode (leftmost is zero)
+ *
+ * This function returns %0 on success and a negative error code on failure.
+ */
+int dbg_check_lpt_nodes(struct ubifs_info *c, struct ubifs_cnode *cnode,
+                       int row, int col)
+{
+       struct ubifs_nnode *nnode, *nn;
+       struct ubifs_cnode *cn;
+       int num, iip = 0, err;
+
+       if (!dbg_is_chk_lprops(c))
+               return 0;
+
+       while (cnode) {
+               ubifs_assert(row >= 0);
+               nnode = cnode->parent;
+               if (cnode->level) {
+                       /* cnode is a nnode */
+                       num = calc_nnode_num(row, col);
+                       if (cnode->num != num) {
+                               ubifs_err("nnode num %d expected %d parent num %d iip %d",
+                                         cnode->num, num,
+                                         (nnode ? nnode->num : 0), cnode->iip);
+                               return -EINVAL;
+                       }
+                       nn = (struct ubifs_nnode *)cnode;
+                       while (iip < UBIFS_LPT_FANOUT) {
+                               cn = nn->nbranch[iip].cnode;
+                               if (cn) {
+                                       /* Go down */
+                                       row += 1;
+                                       col <<= UBIFS_LPT_FANOUT_SHIFT;
+                                       col += iip;
+                                       iip = 0;
+                                       cnode = cn;
+                                       break;
+                               }
+                               /* Go right */
+                               iip += 1;
+                       }
+                       if (iip < UBIFS_LPT_FANOUT)
+                               continue;
+               } else {
+                       struct ubifs_pnode *pnode;
+
+                       /* cnode is a pnode */
+                       pnode = (struct ubifs_pnode *)cnode;
+                       err = dbg_chk_pnode(c, pnode, col);
+                       if (err)
+                               return err;
+               }
+               /* Go up and to the right */
+               row -= 1;
+               col >>= UBIFS_LPT_FANOUT_SHIFT;
+               iip = cnode->iip + 1;
+               cnode = (struct ubifs_cnode *)nnode;
+       }
        return 0;
 }
index c0af8187acac7dfb1e96acce5c35c9f7c22963c6..cad422e06623a48254b42c51af64f764a8aeff94 100644 (file)
  *
  * Copyright (C) 2006-2008 Nokia Corporation.
  *
- * This program is free software; you can redistribute it and/or modify it
- * under the terms of the GNU General Public License version 2 as published by
- * the Free Software Foundation.
+ * SPDX-License-Identifier:    GPL-2.0+
  *
- * This program is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
- * more details.
+ * Authors: Adrian Hunter
+ *          Artem Bityutskiy (Битюцкий Артём)
+ */
+
+/*
+ * This file implements commit-related functionality of the LEB properties
+ * subsystem.
+ */
+
+#define __UBOOT__
+#ifndef __UBOOT__
+#include <linux/crc16.h>
+#include <linux/slab.h>
+#include <linux/random.h>
+#else
+#include <linux/compat.h>
+#include <linux/err.h>
+#include "crc16.h"
+#endif
+#include "ubifs.h"
+
+#ifndef __UBOOT__
+static int dbg_populate_lsave(struct ubifs_info *c);
+#endif
+
+/**
+ * first_dirty_cnode - find first dirty cnode.
+ * @c: UBIFS file-system description object
+ * @nnode: nnode at which to start
+ *
+ * This function returns the first dirty cnode or %NULL if there is not one.
+ */
+static struct ubifs_cnode *first_dirty_cnode(struct ubifs_nnode *nnode)
+{
+       ubifs_assert(nnode);
+       while (1) {
+               int i, cont = 0;
+
+               for (i = 0; i < UBIFS_LPT_FANOUT; i++) {
+                       struct ubifs_cnode *cnode;
+
+                       cnode = nnode->nbranch[i].cnode;
+                       if (cnode &&
+                           test_bit(DIRTY_CNODE, &cnode->flags)) {
+                               if (cnode->level == 0)
+                                       return cnode;
+                               nnode = (struct ubifs_nnode *)cnode;
+                               cont = 1;
+                               break;
+                       }
+               }
+               if (!cont)
+                       return (struct ubifs_cnode *)nnode;
+       }
+}
+
+/**
+ * next_dirty_cnode - find next dirty cnode.
+ * @cnode: cnode from which to begin searching
+ *
+ * This function returns the next dirty cnode or %NULL if there is not one.
+ */
+static struct ubifs_cnode *next_dirty_cnode(struct ubifs_cnode *cnode)
+{
+       struct ubifs_nnode *nnode;
+       int i;
+
+       ubifs_assert(cnode);
+       nnode = cnode->parent;
+       if (!nnode)
+               return NULL;
+       for (i = cnode->iip + 1; i < UBIFS_LPT_FANOUT; i++) {
+               cnode = nnode->nbranch[i].cnode;
+               if (cnode && test_bit(DIRTY_CNODE, &cnode->flags)) {
+                       if (cnode->level == 0)
+                               return cnode; /* cnode is a pnode */
+                       /* cnode is a nnode */
+                       return first_dirty_cnode((struct ubifs_nnode *)cnode);
+               }
+       }
+       return (struct ubifs_cnode *)nnode;
+}
+
+/**
+ * get_cnodes_to_commit - create list of dirty cnodes to commit.
+ * @c: UBIFS file-system description object
+ *
+ * This function returns the number of cnodes to commit.
+ */
+static int get_cnodes_to_commit(struct ubifs_info *c)
+{
+       struct ubifs_cnode *cnode, *cnext;
+       int cnt = 0;
+
+       if (!c->nroot)
+               return 0;
+
+       if (!test_bit(DIRTY_CNODE, &c->nroot->flags))
+               return 0;
+
+       c->lpt_cnext = first_dirty_cnode(c->nroot);
+       cnode = c->lpt_cnext;
+       if (!cnode)
+               return 0;
+       cnt += 1;
+       while (1) {
+               ubifs_assert(!test_bit(COW_CNODE, &cnode->flags));
+               __set_bit(COW_CNODE, &cnode->flags);
+               cnext = next_dirty_cnode(cnode);
+               if (!cnext) {
+                       cnode->cnext = c->lpt_cnext;
+                       break;
+               }
+               cnode->cnext = cnext;
+               cnode = cnext;
+               cnt += 1;
+       }
+       dbg_cmt("committing %d cnodes", cnt);
+       dbg_lp("committing %d cnodes", cnt);
+       ubifs_assert(cnt == c->dirty_nn_cnt + c->dirty_pn_cnt);
+       return cnt;
+}
+
+/**
+ * upd_ltab - update LPT LEB properties.
+ * @c: UBIFS file-system description object
+ * @lnum: LEB number
+ * @free: amount of free space
+ * @dirty: amount of dirty space to add
+ */
+static void upd_ltab(struct ubifs_info *c, int lnum, int free, int dirty)
+{
+       dbg_lp("LEB %d free %d dirty %d to %d +%d",
+              lnum, c->ltab[lnum - c->lpt_first].free,
+              c->ltab[lnum - c->lpt_first].dirty, free, dirty);
+       ubifs_assert(lnum >= c->lpt_first && lnum <= c->lpt_last);
+       c->ltab[lnum - c->lpt_first].free = free;
+       c->ltab[lnum - c->lpt_first].dirty += dirty;
+}
+
+/**
+ * alloc_lpt_leb - allocate an LPT LEB that is empty.
+ * @c: UBIFS file-system description object
+ * @lnum: LEB number is passed and returned here
+ *
+ * This function finds the next empty LEB in the ltab starting from @lnum. If a
+ * an empty LEB is found it is returned in @lnum and the function returns %0.
+ * Otherwise the function returns -ENOSPC.  Note however, that LPT is designed
+ * never to run out of space.
+ */
+static int alloc_lpt_leb(struct ubifs_info *c, int *lnum)
+{
+       int i, n;
+
+       n = *lnum - c->lpt_first + 1;
+       for (i = n; i < c->lpt_lebs; i++) {
+               if (c->ltab[i].tgc || c->ltab[i].cmt)
+                       continue;
+               if (c->ltab[i].free == c->leb_size) {
+                       c->ltab[i].cmt = 1;
+                       *lnum = i + c->lpt_first;
+                       return 0;
+               }
+       }
+
+       for (i = 0; i < n; i++) {
+               if (c->ltab[i].tgc || c->ltab[i].cmt)
+                       continue;
+               if (c->ltab[i].free == c->leb_size) {
+                       c->ltab[i].cmt = 1;
+                       *lnum = i + c->lpt_first;
+                       return 0;
+               }
+       }
+       return -ENOSPC;
+}
+
+/**
+ * layout_cnodes - layout cnodes for commit.
+ * @c: UBIFS file-system description object
+ *
+ * This function returns %0 on success and a negative error code on failure.
+ */
+static int layout_cnodes(struct ubifs_info *c)
+{
+       int lnum, offs, len, alen, done_lsave, done_ltab, err;
+       struct ubifs_cnode *cnode;
+
+       err = dbg_chk_lpt_sz(c, 0, 0);
+       if (err)
+               return err;
+       cnode = c->lpt_cnext;
+       if (!cnode)
+               return 0;
+       lnum = c->nhead_lnum;
+       offs = c->nhead_offs;
+       /* Try to place lsave and ltab nicely */
+       done_lsave = !c->big_lpt;
+       done_ltab = 0;
+       if (!done_lsave && offs + c->lsave_sz <= c->leb_size) {
+               done_lsave = 1;
+               c->lsave_lnum = lnum;
+               c->lsave_offs = offs;
+               offs += c->lsave_sz;
+               dbg_chk_lpt_sz(c, 1, c->lsave_sz);
+       }
+
+       if (offs + c->ltab_sz <= c->leb_size) {
+               done_ltab = 1;
+               c->ltab_lnum = lnum;
+               c->ltab_offs = offs;
+               offs += c->ltab_sz;
+               dbg_chk_lpt_sz(c, 1, c->ltab_sz);
+       }
+
+       do {
+               if (cnode->level) {
+                       len = c->nnode_sz;
+                       c->dirty_nn_cnt -= 1;
+               } else {
+                       len = c->pnode_sz;
+                       c->dirty_pn_cnt -= 1;
+               }
+               while (offs + len > c->leb_size) {
+                       alen = ALIGN(offs, c->min_io_size);
+                       upd_ltab(c, lnum, c->leb_size - alen, alen - offs);
+                       dbg_chk_lpt_sz(c, 2, c->leb_size - offs);
+                       err = alloc_lpt_leb(c, &lnum);
+                       if (err)
+                               goto no_space;
+                       offs = 0;
+                       ubifs_assert(lnum >= c->lpt_first &&
+                                    lnum <= c->lpt_last);
+                       /* Try to place lsave and ltab nicely */
+                       if (!done_lsave) {
+                               done_lsave = 1;
+                               c->lsave_lnum = lnum;
+                               c->lsave_offs = offs;
+                               offs += c->lsave_sz;
+                               dbg_chk_lpt_sz(c, 1, c->lsave_sz);
+                               continue;
+                       }
+                       if (!done_ltab) {
+                               done_ltab = 1;
+                               c->ltab_lnum = lnum;
+                               c->ltab_offs = offs;
+                               offs += c->ltab_sz;
+                               dbg_chk_lpt_sz(c, 1, c->ltab_sz);
+                               continue;
+                       }
+                       break;
+               }
+               if (cnode->parent) {
+                       cnode->parent->nbranch[cnode->iip].lnum = lnum;
+                       cnode->parent->nbranch[cnode->iip].offs = offs;
+               } else {
+                       c->lpt_lnum = lnum;
+                       c->lpt_offs = offs;
+               }
+               offs += len;
+               dbg_chk_lpt_sz(c, 1, len);
+               cnode = cnode->cnext;
+       } while (cnode && cnode != c->lpt_cnext);
+
+       /* Make sure to place LPT's save table */
+       if (!done_lsave) {
+               if (offs + c->lsave_sz > c->leb_size) {
+                       alen = ALIGN(offs, c->min_io_size);
+                       upd_ltab(c, lnum, c->leb_size - alen, alen - offs);
+                       dbg_chk_lpt_sz(c, 2, c->leb_size - offs);
+                       err = alloc_lpt_leb(c, &lnum);
+                       if (err)
+                               goto no_space;
+                       offs = 0;
+                       ubifs_assert(lnum >= c->lpt_first &&
+                                    lnum <= c->lpt_last);
+               }
+               done_lsave = 1;
+               c->lsave_lnum = lnum;
+               c->lsave_offs = offs;
+               offs += c->lsave_sz;
+               dbg_chk_lpt_sz(c, 1, c->lsave_sz);
+       }
+
+       /* Make sure to place LPT's own lprops table */
+       if (!done_ltab) {
+               if (offs + c->ltab_sz > c->leb_size) {
+                       alen = ALIGN(offs, c->min_io_size);
+                       upd_ltab(c, lnum, c->leb_size - alen, alen - offs);
+                       dbg_chk_lpt_sz(c, 2, c->leb_size - offs);
+                       err = alloc_lpt_leb(c, &lnum);
+                       if (err)
+                               goto no_space;
+                       offs = 0;
+                       ubifs_assert(lnum >= c->lpt_first &&
+                                    lnum <= c->lpt_last);
+               }
+               done_ltab = 1;
+               c->ltab_lnum = lnum;
+               c->ltab_offs = offs;
+               offs += c->ltab_sz;
+               dbg_chk_lpt_sz(c, 1, c->ltab_sz);
+       }
+
+       alen = ALIGN(offs, c->min_io_size);
+       upd_ltab(c, lnum, c->leb_size - alen, alen - offs);
+       dbg_chk_lpt_sz(c, 4, alen - offs);
+       err = dbg_chk_lpt_sz(c, 3, alen);
+       if (err)
+               return err;
+       return 0;
+
+no_space:
+       ubifs_err("LPT out of space at LEB %d:%d needing %d, done_ltab %d, done_lsave %d",
+                 lnum, offs, len, done_ltab, done_lsave);
+       ubifs_dump_lpt_info(c);
+       ubifs_dump_lpt_lebs(c);
+       dump_stack();
+       return err;
+}
+
+#ifndef __UBOOT__
+/**
+ * realloc_lpt_leb - allocate an LPT LEB that is empty.
+ * @c: UBIFS file-system description object
+ * @lnum: LEB number is passed and returned here
+ *
+ * This function duplicates exactly the results of the function alloc_lpt_leb.
+ * It is used during end commit to reallocate the same LEB numbers that were
+ * allocated by alloc_lpt_leb during start commit.
+ *
+ * This function finds the next LEB that was allocated by the alloc_lpt_leb
+ * function starting from @lnum. If a LEB is found it is returned in @lnum and
+ * the function returns %0. Otherwise the function returns -ENOSPC.
+ * Note however, that LPT is designed never to run out of space.
+ */
+static int realloc_lpt_leb(struct ubifs_info *c, int *lnum)
+{
+       int i, n;
+
+       n = *lnum - c->lpt_first + 1;
+       for (i = n; i < c->lpt_lebs; i++)
+               if (c->ltab[i].cmt) {
+                       c->ltab[i].cmt = 0;
+                       *lnum = i + c->lpt_first;
+                       return 0;
+               }
+
+       for (i = 0; i < n; i++)
+               if (c->ltab[i].cmt) {
+                       c->ltab[i].cmt = 0;
+                       *lnum = i + c->lpt_first;
+                       return 0;
+               }
+       return -ENOSPC;
+}
+
+/**
+ * write_cnodes - write cnodes for commit.
+ * @c: UBIFS file-system description object
+ *
+ * This function returns %0 on success and a negative error code on failure.
+ */
+static int write_cnodes(struct ubifs_info *c)
+{
+       int lnum, offs, len, from, err, wlen, alen, done_ltab, done_lsave;
+       struct ubifs_cnode *cnode;
+       void *buf = c->lpt_buf;
+
+       cnode = c->lpt_cnext;
+       if (!cnode)
+               return 0;
+       lnum = c->nhead_lnum;
+       offs = c->nhead_offs;
+       from = offs;
+       /* Ensure empty LEB is unmapped */
+       if (offs == 0) {
+               err = ubifs_leb_unmap(c, lnum);
+               if (err)
+                       return err;
+       }
+       /* Try to place lsave and ltab nicely */
+       done_lsave = !c->big_lpt;
+       done_ltab = 0;
+       if (!done_lsave && offs + c->lsave_sz <= c->leb_size) {
+               done_lsave = 1;
+               ubifs_pack_lsave(c, buf + offs, c->lsave);
+               offs += c->lsave_sz;
+               dbg_chk_lpt_sz(c, 1, c->lsave_sz);
+       }
+
+       if (offs + c->ltab_sz <= c->leb_size) {
+               done_ltab = 1;
+               ubifs_pack_ltab(c, buf + offs, c->ltab_cmt);
+               offs += c->ltab_sz;
+               dbg_chk_lpt_sz(c, 1, c->ltab_sz);
+       }
+
+       /* Loop for each cnode */
+       do {
+               if (cnode->level)
+                       len = c->nnode_sz;
+               else
+                       len = c->pnode_sz;
+               while (offs + len > c->leb_size) {
+                       wlen = offs - from;
+                       if (wlen) {
+                               alen = ALIGN(wlen, c->min_io_size);
+                               memset(buf + offs, 0xff, alen - wlen);
+                               err = ubifs_leb_write(c, lnum, buf + from, from,
+                                                      alen);
+                               if (err)
+                                       return err;
+                       }
+                       dbg_chk_lpt_sz(c, 2, c->leb_size - offs);
+                       err = realloc_lpt_leb(c, &lnum);
+                       if (err)
+                               goto no_space;
+                       offs = from = 0;
+                       ubifs_assert(lnum >= c->lpt_first &&
+                                    lnum <= c->lpt_last);
+                       err = ubifs_leb_unmap(c, lnum);
+                       if (err)
+                               return err;
+                       /* Try to place lsave and ltab nicely */
+                       if (!done_lsave) {
+                               done_lsave = 1;
+                               ubifs_pack_lsave(c, buf + offs, c->lsave);
+                               offs += c->lsave_sz;
+                               dbg_chk_lpt_sz(c, 1, c->lsave_sz);
+                               continue;
+                       }
+                       if (!done_ltab) {
+                               done_ltab = 1;
+                               ubifs_pack_ltab(c, buf + offs, c->ltab_cmt);
+                               offs += c->ltab_sz;
+                               dbg_chk_lpt_sz(c, 1, c->ltab_sz);
+                               continue;
+                       }
+                       break;
+               }
+               if (cnode->level)
+                       ubifs_pack_nnode(c, buf + offs,
+                                        (struct ubifs_nnode *)cnode);
+               else
+                       ubifs_pack_pnode(c, buf + offs,
+                                        (struct ubifs_pnode *)cnode);
+               /*
+                * The reason for the barriers is the same as in case of TNC.
+                * See comment in 'write_index()'. 'dirty_cow_nnode()' and
+                * 'dirty_cow_pnode()' are the functions for which this is
+                * important.
+                */
+               clear_bit(DIRTY_CNODE, &cnode->flags);
+               smp_mb__before_clear_bit();
+               clear_bit(COW_CNODE, &cnode->flags);
+               smp_mb__after_clear_bit();
+               offs += len;
+               dbg_chk_lpt_sz(c, 1, len);
+               cnode = cnode->cnext;
+       } while (cnode && cnode != c->lpt_cnext);
+
+       /* Make sure to place LPT's save table */
+       if (!done_lsave) {
+               if (offs + c->lsave_sz > c->leb_size) {
+                       wlen = offs - from;
+                       alen = ALIGN(wlen, c->min_io_size);
+                       memset(buf + offs, 0xff, alen - wlen);
+                       err = ubifs_leb_write(c, lnum, buf + from, from, alen);
+                       if (err)
+                               return err;
+                       dbg_chk_lpt_sz(c, 2, c->leb_size - offs);
+                       err = realloc_lpt_leb(c, &lnum);
+                       if (err)
+                               goto no_space;
+                       offs = from = 0;
+                       ubifs_assert(lnum >= c->lpt_first &&
+                                    lnum <= c->lpt_last);
+                       err = ubifs_leb_unmap(c, lnum);
+                       if (err)
+                               return err;
+               }
+               done_lsave = 1;
+               ubifs_pack_lsave(c, buf + offs, c->lsave);
+               offs += c->lsave_sz;
+               dbg_chk_lpt_sz(c, 1, c->lsave_sz);
+       }
+
+       /* Make sure to place LPT's own lprops table */
+       if (!done_ltab) {
+               if (offs + c->ltab_sz > c->leb_size) {
+                       wlen = offs - from;
+                       alen = ALIGN(wlen, c->min_io_size);
+                       memset(buf + offs, 0xff, alen - wlen);
+                       err = ubifs_leb_write(c, lnum, buf + from, from, alen);
+                       if (err)
+                               return err;
+                       dbg_chk_lpt_sz(c, 2, c->leb_size - offs);
+                       err = realloc_lpt_leb(c, &lnum);
+                       if (err)
+                               goto no_space;
+                       offs = from = 0;
+                       ubifs_assert(lnum >= c->lpt_first &&
+                                    lnum <= c->lpt_last);
+                       err = ubifs_leb_unmap(c, lnum);
+                       if (err)
+                               return err;
+               }
+               done_ltab = 1;
+               ubifs_pack_ltab(c, buf + offs, c->ltab_cmt);
+               offs += c->ltab_sz;
+               dbg_chk_lpt_sz(c, 1, c->ltab_sz);
+       }
+
+       /* Write remaining data in buffer */
+       wlen = offs - from;
+       alen = ALIGN(wlen, c->min_io_size);
+       memset(buf + offs, 0xff, alen - wlen);
+       err = ubifs_leb_write(c, lnum, buf + from, from, alen);
+       if (err)
+               return err;
+
+       dbg_chk_lpt_sz(c, 4, alen - wlen);
+       err = dbg_chk_lpt_sz(c, 3, ALIGN(offs, c->min_io_size));
+       if (err)
+               return err;
+
+       c->nhead_lnum = lnum;
+       c->nhead_offs = ALIGN(offs, c->min_io_size);
+
+       dbg_lp("LPT root is at %d:%d", c->lpt_lnum, c->lpt_offs);
+       dbg_lp("LPT head is at %d:%d", c->nhead_lnum, c->nhead_offs);
+       dbg_lp("LPT ltab is at %d:%d", c->ltab_lnum, c->ltab_offs);
+       if (c->big_lpt)
+               dbg_lp("LPT lsave is at %d:%d", c->lsave_lnum, c->lsave_offs);
+
+       return 0;
+
+no_space:
+       ubifs_err("LPT out of space mismatch at LEB %d:%d needing %d, done_ltab %d, done_lsave %d",
+                 lnum, offs, len, done_ltab, done_lsave);
+       ubifs_dump_lpt_info(c);
+       ubifs_dump_lpt_lebs(c);
+       dump_stack();
+       return err;
+}
+#endif
+
+/**
+ * next_pnode_to_dirty - find next pnode to dirty.
+ * @c: UBIFS file-system description object
+ * @pnode: pnode
+ *
+ * This function returns the next pnode to dirty or %NULL if there are no more
+ * pnodes.  Note that pnodes that have never been written (lnum == 0) are
+ * skipped.
+ */
+static struct ubifs_pnode *next_pnode_to_dirty(struct ubifs_info *c,
+                                              struct ubifs_pnode *pnode)
+{
+       struct ubifs_nnode *nnode;
+       int iip;
+
+       /* Try to go right */
+       nnode = pnode->parent;
+       for (iip = pnode->iip + 1; iip < UBIFS_LPT_FANOUT; iip++) {
+               if (nnode->nbranch[iip].lnum)
+                       return ubifs_get_pnode(c, nnode, iip);
+       }
+
+       /* Go up while can't go right */
+       do {
+               iip = nnode->iip + 1;
+               nnode = nnode->parent;
+               if (!nnode)
+                       return NULL;
+               for (; iip < UBIFS_LPT_FANOUT; iip++) {
+                       if (nnode->nbranch[iip].lnum)
+                               break;
+               }
+       } while (iip >= UBIFS_LPT_FANOUT);
+
+       /* Go right */
+       nnode = ubifs_get_nnode(c, nnode, iip);
+       if (IS_ERR(nnode))
+               return (void *)nnode;
+
+       /* Go down to level 1 */
+       while (nnode->level > 1) {
+               for (iip = 0; iip < UBIFS_LPT_FANOUT; iip++) {
+                       if (nnode->nbranch[iip].lnum)
+                               break;
+               }
+               if (iip >= UBIFS_LPT_FANOUT) {
+                       /*
+                        * Should not happen, but we need to keep going
+                        * if it does.
+                        */
+                       iip = 0;
+               }
+               nnode = ubifs_get_nnode(c, nnode, iip);
+               if (IS_ERR(nnode))
+                       return (void *)nnode;
+       }
+
+       for (iip = 0; iip < UBIFS_LPT_FANOUT; iip++)
+               if (nnode->nbranch[iip].lnum)
+                       break;
+       if (iip >= UBIFS_LPT_FANOUT)
+               /* Should not happen, but we need to keep going if it does */
+               iip = 0;
+       return ubifs_get_pnode(c, nnode, iip);
+}
+
+/**
+ * pnode_lookup - lookup a pnode in the LPT.
+ * @c: UBIFS file-system description object
+ * @i: pnode number (0 to main_lebs - 1)
+ *
+ * This function returns a pointer to the pnode on success or a negative
+ * error code on failure.
+ */
+static struct ubifs_pnode *pnode_lookup(struct ubifs_info *c, int i)
+{
+       int err, h, iip, shft;
+       struct ubifs_nnode *nnode;
+
+       if (!c->nroot) {
+               err = ubifs_read_nnode(c, NULL, 0);
+               if (err)
+                       return ERR_PTR(err);
+       }
+       i <<= UBIFS_LPT_FANOUT_SHIFT;
+       nnode = c->nroot;
+       shft = c->lpt_hght * UBIFS_LPT_FANOUT_SHIFT;
+       for (h = 1; h < c->lpt_hght; h++) {
+               iip = ((i >> shft) & (UBIFS_LPT_FANOUT - 1));
+               shft -= UBIFS_LPT_FANOUT_SHIFT;
+               nnode = ubifs_get_nnode(c, nnode, iip);
+               if (IS_ERR(nnode))
+                       return ERR_CAST(nnode);
+       }
+       iip = ((i >> shft) & (UBIFS_LPT_FANOUT - 1));
+       return ubifs_get_pnode(c, nnode, iip);
+}
+
+/**
+ * add_pnode_dirt - add dirty space to LPT LEB properties.
+ * @c: UBIFS file-system description object
+ * @pnode: pnode for which to add dirt
+ */
+static void add_pnode_dirt(struct ubifs_info *c, struct ubifs_pnode *pnode)
+{
+       ubifs_add_lpt_dirt(c, pnode->parent->nbranch[pnode->iip].lnum,
+                          c->pnode_sz);
+}
+
+/**
+ * do_make_pnode_dirty - mark a pnode dirty.
+ * @c: UBIFS file-system description object
+ * @pnode: pnode to mark dirty
+ */
+static void do_make_pnode_dirty(struct ubifs_info *c, struct ubifs_pnode *pnode)
+{
+       /* Assumes cnext list is empty i.e. not called during commit */
+       if (!test_and_set_bit(DIRTY_CNODE, &pnode->flags)) {
+               struct ubifs_nnode *nnode;
+
+               c->dirty_pn_cnt += 1;
+               add_pnode_dirt(c, pnode);
+               /* Mark parent and ancestors dirty too */
+               nnode = pnode->parent;
+               while (nnode) {
+                       if (!test_and_set_bit(DIRTY_CNODE, &nnode->flags)) {
+                               c->dirty_nn_cnt += 1;
+                               ubifs_add_nnode_dirt(c, nnode);
+                               nnode = nnode->parent;
+                       } else
+                               break;
+               }
+       }
+}
+
+/**
+ * make_tree_dirty - mark the entire LEB properties tree dirty.
+ * @c: UBIFS file-system description object
+ *
+ * This function is used by the "small" LPT model to cause the entire LEB
+ * properties tree to be written.  The "small" LPT model does not use LPT
+ * garbage collection because it is more efficient to write the entire tree
+ * (because it is small).
+ *
+ * This function returns %0 on success and a negative error code on failure.
+ */
+static int make_tree_dirty(struct ubifs_info *c)
+{
+       struct ubifs_pnode *pnode;
+
+       pnode = pnode_lookup(c, 0);
+       if (IS_ERR(pnode))
+               return PTR_ERR(pnode);
+
+       while (pnode) {
+               do_make_pnode_dirty(c, pnode);
+               pnode = next_pnode_to_dirty(c, pnode);
+               if (IS_ERR(pnode))
+                       return PTR_ERR(pnode);
+       }
+       return 0;
+}
+
+/**
+ * need_write_all - determine if the LPT area is running out of free space.
+ * @c: UBIFS file-system description object
+ *
+ * This function returns %1 if the LPT area is running out of free space and %0
+ * if it is not.
+ */
+static int need_write_all(struct ubifs_info *c)
+{
+       long long free = 0;
+       int i;
+
+       for (i = 0; i < c->lpt_lebs; i++) {
+               if (i + c->lpt_first == c->nhead_lnum)
+                       free += c->leb_size - c->nhead_offs;
+               else if (c->ltab[i].free == c->leb_size)
+                       free += c->leb_size;
+               else if (c->ltab[i].free + c->ltab[i].dirty == c->leb_size)
+                       free += c->leb_size;
+       }
+       /* Less than twice the size left */
+       if (free <= c->lpt_sz * 2)
+               return 1;
+       return 0;
+}
+
+/**
+ * lpt_tgc_start - start trivial garbage collection of LPT LEBs.
+ * @c: UBIFS file-system description object
+ *
+ * LPT trivial garbage collection is where a LPT LEB contains only dirty and
+ * free space and so may be reused as soon as the next commit is completed.
+ * This function is called during start commit to mark LPT LEBs for trivial GC.
+ */
+static void lpt_tgc_start(struct ubifs_info *c)
+{
+       int i;
+
+       for (i = 0; i < c->lpt_lebs; i++) {
+               if (i + c->lpt_first == c->nhead_lnum)
+                       continue;
+               if (c->ltab[i].dirty > 0 &&
+                   c->ltab[i].free + c->ltab[i].dirty == c->leb_size) {
+                       c->ltab[i].tgc = 1;
+                       c->ltab[i].free = c->leb_size;
+                       c->ltab[i].dirty = 0;
+                       dbg_lp("LEB %d", i + c->lpt_first);
+               }
+       }
+}
+
+/**
+ * lpt_tgc_end - end trivial garbage collection of LPT LEBs.
+ * @c: UBIFS file-system description object
+ *
+ * LPT trivial garbage collection is where a LPT LEB contains only dirty and
+ * free space and so may be reused as soon as the next commit is completed.
+ * This function is called after the commit is completed (master node has been
+ * written) and un-maps LPT LEBs that were marked for trivial GC.
+ */
+static int lpt_tgc_end(struct ubifs_info *c)
+{
+       int i, err;
+
+       for (i = 0; i < c->lpt_lebs; i++)
+               if (c->ltab[i].tgc) {
+                       err = ubifs_leb_unmap(c, i + c->lpt_first);
+                       if (err)
+                               return err;
+                       c->ltab[i].tgc = 0;
+                       dbg_lp("LEB %d", i + c->lpt_first);
+               }
+       return 0;
+}
+
+/**
+ * populate_lsave - fill the lsave array with important LEB numbers.
+ * @c: the UBIFS file-system description object
+ *
+ * This function is only called for the "big" model. It records a small number
+ * of LEB numbers of important LEBs.  Important LEBs are ones that are (from
+ * most important to least important): empty, freeable, freeable index, dirty
+ * index, dirty or free. Upon mount, we read this list of LEB numbers and bring
+ * their pnodes into memory.  That will stop us from having to scan the LPT
+ * straight away. For the "small" model we assume that scanning the LPT is no
+ * big deal.
+ */
+static void populate_lsave(struct ubifs_info *c)
+{
+       struct ubifs_lprops *lprops;
+       struct ubifs_lpt_heap *heap;
+       int i, cnt = 0;
+
+       ubifs_assert(c->big_lpt);
+       if (!(c->lpt_drty_flgs & LSAVE_DIRTY)) {
+               c->lpt_drty_flgs |= LSAVE_DIRTY;
+               ubifs_add_lpt_dirt(c, c->lsave_lnum, c->lsave_sz);
+       }
+
+#ifndef __UBOOT__
+       if (dbg_populate_lsave(c))
+               return;
+#endif
+
+       list_for_each_entry(lprops, &c->empty_list, list) {
+               c->lsave[cnt++] = lprops->lnum;
+               if (cnt >= c->lsave_cnt)
+                       return;
+       }
+       list_for_each_entry(lprops, &c->freeable_list, list) {
+               c->lsave[cnt++] = lprops->lnum;
+               if (cnt >= c->lsave_cnt)
+                       return;
+       }
+       list_for_each_entry(lprops, &c->frdi_idx_list, list) {
+               c->lsave[cnt++] = lprops->lnum;
+               if (cnt >= c->lsave_cnt)
+                       return;
+       }
+       heap = &c->lpt_heap[LPROPS_DIRTY_IDX - 1];
+       for (i = 0; i < heap->cnt; i++) {
+               c->lsave[cnt++] = heap->arr[i]->lnum;
+               if (cnt >= c->lsave_cnt)
+                       return;
+       }
+       heap = &c->lpt_heap[LPROPS_DIRTY - 1];
+       for (i = 0; i < heap->cnt; i++) {
+               c->lsave[cnt++] = heap->arr[i]->lnum;
+               if (cnt >= c->lsave_cnt)
+                       return;
+       }
+       heap = &c->lpt_heap[LPROPS_FREE - 1];
+       for (i = 0; i < heap->cnt; i++) {
+               c->lsave[cnt++] = heap->arr[i]->lnum;
+               if (cnt >= c->lsave_cnt)
+                       return;
+       }
+       /* Fill it up completely */
+       while (cnt < c->lsave_cnt)
+               c->lsave[cnt++] = c->main_first;
+}
+
+/**
+ * nnode_lookup - lookup a nnode in the LPT.
+ * @c: UBIFS file-system description object
+ * @i: nnode number
+ *
+ * This function returns a pointer to the nnode on success or a negative
+ * error code on failure.
+ */
+static struct ubifs_nnode *nnode_lookup(struct ubifs_info *c, int i)
+{
+       int err, iip;
+       struct ubifs_nnode *nnode;
+
+       if (!c->nroot) {
+               err = ubifs_read_nnode(c, NULL, 0);
+               if (err)
+                       return ERR_PTR(err);
+       }
+       nnode = c->nroot;
+       while (1) {
+               iip = i & (UBIFS_LPT_FANOUT - 1);
+               i >>= UBIFS_LPT_FANOUT_SHIFT;
+               if (!i)
+                       break;
+               nnode = ubifs_get_nnode(c, nnode, iip);
+               if (IS_ERR(nnode))
+                       return nnode;
+       }
+       return nnode;
+}
+
+/**
+ * make_nnode_dirty - find a nnode and, if found, make it dirty.
+ * @c: UBIFS file-system description object
+ * @node_num: nnode number of nnode to make dirty
+ * @lnum: LEB number where nnode was written
+ * @offs: offset where nnode was written
+ *
+ * This function is used by LPT garbage collection.  LPT garbage collection is
+ * used only for the "big" LPT model (c->big_lpt == 1).  Garbage collection
+ * simply involves marking all the nodes in the LEB being garbage-collected as
+ * dirty.  The dirty nodes are written next commit, after which the LEB is free
+ * to be reused.
+ *
+ * This function returns %0 on success and a negative error code on failure.
+ */
+static int make_nnode_dirty(struct ubifs_info *c, int node_num, int lnum,
+                           int offs)
+{
+       struct ubifs_nnode *nnode;
+
+       nnode = nnode_lookup(c, node_num);
+       if (IS_ERR(nnode))
+               return PTR_ERR(nnode);
+       if (nnode->parent) {
+               struct ubifs_nbranch *branch;
+
+               branch = &nnode->parent->nbranch[nnode->iip];
+               if (branch->lnum != lnum || branch->offs != offs)
+                       return 0; /* nnode is obsolete */
+       } else if (c->lpt_lnum != lnum || c->lpt_offs != offs)
+                       return 0; /* nnode is obsolete */
+       /* Assumes cnext list is empty i.e. not called during commit */
+       if (!test_and_set_bit(DIRTY_CNODE, &nnode->flags)) {
+               c->dirty_nn_cnt += 1;
+               ubifs_add_nnode_dirt(c, nnode);
+               /* Mark parent and ancestors dirty too */
+               nnode = nnode->parent;
+               while (nnode) {
+                       if (!test_and_set_bit(DIRTY_CNODE, &nnode->flags)) {
+                               c->dirty_nn_cnt += 1;
+                               ubifs_add_nnode_dirt(c, nnode);
+                               nnode = nnode->parent;
+                       } else
+                               break;
+               }
+       }
+       return 0;
+}
+
+/**
+ * make_pnode_dirty - find a pnode and, if found, make it dirty.
+ * @c: UBIFS file-system description object
+ * @node_num: pnode number of pnode to make dirty
+ * @lnum: LEB number where pnode was written
+ * @offs: offset where pnode was written
+ *
+ * This function is used by LPT garbage collection.  LPT garbage collection is
+ * used only for the "big" LPT model (c->big_lpt == 1).  Garbage collection
+ * simply involves marking all the nodes in the LEB being garbage-collected as
+ * dirty.  The dirty nodes are written next commit, after which the LEB is free
+ * to be reused.
+ *
+ * This function returns %0 on success and a negative error code on failure.
+ */
+static int make_pnode_dirty(struct ubifs_info *c, int node_num, int lnum,
+                           int offs)
+{
+       struct ubifs_pnode *pnode;
+       struct ubifs_nbranch *branch;
+
+       pnode = pnode_lookup(c, node_num);
+       if (IS_ERR(pnode))
+               return PTR_ERR(pnode);
+       branch = &pnode->parent->nbranch[pnode->iip];
+       if (branch->lnum != lnum || branch->offs != offs)
+               return 0;
+       do_make_pnode_dirty(c, pnode);
+       return 0;
+}
+
+/**
+ * make_ltab_dirty - make ltab node dirty.
+ * @c: UBIFS file-system description object
+ * @lnum: LEB number where ltab was written
+ * @offs: offset where ltab was written
+ *
+ * This function is used by LPT garbage collection.  LPT garbage collection is
+ * used only for the "big" LPT model (c->big_lpt == 1).  Garbage collection
+ * simply involves marking all the nodes in the LEB being garbage-collected as
+ * dirty.  The dirty nodes are written next commit, after which the LEB is free
+ * to be reused.
+ *
+ * This function returns %0 on success and a negative error code on failure.
+ */
+static int make_ltab_dirty(struct ubifs_info *c, int lnum, int offs)
+{
+       if (lnum != c->ltab_lnum || offs != c->ltab_offs)
+               return 0; /* This ltab node is obsolete */
+       if (!(c->lpt_drty_flgs & LTAB_DIRTY)) {
+               c->lpt_drty_flgs |= LTAB_DIRTY;
+               ubifs_add_lpt_dirt(c, c->ltab_lnum, c->ltab_sz);
+       }
+       return 0;
+}
+
+/**
+ * make_lsave_dirty - make lsave node dirty.
+ * @c: UBIFS file-system description object
+ * @lnum: LEB number where lsave was written
+ * @offs: offset where lsave was written
+ *
+ * This function is used by LPT garbage collection.  LPT garbage collection is
+ * used only for the "big" LPT model (c->big_lpt == 1).  Garbage collection
+ * simply involves marking all the nodes in the LEB being garbage-collected as
+ * dirty.  The dirty nodes are written next commit, after which the LEB is free
+ * to be reused.
+ *
+ * This function returns %0 on success and a negative error code on failure.
+ */
+static int make_lsave_dirty(struct ubifs_info *c, int lnum, int offs)
+{
+       if (lnum != c->lsave_lnum || offs != c->lsave_offs)
+               return 0; /* This lsave node is obsolete */
+       if (!(c->lpt_drty_flgs & LSAVE_DIRTY)) {
+               c->lpt_drty_flgs |= LSAVE_DIRTY;
+               ubifs_add_lpt_dirt(c, c->lsave_lnum, c->lsave_sz);
+       }
+       return 0;
+}
+
+/**
+ * make_node_dirty - make node dirty.
+ * @c: UBIFS file-system description object
+ * @node_type: LPT node type
+ * @node_num: node number
+ * @lnum: LEB number where node was written
+ * @offs: offset where node was written
+ *
+ * This function is used by LPT garbage collection.  LPT garbage collection is
+ * used only for the "big" LPT model (c->big_lpt == 1).  Garbage collection
+ * simply involves marking all the nodes in the LEB being garbage-collected as
+ * dirty.  The dirty nodes are written next commit, after which the LEB is free
+ * to be reused.
  *
- * You should have received a copy of the GNU General Public License along with
- * this program; if not, write to the Free Software Foundation, Inc., 51
- * Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
+ * This function returns %0 on success and a negative error code on failure.
+ */
+static int make_node_dirty(struct ubifs_info *c, int node_type, int node_num,
+                          int lnum, int offs)
+{
+       switch (node_type) {
+       case UBIFS_LPT_NNODE:
+               return make_nnode_dirty(c, node_num, lnum, offs);
+       case UBIFS_LPT_PNODE:
+               return make_pnode_dirty(c, node_num, lnum, offs);
+       case UBIFS_LPT_LTAB:
+               return make_ltab_dirty(c, lnum, offs);
+       case UBIFS_LPT_LSAVE:
+               return make_lsave_dirty(c, lnum, offs);
+       }
+       return -EINVAL;
+}
+
+/**
+ * get_lpt_node_len - return the length of a node based on its type.
+ * @c: UBIFS file-system description object
+ * @node_type: LPT node type
+ */
+static int get_lpt_node_len(const struct ubifs_info *c, int node_type)
+{
+       switch (node_type) {
+       case UBIFS_LPT_NNODE:
+               return c->nnode_sz;
+       case UBIFS_LPT_PNODE:
+               return c->pnode_sz;
+       case UBIFS_LPT_LTAB:
+               return c->ltab_sz;
+       case UBIFS_LPT_LSAVE:
+               return c->lsave_sz;
+       }
+       return 0;
+}
+
+/**
+ * get_pad_len - return the length of padding in a buffer.
+ * @c: UBIFS file-system description object
+ * @buf: buffer
+ * @len: length of buffer
+ */
+static int get_pad_len(const struct ubifs_info *c, uint8_t *buf, int len)
+{
+       int offs, pad_len;
+
+       if (c->min_io_size == 1)
+               return 0;
+       offs = c->leb_size - len;
+       pad_len = ALIGN(offs, c->min_io_size) - offs;
+       return pad_len;
+}
+
+/**
+ * get_lpt_node_type - return type (and node number) of a node in a buffer.
+ * @c: UBIFS file-system description object
+ * @buf: buffer
+ * @node_num: node number is returned here
+ */
+static int get_lpt_node_type(const struct ubifs_info *c, uint8_t *buf,
+                            int *node_num)
+{
+       uint8_t *addr = buf + UBIFS_LPT_CRC_BYTES;
+       int pos = 0, node_type;
+
+       node_type = ubifs_unpack_bits(&addr, &pos, UBIFS_LPT_TYPE_BITS);
+       *node_num = ubifs_unpack_bits(&addr, &pos, c->pcnt_bits);
+       return node_type;
+}
+
+/**
+ * is_a_node - determine if a buffer contains a node.
+ * @c: UBIFS file-system description object
+ * @buf: buffer
+ * @len: length of buffer
  *
- * Authors: Adrian Hunter
- *          Artem Bityutskiy (Битюцкий Артём)
+ * This function returns %1 if the buffer contains a node or %0 if it does not.
  */
+static int is_a_node(const struct ubifs_info *c, uint8_t *buf, int len)
+{
+       uint8_t *addr = buf + UBIFS_LPT_CRC_BYTES;
+       int pos = 0, node_type, node_len;
+       uint16_t crc, calc_crc;
 
-/*
- * This file implements commit-related functionality of the LEB properties
- * subsystem.
+       if (len < UBIFS_LPT_CRC_BYTES + (UBIFS_LPT_TYPE_BITS + 7) / 8)
+               return 0;
+       node_type = ubifs_unpack_bits(&addr, &pos, UBIFS_LPT_TYPE_BITS);
+       if (node_type == UBIFS_LPT_NOT_A_NODE)
+               return 0;
+       node_len = get_lpt_node_len(c, node_type);
+       if (!node_len || node_len > len)
+               return 0;
+       pos = 0;
+       addr = buf;
+       crc = ubifs_unpack_bits(&addr, &pos, UBIFS_LPT_CRC_BITS);
+       calc_crc = crc16(-1, buf + UBIFS_LPT_CRC_BYTES,
+                        node_len - UBIFS_LPT_CRC_BYTES);
+       if (crc != calc_crc)
+               return 0;
+       return 1;
+}
+
+/**
+ * lpt_gc_lnum - garbage collect a LPT LEB.
+ * @c: UBIFS file-system description object
+ * @lnum: LEB number to garbage collect
+ *
+ * LPT garbage collection is used only for the "big" LPT model
+ * (c->big_lpt == 1).  Garbage collection simply involves marking all the nodes
+ * in the LEB being garbage-collected as dirty.  The dirty nodes are written
+ * next commit, after which the LEB is free to be reused.
+ *
+ * This function returns %0 on success and a negative error code on failure.
  */
+static int lpt_gc_lnum(struct ubifs_info *c, int lnum)
+{
+       int err, len = c->leb_size, node_type, node_num, node_len, offs;
+       void *buf = c->lpt_buf;
 
-#include "crc16.h"
-#include "ubifs.h"
+       dbg_lp("LEB %d", lnum);
+
+       err = ubifs_leb_read(c, lnum, buf, 0, c->leb_size, 1);
+       if (err)
+               return err;
+
+       while (1) {
+               if (!is_a_node(c, buf, len)) {
+                       int pad_len;
+
+                       pad_len = get_pad_len(c, buf, len);
+                       if (pad_len) {
+                               buf += pad_len;
+                               len -= pad_len;
+                               continue;
+                       }
+                       return 0;
+               }
+               node_type = get_lpt_node_type(c, buf, &node_num);
+               node_len = get_lpt_node_len(c, node_type);
+               offs = c->leb_size - len;
+               ubifs_assert(node_len != 0);
+               mutex_lock(&c->lp_mutex);
+               err = make_node_dirty(c, node_type, node_num, lnum, offs);
+               mutex_unlock(&c->lp_mutex);
+               if (err)
+                       return err;
+               buf += node_len;
+               len -= node_len;
+       }
+       return 0;
+}
+
+/**
+ * lpt_gc - LPT garbage collection.
+ * @c: UBIFS file-system description object
+ *
+ * Select a LPT LEB for LPT garbage collection and call 'lpt_gc_lnum()'.
+ * Returns %0 on success and a negative error code on failure.
+ */
+static int lpt_gc(struct ubifs_info *c)
+{
+       int i, lnum = -1, dirty = 0;
+
+       mutex_lock(&c->lp_mutex);
+       for (i = 0; i < c->lpt_lebs; i++) {
+               ubifs_assert(!c->ltab[i].tgc);
+               if (i + c->lpt_first == c->nhead_lnum ||
+                   c->ltab[i].free + c->ltab[i].dirty == c->leb_size)
+                       continue;
+               if (c->ltab[i].dirty > dirty) {
+                       dirty = c->ltab[i].dirty;
+                       lnum = i + c->lpt_first;
+               }
+       }
+       mutex_unlock(&c->lp_mutex);
+       if (lnum == -1)
+               return -ENOSPC;
+       return lpt_gc_lnum(c, lnum);
+}
+
+/**
+ * ubifs_lpt_start_commit - UBIFS commit starts.
+ * @c: the UBIFS file-system description object
+ *
+ * This function has to be called when UBIFS starts the commit operation.
+ * This function "freezes" all currently dirty LEB properties and does not
+ * change them anymore. Further changes are saved and tracked separately
+ * because they are not part of this commit. This function returns zero in case
+ * of success and a negative error code in case of failure.
+ */
+int ubifs_lpt_start_commit(struct ubifs_info *c)
+{
+       int err, cnt;
+
+       dbg_lp("");
+
+       mutex_lock(&c->lp_mutex);
+       err = dbg_chk_lpt_free_spc(c);
+       if (err)
+               goto out;
+       err = dbg_check_ltab(c);
+       if (err)
+               goto out;
+
+       if (c->check_lpt_free) {
+               /*
+                * We ensure there is enough free space in
+                * ubifs_lpt_post_commit() by marking nodes dirty. That
+                * information is lost when we unmount, so we also need
+                * to check free space once after mounting also.
+                */
+               c->check_lpt_free = 0;
+               while (need_write_all(c)) {
+                       mutex_unlock(&c->lp_mutex);
+                       err = lpt_gc(c);
+                       if (err)
+                               return err;
+                       mutex_lock(&c->lp_mutex);
+               }
+       }
+
+       lpt_tgc_start(c);
+
+       if (!c->dirty_pn_cnt) {
+               dbg_cmt("no cnodes to commit");
+               err = 0;
+               goto out;
+       }
+
+       if (!c->big_lpt && need_write_all(c)) {
+               /* If needed, write everything */
+               err = make_tree_dirty(c);
+               if (err)
+                       goto out;
+               lpt_tgc_start(c);
+       }
+
+       if (c->big_lpt)
+               populate_lsave(c);
+
+       cnt = get_cnodes_to_commit(c);
+       ubifs_assert(cnt != 0);
+
+       err = layout_cnodes(c);
+       if (err)
+               goto out;
+
+       /* Copy the LPT's own lprops for end commit to write */
+       memcpy(c->ltab_cmt, c->ltab,
+              sizeof(struct ubifs_lpt_lprops) * c->lpt_lebs);
+       c->lpt_drty_flgs &= ~(LTAB_DIRTY | LSAVE_DIRTY);
+
+out:
+       mutex_unlock(&c->lp_mutex);
+       return err;
+}
 
 /**
  * free_obsolete_cnodes - free obsolete cnodes for commit end.
@@ -50,6 +1314,65 @@ static void free_obsolete_cnodes(struct ubifs_info *c)
        c->lpt_cnext = NULL;
 }
 
+#ifndef __UBOOT__
+/**
+ * ubifs_lpt_end_commit - finish the commit operation.
+ * @c: the UBIFS file-system description object
+ *
+ * This function has to be called when the commit operation finishes. It
+ * flushes the changes which were "frozen" by 'ubifs_lprops_start_commit()' to
+ * the media. Returns zero in case of success and a negative error code in case
+ * of failure.
+ */
+int ubifs_lpt_end_commit(struct ubifs_info *c)
+{
+       int err;
+
+       dbg_lp("");
+
+       if (!c->lpt_cnext)
+               return 0;
+
+       err = write_cnodes(c);
+       if (err)
+               return err;
+
+       mutex_lock(&c->lp_mutex);
+       free_obsolete_cnodes(c);
+       mutex_unlock(&c->lp_mutex);
+
+       return 0;
+}
+#endif
+
+/**
+ * ubifs_lpt_post_commit - post commit LPT trivial GC and LPT GC.
+ * @c: UBIFS file-system description object
+ *
+ * LPT trivial GC is completed after a commit. Also LPT GC is done after a
+ * commit for the "big" LPT model.
+ */
+int ubifs_lpt_post_commit(struct ubifs_info *c)
+{
+       int err;
+
+       mutex_lock(&c->lp_mutex);
+       err = lpt_tgc_end(c);
+       if (err)
+               goto out;
+       if (c->big_lpt)
+               while (need_write_all(c)) {
+                       mutex_unlock(&c->lp_mutex);
+                       err = lpt_gc(c);
+                       if (err)
+                               return err;
+                       mutex_lock(&c->lp_mutex);
+               }
+out:
+       mutex_unlock(&c->lp_mutex);
+       return err;
+}
+
 /**
  * first_nnode - find the first nnode in memory.
  * @c: UBIFS file-system description object
@@ -169,3 +1492,549 @@ void ubifs_lpt_free(struct ubifs_info *c, int wr_only)
        vfree(c->ltab);
        kfree(c->lpt_nod_buf);
 }
+
+#ifndef __UBOOT__
+/*
+ * Everything below is related to debugging.
+ */
+
+/**
+ * dbg_is_all_ff - determine if a buffer contains only 0xFF bytes.
+ * @buf: buffer
+ * @len: buffer length
+ */
+static int dbg_is_all_ff(uint8_t *buf, int len)
+{
+       int i;
+
+       for (i = 0; i < len; i++)
+               if (buf[i] != 0xff)
+                       return 0;
+       return 1;
+}
+
+/**
+ * dbg_is_nnode_dirty - determine if a nnode is dirty.
+ * @c: the UBIFS file-system description object
+ * @lnum: LEB number where nnode was written
+ * @offs: offset where nnode was written
+ */
+static int dbg_is_nnode_dirty(struct ubifs_info *c, int lnum, int offs)
+{
+       struct ubifs_nnode *nnode;
+       int hght;
+
+       /* Entire tree is in memory so first_nnode / next_nnode are OK */
+       nnode = first_nnode(c, &hght);
+       for (; nnode; nnode = next_nnode(c, nnode, &hght)) {
+               struct ubifs_nbranch *branch;
+
+               cond_resched();
+               if (nnode->parent) {
+                       branch = &nnode->parent->nbranch[nnode->iip];
+                       if (branch->lnum != lnum || branch->offs != offs)
+                               continue;
+                       if (test_bit(DIRTY_CNODE, &nnode->flags))
+                               return 1;
+                       return 0;
+               } else {
+                       if (c->lpt_lnum != lnum || c->lpt_offs != offs)
+                               continue;
+                       if (test_bit(DIRTY_CNODE, &nnode->flags))
+                               return 1;
+                       return 0;
+               }
+       }
+       return 1;
+}
+
+/**
+ * dbg_is_pnode_dirty - determine if a pnode is dirty.
+ * @c: the UBIFS file-system description object
+ * @lnum: LEB number where pnode was written
+ * @offs: offset where pnode was written
+ */
+static int dbg_is_pnode_dirty(struct ubifs_info *c, int lnum, int offs)
+{
+       int i, cnt;
+
+       cnt = DIV_ROUND_UP(c->main_lebs, UBIFS_LPT_FANOUT);
+       for (i = 0; i < cnt; i++) {
+               struct ubifs_pnode *pnode;
+               struct ubifs_nbranch *branch;
+
+               cond_resched();
+               pnode = pnode_lookup(c, i);
+               if (IS_ERR(pnode))
+                       return PTR_ERR(pnode);
+               branch = &pnode->parent->nbranch[pnode->iip];
+               if (branch->lnum != lnum || branch->offs != offs)
+                       continue;
+               if (test_bit(DIRTY_CNODE, &pnode->flags))
+                       return 1;
+               return 0;
+       }
+       return 1;
+}
+
+/**
+ * dbg_is_ltab_dirty - determine if a ltab node is dirty.
+ * @c: the UBIFS file-system description object
+ * @lnum: LEB number where ltab node was written
+ * @offs: offset where ltab node was written
+ */
+static int dbg_is_ltab_dirty(struct ubifs_info *c, int lnum, int offs)
+{
+       if (lnum != c->ltab_lnum || offs != c->ltab_offs)
+               return 1;
+       return (c->lpt_drty_flgs & LTAB_DIRTY) != 0;
+}
+
+/**
+ * dbg_is_lsave_dirty - determine if a lsave node is dirty.
+ * @c: the UBIFS file-system description object
+ * @lnum: LEB number where lsave node was written
+ * @offs: offset where lsave node was written
+ */
+static int dbg_is_lsave_dirty(struct ubifs_info *c, int lnum, int offs)
+{
+       if (lnum != c->lsave_lnum || offs != c->lsave_offs)
+               return 1;
+       return (c->lpt_drty_flgs & LSAVE_DIRTY) != 0;
+}
+
+/**
+ * dbg_is_node_dirty - determine if a node is dirty.
+ * @c: the UBIFS file-system description object
+ * @node_type: node type
+ * @lnum: LEB number where node was written
+ * @offs: offset where node was written
+ */
+static int dbg_is_node_dirty(struct ubifs_info *c, int node_type, int lnum,
+                            int offs)
+{
+       switch (node_type) {
+       case UBIFS_LPT_NNODE:
+               return dbg_is_nnode_dirty(c, lnum, offs);
+       case UBIFS_LPT_PNODE:
+               return dbg_is_pnode_dirty(c, lnum, offs);
+       case UBIFS_LPT_LTAB:
+               return dbg_is_ltab_dirty(c, lnum, offs);
+       case UBIFS_LPT_LSAVE:
+               return dbg_is_lsave_dirty(c, lnum, offs);
+       }
+       return 1;
+}
+
+/**
+ * dbg_check_ltab_lnum - check the ltab for a LPT LEB number.
+ * @c: the UBIFS file-system description object
+ * @lnum: LEB number where node was written
+ * @offs: offset where node was written
+ *
+ * This function returns %0 on success and a negative error code on failure.
+ */
+static int dbg_check_ltab_lnum(struct ubifs_info *c, int lnum)
+{
+       int err, len = c->leb_size, dirty = 0, node_type, node_num, node_len;
+       int ret;
+       void *buf, *p;
+
+       if (!dbg_is_chk_lprops(c))
+               return 0;
+
+       buf = p = __vmalloc(c->leb_size, GFP_NOFS, PAGE_KERNEL);
+       if (!buf) {
+               ubifs_err("cannot allocate memory for ltab checking");
+               return 0;
+       }
+
+       dbg_lp("LEB %d", lnum);
+
+       err = ubifs_leb_read(c, lnum, buf, 0, c->leb_size, 1);
+       if (err)
+               goto out;
+
+       while (1) {
+               if (!is_a_node(c, p, len)) {
+                       int i, pad_len;
+
+                       pad_len = get_pad_len(c, p, len);
+                       if (pad_len) {
+                               p += pad_len;
+                               len -= pad_len;
+                               dirty += pad_len;
+                               continue;
+                       }
+                       if (!dbg_is_all_ff(p, len)) {
+                               ubifs_err("invalid empty space in LEB %d at %d",
+                                         lnum, c->leb_size - len);
+                               err = -EINVAL;
+                       }
+                       i = lnum - c->lpt_first;
+                       if (len != c->ltab[i].free) {
+                               ubifs_err("invalid free space in LEB %d (free %d, expected %d)",
+                                         lnum, len, c->ltab[i].free);
+                               err = -EINVAL;
+                       }
+                       if (dirty != c->ltab[i].dirty) {
+                               ubifs_err("invalid dirty space in LEB %d (dirty %d, expected %d)",
+                                         lnum, dirty, c->ltab[i].dirty);
+                               err = -EINVAL;
+                       }
+                       goto out;
+               }
+               node_type = get_lpt_node_type(c, p, &node_num);
+               node_len = get_lpt_node_len(c, node_type);
+               ret = dbg_is_node_dirty(c, node_type, lnum, c->leb_size - len);
+               if (ret == 1)
+                       dirty += node_len;
+               p += node_len;
+               len -= node_len;
+       }
+
+       err = 0;
+out:
+       vfree(buf);
+       return err;
+}
+
+/**
+ * dbg_check_ltab - check the free and dirty space in the ltab.
+ * @c: the UBIFS file-system description object
+ *
+ * This function returns %0 on success and a negative error code on failure.
+ */
+int dbg_check_ltab(struct ubifs_info *c)
+{
+       int lnum, err, i, cnt;
+
+       if (!dbg_is_chk_lprops(c))
+               return 0;
+
+       /* Bring the entire tree into memory */
+       cnt = DIV_ROUND_UP(c->main_lebs, UBIFS_LPT_FANOUT);
+       for (i = 0; i < cnt; i++) {
+               struct ubifs_pnode *pnode;
+
+               pnode = pnode_lookup(c, i);
+               if (IS_ERR(pnode))
+                       return PTR_ERR(pnode);
+               cond_resched();
+       }
+
+       /* Check nodes */
+       err = dbg_check_lpt_nodes(c, (struct ubifs_cnode *)c->nroot, 0, 0);
+       if (err)
+               return err;
+
+       /* Check each LEB */
+       for (lnum = c->lpt_first; lnum <= c->lpt_last; lnum++) {
+               err = dbg_check_ltab_lnum(c, lnum);
+               if (err) {
+                       ubifs_err("failed at LEB %d", lnum);
+                       return err;
+               }
+       }
+
+       dbg_lp("succeeded");
+       return 0;
+}
+
+/**
+ * dbg_chk_lpt_free_spc - check LPT free space is enough to write entire LPT.
+ * @c: the UBIFS file-system description object
+ *
+ * This function returns %0 on success and a negative error code on failure.
+ */
+int dbg_chk_lpt_free_spc(struct ubifs_info *c)
+{
+       long long free = 0;
+       int i;
+
+       if (!dbg_is_chk_lprops(c))
+               return 0;
+
+       for (i = 0; i < c->lpt_lebs; i++) {
+               if (c->ltab[i].tgc || c->ltab[i].cmt)
+                       continue;
+               if (i + c->lpt_first == c->nhead_lnum)
+                       free += c->leb_size - c->nhead_offs;
+               else if (c->ltab[i].free == c->leb_size)
+                       free += c->leb_size;
+       }
+       if (free < c->lpt_sz) {
+               ubifs_err("LPT space error: free %lld lpt_sz %lld",
+                         free, c->lpt_sz);
+               ubifs_dump_lpt_info(c);
+               ubifs_dump_lpt_lebs(c);
+               dump_stack();
+               return -EINVAL;
+       }
+       return 0;
+}
+
+/**
+ * dbg_chk_lpt_sz - check LPT does not write more than LPT size.
+ * @c: the UBIFS file-system description object
+ * @action: what to do
+ * @len: length written
+ *
+ * This function returns %0 on success and a negative error code on failure.
+ * The @action argument may be one of:
+ *   o %0 - LPT debugging checking starts, initialize debugging variables;
+ *   o %1 - wrote an LPT node, increase LPT size by @len bytes;
+ *   o %2 - switched to a different LEB and wasted @len bytes;
+ *   o %3 - check that we've written the right number of bytes.
+ *   o %4 - wasted @len bytes;
+ */
+int dbg_chk_lpt_sz(struct ubifs_info *c, int action, int len)
+{
+       struct ubifs_debug_info *d = c->dbg;
+       long long chk_lpt_sz, lpt_sz;
+       int err = 0;
+
+       if (!dbg_is_chk_lprops(c))
+               return 0;
+
+       switch (action) {
+       case 0:
+               d->chk_lpt_sz = 0;
+               d->chk_lpt_sz2 = 0;
+               d->chk_lpt_lebs = 0;
+               d->chk_lpt_wastage = 0;
+               if (c->dirty_pn_cnt > c->pnode_cnt) {
+                       ubifs_err("dirty pnodes %d exceed max %d",
+                                 c->dirty_pn_cnt, c->pnode_cnt);
+                       err = -EINVAL;
+               }
+               if (c->dirty_nn_cnt > c->nnode_cnt) {
+                       ubifs_err("dirty nnodes %d exceed max %d",
+                                 c->dirty_nn_cnt, c->nnode_cnt);
+                       err = -EINVAL;
+               }
+               return err;
+       case 1:
+               d->chk_lpt_sz += len;
+               return 0;
+       case 2:
+               d->chk_lpt_sz += len;
+               d->chk_lpt_wastage += len;
+               d->chk_lpt_lebs += 1;
+               return 0;
+       case 3:
+               chk_lpt_sz = c->leb_size;
+               chk_lpt_sz *= d->chk_lpt_lebs;
+               chk_lpt_sz += len - c->nhead_offs;
+               if (d->chk_lpt_sz != chk_lpt_sz) {
+                       ubifs_err("LPT wrote %lld but space used was %lld",
+                                 d->chk_lpt_sz, chk_lpt_sz);
+                       err = -EINVAL;
+               }
+               if (d->chk_lpt_sz > c->lpt_sz) {
+                       ubifs_err("LPT wrote %lld but lpt_sz is %lld",
+                                 d->chk_lpt_sz, c->lpt_sz);
+                       err = -EINVAL;
+               }
+               if (d->chk_lpt_sz2 && d->chk_lpt_sz != d->chk_lpt_sz2) {
+                       ubifs_err("LPT layout size %lld but wrote %lld",
+                                 d->chk_lpt_sz, d->chk_lpt_sz2);
+                       err = -EINVAL;
+               }
+               if (d->chk_lpt_sz2 && d->new_nhead_offs != len) {
+                       ubifs_err("LPT new nhead offs: expected %d was %d",
+                                 d->new_nhead_offs, len);
+                       err = -EINVAL;
+               }
+               lpt_sz = (long long)c->pnode_cnt * c->pnode_sz;
+               lpt_sz += (long long)c->nnode_cnt * c->nnode_sz;
+               lpt_sz += c->ltab_sz;
+               if (c->big_lpt)
+                       lpt_sz += c->lsave_sz;
+               if (d->chk_lpt_sz - d->chk_lpt_wastage > lpt_sz) {
+                       ubifs_err("LPT chk_lpt_sz %lld + waste %lld exceeds %lld",
+                                 d->chk_lpt_sz, d->chk_lpt_wastage, lpt_sz);
+                       err = -EINVAL;
+               }
+               if (err) {
+                       ubifs_dump_lpt_info(c);
+                       ubifs_dump_lpt_lebs(c);
+                       dump_stack();
+               }
+               d->chk_lpt_sz2 = d->chk_lpt_sz;
+               d->chk_lpt_sz = 0;
+               d->chk_lpt_wastage = 0;
+               d->chk_lpt_lebs = 0;
+               d->new_nhead_offs = len;
+               return err;
+       case 4:
+               d->chk_lpt_sz += len;
+               d->chk_lpt_wastage += len;
+               return 0;
+       default:
+               return -EINVAL;
+       }
+}
+
+/**
+ * ubifs_dump_lpt_leb - dump an LPT LEB.
+ * @c: UBIFS file-system description object
+ * @lnum: LEB number to dump
+ *
+ * This function dumps an LEB from LPT area. Nodes in this area are very
+ * different to nodes in the main area (e.g., they do not have common headers,
+ * they do not have 8-byte alignments, etc), so we have a separate function to
+ * dump LPT area LEBs. Note, LPT has to be locked by the caller.
+ */
+static void dump_lpt_leb(const struct ubifs_info *c, int lnum)
+{
+       int err, len = c->leb_size, node_type, node_num, node_len, offs;
+       void *buf, *p;
+
+       pr_err("(pid %d) start dumping LEB %d\n", current->pid, lnum);
+       buf = p = __vmalloc(c->leb_size, GFP_NOFS, PAGE_KERNEL);
+       if (!buf) {
+               ubifs_err("cannot allocate memory to dump LPT");
+               return;
+       }
+
+       err = ubifs_leb_read(c, lnum, buf, 0, c->leb_size, 1);
+       if (err)
+               goto out;
+
+       while (1) {
+               offs = c->leb_size - len;
+               if (!is_a_node(c, p, len)) {
+                       int pad_len;
+
+                       pad_len = get_pad_len(c, p, len);
+                       if (pad_len) {
+                               pr_err("LEB %d:%d, pad %d bytes\n",
+                                      lnum, offs, pad_len);
+                               p += pad_len;
+                               len -= pad_len;
+                               continue;
+                       }
+                       if (len)
+                               pr_err("LEB %d:%d, free %d bytes\n",
+                                      lnum, offs, len);
+                       break;
+               }
+
+               node_type = get_lpt_node_type(c, p, &node_num);
+               switch (node_type) {
+               case UBIFS_LPT_PNODE:
+               {
+                       node_len = c->pnode_sz;
+                       if (c->big_lpt)
+                               pr_err("LEB %d:%d, pnode num %d\n",
+                                      lnum, offs, node_num);
+                       else
+                               pr_err("LEB %d:%d, pnode\n", lnum, offs);
+                       break;
+               }
+               case UBIFS_LPT_NNODE:
+               {
+                       int i;
+                       struct ubifs_nnode nnode;
+
+                       node_len = c->nnode_sz;
+                       if (c->big_lpt)
+                               pr_err("LEB %d:%d, nnode num %d, ",
+                                      lnum, offs, node_num);
+                       else
+                               pr_err("LEB %d:%d, nnode, ",
+                                      lnum, offs);
+                       err = ubifs_unpack_nnode(c, p, &nnode);
+                       for (i = 0; i < UBIFS_LPT_FANOUT; i++) {
+                               pr_cont("%d:%d", nnode.nbranch[i].lnum,
+                                      nnode.nbranch[i].offs);
+                               if (i != UBIFS_LPT_FANOUT - 1)
+                                       pr_cont(", ");
+                       }
+                       pr_cont("\n");
+                       break;
+               }
+               case UBIFS_LPT_LTAB:
+                       node_len = c->ltab_sz;
+                       pr_err("LEB %d:%d, ltab\n", lnum, offs);
+                       break;
+               case UBIFS_LPT_LSAVE:
+                       node_len = c->lsave_sz;
+                       pr_err("LEB %d:%d, lsave len\n", lnum, offs);
+                       break;
+               default:
+                       ubifs_err("LPT node type %d not recognized", node_type);
+                       goto out;
+               }
+
+               p += node_len;
+               len -= node_len;
+       }
+
+       pr_err("(pid %d) finish dumping LEB %d\n", current->pid, lnum);
+out:
+       vfree(buf);
+       return;
+}
+
+/**
+ * ubifs_dump_lpt_lebs - dump LPT lebs.
+ * @c: UBIFS file-system description object
+ *
+ * This function dumps all LPT LEBs. The caller has to make sure the LPT is
+ * locked.
+ */
+void ubifs_dump_lpt_lebs(const struct ubifs_info *c)
+{
+       int i;
+
+       pr_err("(pid %d) start dumping all LPT LEBs\n", current->pid);
+       for (i = 0; i < c->lpt_lebs; i++)
+               dump_lpt_leb(c, i + c->lpt_first);
+       pr_err("(pid %d) finish dumping all LPT LEBs\n", current->pid);
+}
+
+/**
+ * dbg_populate_lsave - debugging version of 'populate_lsave()'
+ * @c: UBIFS file-system description object
+ *
+ * This is a debugging version for 'populate_lsave()' which populates lsave
+ * with random LEBs instead of useful LEBs, which is good for test coverage.
+ * Returns zero if lsave has not been populated (this debugging feature is
+ * disabled) an non-zero if lsave has been populated.
+ */
+static int dbg_populate_lsave(struct ubifs_info *c)
+{
+       struct ubifs_lprops *lprops;
+       struct ubifs_lpt_heap *heap;
+       int i;
+
+       if (!dbg_is_chk_gen(c))
+               return 0;
+       if (prandom_u32() & 3)
+               return 0;
+
+       for (i = 0; i < c->lsave_cnt; i++)
+               c->lsave[i] = c->main_first;
+
+       list_for_each_entry(lprops, &c->empty_list, list)
+               c->lsave[prandom_u32() % c->lsave_cnt] = lprops->lnum;
+       list_for_each_entry(lprops, &c->freeable_list, list)
+               c->lsave[prandom_u32() % c->lsave_cnt] = lprops->lnum;
+       list_for_each_entry(lprops, &c->frdi_idx_list, list)
+               c->lsave[prandom_u32() % c->lsave_cnt] = lprops->lnum;
+
+       heap = &c->lpt_heap[LPROPS_DIRTY_IDX - 1];
+       for (i = 0; i < heap->cnt; i++)
+               c->lsave[prandom_u32() % c->lsave_cnt] = heap->arr[i]->lnum;
+       heap = &c->lpt_heap[LPROPS_DIRTY - 1];
+       for (i = 0; i < heap->cnt; i++)
+               c->lsave[prandom_u32() % c->lsave_cnt] = heap->arr[i]->lnum;
+       heap = &c->lpt_heap[LPROPS_FREE - 1];
+       for (i = 0; i < heap->cnt; i++)
+               c->lsave[prandom_u32() % c->lsave_cnt] = heap->arr[i]->lnum;
+
+       return 1;
+}
+#endif
index 3f2926e870b00817cc524c3f428306b514ac19de..00ca855e818cad8f7c51017f7fa32e907972e885 100644 (file)
@@ -3,18 +3,7 @@
  *
  * Copyright (C) 2006-2008 Nokia Corporation.
  *
- * This program is free software; you can redistribute it and/or modify it
- * under the terms of the GNU General Public License version 2 as published by
- * the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
- * more details.
- *
- * You should have received a copy of the GNU General Public License along with
- * this program; if not, write to the Free Software Foundation, Inc., 51
- * Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
+ * SPDX-License-Identifier:    GPL-2.0+
  *
  * Authors: Artem Bityutskiy (Битюцкий Артём)
  *          Adrian Hunter
 
 /* This file implements reading and writing the master node */
 
+#define __UBOOT__
 #include "ubifs.h"
+#ifdef __UBOOT__
+#include <linux/compat.h>
+#include <linux/err.h>
+#include <ubi_uboot.h>
+#endif
 
 /**
  * scan_for_master - search the valid master node.
  * @c: UBIFS file-system description object
  *
  * This function scans the master node LEBs and search for the latest master
- * node. Returns zero in case of success and a negative error code in case of
+ * node. Returns zero in case of success, %-EUCLEAN if there master area is
+ * corrupted and requires recovery, and a negative error code in case of
  * failure.
  */
 static int scan_for_master(struct ubifs_info *c)
@@ -40,7 +36,7 @@ static int scan_for_master(struct ubifs_info *c)
 
        lnum = UBIFS_MST_LNUM;
 
-       sleb = ubifs_scan(c, lnum, 0, c->sbuf);
+       sleb = ubifs_scan(c, lnum, 0, c->sbuf, 1);
        if (IS_ERR(sleb))
                return PTR_ERR(sleb);
        nodes_cnt = sleb->nodes_cnt;
@@ -48,7 +44,7 @@ static int scan_for_master(struct ubifs_info *c)
                snod = list_entry(sleb->nodes.prev, struct ubifs_scan_node,
                                  list);
                if (snod->type != UBIFS_MST_NODE)
-                       goto out;
+                       goto out_dump;
                memcpy(c->mst_node, snod->node, snod->len);
                offs = snod->offs;
        }
@@ -56,7 +52,7 @@ static int scan_for_master(struct ubifs_info *c)
 
        lnum += 1;
 
-       sleb = ubifs_scan(c, lnum, 0, c->sbuf);
+       sleb = ubifs_scan(c, lnum, 0, c->sbuf, 1);
        if (IS_ERR(sleb))
                return PTR_ERR(sleb);
        if (sleb->nodes_cnt != nodes_cnt)
@@ -65,7 +61,7 @@ static int scan_for_master(struct ubifs_info *c)
                goto out;
        snod = list_entry(sleb->nodes.prev, struct ubifs_scan_node, list);
        if (snod->type != UBIFS_MST_NODE)
-               goto out;
+               goto out_dump;
        if (snod->offs != offs)
                goto out;
        if (memcmp((void *)c->mst_node + UBIFS_CH_SZ,
@@ -78,6 +74,12 @@ static int scan_for_master(struct ubifs_info *c)
 
 out:
        ubifs_scan_destroy(sleb);
+       return -EUCLEAN;
+
+out_dump:
+       ubifs_err("unexpected node type %d master LEB %d:%d",
+                 snod->type, lnum, snod->offs);
+       ubifs_scan_destroy(sleb);
        return -EINVAL;
 }
 
@@ -141,7 +143,7 @@ static int validate_master(const struct ubifs_info *c)
        }
 
        main_sz = (long long)c->main_lebs * c->leb_size;
-       if (c->old_idx_sz & 7 || c->old_idx_sz >= main_sz) {
+       if (c->bi.old_idx_sz & 7 || c->bi.old_idx_sz >= main_sz) {
                err = 9;
                goto out;
        }
@@ -211,7 +213,7 @@ static int validate_master(const struct ubifs_info *c)
        }
 
        if (c->lst.total_dead + c->lst.total_dark +
-           c->lst.total_used + c->old_idx_sz > main_sz) {
+           c->lst.total_used + c->bi.old_idx_sz > main_sz) {
                err = 21;
                goto out;
        }
@@ -234,7 +236,7 @@ static int validate_master(const struct ubifs_info *c)
 
 out:
        ubifs_err("bad master node at offset %d error %d", c->mst_offs, err);
-       dbg_dump_node(c, c->mst_node);
+       ubifs_dump_node(c, c->mst_node);
        return -EINVAL;
 }
 
@@ -256,7 +258,8 @@ int ubifs_read_master(struct ubifs_info *c)
 
        err = scan_for_master(c);
        if (err) {
-               err = ubifs_recover_master_node(c);
+               if (err == -EUCLEAN)
+                       err = ubifs_recover_master_node(c);
                if (err)
                        /*
                         * Note, we do not free 'c->mst_node' here because the
@@ -278,7 +281,7 @@ int ubifs_read_master(struct ubifs_info *c)
        c->gc_lnum         = le32_to_cpu(c->mst_node->gc_lnum);
        c->ihead_lnum      = le32_to_cpu(c->mst_node->ihead_lnum);
        c->ihead_offs      = le32_to_cpu(c->mst_node->ihead_offs);
-       c->old_idx_sz      = le64_to_cpu(c->mst_node->index_size);
+       c->bi.old_idx_sz   = le64_to_cpu(c->mst_node->index_size);
        c->lpt_lnum        = le32_to_cpu(c->mst_node->lpt_lnum);
        c->lpt_offs        = le32_to_cpu(c->mst_node->lpt_offs);
        c->nhead_lnum      = le32_to_cpu(c->mst_node->nhead_lnum);
@@ -297,7 +300,7 @@ int ubifs_read_master(struct ubifs_info *c)
        c->lst.total_dead  = le64_to_cpu(c->mst_node->total_dead);
        c->lst.total_dark  = le64_to_cpu(c->mst_node->total_dark);
 
-       c->calc_idx_sz = c->old_idx_sz;
+       c->calc_idx_sz = c->bi.old_idx_sz;
 
        if (c->mst_node->flags & cpu_to_le32(UBIFS_MST_NO_ORPHS))
                c->no_orphs = 1;
@@ -309,7 +312,7 @@ int ubifs_read_master(struct ubifs_info *c)
                if (c->leb_cnt < old_leb_cnt ||
                    c->leb_cnt < UBIFS_MIN_LEB_CNT) {
                        ubifs_err("bad leb_cnt on master node");
-                       dbg_dump_node(c, c->mst_node);
+                       ubifs_dump_node(c, c->mst_node);
                        return -EINVAL;
                }
 
@@ -335,7 +338,58 @@ int ubifs_read_master(struct ubifs_info *c)
        if (err)
                return err;
 
+#ifndef __UBOOT__
        err = dbg_old_index_check_init(c, &c->zroot);
+#endif
+
+       return err;
+}
+
+#ifndef __UBOOT__
+/**
+ * ubifs_write_master - write master node.
+ * @c: UBIFS file-system description object
+ *
+ * This function writes the master node. The caller has to take the
+ * @c->mst_mutex lock before calling this function. Returns zero in case of
+ * success and a negative error code in case of failure. The master node is
+ * written twice to enable recovery.
+ */
+int ubifs_write_master(struct ubifs_info *c)
+{
+       int err, lnum, offs, len;
+
+       ubifs_assert(!c->ro_media && !c->ro_mount);
+       if (c->ro_error)
+               return -EROFS;
+
+       lnum = UBIFS_MST_LNUM;
+       offs = c->mst_offs + c->mst_node_alsz;
+       len = UBIFS_MST_NODE_SZ;
+
+       if (offs + UBIFS_MST_NODE_SZ > c->leb_size) {
+               err = ubifs_leb_unmap(c, lnum);
+               if (err)
+                       return err;
+               offs = 0;
+       }
+
+       c->mst_offs = offs;
+       c->mst_node->highest_inum = cpu_to_le64(c->highest_inum);
+
+       err = ubifs_write_node(c, c->mst_node, len, lnum, offs);
+       if (err)
+               return err;
+
+       lnum += 1;
+
+       if (offs == 0) {
+               err = ubifs_leb_unmap(c, lnum);
+               if (err)
+                       return err;
+       }
+       err = ubifs_write_node(c, c->mst_node, len, lnum, offs);
 
        return err;
 }
+#endif
index 609232e9314a8b169cd6c82d0984a96051b014b2..4316d3c8ae2b1ecf8ceeef53c2c9200098001452 100644 (file)
@@ -3,18 +3,7 @@
  *
  * Copyright (C) 2006-2008 Nokia Corporation
  *
- * This program is free software; you can redistribute it and/or modify it
- * under the terms of the GNU General Public License version 2 as published by
- * the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
- * more details.
- *
- * You should have received a copy of the GNU General Public License along with
- * this program; if not, write to the Free Software Foundation, Inc., 51
- * Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
+ * SPDX-License-Identifier:    GPL-2.0+
  *
  * Authors: Artem Bityutskiy (Битюцкий Артём)
  *          Adrian Hunter
@@ -27,6 +16,7 @@
 #ifndef __UBIFS_MISC_H__
 #define __UBIFS_MISC_H__
 
+#define __UBOOT__
 /**
  * ubifs_zn_dirty - check if znode is dirty.
  * @znode: znode to check
@@ -38,6 +28,29 @@ static inline int ubifs_zn_dirty(const struct ubifs_znode *znode)
        return !!test_bit(DIRTY_ZNODE, &znode->flags);
 }
 
+/**
+ * ubifs_zn_obsolete - check if znode is obsolete.
+ * @znode: znode to check
+ *
+ * This helper function returns %1 if @znode is obsolete and %0 otherwise.
+ */
+static inline int ubifs_zn_obsolete(const struct ubifs_znode *znode)
+{
+       return !!test_bit(OBSOLETE_ZNODE, &znode->flags);
+}
+
+/**
+ * ubifs_zn_cow - check if znode has to be copied on write.
+ * @znode: znode to check
+ *
+ * This helper function returns %1 if @znode is has COW flag set and %0
+ * otherwise.
+ */
+static inline int ubifs_zn_cow(const struct ubifs_znode *znode)
+{
+       return !!test_bit(COW_ZNODE, &znode->flags);
+}
+
 /**
  * ubifs_wake_up_bgt - wake up background thread.
  * @c: UBIFS file-system description object
@@ -121,82 +134,27 @@ static inline int ubifs_wbuf_sync(struct ubifs_wbuf *wbuf)
        return err;
 }
 
+#ifndef __UBOOT__
 /**
- * ubifs_leb_unmap - unmap an LEB.
- * @c: UBIFS file-system description object
- * @lnum: LEB number to unmap
- *
- * This function returns %0 on success and a negative error code on failure.
- */
-static inline int ubifs_leb_unmap(const struct ubifs_info *c, int lnum)
-{
-       int err;
-
-       if (c->ro_media)
-               return -EROFS;
-       err = ubi_leb_unmap(c->ubi, lnum);
-       if (err) {
-               ubifs_err("unmap LEB %d failed, error %d", lnum, err);
-               return err;
-       }
-
-       return 0;
-}
-
-/**
- * ubifs_leb_write - write to a LEB.
- * @c: UBIFS file-system description object
- * @lnum: LEB number to write
- * @buf: buffer to write from
- * @offs: offset within LEB to write to
- * @len: length to write
- * @dtype: data type
- *
- * This function returns %0 on success and a negative error code on failure.
- */
-static inline int ubifs_leb_write(const struct ubifs_info *c, int lnum,
-                                 const void *buf, int offs, int len, int dtype)
-{
-       int err;
-
-       if (c->ro_media)
-               return -EROFS;
-       err = ubi_leb_write(c->ubi, lnum, buf, offs, len, dtype);
-       if (err) {
-               ubifs_err("writing %d bytes at %d:%d, error %d",
-                         len, lnum, offs, err);
-               return err;
-       }
-
-       return 0;
-}
-
-/**
- * ubifs_leb_change - atomic LEB change.
- * @c: UBIFS file-system description object
- * @lnum: LEB number to write
- * @buf: buffer to write from
- * @len: length to write
- * @dtype: data type
+ * ubifs_encode_dev - encode device node IDs.
+ * @dev: UBIFS device node information
+ * @rdev: device IDs to encode
  *
- * This function returns %0 on success and a negative error code on failure.
+ * This is a helper function which encodes major/minor numbers of a device node
+ * into UBIFS device node description. We use standard Linux "new" and "huge"
+ * encodings.
  */
-static inline int ubifs_leb_change(const struct ubifs_info *c, int lnum,
-                                  const void *buf, int len, int dtype)
+static inline int ubifs_encode_dev(union ubifs_dev_desc *dev, dev_t rdev)
 {
-       int err;
-
-       if (c->ro_media)
-               return -EROFS;
-       err = ubi_leb_change(c->ubi, lnum, buf, len, dtype);
-       if (err) {
-               ubifs_err("changing %d bytes in LEB %d, error %d",
-                         len, lnum, err);
-               return err;
+       if (new_valid_dev(rdev)) {
+               dev->new = cpu_to_le32(new_encode_dev(rdev));
+               return sizeof(dev->new);
+       } else {
+               dev->huge = cpu_to_le64(huge_encode_dev(rdev));
+               return sizeof(dev->huge);
        }
-
-       return 0;
 }
+#endif
 
 /**
  * ubifs_add_dirt - add dirty space to LEB properties.
@@ -260,8 +218,24 @@ struct ubifs_branch *ubifs_idx_branch(const struct ubifs_info *c,
 static inline void *ubifs_idx_key(const struct ubifs_info *c,
                                  const struct ubifs_idx_node *idx)
 {
-       const __u8 *branch = idx->branches;
-       return (void *)((struct ubifs_branch *)branch)->key;
+#ifndef __UBOOT__
+       return (void *)((struct ubifs_branch *)idx->branches)->key;
+#else
+       struct ubifs_branch *tmp;
+
+       tmp = (struct ubifs_branch *)idx->branches;
+       return (void *)tmp->key;
+#endif
+}
+
+/**
+ * ubifs_current_time - round current time to time granularity.
+ * @inode: inode
+ */
+static inline struct timespec ubifs_current_time(struct inode *inode)
+{
+       return (inode->i_sb->s_time_gran < NSEC_PER_SEC) ?
+               current_fs_time(inode->i_sb) : CURRENT_TIME_SEC;
 }
 
 /**
@@ -308,4 +282,21 @@ static inline void ubifs_release_lprops(struct ubifs_info *c)
        mutex_unlock(&c->lp_mutex);
 }
 
+/**
+ * ubifs_next_log_lnum - switch to the next log LEB.
+ * @c: UBIFS file-system description object
+ * @lnum: current log LEB
+ *
+ * This helper function returns the log LEB number which goes next after LEB
+ * 'lnum'.
+ */
+static inline int ubifs_next_log_lnum(const struct ubifs_info *c, int lnum)
+{
+       lnum += 1;
+       if (lnum > c->log_last)
+               lnum = UBIFS_LOG_LNUM;
+
+       return lnum;
+}
+
 #endif /* __UBIFS_MISC_H__ */
index d091031b8b75471733d7ccebf8ee0fcbba5e1a96..4e42879b5d6bc7513f45230447b163829b827951 100644 (file)
@@ -3,22 +3,12 @@
  *
  * Copyright (C) 2006-2008 Nokia Corporation.
  *
- * This program is free software; you can redistribute it and/or modify it
- * under the terms of the GNU General Public License version 2 as published by
- * the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
- * more details.
- *
- * You should have received a copy of the GNU General Public License along with
- * this program; if not, write to the Free Software Foundation, Inc., 51
- * Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
+ * SPDX-License-Identifier:    GPL-2.0+
  *
  * Author: Adrian Hunter
  */
 
+#include <linux/err.h>
 #include "ubifs.h"
 
 /*
  * than the maximum number of orphans allowed.
  */
 
+static int dbg_check_orphans(struct ubifs_info *c);
+
+/**
+ * ubifs_add_orphan - add an orphan.
+ * @c: UBIFS file-system description object
+ * @inum: orphan inode number
+ *
+ * Add an orphan. This function is called when an inodes link count drops to
+ * zero.
+ */
+int ubifs_add_orphan(struct ubifs_info *c, ino_t inum)
+{
+       struct ubifs_orphan *orphan, *o;
+       struct rb_node **p, *parent = NULL;
+
+       orphan = kzalloc(sizeof(struct ubifs_orphan), GFP_NOFS);
+       if (!orphan)
+               return -ENOMEM;
+       orphan->inum = inum;
+       orphan->new = 1;
+
+       spin_lock(&c->orphan_lock);
+       if (c->tot_orphans >= c->max_orphans) {
+               spin_unlock(&c->orphan_lock);
+               kfree(orphan);
+               return -ENFILE;
+       }
+       p = &c->orph_tree.rb_node;
+       while (*p) {
+               parent = *p;
+               o = rb_entry(parent, struct ubifs_orphan, rb);
+               if (inum < o->inum)
+                       p = &(*p)->rb_left;
+               else if (inum > o->inum)
+                       p = &(*p)->rb_right;
+               else {
+                       ubifs_err("orphaned twice");
+                       spin_unlock(&c->orphan_lock);
+                       kfree(orphan);
+                       return 0;
+               }
+       }
+       c->tot_orphans += 1;
+       c->new_orphans += 1;
+       rb_link_node(&orphan->rb, parent, p);
+       rb_insert_color(&orphan->rb, &c->orph_tree);
+       list_add_tail(&orphan->list, &c->orph_list);
+       list_add_tail(&orphan->new_list, &c->orph_new);
+       spin_unlock(&c->orphan_lock);
+       dbg_gen("ino %lu", (unsigned long)inum);
+       return 0;
+}
+
+/**
+ * ubifs_delete_orphan - delete an orphan.
+ * @c: UBIFS file-system description object
+ * @inum: orphan inode number
+ *
+ * Delete an orphan. This function is called when an inode is deleted.
+ */
+void ubifs_delete_orphan(struct ubifs_info *c, ino_t inum)
+{
+       struct ubifs_orphan *o;
+       struct rb_node *p;
+
+       spin_lock(&c->orphan_lock);
+       p = c->orph_tree.rb_node;
+       while (p) {
+               o = rb_entry(p, struct ubifs_orphan, rb);
+               if (inum < o->inum)
+                       p = p->rb_left;
+               else if (inum > o->inum)
+                       p = p->rb_right;
+               else {
+                       if (o->del) {
+                               spin_unlock(&c->orphan_lock);
+                               dbg_gen("deleted twice ino %lu",
+                                       (unsigned long)inum);
+                               return;
+                       }
+                       if (o->cmt) {
+                               o->del = 1;
+                               o->dnext = c->orph_dnext;
+                               c->orph_dnext = o;
+                               spin_unlock(&c->orphan_lock);
+                               dbg_gen("delete later ino %lu",
+                                       (unsigned long)inum);
+                               return;
+                       }
+                       rb_erase(p, &c->orph_tree);
+                       list_del(&o->list);
+                       c->tot_orphans -= 1;
+                       if (o->new) {
+                               list_del(&o->new_list);
+                               c->new_orphans -= 1;
+                       }
+                       spin_unlock(&c->orphan_lock);
+                       kfree(o);
+                       dbg_gen("inum %lu", (unsigned long)inum);
+                       return;
+               }
+       }
+       spin_unlock(&c->orphan_lock);
+       ubifs_err("missing orphan ino %lu", (unsigned long)inum);
+       dump_stack();
+}
+
+/**
+ * ubifs_orphan_start_commit - start commit of orphans.
+ * @c: UBIFS file-system description object
+ *
+ * Start commit of orphans.
+ */
+int ubifs_orphan_start_commit(struct ubifs_info *c)
+{
+       struct ubifs_orphan *orphan, **last;
+
+       spin_lock(&c->orphan_lock);
+       last = &c->orph_cnext;
+       list_for_each_entry(orphan, &c->orph_new, new_list) {
+               ubifs_assert(orphan->new);
+               ubifs_assert(!orphan->cmt);
+               orphan->new = 0;
+               orphan->cmt = 1;
+               *last = orphan;
+               last = &orphan->cnext;
+       }
+       *last = NULL;
+       c->cmt_orphans = c->new_orphans;
+       c->new_orphans = 0;
+       dbg_cmt("%d orphans to commit", c->cmt_orphans);
+       INIT_LIST_HEAD(&c->orph_new);
+       if (c->tot_orphans == 0)
+               c->no_orphs = 1;
+       else
+               c->no_orphs = 0;
+       spin_unlock(&c->orphan_lock);
+       return 0;
+}
+
+/**
+ * avail_orphs - calculate available space.
+ * @c: UBIFS file-system description object
+ *
+ * This function returns the number of orphans that can be written in the
+ * available space.
+ */
+static int avail_orphs(struct ubifs_info *c)
+{
+       int avail_lebs, avail, gap;
+
+       avail_lebs = c->orph_lebs - (c->ohead_lnum - c->orph_first) - 1;
+       avail = avail_lebs *
+              ((c->leb_size - UBIFS_ORPH_NODE_SZ) / sizeof(__le64));
+       gap = c->leb_size - c->ohead_offs;
+       if (gap >= UBIFS_ORPH_NODE_SZ + sizeof(__le64))
+               avail += (gap - UBIFS_ORPH_NODE_SZ) / sizeof(__le64);
+       return avail;
+}
+
 /**
  * tot_avail_orphs - calculate total space.
  * @c: UBIFS file-system description object
@@ -69,6 +219,256 @@ static int tot_avail_orphs(struct ubifs_info *c)
        return avail / 2;
 }
 
+/**
+ * do_write_orph_node - write a node to the orphan head.
+ * @c: UBIFS file-system description object
+ * @len: length of node
+ * @atomic: write atomically
+ *
+ * This function writes a node to the orphan head from the orphan buffer. If
+ * %atomic is not zero, then the write is done atomically. On success, %0 is
+ * returned, otherwise a negative error code is returned.
+ */
+static int do_write_orph_node(struct ubifs_info *c, int len, int atomic)
+{
+       int err = 0;
+
+       if (atomic) {
+               ubifs_assert(c->ohead_offs == 0);
+               ubifs_prepare_node(c, c->orph_buf, len, 1);
+               len = ALIGN(len, c->min_io_size);
+               err = ubifs_leb_change(c, c->ohead_lnum, c->orph_buf, len);
+       } else {
+               if (c->ohead_offs == 0) {
+                       /* Ensure LEB has been unmapped */
+                       err = ubifs_leb_unmap(c, c->ohead_lnum);
+                       if (err)
+                               return err;
+               }
+               err = ubifs_write_node(c, c->orph_buf, len, c->ohead_lnum,
+                                      c->ohead_offs);
+       }
+       return err;
+}
+
+/**
+ * write_orph_node - write an orphan node.
+ * @c: UBIFS file-system description object
+ * @atomic: write atomically
+ *
+ * This function builds an orphan node from the cnext list and writes it to the
+ * orphan head. On success, %0 is returned, otherwise a negative error code
+ * is returned.
+ */
+static int write_orph_node(struct ubifs_info *c, int atomic)
+{
+       struct ubifs_orphan *orphan, *cnext;
+       struct ubifs_orph_node *orph;
+       int gap, err, len, cnt, i;
+
+       ubifs_assert(c->cmt_orphans > 0);
+       gap = c->leb_size - c->ohead_offs;
+       if (gap < UBIFS_ORPH_NODE_SZ + sizeof(__le64)) {
+               c->ohead_lnum += 1;
+               c->ohead_offs = 0;
+               gap = c->leb_size;
+               if (c->ohead_lnum > c->orph_last) {
+                       /*
+                        * We limit the number of orphans so that this should
+                        * never happen.
+                        */
+                       ubifs_err("out of space in orphan area");
+                       return -EINVAL;
+               }
+       }
+       cnt = (gap - UBIFS_ORPH_NODE_SZ) / sizeof(__le64);
+       if (cnt > c->cmt_orphans)
+               cnt = c->cmt_orphans;
+       len = UBIFS_ORPH_NODE_SZ + cnt * sizeof(__le64);
+       ubifs_assert(c->orph_buf);
+       orph = c->orph_buf;
+       orph->ch.node_type = UBIFS_ORPH_NODE;
+       spin_lock(&c->orphan_lock);
+       cnext = c->orph_cnext;
+       for (i = 0; i < cnt; i++) {
+               orphan = cnext;
+               ubifs_assert(orphan->cmt);
+               orph->inos[i] = cpu_to_le64(orphan->inum);
+               orphan->cmt = 0;
+               cnext = orphan->cnext;
+               orphan->cnext = NULL;
+       }
+       c->orph_cnext = cnext;
+       c->cmt_orphans -= cnt;
+       spin_unlock(&c->orphan_lock);
+       if (c->cmt_orphans)
+               orph->cmt_no = cpu_to_le64(c->cmt_no);
+       else
+               /* Mark the last node of the commit */
+               orph->cmt_no = cpu_to_le64((c->cmt_no) | (1ULL << 63));
+       ubifs_assert(c->ohead_offs + len <= c->leb_size);
+       ubifs_assert(c->ohead_lnum >= c->orph_first);
+       ubifs_assert(c->ohead_lnum <= c->orph_last);
+       err = do_write_orph_node(c, len, atomic);
+       c->ohead_offs += ALIGN(len, c->min_io_size);
+       c->ohead_offs = ALIGN(c->ohead_offs, 8);
+       return err;
+}
+
+/**
+ * write_orph_nodes - write orphan nodes until there are no more to commit.
+ * @c: UBIFS file-system description object
+ * @atomic: write atomically
+ *
+ * This function writes orphan nodes for all the orphans to commit. On success,
+ * %0 is returned, otherwise a negative error code is returned.
+ */
+static int write_orph_nodes(struct ubifs_info *c, int atomic)
+{
+       int err;
+
+       while (c->cmt_orphans > 0) {
+               err = write_orph_node(c, atomic);
+               if (err)
+                       return err;
+       }
+       if (atomic) {
+               int lnum;
+
+               /* Unmap any unused LEBs after consolidation */
+               lnum = c->ohead_lnum + 1;
+               for (lnum = c->ohead_lnum + 1; lnum <= c->orph_last; lnum++) {
+                       err = ubifs_leb_unmap(c, lnum);
+                       if (err)
+                               return err;
+               }
+       }
+       return 0;
+}
+
+/**
+ * consolidate - consolidate the orphan area.
+ * @c: UBIFS file-system description object
+ *
+ * This function enables consolidation by putting all the orphans into the list
+ * to commit. The list is in the order that the orphans were added, and the
+ * LEBs are written atomically in order, so at no time can orphans be lost by
+ * an unclean unmount.
+ *
+ * This function returns %0 on success and a negative error code on failure.
+ */
+static int consolidate(struct ubifs_info *c)
+{
+       int tot_avail = tot_avail_orphs(c), err = 0;
+
+       spin_lock(&c->orphan_lock);
+       dbg_cmt("there is space for %d orphans and there are %d",
+               tot_avail, c->tot_orphans);
+       if (c->tot_orphans - c->new_orphans <= tot_avail) {
+               struct ubifs_orphan *orphan, **last;
+               int cnt = 0;
+
+               /* Change the cnext list to include all non-new orphans */
+               last = &c->orph_cnext;
+               list_for_each_entry(orphan, &c->orph_list, list) {
+                       if (orphan->new)
+                               continue;
+                       orphan->cmt = 1;
+                       *last = orphan;
+                       last = &orphan->cnext;
+                       cnt += 1;
+               }
+               *last = NULL;
+               ubifs_assert(cnt == c->tot_orphans - c->new_orphans);
+               c->cmt_orphans = cnt;
+               c->ohead_lnum = c->orph_first;
+               c->ohead_offs = 0;
+       } else {
+               /*
+                * We limit the number of orphans so that this should
+                * never happen.
+                */
+               ubifs_err("out of space in orphan area");
+               err = -EINVAL;
+       }
+       spin_unlock(&c->orphan_lock);
+       return err;
+}
+
+/**
+ * commit_orphans - commit orphans.
+ * @c: UBIFS file-system description object
+ *
+ * This function commits orphans to flash. On success, %0 is returned,
+ * otherwise a negative error code is returned.
+ */
+static int commit_orphans(struct ubifs_info *c)
+{
+       int avail, atomic = 0, err;
+
+       ubifs_assert(c->cmt_orphans > 0);
+       avail = avail_orphs(c);
+       if (avail < c->cmt_orphans) {
+               /* Not enough space to write new orphans, so consolidate */
+               err = consolidate(c);
+               if (err)
+                       return err;
+               atomic = 1;
+       }
+       err = write_orph_nodes(c, atomic);
+       return err;
+}
+
+/**
+ * erase_deleted - erase the orphans marked for deletion.
+ * @c: UBIFS file-system description object
+ *
+ * During commit, the orphans being committed cannot be deleted, so they are
+ * marked for deletion and deleted by this function. Also, the recovery
+ * adds killed orphans to the deletion list, and therefore they are deleted
+ * here too.
+ */
+static void erase_deleted(struct ubifs_info *c)
+{
+       struct ubifs_orphan *orphan, *dnext;
+
+       spin_lock(&c->orphan_lock);
+       dnext = c->orph_dnext;
+       while (dnext) {
+               orphan = dnext;
+               dnext = orphan->dnext;
+               ubifs_assert(!orphan->new);
+               ubifs_assert(orphan->del);
+               rb_erase(&orphan->rb, &c->orph_tree);
+               list_del(&orphan->list);
+               c->tot_orphans -= 1;
+               dbg_gen("deleting orphan ino %lu", (unsigned long)orphan->inum);
+               kfree(orphan);
+       }
+       c->orph_dnext = NULL;
+       spin_unlock(&c->orphan_lock);
+}
+
+/**
+ * ubifs_orphan_end_commit - end commit of orphans.
+ * @c: UBIFS file-system description object
+ *
+ * End commit of orphans.
+ */
+int ubifs_orphan_end_commit(struct ubifs_info *c)
+{
+       int err;
+
+       if (c->cmt_orphans != 0) {
+               err = commit_orphans(c);
+               if (err)
+                       return err;
+       }
+       erase_deleted(c);
+       err = dbg_check_orphans(c);
+       return err;
+}
+
 /**
  * ubifs_clear_orphans - erase all LEBs used for orphans.
  * @c: UBIFS file-system description object
@@ -128,6 +528,7 @@ static int insert_dead_orphan(struct ubifs_info *c, ino_t inum)
        rb_link_node(&orphan->rb, parent, p);
        rb_insert_color(&orphan->rb, &c->orph_tree);
        list_add_tail(&orphan->list, &c->orph_list);
+       orphan->del = 1;
        orphan->dnext = c->orph_dnext;
        c->orph_dnext = orphan;
        dbg_mnt("ino %lu, new %d, tot %d", (unsigned long)inum,
@@ -159,9 +560,9 @@ static int do_kill_orphans(struct ubifs_info *c, struct ubifs_scan_leb *sleb,
 
        list_for_each_entry(snod, &sleb->nodes, list) {
                if (snod->type != UBIFS_ORPH_NODE) {
-                       ubifs_err("invalid node type %d in orphan area at "
-                                 "%d:%d", snod->type, sleb->lnum, snod->offs);
-                       dbg_dump_node(c, snod->node);
+                       ubifs_err("invalid node type %d in orphan area at %d:%d",
+                                 snod->type, sleb->lnum, snod->offs);
+                       ubifs_dump_node(c, snod->node);
                        return -EINVAL;
                }
 
@@ -186,10 +587,9 @@ static int do_kill_orphans(struct ubifs_info *c, struct ubifs_scan_leb *sleb,
                         * number. That makes this orphan node, out of date.
                         */
                        if (!first) {
-                               ubifs_err("out of order commit number %llu in "
-                                         "orphan node at %d:%d",
+                               ubifs_err("out of order commit number %llu in orphan node at %d:%d",
                                          cmt_no, sleb->lnum, snod->offs);
-                               dbg_dump_node(c, snod->node);
+                               ubifs_dump_node(c, snod->node);
                                return -EINVAL;
                        }
                        dbg_rcvry("out of date LEB %d", sleb->lnum);
@@ -262,9 +662,11 @@ static int kill_orphans(struct ubifs_info *c)
                struct ubifs_scan_leb *sleb;
 
                dbg_rcvry("LEB %d", lnum);
-               sleb = ubifs_scan(c, lnum, 0, c->sbuf);
+               sleb = ubifs_scan(c, lnum, 0, c->sbuf, 1);
                if (IS_ERR(sleb)) {
-                       sleb = ubifs_recover_leb(c, lnum, 0, c->sbuf, 0);
+                       if (PTR_ERR(sleb) == -EUCLEAN)
+                               sleb = ubifs_recover_leb(c, lnum, 0,
+                                                        c->sbuf, -1);
                        if (IS_ERR(sleb)) {
                                err = PTR_ERR(sleb);
                                break;
@@ -314,3 +716,232 @@ int ubifs_mount_orphans(struct ubifs_info *c, int unclean, int read_only)
 
        return err;
 }
+
+/*
+ * Everything below is related to debugging.
+ */
+
+struct check_orphan {
+       struct rb_node rb;
+       ino_t inum;
+};
+
+struct check_info {
+       unsigned long last_ino;
+       unsigned long tot_inos;
+       unsigned long missing;
+       unsigned long long leaf_cnt;
+       struct ubifs_ino_node *node;
+       struct rb_root root;
+};
+
+static int dbg_find_orphan(struct ubifs_info *c, ino_t inum)
+{
+       struct ubifs_orphan *o;
+       struct rb_node *p;
+
+       spin_lock(&c->orphan_lock);
+       p = c->orph_tree.rb_node;
+       while (p) {
+               o = rb_entry(p, struct ubifs_orphan, rb);
+               if (inum < o->inum)
+                       p = p->rb_left;
+               else if (inum > o->inum)
+                       p = p->rb_right;
+               else {
+                       spin_unlock(&c->orphan_lock);
+                       return 1;
+               }
+       }
+       spin_unlock(&c->orphan_lock);
+       return 0;
+}
+
+static int dbg_ins_check_orphan(struct rb_root *root, ino_t inum)
+{
+       struct check_orphan *orphan, *o;
+       struct rb_node **p, *parent = NULL;
+
+       orphan = kzalloc(sizeof(struct check_orphan), GFP_NOFS);
+       if (!orphan)
+               return -ENOMEM;
+       orphan->inum = inum;
+
+       p = &root->rb_node;
+       while (*p) {
+               parent = *p;
+               o = rb_entry(parent, struct check_orphan, rb);
+               if (inum < o->inum)
+                       p = &(*p)->rb_left;
+               else if (inum > o->inum)
+                       p = &(*p)->rb_right;
+               else {
+                       kfree(orphan);
+                       return 0;
+               }
+       }
+       rb_link_node(&orphan->rb, parent, p);
+       rb_insert_color(&orphan->rb, root);
+       return 0;
+}
+
+static int dbg_find_check_orphan(struct rb_root *root, ino_t inum)
+{
+       struct check_orphan *o;
+       struct rb_node *p;
+
+       p = root->rb_node;
+       while (p) {
+               o = rb_entry(p, struct check_orphan, rb);
+               if (inum < o->inum)
+                       p = p->rb_left;
+               else if (inum > o->inum)
+                       p = p->rb_right;
+               else
+                       return 1;
+       }
+       return 0;
+}
+
+static void dbg_free_check_tree(struct rb_root *root)
+{
+       struct check_orphan *o, *n;
+
+       rbtree_postorder_for_each_entry_safe(o, n, root, rb)
+               kfree(o);
+}
+
+static int dbg_orphan_check(struct ubifs_info *c, struct ubifs_zbranch *zbr,
+                           void *priv)
+{
+       struct check_info *ci = priv;
+       ino_t inum;
+       int err;
+
+       inum = key_inum(c, &zbr->key);
+       if (inum != ci->last_ino) {
+               /* Lowest node type is the inode node, so it comes first */
+               if (key_type(c, &zbr->key) != UBIFS_INO_KEY)
+                       ubifs_err("found orphan node ino %lu, type %d",
+                                 (unsigned long)inum, key_type(c, &zbr->key));
+               ci->last_ino = inum;
+               ci->tot_inos += 1;
+               err = ubifs_tnc_read_node(c, zbr, ci->node);
+               if (err) {
+                       ubifs_err("node read failed, error %d", err);
+                       return err;
+               }
+               if (ci->node->nlink == 0)
+                       /* Must be recorded as an orphan */
+                       if (!dbg_find_check_orphan(&ci->root, inum) &&
+                           !dbg_find_orphan(c, inum)) {
+                               ubifs_err("missing orphan, ino %lu",
+                                         (unsigned long)inum);
+                               ci->missing += 1;
+                       }
+       }
+       ci->leaf_cnt += 1;
+       return 0;
+}
+
+static int dbg_read_orphans(struct check_info *ci, struct ubifs_scan_leb *sleb)
+{
+       struct ubifs_scan_node *snod;
+       struct ubifs_orph_node *orph;
+       ino_t inum;
+       int i, n, err;
+
+       list_for_each_entry(snod, &sleb->nodes, list) {
+               cond_resched();
+               if (snod->type != UBIFS_ORPH_NODE)
+                       continue;
+               orph = snod->node;
+               n = (le32_to_cpu(orph->ch.len) - UBIFS_ORPH_NODE_SZ) >> 3;
+               for (i = 0; i < n; i++) {
+                       inum = le64_to_cpu(orph->inos[i]);
+                       err = dbg_ins_check_orphan(&ci->root, inum);
+                       if (err)
+                               return err;
+               }
+       }
+       return 0;
+}
+
+static int dbg_scan_orphans(struct ubifs_info *c, struct check_info *ci)
+{
+       int lnum, err = 0;
+       void *buf;
+
+       /* Check no-orphans flag and skip this if no orphans */
+       if (c->no_orphs)
+               return 0;
+
+       buf = __vmalloc(c->leb_size, GFP_NOFS, PAGE_KERNEL);
+       if (!buf) {
+               ubifs_err("cannot allocate memory to check orphans");
+               return 0;
+       }
+
+       for (lnum = c->orph_first; lnum <= c->orph_last; lnum++) {
+               struct ubifs_scan_leb *sleb;
+
+               sleb = ubifs_scan(c, lnum, 0, buf, 0);
+               if (IS_ERR(sleb)) {
+                       err = PTR_ERR(sleb);
+                       break;
+               }
+
+               err = dbg_read_orphans(ci, sleb);
+               ubifs_scan_destroy(sleb);
+               if (err)
+                       break;
+       }
+
+       vfree(buf);
+       return err;
+}
+
+static int dbg_check_orphans(struct ubifs_info *c)
+{
+       struct check_info ci;
+       int err;
+
+       if (!dbg_is_chk_orph(c))
+               return 0;
+
+       ci.last_ino = 0;
+       ci.tot_inos = 0;
+       ci.missing  = 0;
+       ci.leaf_cnt = 0;
+       ci.root = RB_ROOT;
+       ci.node = kmalloc(UBIFS_MAX_INO_NODE_SZ, GFP_NOFS);
+       if (!ci.node) {
+               ubifs_err("out of memory");
+               return -ENOMEM;
+       }
+
+       err = dbg_scan_orphans(c, &ci);
+       if (err)
+               goto out;
+
+       err = dbg_walk_index(c, &dbg_orphan_check, NULL, &ci);
+       if (err) {
+               ubifs_err("cannot scan TNC, error %d", err);
+               goto out;
+       }
+
+       if (ci.missing) {
+               ubifs_err("%lu missing orphan(s)", ci.missing);
+               err = -EINVAL;
+               goto out;
+       }
+
+       dbg_cmt("last inode number is %lu", ci.last_ino);
+       dbg_cmt("total number of inodes is %lu", ci.tot_inos);
+       dbg_cmt("total number of leaf nodes is %llu", ci.leaf_cnt);
+
+out:
+       dbg_free_check_tree(&ci.root);
+       kfree(ci.node);
+       return err;
+}
index 744465005caf036f52e2347e6214c6d3cb53b999..f54a440cd51d8a66fecb2ba3232eb570e3bd3efe 100644 (file)
@@ -3,18 +3,7 @@
  *
  * Copyright (C) 2006-2008 Nokia Corporation
  *
- * This program is free software; you can redistribute it and/or modify it
- * under the terms of the GNU General Public License version 2 as published by
- * the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
- * more details.
- *
- * You should have received a copy of the GNU General Public License along with
- * this program; if not, write to the Free Software Foundation, Inc., 51
- * Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
+ * SPDX-License-Identifier:    GPL-2.0+
  *
  * Authors: Adrian Hunter
  *          Artem Bityutskiy (Битюцкий Артём)
 /*
  * This file implements functions needed to recover from unclean un-mounts.
  * When UBIFS is mounted, it checks a flag on the master node to determine if
- * an un-mount was completed sucessfully. If not, the process of mounting
- * incorparates additional checking and fixing of on-flash data structures.
+ * an un-mount was completed successfully. If not, the process of mounting
+ * incorporates additional checking and fixing of on-flash data structures.
  * UBIFS always cleans away all remnants of an unclean un-mount, so that
  * errors do not accumulate. However UBIFS defers recovery if it is mounted
  * read-only, and the flash is not modified in that case.
+ *
+ * The general UBIFS approach to the recovery is that it recovers from
+ * corruptions which could be caused by power cuts, but it refuses to recover
+ * from corruption caused by other reasons. And UBIFS tries to distinguish
+ * between these 2 reasons of corruptions and silently recover in the former
+ * case and loudly complain in the latter case.
+ *
+ * UBIFS writes only to erased LEBs, so it writes only to the flash space
+ * containing only 0xFFs. UBIFS also always writes strictly from the beginning
+ * of the LEB to the end. And UBIFS assumes that the underlying flash media
+ * writes in @c->max_write_size bytes at a time.
+ *
+ * Hence, if UBIFS finds a corrupted node at offset X, it expects only the min.
+ * I/O unit corresponding to offset X to contain corrupted data, all the
+ * following min. I/O units have to contain empty space (all 0xFFs). If this is
+ * not true, the corruption cannot be the result of a power cut, and UBIFS
+ * refuses to mount.
  */
 
+#define __UBOOT__
+#ifndef __UBOOT__
+#include <linux/crc32.h>
+#include <linux/slab.h>
+#else
+#include <linux/err.h>
+#endif
 #include "ubifs.h"
 
 /**
@@ -51,6 +64,25 @@ static int is_empty(void *buf, int len)
        return 1;
 }
 
+/**
+ * first_non_ff - find offset of the first non-0xff byte.
+ * @buf: buffer to search in
+ * @len: length of buffer
+ *
+ * This function returns offset of the first non-0xff byte in @buf or %-1 if
+ * the buffer contains only 0xff bytes.
+ */
+static int first_non_ff(void *buf, int len)
+{
+       uint8_t *p = buf;
+       int i;
+
+       for (i = 0; i < len; i++)
+               if (*p++ != 0xff)
+                       return i;
+       return -1;
+}
+
 /**
  * get_master_node - get the last valid master node allowing for corruption.
  * @c: UBIFS file-system description object
@@ -79,7 +111,7 @@ static int get_master_node(const struct ubifs_info *c, int lnum, void **pbuf,
        if (!sbuf)
                return -ENOMEM;
 
-       err = ubi_read(c->ubi, lnum, sbuf, 0, c->leb_size);
+       err = ubifs_leb_read(c, lnum, sbuf, 0, c->leb_size, 0);
        if (err && err != -EBADMSG)
                goto out_free;
 
@@ -175,10 +207,10 @@ static int write_rcvrd_mst_node(struct ubifs_info *c,
        mst->flags |= cpu_to_le32(UBIFS_MST_RCVRY);
 
        ubifs_prepare_node(c, mst, UBIFS_MST_NODE_SZ, 1);
-       err = ubi_leb_change(c->ubi, lnum, mst, sz, UBI_SHORTTERM);
+       err = ubifs_leb_change(c, lnum, mst, sz);
        if (err)
                goto out;
-       err = ubi_leb_change(c->ubi, lnum + 1, mst, sz, UBI_SHORTTERM);
+       err = ubifs_leb_change(c, lnum + 1, mst, sz);
        if (err)
                goto out;
 out:
@@ -236,7 +268,8 @@ int ubifs_recover_master_node(struct ubifs_info *c)
                                if (cor1)
                                        goto out_err;
                                mst = mst1;
-                       } else if (offs1 == 0 && offs2 + sz >= c->leb_size) {
+                       } else if (offs1 == 0 &&
+                                  c->leb_size - offs2 - sz < sz) {
                                /* 1st LEB was unmapped and written, 2nd not */
                                if (cor1)
                                        goto out_err;
@@ -266,12 +299,12 @@ int ubifs_recover_master_node(struct ubifs_info *c)
                mst = mst2;
        }
 
-       dbg_rcvry("recovered master node from LEB %d",
+       ubifs_msg("recovered master node from LEB %d",
                  (mst == mst1 ? UBIFS_MST_LNUM : UBIFS_MST_LNUM + 1));
 
        memcpy(c->mst_node, mst, UBIFS_MST_NODE_SZ);
 
-       if ((c->vfs_sb->s_flags & MS_RDONLY)) {
+       if (c->ro_mount) {
                /* Read-only mode. Keep a copy for switching to rw mode */
                c->rcvrd_mst_node = kmalloc(sz, GFP_KERNEL);
                if (!c->rcvrd_mst_node) {
@@ -279,6 +312,40 @@ int ubifs_recover_master_node(struct ubifs_info *c)
                        goto out_free;
                }
                memcpy(c->rcvrd_mst_node, c->mst_node, UBIFS_MST_NODE_SZ);
+
+               /*
+                * We had to recover the master node, which means there was an
+                * unclean reboot. However, it is possible that the master node
+                * is clean at this point, i.e., %UBIFS_MST_DIRTY is not set.
+                * E.g., consider the following chain of events:
+                *
+                * 1. UBIFS was cleanly unmounted, so the master node is clean
+                * 2. UBIFS is being mounted R/W and starts changing the master
+                *    node in the first (%UBIFS_MST_LNUM). A power cut happens,
+                *    so this LEB ends up with some amount of garbage at the
+                *    end.
+                * 3. UBIFS is being mounted R/O. We reach this place and
+                *    recover the master node from the second LEB
+                *    (%UBIFS_MST_LNUM + 1). But we cannot update the media
+                *    because we are being mounted R/O. We have to defer the
+                *    operation.
+                * 4. However, this master node (@c->mst_node) is marked as
+                *    clean (since the step 1). And if we just return, the
+                *    mount code will be confused and won't recover the master
+                *    node when it is re-mounter R/W later.
+                *
+                *    Thus, to force the recovery by marking the master node as
+                *    dirty.
+                */
+               c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
+#ifndef __UBOOT__
+       } else {
+               /* Write the recovered master node */
+               c->max_sqnum = le64_to_cpu(mst->ch.sqnum) - 1;
+               err = write_rcvrd_mst_node(c, c->mst_node);
+               if (err)
+                       goto out_free;
+#endif
        }
 
        vfree(buf2);
@@ -291,12 +358,12 @@ out_err:
 out_free:
        ubifs_err("failed to recover master node");
        if (mst1) {
-               dbg_err("dumping first master node");
-               dbg_dump_node(c, mst1);
+               ubifs_err("dumping first master node");
+               ubifs_dump_node(c, mst1);
        }
        if (mst2) {
-               dbg_err("dumping second master node");
-               dbg_dump_node(c, mst2);
+               ubifs_err("dumping second master node");
+               ubifs_dump_node(c, mst2);
        }
        vfree(buf2);
        vfree(buf1);
@@ -335,44 +402,23 @@ int ubifs_write_rcvrd_mst_node(struct ubifs_info *c)
  * @offs: offset to check
  *
  * This function returns %1 if @offs was in the last write to the LEB whose data
- * is in @buf, otherwise %0 is returned.  The determination is made by checking
- * for subsequent empty space starting from the next min_io_size boundary (or a
- * bit less than the common header size if min_io_size is one).
+ * is in @buf, otherwise %0 is returned. The determination is made by checking
+ * for subsequent empty space starting from the next @c->max_write_size
+ * boundary.
  */
 static int is_last_write(const struct ubifs_info *c, void *buf, int offs)
 {
-       int empty_offs;
-       int check_len;
+       int empty_offs, check_len;
        uint8_t *p;
 
-       if (c->min_io_size == 1) {
-               check_len = c->leb_size - offs;
-               p = buf + check_len;
-               for (; check_len > 0; check_len--)
-                       if (*--p != 0xff)
-                               break;
-               /*
-                * 'check_len' is the size of the corruption which cannot be
-                * more than the size of 1 node if it was caused by an unclean
-                * unmount.
-                */
-               if (check_len > UBIFS_MAX_NODE_SZ)
-                       return 0;
-               return 1;
-       }
-
        /*
-        * Round up to the next c->min_io_size boundary i.e. 'offs' is in the
-        * last wbuf written. After that should be empty space.
+        * Round up to the next @c->max_write_size boundary i.e. @offs is in
+        * the last wbuf written. After that should be empty space.
         */
-       empty_offs = ALIGN(offs + 1, c->min_io_size);
+       empty_offs = ALIGN(offs + 1, c->max_write_size);
        check_len = c->leb_size - empty_offs;
        p = buf + empty_offs - offs;
-
-       for (; check_len > 0; check_len--)
-               if (*p++ != 0xff)
-                       return 0;
-       return 1;
+       return is_empty(p, check_len);
 }
 
 /**
@@ -385,7 +431,7 @@ static int is_last_write(const struct ubifs_info *c, void *buf, int offs)
  *
  * This function pads up to the next min_io_size boundary (if there is one) and
  * sets empty space to all 0xff. @buf, @offs and @len are updated to the next
- * min_io_size boundary (if there is one).
+ * @c->min_io_size boundary.
  */
 static void clean_buf(const struct ubifs_info *c, void **buf, int lnum,
                      int *offs, int *len)
@@ -395,11 +441,6 @@ static void clean_buf(const struct ubifs_info *c, void **buf, int lnum,
        lnum = lnum;
        dbg_rcvry("cleaning corruption at %d:%d", lnum, *offs);
 
-       if (c->min_io_size == 1) {
-               memset(*buf, 0xff, c->leb_size - *offs);
-               return;
-       }
-
        ubifs_assert(!(*offs & 7));
        empty_offs = ALIGN(*offs, c->min_io_size);
        pad_len = empty_offs - *offs;
@@ -429,7 +470,7 @@ static int no_more_nodes(const struct ubifs_info *c, void *buf, int len,
        int skip, dlen = le32_to_cpu(ch->len);
 
        /* Check for empty space after the corrupt node's common header */
-       skip = ALIGN(offs + UBIFS_CH_SZ, c->min_io_size) - offs;
+       skip = ALIGN(offs + UBIFS_CH_SZ, c->max_write_size) - offs;
        if (is_empty(buf + skip, len - skip))
                return 1;
        /*
@@ -441,7 +482,7 @@ static int no_more_nodes(const struct ubifs_info *c, void *buf, int len,
                return 0;
        }
        /* Now we know the corrupt node's length we can skip over it */
-       skip = ALIGN(offs + dlen, c->min_io_size) - offs;
+       skip = ALIGN(offs + dlen, c->max_write_size) - offs;
        /* After which there should be empty space */
        if (is_empty(buf + skip, len - skip))
                return 1;
@@ -469,7 +510,7 @@ static int fix_unclean_leb(struct ubifs_info *c, struct ubifs_scan_leb *sleb,
                endpt = snod->offs + snod->len;
        }
 
-       if ((c->vfs_sb->s_flags & MS_RDONLY) && !c->remounting_rw) {
+       if (c->ro_mount && !c->remounting_rw) {
                /* Add to recovery list */
                struct ubifs_unclean_leb *ucleb;
 
@@ -481,21 +522,55 @@ static int fix_unclean_leb(struct ubifs_info *c, struct ubifs_scan_leb *sleb,
                ucleb->lnum = lnum;
                ucleb->endpt = endpt;
                list_add_tail(&ucleb->list, &c->unclean_leb_list);
+#ifndef __UBOOT__
+       } else {
+               /* Write the fixed LEB back to flash */
+               int err;
+
+               dbg_rcvry("fixing LEB %d start %d endpt %d",
+                         lnum, start, sleb->endpt);
+               if (endpt == 0) {
+                       err = ubifs_leb_unmap(c, lnum);
+                       if (err)
+                               return err;
+               } else {
+                       int len = ALIGN(endpt, c->min_io_size);
+
+                       if (start) {
+                               err = ubifs_leb_read(c, lnum, sleb->buf, 0,
+                                                    start, 1);
+                               if (err)
+                                       return err;
+                       }
+                       /* Pad to min_io_size */
+                       if (len > endpt) {
+                               int pad_len = len - ALIGN(endpt, 8);
+
+                               if (pad_len > 0) {
+                                       void *buf = sleb->buf + len - pad_len;
+
+                                       ubifs_pad(c, buf, pad_len);
+                               }
+                       }
+                       err = ubifs_leb_change(c, lnum, sleb->buf, len);
+                       if (err)
+                               return err;
+               }
+#endif
        }
        return 0;
 }
 
 /**
- * drop_incomplete_group - drop nodes from an incomplete group.
+ * drop_last_group - drop the last group of nodes.
  * @sleb: scanned LEB information
  * @offs: offset of dropped nodes is returned here
  *
- * This function returns %1 if nodes are dropped and %0 otherwise.
+ * This is a helper function for 'ubifs_recover_leb()' which drops the last
+ * group of nodes of the scanned LEB.
  */
-static int drop_incomplete_group(struct ubifs_scan_leb *sleb, int *offs)
+static void drop_last_group(struct ubifs_scan_leb *sleb, int *offs)
 {
-       int dropped = 0;
-
        while (!list_empty(&sleb->nodes)) {
                struct ubifs_scan_node *snod;
                struct ubifs_ch *ch;
@@ -504,15 +579,41 @@ static int drop_incomplete_group(struct ubifs_scan_leb *sleb, int *offs)
                                  list);
                ch = snod->node;
                if (ch->group_type != UBIFS_IN_NODE_GROUP)
-                       return dropped;
-               dbg_rcvry("dropping node at %d:%d", sleb->lnum, snod->offs);
+                       break;
+
+               dbg_rcvry("dropping grouped node at %d:%d",
+                         sleb->lnum, snod->offs);
+               *offs = snod->offs;
+               list_del(&snod->list);
+               kfree(snod);
+               sleb->nodes_cnt -= 1;
+       }
+}
+
+/**
+ * drop_last_node - drop the last node.
+ * @sleb: scanned LEB information
+ * @offs: offset of dropped nodes is returned here
+ * @grouped: non-zero if whole group of nodes have to be dropped
+ *
+ * This is a helper function for 'ubifs_recover_leb()' which drops the last
+ * node of the scanned LEB.
+ */
+static void drop_last_node(struct ubifs_scan_leb *sleb, int *offs)
+{
+       struct ubifs_scan_node *snod;
+
+       if (!list_empty(&sleb->nodes)) {
+               snod = list_entry(sleb->nodes.prev, struct ubifs_scan_node,
+                                 list);
+
+               dbg_rcvry("dropping last node at %d:%d",
+                         sleb->lnum, snod->offs);
                *offs = snod->offs;
                list_del(&snod->list);
                kfree(snod);
                sleb->nodes_cnt -= 1;
-               dropped = 1;
        }
-       return dropped;
 }
 
 /**
@@ -521,33 +622,30 @@ static int drop_incomplete_group(struct ubifs_scan_leb *sleb, int *offs)
  * @lnum: LEB number
  * @offs: offset
  * @sbuf: LEB-sized buffer to use
- * @grouped: nodes may be grouped for recovery
+ * @jhead: journal head number this LEB belongs to (%-1 if the LEB does not
+ *         belong to any journal head)
  *
  * This function does a scan of a LEB, but caters for errors that might have
  * been caused by the unclean unmount from which we are attempting to recover.
- *
- * This function returns %0 on success and a negative error code on failure.
+ * Returns %0 in case of success, %-EUCLEAN if an unrecoverable corruption is
+ * found, and a negative error code in case of failure.
  */
 struct ubifs_scan_leb *ubifs_recover_leb(struct ubifs_info *c, int lnum,
-                                        int offs, void *sbuf, int grouped)
+                                        int offs, void *sbuf, int jhead)
 {
-       int err, len = c->leb_size - offs, need_clean = 0, quiet = 1;
-       int empty_chkd = 0, start = offs;
+       int ret = 0, err, len = c->leb_size - offs, start = offs, min_io_unit;
+       int grouped = jhead == -1 ? 0 : c->jheads[jhead].grouped;
        struct ubifs_scan_leb *sleb;
        void *buf = sbuf + offs;
 
-       dbg_rcvry("%d:%d", lnum, offs);
+       dbg_rcvry("%d:%d, jhead %d, grouped %d", lnum, offs, jhead, grouped);
 
        sleb = ubifs_start_scan(c, lnum, offs, sbuf);
        if (IS_ERR(sleb))
                return sleb;
 
-       if (sleb->ecc)
-               need_clean = 1;
-
+       ubifs_assert(len >= 8);
        while (len >= 8) {
-               int ret;
-
                dbg_scan("look at LEB %d:%d (%d bytes left)",
                         lnum, offs, len);
 
@@ -557,8 +655,7 @@ struct ubifs_scan_leb *ubifs_recover_leb(struct ubifs_info *c, int lnum,
                 * Scan quietly until there is an error from which we cannot
                 * recover
                 */
-               ret = ubifs_scan_a_node(c, buf, len, lnum, offs, quiet);
-
+               ret = ubifs_scan_a_node(c, buf, len, lnum, offs, 1);
                if (ret == SCANNED_A_NODE) {
                        /* A valid node, and not a padding node */
                        struct ubifs_ch *ch = buf;
@@ -571,98 +668,127 @@ struct ubifs_scan_leb *ubifs_recover_leb(struct ubifs_info *c, int lnum,
                        offs += node_len;
                        buf += node_len;
                        len -= node_len;
-                       continue;
-               }
-
-               if (ret > 0) {
+               } else if (ret > 0) {
                        /* Padding bytes or a valid padding node */
                        offs += ret;
                        buf += ret;
                        len -= ret;
-                       continue;
-               }
-
-               if (ret == SCANNED_EMPTY_SPACE) {
-                       if (!is_empty(buf, len)) {
-                               if (!is_last_write(c, buf, offs))
-                                       break;
-                               clean_buf(c, &buf, lnum, &offs, &len);
-                               need_clean = 1;
-                       }
-                       empty_chkd = 1;
+               } else if (ret == SCANNED_EMPTY_SPACE ||
+                          ret == SCANNED_GARBAGE     ||
+                          ret == SCANNED_A_BAD_PAD_NODE ||
+                          ret == SCANNED_A_CORRUPT_NODE) {
+                       dbg_rcvry("found corruption (%d) at %d:%d",
+                                 ret, lnum, offs);
                        break;
+               } else {
+                       ubifs_err("unexpected return value %d", ret);
+                       err = -EINVAL;
+                       goto error;
                }
+       }
 
-               if (ret == SCANNED_GARBAGE || ret == SCANNED_A_BAD_PAD_NODE)
-                       if (is_last_write(c, buf, offs)) {
-                               clean_buf(c, &buf, lnum, &offs, &len);
-                               need_clean = 1;
-                               empty_chkd = 1;
-                               break;
-                       }
-
-               if (ret == SCANNED_A_CORRUPT_NODE)
-                       if (no_more_nodes(c, buf, len, lnum, offs)) {
-                               clean_buf(c, &buf, lnum, &offs, &len);
-                               need_clean = 1;
-                               empty_chkd = 1;
-                               break;
-                       }
-
-               if (quiet) {
-                       /* Redo the last scan but noisily */
-                       quiet = 0;
-                       continue;
-               }
+       if (ret == SCANNED_GARBAGE || ret == SCANNED_A_BAD_PAD_NODE) {
+               if (!is_last_write(c, buf, offs))
+                       goto corrupted_rescan;
+       } else if (ret == SCANNED_A_CORRUPT_NODE) {
+               if (!no_more_nodes(c, buf, len, lnum, offs))
+                       goto corrupted_rescan;
+       } else if (!is_empty(buf, len)) {
+               if (!is_last_write(c, buf, offs)) {
+                       int corruption = first_non_ff(buf, len);
 
-               switch (ret) {
-               case SCANNED_GARBAGE:
-                       dbg_err("garbage");
-                       goto corrupted;
-               case SCANNED_A_CORRUPT_NODE:
-               case SCANNED_A_BAD_PAD_NODE:
-                       dbg_err("bad node");
-                       goto corrupted;
-               default:
-                       dbg_err("unknown");
+                       /*
+                        * See header comment for this file for more
+                        * explanations about the reasons we have this check.
+                        */
+                       ubifs_err("corrupt empty space LEB %d:%d, corruption starts at %d",
+                                 lnum, offs, corruption);
+                       /* Make sure we dump interesting non-0xFF data */
+                       offs += corruption;
+                       buf += corruption;
                        goto corrupted;
                }
        }
 
-       if (!empty_chkd && !is_empty(buf, len)) {
-               if (is_last_write(c, buf, offs)) {
-                       clean_buf(c, &buf, lnum, &offs, &len);
-                       need_clean = 1;
-               } else {
-                       ubifs_err("corrupt empty space at LEB %d:%d",
-                                 lnum, offs);
-                       goto corrupted;
-               }
-       }
+       min_io_unit = round_down(offs, c->min_io_size);
+       if (grouped)
+               /*
+                * If nodes are grouped, always drop the incomplete group at
+                * the end.
+                */
+               drop_last_group(sleb, &offs);
 
-       /* Drop nodes from incomplete group */
-       if (grouped && drop_incomplete_group(sleb, &offs)) {
-               buf = sbuf + offs;
-               len = c->leb_size - offs;
-               clean_buf(c, &buf, lnum, &offs, &len);
-               need_clean = 1;
+       if (jhead == GCHD) {
+               /*
+                * If this LEB belongs to the GC head then while we are in the
+                * middle of the same min. I/O unit keep dropping nodes. So
+                * basically, what we want is to make sure that the last min.
+                * I/O unit where we saw the corruption is dropped completely
+                * with all the uncorrupted nodes which may possibly sit there.
+                *
+                * In other words, let's name the min. I/O unit where the
+                * corruption starts B, and the previous min. I/O unit A. The
+                * below code tries to deal with a situation when half of B
+                * contains valid nodes or the end of a valid node, and the
+                * second half of B contains corrupted data or garbage. This
+                * means that UBIFS had been writing to B just before the power
+                * cut happened. I do not know how realistic is this scenario
+                * that half of the min. I/O unit had been written successfully
+                * and the other half not, but this is possible in our 'failure
+                * mode emulation' infrastructure at least.
+                *
+                * So what is the problem, why we need to drop those nodes? Why
+                * can't we just clean-up the second half of B by putting a
+                * padding node there? We can, and this works fine with one
+                * exception which was reproduced with power cut emulation
+                * testing and happens extremely rarely.
+                *
+                * Imagine the file-system is full, we run GC which starts
+                * moving valid nodes from LEB X to LEB Y (obviously, LEB Y is
+                * the current GC head LEB). The @c->gc_lnum is -1, which means
+                * that GC will retain LEB X and will try to continue. Imagine
+                * that LEB X is currently the dirtiest LEB, and the amount of
+                * used space in LEB Y is exactly the same as amount of free
+                * space in LEB X.
+                *
+                * And a power cut happens when nodes are moved from LEB X to
+                * LEB Y. We are here trying to recover LEB Y which is the GC
+                * head LEB. We find the min. I/O unit B as described above.
+                * Then we clean-up LEB Y by padding min. I/O unit. And later
+                * 'ubifs_rcvry_gc_commit()' function fails, because it cannot
+                * find a dirty LEB which could be GC'd into LEB Y! Even LEB X
+                * does not match because the amount of valid nodes there does
+                * not fit the free space in LEB Y any more! And this is
+                * because of the padding node which we added to LEB Y. The
+                * user-visible effect of this which I once observed and
+                * analysed is that we cannot mount the file-system with
+                * -ENOSPC error.
+                *
+                * So obviously, to make sure that situation does not happen we
+                * should free min. I/O unit B in LEB Y completely and the last
+                * used min. I/O unit in LEB Y should be A. This is basically
+                * what the below code tries to do.
+                */
+               while (offs > min_io_unit)
+                       drop_last_node(sleb, &offs);
        }
 
-       if (offs % c->min_io_size) {
-               clean_buf(c, &buf, lnum, &offs, &len);
-               need_clean = 1;
-       }
+       buf = sbuf + offs;
+       len = c->leb_size - offs;
 
+       clean_buf(c, &buf, lnum, &offs, &len);
        ubifs_end_scan(c, sleb, lnum, offs);
 
-       if (need_clean) {
-               err = fix_unclean_leb(c, sleb, start);
-               if (err)
-                       goto error;
-       }
+       err = fix_unclean_leb(c, sleb, start);
+       if (err)
+               goto error;
 
        return sleb;
 
+corrupted_rescan:
+       /* Re-scan the corrupted data with verbose messages */
+       ubifs_err("corruption %d", ret);
+       ubifs_scan_a_node(c, buf, len, lnum, offs, 1);
 corrupted:
        ubifs_scanned_corruption(c, lnum, offs, buf);
        err = -EUCLEAN;
@@ -693,22 +819,23 @@ static int get_cs_sqnum(struct ubifs_info *c, int lnum, int offs,
                return -ENOMEM;
        if (c->leb_size - offs < UBIFS_CS_NODE_SZ)
                goto out_err;
-       err = ubi_read(c->ubi, lnum, (void *)cs_node, offs, UBIFS_CS_NODE_SZ);
+       err = ubifs_leb_read(c, lnum, (void *)cs_node, offs,
+                            UBIFS_CS_NODE_SZ, 0);
        if (err && err != -EBADMSG)
                goto out_free;
        ret = ubifs_scan_a_node(c, cs_node, UBIFS_CS_NODE_SZ, lnum, offs, 0);
        if (ret != SCANNED_A_NODE) {
-               dbg_err("Not a valid node");
+               ubifs_err("Not a valid node");
                goto out_err;
        }
        if (cs_node->ch.node_type != UBIFS_CS_NODE) {
-               dbg_err("Node a CS node, type is %d", cs_node->ch.node_type);
+               ubifs_err("Node a CS node, type is %d", cs_node->ch.node_type);
                goto out_err;
        }
        if (le64_to_cpu(cs_node->cmt_no) != c->cmt_no) {
-               dbg_err("CS node cmt_no %llu != current cmt_no %llu",
-                       (unsigned long long)le64_to_cpu(cs_node->cmt_no),
-                       c->cmt_no);
+               ubifs_err("CS node cmt_no %llu != current cmt_no %llu",
+                         (unsigned long long)le64_to_cpu(cs_node->cmt_no),
+                         c->cmt_no);
                goto out_err;
        }
        *cs_sqnum = le64_to_cpu(cs_node->ch.sqnum);
@@ -732,7 +859,8 @@ out_free:
  * @sbuf: LEB-sized buffer to use
  *
  * This function does a scan of a LEB, but caters for errors that might have
- * been caused by the unclean unmount from which we are attempting to recover.
+ * been caused by unclean reboots from which we are attempting to recover
+ * (assume that only the last log LEB can be corrupted by an unclean reboot).
  *
  * This function returns %0 on success and a negative error code on failure.
  */
@@ -751,7 +879,7 @@ struct ubifs_scan_leb *ubifs_recover_log_leb(struct ubifs_info *c, int lnum,
                 * We can only recover at the end of the log, so check that the
                 * next log LEB is empty or out of date.
                 */
-               sleb = ubifs_scan(c, next_lnum, 0, sbuf);
+               sleb = ubifs_scan(c, next_lnum, 0, sbuf, 0);
                if (IS_ERR(sleb))
                        return sleb;
                if (sleb->nodes_cnt) {
@@ -770,15 +898,15 @@ struct ubifs_scan_leb *ubifs_recover_log_leb(struct ubifs_info *c, int lnum,
                                }
                        }
                        if (snod->sqnum > cs_sqnum) {
-                               ubifs_err("unrecoverable log corruption "
-                                         "in LEB %d", lnum);
+                               ubifs_err("unrecoverable log corruption in LEB %d",
+                                         lnum);
                                ubifs_scan_destroy(sleb);
                                return ERR_PTR(-EUCLEAN);
                        }
                }
                ubifs_scan_destroy(sleb);
        }
-       return ubifs_recover_leb(c, lnum, offs, sbuf, 0);
+       return ubifs_recover_leb(c, lnum, offs, sbuf, -1);
 }
 
 /**
@@ -792,15 +920,10 @@ struct ubifs_scan_leb *ubifs_recover_log_leb(struct ubifs_info *c, int lnum,
  *
  * This function returns %0 on success and a negative error code on failure.
  */
-static int recover_head(const struct ubifs_info *c, int lnum, int offs,
-                       void *sbuf)
+static int recover_head(struct ubifs_info *c, int lnum, int offs, void *sbuf)
 {
-       int len, err, need_clean = 0;
+       int len = c->max_write_size, err;
 
-       if (c->min_io_size > 1)
-               len = c->min_io_size;
-       else
-               len = 512;
        if (offs + len > c->leb_size)
                len = c->leb_size - offs;
 
@@ -808,27 +931,15 @@ static int recover_head(const struct ubifs_info *c, int lnum, int offs,
                return 0;
 
        /* Read at the head location and check it is empty flash */
-       err = ubi_read(c->ubi, lnum, sbuf, offs, len);
-       if (err)
-               need_clean = 1;
-       else {
-               uint8_t *p = sbuf;
-
-               while (len--)
-                       if (*p++ != 0xff) {
-                               need_clean = 1;
-                               break;
-                       }
-       }
-
-       if (need_clean) {
+       err = ubifs_leb_read(c, lnum, sbuf, offs, len, 1);
+       if (err || !is_empty(sbuf, len)) {
                dbg_rcvry("cleaning head at %d:%d", lnum, offs);
                if (offs == 0)
                        return ubifs_leb_unmap(c, lnum);
-               err = ubi_read(c->ubi, lnum, sbuf, 0, offs);
+               err = ubifs_leb_read(c, lnum, sbuf, 0, offs, 1);
                if (err)
                        return err;
-               return ubi_leb_change(c->ubi, lnum, sbuf, offs, UBI_UNKNOWN);
+               return ubifs_leb_change(c, lnum, sbuf, offs);
        }
 
        return 0;
@@ -851,11 +962,11 @@ static int recover_head(const struct ubifs_info *c, int lnum, int offs,
  *
  * This function returns %0 on success and a negative error code on failure.
  */
-int ubifs_recover_inl_heads(const struct ubifs_info *c, void *sbuf)
+int ubifs_recover_inl_heads(struct ubifs_info *c, void *sbuf)
 {
        int err;
 
-       ubifs_assert(!(c->vfs_sb->s_flags & MS_RDONLY) || c->remounting_rw);
+       ubifs_assert(!c->ro_mount || c->remounting_rw);
 
        dbg_rcvry("checking index head at %d:%d", c->ihead_lnum, c->ihead_offs);
        err = recover_head(c, c->ihead_lnum, c->ihead_offs, sbuf);
@@ -871,7 +982,7 @@ int ubifs_recover_inl_heads(const struct ubifs_info *c, void *sbuf)
 }
 
 /**
- *  clean_an_unclean_leb - read and write a LEB to remove corruption.
+ * clean_an_unclean_leb - read and write a LEB to remove corruption.
  * @c: UBIFS file-system description object
  * @ucleb: unclean LEB information
  * @sbuf: LEB-sized buffer to use
@@ -882,7 +993,7 @@ int ubifs_recover_inl_heads(const struct ubifs_info *c, void *sbuf)
  *
  * This function returns %0 on success and a negative error code on failure.
  */
-static int clean_an_unclean_leb(const struct ubifs_info *c,
+static int clean_an_unclean_leb(struct ubifs_info *c,
                                struct ubifs_unclean_leb *ucleb, void *sbuf)
 {
        int err, lnum = ucleb->lnum, offs = 0, len = ucleb->endpt, quiet = 1;
@@ -898,7 +1009,7 @@ static int clean_an_unclean_leb(const struct ubifs_info *c,
                return 0;
        }
 
-       err = ubi_read(c->ubi, lnum, buf, offs, len);
+       err = ubifs_leb_read(c, lnum, buf, offs, len, 0);
        if (err && err != -EBADMSG)
                return err;
 
@@ -958,7 +1069,7 @@ static int clean_an_unclean_leb(const struct ubifs_info *c,
        }
 
        /* Write back the LEB atomically */
-       err = ubi_leb_change(c->ubi, lnum, sbuf, len, UBI_UNKNOWN);
+       err = ubifs_leb_change(c, lnum, sbuf, len);
        if (err)
                return err;
 
@@ -978,7 +1089,7 @@ static int clean_an_unclean_leb(const struct ubifs_info *c,
  *
  * This function returns %0 on success and a negative error code on failure.
  */
-int ubifs_clean_lebs(const struct ubifs_info *c, void *sbuf)
+int ubifs_clean_lebs(struct ubifs_info *c, void *sbuf)
 {
        dbg_rcvry("recovery");
        while (!list_empty(&c->unclean_leb_list)) {
@@ -996,6 +1107,140 @@ int ubifs_clean_lebs(const struct ubifs_info *c, void *sbuf)
        return 0;
 }
 
+#ifndef __UBOOT__
+/**
+ * grab_empty_leb - grab an empty LEB to use as GC LEB and run commit.
+ * @c: UBIFS file-system description object
+ *
+ * This is a helper function for 'ubifs_rcvry_gc_commit()' which grabs an empty
+ * LEB to be used as GC LEB (@c->gc_lnum), and then runs the commit. Returns
+ * zero in case of success and a negative error code in case of failure.
+ */
+static int grab_empty_leb(struct ubifs_info *c)
+{
+       int lnum, err;
+
+       /*
+        * Note, it is very important to first search for an empty LEB and then
+        * run the commit, not vice-versa. The reason is that there might be
+        * only one empty LEB at the moment, the one which has been the
+        * @c->gc_lnum just before the power cut happened. During the regular
+        * UBIFS operation (not now) @c->gc_lnum is marked as "taken", so no
+        * one but GC can grab it. But at this moment this single empty LEB is
+        * not marked as taken, so if we run commit - what happens? Right, the
+        * commit will grab it and write the index there. Remember that the
+        * index always expands as long as there is free space, and it only
+        * starts consolidating when we run out of space.
+        *
+        * IOW, if we run commit now, we might not be able to find a free LEB
+        * after this.
+        */
+       lnum = ubifs_find_free_leb_for_idx(c);
+       if (lnum < 0) {
+               ubifs_err("could not find an empty LEB");
+               ubifs_dump_lprops(c);
+               ubifs_dump_budg(c, &c->bi);
+               return lnum;
+       }
+
+       /* Reset the index flag */
+       err = ubifs_change_one_lp(c, lnum, LPROPS_NC, LPROPS_NC, 0,
+                                 LPROPS_INDEX, 0);
+       if (err)
+               return err;
+
+       c->gc_lnum = lnum;
+       dbg_rcvry("found empty LEB %d, run commit", lnum);
+
+       return ubifs_run_commit(c);
+}
+
+/**
+ * ubifs_rcvry_gc_commit - recover the GC LEB number and run the commit.
+ * @c: UBIFS file-system description object
+ *
+ * Out-of-place garbage collection requires always one empty LEB with which to
+ * start garbage collection. The LEB number is recorded in c->gc_lnum and is
+ * written to the master node on unmounting. In the case of an unclean unmount
+ * the value of gc_lnum recorded in the master node is out of date and cannot
+ * be used. Instead, recovery must allocate an empty LEB for this purpose.
+ * However, there may not be enough empty space, in which case it must be
+ * possible to GC the dirtiest LEB into the GC head LEB.
+ *
+ * This function also runs the commit which causes the TNC updates from
+ * size-recovery and orphans to be written to the flash. That is important to
+ * ensure correct replay order for subsequent mounts.
+ *
+ * This function returns %0 on success and a negative error code on failure.
+ */
+int ubifs_rcvry_gc_commit(struct ubifs_info *c)
+{
+       struct ubifs_wbuf *wbuf = &c->jheads[GCHD].wbuf;
+       struct ubifs_lprops lp;
+       int err;
+
+       dbg_rcvry("GC head LEB %d, offs %d", wbuf->lnum, wbuf->offs);
+
+       c->gc_lnum = -1;
+       if (wbuf->lnum == -1 || wbuf->offs == c->leb_size)
+               return grab_empty_leb(c);
+
+       err = ubifs_find_dirty_leb(c, &lp, wbuf->offs, 2);
+       if (err) {
+               if (err != -ENOSPC)
+                       return err;
+
+               dbg_rcvry("could not find a dirty LEB");
+               return grab_empty_leb(c);
+       }
+
+       ubifs_assert(!(lp.flags & LPROPS_INDEX));
+       ubifs_assert(lp.free + lp.dirty >= wbuf->offs);
+
+       /*
+        * We run the commit before garbage collection otherwise subsequent
+        * mounts will see the GC and orphan deletion in a different order.
+        */
+       dbg_rcvry("committing");
+       err = ubifs_run_commit(c);
+       if (err)
+               return err;
+
+       dbg_rcvry("GC'ing LEB %d", lp.lnum);
+       mutex_lock_nested(&wbuf->io_mutex, wbuf->jhead);
+       err = ubifs_garbage_collect_leb(c, &lp);
+       if (err >= 0) {
+               int err2 = ubifs_wbuf_sync_nolock(wbuf);
+
+               if (err2)
+                       err = err2;
+       }
+       mutex_unlock(&wbuf->io_mutex);
+       if (err < 0) {
+               ubifs_err("GC failed, error %d", err);
+               if (err == -EAGAIN)
+                       err = -EINVAL;
+               return err;
+       }
+
+       ubifs_assert(err == LEB_RETAINED);
+       if (err != LEB_RETAINED)
+               return -EINVAL;
+
+       err = ubifs_leb_unmap(c, c->gc_lnum);
+       if (err)
+               return err;
+
+       dbg_rcvry("allocated LEB %d for GC", lp.lnum);
+       return 0;
+}
+#else
+int ubifs_rcvry_gc_commit(struct ubifs_info *c)
+{
+       return 0;
+}
+#endif
+
 /**
  * struct size_entry - inode size information for recovery.
  * @rb: link in the RB-tree of sizes
@@ -1089,6 +1334,23 @@ static void remove_ino(struct ubifs_info *c, ino_t inum)
        kfree(e);
 }
 
+/**
+ * ubifs_destroy_size_tree - free resources related to the size tree.
+ * @c: UBIFS file-system description object
+ */
+void ubifs_destroy_size_tree(struct ubifs_info *c)
+{
+       struct size_entry *e, *n;
+
+       rbtree_postorder_for_each_entry_safe(e, n, &c->size_tree, rb) {
+               if (e->inode)
+                       iput(e->inode);
+               kfree(e);
+       }
+
+       c->size_tree = RB_ROOT;
+}
+
 /**
  * ubifs_recover_size_accum - accumulate inode sizes for recovery.
  * @c: UBIFS file-system description object
@@ -1157,6 +1419,64 @@ int ubifs_recover_size_accum(struct ubifs_info *c, union ubifs_key *key,
        return 0;
 }
 
+#ifndef __UBOOT__
+/**
+ * fix_size_in_place - fix inode size in place on flash.
+ * @c: UBIFS file-system description object
+ * @e: inode size information for recovery
+ */
+static int fix_size_in_place(struct ubifs_info *c, struct size_entry *e)
+{
+       struct ubifs_ino_node *ino = c->sbuf;
+       unsigned char *p;
+       union ubifs_key key;
+       int err, lnum, offs, len;
+       loff_t i_size;
+       uint32_t crc;
+
+       /* Locate the inode node LEB number and offset */
+       ino_key_init(c, &key, e->inum);
+       err = ubifs_tnc_locate(c, &key, ino, &lnum, &offs);
+       if (err)
+               goto out;
+       /*
+        * If the size recorded on the inode node is greater than the size that
+        * was calculated from nodes in the journal then don't change the inode.
+        */
+       i_size = le64_to_cpu(ino->size);
+       if (i_size >= e->d_size)
+               return 0;
+       /* Read the LEB */
+       err = ubifs_leb_read(c, lnum, c->sbuf, 0, c->leb_size, 1);
+       if (err)
+               goto out;
+       /* Change the size field and recalculate the CRC */
+       ino = c->sbuf + offs;
+       ino->size = cpu_to_le64(e->d_size);
+       len = le32_to_cpu(ino->ch.len);
+       crc = crc32(UBIFS_CRC32_INIT, (void *)ino + 8, len - 8);
+       ino->ch.crc = cpu_to_le32(crc);
+       /* Work out where data in the LEB ends and free space begins */
+       p = c->sbuf;
+       len = c->leb_size - 1;
+       while (p[len] == 0xff)
+               len -= 1;
+       len = ALIGN(len + 1, c->min_io_size);
+       /* Atomically write the fixed LEB back again */
+       err = ubifs_leb_change(c, lnum, c->sbuf, len);
+       if (err)
+               goto out;
+       dbg_rcvry("inode %lu at %d:%d size %lld -> %lld",
+                 (unsigned long)e->inum, lnum, offs, i_size, e->d_size);
+       return 0;
+
+out:
+       ubifs_warn("inode %lu failed to fix size %lld -> %lld error %d",
+                  (unsigned long)e->inum, e->i_size, e->d_size, err);
+       return err;
+}
+#endif
+
 /**
  * ubifs_recover_size - recover inode size.
  * @c: UBIFS file-system description object
@@ -1196,30 +1516,48 @@ int ubifs_recover_size(struct ubifs_info *c)
                                e->i_size = le64_to_cpu(ino->size);
                        }
                }
+
                if (e->exists && e->i_size < e->d_size) {
-                       if (!e->inode && (c->vfs_sb->s_flags & MS_RDONLY)) {
+                       if (c->ro_mount) {
                                /* Fix the inode size and pin it in memory */
                                struct inode *inode;
+                               struct ubifs_inode *ui;
+
+                               ubifs_assert(!e->inode);
 
                                inode = ubifs_iget(c->vfs_sb, e->inum);
                                if (IS_ERR(inode))
                                        return PTR_ERR(inode);
+
+                               ui = ubifs_inode(inode);
                                if (inode->i_size < e->d_size) {
                                        dbg_rcvry("ino %lu size %lld -> %lld",
                                                  (unsigned long)e->inum,
-                                                 e->d_size, inode->i_size);
+                                                 inode->i_size, e->d_size);
                                        inode->i_size = e->d_size;
-                                       ubifs_inode(inode)->ui_size = e->d_size;
+                                       ui->ui_size = e->d_size;
+                                       ui->synced_i_size = e->d_size;
                                        e->inode = inode;
                                        this = rb_next(this);
                                        continue;
                                }
                                iput(inode);
+#ifndef __UBOOT__
+                       } else {
+                               /* Fix the size in place */
+                               err = fix_size_in_place(c, e);
+                               if (err)
+                                       return err;
+                               if (e->inode)
+                                       iput(e->inode);
+#endif
                        }
                }
+
                this = rb_next(this);
                rb_erase(&e->rb, &c->size_tree);
                kfree(e);
        }
+
        return 0;
 }
index da33a14ab6616f458386d756106f95c7dc2beea1..6393b15b1852d95300181248e114736030b9b2ac 100644 (file)
@@ -3,18 +3,7 @@
  *
  * Copyright (C) 2006-2008 Nokia Corporation.
  *
- * This program is free software; you can redistribute it and/or modify it
- * under the terms of the GNU General Public License version 2 as published by
- * the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
- * more details.
- *
- * You should have received a copy of the GNU General Public License along with
- * this program; if not, write to the Free Software Foundation, Inc., 51
- * Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
+ * SPDX-License-Identifier:    GPL-2.0+
  *
  * Authors: Adrian Hunter
  *          Artem Bityutskiy (Битюцкий Артём)
  * larger is the journal, the more memory its index may consume.
  */
 
+#define __UBOOT__
+#ifdef __UBOOT__
+#include <linux/compat.h>
+#include <linux/err.h>
+#endif
 #include "ubifs.h"
-
-/*
- * Replay flags.
- *
- * REPLAY_DELETION: node was deleted
- * REPLAY_REF: node is a reference node
- */
-enum {
-       REPLAY_DELETION = 1,
-       REPLAY_REF = 2,
-};
+#include <linux/list_sort.h>
 
 /**
- * struct replay_entry - replay tree entry.
+ * struct replay_entry - replay list entry.
  * @lnum: logical eraseblock number of the node
  * @offs: node offset
  * @len: node length
+ * @deletion: non-zero if this entry corresponds to a node deletion
  * @sqnum: node sequence number
- * @flags: replay flags
- * @rb: links the replay tree
+ * @list: links the replay list
  * @key: node key
  * @nm: directory entry name
  * @old_size: truncation old size
  * @new_size: truncation new size
- * @free: amount of free space in a bud
- * @dirty: amount of dirty space in a bud from padding and deletion nodes
  *
- * UBIFS journal replay must compare node sequence numbers, which means it must
- * build a tree of node information to insert into the TNC.
+ * The replay process first scans all buds and builds the replay list, then
+ * sorts the replay list in nodes sequence number order, and then inserts all
+ * the replay entries to the TNC.
  */
 struct replay_entry {
        int lnum;
        int offs;
        int len;
+       unsigned int deletion:1;
        unsigned long long sqnum;
-       int flags;
-       struct rb_node rb;
+       struct list_head list;
        union ubifs_key key;
        union {
                struct qstr nm;
@@ -77,10 +60,6 @@ struct replay_entry {
                        loff_t old_size;
                        loff_t new_size;
                };
-               struct {
-                       int free;
-                       int dirty;
-               };
        };
 };
 
@@ -88,82 +67,116 @@ struct replay_entry {
  * struct bud_entry - entry in the list of buds to replay.
  * @list: next bud in the list
  * @bud: bud description object
- * @free: free bytes in the bud
  * @sqnum: reference node sequence number
+ * @free: free bytes in the bud
+ * @dirty: dirty bytes in the bud
  */
 struct bud_entry {
        struct list_head list;
        struct ubifs_bud *bud;
-       int free;
        unsigned long long sqnum;
+       int free;
+       int dirty;
 };
 
+#ifndef __UBOOT__
 /**
  * set_bud_lprops - set free and dirty space used by a bud.
  * @c: UBIFS file-system description object
- * @r: replay entry of bud
+ * @b: bud entry which describes the bud
+ *
+ * This function makes sure the LEB properties of bud @b are set correctly
+ * after the replay. Returns zero in case of success and a negative error code
+ * in case of failure.
  */
-static int set_bud_lprops(struct ubifs_info *c, struct replay_entry *r)
+static int set_bud_lprops(struct ubifs_info *c, struct bud_entry *b)
 {
        const struct ubifs_lprops *lp;
        int err = 0, dirty;
 
        ubifs_get_lprops(c);
 
-       lp = ubifs_lpt_lookup_dirty(c, r->lnum);
+       lp = ubifs_lpt_lookup_dirty(c, b->bud->lnum);
        if (IS_ERR(lp)) {
                err = PTR_ERR(lp);
                goto out;
        }
 
        dirty = lp->dirty;
-       if (r->offs == 0 && (lp->free != c->leb_size || lp->dirty != 0)) {
+       if (b->bud->start == 0 && (lp->free != c->leb_size || lp->dirty != 0)) {
                /*
                 * The LEB was added to the journal with a starting offset of
                 * zero which means the LEB must have been empty. The LEB
-                * property values should be lp->free == c->leb_size and
-                * lp->dirty == 0, but that is not the case. The reason is that
-                * the LEB was garbage collected. The garbage collector resets
-                * the free and dirty space without recording it anywhere except
-                * lprops, so if there is not a commit then lprops does not have
-                * that information next time the file system is mounted.
+                * property values should be @lp->free == @c->leb_size and
+                * @lp->dirty == 0, but that is not the case. The reason is that
+                * the LEB had been garbage collected before it became the bud,
+                * and there was not commit inbetween. The garbage collector
+                * resets the free and dirty space without recording it
+                * anywhere except lprops, so if there was no commit then
+                * lprops does not have that information.
                 *
                 * We do not need to adjust free space because the scan has told
                 * us the exact value which is recorded in the replay entry as
-                * r->free.
+                * @b->free.
                 *
                 * However we do need to subtract from the dirty space the
                 * amount of space that the garbage collector reclaimed, which
                 * is the whole LEB minus the amount of space that was free.
                 */
-               dbg_mnt("bud LEB %d was GC'd (%d free, %d dirty)", r->lnum,
+               dbg_mnt("bud LEB %d was GC'd (%d free, %d dirty)", b->bud->lnum,
                        lp->free, lp->dirty);
-               dbg_gc("bud LEB %d was GC'd (%d free, %d dirty)", r->lnum,
+               dbg_gc("bud LEB %d was GC'd (%d free, %d dirty)", b->bud->lnum,
                        lp->free, lp->dirty);
                dirty -= c->leb_size - lp->free;
                /*
                 * If the replay order was perfect the dirty space would now be
-                * zero. The order is not perfect because the the journal heads
+                * zero. The order is not perfect because the journal heads
                 * race with each other. This is not a problem but is does mean
                 * that the dirty space may temporarily exceed c->leb_size
                 * during the replay.
                 */
                if (dirty != 0)
-                       dbg_msg("LEB %d lp: %d free %d dirty "
-                               "replay: %d free %d dirty", r->lnum, lp->free,
-                               lp->dirty, r->free, r->dirty);
+                       dbg_mnt("LEB %d lp: %d free %d dirty replay: %d free %d dirty",
+                               b->bud->lnum, lp->free, lp->dirty, b->free,
+                               b->dirty);
        }
-       lp = ubifs_change_lp(c, lp, r->free, dirty + r->dirty,
+       lp = ubifs_change_lp(c, lp, b->free, dirty + b->dirty,
                             lp->flags | LPROPS_TAKEN, 0);
        if (IS_ERR(lp)) {
                err = PTR_ERR(lp);
                goto out;
        }
+
+       /* Make sure the journal head points to the latest bud */
+       err = ubifs_wbuf_seek_nolock(&c->jheads[b->bud->jhead].wbuf,
+                                    b->bud->lnum, c->leb_size - b->free);
+
 out:
        ubifs_release_lprops(c);
        return err;
 }
 
+/**
+ * set_buds_lprops - set free and dirty space for all replayed buds.
+ * @c: UBIFS file-system description object
+ *
+ * This function sets LEB properties for all replayed buds. Returns zero in
+ * case of success and a negative error code in case of failure.
+ */
+static int set_buds_lprops(struct ubifs_info *c)
+{
+       struct bud_entry *b;
+       int err;
+
+       list_for_each_entry(b, &c->replay_buds, list) {
+               err = set_bud_lprops(c, b);
+               if (err)
+                       return err;
+       }
+
+       return 0;
+}
+
 /**
  * trun_remove_range - apply a replay entry for a truncation to the TNC.
  * @c: UBIFS file-system description object
@@ -200,24 +213,22 @@ static int trun_remove_range(struct ubifs_info *c, struct replay_entry *r)
  */
 static int apply_replay_entry(struct ubifs_info *c, struct replay_entry *r)
 {
-       int err, deletion = ((r->flags & REPLAY_DELETION) != 0);
+       int err;
 
-       dbg_mnt("LEB %d:%d len %d flgs %d sqnum %llu %s", r->lnum,
-               r->offs, r->len, r->flags, r->sqnum, DBGKEY(&r->key));
+       dbg_mntk(&r->key, "LEB %d:%d len %d deletion %d sqnum %llu key ",
+                r->lnum, r->offs, r->len, r->deletion, r->sqnum);
 
        /* Set c->replay_sqnum to help deal with dangling branches. */
        c->replay_sqnum = r->sqnum;
 
-       if (r->flags & REPLAY_REF)
-               err = set_bud_lprops(c, r);
-       else if (is_hash_key(c, &r->key)) {
-               if (deletion)
+       if (is_hash_key(c, &r->key)) {
+               if (r->deletion)
                        err = ubifs_tnc_remove_nm(c, &r->key, &r->nm);
                else
                        err = ubifs_tnc_add_nm(c, &r->key, r->lnum, r->offs,
                                               r->len, &r->nm);
        } else {
-               if (deletion)
+               if (r->deletion)
                        switch (key_type(c, &r->key)) {
                        case UBIFS_INO_KEY:
                        {
@@ -240,7 +251,7 @@ static int apply_replay_entry(struct ubifs_info *c, struct replay_entry *r)
                        return err;
 
                if (c->need_recovery)
-                       err = ubifs_recover_size_accum(c, &r->key, deletion,
+                       err = ubifs_recover_size_accum(c, &r->key, r->deletion,
                                                       r->new_size);
        }
 
@@ -248,68 +259,77 @@ static int apply_replay_entry(struct ubifs_info *c, struct replay_entry *r)
 }
 
 /**
- * destroy_replay_tree - destroy the replay.
- * @c: UBIFS file-system description object
+ * replay_entries_cmp - compare 2 replay entries.
+ * @priv: UBIFS file-system description object
+ * @a: first replay entry
+ * @a: second replay entry
  *
- * Destroy the replay tree.
+ * This is a comparios function for 'list_sort()' which compares 2 replay
+ * entries @a and @b by comparing their sequence numer.  Returns %1 if @a has
+ * greater sequence number and %-1 otherwise.
  */
-static void destroy_replay_tree(struct ubifs_info *c)
+static int replay_entries_cmp(void *priv, struct list_head *a,
+                             struct list_head *b)
 {
-       struct rb_node *this = c->replay_tree.rb_node;
-       struct replay_entry *r;
-
-       while (this) {
-               if (this->rb_left) {
-                       this = this->rb_left;
-                       continue;
-               } else if (this->rb_right) {
-                       this = this->rb_right;
-                       continue;
-               }
-               r = rb_entry(this, struct replay_entry, rb);
-               this = rb_parent(this);
-               if (this) {
-                       if (this->rb_left == &r->rb)
-                               this->rb_left = NULL;
-                       else
-                               this->rb_right = NULL;
-               }
-               if (is_hash_key(c, &r->key))
-                       kfree((void *)r->nm.name);
-               kfree(r);
-       }
-       c->replay_tree = RB_ROOT;
+       struct replay_entry *ra, *rb;
+
+       cond_resched();
+       if (a == b)
+               return 0;
+
+       ra = list_entry(a, struct replay_entry, list);
+       rb = list_entry(b, struct replay_entry, list);
+       ubifs_assert(ra->sqnum != rb->sqnum);
+       if (ra->sqnum > rb->sqnum)
+               return 1;
+       return -1;
 }
 
 /**
- * apply_replay_tree - apply the replay tree to the TNC.
+ * apply_replay_list - apply the replay list to the TNC.
  * @c: UBIFS file-system description object
  *
- * Apply the replay tree.
- * Returns zero in case of success and a negative error code in case of
- * failure.
+ * Apply all entries in the replay list to the TNC. Returns zero in case of
+ * success and a negative error code in case of failure.
  */
-static int apply_replay_tree(struct ubifs_info *c)
+static int apply_replay_list(struct ubifs_info *c)
 {
-       struct rb_node *this = rb_first(&c->replay_tree);
+       struct replay_entry *r;
+       int err;
 
-       while (this) {
-               struct replay_entry *r;
-               int err;
+       list_sort(c, &c->replay_list, &replay_entries_cmp);
 
+       list_for_each_entry(r, &c->replay_list, list) {
                cond_resched();
 
-               r = rb_entry(this, struct replay_entry, rb);
                err = apply_replay_entry(c, r);
                if (err)
                        return err;
-               this = rb_next(this);
        }
+
        return 0;
 }
 
 /**
- * insert_node - insert a node to the replay tree.
+ * destroy_replay_list - destroy the replay.
+ * @c: UBIFS file-system description object
+ *
+ * Destroy the replay list.
+ */
+static void destroy_replay_list(struct ubifs_info *c)
+{
+       struct replay_entry *r, *tmp;
+
+       list_for_each_entry_safe(r, tmp, &c->replay_list, list) {
+               if (is_hash_key(c, &r->key))
+                       kfree(r->nm.name);
+               list_del(&r->list);
+               kfree(r);
+       }
+}
+
+/**
+ * insert_node - insert a node to the replay list
  * @c: UBIFS file-system description object
  * @lnum: node logical eraseblock number
  * @offs: node offset
@@ -321,39 +341,25 @@ static int apply_replay_tree(struct ubifs_info *c)
  * @old_size: truncation old size
  * @new_size: truncation new size
  *
- * This function inserts a scanned non-direntry node to the replay tree. The
- * replay tree is an RB-tree containing @struct replay_entry elements which are
- * indexed by the sequence number. The replay tree is applied at the very end
- * of the replay process. Since the tree is sorted in sequence number order,
- * the older modifications are applied first. This function returns zero in
- * case of success and a negative error code in case of failure.
+ * This function inserts a scanned non-direntry node to the replay list. The
+ * replay list contains @struct replay_entry elements, and we sort this list in
+ * sequence number order before applying it. The replay list is applied at the
+ * very end of the replay process. Since the list is sorted in sequence number
+ * order, the older modifications are applied first. This function returns zero
+ * in case of success and a negative error code in case of failure.
  */
 static int insert_node(struct ubifs_info *c, int lnum, int offs, int len,
                       union ubifs_key *key, unsigned long long sqnum,
                       int deletion, int *used, loff_t old_size,
                       loff_t new_size)
 {
-       struct rb_node **p = &c->replay_tree.rb_node, *parent = NULL;
        struct replay_entry *r;
 
+       dbg_mntk(key, "add LEB %d:%d, key ", lnum, offs);
+
        if (key_inum(c, key) >= c->highest_inum)
                c->highest_inum = key_inum(c, key);
 
-       dbg_mnt("add LEB %d:%d, key %s", lnum, offs, DBGKEY(key));
-       while (*p) {
-               parent = *p;
-               r = rb_entry(parent, struct replay_entry, rb);
-               if (sqnum < r->sqnum) {
-                       p = &(*p)->rb_left;
-                       continue;
-               } else if (sqnum > r->sqnum) {
-                       p = &(*p)->rb_right;
-                       continue;
-               }
-               ubifs_err("duplicate sqnum in replay");
-               return -EINVAL;
-       }
-
        r = kzalloc(sizeof(struct replay_entry), GFP_KERNEL);
        if (!r)
                return -ENOMEM;
@@ -363,19 +369,18 @@ static int insert_node(struct ubifs_info *c, int lnum, int offs, int len,
        r->lnum = lnum;
        r->offs = offs;
        r->len = len;
+       r->deletion = !!deletion;
        r->sqnum = sqnum;
-       r->flags = (deletion ? REPLAY_DELETION : 0);
+       key_copy(c, key, &r->key);
        r->old_size = old_size;
        r->new_size = new_size;
-       key_copy(c, key, &r->key);
 
-       rb_link_node(&r->rb, parent, p);
-       rb_insert_color(&r->rb, &c->replay_tree);
+       list_add_tail(&r->list, &c->replay_list);
        return 0;
 }
 
 /**
- * insert_dent - insert a directory entry node into the replay tree.
+ * insert_dent - insert a directory entry node into the replay list.
  * @c: UBIFS file-system description object
  * @lnum: node logical eraseblock number
  * @offs: node offset
@@ -387,43 +392,25 @@ static int insert_node(struct ubifs_info *c, int lnum, int offs, int len,
  * @deletion: non-zero if this is a deletion
  * @used: number of bytes in use in a LEB
  *
- * This function inserts a scanned directory entry node to the replay tree.
- * Returns zero in case of success and a negative error code in case of
- * failure.
- *
- * This function is also used for extended attribute entries because they are
- * implemented as directory entry nodes.
+ * This function inserts a scanned directory entry node or an extended
+ * attribute entry to the replay list. Returns zero in case of success and a
+ * negative error code in case of failure.
  */
 static int insert_dent(struct ubifs_info *c, int lnum, int offs, int len,
                       union ubifs_key *key, const char *name, int nlen,
                       unsigned long long sqnum, int deletion, int *used)
 {
-       struct rb_node **p = &c->replay_tree.rb_node, *parent = NULL;
        struct replay_entry *r;
        char *nbuf;
 
+       dbg_mntk(key, "add LEB %d:%d, key ", lnum, offs);
        if (key_inum(c, key) >= c->highest_inum)
                c->highest_inum = key_inum(c, key);
 
-       dbg_mnt("add LEB %d:%d, key %s", lnum, offs, DBGKEY(key));
-       while (*p) {
-               parent = *p;
-               r = rb_entry(parent, struct replay_entry, rb);
-               if (sqnum < r->sqnum) {
-                       p = &(*p)->rb_left;
-                       continue;
-               }
-               if (sqnum > r->sqnum) {
-                       p = &(*p)->rb_right;
-                       continue;
-               }
-               ubifs_err("duplicate sqnum in replay");
-               return -EINVAL;
-       }
-
        r = kzalloc(sizeof(struct replay_entry), GFP_KERNEL);
        if (!r)
                return -ENOMEM;
+
        nbuf = kmalloc(nlen + 1, GFP_KERNEL);
        if (!nbuf) {
                kfree(r);
@@ -435,19 +422,18 @@ static int insert_dent(struct ubifs_info *c, int lnum, int offs, int len,
        r->lnum = lnum;
        r->offs = offs;
        r->len = len;
+       r->deletion = !!deletion;
        r->sqnum = sqnum;
+       key_copy(c, key, &r->key);
        r->nm.len = nlen;
        memcpy(nbuf, name, nlen);
        nbuf[nlen] = '\0';
        r->nm.name = nbuf;
-       r->flags = (deletion ? REPLAY_DELETION : 0);
-       key_copy(c, key, &r->key);
 
-       ubifs_assert(!*p);
-       rb_link_node(&r->rb, parent, p);
-       rb_insert_color(&r->rb, &c->replay_tree);
+       list_add_tail(&r->list, &c->replay_list);
        return 0;
 }
+#endif
 
 /**
  * ubifs_validate_entry - validate directory or extended attribute entry node.
@@ -466,7 +452,7 @@ int ubifs_validate_entry(struct ubifs_info *c,
        if (le32_to_cpu(dent->ch.len) != nlen + UBIFS_DENT_NODE_SZ + 1 ||
            dent->type >= UBIFS_ITYPES_CNT ||
            nlen > UBIFS_MAX_NLEN || dent->name[nlen] != 0 ||
-           strnlen((char *)dent->name, nlen) != nlen ||
+           strnlen(dent->name, nlen) != nlen ||
            le64_to_cpu(dent->inum) > MAX_INUM) {
                ubifs_err("bad %s node", key_type == UBIFS_DENT_KEY ?
                          "directory entry" : "extended attribute entry");
@@ -481,32 +467,94 @@ int ubifs_validate_entry(struct ubifs_info *c,
        return 0;
 }
 
+#ifndef __UBOOT__
+/**
+ * is_last_bud - check if the bud is the last in the journal head.
+ * @c: UBIFS file-system description object
+ * @bud: bud description object
+ *
+ * This function checks if bud @bud is the last bud in its journal head. This
+ * information is then used by 'replay_bud()' to decide whether the bud can
+ * have corruptions or not. Indeed, only last buds can be corrupted by power
+ * cuts. Returns %1 if this is the last bud, and %0 if not.
+ */
+static int is_last_bud(struct ubifs_info *c, struct ubifs_bud *bud)
+{
+       struct ubifs_jhead *jh = &c->jheads[bud->jhead];
+       struct ubifs_bud *next;
+       uint32_t data;
+       int err;
+
+       if (list_is_last(&bud->list, &jh->buds_list))
+               return 1;
+
+       /*
+        * The following is a quirk to make sure we work correctly with UBIFS
+        * images used with older UBIFS.
+        *
+        * Normally, the last bud will be the last in the journal head's list
+        * of bud. However, there is one exception if the UBIFS image belongs
+        * to older UBIFS. This is fairly unlikely: one would need to use old
+        * UBIFS, then have a power cut exactly at the right point, and then
+        * try to mount this image with new UBIFS.
+        *
+        * The exception is: it is possible to have 2 buds A and B, A goes
+        * before B, and B is the last, bud B is contains no data, and bud A is
+        * corrupted at the end. The reason is that in older versions when the
+        * journal code switched the next bud (from A to B), it first added a
+        * log reference node for the new bud (B), and only after this it
+        * synchronized the write-buffer of current bud (A). But later this was
+        * changed and UBIFS started to always synchronize the write-buffer of
+        * the bud (A) before writing the log reference for the new bud (B).
+        *
+        * But because older UBIFS always synchronized A's write-buffer before
+        * writing to B, we can recognize this exceptional situation but
+        * checking the contents of bud B - if it is empty, then A can be
+        * treated as the last and we can recover it.
+        *
+        * TODO: remove this piece of code in a couple of years (today it is
+        * 16.05.2011).
+        */
+       next = list_entry(bud->list.next, struct ubifs_bud, list);
+       if (!list_is_last(&next->list, &jh->buds_list))
+               return 0;
+
+       err = ubifs_leb_read(c, next->lnum, (char *)&data, next->start, 4, 1);
+       if (err)
+               return 0;
+
+       return data == 0xFFFFFFFF;
+}
+
 /**
  * replay_bud - replay a bud logical eraseblock.
  * @c: UBIFS file-system description object
- * @lnum: bud logical eraseblock number to replay
- * @offs: bud start offset
- * @jhead: journal head to which this bud belongs
- * @free: amount of free space in the bud is returned here
- * @dirty: amount of dirty space from padding and deletion nodes is returned
- * here
+ * @b: bud entry which describes the bud
  *
- * This function returns zero in case of success and a negative error code in
- * case of failure.
+ * This function replays bud @bud, recovers it if needed, and adds all nodes
+ * from this bud to the replay list. Returns zero in case of success and a
+ * negative error code in case of failure.
  */
-static int replay_bud(struct ubifs_info *c, int lnum, int offs, int jhead,
-                     int *free, int *dirty)
+static int replay_bud(struct ubifs_info *c, struct bud_entry *b)
 {
-       int err = 0, used = 0;
+       int is_last = is_last_bud(c, b->bud);
+       int err = 0, used = 0, lnum = b->bud->lnum, offs = b->bud->start;
        struct ubifs_scan_leb *sleb;
        struct ubifs_scan_node *snod;
-       struct ubifs_bud *bud;
 
-       dbg_mnt("replay bud LEB %d, head %d", lnum, jhead);
-       if (c->need_recovery)
-               sleb = ubifs_recover_leb(c, lnum, offs, c->sbuf, jhead != GCHD);
+       dbg_mnt("replay bud LEB %d, head %d, offs %d, is_last %d",
+               lnum, b->bud->jhead, offs, is_last);
+
+       if (c->need_recovery && is_last)
+               /*
+                * Recover only last LEBs in the journal heads, because power
+                * cuts may cause corruptions only in these LEBs, because only
+                * these LEBs could possibly be written to at the power cut
+                * time.
+                */
+               sleb = ubifs_recover_leb(c, lnum, offs, c->sbuf, b->bud->jhead);
        else
-               sleb = ubifs_scan(c, lnum, offs, c->sbuf);
+               sleb = ubifs_scan(c, lnum, offs, c->sbuf, 0);
        if (IS_ERR(sleb))
                return PTR_ERR(sleb);
 
@@ -580,7 +628,7 @@ static int replay_bud(struct ubifs_info *c, int lnum, int offs, int jhead,
                                goto out_dump;
 
                        err = insert_dent(c, lnum, snod->offs, snod->len,
-                                         &snod->key, (char *)dent->name,
+                                         &snod->key, dent->name,
                                          le16_to_cpu(dent->nlen), snod->sqnum,
                                          !le64_to_cpu(dent->inum), &used);
                        break;
@@ -620,15 +668,14 @@ static int replay_bud(struct ubifs_info *c, int lnum, int offs, int jhead,
                        goto out;
        }
 
-       bud = ubifs_search_bud(c, lnum);
-       if (!bud)
-               BUG();
-
+       ubifs_assert(ubifs_search_bud(c, lnum));
        ubifs_assert(sleb->endpt - offs >= used);
        ubifs_assert(sleb->endpt % c->min_io_size == 0);
 
-       *dirty = sleb->endpt - offs - used;
-       *free = c->leb_size - sleb->endpt;
+       b->dirty = sleb->endpt - offs - used;
+       b->free = c->leb_size - sleb->endpt;
+       dbg_mnt("bud LEB %d replied: dirty %d, free %d",
+               lnum, b->dirty, b->free);
 
 out:
        ubifs_scan_destroy(sleb);
@@ -636,60 +683,11 @@ out:
 
 out_dump:
        ubifs_err("bad node is at LEB %d:%d", lnum, snod->offs);
-       dbg_dump_node(c, snod->node);
+       ubifs_dump_node(c, snod->node);
        ubifs_scan_destroy(sleb);
        return -EINVAL;
 }
 
-/**
- * insert_ref_node - insert a reference node to the replay tree.
- * @c: UBIFS file-system description object
- * @lnum: node logical eraseblock number
- * @offs: node offset
- * @sqnum: sequence number
- * @free: amount of free space in bud
- * @dirty: amount of dirty space from padding and deletion nodes
- *
- * This function inserts a reference node to the replay tree and returns zero
- * in case of success or a negative error code in case of failure.
- */
-static int insert_ref_node(struct ubifs_info *c, int lnum, int offs,
-                          unsigned long long sqnum, int free, int dirty)
-{
-       struct rb_node **p = &c->replay_tree.rb_node, *parent = NULL;
-       struct replay_entry *r;
-
-       dbg_mnt("add ref LEB %d:%d", lnum, offs);
-       while (*p) {
-               parent = *p;
-               r = rb_entry(parent, struct replay_entry, rb);
-               if (sqnum < r->sqnum) {
-                       p = &(*p)->rb_left;
-                       continue;
-               } else if (sqnum > r->sqnum) {
-                       p = &(*p)->rb_right;
-                       continue;
-               }
-               ubifs_err("duplicate sqnum in replay tree");
-               return -EINVAL;
-       }
-
-       r = kzalloc(sizeof(struct replay_entry), GFP_KERNEL);
-       if (!r)
-               return -ENOMEM;
-
-       r->lnum = lnum;
-       r->offs = offs;
-       r->sqnum = sqnum;
-       r->flags = REPLAY_REF;
-       r->free = free;
-       r->dirty = dirty;
-
-       rb_link_node(&r->rb, parent, p);
-       rb_insert_color(&r->rb, &c->replay_tree);
-       return 0;
-}
-
 /**
  * replay_buds - replay all buds.
  * @c: UBIFS file-system description object
@@ -700,17 +698,16 @@ static int insert_ref_node(struct ubifs_info *c, int lnum, int offs,
 static int replay_buds(struct ubifs_info *c)
 {
        struct bud_entry *b;
-       int err, uninitialized_var(free), uninitialized_var(dirty);
+       int err;
+       unsigned long long prev_sqnum = 0;
 
        list_for_each_entry(b, &c->replay_buds, list) {
-               err = replay_bud(c, b->bud->lnum, b->bud->start, b->bud->jhead,
-                                &free, &dirty);
-               if (err)
-                       return err;
-               err = insert_ref_node(c, b->bud->lnum, b->bud->start, b->sqnum,
-                                     free, dirty);
+               err = replay_bud(c, b);
                if (err)
                        return err;
+
+               ubifs_assert(b->sqnum > prev_sqnum);
+               prev_sqnum = b->sqnum;
        }
 
        return 0;
@@ -831,10 +828,16 @@ static int replay_log_leb(struct ubifs_info *c, int lnum, int offs, void *sbuf)
        const struct ubifs_cs_node *node;
 
        dbg_mnt("replay log LEB %d:%d", lnum, offs);
-       sleb = ubifs_scan(c, lnum, offs, sbuf);
+       sleb = ubifs_scan(c, lnum, offs, sbuf, c->need_recovery);
        if (IS_ERR(sleb)) {
-               if (c->need_recovery)
-                       sleb = ubifs_recover_log_leb(c, lnum, offs, sbuf);
+               if (PTR_ERR(sleb) != -EUCLEAN || !c->need_recovery)
+                       return PTR_ERR(sleb);
+               /*
+                * Note, the below function will recover this log LEB only if
+                * it is the last, because unclean reboots can possibly corrupt
+                * only the tail of the log.
+                */
+               sleb = ubifs_recover_log_leb(c, lnum, offs, sbuf);
                if (IS_ERR(sleb))
                        return PTR_ERR(sleb);
        }
@@ -845,7 +848,6 @@ static int replay_log_leb(struct ubifs_info *c, int lnum, int offs, void *sbuf)
        }
 
        node = sleb->buf;
-
        snod = list_entry(sleb->nodes.next, struct ubifs_scan_node, list);
        if (c->cs_sqnum == 0) {
                /*
@@ -856,16 +858,15 @@ static int replay_log_leb(struct ubifs_info *c, int lnum, int offs, void *sbuf)
                 * numbers.
                 */
                if (snod->type != UBIFS_CS_NODE) {
-                       dbg_err("first log node at LEB %d:%d is not CS node",
-                               lnum, offs);
+                       ubifs_err("first log node at LEB %d:%d is not CS node",
+                                 lnum, offs);
                        goto out_dump;
                }
                if (le64_to_cpu(node->cmt_no) != c->cmt_no) {
-                       dbg_err("first CS node at LEB %d:%d has wrong "
-                               "commit number %llu expected %llu",
-                               lnum, offs,
-                               (unsigned long long)le64_to_cpu(node->cmt_no),
-                               c->cmt_no);
+                       ubifs_err("first CS node at LEB %d:%d has wrong commit number %llu expected %llu",
+                                 lnum, offs,
+                                 (unsigned long long)le64_to_cpu(node->cmt_no),
+                                 c->cmt_no);
                        goto out_dump;
                }
 
@@ -887,12 +888,11 @@ static int replay_log_leb(struct ubifs_info *c, int lnum, int offs, void *sbuf)
 
        /* Make sure the first node sits at offset zero of the LEB */
        if (snod->offs != 0) {
-               dbg_err("first node is not at zero offset");
+               ubifs_err("first node is not at zero offset");
                goto out_dump;
        }
 
        list_for_each_entry(snod, &sleb->nodes, list) {
-
                cond_resched();
 
                if (snod->sqnum >= SQNUM_WATERMARK) {
@@ -901,8 +901,8 @@ static int replay_log_leb(struct ubifs_info *c, int lnum, int offs, void *sbuf)
                }
 
                if (snod->sqnum < c->cs_sqnum) {
-                       dbg_err("bad sqnum %llu, commit sqnum %llu",
-                               snod->sqnum, c->cs_sqnum);
+                       ubifs_err("bad sqnum %llu, commit sqnum %llu",
+                                 snod->sqnum, c->cs_sqnum);
                        goto out_dump;
                }
 
@@ -952,9 +952,9 @@ out:
        return err;
 
 out_dump:
-       ubifs_err("log error detected while replying the log at LEB %d:%d",
+       ubifs_err("log error detected while replaying the log at LEB %d:%d",
                  lnum, offs + snod->offs);
-       dbg_dump_node(c, snod->node);
+       ubifs_dump_node(c, snod->node);
        ubifs_scan_destroy(sleb);
        return -EINVAL;
 }
@@ -1004,67 +1004,64 @@ out:
  */
 int ubifs_replay_journal(struct ubifs_info *c)
 {
-       int err, i, lnum, offs, _free;
-       void *sbuf = NULL;
+       int err, lnum, free;
 
        BUILD_BUG_ON(UBIFS_TRUN_KEY > 5);
 
        /* Update the status of the index head in lprops to 'taken' */
-       _free = take_ihead(c);
-       if (_free < 0)
-               return _free; /* Error code */
+       free = take_ihead(c);
+       if (free < 0)
+               return free; /* Error code */
 
-       if (c->ihead_offs != c->leb_size - _free) {
+       if (c->ihead_offs != c->leb_size - free) {
                ubifs_err("bad index head LEB %d:%d", c->ihead_lnum,
                          c->ihead_offs);
                return -EINVAL;
        }
 
-       sbuf = vmalloc(c->leb_size);
-       if (!sbuf)
-               return -ENOMEM;
-
        dbg_mnt("start replaying the journal");
-
        c->replaying = 1;
-
        lnum = c->ltail_lnum = c->lhead_lnum;
-       offs = c->lhead_offs;
 
-       for (i = 0; i < c->log_lebs; i++, lnum++) {
-               if (lnum >= UBIFS_LOG_LNUM + c->log_lebs) {
-                       /*
-                        * The log is logically circular, we reached the last
-                        * LEB, switch to the first one.
-                        */
-                       lnum = UBIFS_LOG_LNUM;
-                       offs = 0;
-               }
-               err = replay_log_leb(c, lnum, offs, sbuf);
+       do {
+               err = replay_log_leb(c, lnum, 0, c->sbuf);
                if (err == 1)
                        /* We hit the end of the log */
                        break;
                if (err)
                        goto out;
-               offs = 0;
-       }
+               lnum = ubifs_next_log_lnum(c, lnum);
+       } while (lnum != c->ltail_lnum);
 
        err = replay_buds(c);
        if (err)
                goto out;
 
-       err = apply_replay_tree(c);
+       err = apply_replay_list(c);
        if (err)
                goto out;
 
+       err = set_buds_lprops(c);
+       if (err)
+               goto out;
+
+       /*
+        * UBIFS budgeting calculations use @c->bi.uncommitted_idx variable
+        * to roughly estimate index growth. Things like @c->bi.min_idx_lebs
+        * depend on it. This means we have to initialize it to make sure
+        * budgeting works properly.
+        */
+       c->bi.uncommitted_idx = atomic_long_read(&c->dirty_zn_cnt);
+       c->bi.uncommitted_idx *= c->max_idx_node_sz;
+
        ubifs_assert(c->bud_bytes <= c->max_bud_bytes || c->need_recovery);
-       dbg_mnt("finished, log head LEB %d:%d, max_sqnum %llu, "
-               "highest_inum %lu", c->lhead_lnum, c->lhead_offs, c->max_sqnum,
+       dbg_mnt("finished, log head LEB %d:%d, max_sqnum %llu, highest_inum %lu",
+               c->lhead_lnum, c->lhead_offs, c->max_sqnum,
                (unsigned long)c->highest_inum);
 out:
-       destroy_replay_tree(c);
+       destroy_replay_list(c);
        destroy_bud_list(c);
-       vfree(sbuf);
        c->replaying = 0;
        return err;
 }
+#endif
index 00c9cd31a017e7520b6181d54b9ba72529674853..fc0194aa267023866ebef30eecffb74a422bb14e 100644 (file)
@@ -3,18 +3,7 @@
  *
  * Copyright (C) 2006-2008 Nokia Corporation.
  *
- * This program is free software; you can redistribute it and/or modify it
- * under the terms of the GNU General Public License version 2 as published by
- * the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
- * more details.
- *
- * You should have received a copy of the GNU General Public License along with
- * this program; if not, write to the Free Software Foundation, Inc., 51
- * Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
+ * SPDX-License-Identifier:    GPL-2.0+
  *
  * Authors: Artem Bityutskiy (Битюцкий Артём)
  *          Adrian Hunter
  */
 
 #include "ubifs.h"
+#define __UBOOT__
+#ifndef __UBOOT__
+#include <linux/slab.h>
+#include <linux/random.h>
+#include <linux/math64.h>
+#else
+
+#include <linux/compat.h>
+#include <linux/err.h>
+#include <ubi_uboot.h>
+#include <linux/stat.h>
+#endif
 
 /*
  * Default journal size in logical eraseblocks as a percent of total
 /* Default time granularity in nanoseconds */
 #define DEFAULT_TIME_GRAN 1000000000
 
+#ifndef __UBOOT__
+/**
+ * create_default_filesystem - format empty UBI volume.
+ * @c: UBIFS file-system description object
+ *
+ * This function creates default empty file-system. Returns zero in case of
+ * success and a negative error code in case of failure.
+ */
+static int create_default_filesystem(struct ubifs_info *c)
+{
+       struct ubifs_sb_node *sup;
+       struct ubifs_mst_node *mst;
+       struct ubifs_idx_node *idx;
+       struct ubifs_branch *br;
+       struct ubifs_ino_node *ino;
+       struct ubifs_cs_node *cs;
+       union ubifs_key key;
+       int err, tmp, jnl_lebs, log_lebs, max_buds, main_lebs, main_first;
+       int lpt_lebs, lpt_first, orph_lebs, big_lpt, ino_waste, sup_flags = 0;
+       int min_leb_cnt = UBIFS_MIN_LEB_CNT;
+       long long tmp64, main_bytes;
+       __le64 tmp_le64;
+
+       /* Some functions called from here depend on the @c->key_len filed */
+       c->key_len = UBIFS_SK_LEN;
+
+       /*
+        * First of all, we have to calculate default file-system geometry -
+        * log size, journal size, etc.
+        */
+       if (c->leb_cnt < 0x7FFFFFFF / DEFAULT_JNL_PERCENT)
+               /* We can first multiply then divide and have no overflow */
+               jnl_lebs = c->leb_cnt * DEFAULT_JNL_PERCENT / 100;
+       else
+               jnl_lebs = (c->leb_cnt / 100) * DEFAULT_JNL_PERCENT;
+
+       if (jnl_lebs < UBIFS_MIN_JNL_LEBS)
+               jnl_lebs = UBIFS_MIN_JNL_LEBS;
+       if (jnl_lebs * c->leb_size > DEFAULT_MAX_JNL)
+               jnl_lebs = DEFAULT_MAX_JNL / c->leb_size;
+
+       /*
+        * The log should be large enough to fit reference nodes for all bud
+        * LEBs. Because buds do not have to start from the beginning of LEBs
+        * (half of the LEB may contain committed data), the log should
+        * generally be larger, make it twice as large.
+        */
+       tmp = 2 * (c->ref_node_alsz * jnl_lebs) + c->leb_size - 1;
+       log_lebs = tmp / c->leb_size;
+       /* Plus one LEB reserved for commit */
+       log_lebs += 1;
+       if (c->leb_cnt - min_leb_cnt > 8) {
+               /* And some extra space to allow writes while committing */
+               log_lebs += 1;
+               min_leb_cnt += 1;
+       }
+
+       max_buds = jnl_lebs - log_lebs;
+       if (max_buds < UBIFS_MIN_BUD_LEBS)
+               max_buds = UBIFS_MIN_BUD_LEBS;
+
+       /*
+        * Orphan nodes are stored in a separate area. One node can store a lot
+        * of orphan inode numbers, but when new orphan comes we just add a new
+        * orphan node. At some point the nodes are consolidated into one
+        * orphan node.
+        */
+       orph_lebs = UBIFS_MIN_ORPH_LEBS;
+       if (c->leb_cnt - min_leb_cnt > 1)
+               /*
+                * For debugging purposes it is better to have at least 2
+                * orphan LEBs, because the orphan subsystem would need to do
+                * consolidations and would be stressed more.
+                */
+               orph_lebs += 1;
+
+       main_lebs = c->leb_cnt - UBIFS_SB_LEBS - UBIFS_MST_LEBS - log_lebs;
+       main_lebs -= orph_lebs;
+
+       lpt_first = UBIFS_LOG_LNUM + log_lebs;
+       c->lsave_cnt = DEFAULT_LSAVE_CNT;
+       c->max_leb_cnt = c->leb_cnt;
+       err = ubifs_create_dflt_lpt(c, &main_lebs, lpt_first, &lpt_lebs,
+                                   &big_lpt);
+       if (err)
+               return err;
+
+       dbg_gen("LEB Properties Tree created (LEBs %d-%d)", lpt_first,
+               lpt_first + lpt_lebs - 1);
+
+       main_first = c->leb_cnt - main_lebs;
+
+       /* Create default superblock */
+       tmp = ALIGN(UBIFS_SB_NODE_SZ, c->min_io_size);
+       sup = kzalloc(tmp, GFP_KERNEL);
+       if (!sup)
+               return -ENOMEM;
+
+       tmp64 = (long long)max_buds * c->leb_size;
+       if (big_lpt)
+               sup_flags |= UBIFS_FLG_BIGLPT;
+
+       sup->ch.node_type  = UBIFS_SB_NODE;
+       sup->key_hash      = UBIFS_KEY_HASH_R5;
+       sup->flags         = cpu_to_le32(sup_flags);
+       sup->min_io_size   = cpu_to_le32(c->min_io_size);
+       sup->leb_size      = cpu_to_le32(c->leb_size);
+       sup->leb_cnt       = cpu_to_le32(c->leb_cnt);
+       sup->max_leb_cnt   = cpu_to_le32(c->max_leb_cnt);
+       sup->max_bud_bytes = cpu_to_le64(tmp64);
+       sup->log_lebs      = cpu_to_le32(log_lebs);
+       sup->lpt_lebs      = cpu_to_le32(lpt_lebs);
+       sup->orph_lebs     = cpu_to_le32(orph_lebs);
+       sup->jhead_cnt     = cpu_to_le32(DEFAULT_JHEADS_CNT);
+       sup->fanout        = cpu_to_le32(DEFAULT_FANOUT);
+       sup->lsave_cnt     = cpu_to_le32(c->lsave_cnt);
+       sup->fmt_version   = cpu_to_le32(UBIFS_FORMAT_VERSION);
+       sup->time_gran     = cpu_to_le32(DEFAULT_TIME_GRAN);
+       if (c->mount_opts.override_compr)
+               sup->default_compr = cpu_to_le16(c->mount_opts.compr_type);
+       else
+               sup->default_compr = cpu_to_le16(UBIFS_COMPR_LZO);
+
+       generate_random_uuid(sup->uuid);
+
+       main_bytes = (long long)main_lebs * c->leb_size;
+       tmp64 = div_u64(main_bytes * DEFAULT_RP_PERCENT, 100);
+       if (tmp64 > DEFAULT_MAX_RP_SIZE)
+               tmp64 = DEFAULT_MAX_RP_SIZE;
+       sup->rp_size = cpu_to_le64(tmp64);
+       sup->ro_compat_version = cpu_to_le32(UBIFS_RO_COMPAT_VERSION);
+
+       err = ubifs_write_node(c, sup, UBIFS_SB_NODE_SZ, 0, 0);
+       kfree(sup);
+       if (err)
+               return err;
+
+       dbg_gen("default superblock created at LEB 0:0");
+
+       /* Create default master node */
+       mst = kzalloc(c->mst_node_alsz, GFP_KERNEL);
+       if (!mst)
+               return -ENOMEM;
+
+       mst->ch.node_type = UBIFS_MST_NODE;
+       mst->log_lnum     = cpu_to_le32(UBIFS_LOG_LNUM);
+       mst->highest_inum = cpu_to_le64(UBIFS_FIRST_INO);
+       mst->cmt_no       = 0;
+       mst->root_lnum    = cpu_to_le32(main_first + DEFAULT_IDX_LEB);
+       mst->root_offs    = 0;
+       tmp = ubifs_idx_node_sz(c, 1);
+       mst->root_len     = cpu_to_le32(tmp);
+       mst->gc_lnum      = cpu_to_le32(main_first + DEFAULT_GC_LEB);
+       mst->ihead_lnum   = cpu_to_le32(main_first + DEFAULT_IDX_LEB);
+       mst->ihead_offs   = cpu_to_le32(ALIGN(tmp, c->min_io_size));
+       mst->index_size   = cpu_to_le64(ALIGN(tmp, 8));
+       mst->lpt_lnum     = cpu_to_le32(c->lpt_lnum);
+       mst->lpt_offs     = cpu_to_le32(c->lpt_offs);
+       mst->nhead_lnum   = cpu_to_le32(c->nhead_lnum);
+       mst->nhead_offs   = cpu_to_le32(c->nhead_offs);
+       mst->ltab_lnum    = cpu_to_le32(c->ltab_lnum);
+       mst->ltab_offs    = cpu_to_le32(c->ltab_offs);
+       mst->lsave_lnum   = cpu_to_le32(c->lsave_lnum);
+       mst->lsave_offs   = cpu_to_le32(c->lsave_offs);
+       mst->lscan_lnum   = cpu_to_le32(main_first);
+       mst->empty_lebs   = cpu_to_le32(main_lebs - 2);
+       mst->idx_lebs     = cpu_to_le32(1);
+       mst->leb_cnt      = cpu_to_le32(c->leb_cnt);
+
+       /* Calculate lprops statistics */
+       tmp64 = main_bytes;
+       tmp64 -= ALIGN(ubifs_idx_node_sz(c, 1), c->min_io_size);
+       tmp64 -= ALIGN(UBIFS_INO_NODE_SZ, c->min_io_size);
+       mst->total_free = cpu_to_le64(tmp64);
+
+       tmp64 = ALIGN(ubifs_idx_node_sz(c, 1), c->min_io_size);
+       ino_waste = ALIGN(UBIFS_INO_NODE_SZ, c->min_io_size) -
+                         UBIFS_INO_NODE_SZ;
+       tmp64 += ino_waste;
+       tmp64 -= ALIGN(ubifs_idx_node_sz(c, 1), 8);
+       mst->total_dirty = cpu_to_le64(tmp64);
+
+       /*  The indexing LEB does not contribute to dark space */
+       tmp64 = ((long long)(c->main_lebs - 1) * c->dark_wm);
+       mst->total_dark = cpu_to_le64(tmp64);
+
+       mst->total_used = cpu_to_le64(UBIFS_INO_NODE_SZ);
+
+       err = ubifs_write_node(c, mst, UBIFS_MST_NODE_SZ, UBIFS_MST_LNUM, 0);
+       if (err) {
+               kfree(mst);
+               return err;
+       }
+       err = ubifs_write_node(c, mst, UBIFS_MST_NODE_SZ, UBIFS_MST_LNUM + 1,
+                              0);
+       kfree(mst);
+       if (err)
+               return err;
+
+       dbg_gen("default master node created at LEB %d:0", UBIFS_MST_LNUM);
+
+       /* Create the root indexing node */
+       tmp = ubifs_idx_node_sz(c, 1);
+       idx = kzalloc(ALIGN(tmp, c->min_io_size), GFP_KERNEL);
+       if (!idx)
+               return -ENOMEM;
+
+       c->key_fmt = UBIFS_SIMPLE_KEY_FMT;
+       c->key_hash = key_r5_hash;
+
+       idx->ch.node_type = UBIFS_IDX_NODE;
+       idx->child_cnt = cpu_to_le16(1);
+       ino_key_init(c, &key, UBIFS_ROOT_INO);
+       br = ubifs_idx_branch(c, idx, 0);
+       key_write_idx(c, &key, &br->key);
+       br->lnum = cpu_to_le32(main_first + DEFAULT_DATA_LEB);
+       br->len  = cpu_to_le32(UBIFS_INO_NODE_SZ);
+       err = ubifs_write_node(c, idx, tmp, main_first + DEFAULT_IDX_LEB, 0);
+       kfree(idx);
+       if (err)
+               return err;
+
+       dbg_gen("default root indexing node created LEB %d:0",
+               main_first + DEFAULT_IDX_LEB);
+
+       /* Create default root inode */
+       tmp = ALIGN(UBIFS_INO_NODE_SZ, c->min_io_size);
+       ino = kzalloc(tmp, GFP_KERNEL);
+       if (!ino)
+               return -ENOMEM;
+
+       ino_key_init_flash(c, &ino->key, UBIFS_ROOT_INO);
+       ino->ch.node_type = UBIFS_INO_NODE;
+       ino->creat_sqnum = cpu_to_le64(++c->max_sqnum);
+       ino->nlink = cpu_to_le32(2);
+       tmp_le64 = cpu_to_le64(CURRENT_TIME_SEC.tv_sec);
+       ino->atime_sec   = tmp_le64;
+       ino->ctime_sec   = tmp_le64;
+       ino->mtime_sec   = tmp_le64;
+       ino->atime_nsec  = 0;
+       ino->ctime_nsec  = 0;
+       ino->mtime_nsec  = 0;
+       ino->mode = cpu_to_le32(S_IFDIR | S_IRUGO | S_IWUSR | S_IXUGO);
+       ino->size = cpu_to_le64(UBIFS_INO_NODE_SZ);
+
+       /* Set compression enabled by default */
+       ino->flags = cpu_to_le32(UBIFS_COMPR_FL);
+
+       err = ubifs_write_node(c, ino, UBIFS_INO_NODE_SZ,
+                              main_first + DEFAULT_DATA_LEB, 0);
+       kfree(ino);
+       if (err)
+               return err;
+
+       dbg_gen("root inode created at LEB %d:0",
+               main_first + DEFAULT_DATA_LEB);
+
+       /*
+        * The first node in the log has to be the commit start node. This is
+        * always the case during normal file-system operation. Write a fake
+        * commit start node to the log.
+        */
+       tmp = ALIGN(UBIFS_CS_NODE_SZ, c->min_io_size);
+       cs = kzalloc(tmp, GFP_KERNEL);
+       if (!cs)
+               return -ENOMEM;
+
+       cs->ch.node_type = UBIFS_CS_NODE;
+       err = ubifs_write_node(c, cs, UBIFS_CS_NODE_SZ, UBIFS_LOG_LNUM, 0);
+       kfree(cs);
+
+       ubifs_msg("default file-system created");
+       return 0;
+}
+#endif
+
 /**
  * validate_sb - validate superblock node.
  * @c: UBIFS file-system description object
@@ -114,9 +391,8 @@ static int validate_sb(struct ubifs_info *c, struct ubifs_sb_node *sup)
        min_leb_cnt += c->lpt_lebs + c->orph_lebs + c->jhead_cnt + 6;
 
        if (c->leb_cnt < min_leb_cnt || c->leb_cnt > c->vi.size) {
-               ubifs_err("bad LEB count: %d in superblock, %d on UBI volume, "
-                         "%d minimum required", c->leb_cnt, c->vi.size,
-                         min_leb_cnt);
+               ubifs_err("bad LEB count: %d in superblock, %d on UBI volume, %d minimum required",
+                         c->leb_cnt, c->vi.size, min_leb_cnt);
                goto failed;
        }
 
@@ -127,13 +403,22 @@ static int validate_sb(struct ubifs_info *c, struct ubifs_sb_node *sup)
        }
 
        if (c->main_lebs < UBIFS_MIN_MAIN_LEBS) {
-               err = 7;
+               ubifs_err("too few main LEBs count %d, must be at least %d",
+                         c->main_lebs, UBIFS_MIN_MAIN_LEBS);
+               goto failed;
+       }
+
+       max_bytes = (long long)c->leb_size * UBIFS_MIN_BUD_LEBS;
+       if (c->max_bud_bytes < max_bytes) {
+               ubifs_err("too small journal (%lld bytes), must be at least %lld bytes",
+                         c->max_bud_bytes, max_bytes);
                goto failed;
        }
 
-       if (c->max_bud_bytes < (long long)c->leb_size * UBIFS_MIN_BUD_LEBS ||
-           c->max_bud_bytes > (long long)c->leb_size * c->main_lebs) {
-               err = 8;
+       max_bytes = (long long)c->leb_size * c->main_lebs;
+       if (c->max_bud_bytes > max_bytes) {
+               ubifs_err("too large journal size (%lld bytes), only %lld bytes available in the main area",
+                         c->max_bud_bytes, max_bytes);
                goto failed;
        }
 
@@ -167,7 +452,6 @@ static int validate_sb(struct ubifs_info *c, struct ubifs_sb_node *sup)
                goto failed;
        }
 
-       max_bytes = c->main_lebs * (long long)c->leb_size;
        if (c->rp_size < 0 || max_bytes < c->rp_size) {
                err = 14;
                goto failed;
@@ -183,7 +467,7 @@ static int validate_sb(struct ubifs_info *c, struct ubifs_sb_node *sup)
 
 failed:
        ubifs_err("bad superblock, error %d", err);
-       dbg_dump_node(c, sup);
+       ubifs_dump_node(c, sup);
        return -EINVAL;
 }
 
@@ -192,7 +476,8 @@ failed:
  * @c: UBIFS file-system description object
  *
  * This function returns a pointer to the superblock node or a negative error
- * code.
+ * code. Note, the user of this function is responsible of kfree()'ing the
+ * returned superblock buffer.
  */
 struct ubifs_sb_node *ubifs_read_sb_node(struct ubifs_info *c)
 {
@@ -213,6 +498,21 @@ struct ubifs_sb_node *ubifs_read_sb_node(struct ubifs_info *c)
        return sup;
 }
 
+/**
+ * ubifs_write_sb_node - write superblock node.
+ * @c: UBIFS file-system description object
+ * @sup: superblock node read with 'ubifs_read_sb_node()'
+ *
+ * This function returns %0 on success and a negative error code on failure.
+ */
+int ubifs_write_sb_node(struct ubifs_info *c, struct ubifs_sb_node *sup)
+{
+       int len = ALIGN(UBIFS_SB_NODE_SZ, c->min_io_size);
+
+       ubifs_prepare_node(c, sup, UBIFS_SB_NODE_SZ, 1);
+       return ubifs_leb_change(c, UBIFS_SB_LNUM, sup, len);
+}
+
 /**
  * ubifs_read_superblock - read superblock.
  * @c: UBIFS file-system description object
@@ -227,8 +527,14 @@ int ubifs_read_superblock(struct ubifs_info *c)
        struct ubifs_sb_node *sup;
 
        if (c->empty) {
+#ifndef __UBOOT__
+               err = create_default_filesystem(c);
+               if (err)
+                       return err;
+#else
                printf("No UBIFS filesystem found!\n");
                return -1;
+#endif
        }
 
        sup = ubifs_read_sb_node(c);
@@ -243,16 +549,12 @@ int ubifs_read_superblock(struct ubifs_info *c)
         * due to the unavailability of time-travelling equipment.
         */
        if (c->fmt_version > UBIFS_FORMAT_VERSION) {
-               struct super_block *sb = c->vfs_sb;
-               int mounting_ro = sb->s_flags & MS_RDONLY;
-
-               ubifs_assert(!c->ro_media || mounting_ro);
-               if (!mounting_ro ||
+               ubifs_assert(!c->ro_media || c->ro_mount);
+               if (!c->ro_mount ||
                    c->ro_compat_version > UBIFS_RO_COMPAT_VERSION) {
-                       ubifs_err("on-flash format version is w%d/r%d, but "
-                                 "software only supports up to version "
-                                 "w%d/r%d", c->fmt_version,
-                                 c->ro_compat_version, UBIFS_FORMAT_VERSION,
+                       ubifs_err("on-flash format version is w%d/r%d, but software only supports up to version w%d/r%d",
+                                 c->fmt_version, c->ro_compat_version,
+                                 UBIFS_FORMAT_VERSION,
                                  UBIFS_RO_COMPAT_VERSION);
                        if (c->ro_compat_version <= UBIFS_RO_COMPAT_VERSION) {
                                ubifs_msg("only R/O mounting is possible");
@@ -310,22 +612,41 @@ int ubifs_read_superblock(struct ubifs_info *c)
        c->jhead_cnt     = le32_to_cpu(sup->jhead_cnt) + NONDATA_JHEADS_CNT;
        c->fanout        = le32_to_cpu(sup->fanout);
        c->lsave_cnt     = le32_to_cpu(sup->lsave_cnt);
-       c->default_compr = le16_to_cpu(sup->default_compr);
        c->rp_size       = le64_to_cpu(sup->rp_size);
-       c->rp_uid        = le32_to_cpu(sup->rp_uid);
-       c->rp_gid        = le32_to_cpu(sup->rp_gid);
+#ifndef __UBOOT__
+       c->rp_uid        = make_kuid(&init_user_ns, le32_to_cpu(sup->rp_uid));
+       c->rp_gid        = make_kgid(&init_user_ns, le32_to_cpu(sup->rp_gid));
+#else
+       c->rp_uid.val    = le32_to_cpu(sup->rp_uid);
+       c->rp_gid.val    = le32_to_cpu(sup->rp_gid);
+#endif
        sup_flags        = le32_to_cpu(sup->flags);
+       if (!c->mount_opts.override_compr)
+               c->default_compr = le16_to_cpu(sup->default_compr);
 
        c->vfs_sb->s_time_gran = le32_to_cpu(sup->time_gran);
        memcpy(&c->uuid, &sup->uuid, 16);
        c->big_lpt = !!(sup_flags & UBIFS_FLG_BIGLPT);
+       c->space_fixup = !!(sup_flags & UBIFS_FLG_SPACE_FIXUP);
 
        /* Automatically increase file system size to the maximum size */
        c->old_leb_cnt = c->leb_cnt;
        if (c->leb_cnt < c->vi.size && c->leb_cnt < c->max_leb_cnt) {
                c->leb_cnt = min_t(int, c->max_leb_cnt, c->vi.size);
-               dbg_mnt("Auto resizing (ro) from %d LEBs to %d LEBs",
-                       c->old_leb_cnt, c->leb_cnt);
+               if (c->ro_mount)
+                       dbg_mnt("Auto resizing (ro) from %d LEBs to %d LEBs",
+                               c->old_leb_cnt, c->leb_cnt);
+#ifndef __UBOOT__
+               else {
+                       dbg_mnt("Auto resizing (sb) from %d LEBs to %d LEBs",
+                               c->old_leb_cnt, c->leb_cnt);
+                       sup->leb_cnt = cpu_to_le32(c->leb_cnt);
+                       err = ubifs_write_sb_node(c, sup);
+                       if (err)
+                               goto out;
+                       c->old_leb_cnt = c->leb_cnt;
+               }
+#endif
        }
 
        c->log_bytes = (long long)c->log_lebs * c->leb_size;
@@ -337,10 +658,162 @@ int ubifs_read_superblock(struct ubifs_info *c)
        c->main_lebs = c->leb_cnt - UBIFS_SB_LEBS - UBIFS_MST_LEBS;
        c->main_lebs -= c->log_lebs + c->lpt_lebs + c->orph_lebs;
        c->main_first = c->leb_cnt - c->main_lebs;
-       c->report_rp_size = ubifs_reported_space(c, c->rp_size);
 
        err = validate_sb(c, sup);
 out:
        kfree(sup);
        return err;
 }
+
+/**
+ * fixup_leb - fixup/unmap an LEB containing free space.
+ * @c: UBIFS file-system description object
+ * @lnum: the LEB number to fix up
+ * @len: number of used bytes in LEB (starting at offset 0)
+ *
+ * This function reads the contents of the given LEB number @lnum, then fixes
+ * it up, so that empty min. I/O units in the end of LEB are actually erased on
+ * flash (rather than being just all-0xff real data). If the LEB is completely
+ * empty, it is simply unmapped.
+ */
+static int fixup_leb(struct ubifs_info *c, int lnum, int len)
+{
+       int err;
+
+       ubifs_assert(len >= 0);
+       ubifs_assert(len % c->min_io_size == 0);
+       ubifs_assert(len < c->leb_size);
+
+       if (len == 0) {
+               dbg_mnt("unmap empty LEB %d", lnum);
+               return ubifs_leb_unmap(c, lnum);
+       }
+
+       dbg_mnt("fixup LEB %d, data len %d", lnum, len);
+       err = ubifs_leb_read(c, lnum, c->sbuf, 0, len, 1);
+       if (err)
+               return err;
+
+       return ubifs_leb_change(c, lnum, c->sbuf, len);
+}
+
+/**
+ * fixup_free_space - find & remap all LEBs containing free space.
+ * @c: UBIFS file-system description object
+ *
+ * This function walks through all LEBs in the filesystem and fiexes up those
+ * containing free/empty space.
+ */
+static int fixup_free_space(struct ubifs_info *c)
+{
+       int lnum, err = 0;
+       struct ubifs_lprops *lprops;
+
+       ubifs_get_lprops(c);
+
+       /* Fixup LEBs in the master area */
+       for (lnum = UBIFS_MST_LNUM; lnum < UBIFS_LOG_LNUM; lnum++) {
+               err = fixup_leb(c, lnum, c->mst_offs + c->mst_node_alsz);
+               if (err)
+                       goto out;
+       }
+
+       /* Unmap unused log LEBs */
+       lnum = ubifs_next_log_lnum(c, c->lhead_lnum);
+       while (lnum != c->ltail_lnum) {
+               err = fixup_leb(c, lnum, 0);
+               if (err)
+                       goto out;
+               lnum = ubifs_next_log_lnum(c, lnum);
+       }
+
+       /*
+        * Fixup the log head which contains the only a CS node at the
+        * beginning.
+        */
+       err = fixup_leb(c, c->lhead_lnum,
+                       ALIGN(UBIFS_CS_NODE_SZ, c->min_io_size));
+       if (err)
+               goto out;
+
+       /* Fixup LEBs in the LPT area */
+       for (lnum = c->lpt_first; lnum <= c->lpt_last; lnum++) {
+               int free = c->ltab[lnum - c->lpt_first].free;
+
+               if (free > 0) {
+                       err = fixup_leb(c, lnum, c->leb_size - free);
+                       if (err)
+                               goto out;
+               }
+       }
+
+       /* Unmap LEBs in the orphans area */
+       for (lnum = c->orph_first; lnum <= c->orph_last; lnum++) {
+               err = fixup_leb(c, lnum, 0);
+               if (err)
+                       goto out;
+       }
+
+       /* Fixup LEBs in the main area */
+       for (lnum = c->main_first; lnum < c->leb_cnt; lnum++) {
+               lprops = ubifs_lpt_lookup(c, lnum);
+               if (IS_ERR(lprops)) {
+                       err = PTR_ERR(lprops);
+                       goto out;
+               }
+
+               if (lprops->free > 0) {
+                       err = fixup_leb(c, lnum, c->leb_size - lprops->free);
+                       if (err)
+                               goto out;
+               }
+       }
+
+out:
+       ubifs_release_lprops(c);
+       return err;
+}
+
+/**
+ * ubifs_fixup_free_space - find & fix all LEBs with free space.
+ * @c: UBIFS file-system description object
+ *
+ * This function fixes up LEBs containing free space on first mount, if the
+ * appropriate flag was set when the FS was created. Each LEB with one or more
+ * empty min. I/O unit (i.e. free-space-count > 0) is re-written, to make sure
+ * the free space is actually erased. E.g., this is necessary for some NAND
+ * chips, since the free space may have been programmed like real "0xff" data
+ * (generating a non-0xff ECC), causing future writes to the not-really-erased
+ * NAND pages to behave badly. After the space is fixed up, the superblock flag
+ * is cleared, so that this is skipped for all future mounts.
+ */
+int ubifs_fixup_free_space(struct ubifs_info *c)
+{
+       int err;
+       struct ubifs_sb_node *sup;
+
+       ubifs_assert(c->space_fixup);
+       ubifs_assert(!c->ro_mount);
+
+       ubifs_msg("start fixing up free space");
+
+       err = fixup_free_space(c);
+       if (err)
+               return err;
+
+       sup = ubifs_read_sb_node(c);
+       if (IS_ERR(sup))
+               return PTR_ERR(sup);
+
+       /* Free-space fixup is no longer required */
+       c->space_fixup = 0;
+       sup->flags &= cpu_to_le32(~UBIFS_FLG_SPACE_FIXUP);
+
+       err = ubifs_write_sb_node(c, sup);
+       kfree(sup);
+       if (err)
+               return err;
+
+       ubifs_msg("free space fixup complete");
+       return err;
+}
index 0ed82479b44b4daca156a80107fee1b040ec10a7..5523d4e6608f39d16e98641eb584128fc67a6e09 100644 (file)
@@ -3,18 +3,7 @@
  *
  * Copyright (C) 2006-2008 Nokia Corporation
  *
- * This program is free software; you can redistribute it and/or modify it
- * under the terms of the GNU General Public License version 2 as published by
- * the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
- * more details.
- *
- * You should have received a copy of the GNU General Public License along with
- * this program; if not, write to the Free Software Foundation, Inc., 51
- * Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
+ * SPDX-License-Identifier:    GPL-2.0+
  *
  * Authors: Adrian Hunter
  *          Artem Bityutskiy (Битюцкий Артём)
  * debugging functions.
  */
 
+#define __UBOOT__
+#ifdef __UBOOT__
+#include <linux/err.h>
+#endif
 #include "ubifs.h"
 
 /**
@@ -75,7 +68,7 @@ int ubifs_scan_a_node(const struct ubifs_info *c, void *buf, int len, int lnum,
        magic = le32_to_cpu(ch->magic);
 
        if (magic == 0xFFFFFFFF) {
-               dbg_scan("hit empty space");
+               dbg_scan("hit empty space at LEB %d:%d", lnum, offs);
                return SCANNED_EMPTY_SPACE;
        }
 
@@ -85,7 +78,8 @@ int ubifs_scan_a_node(const struct ubifs_info *c, void *buf, int len, int lnum,
        if (len < UBIFS_CH_SZ)
                return SCANNED_GARBAGE;
 
-       dbg_scan("scanning %s", dbg_ntype(ch->node_type));
+       dbg_scan("scanning %s at LEB %d:%d",
+                dbg_ntype(ch->node_type), lnum, offs);
 
        if (ubifs_check_node(c, buf, lnum, offs, quiet, 1))
                return SCANNED_A_CORRUPT_NODE;
@@ -101,22 +95,21 @@ int ubifs_scan_a_node(const struct ubifs_info *c, void *buf, int len, int lnum,
                        if (!quiet) {
                                ubifs_err("bad pad node at LEB %d:%d",
                                          lnum, offs);
-                               dbg_dump_node(c, pad);
+                               ubifs_dump_node(c, pad);
                        }
                        return SCANNED_A_BAD_PAD_NODE;
                }
 
                /* Make the node pads to 8-byte boundary */
                if ((node_len + pad_len) & 7) {
-                       if (!quiet) {
-                               dbg_err("bad padding length %d - %d",
-                                       offs, offs + node_len + pad_len);
-                       }
+                       if (!quiet)
+                               ubifs_err("bad padding length %d - %d",
+                                         offs, offs + node_len + pad_len);
                        return SCANNED_A_BAD_PAD_NODE;
                }
 
-               dbg_scan("%d bytes padded, offset now %d",
-                        pad_len, ALIGN(offs + node_len + pad_len, 8));
+               dbg_scan("%d bytes padded at LEB %d:%d, offset now %d", pad_len,
+                        lnum, offs, ALIGN(offs + node_len + pad_len, 8));
 
                return node_len + pad_len;
        }
@@ -149,10 +142,10 @@ struct ubifs_scan_leb *ubifs_start_scan(const struct ubifs_info *c, int lnum,
        INIT_LIST_HEAD(&sleb->nodes);
        sleb->buf = sbuf;
 
-       err = ubi_read(c->ubi, lnum, sbuf + offs, offs, c->leb_size - offs);
+       err = ubifs_leb_read(c, lnum, sbuf + offs, offs, c->leb_size - offs, 0);
        if (err && err != -EBADMSG) {
-               ubifs_err("cannot read %d bytes from LEB %d:%d,"
-                         " error %d", c->leb_size - offs, lnum, offs, err);
+               ubifs_err("cannot read %d bytes from LEB %d:%d, error %d",
+                         c->leb_size - offs, lnum, offs, err);
                kfree(sleb);
                return ERR_PTR(err);
        }
@@ -198,7 +191,7 @@ int ubifs_add_snod(const struct ubifs_info *c, struct ubifs_scan_leb *sleb,
        struct ubifs_ino_node *ino = buf;
        struct ubifs_scan_node *snod;
 
-       snod = kzalloc(sizeof(struct ubifs_scan_node), GFP_NOFS);
+       snod = kmalloc(sizeof(struct ubifs_scan_node), GFP_NOFS);
        if (!snod)
                return -ENOMEM;
 
@@ -213,13 +206,15 @@ int ubifs_add_snod(const struct ubifs_info *c, struct ubifs_scan_leb *sleb,
        case UBIFS_DENT_NODE:
        case UBIFS_XENT_NODE:
        case UBIFS_DATA_NODE:
-       case UBIFS_TRUN_NODE:
                /*
                 * The key is in the same place in all keyed
                 * nodes.
                 */
                key_read(c, &ino->key, &snod->key);
                break;
+       default:
+               invalid_key_init(c, &snod->key);
+               break;
        }
        list_add_tail(&snod->list, &sleb->nodes);
        sleb->nodes_cnt += 1;
@@ -238,13 +233,11 @@ void ubifs_scanned_corruption(const struct ubifs_info *c, int lnum, int offs,
 {
        int len;
 
-       ubifs_err("corrupted data at LEB %d:%d", lnum, offs);
-       if (dbg_failure_mode)
-               return;
+       ubifs_err("corruption at LEB %d:%d", lnum, offs);
        len = c->leb_size - offs;
-       if (len > 4096)
-               len = 4096;
-       dbg_err("first %d bytes from LEB %d:%d", len, lnum, offs);
+       if (len > 8192)
+               len = 8192;
+       ubifs_err("first %d bytes from LEB %d:%d", len, lnum, offs);
        print_hex_dump(KERN_DEBUG, "", DUMP_PREFIX_OFFSET, 32, 4, buf, len, 1);
 }
 
@@ -253,13 +246,19 @@ void ubifs_scanned_corruption(const struct ubifs_info *c, int lnum, int offs,
  * @c: UBIFS file-system description object
  * @lnum: logical eraseblock number
  * @offs: offset to start at (usually zero)
- * @sbuf: scan buffer (must be c->leb_size)
+ * @sbuf: scan buffer (must be of @c->leb_size bytes in size)
+ * @quiet: print no messages
  *
  * This function scans LEB number @lnum and returns complete information about
- * its contents. Returns an error code in case of failure.
+ * its contents. Returns the scaned information in case of success and,
+ * %-EUCLEAN if the LEB neads recovery, and other negative error codes in case
+ * of failure.
+ *
+ * If @quiet is non-zero, this function does not print large and scary
+ * error messages and flash dumps in case of errors.
  */
 struct ubifs_scan_leb *ubifs_scan(const struct ubifs_info *c, int lnum,
-                                 int offs, void *sbuf)
+                                 int offs, void *sbuf, int quiet)
 {
        void *buf = sbuf + offs;
        int err, len = c->leb_size - offs;
@@ -278,8 +277,7 @@ struct ubifs_scan_leb *ubifs_scan(const struct ubifs_info *c, int lnum,
 
                cond_resched();
 
-               ret = ubifs_scan_a_node(c, buf, len, lnum, offs, 0);
-
+               ret = ubifs_scan_a_node(c, buf, len, lnum, offs, quiet);
                if (ret > 0) {
                        /* Padding bytes or a valid padding node */
                        offs += ret;
@@ -294,17 +292,18 @@ struct ubifs_scan_leb *ubifs_scan(const struct ubifs_info *c, int lnum,
 
                switch (ret) {
                case SCANNED_GARBAGE:
-                       dbg_err("garbage");
+                       ubifs_err("garbage");
                        goto corrupted;
                case SCANNED_A_NODE:
                        break;
                case SCANNED_A_CORRUPT_NODE:
                case SCANNED_A_BAD_PAD_NODE:
-                       dbg_err("bad node");
+                       ubifs_err("bad node");
                        goto corrupted;
                default:
-                       dbg_err("unknown");
-                       goto corrupted;
+                       ubifs_err("unknown");
+                       err = -EINVAL;
+                       goto error;
                }
 
                err = ubifs_add_snod(c, sleb, buf, offs);
@@ -317,8 +316,12 @@ struct ubifs_scan_leb *ubifs_scan(const struct ubifs_info *c, int lnum,
                len -= node_len;
        }
 
-       if (offs % c->min_io_size)
+       if (offs % c->min_io_size) {
+               if (!quiet)
+                       ubifs_err("empty space starts at non-aligned offset %d",
+                                 offs);
                goto corrupted;
+       }
 
        ubifs_end_scan(c, sleb, lnum, offs);
 
@@ -327,18 +330,25 @@ struct ubifs_scan_leb *ubifs_scan(const struct ubifs_info *c, int lnum,
                        break;
        for (; len; offs++, buf++, len--)
                if (*(uint8_t *)buf != 0xff) {
-                       ubifs_err("corrupt empty space at LEB %d:%d",
-                                 lnum, offs);
+                       if (!quiet)
+                               ubifs_err("corrupt empty space at LEB %d:%d",
+                                         lnum, offs);
                        goto corrupted;
                }
 
        return sleb;
 
 corrupted:
-       ubifs_scanned_corruption(c, lnum, offs, buf);
+       if (!quiet) {
+               ubifs_scanned_corruption(c, lnum, offs, buf);
+               ubifs_err("LEB %d scanning failed", lnum);
+       }
        err = -EUCLEAN;
+       ubifs_scan_destroy(sleb);
+       return ERR_PTR(err);
+
 error:
-       ubifs_err("LEB %d scanning failed", lnum);
+       ubifs_err("LEB %d scanning failed, error %d", lnum, err);
        ubifs_scan_destroy(sleb);
        return ERR_PTR(err);
 }
index 748ab6792d0d2055abb82e490121cd9540b1043d..9c87db40d3838c4dd271460884b4c293c3f4e14e 100644 (file)
@@ -3,18 +3,7 @@
  *
  * Copyright (C) 2006-2008 Nokia Corporation.
  *
- * This program is free software; you can redistribute it and/or modify it
- * under the terms of the GNU General Public License version 2 as published by
- * the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
- * more details.
- *
- * You should have received a copy of the GNU General Public License along with
- * this program; if not, write to the Free Software Foundation, Inc., 51
- * Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
+ * SPDX-License-Identifier:    GPL-2.0+
  *
  * Authors: Artem Bityutskiy (Битюцкий Артём)
  *          Adrian Hunter
  * corresponding subsystems, but most of it is here.
  */
 
-#include "ubifs.h"
+#define __UBOOT__
+#ifndef __UBOOT__
+#include <linux/init.h>
+#include <linux/slab.h>
+#include <linux/module.h>
+#include <linux/ctype.h>
+#include <linux/kthread.h>
+#include <linux/parser.h>
+#include <linux/seq_file.h>
+#include <linux/mount.h>
 #include <linux/math64.h>
+#include <linux/writeback.h>
+#else
 
-#define INODE_LOCKED_MAX       64
+#include <linux/compat.h>
+#include <linux/stat.h>
+#include <linux/err.h>
+#include "ubifs.h"
+#include <ubi_uboot.h>
+#include <mtd/ubi-user.h>
 
-struct super_block *ubifs_sb;
-static struct inode *inodes_locked_down[INODE_LOCKED_MAX];
+struct dentry;
+struct file;
+struct iattr;
+struct kstat;
+struct vfsmount;
 
-/* shrinker.c */
+#define INODE_LOCKED_MAX       64
 
-/* List of all UBIFS file-system instances */
-struct list_head ubifs_infos;
+struct super_block *ubifs_sb;
+LIST_HEAD(super_blocks);
 
-/* linux/fs/super.c */
+static struct inode *inodes_locked_down[INODE_LOCKED_MAX];
 
-static int sb_set(struct super_block *sb, void *data)
+int set_anon_super(struct super_block *s, void *data)
 {
-       dev_t *dev = data;
-
-       sb->s_dev = *dev;
        return 0;
 }
 
-/**
- *     sget    -       find or create a superblock
- *     @type:  filesystem type superblock should belong to
- *     @test:  comparison callback
- *     @set:   setup callback
- *     @data:  argument to each of them
- */
-struct super_block *sget(struct file_system_type *type,
-                       int (*test)(struct super_block *,void *),
-                       int (*set)(struct super_block *,void *),
-                       void *data)
-{
-       struct super_block *s = NULL;
-       int err;
-
-       s = kzalloc(sizeof(struct super_block),  GFP_USER);
-       if (!s) {
-               err = -ENOMEM;
-               return ERR_PTR(err);
-       }
-
-       INIT_LIST_HEAD(&s->s_instances);
-       INIT_LIST_HEAD(&s->s_inodes);
-       s->s_time_gran = 1000000000;
-
-       err = set(s, data);
-       if (err) {
-               return ERR_PTR(err);
-       }
-       s->s_type = type;
-       strncpy(s->s_id, type->name, sizeof(s->s_id));
-       list_add(&s->s_instances, &type->fs_supers);
-       return s;
-}
-
-/**
- * validate_inode - validate inode.
- * @c: UBIFS file-system description object
- * @inode: the inode to validate
- *
- * This is a helper function for 'ubifs_iget()' which validates various fields
- * of a newly built inode to make sure they contain sane values and prevent
- * possible vulnerabilities. Returns zero if the inode is all right and
- * a non-zero error code if not.
- */
-static int validate_inode(struct ubifs_info *c, const struct inode *inode)
-{
-       int err;
-       const struct ubifs_inode *ui = ubifs_inode(inode);
-
-       if (inode->i_size > c->max_inode_sz) {
-               ubifs_err("inode is too large (%lld)",
-                         (long long)inode->i_size);
-               return 1;
-       }
-
-       if (ui->compr_type < 0 || ui->compr_type >= UBIFS_COMPR_TYPES_CNT) {
-               ubifs_err("unknown compression type %d", ui->compr_type);
-               return 2;
-       }
-
-       if (ui->data_len < 0 || ui->data_len > UBIFS_MAX_INO_DATA)
-               return 4;
-
-       if (!ubifs_compr_present(ui->compr_type)) {
-               ubifs_warn("inode %lu uses '%s' compression, but it was not "
-                          "compiled in", inode->i_ino,
-                          ubifs_compr_name(ui->compr_type));
-       }
-
-       err = dbg_check_dir_size(c, inode);
-       return err;
-}
-
 struct inode *iget_locked(struct super_block *sb, unsigned long ino)
 {
        struct inode *inode;
@@ -138,6 +69,10 @@ struct inode *iget_locked(struct super_block *sb, unsigned long ino)
        return inode;
 }
 
+void iget_failed(struct inode *inode)
+{
+}
+
 int ubifs_iput(struct inode *inode)
 {
        list_del_init(&inode->i_sb_list);
@@ -179,6 +114,125 @@ void iput(struct inode *inode)
        inodes_locked_down[i] = ino;
 }
 
+/* from fs/inode.c */
+/**
+ * clear_nlink - directly zero an inode's link count
+ * @inode: inode
+ *
+ * This is a low-level filesystem helper to replace any
+ * direct filesystem manipulation of i_nlink.  See
+ * drop_nlink() for why we care about i_nlink hitting zero.
+ */
+void clear_nlink(struct inode *inode)
+{
+       if (inode->i_nlink) {
+               inode->__i_nlink = 0;
+               atomic_long_inc(&inode->i_sb->s_remove_count);
+       }
+}
+EXPORT_SYMBOL(clear_nlink);
+
+/**
+ * set_nlink - directly set an inode's link count
+ * @inode: inode
+ * @nlink: new nlink (should be non-zero)
+ *
+ * This is a low-level filesystem helper to replace any
+ * direct filesystem manipulation of i_nlink.
+ */
+void set_nlink(struct inode *inode, unsigned int nlink)
+{
+       if (!nlink) {
+               clear_nlink(inode);
+       } else {
+               /* Yes, some filesystems do change nlink from zero to one */
+               if (inode->i_nlink == 0)
+                       atomic_long_dec(&inode->i_sb->s_remove_count);
+
+               inode->__i_nlink = nlink;
+       }
+}
+EXPORT_SYMBOL(set_nlink);
+
+/* from include/linux/fs.h */
+static inline void i_uid_write(struct inode *inode, uid_t uid)
+{
+       inode->i_uid.val = uid;
+}
+
+static inline void i_gid_write(struct inode *inode, gid_t gid)
+{
+       inode->i_gid.val = gid;
+}
+
+void unlock_new_inode(struct inode *inode)
+{
+       return;
+}
+#endif
+
+/*
+ * Maximum amount of memory we may 'kmalloc()' without worrying that we are
+ * allocating too much.
+ */
+#define UBIFS_KMALLOC_OK (128*1024)
+
+/* Slab cache for UBIFS inodes */
+struct kmem_cache *ubifs_inode_slab;
+
+#ifndef __UBOOT__
+/* UBIFS TNC shrinker description */
+static struct shrinker ubifs_shrinker_info = {
+       .scan_objects = ubifs_shrink_scan,
+       .count_objects = ubifs_shrink_count,
+       .seeks = DEFAULT_SEEKS,
+};
+#endif
+
+/**
+ * validate_inode - validate inode.
+ * @c: UBIFS file-system description object
+ * @inode: the inode to validate
+ *
+ * This is a helper function for 'ubifs_iget()' which validates various fields
+ * of a newly built inode to make sure they contain sane values and prevent
+ * possible vulnerabilities. Returns zero if the inode is all right and
+ * a non-zero error code if not.
+ */
+static int validate_inode(struct ubifs_info *c, const struct inode *inode)
+{
+       int err;
+       const struct ubifs_inode *ui = ubifs_inode(inode);
+
+       if (inode->i_size > c->max_inode_sz) {
+               ubifs_err("inode is too large (%lld)",
+                         (long long)inode->i_size);
+               return 1;
+       }
+
+       if (ui->compr_type < 0 || ui->compr_type >= UBIFS_COMPR_TYPES_CNT) {
+               ubifs_err("unknown compression type %d", ui->compr_type);
+               return 2;
+       }
+
+       if (ui->xattr_names + ui->xattr_cnt > XATTR_LIST_MAX)
+               return 3;
+
+       if (ui->data_len < 0 || ui->data_len > UBIFS_MAX_INO_DATA)
+               return 4;
+
+       if (ui->xattr && !S_ISREG(inode->i_mode))
+               return 5;
+
+       if (!ubifs_compr_present(ui->compr_type)) {
+               ubifs_warn("inode %lu uses '%s' compression, but it was not compiled in",
+                          inode->i_ino, ubifs_compr_name(ui->compr_type));
+       }
+
+       err = dbg_check_dir(c, inode);
+       return err;
+}
+
 struct inode *ubifs_iget(struct super_block *sb, unsigned long inum)
 {
        int err;
@@ -187,10 +241,13 @@ struct inode *ubifs_iget(struct super_block *sb, unsigned long inum)
        struct ubifs_info *c = sb->s_fs_info;
        struct inode *inode;
        struct ubifs_inode *ui;
+#ifdef __UBOOT__
        int i;
+#endif
 
        dbg_gen("inode %lu", inum);
 
+#ifdef __UBOOT__
        /*
         * U-Boot special handling of locked down inodes via recovery
         * e.g. ubifs_recover_size()
@@ -211,6 +268,7 @@ struct inode *ubifs_iget(struct super_block *sb, unsigned long inum)
                        return inodes_locked_down[i];
                }
        }
+#endif
 
        inode = iget_locked(sb, inum);
        if (!inode)
@@ -232,9 +290,9 @@ struct inode *ubifs_iget(struct super_block *sb, unsigned long inum)
                goto out_ino;
 
        inode->i_flags |= (S_NOCMTIME | S_NOATIME);
-       inode->i_nlink = le32_to_cpu(ino->nlink);
-       inode->i_uid   = le32_to_cpu(ino->uid);
-       inode->i_gid   = le32_to_cpu(ino->gid);
+       set_nlink(inode, le32_to_cpu(ino->nlink));
+       i_uid_write(inode, le32_to_cpu(ino->uid));
+       i_gid_write(inode, le32_to_cpu(ino->gid));
        inode->i_atime.tv_sec  = (int64_t)le64_to_cpu(ino->atime_sec);
        inode->i_atime.tv_nsec = le32_to_cpu(ino->atime_nsec);
        inode->i_mtime.tv_sec  = (int64_t)le64_to_cpu(ino->mtime_sec);
@@ -248,12 +306,101 @@ struct inode *ubifs_iget(struct super_block *sb, unsigned long inum)
        ui->flags       = le32_to_cpu(ino->flags);
        ui->compr_type  = le16_to_cpu(ino->compr_type);
        ui->creat_sqnum = le64_to_cpu(ino->creat_sqnum);
+       ui->xattr_cnt   = le32_to_cpu(ino->xattr_cnt);
+       ui->xattr_size  = le32_to_cpu(ino->xattr_size);
+       ui->xattr_names = le32_to_cpu(ino->xattr_names);
        ui->synced_i_size = ui->ui_size = inode->i_size;
 
+       ui->xattr = (ui->flags & UBIFS_XATTR_FL) ? 1 : 0;
+
        err = validate_inode(c, inode);
        if (err)
                goto out_invalid;
 
+#ifndef __UBOOT__
+       /* Disable read-ahead */
+       inode->i_mapping->backing_dev_info = &c->bdi;
+
+       switch (inode->i_mode & S_IFMT) {
+       case S_IFREG:
+               inode->i_mapping->a_ops = &ubifs_file_address_operations;
+               inode->i_op = &ubifs_file_inode_operations;
+               inode->i_fop = &ubifs_file_operations;
+               if (ui->xattr) {
+                       ui->data = kmalloc(ui->data_len + 1, GFP_NOFS);
+                       if (!ui->data) {
+                               err = -ENOMEM;
+                               goto out_ino;
+                       }
+                       memcpy(ui->data, ino->data, ui->data_len);
+                       ((char *)ui->data)[ui->data_len] = '\0';
+               } else if (ui->data_len != 0) {
+                       err = 10;
+                       goto out_invalid;
+               }
+               break;
+       case S_IFDIR:
+               inode->i_op  = &ubifs_dir_inode_operations;
+               inode->i_fop = &ubifs_dir_operations;
+               if (ui->data_len != 0) {
+                       err = 11;
+                       goto out_invalid;
+               }
+               break;
+       case S_IFLNK:
+               inode->i_op = &ubifs_symlink_inode_operations;
+               if (ui->data_len <= 0 || ui->data_len > UBIFS_MAX_INO_DATA) {
+                       err = 12;
+                       goto out_invalid;
+               }
+               ui->data = kmalloc(ui->data_len + 1, GFP_NOFS);
+               if (!ui->data) {
+                       err = -ENOMEM;
+                       goto out_ino;
+               }
+               memcpy(ui->data, ino->data, ui->data_len);
+               ((char *)ui->data)[ui->data_len] = '\0';
+               break;
+       case S_IFBLK:
+       case S_IFCHR:
+       {
+               dev_t rdev;
+               union ubifs_dev_desc *dev;
+
+               ui->data = kmalloc(sizeof(union ubifs_dev_desc), GFP_NOFS);
+               if (!ui->data) {
+                       err = -ENOMEM;
+                       goto out_ino;
+               }
+
+               dev = (union ubifs_dev_desc *)ino->data;
+               if (ui->data_len == sizeof(dev->new))
+                       rdev = new_decode_dev(le32_to_cpu(dev->new));
+               else if (ui->data_len == sizeof(dev->huge))
+                       rdev = huge_decode_dev(le64_to_cpu(dev->huge));
+               else {
+                       err = 13;
+                       goto out_invalid;
+               }
+               memcpy(ui->data, ino->data, ui->data_len);
+               inode->i_op = &ubifs_file_inode_operations;
+               init_special_inode(inode, inode->i_mode, rdev);
+               break;
+       }
+       case S_IFSOCK:
+       case S_IFIFO:
+               inode->i_op = &ubifs_file_inode_operations;
+               init_special_inode(inode, inode->i_mode, 0);
+               if (ui->data_len != 0) {
+                       err = 14;
+                       goto out_invalid;
+               }
+               break;
+       default:
+               err = 15;
+               goto out_invalid;
+       }
+#else
        if ((inode->i_mode & S_IFMT) == S_IFLNK) {
                if (ui->data_len <= 0 || ui->data_len > UBIFS_MAX_INO_DATA) {
                        err = 12;
@@ -267,23 +414,258 @@ struct inode *ubifs_iget(struct super_block *sb, unsigned long inum)
                memcpy(ui->data, ino->data, ui->data_len);
                ((char *)ui->data)[ui->data_len] = '\0';
        }
+#endif
 
        kfree(ino);
-       inode->i_state &= ~(I_LOCK | I_NEW);
+#ifndef __UBOOT__
+       ubifs_set_inode_flags(inode);
+#endif
+       unlock_new_inode(inode);
        return inode;
 
 out_invalid:
        ubifs_err("inode %lu validation failed, error %d", inode->i_ino, err);
-       dbg_dump_node(c, ino);
-       dbg_dump_inode(c, inode);
+       ubifs_dump_node(c, ino);
+       ubifs_dump_inode(c, inode);
        err = -EINVAL;
 out_ino:
        kfree(ino);
 out:
        ubifs_err("failed to read inode %lu, error %d", inode->i_ino, err);
+       iget_failed(inode);
        return ERR_PTR(err);
 }
 
+static struct inode *ubifs_alloc_inode(struct super_block *sb)
+{
+       struct ubifs_inode *ui;
+
+       ui = kmem_cache_alloc(ubifs_inode_slab, GFP_NOFS);
+       if (!ui)
+               return NULL;
+
+       memset((void *)ui + sizeof(struct inode), 0,
+              sizeof(struct ubifs_inode) - sizeof(struct inode));
+       mutex_init(&ui->ui_mutex);
+       spin_lock_init(&ui->ui_lock);
+       return &ui->vfs_inode;
+};
+
+#ifndef __UBOOT__
+static void ubifs_i_callback(struct rcu_head *head)
+{
+       struct inode *inode = container_of(head, struct inode, i_rcu);
+       struct ubifs_inode *ui = ubifs_inode(inode);
+       kmem_cache_free(ubifs_inode_slab, ui);
+}
+
+static void ubifs_destroy_inode(struct inode *inode)
+{
+       struct ubifs_inode *ui = ubifs_inode(inode);
+
+       kfree(ui->data);
+       call_rcu(&inode->i_rcu, ubifs_i_callback);
+}
+
+/*
+ * Note, Linux write-back code calls this without 'i_mutex'.
+ */
+static int ubifs_write_inode(struct inode *inode, struct writeback_control *wbc)
+{
+       int err = 0;
+       struct ubifs_info *c = inode->i_sb->s_fs_info;
+       struct ubifs_inode *ui = ubifs_inode(inode);
+
+       ubifs_assert(!ui->xattr);
+       if (is_bad_inode(inode))
+               return 0;
+
+       mutex_lock(&ui->ui_mutex);
+       /*
+        * Due to races between write-back forced by budgeting
+        * (see 'sync_some_inodes()') and background write-back, the inode may
+        * have already been synchronized, do not do this again. This might
+        * also happen if it was synchronized in an VFS operation, e.g.
+        * 'ubifs_link()'.
+        */
+       if (!ui->dirty) {
+               mutex_unlock(&ui->ui_mutex);
+               return 0;
+       }
+
+       /*
+        * As an optimization, do not write orphan inodes to the media just
+        * because this is not needed.
+        */
+       dbg_gen("inode %lu, mode %#x, nlink %u",
+               inode->i_ino, (int)inode->i_mode, inode->i_nlink);
+       if (inode->i_nlink) {
+               err = ubifs_jnl_write_inode(c, inode);
+               if (err)
+                       ubifs_err("can't write inode %lu, error %d",
+                                 inode->i_ino, err);
+               else
+                       err = dbg_check_inode_size(c, inode, ui->ui_size);
+       }
+
+       ui->dirty = 0;
+       mutex_unlock(&ui->ui_mutex);
+       ubifs_release_dirty_inode_budget(c, ui);
+       return err;
+}
+
+static void ubifs_evict_inode(struct inode *inode)
+{
+       int err;
+       struct ubifs_info *c = inode->i_sb->s_fs_info;
+       struct ubifs_inode *ui = ubifs_inode(inode);
+
+       if (ui->xattr)
+               /*
+                * Extended attribute inode deletions are fully handled in
+                * 'ubifs_removexattr()'. These inodes are special and have
+                * limited usage, so there is nothing to do here.
+                */
+               goto out;
+
+       dbg_gen("inode %lu, mode %#x", inode->i_ino, (int)inode->i_mode);
+       ubifs_assert(!atomic_read(&inode->i_count));
+
+       truncate_inode_pages(&inode->i_data, 0);
+
+       if (inode->i_nlink)
+               goto done;
+
+       if (is_bad_inode(inode))
+               goto out;
+
+       ui->ui_size = inode->i_size = 0;
+       err = ubifs_jnl_delete_inode(c, inode);
+       if (err)
+               /*
+                * Worst case we have a lost orphan inode wasting space, so a
+                * simple error message is OK here.
+                */
+               ubifs_err("can't delete inode %lu, error %d",
+                         inode->i_ino, err);
+
+out:
+       if (ui->dirty)
+               ubifs_release_dirty_inode_budget(c, ui);
+       else {
+               /* We've deleted something - clean the "no space" flags */
+               c->bi.nospace = c->bi.nospace_rp = 0;
+               smp_wmb();
+       }
+done:
+       clear_inode(inode);
+}
+#endif
+
+static void ubifs_dirty_inode(struct inode *inode, int flags)
+{
+       struct ubifs_inode *ui = ubifs_inode(inode);
+
+       ubifs_assert(mutex_is_locked(&ui->ui_mutex));
+       if (!ui->dirty) {
+               ui->dirty = 1;
+               dbg_gen("inode %lu",  inode->i_ino);
+       }
+}
+
+#ifndef __UBOOT__
+static int ubifs_statfs(struct dentry *dentry, struct kstatfs *buf)
+{
+       struct ubifs_info *c = dentry->d_sb->s_fs_info;
+       unsigned long long free;
+       __le32 *uuid = (__le32 *)c->uuid;
+
+       free = ubifs_get_free_space(c);
+       dbg_gen("free space %lld bytes (%lld blocks)",
+               free, free >> UBIFS_BLOCK_SHIFT);
+
+       buf->f_type = UBIFS_SUPER_MAGIC;
+       buf->f_bsize = UBIFS_BLOCK_SIZE;
+       buf->f_blocks = c->block_cnt;
+       buf->f_bfree = free >> UBIFS_BLOCK_SHIFT;
+       if (free > c->report_rp_size)
+               buf->f_bavail = (free - c->report_rp_size) >> UBIFS_BLOCK_SHIFT;
+       else
+               buf->f_bavail = 0;
+       buf->f_files = 0;
+       buf->f_ffree = 0;
+       buf->f_namelen = UBIFS_MAX_NLEN;
+       buf->f_fsid.val[0] = le32_to_cpu(uuid[0]) ^ le32_to_cpu(uuid[2]);
+       buf->f_fsid.val[1] = le32_to_cpu(uuid[1]) ^ le32_to_cpu(uuid[3]);
+       ubifs_assert(buf->f_bfree <= c->block_cnt);
+       return 0;
+}
+
+static int ubifs_show_options(struct seq_file *s, struct dentry *root)
+{
+       struct ubifs_info *c = root->d_sb->s_fs_info;
+
+       if (c->mount_opts.unmount_mode == 2)
+               seq_printf(s, ",fast_unmount");
+       else if (c->mount_opts.unmount_mode == 1)
+               seq_printf(s, ",norm_unmount");
+
+       if (c->mount_opts.bulk_read == 2)
+               seq_printf(s, ",bulk_read");
+       else if (c->mount_opts.bulk_read == 1)
+               seq_printf(s, ",no_bulk_read");
+
+       if (c->mount_opts.chk_data_crc == 2)
+               seq_printf(s, ",chk_data_crc");
+       else if (c->mount_opts.chk_data_crc == 1)
+               seq_printf(s, ",no_chk_data_crc");
+
+       if (c->mount_opts.override_compr) {
+               seq_printf(s, ",compr=%s",
+                          ubifs_compr_name(c->mount_opts.compr_type));
+       }
+
+       return 0;
+}
+
+static int ubifs_sync_fs(struct super_block *sb, int wait)
+{
+       int i, err;
+       struct ubifs_info *c = sb->s_fs_info;
+
+       /*
+        * Zero @wait is just an advisory thing to help the file system shove
+        * lots of data into the queues, and there will be the second
+        * '->sync_fs()' call, with non-zero @wait.
+        */
+       if (!wait)
+               return 0;
+
+       /*
+        * Synchronize write buffers, because 'ubifs_run_commit()' does not
+        * do this if it waits for an already running commit.
+        */
+       for (i = 0; i < c->jhead_cnt; i++) {
+               err = ubifs_wbuf_sync(&c->jheads[i].wbuf);
+               if (err)
+                       return err;
+       }
+
+       /*
+        * Strictly speaking, it is not necessary to commit the journal here,
+        * synchronizing write-buffers would be enough. But committing makes
+        * UBIFS free space predictions much more accurate, so we want to let
+        * the user be able to get more accurate results of 'statfs()' after
+        * they synchronize the file system.
+        */
+       err = ubifs_run_commit(c);
+       if (err)
+               return err;
+
+       return ubi_sync(c->vi.ubi_num);
+}
+#endif
+
 /**
  * init_constants_early - initialize UBIFS constants.
  * @c: UBIFS file-system description object
@@ -312,9 +694,12 @@ static int init_constants_early(struct ubifs_info *c)
 
        c->leb_cnt = c->vi.size;
        c->leb_size = c->vi.usable_leb_size;
+       c->leb_start = c->di.leb_start;
        c->half_leb_size = c->leb_size / 2;
        c->min_io_size = c->di.min_io_size;
        c->min_io_shift = fls(c->min_io_size) - 1;
+       c->max_write_size = c->di.max_write_size;
+       c->max_write_shift = fls(c->max_write_size) - 1;
 
        if (c->leb_size < UBIFS_MIN_LEB_SZ) {
                ubifs_err("too small LEBs (%d bytes), min. is %d bytes",
@@ -333,6 +718,18 @@ static int init_constants_early(struct ubifs_info *c)
                return -EINVAL;
        }
 
+       /*
+        * Maximum write size has to be greater or equivalent to min. I/O
+        * size, and be multiple of min. I/O size.
+        */
+       if (c->max_write_size < c->min_io_size ||
+           c->max_write_size % c->min_io_size ||
+           !is_power_of_2(c->max_write_size)) {
+               ubifs_err("bad write buffer size %d for %d min. I/O unit",
+                         c->max_write_size, c->min_io_size);
+               return -EINVAL;
+       }
+
        /*
         * UBIFS aligns all node to 8-byte boundary, so to make function in
         * io.c simpler, assume minimum I/O unit size to be 8 bytes if it is
@@ -341,6 +738,10 @@ static int init_constants_early(struct ubifs_info *c)
        if (c->min_io_size < 8) {
                c->min_io_size = 8;
                c->min_io_shift = 3;
+               if (c->max_write_size < c->min_io_size) {
+                       c->max_write_size = c->min_io_size;
+                       c->max_write_shift = c->min_io_shift;
+               }
        }
 
        c->ref_node_alsz = ALIGN(UBIFS_REF_NODE_SZ, c->min_io_size);
@@ -393,9 +794,33 @@ static int init_constants_early(struct ubifs_info *c)
         */
        c->leb_overhead = c->leb_size % UBIFS_MAX_DATA_NODE_SZ;
 
+       /* Buffer size for bulk-reads */
+       c->max_bu_buf_len = UBIFS_MAX_BULK_READ * UBIFS_MAX_DATA_NODE_SZ;
+       if (c->max_bu_buf_len > c->leb_size)
+               c->max_bu_buf_len = c->leb_size;
        return 0;
 }
 
+/**
+ * bud_wbuf_callback - bud LEB write-buffer synchronization call-back.
+ * @c: UBIFS file-system description object
+ * @lnum: LEB the write-buffer was synchronized to
+ * @free: how many free bytes left in this LEB
+ * @pad: how many bytes were padded
+ *
+ * This is a callback function which is called by the I/O unit when the
+ * write-buffer is synchronized. We need this to correctly maintain space
+ * accounting in bud logical eraseblocks. This function returns zero in case of
+ * success and a negative error code in case of failure.
+ *
+ * This function actually belongs to the journal, but we keep it here because
+ * we want to keep it static.
+ */
+static int bud_wbuf_callback(struct ubifs_info *c, int lnum, int free, int pad)
+{
+       return ubifs_update_one_lp(c, lnum, free, pad, 0, 0);
+}
+
 /*
  * init_constants_sb - initialize UBIFS constants.
  * @c: UBIFS file-system description object
@@ -426,8 +851,8 @@ static int init_constants_sb(struct ubifs_info *c)
        tmp = UBIFS_CS_NODE_SZ + UBIFS_REF_NODE_SZ * c->jhead_cnt;
        tmp = ALIGN(tmp, c->min_io_size);
        if (tmp > c->leb_size) {
-               dbg_err("too small LEB size %d, at least %d needed",
-                       c->leb_size, tmp);
+               ubifs_err("too small LEB size %d, at least %d needed",
+                         c->leb_size, tmp);
                return -EINVAL;
        }
 
@@ -441,8 +866,8 @@ static int init_constants_sb(struct ubifs_info *c)
        tmp /= c->leb_size;
        tmp += 1;
        if (c->log_lebs < tmp) {
-               dbg_err("too small log %d LEBs, required min. %d LEBs",
-                       c->log_lebs, tmp);
+               ubifs_err("too small log %d LEBs, required min. %d LEBs",
+                         c->log_lebs, tmp);
                return -EINVAL;
        }
 
@@ -451,11 +876,11 @@ static int init_constants_sb(struct ubifs_info *c)
         * be compressed and direntries are of the maximum size.
         *
         * Note, data, which may be stored in inodes is budgeted separately, so
-        * it is not included into 'c->inode_budget'.
+        * it is not included into 'c->bi.inode_budget'.
         */
-       c->page_budget = UBIFS_MAX_DATA_NODE_SZ * UBIFS_BLOCKS_PER_PAGE;
-       c->inode_budget = UBIFS_INO_NODE_SZ;
-       c->dent_budget = UBIFS_MAX_DENT_NODE_SZ;
+       c->bi.page_budget = UBIFS_MAX_DATA_NODE_SZ * UBIFS_BLOCKS_PER_PAGE;
+       c->bi.inode_budget = UBIFS_INO_NODE_SZ;
+       c->bi.dent_budget = UBIFS_MAX_DENT_NODE_SZ;
 
        /*
         * When the amount of flash space used by buds becomes
@@ -482,6 +907,8 @@ static int init_constants_sb(struct ubifs_info *c)
        if (err)
                return err;
 
+       /* Initialize effective LEB size used in budgeting calculations */
+       c->idx_leb_size = c->leb_size - c->max_idx_node_sz;
        return 0;
 }
 
@@ -497,7 +924,8 @@ static void init_constants_master(struct ubifs_info *c)
 {
        long long tmp64;
 
-       c->min_idx_lebs = ubifs_calc_min_idx_lebs(c);
+       c->bi.min_idx_lebs = ubifs_calc_min_idx_lebs(c);
+       c->report_rp_size = ubifs_reported_space(c, c->rp_size);
 
        /*
         * Calculate total amount of FS blocks. This number is not used
@@ -514,6 +942,88 @@ static void init_constants_master(struct ubifs_info *c)
        c->block_cnt = tmp64 >> UBIFS_BLOCK_SHIFT;
 }
 
+/**
+ * take_gc_lnum - reserve GC LEB.
+ * @c: UBIFS file-system description object
+ *
+ * This function ensures that the LEB reserved for garbage collection is marked
+ * as "taken" in lprops. We also have to set free space to LEB size and dirty
+ * space to zero, because lprops may contain out-of-date information if the
+ * file-system was un-mounted before it has been committed. This function
+ * returns zero in case of success and a negative error code in case of
+ * failure.
+ */
+static int take_gc_lnum(struct ubifs_info *c)
+{
+       int err;
+
+       if (c->gc_lnum == -1) {
+               ubifs_err("no LEB for GC");
+               return -EINVAL;
+       }
+
+       /* And we have to tell lprops that this LEB is taken */
+       err = ubifs_change_one_lp(c, c->gc_lnum, c->leb_size, 0,
+                                 LPROPS_TAKEN, 0, 0);
+       return err;
+}
+
+/**
+ * alloc_wbufs - allocate write-buffers.
+ * @c: UBIFS file-system description object
+ *
+ * This helper function allocates and initializes UBIFS write-buffers. Returns
+ * zero in case of success and %-ENOMEM in case of failure.
+ */
+static int alloc_wbufs(struct ubifs_info *c)
+{
+       int i, err;
+
+       c->jheads = kzalloc(c->jhead_cnt * sizeof(struct ubifs_jhead),
+                          GFP_KERNEL);
+       if (!c->jheads)
+               return -ENOMEM;
+
+       /* Initialize journal heads */
+       for (i = 0; i < c->jhead_cnt; i++) {
+               INIT_LIST_HEAD(&c->jheads[i].buds_list);
+               err = ubifs_wbuf_init(c, &c->jheads[i].wbuf);
+               if (err)
+                       return err;
+
+               c->jheads[i].wbuf.sync_callback = &bud_wbuf_callback;
+               c->jheads[i].wbuf.jhead = i;
+               c->jheads[i].grouped = 1;
+       }
+
+       /*
+        * Garbage Collector head does not need to be synchronized by timer.
+        * Also GC head nodes are not grouped.
+        */
+       c->jheads[GCHD].wbuf.no_timer = 1;
+       c->jheads[GCHD].grouped = 0;
+
+       return 0;
+}
+
+/**
+ * free_wbufs - free write-buffers.
+ * @c: UBIFS file-system description object
+ */
+static void free_wbufs(struct ubifs_info *c)
+{
+       int i;
+
+       if (c->jheads) {
+               for (i = 0; i < c->jhead_cnt; i++) {
+                       kfree(c->jheads[i].wbuf.buf);
+                       kfree(c->jheads[i].wbuf.inodes);
+               }
+               kfree(c->jheads);
+               c->jheads = NULL;
+       }
+}
+
 /**
  * free_orphans - free orphans.
  * @c: UBIFS file-system description object
@@ -533,13 +1043,27 @@ static void free_orphans(struct ubifs_info *c)
                orph = list_entry(c->orph_list.next, struct ubifs_orphan, list);
                list_del(&orph->list);
                kfree(orph);
-               dbg_err("orphan list not empty at unmount");
+               ubifs_err("orphan list not empty at unmount");
        }
 
        vfree(c->orph_buf);
        c->orph_buf = NULL;
 }
 
+#ifndef __UBOOT__
+/**
+ * free_buds - free per-bud objects.
+ * @c: UBIFS file-system description object
+ */
+static void free_buds(struct ubifs_info *c)
+{
+       struct ubifs_bud *bud, *n;
+
+       rbtree_postorder_for_each_entry_safe(bud, n, &c->buds, rb)
+               kfree(bud);
+}
+#endif
+
 /**
  * check_volume_empty - check if the UBI volume is empty.
  * @c: UBIFS file-system description object
@@ -555,7 +1079,7 @@ static int check_volume_empty(struct ubifs_info *c)
 
        c->empty = 1;
        for (lnum = 0; lnum < c->leb_cnt; lnum++) {
-               err = ubi_is_mapped(c->ubi, lnum);
+               err = ubifs_is_mapped(c, lnum);
                if (unlikely(err < 0))
                        return err;
                if (err == 1) {
@@ -569,23 +1093,258 @@ static int check_volume_empty(struct ubifs_info *c)
        return 0;
 }
 
+/*
+ * UBIFS mount options.
+ *
+ * Opt_fast_unmount: do not run a journal commit before un-mounting
+ * Opt_norm_unmount: run a journal commit before un-mounting
+ * Opt_bulk_read: enable bulk-reads
+ * Opt_no_bulk_read: disable bulk-reads
+ * Opt_chk_data_crc: check CRCs when reading data nodes
+ * Opt_no_chk_data_crc: do not check CRCs when reading data nodes
+ * Opt_override_compr: override default compressor
+ * Opt_err: just end of array marker
+ */
+enum {
+       Opt_fast_unmount,
+       Opt_norm_unmount,
+       Opt_bulk_read,
+       Opt_no_bulk_read,
+       Opt_chk_data_crc,
+       Opt_no_chk_data_crc,
+       Opt_override_compr,
+       Opt_err,
+};
+
+#ifndef __UBOOT__
+static const match_table_t tokens = {
+       {Opt_fast_unmount, "fast_unmount"},
+       {Opt_norm_unmount, "norm_unmount"},
+       {Opt_bulk_read, "bulk_read"},
+       {Opt_no_bulk_read, "no_bulk_read"},
+       {Opt_chk_data_crc, "chk_data_crc"},
+       {Opt_no_chk_data_crc, "no_chk_data_crc"},
+       {Opt_override_compr, "compr=%s"},
+       {Opt_err, NULL},
+};
+
+/**
+ * parse_standard_option - parse a standard mount option.
+ * @option: the option to parse
+ *
+ * Normally, standard mount options like "sync" are passed to file-systems as
+ * flags. However, when a "rootflags=" kernel boot parameter is used, they may
+ * be present in the options string. This function tries to deal with this
+ * situation and parse standard options. Returns 0 if the option was not
+ * recognized, and the corresponding integer flag if it was.
+ *
+ * UBIFS is only interested in the "sync" option, so do not check for anything
+ * else.
+ */
+static int parse_standard_option(const char *option)
+{
+       ubifs_msg("parse %s", option);
+       if (!strcmp(option, "sync"))
+               return MS_SYNCHRONOUS;
+       return 0;
+}
+
+/**
+ * ubifs_parse_options - parse mount parameters.
+ * @c: UBIFS file-system description object
+ * @options: parameters to parse
+ * @is_remount: non-zero if this is FS re-mount
+ *
+ * This function parses UBIFS mount options and returns zero in case success
+ * and a negative error code in case of failure.
+ */
+static int ubifs_parse_options(struct ubifs_info *c, char *options,
+                              int is_remount)
+{
+       char *p;
+       substring_t args[MAX_OPT_ARGS];
+
+       if (!options)
+               return 0;
+
+       while ((p = strsep(&options, ","))) {
+               int token;
+
+               if (!*p)
+                       continue;
+
+               token = match_token(p, tokens, args);
+               switch (token) {
+               /*
+                * %Opt_fast_unmount and %Opt_norm_unmount options are ignored.
+                * We accept them in order to be backward-compatible. But this
+                * should be removed at some point.
+                */
+               case Opt_fast_unmount:
+                       c->mount_opts.unmount_mode = 2;
+                       break;
+               case Opt_norm_unmount:
+                       c->mount_opts.unmount_mode = 1;
+                       break;
+               case Opt_bulk_read:
+                       c->mount_opts.bulk_read = 2;
+                       c->bulk_read = 1;
+                       break;
+               case Opt_no_bulk_read:
+                       c->mount_opts.bulk_read = 1;
+                       c->bulk_read = 0;
+                       break;
+               case Opt_chk_data_crc:
+                       c->mount_opts.chk_data_crc = 2;
+                       c->no_chk_data_crc = 0;
+                       break;
+               case Opt_no_chk_data_crc:
+                       c->mount_opts.chk_data_crc = 1;
+                       c->no_chk_data_crc = 1;
+                       break;
+               case Opt_override_compr:
+               {
+                       char *name = match_strdup(&args[0]);
+
+                       if (!name)
+                               return -ENOMEM;
+                       if (!strcmp(name, "none"))
+                               c->mount_opts.compr_type = UBIFS_COMPR_NONE;
+                       else if (!strcmp(name, "lzo"))
+                               c->mount_opts.compr_type = UBIFS_COMPR_LZO;
+                       else if (!strcmp(name, "zlib"))
+                               c->mount_opts.compr_type = UBIFS_COMPR_ZLIB;
+                       else {
+                               ubifs_err("unknown compressor \"%s\"", name);
+                               kfree(name);
+                               return -EINVAL;
+                       }
+                       kfree(name);
+                       c->mount_opts.override_compr = 1;
+                       c->default_compr = c->mount_opts.compr_type;
+                       break;
+               }
+               default:
+               {
+                       unsigned long flag;
+                       struct super_block *sb = c->vfs_sb;
+
+                       flag = parse_standard_option(p);
+                       if (!flag) {
+                               ubifs_err("unrecognized mount option \"%s\" or missing value",
+                                         p);
+                               return -EINVAL;
+                       }
+                       sb->s_flags |= flag;
+                       break;
+               }
+               }
+       }
+
+       return 0;
+}
+
+/**
+ * destroy_journal - destroy journal data structures.
+ * @c: UBIFS file-system description object
+ *
+ * This function destroys journal data structures including those that may have
+ * been created by recovery functions.
+ */
+static void destroy_journal(struct ubifs_info *c)
+{
+       while (!list_empty(&c->unclean_leb_list)) {
+               struct ubifs_unclean_leb *ucleb;
+
+               ucleb = list_entry(c->unclean_leb_list.next,
+                                  struct ubifs_unclean_leb, list);
+               list_del(&ucleb->list);
+               kfree(ucleb);
+       }
+       while (!list_empty(&c->old_buds)) {
+               struct ubifs_bud *bud;
+
+               bud = list_entry(c->old_buds.next, struct ubifs_bud, list);
+               list_del(&bud->list);
+               kfree(bud);
+       }
+       ubifs_destroy_idx_gc(c);
+       ubifs_destroy_size_tree(c);
+       ubifs_tnc_close(c);
+       free_buds(c);
+}
+#endif
+
+/**
+ * bu_init - initialize bulk-read information.
+ * @c: UBIFS file-system description object
+ */
+static void bu_init(struct ubifs_info *c)
+{
+       ubifs_assert(c->bulk_read == 1);
+
+       if (c->bu.buf)
+               return; /* Already initialized */
+
+again:
+       c->bu.buf = kmalloc(c->max_bu_buf_len, GFP_KERNEL | __GFP_NOWARN);
+       if (!c->bu.buf) {
+               if (c->max_bu_buf_len > UBIFS_KMALLOC_OK) {
+                       c->max_bu_buf_len = UBIFS_KMALLOC_OK;
+                       goto again;
+               }
+
+               /* Just disable bulk-read */
+               ubifs_warn("cannot allocate %d bytes of memory for bulk-read, disabling it",
+                          c->max_bu_buf_len);
+               c->mount_opts.bulk_read = 1;
+               c->bulk_read = 0;
+               return;
+       }
+}
+
+#ifndef __UBOOT__
+/**
+ * check_free_space - check if there is enough free space to mount.
+ * @c: UBIFS file-system description object
+ *
+ * This function makes sure UBIFS has enough free space to be mounted in
+ * read/write mode. UBIFS must always have some free space to allow deletions.
+ */
+static int check_free_space(struct ubifs_info *c)
+{
+       ubifs_assert(c->dark_wm > 0);
+       if (c->lst.total_free + c->lst.total_dirty < c->dark_wm) {
+               ubifs_err("insufficient free space to mount in R/W mode");
+               ubifs_dump_budg(c, &c->bi);
+               ubifs_dump_lprops(c);
+               return -ENOSPC;
+       }
+       return 0;
+}
+#endif
+
 /**
  * mount_ubifs - mount UBIFS file-system.
  * @c: UBIFS file-system description object
  *
  * This function mounts UBIFS file system. Returns zero in case of success and
  * a negative error code in case of failure.
- *
- * Note, the function does not de-allocate resources it it fails half way
- * through, and the caller has to do this instead.
  */
 static int mount_ubifs(struct ubifs_info *c)
 {
-       struct super_block *sb = c->vfs_sb;
-       int err, mounted_read_only = (sb->s_flags & MS_RDONLY);
-       long long x;
+       int err;
+       long long x, y;
        size_t sz;
 
+       c->ro_mount = !!(c->vfs_sb->s_flags & MS_RDONLY);
+#ifdef __UBOOT__
+       if (!c->ro_mount) {
+               printf("UBIFS: only ro mode in U-Boot allowed.\n");
+               return -EACCES;
+       }
+#endif
+
        err = init_constants_early(c);
        if (err)
                return err;
@@ -598,7 +1357,7 @@ static int mount_ubifs(struct ubifs_info *c)
        if (err)
                goto out_free;
 
-       if (c->empty && (mounted_read_only || c->ro_media)) {
+       if (c->empty && (c->ro_mount || c->ro_media)) {
                /*
                 * This UBI volume is empty, and read-only, or the file system
                 * is mounted read-only - we cannot format it.
@@ -609,7 +1368,7 @@ static int mount_ubifs(struct ubifs_info *c)
                goto out_free;
        }
 
-       if (c->ro_media && !mounted_read_only) {
+       if (c->ro_media && !c->ro_mount) {
                ubifs_err("cannot mount read-write - read-only media");
                err = -EROFS;
                goto out_free;
@@ -629,11 +1388,27 @@ static int mount_ubifs(struct ubifs_info *c)
        if (!c->sbuf)
                goto out_free;
 
-       /*
-        * We have to check all CRCs, even for data nodes, when we mount the FS
-        * (specifically, when we are replaying).
-        */
-       c->always_chk_crc = 1;
+#ifndef __UBOOT__
+       if (!c->ro_mount) {
+               c->ileb_buf = vmalloc(c->leb_size);
+               if (!c->ileb_buf)
+                       goto out_free;
+       }
+#endif
+
+       if (c->bulk_read == 1)
+               bu_init(c);
+
+#ifndef __UBOOT__
+       if (!c->ro_mount) {
+               c->write_reserve_buf = kmalloc(COMPRESSED_DATA_NODE_BUF_SZ,
+                                              GFP_KERNEL);
+               if (!c->write_reserve_buf)
+                       goto out_free;
+       }
+#endif
+
+       c->mounting = 1;
 
        err = ubifs_read_superblock(c);
        if (err)
@@ -646,11 +1421,10 @@ static int mount_ubifs(struct ubifs_info *c)
        if (!ubifs_compr_present(c->default_compr)) {
                ubifs_err("'compressor \"%s\" is not compiled in",
                          ubifs_compr_name(c->default_compr));
+               err = -ENOTSUPP;
                goto out_free;
        }
 
-       dbg_failure_mode_registration(c);
-
        err = init_constants_sb(c);
        if (err)
                goto out_free;
@@ -663,7 +1437,25 @@ static int mount_ubifs(struct ubifs_info *c)
                goto out_free;
        }
 
+       err = alloc_wbufs(c);
+       if (err)
+               goto out_cbuf;
+
        sprintf(c->bgt_name, BGT_NAME_PATTERN, c->vi.ubi_num, c->vi.vol_id);
+#ifndef __UBOOT__
+       if (!c->ro_mount) {
+               /* Create background thread */
+               c->bgt = kthread_create(ubifs_bg_thread, c, "%s", c->bgt_name);
+               if (IS_ERR(c->bgt)) {
+                       err = PTR_ERR(c->bgt);
+                       c->bgt = NULL;
+                       ubifs_err("cannot spawn \"%s\", error %d",
+                                 c->bgt_name, err);
+                       goto out_wbufs;
+               }
+               wake_up_process(c->bgt);
+       }
+#endif
 
        err = ubifs_read_master(c);
        if (err)
@@ -676,204 +1468,694 @@ static int mount_ubifs(struct ubifs_info *c)
                c->need_recovery = 1;
        }
 
-       err = ubifs_lpt_init(c, 1, !mounted_read_only);
+#ifndef __UBOOT__
+       if (c->need_recovery && !c->ro_mount) {
+               err = ubifs_recover_inl_heads(c, c->sbuf);
+               if (err)
+                       goto out_master;
+       }
+#endif
+
+       err = ubifs_lpt_init(c, 1, !c->ro_mount);
        if (err)
-               goto out_lpt;
+               goto out_master;
+
+#ifndef __UBOOT__
+       if (!c->ro_mount && c->space_fixup) {
+               err = ubifs_fixup_free_space(c);
+               if (err)
+                       goto out_lpt;
+       }
+
+       if (!c->ro_mount) {
+               /*
+                * Set the "dirty" flag so that if we reboot uncleanly we
+                * will notice this immediately on the next mount.
+                */
+               c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
+               err = ubifs_write_master(c);
+               if (err)
+                       goto out_lpt;
+       }
+#endif
 
-       err = dbg_check_idx_size(c, c->old_idx_sz);
+       err = dbg_check_idx_size(c, c->bi.old_idx_sz);
        if (err)
                goto out_lpt;
 
+#ifndef __UBOOT__
        err = ubifs_replay_journal(c);
        if (err)
                goto out_journal;
+#endif
+
+       /* Calculate 'min_idx_lebs' after journal replay */
+       c->bi.min_idx_lebs = ubifs_calc_min_idx_lebs(c);
 
-       err = ubifs_mount_orphans(c, c->need_recovery, mounted_read_only);
+       err = ubifs_mount_orphans(c, c->need_recovery, c->ro_mount);
        if (err)
                goto out_orphans;
 
-       if (c->need_recovery) {
+       if (!c->ro_mount) {
+#ifndef __UBOOT__
+               int lnum;
+
+               err = check_free_space(c);
+               if (err)
+                       goto out_orphans;
+
+               /* Check for enough log space */
+               lnum = c->lhead_lnum + 1;
+               if (lnum >= UBIFS_LOG_LNUM + c->log_lebs)
+                       lnum = UBIFS_LOG_LNUM;
+               if (lnum == c->ltail_lnum) {
+                       err = ubifs_consolidate_log(c);
+                       if (err)
+                               goto out_orphans;
+               }
+
+               if (c->need_recovery) {
+                       err = ubifs_recover_size(c);
+                       if (err)
+                               goto out_orphans;
+                       err = ubifs_rcvry_gc_commit(c);
+                       if (err)
+                               goto out_orphans;
+               } else {
+                       err = take_gc_lnum(c);
+                       if (err)
+                               goto out_orphans;
+
+                       /*
+                        * GC LEB may contain garbage if there was an unclean
+                        * reboot, and it should be un-mapped.
+                        */
+                       err = ubifs_leb_unmap(c, c->gc_lnum);
+                       if (err)
+                               goto out_orphans;
+               }
+
+               err = dbg_check_lprops(c);
+               if (err)
+                       goto out_orphans;
+#endif
+       } else if (c->need_recovery) {
                err = ubifs_recover_size(c);
                if (err)
                        goto out_orphans;
+       } else {
+               /*
+                * Even if we mount read-only, we have to set space in GC LEB
+                * to proper value because this affects UBIFS free space
+                * reporting. We do not want to have a situation when
+                * re-mounting from R/O to R/W changes amount of free space.
+                */
+               err = take_gc_lnum(c);
+               if (err)
+                       goto out_orphans;
        }
 
+#ifndef __UBOOT__
        spin_lock(&ubifs_infos_lock);
        list_add_tail(&c->infos_list, &ubifs_infos);
        spin_unlock(&ubifs_infos_lock);
+#endif
 
        if (c->need_recovery) {
-               if (mounted_read_only)
+               if (c->ro_mount)
                        ubifs_msg("recovery deferred");
                else {
                        c->need_recovery = 0;
                        ubifs_msg("recovery completed");
+                       /*
+                        * GC LEB has to be empty and taken at this point. But
+                        * the journal head LEBs may also be accounted as
+                        * "empty taken" if they are empty.
+                        */
+                       ubifs_assert(c->lst.taken_empty_lebs > 0);
                }
-       }
+       } else
+               ubifs_assert(c->lst.taken_empty_lebs > 0);
 
        err = dbg_check_filesystem(c);
        if (err)
                goto out_infos;
 
-       c->always_chk_crc = 0;
+       err = dbg_debugfs_init_fs(c);
+       if (err)
+               goto out_infos;
+
+       c->mounting = 0;
 
-       ubifs_msg("mounted UBI device %d, volume %d, name \"%s\"",
-                 c->vi.ubi_num, c->vi.vol_id, c->vi.name);
-       if (mounted_read_only)
-               ubifs_msg("mounted read-only");
+       ubifs_msg("mounted UBI device %d, volume %d, name \"%s\"%s",
+                 c->vi.ubi_num, c->vi.vol_id, c->vi.name,
+                 c->ro_mount ? ", R/O mode" : "");
        x = (long long)c->main_lebs * c->leb_size;
-       ubifs_msg("file system size:   %lld bytes (%lld KiB, %lld MiB, %d "
-                 "LEBs)", x, x >> 10, x >> 20, c->main_lebs);
-       x = (long long)c->log_lebs * c->leb_size + c->max_bud_bytes;
-       ubifs_msg("journal size:       %lld bytes (%lld KiB, %lld MiB, %d "
-                 "LEBs)", x, x >> 10, x >> 20, c->log_lebs + c->max_bud_cnt);
-       ubifs_msg("media format:       w%d/r%d (latest is w%d/r%d)",
+       y = (long long)c->log_lebs * c->leb_size + c->max_bud_bytes;
+       ubifs_msg("LEB size: %d bytes (%d KiB), min./max. I/O unit sizes: %d bytes/%d bytes",
+                 c->leb_size, c->leb_size >> 10, c->min_io_size,
+                 c->max_write_size);
+       ubifs_msg("FS size: %lld bytes (%lld MiB, %d LEBs), journal size %lld bytes (%lld MiB, %d LEBs)",
+                 x, x >> 20, c->main_lebs,
+                 y, y >> 20, c->log_lebs + c->max_bud_cnt);
+       ubifs_msg("reserved for root: %llu bytes (%llu KiB)",
+                 c->report_rp_size, c->report_rp_size >> 10);
+       ubifs_msg("media format: w%d/r%d (latest is w%d/r%d), UUID %pUB%s",
                  c->fmt_version, c->ro_compat_version,
-                 UBIFS_FORMAT_VERSION, UBIFS_RO_COMPAT_VERSION);
-       ubifs_msg("default compressor: %s", ubifs_compr_name(c->default_compr));
-       ubifs_msg("reserved for root:  %llu bytes (%llu KiB)",
-               c->report_rp_size, c->report_rp_size >> 10);
-
-       dbg_msg("min. I/O unit size:  %d bytes", c->min_io_size);
-       dbg_msg("LEB size:            %d bytes (%d KiB)",
-               c->leb_size, c->leb_size >> 10);
-       dbg_msg("data journal heads:  %d",
+                 UBIFS_FORMAT_VERSION, UBIFS_RO_COMPAT_VERSION, c->uuid,
+                 c->big_lpt ? ", big LPT model" : ", small LPT model");
+
+       dbg_gen("default compressor:  %s", ubifs_compr_name(c->default_compr));
+       dbg_gen("data journal heads:  %d",
                c->jhead_cnt - NONDATA_JHEADS_CNT);
-       dbg_msg("UUID:                %02X%02X%02X%02X-%02X%02X"
-              "-%02X%02X-%02X%02X-%02X%02X%02X%02X%02X%02X",
-              c->uuid[0], c->uuid[1], c->uuid[2], c->uuid[3],
-              c->uuid[4], c->uuid[5], c->uuid[6], c->uuid[7],
-              c->uuid[8], c->uuid[9], c->uuid[10], c->uuid[11],
-              c->uuid[12], c->uuid[13], c->uuid[14], c->uuid[15]);
-       dbg_msg("big_lpt              %d", c->big_lpt);
-       dbg_msg("log LEBs:            %d (%d - %d)",
+       dbg_gen("log LEBs:            %d (%d - %d)",
                c->log_lebs, UBIFS_LOG_LNUM, c->log_last);
-       dbg_msg("LPT area LEBs:       %d (%d - %d)",
+       dbg_gen("LPT area LEBs:       %d (%d - %d)",
                c->lpt_lebs, c->lpt_first, c->lpt_last);
-       dbg_msg("orphan area LEBs:    %d (%d - %d)",
+       dbg_gen("orphan area LEBs:    %d (%d - %d)",
                c->orph_lebs, c->orph_first, c->orph_last);
-       dbg_msg("main area LEBs:      %d (%d - %d)",
+       dbg_gen("main area LEBs:      %d (%d - %d)",
                c->main_lebs, c->main_first, c->leb_cnt - 1);
-       dbg_msg("index LEBs:          %d", c->lst.idx_lebs);
-       dbg_msg("total index bytes:   %lld (%lld KiB, %lld MiB)",
-               c->old_idx_sz, c->old_idx_sz >> 10, c->old_idx_sz >> 20);
-       dbg_msg("key hash type:       %d", c->key_hash_type);
-       dbg_msg("tree fanout:         %d", c->fanout);
-       dbg_msg("reserved GC LEB:     %d", c->gc_lnum);
-       dbg_msg("first main LEB:      %d", c->main_first);
-       dbg_msg("max. znode size      %d", c->max_znode_sz);
-       dbg_msg("max. index node size %d", c->max_idx_node_sz);
-       dbg_msg("node sizes:          data %zu, inode %zu, dentry %zu",
+       dbg_gen("index LEBs:          %d", c->lst.idx_lebs);
+       dbg_gen("total index bytes:   %lld (%lld KiB, %lld MiB)",
+               c->bi.old_idx_sz, c->bi.old_idx_sz >> 10,
+               c->bi.old_idx_sz >> 20);
+       dbg_gen("key hash type:       %d", c->key_hash_type);
+       dbg_gen("tree fanout:         %d", c->fanout);
+       dbg_gen("reserved GC LEB:     %d", c->gc_lnum);
+       dbg_gen("max. znode size      %d", c->max_znode_sz);
+       dbg_gen("max. index node size %d", c->max_idx_node_sz);
+       dbg_gen("node sizes:          data %zu, inode %zu, dentry %zu",
                UBIFS_DATA_NODE_SZ, UBIFS_INO_NODE_SZ, UBIFS_DENT_NODE_SZ);
-       dbg_msg("node sizes:          trun %zu, sb %zu, master %zu",
+       dbg_gen("node sizes:          trun %zu, sb %zu, master %zu",
                UBIFS_TRUN_NODE_SZ, UBIFS_SB_NODE_SZ, UBIFS_MST_NODE_SZ);
-       dbg_msg("node sizes:          ref %zu, cmt. start %zu, orph %zu",
+       dbg_gen("node sizes:          ref %zu, cmt. start %zu, orph %zu",
                UBIFS_REF_NODE_SZ, UBIFS_CS_NODE_SZ, UBIFS_ORPH_NODE_SZ);
-       dbg_msg("max. node sizes:     data %zu, inode %zu dentry %zu",
+       dbg_gen("max. node sizes:     data %zu, inode %zu dentry %zu, idx %d",
                UBIFS_MAX_DATA_NODE_SZ, UBIFS_MAX_INO_NODE_SZ,
-               UBIFS_MAX_DENT_NODE_SZ);
-       dbg_msg("dead watermark:      %d", c->dead_wm);
-       dbg_msg("dark watermark:      %d", c->dark_wm);
-       dbg_msg("LEB overhead:        %d", c->leb_overhead);
+               UBIFS_MAX_DENT_NODE_SZ, ubifs_idx_node_sz(c, c->fanout));
+       dbg_gen("dead watermark:      %d", c->dead_wm);
+       dbg_gen("dark watermark:      %d", c->dark_wm);
+       dbg_gen("LEB overhead:        %d", c->leb_overhead);
        x = (long long)c->main_lebs * c->dark_wm;
-       dbg_msg("max. dark space:     %lld (%lld KiB, %lld MiB)",
+       dbg_gen("max. dark space:     %lld (%lld KiB, %lld MiB)",
                x, x >> 10, x >> 20);
-       dbg_msg("maximum bud bytes:   %lld (%lld KiB, %lld MiB)",
+       dbg_gen("maximum bud bytes:   %lld (%lld KiB, %lld MiB)",
                c->max_bud_bytes, c->max_bud_bytes >> 10,
                c->max_bud_bytes >> 20);
-       dbg_msg("BG commit bud bytes: %lld (%lld KiB, %lld MiB)",
+       dbg_gen("BG commit bud bytes: %lld (%lld KiB, %lld MiB)",
                c->bg_bud_bytes, c->bg_bud_bytes >> 10,
                c->bg_bud_bytes >> 20);
-       dbg_msg("current bud bytes    %lld (%lld KiB, %lld MiB)",
+       dbg_gen("current bud bytes    %lld (%lld KiB, %lld MiB)",
                c->bud_bytes, c->bud_bytes >> 10, c->bud_bytes >> 20);
-       dbg_msg("max. seq. number:    %llu", c->max_sqnum);
-       dbg_msg("commit number:       %llu", c->cmt_no);
+       dbg_gen("max. seq. number:    %llu", c->max_sqnum);
+       dbg_gen("commit number:       %llu", c->cmt_no);
+
+       return 0;
+
+out_infos:
+       spin_lock(&ubifs_infos_lock);
+       list_del(&c->infos_list);
+       spin_unlock(&ubifs_infos_lock);
+out_orphans:
+       free_orphans(c);
+#ifndef __UBOOT__
+out_journal:
+       destroy_journal(c);
+#endif
+out_lpt:
+       ubifs_lpt_free(c, 0);
+out_master:
+       kfree(c->mst_node);
+       kfree(c->rcvrd_mst_node);
+       if (c->bgt)
+               kthread_stop(c->bgt);
+#ifndef __UBOOT__
+out_wbufs:
+#endif
+       free_wbufs(c);
+out_cbuf:
+       kfree(c->cbuf);
+out_free:
+       kfree(c->write_reserve_buf);
+       kfree(c->bu.buf);
+       vfree(c->ileb_buf);
+       vfree(c->sbuf);
+       kfree(c->bottom_up_buf);
+       ubifs_debugging_exit(c);
+       return err;
+}
+
+/**
+ * ubifs_umount - un-mount UBIFS file-system.
+ * @c: UBIFS file-system description object
+ *
+ * Note, this function is called to free allocated resourced when un-mounting,
+ * as well as free resources when an error occurred while we were half way
+ * through mounting (error path cleanup function). So it has to make sure the
+ * resource was actually allocated before freeing it.
+ */
+#ifndef __UBOOT__
+static void ubifs_umount(struct ubifs_info *c)
+#else
+void ubifs_umount(struct ubifs_info *c)
+#endif
+{
+       dbg_gen("un-mounting UBI device %d, volume %d", c->vi.ubi_num,
+               c->vi.vol_id);
+
+       dbg_debugfs_exit_fs(c);
+       spin_lock(&ubifs_infos_lock);
+       list_del(&c->infos_list);
+       spin_unlock(&ubifs_infos_lock);
+
+#ifndef __UBOOT__
+       if (c->bgt)
+               kthread_stop(c->bgt);
+
+       destroy_journal(c);
+#endif
+       free_wbufs(c);
+       free_orphans(c);
+       ubifs_lpt_free(c, 0);
+
+       kfree(c->cbuf);
+       kfree(c->rcvrd_mst_node);
+       kfree(c->mst_node);
+       kfree(c->write_reserve_buf);
+       kfree(c->bu.buf);
+       vfree(c->ileb_buf);
+       vfree(c->sbuf);
+       kfree(c->bottom_up_buf);
+       ubifs_debugging_exit(c);
+#ifdef __UBOOT__
+       /* Finally free U-Boot's global copy of superblock */
+       if (ubifs_sb != NULL) {
+               free(ubifs_sb->s_fs_info);
+               free(ubifs_sb);
+       }
+#endif
+}
+
+#ifndef __UBOOT__
+/**
+ * ubifs_remount_rw - re-mount in read-write mode.
+ * @c: UBIFS file-system description object
+ *
+ * UBIFS avoids allocating many unnecessary resources when mounted in read-only
+ * mode. This function allocates the needed resources and re-mounts UBIFS in
+ * read-write mode.
+ */
+static int ubifs_remount_rw(struct ubifs_info *c)
+{
+       int err, lnum;
+
+       if (c->rw_incompat) {
+               ubifs_err("the file-system is not R/W-compatible");
+               ubifs_msg("on-flash format version is w%d/r%d, but software only supports up to version w%d/r%d",
+                         c->fmt_version, c->ro_compat_version,
+                         UBIFS_FORMAT_VERSION, UBIFS_RO_COMPAT_VERSION);
+               return -EROFS;
+       }
+
+       mutex_lock(&c->umount_mutex);
+       dbg_save_space_info(c);
+       c->remounting_rw = 1;
+       c->ro_mount = 0;
+
+       if (c->space_fixup) {
+               err = ubifs_fixup_free_space(c);
+               if (err)
+                       return err;
+       }
+
+       err = check_free_space(c);
+       if (err)
+               goto out;
+
+       if (c->old_leb_cnt != c->leb_cnt) {
+               struct ubifs_sb_node *sup;
+
+               sup = ubifs_read_sb_node(c);
+               if (IS_ERR(sup)) {
+                       err = PTR_ERR(sup);
+                       goto out;
+               }
+               sup->leb_cnt = cpu_to_le32(c->leb_cnt);
+               err = ubifs_write_sb_node(c, sup);
+               kfree(sup);
+               if (err)
+                       goto out;
+       }
+
+       if (c->need_recovery) {
+               ubifs_msg("completing deferred recovery");
+               err = ubifs_write_rcvrd_mst_node(c);
+               if (err)
+                       goto out;
+               err = ubifs_recover_size(c);
+               if (err)
+                       goto out;
+               err = ubifs_clean_lebs(c, c->sbuf);
+               if (err)
+                       goto out;
+               err = ubifs_recover_inl_heads(c, c->sbuf);
+               if (err)
+                       goto out;
+       } else {
+               /* A readonly mount is not allowed to have orphans */
+               ubifs_assert(c->tot_orphans == 0);
+               err = ubifs_clear_orphans(c);
+               if (err)
+                       goto out;
+       }
+
+       if (!(c->mst_node->flags & cpu_to_le32(UBIFS_MST_DIRTY))) {
+               c->mst_node->flags |= cpu_to_le32(UBIFS_MST_DIRTY);
+               err = ubifs_write_master(c);
+               if (err)
+                       goto out;
+       }
+
+       c->ileb_buf = vmalloc(c->leb_size);
+       if (!c->ileb_buf) {
+               err = -ENOMEM;
+               goto out;
+       }
+
+       c->write_reserve_buf = kmalloc(COMPRESSED_DATA_NODE_BUF_SZ, GFP_KERNEL);
+       if (!c->write_reserve_buf) {
+               err = -ENOMEM;
+               goto out;
+       }
+
+       err = ubifs_lpt_init(c, 0, 1);
+       if (err)
+               goto out;
+
+       /* Create background thread */
+       c->bgt = kthread_create(ubifs_bg_thread, c, "%s", c->bgt_name);
+       if (IS_ERR(c->bgt)) {
+               err = PTR_ERR(c->bgt);
+               c->bgt = NULL;
+               ubifs_err("cannot spawn \"%s\", error %d",
+                         c->bgt_name, err);
+               goto out;
+       }
+       wake_up_process(c->bgt);
+
+       c->orph_buf = vmalloc(c->leb_size);
+       if (!c->orph_buf) {
+               err = -ENOMEM;
+               goto out;
+       }
 
-       return 0;
+       /* Check for enough log space */
+       lnum = c->lhead_lnum + 1;
+       if (lnum >= UBIFS_LOG_LNUM + c->log_lebs)
+               lnum = UBIFS_LOG_LNUM;
+       if (lnum == c->ltail_lnum) {
+               err = ubifs_consolidate_log(c);
+               if (err)
+                       goto out;
+       }
 
-out_infos:
-       spin_lock(&ubifs_infos_lock);
-       list_del(&c->infos_list);
-       spin_unlock(&ubifs_infos_lock);
-out_orphans:
-       free_orphans(c);
-out_journal:
-out_lpt:
-       ubifs_lpt_free(c, 0);
-out_master:
-       kfree(c->mst_node);
-       kfree(c->rcvrd_mst_node);
-       if (c->bgt)
+       if (c->need_recovery)
+               err = ubifs_rcvry_gc_commit(c);
+       else
+               err = ubifs_leb_unmap(c, c->gc_lnum);
+       if (err)
+               goto out;
+
+       dbg_gen("re-mounted read-write");
+       c->remounting_rw = 0;
+
+       if (c->need_recovery) {
+               c->need_recovery = 0;
+               ubifs_msg("deferred recovery completed");
+       } else {
+               /*
+                * Do not run the debugging space check if the were doing
+                * recovery, because when we saved the information we had the
+                * file-system in a state where the TNC and lprops has been
+                * modified in memory, but all the I/O operations (including a
+                * commit) were deferred. So the file-system was in
+                * "non-committed" state. Now the file-system is in committed
+                * state, and of course the amount of free space will change
+                * because, for example, the old index size was imprecise.
+                */
+               err = dbg_check_space_info(c);
+       }
+
+       mutex_unlock(&c->umount_mutex);
+       return err;
+
+out:
+       c->ro_mount = 1;
+       vfree(c->orph_buf);
+       c->orph_buf = NULL;
+       if (c->bgt) {
                kthread_stop(c->bgt);
-       kfree(c->cbuf);
-out_free:
+               c->bgt = NULL;
+       }
+       free_wbufs(c);
+       kfree(c->write_reserve_buf);
+       c->write_reserve_buf = NULL;
        vfree(c->ileb_buf);
-       vfree(c->sbuf);
-       kfree(c->bottom_up_buf);
-       ubifs_debugging_exit(c);
+       c->ileb_buf = NULL;
+       ubifs_lpt_free(c, 1);
+       c->remounting_rw = 0;
+       mutex_unlock(&c->umount_mutex);
        return err;
 }
 
 /**
- * ubifs_umount - un-mount UBIFS file-system.
+ * ubifs_remount_ro - re-mount in read-only mode.
  * @c: UBIFS file-system description object
  *
- * Note, this function is called to free allocated resourced when un-mounting,
- * as well as free resources when an error occurred while we were half way
- * through mounting (error path cleanup function). So it has to make sure the
- * resource was actually allocated before freeing it.
+ * We assume VFS has stopped writing. Possibly the background thread could be
+ * running a commit, however kthread_stop will wait in that case.
  */
-void ubifs_umount(struct ubifs_info *c)
+static void ubifs_remount_ro(struct ubifs_info *c)
 {
-       dbg_gen("un-mounting UBI device %d, volume %d", c->vi.ubi_num,
-               c->vi.vol_id);
+       int i, err;
 
-       spin_lock(&ubifs_infos_lock);
-       list_del(&c->infos_list);
-       spin_unlock(&ubifs_infos_lock);
+       ubifs_assert(!c->need_recovery);
+       ubifs_assert(!c->ro_mount);
 
-       if (c->bgt)
+       mutex_lock(&c->umount_mutex);
+       if (c->bgt) {
                kthread_stop(c->bgt);
+               c->bgt = NULL;
+       }
 
-       free_orphans(c);
-       ubifs_lpt_free(c, 0);
+       dbg_save_space_info(c);
 
-       kfree(c->cbuf);
-       kfree(c->rcvrd_mst_node);
-       kfree(c->mst_node);
+       for (i = 0; i < c->jhead_cnt; i++)
+               ubifs_wbuf_sync(&c->jheads[i].wbuf);
+
+       c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
+       c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
+       c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
+       err = ubifs_write_master(c);
+       if (err)
+               ubifs_ro_mode(c, err);
+
+       vfree(c->orph_buf);
+       c->orph_buf = NULL;
+       kfree(c->write_reserve_buf);
+       c->write_reserve_buf = NULL;
        vfree(c->ileb_buf);
-       vfree(c->sbuf);
-       kfree(c->bottom_up_buf);
-       ubifs_debugging_exit(c);
+       c->ileb_buf = NULL;
+       ubifs_lpt_free(c, 1);
+       c->ro_mount = 1;
+       err = dbg_check_space_info(c);
+       if (err)
+               ubifs_ro_mode(c, err);
+       mutex_unlock(&c->umount_mutex);
+}
 
-       /* Finally free U-Boot's global copy of superblock */
-       if (ubifs_sb != NULL) {
-               free(ubifs_sb->s_fs_info);
-               free(ubifs_sb);
+static void ubifs_put_super(struct super_block *sb)
+{
+       int i;
+       struct ubifs_info *c = sb->s_fs_info;
+
+       ubifs_msg("un-mount UBI device %d, volume %d", c->vi.ubi_num,
+                 c->vi.vol_id);
+
+       /*
+        * The following asserts are only valid if there has not been a failure
+        * of the media. For example, there will be dirty inodes if we failed
+        * to write them back because of I/O errors.
+        */
+       if (!c->ro_error) {
+               ubifs_assert(c->bi.idx_growth == 0);
+               ubifs_assert(c->bi.dd_growth == 0);
+               ubifs_assert(c->bi.data_growth == 0);
+       }
+
+       /*
+        * The 'c->umount_lock' prevents races between UBIFS memory shrinker
+        * and file system un-mount. Namely, it prevents the shrinker from
+        * picking this superblock for shrinking - it will be just skipped if
+        * the mutex is locked.
+        */
+       mutex_lock(&c->umount_mutex);
+       if (!c->ro_mount) {
+               /*
+                * First of all kill the background thread to make sure it does
+                * not interfere with un-mounting and freeing resources.
+                */
+               if (c->bgt) {
+                       kthread_stop(c->bgt);
+                       c->bgt = NULL;
+               }
+
+               /*
+                * On fatal errors c->ro_error is set to 1, in which case we do
+                * not write the master node.
+                */
+               if (!c->ro_error) {
+                       int err;
+
+                       /* Synchronize write-buffers */
+                       for (i = 0; i < c->jhead_cnt; i++)
+                               ubifs_wbuf_sync(&c->jheads[i].wbuf);
+
+                       /*
+                        * We are being cleanly unmounted which means the
+                        * orphans were killed - indicate this in the master
+                        * node. Also save the reserved GC LEB number.
+                        */
+                       c->mst_node->flags &= ~cpu_to_le32(UBIFS_MST_DIRTY);
+                       c->mst_node->flags |= cpu_to_le32(UBIFS_MST_NO_ORPHS);
+                       c->mst_node->gc_lnum = cpu_to_le32(c->gc_lnum);
+                       err = ubifs_write_master(c);
+                       if (err)
+                               /*
+                                * Recovery will attempt to fix the master area
+                                * next mount, so we just print a message and
+                                * continue to unmount normally.
+                                */
+                               ubifs_err("failed to write master node, error %d",
+                                         err);
+               } else {
+#ifndef __UBOOT__
+                       for (i = 0; i < c->jhead_cnt; i++)
+                               /* Make sure write-buffer timers are canceled */
+                               hrtimer_cancel(&c->jheads[i].wbuf.timer);
+#endif
+               }
+       }
+
+       ubifs_umount(c);
+#ifndef __UBOOT__
+       bdi_destroy(&c->bdi);
+#endif
+       ubi_close_volume(c->ubi);
+       mutex_unlock(&c->umount_mutex);
+}
+#endif
+
+#ifndef __UBOOT__
+static int ubifs_remount_fs(struct super_block *sb, int *flags, char *data)
+{
+       int err;
+       struct ubifs_info *c = sb->s_fs_info;
+
+       dbg_gen("old flags %#lx, new flags %#x", sb->s_flags, *flags);
+
+       err = ubifs_parse_options(c, data, 1);
+       if (err) {
+               ubifs_err("invalid or unknown remount parameter");
+               return err;
+       }
+
+       if (c->ro_mount && !(*flags & MS_RDONLY)) {
+               if (c->ro_error) {
+                       ubifs_msg("cannot re-mount R/W due to prior errors");
+                       return -EROFS;
+               }
+               if (c->ro_media) {
+                       ubifs_msg("cannot re-mount R/W - UBI volume is R/O");
+                       return -EROFS;
+               }
+               err = ubifs_remount_rw(c);
+               if (err)
+                       return err;
+       } else if (!c->ro_mount && (*flags & MS_RDONLY)) {
+               if (c->ro_error) {
+                       ubifs_msg("cannot re-mount R/O due to prior errors");
+                       return -EROFS;
+               }
+               ubifs_remount_ro(c);
+       }
+
+       if (c->bulk_read == 1)
+               bu_init(c);
+       else {
+               dbg_gen("disable bulk-read");
+               kfree(c->bu.buf);
+               c->bu.buf = NULL;
        }
+
+       ubifs_assert(c->lst.taken_empty_lebs > 0);
+       return 0;
 }
+#endif
+
+const struct super_operations ubifs_super_operations = {
+       .alloc_inode   = ubifs_alloc_inode,
+#ifndef __UBOOT__
+       .destroy_inode = ubifs_destroy_inode,
+       .put_super     = ubifs_put_super,
+       .write_inode   = ubifs_write_inode,
+       .evict_inode   = ubifs_evict_inode,
+       .statfs        = ubifs_statfs,
+#endif
+       .dirty_inode   = ubifs_dirty_inode,
+#ifndef __UBOOT__
+       .remount_fs    = ubifs_remount_fs,
+       .show_options  = ubifs_show_options,
+       .sync_fs       = ubifs_sync_fs,
+#endif
+};
 
 /**
  * open_ubi - parse UBI device name string and open the UBI device.
  * @name: UBI volume name
  * @mode: UBI volume open mode
  *
- * There are several ways to specify UBI volumes when mounting UBIFS:
- * o ubiX_Y    - UBI device number X, volume Y;
- * o ubiY      - UBI device number 0, volume Y;
+ * The primary method of mounting UBIFS is by specifying the UBI volume
+ * character device node path. However, UBIFS may also be mounted withoug any
+ * character device node using one of the following methods:
+ *
+ * o ubiX_Y    - mount UBI device number X, volume Y;
+ * o ubiY      - mount UBI device number 0, volume Y;
  * o ubiX:NAME - mount UBI device X, volume with name NAME;
  * o ubi:NAME  - mount UBI device 0, volume with name NAME.
  *
  * Alternative '!' separator may be used instead of ':' (because some shells
  * like busybox may interpret ':' as an NFS host name separator). This function
- * returns ubi volume object in case of success and a negative error code in
- * case of failure.
+ * returns UBI volume description object in case of success and a negative
+ * error code in case of failure.
  */
 static struct ubi_volume_desc *open_ubi(const char *name, int mode)
 {
+#ifndef __UBOOT__
+       struct ubi_volume_desc *ubi;
+#endif
        int dev, vol;
        char *endptr;
 
+#ifndef __UBOOT__
+       /* First, try to open using the device node path method */
+       ubi = ubi_open_volume_path(name, mode);
+       if (!IS_ERR(ubi))
+               return ubi;
+#endif
+
+       /* Try the "nodev" method */
        if (name[0] != 'u' || name[1] != 'b' || name[2] != 'i')
                return ERR_PTR(-EINVAL);
 
@@ -905,78 +2187,100 @@ static struct ubi_volume_desc *open_ubi(const char *name, int mode)
        return ERR_PTR(-EINVAL);
 }
 
-static int ubifs_fill_super(struct super_block *sb, void *data, int silent)
+static struct ubifs_info *alloc_ubifs_info(struct ubi_volume_desc *ubi)
 {
-       struct ubi_volume_desc *ubi = sb->s_fs_info;
        struct ubifs_info *c;
-       struct inode *root;
-       int err;
 
        c = kzalloc(sizeof(struct ubifs_info), GFP_KERNEL);
-       if (!c)
-               return -ENOMEM;
+       if (c) {
+               spin_lock_init(&c->cnt_lock);
+               spin_lock_init(&c->cs_lock);
+               spin_lock_init(&c->buds_lock);
+               spin_lock_init(&c->space_lock);
+               spin_lock_init(&c->orphan_lock);
+               init_rwsem(&c->commit_sem);
+               mutex_init(&c->lp_mutex);
+               mutex_init(&c->tnc_mutex);
+               mutex_init(&c->log_mutex);
+               mutex_init(&c->mst_mutex);
+               mutex_init(&c->umount_mutex);
+               mutex_init(&c->bu_mutex);
+               mutex_init(&c->write_reserve_mutex);
+               init_waitqueue_head(&c->cmt_wq);
+               c->buds = RB_ROOT;
+               c->old_idx = RB_ROOT;
+               c->size_tree = RB_ROOT;
+               c->orph_tree = RB_ROOT;
+               INIT_LIST_HEAD(&c->infos_list);
+               INIT_LIST_HEAD(&c->idx_gc);
+               INIT_LIST_HEAD(&c->replay_list);
+               INIT_LIST_HEAD(&c->replay_buds);
+               INIT_LIST_HEAD(&c->uncat_list);
+               INIT_LIST_HEAD(&c->empty_list);
+               INIT_LIST_HEAD(&c->freeable_list);
+               INIT_LIST_HEAD(&c->frdi_idx_list);
+               INIT_LIST_HEAD(&c->unclean_leb_list);
+               INIT_LIST_HEAD(&c->old_buds);
+               INIT_LIST_HEAD(&c->orph_list);
+               INIT_LIST_HEAD(&c->orph_new);
+               c->no_chk_data_crc = 1;
+
+               c->highest_inum = UBIFS_FIRST_INO;
+               c->lhead_lnum = c->ltail_lnum = UBIFS_LOG_LNUM;
+
+               ubi_get_volume_info(ubi, &c->vi);
+               ubi_get_device_info(c->vi.ubi_num, &c->di);
+       }
+       return c;
+}
 
-       spin_lock_init(&c->cnt_lock);
-       spin_lock_init(&c->cs_lock);
-       spin_lock_init(&c->buds_lock);
-       spin_lock_init(&c->space_lock);
-       spin_lock_init(&c->orphan_lock);
-       init_rwsem(&c->commit_sem);
-       mutex_init(&c->lp_mutex);
-       mutex_init(&c->tnc_mutex);
-       mutex_init(&c->log_mutex);
-       mutex_init(&c->mst_mutex);
-       mutex_init(&c->umount_mutex);
-       init_waitqueue_head(&c->cmt_wq);
-       c->buds = RB_ROOT;
-       c->old_idx = RB_ROOT;
-       c->size_tree = RB_ROOT;
-       c->orph_tree = RB_ROOT;
-       INIT_LIST_HEAD(&c->infos_list);
-       INIT_LIST_HEAD(&c->idx_gc);
-       INIT_LIST_HEAD(&c->replay_list);
-       INIT_LIST_HEAD(&c->replay_buds);
-       INIT_LIST_HEAD(&c->uncat_list);
-       INIT_LIST_HEAD(&c->empty_list);
-       INIT_LIST_HEAD(&c->freeable_list);
-       INIT_LIST_HEAD(&c->frdi_idx_list);
-       INIT_LIST_HEAD(&c->unclean_leb_list);
-       INIT_LIST_HEAD(&c->old_buds);
-       INIT_LIST_HEAD(&c->orph_list);
-       INIT_LIST_HEAD(&c->orph_new);
-
-       c->highest_inum = UBIFS_FIRST_INO;
-       c->lhead_lnum = c->ltail_lnum = UBIFS_LOG_LNUM;
-
-       ubi_get_volume_info(ubi, &c->vi);
-       ubi_get_device_info(c->vi.ubi_num, &c->di);
+static int ubifs_fill_super(struct super_block *sb, void *data, int silent)
+{
+       struct ubifs_info *c = sb->s_fs_info;
+       struct inode *root;
+       int err;
 
+       c->vfs_sb = sb;
        /* Re-open the UBI device in read-write mode */
-       c->ubi = ubi_open_volume(c->vi.ubi_num, c->vi.vol_id, UBI_READONLY);
+       c->ubi = ubi_open_volume(c->vi.ubi_num, c->vi.vol_id, UBI_READWRITE);
        if (IS_ERR(c->ubi)) {
                err = PTR_ERR(c->ubi);
-               goto out_free;
+               goto out;
        }
 
-       c->vfs_sb = sb;
+#ifndef __UBOOT__
+       /*
+        * UBIFS provides 'backing_dev_info' in order to disable read-ahead. For
+        * UBIFS, I/O is not deferred, it is done immediately in readpage,
+        * which means the user would have to wait not just for their own I/O
+        * but the read-ahead I/O as well i.e. completely pointless.
+        *
+        * Read-ahead will be disabled because @c->bdi.ra_pages is 0.
+        */
+       co>bdi.name = "ubifs",
+       c->bdi.capabilities = BDI_CAP_MAP_COPY;
+       err  = bdi_init(&c->bdi);
+       if (err)
+               goto out_close;
+       err = bdi_register(&c->bdi, NULL, "ubifs_%d_%d",
+                          c->vi.ubi_num, c->vi.vol_id);
+       if (err)
+               goto out_bdi;
 
+       err = ubifs_parse_options(c, data, 0);
+       if (err)
+               goto out_bdi;
+
+       sb->s_bdi = &c->bdi;
+#endif
        sb->s_fs_info = c;
        sb->s_magic = UBIFS_SUPER_MAGIC;
        sb->s_blocksize = UBIFS_BLOCK_SIZE;
        sb->s_blocksize_bits = UBIFS_BLOCK_SHIFT;
-       sb->s_dev = c->vi.cdev;
        sb->s_maxbytes = c->max_inode_sz = key_max_inode_size(c);
        if (c->max_inode_sz > MAX_LFS_FILESIZE)
                sb->s_maxbytes = c->max_inode_sz = MAX_LFS_FILESIZE;
-
-       if (c->rw_incompat) {
-               ubifs_err("the file-system is not R/W-compatible");
-               ubifs_msg("on-flash format version is w%d/r%d, but software "
-                         "only supports up to version w%d/r%d", c->fmt_version,
-                         c->ro_compat_version, UBIFS_FORMAT_VERSION,
-                         UBIFS_RO_COMPAT_VERSION);
-               return -EROFS;
-       }
+       sb->s_op = &ubifs_super_operations;
 
        mutex_lock(&c->umount_mutex);
        err = mount_ubifs(c);
@@ -992,7 +2296,15 @@ static int ubifs_fill_super(struct super_block *sb, void *data, int silent)
                goto out_umount;
        }
 
+#ifndef __UBOOT__
+       sb->s_root = d_make_root(root);
+       if (!sb->s_root) {
+               err = -ENOMEM;
+               goto out_umount;
+       }
+#else
        sb->s_root = NULL;
+#endif
 
        mutex_unlock(&c->umount_mutex);
        return 0;
@@ -1001,24 +2313,130 @@ out_umount:
        ubifs_umount(c);
 out_unlock:
        mutex_unlock(&c->umount_mutex);
+#ifndef __UBOOT__
+out_bdi:
+       bdi_destroy(&c->bdi);
+out_close:
+#endif
        ubi_close_volume(c->ubi);
-out_free:
-       kfree(c);
+out:
        return err;
 }
 
 static int sb_test(struct super_block *sb, void *data)
 {
-       dev_t *dev = data;
+       struct ubifs_info *c1 = data;
+       struct ubifs_info *c = sb->s_fs_info;
+
+       return c->vi.cdev == c1->vi.cdev;
+}
+
+static int sb_set(struct super_block *sb, void *data)
+{
+       sb->s_fs_info = data;
+       return set_anon_super(sb, NULL);
+}
+
+static struct super_block *alloc_super(struct file_system_type *type, int flags)
+{
+       struct super_block *s;
+       int err;
+
+       s = kzalloc(sizeof(struct super_block),  GFP_USER);
+       if (!s) {
+               err = -ENOMEM;
+               return ERR_PTR(err);
+       }
+
+       INIT_HLIST_NODE(&s->s_instances);
+       INIT_LIST_HEAD(&s->s_inodes);
+       s->s_time_gran = 1000000000;
+       s->s_flags = flags;
+
+       return s;
+}
+
+/**
+ *     sget    -       find or create a superblock
+ *     @type:  filesystem type superblock should belong to
+ *     @test:  comparison callback
+ *     @set:   setup callback
+ *     @flags: mount flags
+ *     @data:  argument to each of them
+ */
+struct super_block *sget(struct file_system_type *type,
+                       int (*test)(struct super_block *,void *),
+                       int (*set)(struct super_block *,void *),
+                       int flags,
+                       void *data)
+{
+       struct super_block *s = NULL;
+#ifndef __UBOOT__
+       struct super_block *old;
+#endif
+       int err;
 
-       return sb->s_dev == *dev;
+#ifndef __UBOOT__
+retry:
+       spin_lock(&sb_lock);
+       if (test) {
+               hlist_for_each_entry(old, &type->fs_supers, s_instances) {
+                       if (!test(old, data))
+                               continue;
+                       if (!grab_super(old))
+                               goto retry;
+                       if (s) {
+                               up_write(&s->s_umount);
+                               destroy_super(s);
+                               s = NULL;
+                       }
+                       return old;
+               }
+       }
+#endif
+       if (!s) {
+               spin_unlock(&sb_lock);
+               s = alloc_super(type, flags);
+               if (!s)
+                       return ERR_PTR(-ENOMEM);
+#ifndef __UBOOT__
+               goto retry;
+#endif
+       }
+               
+       err = set(s, data);
+       if (err) {
+#ifndef __UBOOT__
+               spin_unlock(&sb_lock);
+               up_write(&s->s_umount);
+               destroy_super(s);
+#endif
+               return ERR_PTR(err);
+       }
+       s->s_type = type;
+#ifndef __UBOOT__
+       strlcpy(s->s_id, type->name, sizeof(s->s_id));
+#else
+       strncpy(s->s_id, type->name, sizeof(s->s_id));
+#endif
+       list_add_tail(&s->s_list, &super_blocks);
+       hlist_add_head(&s->s_instances, &type->fs_supers);
+#ifndef __UBOOT__
+       spin_unlock(&sb_lock);
+       get_filesystem(type);
+       register_shrinker(&s->s_shrink);
+#endif
+       return s;
 }
 
-static int ubifs_get_sb(struct file_system_type *fs_type, int flags,
-                       const char *name, void *data, struct vfsmount *mnt)
+EXPORT_SYMBOL(sget);
+
+
+static struct dentry *ubifs_mount(struct file_system_type *fs_type, int flags,
+                       const char *name, void *data)
 {
        struct ubi_volume_desc *ubi;
-       struct ubi_volume_info vi;
+       struct ubifs_info *c;
        struct super_block *sb;
        int err;
 
@@ -1033,32 +2451,34 @@ static int ubifs_get_sb(struct file_system_type *fs_type, int flags,
        if (IS_ERR(ubi)) {
                ubifs_err("cannot open \"%s\", error %d",
                          name, (int)PTR_ERR(ubi));
-               return PTR_ERR(ubi);
+               return ERR_CAST(ubi);
+       }
+
+       c = alloc_ubifs_info(ubi);
+       if (!c) {
+               err = -ENOMEM;
+               goto out_close;
        }
-       ubi_get_volume_info(ubi, &vi);
 
-       dbg_gen("opened ubi%d_%d", vi.ubi_num, vi.vol_id);
+       dbg_gen("opened ubi%d_%d", c->vi.ubi_num, c->vi.vol_id);
 
-       sb = sget(fs_type, &sb_test, &sb_set, &vi.cdev);
+       sb = sget(fs_type, sb_test, sb_set, flags, c);
        if (IS_ERR(sb)) {
                err = PTR_ERR(sb);
+               kfree(c);
                goto out_close;
        }
 
        if (sb->s_root) {
+               struct ubifs_info *c1 = sb->s_fs_info;
+               kfree(c);
                /* A new mount point for already mounted UBIFS */
                dbg_gen("this ubi volume is already mounted");
-               if ((flags ^ sb->s_flags) & MS_RDONLY) {
+               if (!!(flags & MS_RDONLY) != c1->ro_mount) {
                        err = -EBUSY;
                        goto out_deact;
                }
        } else {
-               sb->s_flags = flags;
-               /*
-                * Pass 'ubi' to 'fill_super()' in sb->s_fs_info where it is
-                * replaced by 'c'.
-                */
-               sb->s_fs_info = ubi;
                err = ubifs_fill_super(sb, data, flags & MS_SILENT ? 1 : 0);
                if (err)
                        goto out_deact;
@@ -1069,17 +2489,53 @@ static int ubifs_get_sb(struct file_system_type *fs_type, int flags,
        /* 'fill_super()' opens ubi again so we must close it here */
        ubi_close_volume(ubi);
 
+#ifdef __UBOOT__
        ubifs_sb = sb;
        return 0;
+#else
+       return dget(sb->s_root);
+#endif
 
 out_deact:
-       up_write(&sb->s_umount);
+#ifndef __UBOOT__
+       deactivate_locked_super(sb);
+#endif
 out_close:
        ubi_close_volume(ubi);
-       return err;
+       return ERR_PTR(err);
+}
+
+static void kill_ubifs_super(struct super_block *s)
+{
+       struct ubifs_info *c = s->s_fs_info;
+#ifndef __UBOOT__
+       kill_anon_super(s);
+#endif
+       kfree(c);
+}
+
+static struct file_system_type ubifs_fs_type = {
+       .name    = "ubifs",
+       .owner   = THIS_MODULE,
+       .mount   = ubifs_mount,
+       .kill_sb = kill_ubifs_super,
+};
+#ifndef __UBOOT__
+MODULE_ALIAS_FS("ubifs");
+
+/*
+ * Inode slab cache constructor.
+ */
+static void inode_slab_ctor(void *obj)
+{
+       struct ubifs_inode *ui = obj;
+       inode_init_once(&ui->vfs_inode);
 }
 
-int __init ubifs_init(void)
+static int __init ubifs_init(void)
+#else
+int ubifs_init(void)
+#endif
 {
        int err;
 
@@ -1135,41 +2591,84 @@ int __init ubifs_init(void)
         * UBIFS_BLOCK_SIZE. It is assumed that both are powers of 2.
         */
        if (PAGE_CACHE_SIZE < UBIFS_BLOCK_SIZE) {
-               ubifs_err("VFS page cache size is %u bytes, but UBIFS requires"
-                         " at least 4096 bytes",
+               ubifs_err("VFS page cache size is %u bytes, but UBIFS requires at least 4096 bytes",
                          (unsigned int)PAGE_CACHE_SIZE);
                return -EINVAL;
        }
 
-       err = -ENOMEM;
+#ifndef __UBOOT__
+       ubifs_inode_slab = kmem_cache_create("ubifs_inode_slab",
+                               sizeof(struct ubifs_inode), 0,
+                               SLAB_MEM_SPREAD | SLAB_RECLAIM_ACCOUNT,
+                               &inode_slab_ctor);
+       if (!ubifs_inode_slab)
+               return -ENOMEM;
+
+       register_shrinker(&ubifs_shrinker_info);
+#endif
 
        err = ubifs_compressors_init();
        if (err)
                goto out_shrinker;
 
+#ifndef __UBOOT__
+       err = dbg_debugfs_init();
+       if (err)
+               goto out_compr;
+
+       err = register_filesystem(&ubifs_fs_type);
+       if (err) {
+               ubifs_err("cannot register file system, error %d", err);
+               goto out_dbg;
+       }
+#endif
        return 0;
 
+#ifndef __UBOOT__
+out_dbg:
+       dbg_debugfs_exit();
+out_compr:
+       ubifs_compressors_exit();
+#endif
 out_shrinker:
+#ifndef __UBOOT__
+       unregister_shrinker(&ubifs_shrinker_info);
+#endif
+       kmem_cache_destroy(ubifs_inode_slab);
        return err;
 }
+/* late_initcall to let compressors initialize first */
+late_initcall(ubifs_init);
 
-/*
- * ubifsmount...
- */
+#ifndef __UBOOT__
+static void __exit ubifs_exit(void)
+{
+       ubifs_assert(list_empty(&ubifs_infos));
+       ubifs_assert(atomic_long_read(&ubifs_clean_zn_cnt) == 0);
 
-static struct file_system_type ubifs_fs_type = {
-       .name    = "ubifs",
-       .owner   = THIS_MODULE,
-       .get_sb  = ubifs_get_sb,
-};
+       dbg_debugfs_exit();
+       ubifs_compressors_exit();
+       unregister_shrinker(&ubifs_shrinker_info);
 
-int ubifs_mount(char *name)
+       /*
+        * Make sure all delayed rcu free inodes are flushed before we
+        * destroy cache.
+        */
+       rcu_barrier();
+       kmem_cache_destroy(ubifs_inode_slab);
+       unregister_filesystem(&ubifs_fs_type);
+}
+module_exit(ubifs_exit);
+
+MODULE_LICENSE("GPL");
+MODULE_VERSION(__stringify(UBIFS_VERSION));
+MODULE_AUTHOR("Artem Bityutskiy, Adrian Hunter");
+MODULE_DESCRIPTION("UBIFS - UBI File System");
+#else
+int uboot_ubifs_mount(char *vol_name)
 {
+       struct dentry *ret;
        int flags;
-       void *data;
-       struct vfsmount *mnt;
-       int ret;
-       struct ubifs_info *c;
 
        /*
         * First unmount if allready mounted
@@ -1177,23 +2676,17 @@ int ubifs_mount(char *name)
        if (ubifs_sb)
                ubifs_umount(ubifs_sb->s_fs_info);
 
-       INIT_LIST_HEAD(&ubifs_infos);
-       INIT_LIST_HEAD(&ubifs_fs_type.fs_supers);
-
        /*
         * Mount in read-only mode
         */
        flags = MS_RDONLY;
-       data = NULL;
-       mnt = NULL;
-       ret = ubifs_get_sb(&ubifs_fs_type, flags, name, data, mnt);
-       if (ret) {
-               ubifs_err("Error reading superblock on volume '%s' errno=%d!\n", name, ret);
+       ret = ubifs_mount(&ubifs_fs_type, flags, vol_name, NULL);
+       if (IS_ERR(ret)) {
+               printf("Error reading superblock on volume '%s' " \
+                       "errno=%d!\n", vol_name, (int)PTR_ERR(ret));
                return -1;
        }
 
-       c = ubifs_sb->s_fs_info;
-       ubi_close_volume(c->ubi);
-
        return 0;
 }
+#endif
index ccda9387bc8f68d428e454042e43f8a682db309a..eda50705574d9c063bb14d77d427e1ff33631973 100644 (file)
@@ -3,18 +3,7 @@
  *
  * Copyright (C) 2006-2008 Nokia Corporation.
  *
- * This program is free software; you can redistribute it and/or modify it
- * under the terms of the GNU General Public License version 2 as published by
- * the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
- * more details.
- *
- * You should have received a copy of the GNU General Public License along with
- * this program; if not, write to the Free Software Foundation, Inc., 51
- * Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
+ * SPDX-License-Identifier:    GPL-2.0+
  *
  * Authors: Adrian Hunter
  *          Artem Bityutskiy (Битюцкий Артём)
  * the mutex locked.
  */
 
+#define __UBOOT__
+#ifndef __UBOOT__
+#include <linux/crc32.h>
+#include <linux/slab.h>
+#else
+#include <linux/compat.h>
+#include <linux/err.h>
+#include <linux/stat.h>
+#endif
 #include "ubifs.h"
 
 /*
@@ -176,27 +174,11 @@ static int ins_clr_old_idx_znode(struct ubifs_info *c,
  */
 void destroy_old_idx(struct ubifs_info *c)
 {
-       struct rb_node *this = c->old_idx.rb_node;
-       struct ubifs_old_idx *old_idx;
+       struct ubifs_old_idx *old_idx, *n;
 
-       while (this) {
-               if (this->rb_left) {
-                       this = this->rb_left;
-                       continue;
-               } else if (this->rb_right) {
-                       this = this->rb_right;
-                       continue;
-               }
-               old_idx = rb_entry(this, struct ubifs_old_idx, rb);
-               this = rb_parent(this);
-               if (this) {
-                       if (this->rb_left == &old_idx->rb)
-                               this->rb_left = NULL;
-                       else
-                               this->rb_right = NULL;
-               }
+       rbtree_postorder_for_each_entry_safe(old_idx, n, &c->old_idx, rb)
                kfree(old_idx);
-       }
+
        c->old_idx = RB_ROOT;
 }
 
@@ -221,7 +203,7 @@ static struct ubifs_znode *copy_znode(struct ubifs_info *c,
        __set_bit(DIRTY_ZNODE, &zn->flags);
        __clear_bit(COW_ZNODE, &zn->flags);
 
-       ubifs_assert(!test_bit(OBSOLETE_ZNODE, &znode->flags));
+       ubifs_assert(!ubifs_zn_obsolete(znode));
        __set_bit(OBSOLETE_ZNODE, &znode->flags);
 
        if (znode->level != 0) {
@@ -269,7 +251,7 @@ static struct ubifs_znode *dirty_cow_znode(struct ubifs_info *c,
        struct ubifs_znode *zn;
        int err;
 
-       if (!test_bit(COW_ZNODE, &znode->flags)) {
+       if (!ubifs_zn_cow(znode)) {
                /* znode is not being committed */
                if (!test_and_set_bit(DIRTY_ZNODE, &znode->flags)) {
                        atomic_long_inc(&c->dirty_zn_cnt);
@@ -337,17 +319,16 @@ static int lnc_add(struct ubifs_info *c, struct ubifs_zbranch *zbr,
 
        err = ubifs_validate_entry(c, dent);
        if (err) {
-               dbg_dump_stack();
-               dbg_dump_node(c, dent);
+               dump_stack();
+               ubifs_dump_node(c, dent);
                return err;
        }
 
-       lnc_node = kmalloc(zbr->len, GFP_NOFS);
+       lnc_node = kmemdup(node, zbr->len, GFP_NOFS);
        if (!lnc_node)
                /* We don't have to have the cache, so no error */
                return 0;
 
-       memcpy(lnc_node, node, zbr->len);
        zbr->leaf = lnc_node;
        return 0;
 }
@@ -371,8 +352,8 @@ static int lnc_add_directly(struct ubifs_info *c, struct ubifs_zbranch *zbr,
 
        err = ubifs_validate_entry(c, node);
        if (err) {
-               dbg_dump_stack();
-               dbg_dump_node(c, node);
+               dump_stack();
+               ubifs_dump_node(c, node);
                return err;
        }
 
@@ -445,8 +426,11 @@ static int tnc_read_node_nm(struct ubifs_info *c, struct ubifs_zbranch *zbr,
  *
  * Note, this function does not check CRC of data nodes if @c->no_chk_data_crc
  * is true (it is controlled by corresponding mount option). However, if
- * @c->always_chk_crc is true, @c->no_chk_data_crc is ignored and CRC is always
- * checked.
+ * @c->mounting or @c->remounting_rw is true (we are mounting or re-mounting to
+ * R/W mode), @c->no_chk_data_crc is ignored and CRC is checked. This is
+ * because during mounting or re-mounting from R/O mode to R/W mode we may read
+ * journal nodes (when replying the journal or doing the recovery) and the
+ * journal nodes may potentially be corrupted, so checking is required.
  */
 static int try_read_node(const struct ubifs_info *c, void *buf, int type,
                         int len, int lnum, int offs)
@@ -457,7 +441,7 @@ static int try_read_node(const struct ubifs_info *c, void *buf, int type,
 
        dbg_io("LEB %d:%d, %s, length %d", lnum, offs, dbg_ntype(type), len);
 
-       err = ubi_read(c->ubi, lnum, buf, offs, len);
+       err = ubifs_leb_read(c, lnum, buf, offs, len, 1);
        if (err) {
                ubifs_err("cannot read node type %d from LEB %d:%d, error %d",
                          type, lnum, offs, err);
@@ -474,7 +458,8 @@ static int try_read_node(const struct ubifs_info *c, void *buf, int type,
        if (node_len != len)
                return 0;
 
-       if (type == UBIFS_DATA_NODE && !c->always_chk_crc && c->no_chk_data_crc)
+       if (type == UBIFS_DATA_NODE && c->no_chk_data_crc && !c->mounting &&
+           !c->remounting_rw)
                return 1;
 
        crc = crc32(UBIFS_CRC32_INIT, buf + 8, node_len - 8);
@@ -500,7 +485,7 @@ static int fallible_read_node(struct ubifs_info *c, const union ubifs_key *key,
 {
        int ret;
 
-       dbg_tnc("LEB %d:%d, key %s", zbr->lnum, zbr->offs, DBGKEY(key));
+       dbg_tnck(key, "LEB %d:%d, key ", zbr->lnum, zbr->offs);
 
        ret = try_read_node(c, node, key_type(c, key), zbr->len, zbr->lnum,
                            zbr->offs);
@@ -514,8 +499,8 @@ static int fallible_read_node(struct ubifs_info *c, const union ubifs_key *key,
                        ret = 0;
        }
        if (ret == 0 && c->replaying)
-               dbg_mnt("dangling branch LEB %d:%d len %d, key %s",
-                       zbr->lnum, zbr->offs, zbr->len, DBGKEY(key));
+               dbg_mntk(key, "dangling branch LEB %d:%d len %d, key ",
+                       zbr->lnum, zbr->offs, zbr->len);
        return ret;
 }
 
@@ -990,9 +975,9 @@ static int fallible_resolve_collision(struct ubifs_info *c,
        if (adding || !o_znode)
                return 0;
 
-       dbg_mnt("dangling match LEB %d:%d len %d %s",
+       dbg_mntk(key, "dangling match LEB %d:%d len %d key ",
                o_znode->zbranch[o_n].lnum, o_znode->zbranch[o_n].offs,
-               o_znode->zbranch[o_n].len, DBGKEY(key));
+               o_znode->zbranch[o_n].len);
        *zn = o_znode;
        *n = o_n;
        return 1;
@@ -1158,8 +1143,8 @@ static struct ubifs_znode *dirty_cow_bottom_up(struct ubifs_info *c,
  *   o exact match, i.e. the found zero-level znode contains key @key, then %1
  *     is returned and slot number of the matched branch is stored in @n;
  *   o not exact match, which means that zero-level znode does not contain
- *     @key, then %0 is returned and slot number of the closed branch is stored
- *     in  @n;
+ *     @key, then %0 is returned and slot number of the closest branch is stored
+ *     in @n;
  *   o @key is so small that it is even less than the lowest key of the
  *     leftmost zero-level node, then %0 is returned and %0 is stored in @n.
  *
@@ -1174,7 +1159,8 @@ int ubifs_lookup_level0(struct ubifs_info *c, const union ubifs_key *key,
        struct ubifs_znode *znode;
        unsigned long time = get_seconds();
 
-       dbg_tnc("search key %s", DBGKEY(key));
+       dbg_tnck(key, "search key ");
+       ubifs_assert(key_type(c, key) < UBIFS_INVALID_KEY);
 
        znode = c->zroot.znode;
        if (unlikely(!znode)) {
@@ -1251,7 +1237,7 @@ int ubifs_lookup_level0(struct ubifs_info *c, const union ubifs_key *key,
         * splitting in the middle of the colliding sequence. Also, when
         * removing the leftmost key, we would have to correct the key of the
         * parent node, which would introduce additional complications. Namely,
-        * if we changed the the leftmost key of the parent znode, the garbage
+        * if we changed the leftmost key of the parent znode, the garbage
         * collector would be unable to find it (GC is doing this when GC'ing
         * indexing LEBs). Although we already have an additional RB-tree where
         * we save such changed znodes (see 'ins_clr_old_idx_znode()') until
@@ -1309,7 +1295,7 @@ static int lookup_level0_dirty(struct ubifs_info *c, const union ubifs_key *key,
        struct ubifs_znode *znode;
        unsigned long time = get_seconds();
 
-       dbg_tnc("search and dirty key %s", DBGKEY(key));
+       dbg_tnck(key, "search and dirty key ");
 
        znode = c->zroot.znode;
        if (unlikely(!znode)) {
@@ -1400,9 +1386,31 @@ static int lookup_level0_dirty(struct ubifs_info *c, const union ubifs_key *key,
  */
 static int maybe_leb_gced(struct ubifs_info *c, int lnum, int gc_seq1)
 {
+#ifndef __UBOOT__
+       int gc_seq2, gced_lnum;
+
+       gced_lnum = c->gced_lnum;
+       smp_rmb();
+       gc_seq2 = c->gc_seq;
+       /* Same seq means no GC */
+       if (gc_seq1 == gc_seq2)
+               return 0;
+       /* Different by more than 1 means we don't know */
+       if (gc_seq1 + 1 != gc_seq2)
+               return 1;
        /*
-        * No garbage collection in the read-only U-Boot implementation
+        * We have seen the sequence number has increased by 1. Now we need to
+        * be sure we read the right LEB number, so read it again.
         */
+       smp_rmb();
+       if (gced_lnum != c->gced_lnum)
+               return 1;
+       /* Finally we can check lnum */
+       if (gced_lnum == lnum)
+               return 1;
+#else
+       /* No garbage collection in the read-only U-Boot implementation */
+#endif
        return 0;
 }
 
@@ -1414,7 +1422,7 @@ static int maybe_leb_gced(struct ubifs_info *c, int lnum, int gc_seq1)
  * @lnum: LEB number is returned here
  * @offs: offset is returned here
  *
- * This function look up and reads node with key @key. The caller has to make
+ * This function looks up and reads node with key @key. The caller has to make
  * sure the @node buffer is large enough to fit the node. Returns zero in case
  * of success, %-ENOENT if the node was not found, and a negative error code in
  * case of failure. The node location can be returned in @lnum and @offs.
@@ -1458,6 +1466,12 @@ again:
        gc_seq1 = c->gc_seq;
        mutex_unlock(&c->tnc_mutex);
 
+       if (ubifs_get_wbuf(c, zbr.lnum)) {
+               /* We do not GC journal heads */
+               err = ubifs_tnc_read_node(c, &zbr, node);
+               return err;
+       }
+
        err = fallible_read_node(c, key, &zbr, node);
        if (err <= 0 || maybe_leb_gced(c, zbr.lnum, gc_seq1)) {
                /*
@@ -1609,6 +1623,51 @@ out:
        return 0;
 }
 
+/**
+ * read_wbuf - bulk-read from a LEB with a wbuf.
+ * @wbuf: wbuf that may overlap the read
+ * @buf: buffer into which to read
+ * @len: read length
+ * @lnum: LEB number from which to read
+ * @offs: offset from which to read
+ *
+ * This functions returns %0 on success or a negative error code on failure.
+ */
+static int read_wbuf(struct ubifs_wbuf *wbuf, void *buf, int len, int lnum,
+                    int offs)
+{
+       const struct ubifs_info *c = wbuf->c;
+       int rlen, overlap;
+
+       dbg_io("LEB %d:%d, length %d", lnum, offs, len);
+       ubifs_assert(wbuf && lnum >= 0 && lnum < c->leb_cnt && offs >= 0);
+       ubifs_assert(!(offs & 7) && offs < c->leb_size);
+       ubifs_assert(offs + len <= c->leb_size);
+
+       spin_lock(&wbuf->lock);
+       overlap = (lnum == wbuf->lnum && offs + len > wbuf->offs);
+       if (!overlap) {
+               /* We may safely unlock the write-buffer and read the data */
+               spin_unlock(&wbuf->lock);
+               return ubifs_leb_read(c, lnum, buf, offs, len, 0);
+       }
+
+       /* Don't read under wbuf */
+       rlen = wbuf->offs - offs;
+       if (rlen < 0)
+               rlen = 0;
+
+       /* Copy the rest from the write-buffer */
+       memcpy(buf + rlen, wbuf->buf + offs + rlen - wbuf->offs, len - rlen);
+       spin_unlock(&wbuf->lock);
+
+       if (rlen > 0)
+               /* Read everything that goes before write-buffer */
+               return ubifs_leb_read(c, lnum, buf, offs, rlen, 0);
+
+       return 0;
+}
+
 /**
  * validate_data_node - validate data nodes for bulk-read.
  * @c: UBIFS file-system description object
@@ -1647,8 +1706,8 @@ static int validate_data_node(struct ubifs_info *c, void *buf,
        if (!keys_eq(c, &zbr->key, &key1)) {
                ubifs_err("bad key in node at LEB %d:%d",
                          zbr->lnum, zbr->offs);
-               dbg_tnc("looked for key %s found node's key %s",
-                       DBGKEY(&zbr->key), DBGKEY1(&key1));
+               dbg_tnck(&zbr->key, "looked for key ");
+               dbg_tnck(&key1, "found node's key ");
                goto out_err;
        }
 
@@ -1658,8 +1717,8 @@ out_err:
        err = -EINVAL;
 out:
        ubifs_err("bad node at LEB %d:%d", zbr->lnum, zbr->offs);
-       dbg_dump_node(c, buf);
-       dbg_dump_stack();
+       ubifs_dump_node(c, buf);
+       dump_stack();
        return err;
 }
 
@@ -1676,6 +1735,7 @@ out:
 int ubifs_tnc_bulk_read(struct ubifs_info *c, struct bu_info *bu)
 {
        int lnum = bu->zbranch[0].lnum, offs = bu->zbranch[0].offs, len, err, i;
+       struct ubifs_wbuf *wbuf;
        void *buf;
 
        len = bu->zbranch[bu->cnt - 1].offs;
@@ -1686,7 +1746,11 @@ int ubifs_tnc_bulk_read(struct ubifs_info *c, struct bu_info *bu)
        }
 
        /* Do the read */
-       err = ubi_read(c->ubi, lnum, bu->buf, offs, len);
+       wbuf = ubifs_get_wbuf(c, lnum);
+       if (wbuf)
+               err = read_wbuf(wbuf, bu->buf, len, lnum, offs);
+       else
+               err = ubifs_leb_read(c, lnum, bu->buf, offs, len, 0);
 
        /* Check for a race with GC */
        if (maybe_leb_gced(c, lnum, bu->gc_seq))
@@ -1695,8 +1759,8 @@ int ubifs_tnc_bulk_read(struct ubifs_info *c, struct bu_info *bu)
        if (err && err != -EBADMSG) {
                ubifs_err("failed to read from LEB %d:%d, error %d",
                          lnum, offs, err);
-               dbg_dump_stack();
-               dbg_tnc("key %s", DBGKEY(&bu->key));
+               dump_stack();
+               dbg_tnck(&bu->key, "key ");
                return err;
        }
 
@@ -1731,7 +1795,7 @@ static int do_lookup_nm(struct ubifs_info *c, const union ubifs_key *key,
        int found, n, err;
        struct ubifs_znode *znode;
 
-       dbg_tnc("name '%.*s' key %s", nm->len, nm->name, DBGKEY(key));
+       dbg_tnck(key, "name '%.*s' key ", nm->len, nm->name);
        mutex_lock(&c->tnc_mutex);
        found = ubifs_lookup_level0(c, key, &znode, &n);
        if (!found) {
@@ -1905,8 +1969,7 @@ again:
        zp = znode->parent;
        if (znode->child_cnt < c->fanout) {
                ubifs_assert(n != c->fanout);
-               dbg_tnc("inserted at %d level %d, key %s", n, znode->level,
-                       DBGKEY(key));
+               dbg_tnck(key, "inserted at %d level %d, key ", n, znode->level);
 
                insert_zbranch(znode, zbr, n);
 
@@ -1921,7 +1984,7 @@ again:
         * Unfortunately, @znode does not have more empty slots and we have to
         * split it.
         */
-       dbg_tnc("splitting level %d, key %s", znode->level, DBGKEY(key));
+       dbg_tnck(key, "splitting level %d, key ", znode->level);
 
        if (znode->alt)
                /*
@@ -2015,7 +2078,7 @@ do_split:
        }
 
        /* Insert new key and branch */
-       dbg_tnc("inserting at %d level %d, key %s", n, zn->level, DBGKEY(key));
+       dbg_tnck(key, "inserting at %d level %d, key ", n, zn->level);
 
        insert_zbranch(zi, zbr, n);
 
@@ -2091,7 +2154,7 @@ int ubifs_tnc_add(struct ubifs_info *c, const union ubifs_key *key, int lnum,
        struct ubifs_znode *znode;
 
        mutex_lock(&c->tnc_mutex);
-       dbg_tnc("%d:%d, len %d, key %s", lnum, offs, len, DBGKEY(key));
+       dbg_tnck(key, "%d:%d, len %d, key ", lnum, offs, len);
        found = lookup_level0_dirty(c, key, &znode, &n);
        if (!found) {
                struct ubifs_zbranch zbr;
@@ -2140,8 +2203,8 @@ int ubifs_tnc_replace(struct ubifs_info *c, const union ubifs_key *key,
        struct ubifs_znode *znode;
 
        mutex_lock(&c->tnc_mutex);
-       dbg_tnc("old LEB %d:%d, new LEB %d:%d, len %d, key %s", old_lnum,
-               old_offs, lnum, offs, len, DBGKEY(key));
+       dbg_tnck(key, "old LEB %d:%d, new LEB %d:%d, len %d, key ", old_lnum,
+                old_offs, lnum, offs, len);
        found = lookup_level0_dirty(c, key, &znode, &n);
        if (found < 0) {
                err = found;
@@ -2223,8 +2286,8 @@ int ubifs_tnc_add_nm(struct ubifs_info *c, const union ubifs_key *key,
        struct ubifs_znode *znode;
 
        mutex_lock(&c->tnc_mutex);
-       dbg_tnc("LEB %d:%d, name '%.*s', key %s", lnum, offs, nm->len, nm->name,
-               DBGKEY(key));
+       dbg_tnck(key, "LEB %d:%d, name '%.*s', key ",
+                lnum, offs, nm->len, nm->name);
        found = lookup_level0_dirty(c, key, &znode, &n);
        if (found < 0) {
                err = found;
@@ -2282,7 +2345,7 @@ int ubifs_tnc_add_nm(struct ubifs_info *c, const union ubifs_key *key,
                         * by passing 'ubifs_tnc_remove_nm()' the same key but
                         * an unmatchable name.
                         */
-                       struct qstr noname = { .len = 0, .name = "" };
+                       struct qstr noname = { .name = "" };
 
                        err = dbg_check_tnc(c, 0);
                        mutex_unlock(&c->tnc_mutex);
@@ -2317,14 +2380,14 @@ static int tnc_delete(struct ubifs_info *c, struct ubifs_znode *znode, int n)
        /* Delete without merge for now */
        ubifs_assert(znode->level == 0);
        ubifs_assert(n >= 0 && n < c->fanout);
-       dbg_tnc("deleting %s", DBGKEY(&znode->zbranch[n].key));
+       dbg_tnck(&znode->zbranch[n].key, "deleting key ");
 
        zbr = &znode->zbranch[n];
        lnc_free(zbr);
 
        err = ubifs_add_dirt(c, zbr->lnum, zbr->len);
        if (err) {
-               dbg_dump_znode(c, znode);
+               ubifs_dump_znode(c, znode);
                return err;
        }
 
@@ -2342,7 +2405,7 @@ static int tnc_delete(struct ubifs_info *c, struct ubifs_znode *znode, int n)
         */
 
        do {
-               ubifs_assert(!test_bit(OBSOLETE_ZNODE, &znode->flags));
+               ubifs_assert(!ubifs_zn_obsolete(znode));
                ubifs_assert(ubifs_zn_dirty(znode));
 
                zp = znode->parent;
@@ -2398,9 +2461,8 @@ static int tnc_delete(struct ubifs_info *c, struct ubifs_znode *znode, int n)
                        c->zroot.offs = zbr->offs;
                        c->zroot.len = zbr->len;
                        c->zroot.znode = znode;
-                       ubifs_assert(!test_bit(OBSOLETE_ZNODE,
-                                    &zp->flags));
-                       ubifs_assert(test_bit(DIRTY_ZNODE, &zp->flags));
+                       ubifs_assert(!ubifs_zn_obsolete(zp));
+                       ubifs_assert(ubifs_zn_dirty(zp));
                        atomic_long_dec(&c->dirty_zn_cnt);
 
                        if (zp->cnext) {
@@ -2428,7 +2490,7 @@ int ubifs_tnc_remove(struct ubifs_info *c, const union ubifs_key *key)
        struct ubifs_znode *znode;
 
        mutex_lock(&c->tnc_mutex);
-       dbg_tnc("key %s", DBGKEY(key));
+       dbg_tnck(key, "key ");
        found = lookup_level0_dirty(c, key, &znode, &n);
        if (found < 0) {
                err = found;
@@ -2459,7 +2521,7 @@ int ubifs_tnc_remove_nm(struct ubifs_info *c, const union ubifs_key *key,
        struct ubifs_znode *znode;
 
        mutex_lock(&c->tnc_mutex);
-       dbg_tnc("%.*s, key %s", nm->len, nm->name, DBGKEY(key));
+       dbg_tnck(key, "%.*s, key ", nm->len, nm->name);
        err = lookup_level0_dirty(c, key, &znode, &n);
        if (err < 0)
                goto out_unlock;
@@ -2476,11 +2538,11 @@ int ubifs_tnc_remove_nm(struct ubifs_info *c, const union ubifs_key *key,
                if (err) {
                        /* Ensure the znode is dirtied */
                        if (znode->cnext || !ubifs_zn_dirty(znode)) {
-                                   znode = dirty_cow_bottom_up(c, znode);
-                                   if (IS_ERR(znode)) {
-                                           err = PTR_ERR(znode);
-                                           goto out_unlock;
-                                   }
+                               znode = dirty_cow_bottom_up(c, znode);
+                               if (IS_ERR(znode)) {
+                                       err = PTR_ERR(znode);
+                                       goto out_unlock;
+                               }
                        }
                        err = tnc_delete(c, znode, n);
                }
@@ -2571,10 +2633,10 @@ int ubifs_tnc_remove_range(struct ubifs_info *c, union ubifs_key *from_key,
                        err = ubifs_add_dirt(c, znode->zbranch[i].lnum,
                                             znode->zbranch[i].len);
                        if (err) {
-                               dbg_dump_znode(c, znode);
+                               ubifs_dump_znode(c, znode);
                                goto out_unlock;
                        }
-                       dbg_tnc("removing %s", DBGKEY(key));
+                       dbg_tnck(key, "removing key ");
                }
                if (k) {
                        for (i = n + 1 + k; i < znode->child_cnt; i++)
@@ -2633,7 +2695,7 @@ int ubifs_tnc_remove_ino(struct ubifs_info *c, ino_t inum)
                dbg_tnc("xent '%s', ino %lu", xent->name,
                        (unsigned long)xattr_inum);
 
-               nm.name = (char *)xent->name;
+               nm.name = xent->name;
                nm.len = le16_to_cpu(xent->nlen);
                err = ubifs_tnc_remove_nm(c, &key1, &nm);
                if (err) {
@@ -2694,7 +2756,7 @@ struct ubifs_dent_node *ubifs_tnc_next_ent(struct ubifs_info *c,
        struct ubifs_zbranch *zbr;
        union ubifs_key *dkey;
 
-       dbg_tnc("%s %s", nm->name ? (char *)nm->name : "(lowest)", DBGKEY(key));
+       dbg_tnck(key, "%s ", nm->name ? (char *)nm->name : "(lowest)");
        ubifs_assert(is_hash_key(c, key));
 
        mutex_lock(&c->tnc_mutex);
@@ -2765,3 +2827,503 @@ out_unlock:
        mutex_unlock(&c->tnc_mutex);
        return ERR_PTR(err);
 }
+
+#ifndef __UBOOT__
+/**
+ * tnc_destroy_cnext - destroy left-over obsolete znodes from a failed commit.
+ * @c: UBIFS file-system description object
+ *
+ * Destroy left-over obsolete znodes from a failed commit.
+ */
+static void tnc_destroy_cnext(struct ubifs_info *c)
+{
+       struct ubifs_znode *cnext;
+
+       if (!c->cnext)
+               return;
+       ubifs_assert(c->cmt_state == COMMIT_BROKEN);
+       cnext = c->cnext;
+       do {
+               struct ubifs_znode *znode = cnext;
+
+               cnext = cnext->cnext;
+               if (ubifs_zn_obsolete(znode))
+                       kfree(znode);
+       } while (cnext && cnext != c->cnext);
+}
+
+/**
+ * ubifs_tnc_close - close TNC subsystem and free all related resources.
+ * @c: UBIFS file-system description object
+ */
+void ubifs_tnc_close(struct ubifs_info *c)
+{
+       tnc_destroy_cnext(c);
+       if (c->zroot.znode) {
+               long n;
+
+               ubifs_destroy_tnc_subtree(c->zroot.znode);
+               n = atomic_long_read(&c->clean_zn_cnt);
+               atomic_long_sub(n, &ubifs_clean_zn_cnt);
+       }
+       kfree(c->gap_lebs);
+       kfree(c->ilebs);
+       destroy_old_idx(c);
+}
+#endif
+
+/**
+ * left_znode - get the znode to the left.
+ * @c: UBIFS file-system description object
+ * @znode: znode
+ *
+ * This function returns a pointer to the znode to the left of @znode or NULL if
+ * there is not one. A negative error code is returned on failure.
+ */
+static struct ubifs_znode *left_znode(struct ubifs_info *c,
+                                     struct ubifs_znode *znode)
+{
+       int level = znode->level;
+
+       while (1) {
+               int n = znode->iip - 1;
+
+               /* Go up until we can go left */
+               znode = znode->parent;
+               if (!znode)
+                       return NULL;
+               if (n >= 0) {
+                       /* Now go down the rightmost branch to 'level' */
+                       znode = get_znode(c, znode, n);
+                       if (IS_ERR(znode))
+                               return znode;
+                       while (znode->level != level) {
+                               n = znode->child_cnt - 1;
+                               znode = get_znode(c, znode, n);
+                               if (IS_ERR(znode))
+                                       return znode;
+                       }
+                       break;
+               }
+       }
+       return znode;
+}
+
+/**
+ * right_znode - get the znode to the right.
+ * @c: UBIFS file-system description object
+ * @znode: znode
+ *
+ * This function returns a pointer to the znode to the right of @znode or NULL
+ * if there is not one. A negative error code is returned on failure.
+ */
+static struct ubifs_znode *right_znode(struct ubifs_info *c,
+                                      struct ubifs_znode *znode)
+{
+       int level = znode->level;
+
+       while (1) {
+               int n = znode->iip + 1;
+
+               /* Go up until we can go right */
+               znode = znode->parent;
+               if (!znode)
+                       return NULL;
+               if (n < znode->child_cnt) {
+                       /* Now go down the leftmost branch to 'level' */
+                       znode = get_znode(c, znode, n);
+                       if (IS_ERR(znode))
+                               return znode;
+                       while (znode->level != level) {
+                               znode = get_znode(c, znode, 0);
+                               if (IS_ERR(znode))
+                                       return znode;
+                       }
+                       break;
+               }
+       }
+       return znode;
+}
+
+/**
+ * lookup_znode - find a particular indexing node from TNC.
+ * @c: UBIFS file-system description object
+ * @key: index node key to lookup
+ * @level: index node level
+ * @lnum: index node LEB number
+ * @offs: index node offset
+ *
+ * This function searches an indexing node by its first key @key and its
+ * address @lnum:@offs. It looks up the indexing tree by pulling all indexing
+ * nodes it traverses to TNC. This function is called for indexing nodes which
+ * were found on the media by scanning, for example when garbage-collecting or
+ * when doing in-the-gaps commit. This means that the indexing node which is
+ * looked for does not have to have exactly the same leftmost key @key, because
+ * the leftmost key may have been changed, in which case TNC will contain a
+ * dirty znode which still refers the same @lnum:@offs. This function is clever
+ * enough to recognize such indexing nodes.
+ *
+ * Note, if a znode was deleted or changed too much, then this function will
+ * not find it. For situations like this UBIFS has the old index RB-tree
+ * (indexed by @lnum:@offs).
+ *
+ * This function returns a pointer to the znode found or %NULL if it is not
+ * found. A negative error code is returned on failure.
+ */
+static struct ubifs_znode *lookup_znode(struct ubifs_info *c,
+                                       union ubifs_key *key, int level,
+                                       int lnum, int offs)
+{
+       struct ubifs_znode *znode, *zn;
+       int n, nn;
+
+       ubifs_assert(key_type(c, key) < UBIFS_INVALID_KEY);
+
+       /*
+        * The arguments have probably been read off flash, so don't assume
+        * they are valid.
+        */
+       if (level < 0)
+               return ERR_PTR(-EINVAL);
+
+       /* Get the root znode */
+       znode = c->zroot.znode;
+       if (!znode) {
+               znode = ubifs_load_znode(c, &c->zroot, NULL, 0);
+               if (IS_ERR(znode))
+                       return znode;
+       }
+       /* Check if it is the one we are looking for */
+       if (c->zroot.lnum == lnum && c->zroot.offs == offs)
+               return znode;
+       /* Descend to the parent level i.e. (level + 1) */
+       if (level >= znode->level)
+               return NULL;
+       while (1) {
+               ubifs_search_zbranch(c, znode, key, &n);
+               if (n < 0) {
+                       /*
+                        * We reached a znode where the leftmost key is greater
+                        * than the key we are searching for. This is the same
+                        * situation as the one described in a huge comment at
+                        * the end of the 'ubifs_lookup_level0()' function. And
+                        * for exactly the same reasons we have to try to look
+                        * left before giving up.
+                        */
+                       znode = left_znode(c, znode);
+                       if (!znode)
+                               return NULL;
+                       if (IS_ERR(znode))
+                               return znode;
+                       ubifs_search_zbranch(c, znode, key, &n);
+                       ubifs_assert(n >= 0);
+               }
+               if (znode->level == level + 1)
+                       break;
+               znode = get_znode(c, znode, n);
+               if (IS_ERR(znode))
+                       return znode;
+       }
+       /* Check if the child is the one we are looking for */
+       if (znode->zbranch[n].lnum == lnum && znode->zbranch[n].offs == offs)
+               return get_znode(c, znode, n);
+       /* If the key is unique, there is nowhere else to look */
+       if (!is_hash_key(c, key))
+               return NULL;
+       /*
+        * The key is not unique and so may be also in the znodes to either
+        * side.
+        */
+       zn = znode;
+       nn = n;
+       /* Look left */
+       while (1) {
+               /* Move one branch to the left */
+               if (n)
+                       n -= 1;
+               else {
+                       znode = left_znode(c, znode);
+                       if (!znode)
+                               break;
+                       if (IS_ERR(znode))
+                               return znode;
+                       n = znode->child_cnt - 1;
+               }
+               /* Check it */
+               if (znode->zbranch[n].lnum == lnum &&
+                   znode->zbranch[n].offs == offs)
+                       return get_znode(c, znode, n);
+               /* Stop if the key is less than the one we are looking for */
+               if (keys_cmp(c, &znode->zbranch[n].key, key) < 0)
+                       break;
+       }
+       /* Back to the middle */
+       znode = zn;
+       n = nn;
+       /* Look right */
+       while (1) {
+               /* Move one branch to the right */
+               if (++n >= znode->child_cnt) {
+                       znode = right_znode(c, znode);
+                       if (!znode)
+                               break;
+                       if (IS_ERR(znode))
+                               return znode;
+                       n = 0;
+               }
+               /* Check it */
+               if (znode->zbranch[n].lnum == lnum &&
+                   znode->zbranch[n].offs == offs)
+                       return get_znode(c, znode, n);
+               /* Stop if the key is greater than the one we are looking for */
+               if (keys_cmp(c, &znode->zbranch[n].key, key) > 0)
+                       break;
+       }
+       return NULL;
+}
+
+/**
+ * is_idx_node_in_tnc - determine if an index node is in the TNC.
+ * @c: UBIFS file-system description object
+ * @key: key of index node
+ * @level: index node level
+ * @lnum: LEB number of index node
+ * @offs: offset of index node
+ *
+ * This function returns %0 if the index node is not referred to in the TNC, %1
+ * if the index node is referred to in the TNC and the corresponding znode is
+ * dirty, %2 if an index node is referred to in the TNC and the corresponding
+ * znode is clean, and a negative error code in case of failure.
+ *
+ * Note, the @key argument has to be the key of the first child. Also note,
+ * this function relies on the fact that 0:0 is never a valid LEB number and
+ * offset for a main-area node.
+ */
+int is_idx_node_in_tnc(struct ubifs_info *c, union ubifs_key *key, int level,
+                      int lnum, int offs)
+{
+       struct ubifs_znode *znode;
+
+       znode = lookup_znode(c, key, level, lnum, offs);
+       if (!znode)
+               return 0;
+       if (IS_ERR(znode))
+               return PTR_ERR(znode);
+
+       return ubifs_zn_dirty(znode) ? 1 : 2;
+}
+
+/**
+ * is_leaf_node_in_tnc - determine if a non-indexing not is in the TNC.
+ * @c: UBIFS file-system description object
+ * @key: node key
+ * @lnum: node LEB number
+ * @offs: node offset
+ *
+ * This function returns %1 if the node is referred to in the TNC, %0 if it is
+ * not, and a negative error code in case of failure.
+ *
+ * Note, this function relies on the fact that 0:0 is never a valid LEB number
+ * and offset for a main-area node.
+ */
+static int is_leaf_node_in_tnc(struct ubifs_info *c, union ubifs_key *key,
+                              int lnum, int offs)
+{
+       struct ubifs_zbranch *zbr;
+       struct ubifs_znode *znode, *zn;
+       int n, found, err, nn;
+       const int unique = !is_hash_key(c, key);
+
+       found = ubifs_lookup_level0(c, key, &znode, &n);
+       if (found < 0)
+               return found; /* Error code */
+       if (!found)
+               return 0;
+       zbr = &znode->zbranch[n];
+       if (lnum == zbr->lnum && offs == zbr->offs)
+               return 1; /* Found it */
+       if (unique)
+               return 0;
+       /*
+        * Because the key is not unique, we have to look left
+        * and right as well
+        */
+       zn = znode;
+       nn = n;
+       /* Look left */
+       while (1) {
+               err = tnc_prev(c, &znode, &n);
+               if (err == -ENOENT)
+                       break;
+               if (err)
+                       return err;
+               if (keys_cmp(c, key, &znode->zbranch[n].key))
+                       break;
+               zbr = &znode->zbranch[n];
+               if (lnum == zbr->lnum && offs == zbr->offs)
+                       return 1; /* Found it */
+       }
+       /* Look right */
+       znode = zn;
+       n = nn;
+       while (1) {
+               err = tnc_next(c, &znode, &n);
+               if (err) {
+                       if (err == -ENOENT)
+                               return 0;
+                       return err;
+               }
+               if (keys_cmp(c, key, &znode->zbranch[n].key))
+                       break;
+               zbr = &znode->zbranch[n];
+               if (lnum == zbr->lnum && offs == zbr->offs)
+                       return 1; /* Found it */
+       }
+       return 0;
+}
+
+/**
+ * ubifs_tnc_has_node - determine whether a node is in the TNC.
+ * @c: UBIFS file-system description object
+ * @key: node key
+ * @level: index node level (if it is an index node)
+ * @lnum: node LEB number
+ * @offs: node offset
+ * @is_idx: non-zero if the node is an index node
+ *
+ * This function returns %1 if the node is in the TNC, %0 if it is not, and a
+ * negative error code in case of failure. For index nodes, @key has to be the
+ * key of the first child. An index node is considered to be in the TNC only if
+ * the corresponding znode is clean or has not been loaded.
+ */
+int ubifs_tnc_has_node(struct ubifs_info *c, union ubifs_key *key, int level,
+                      int lnum, int offs, int is_idx)
+{
+       int err;
+
+       mutex_lock(&c->tnc_mutex);
+       if (is_idx) {
+               err = is_idx_node_in_tnc(c, key, level, lnum, offs);
+               if (err < 0)
+                       goto out_unlock;
+               if (err == 1)
+                       /* The index node was found but it was dirty */
+                       err = 0;
+               else if (err == 2)
+                       /* The index node was found and it was clean */
+                       err = 1;
+               else
+                       BUG_ON(err != 0);
+       } else
+               err = is_leaf_node_in_tnc(c, key, lnum, offs);
+
+out_unlock:
+       mutex_unlock(&c->tnc_mutex);
+       return err;
+}
+
+/**
+ * ubifs_dirty_idx_node - dirty an index node.
+ * @c: UBIFS file-system description object
+ * @key: index node key
+ * @level: index node level
+ * @lnum: index node LEB number
+ * @offs: index node offset
+ *
+ * This function loads and dirties an index node so that it can be garbage
+ * collected. The @key argument has to be the key of the first child. This
+ * function relies on the fact that 0:0 is never a valid LEB number and offset
+ * for a main-area node. Returns %0 on success and a negative error code on
+ * failure.
+ */
+int ubifs_dirty_idx_node(struct ubifs_info *c, union ubifs_key *key, int level,
+                        int lnum, int offs)
+{
+       struct ubifs_znode *znode;
+       int err = 0;
+
+       mutex_lock(&c->tnc_mutex);
+       znode = lookup_znode(c, key, level, lnum, offs);
+       if (!znode)
+               goto out_unlock;
+       if (IS_ERR(znode)) {
+               err = PTR_ERR(znode);
+               goto out_unlock;
+       }
+       znode = dirty_cow_bottom_up(c, znode);
+       if (IS_ERR(znode)) {
+               err = PTR_ERR(znode);
+               goto out_unlock;
+       }
+
+out_unlock:
+       mutex_unlock(&c->tnc_mutex);
+       return err;
+}
+
+/**
+ * dbg_check_inode_size - check if inode size is correct.
+ * @c: UBIFS file-system description object
+ * @inum: inode number
+ * @size: inode size
+ *
+ * This function makes sure that the inode size (@size) is correct and it does
+ * not have any pages beyond @size. Returns zero if the inode is OK, %-EINVAL
+ * if it has a data page beyond @size, and other negative error code in case of
+ * other errors.
+ */
+int dbg_check_inode_size(struct ubifs_info *c, const struct inode *inode,
+                        loff_t size)
+{
+       int err, n;
+       union ubifs_key from_key, to_key, *key;
+       struct ubifs_znode *znode;
+       unsigned int block;
+
+       if (!S_ISREG(inode->i_mode))
+               return 0;
+       if (!dbg_is_chk_gen(c))
+               return 0;
+
+       block = (size + UBIFS_BLOCK_SIZE - 1) >> UBIFS_BLOCK_SHIFT;
+       data_key_init(c, &from_key, inode->i_ino, block);
+       highest_data_key(c, &to_key, inode->i_ino);
+
+       mutex_lock(&c->tnc_mutex);
+       err = ubifs_lookup_level0(c, &from_key, &znode, &n);
+       if (err < 0)
+               goto out_unlock;
+
+       if (err) {
+               err = -EINVAL;
+               key = &from_key;
+               goto out_dump;
+       }
+
+       err = tnc_next(c, &znode, &n);
+       if (err == -ENOENT) {
+               err = 0;
+               goto out_unlock;
+       }
+       if (err < 0)
+               goto out_unlock;
+
+       ubifs_assert(err == 0);
+       key = &znode->zbranch[n].key;
+       if (!key_in_range(c, key, &from_key, &to_key))
+               goto out_unlock;
+
+out_dump:
+       block = key_block(c, key);
+       ubifs_err("inode %lu has size %lld, but there are data at offset %lld",
+                 (unsigned long)inode->i_ino, size,
+                 ((loff_t)block) << UBIFS_BLOCK_SHIFT);
+       mutex_unlock(&c->tnc_mutex);
+       ubifs_dump_inode(c, inode);
+       dump_stack();
+       return -EINVAL;
+
+out_unlock:
+       mutex_unlock(&c->tnc_mutex);
+       return err;
+}
index 955219fa019487910d7fdfb7ca23051a9eab9ee6..81bdad920344b3c23563ef2814bf187fbfab9971 100644 (file)
@@ -3,18 +3,7 @@
  *
  * Copyright (C) 2006-2008 Nokia Corporation.
  *
- * This program is free software; you can redistribute it and/or modify it
- * under the terms of the GNU General Public License version 2 as published by
- * the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
- * more details.
- *
- * You should have received a copy of the GNU General Public License along with
- * this program; if not, write to the Free Software Foundation, Inc., 51
- * Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
+ * SPDX-License-Identifier:    GPL-2.0+
  *
  * Authors: Adrian Hunter
  *          Artem Bityutskiy (Битюцкий Артём)
  * putting it all in one file would make that file too big and unreadable.
  */
 
+#define __UBOOT__
+#ifdef __UBOOT__
+#include <linux/err.h>
+#endif
 #include "ubifs.h"
 
 /**
@@ -218,6 +211,44 @@ struct ubifs_znode *ubifs_tnc_postorder_next(struct ubifs_znode *znode)
        return ubifs_tnc_postorder_first(zn);
 }
 
+/**
+ * ubifs_destroy_tnc_subtree - destroy all znodes connected to a subtree.
+ * @znode: znode defining subtree to destroy
+ *
+ * This function destroys subtree of the TNC tree. Returns number of clean
+ * znodes in the subtree.
+ */
+long ubifs_destroy_tnc_subtree(struct ubifs_znode *znode)
+{
+       struct ubifs_znode *zn = ubifs_tnc_postorder_first(znode);
+       long clean_freed = 0;
+       int n;
+
+       ubifs_assert(zn);
+       while (1) {
+               for (n = 0; n < zn->child_cnt; n++) {
+                       if (!zn->zbranch[n].znode)
+                               continue;
+
+                       if (zn->level > 0 &&
+                           !ubifs_zn_dirty(zn->zbranch[n].znode))
+                               clean_freed += 1;
+
+                       cond_resched();
+                       kfree(zn->zbranch[n].znode);
+               }
+
+               if (zn == znode) {
+                       if (!ubifs_zn_dirty(zn))
+                               clean_freed += 1;
+                       kfree(zn);
+                       return clean_freed;
+               }
+
+               zn = ubifs_tnc_postorder_next(zn);
+       }
+}
+
 /**
  * read_znode - read an indexing node from flash and fill znode.
  * @c: UBIFS file-system description object
@@ -255,10 +286,10 @@ static int read_znode(struct ubifs_info *c, int lnum, int offs, int len,
                lnum, offs, znode->level, znode->child_cnt);
 
        if (znode->child_cnt > c->fanout || znode->level > UBIFS_MAX_LEVELS) {
-               dbg_err("current fanout %d, branch count %d",
-                       c->fanout, znode->child_cnt);
-               dbg_err("max levels %d, znode level %d",
-                       UBIFS_MAX_LEVELS, znode->level);
+               ubifs_err("current fanout %d, branch count %d",
+                         c->fanout, znode->child_cnt);
+               ubifs_err("max levels %d, znode level %d",
+                         UBIFS_MAX_LEVELS, znode->level);
                err = 1;
                goto out_dump;
        }
@@ -278,7 +309,7 @@ static int read_znode(struct ubifs_info *c, int lnum, int offs, int len,
                if (zbr->lnum < c->main_first ||
                    zbr->lnum >= c->leb_cnt || zbr->offs < 0 ||
                    zbr->offs + zbr->len > c->leb_size || zbr->offs & 7) {
-                       dbg_err("bad branch %d", i);
+                       ubifs_err("bad branch %d", i);
                        err = 2;
                        goto out_dump;
                }
@@ -290,8 +321,8 @@ static int read_znode(struct ubifs_info *c, int lnum, int offs, int len,
                case UBIFS_XENT_KEY:
                        break;
                default:
-                       dbg_msg("bad key type at slot %d: %s", i,
-                               DBGKEY(&zbr->key));
+                       ubifs_err("bad key type at slot %d: %d",
+                                 i, key_type(c, &zbr->key));
                        err = 3;
                        goto out_dump;
                }
@@ -302,19 +333,19 @@ static int read_znode(struct ubifs_info *c, int lnum, int offs, int len,
                type = key_type(c, &zbr->key);
                if (c->ranges[type].max_len == 0) {
                        if (zbr->len != c->ranges[type].len) {
-                               dbg_err("bad target node (type %d) length (%d)",
-                                       type, zbr->len);
-                               dbg_err("have to be %d", c->ranges[type].len);
+                               ubifs_err("bad target node (type %d) length (%d)",
+                                         type, zbr->len);
+                               ubifs_err("have to be %d", c->ranges[type].len);
                                err = 4;
                                goto out_dump;
                        }
                } else if (zbr->len < c->ranges[type].min_len ||
                           zbr->len > c->ranges[type].max_len) {
-                       dbg_err("bad target node (type %d) length (%d)",
-                               type, zbr->len);
-                       dbg_err("have to be in range of %d-%d",
-                               c->ranges[type].min_len,
-                               c->ranges[type].max_len);
+                       ubifs_err("bad target node (type %d) length (%d)",
+                                 type, zbr->len);
+                       ubifs_err("have to be in range of %d-%d",
+                                 c->ranges[type].min_len,
+                                 c->ranges[type].max_len);
                        err = 5;
                        goto out_dump;
                }
@@ -332,13 +363,13 @@ static int read_znode(struct ubifs_info *c, int lnum, int offs, int len,
 
                cmp = keys_cmp(c, key1, key2);
                if (cmp > 0) {
-                       dbg_err("bad key order (keys %d and %d)", i, i + 1);
+                       ubifs_err("bad key order (keys %d and %d)", i, i + 1);
                        err = 6;
                        goto out_dump;
                } else if (cmp == 0 && !is_hash_key(c, key1)) {
                        /* These can only be keys with colliding hash */
-                       dbg_err("keys %d and %d are not hashed but equivalent",
-                               i, i + 1);
+                       ubifs_err("keys %d and %d are not hashed but equivalent",
+                                 i, i + 1);
                        err = 7;
                        goto out_dump;
                }
@@ -349,7 +380,7 @@ static int read_znode(struct ubifs_info *c, int lnum, int offs, int len,
 
 out_dump:
        ubifs_err("bad indexing node at LEB %d:%d, error %d", lnum, offs, err);
-       dbg_dump_node(c, idx);
+       ubifs_dump_node(c, idx);
        kfree(idx);
        return -EINVAL;
 }
@@ -385,6 +416,16 @@ struct ubifs_znode *ubifs_load_znode(struct ubifs_info *c,
        if (err)
                goto out;
 
+       atomic_long_inc(&c->clean_zn_cnt);
+
+       /*
+        * Increment the global clean znode counter as well. It is OK that
+        * global and per-FS clean znode counters may be inconsistent for some
+        * short time (because we might be preempted at this point), the global
+        * one is only used in shrinker.
+        */
+       atomic_long_inc(&ubifs_clean_zn_cnt);
+
        zbr->znode = znode;
        znode->parent = parent;
        znode->time = get_seconds();
@@ -412,11 +453,22 @@ int ubifs_tnc_read_node(struct ubifs_info *c, struct ubifs_zbranch *zbr,
 {
        union ubifs_key key1, *key = &zbr->key;
        int err, type = key_type(c, key);
+       struct ubifs_wbuf *wbuf;
 
-       err = ubifs_read_node(c, node, type, zbr->len, zbr->lnum, zbr->offs);
+       /*
+        * 'zbr' has to point to on-flash node. The node may sit in a bud and
+        * may even be in a write buffer, so we have to take care about this.
+        */
+       wbuf = ubifs_get_wbuf(c, zbr->lnum);
+       if (wbuf)
+               err = ubifs_read_node_wbuf(wbuf, node, type, zbr->len,
+                                          zbr->lnum, zbr->offs);
+       else
+               err = ubifs_read_node(c, node, type, zbr->len, zbr->lnum,
+                                     zbr->offs);
 
        if (err) {
-               dbg_tnc("key %s", DBGKEY(key));
+               dbg_tnck(key, "key ");
                return err;
        }
 
@@ -425,9 +477,9 @@ int ubifs_tnc_read_node(struct ubifs_info *c, struct ubifs_zbranch *zbr,
        if (!keys_eq(c, key, &key1)) {
                ubifs_err("bad key in node at LEB %d:%d",
                          zbr->lnum, zbr->offs);
-               dbg_tnc("looked for key %s found node's key %s",
-                       DBGKEY(key), DBGKEY1(&key1));
-               dbg_dump_node(c, node);
+               dbg_tnck(key, "looked for key ");
+               dbg_tnck(&key1, "but found node's key ");
+               ubifs_dump_node(c, node);
                return -EINVAL;
        }
 
index 3eee07e0c4955b91a0aff3e0be7c5e0c6368dc3e..90b8ffacfbc318d32a12ff64545e05beaabf9235 100644 (file)
@@ -3,18 +3,7 @@
  *
  * Copyright (C) 2006-2008 Nokia Corporation.
  *
- * This program is free software; you can redistribute it and/or modify it
- * under the terms of the GNU General Public License version 2 as published by
- * the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
- * more details.
- *
- * You should have received a copy of the GNU General Public License along with
- * this program; if not, write to the Free Software Foundation, Inc., 51
- * Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
+ * SPDX-License-Identifier:    GPL-2.0+
  *
  * Authors: Artem Bityutskiy (Битюцкий Артём)
  *          Adrian Hunter
 /* The key is always at the same position in all keyed nodes */
 #define UBIFS_KEY_OFFSET offsetof(struct ubifs_ino_node, key)
 
+/* Garbage collector journal head number */
+#define UBIFS_GC_HEAD   0
+/* Base journal head number */
+#define UBIFS_BASE_HEAD 1
+/* Data journal head number */
+#define UBIFS_DATA_HEAD 2
+
 /*
  * LEB Properties Tree node types.
  *
@@ -401,9 +397,11 @@ enum {
  * Superblock flags.
  *
  * UBIFS_FLG_BIGLPT: if "big" LPT model is used if set
+ * UBIFS_FLG_SPACE_FIXUP: first-mount "fixup" of free space within LEBs needed
  */
 enum {
        UBIFS_FLG_BIGLPT = 0x02,
+       UBIFS_FLG_SPACE_FIXUP = 0x04,
 };
 
 /**
@@ -427,7 +425,7 @@ struct ubifs_ch {
        __u8 node_type;
        __u8 group_type;
        __u8 padding[2];
-} __attribute__ ((packed));
+} __packed;
 
 /**
  * union ubifs_dev_desc - device node descriptor.
@@ -441,7 +439,7 @@ struct ubifs_ch {
 union ubifs_dev_desc {
        __le32 new;
        __le64 huge;
-} __attribute__ ((packed));
+} __packed;
 
 /**
  * struct ubifs_ino_node - inode node.
@@ -502,7 +500,7 @@ struct ubifs_ino_node {
        __le16 compr_type;
        __u8 padding2[26]; /* Watch 'zero_ino_node_unused()' if changing! */
        __u8 data[];
-} __attribute__ ((packed));
+} __packed;
 
 /**
  * struct ubifs_dent_node - directory entry node.
@@ -526,8 +524,12 @@ struct ubifs_dent_node {
        __u8 type;
        __le16 nlen;
        __u8 padding2[4]; /* Watch 'zero_dent_node_unused()' if changing! */
+#ifndef __UBOOT__
        __u8 name[];
-} __attribute__ ((packed));
+#else
+       char name[];
+#endif
+} __packed;
 
 /**
  * struct ubifs_data_node - data node.
@@ -548,7 +550,7 @@ struct ubifs_data_node {
        __le16 compr_type;
        __u8 padding[2]; /* Watch 'zero_data_node_unused()' if changing! */
        __u8 data[];
-} __attribute__ ((packed));
+} __packed;
 
 /**
  * struct ubifs_trun_node - truncation node.
@@ -568,7 +570,7 @@ struct ubifs_trun_node {
        __u8 padding[12]; /* Watch 'zero_trun_node_unused()' if changing! */
        __le64 old_size;
        __le64 new_size;
-} __attribute__ ((packed));
+} __packed;
 
 /**
  * struct ubifs_pad_node - padding node.
@@ -579,7 +581,7 @@ struct ubifs_trun_node {
 struct ubifs_pad_node {
        struct ubifs_ch ch;
        __le32 pad_len;
-} __attribute__ ((packed));
+} __packed;
 
 /**
  * struct ubifs_sb_node - superblock node.
@@ -637,7 +639,7 @@ struct ubifs_sb_node {
        __u8 uuid[16];
        __le32 ro_compat_version;
        __u8 padding2[3968];
-} __attribute__ ((packed));
+} __packed;
 
 /**
  * struct ubifs_mst_node - master node.
@@ -704,7 +706,7 @@ struct ubifs_mst_node {
        __le32 idx_lebs;
        __le32 leb_cnt;
        __u8 padding[344];
-} __attribute__ ((packed));
+} __packed;
 
 /**
  * struct ubifs_ref_node - logical eraseblock reference node.
@@ -720,7 +722,7 @@ struct ubifs_ref_node {
        __le32 offs;
        __le32 jhead;
        __u8 padding[28];
-} __attribute__ ((packed));
+} __packed;
 
 /**
  * struct ubifs_branch - key/reference/length branch
@@ -733,8 +735,12 @@ struct ubifs_branch {
        __le32 lnum;
        __le32 offs;
        __le32 len;
+#ifndef __UBOOT__
        __u8 key[];
-} __attribute__ ((packed));
+#else
+       char key[];
+#endif
+} __packed;
 
 /**
  * struct ubifs_idx_node - indexing node.
@@ -747,8 +753,12 @@ struct ubifs_idx_node {
        struct ubifs_ch ch;
        __le16 child_cnt;
        __le16 level;
+#ifndef __UBOOT__
        __u8 branches[];
-} __attribute__ ((packed));
+#else
+       char branches[];
+#endif
+} __packed;
 
 /**
  * struct ubifs_cs_node - commit start node.
@@ -758,7 +768,7 @@ struct ubifs_idx_node {
 struct ubifs_cs_node {
        struct ubifs_ch ch;
        __le64 cmt_no;
-} __attribute__ ((packed));
+} __packed;
 
 /**
  * struct ubifs_orph_node - orphan node.
@@ -770,6 +780,6 @@ struct ubifs_orph_node {
        struct ubifs_ch ch;
        __le64 cmt_no;
        __le64 inos[];
-} __attribute__ ((packed));
+} __packed;
 
 #endif /* __UBIFS_MEDIA_H__ */
index 273c0a96383538ccb35421a2694d3f0e272612df..b91a6fd26f261118516ba6c49938c9ac2419ed49 100644 (file)
 #include "ubifs.h"
 #include <u-boot/zlib.h>
 
+#define __UBOOT__
+#include <linux/err.h>
+#include <linux/lzo.h>
+
 DECLARE_GLOBAL_DATA_PTR;
 
 /* compress.c */
@@ -44,20 +48,27 @@ static int gzip_decompress(const unsigned char *in, size_t in_len,
 /* Fake description object for the "none" compressor */
 static struct ubifs_compressor none_compr = {
        .compr_type = UBIFS_COMPR_NONE,
-       .name = "no compression",
+       .name = "none",
        .capi_name = "",
        .decompress = NULL,
 };
 
 static struct ubifs_compressor lzo_compr = {
        .compr_type = UBIFS_COMPR_LZO,
-       .name = "LZO",
+#ifndef __UBOOT__
+       .comp_mutex = &lzo_mutex,
+#endif
+       .name = "lzo",
        .capi_name = "lzo",
        .decompress = lzo1x_decompress_safe,
 };
 
 static struct ubifs_compressor zlib_compr = {
        .compr_type = UBIFS_COMPR_ZLIB,
+#ifndef __UBOOT__
+       .comp_mutex = &deflate_mutex,
+       .decomp_mutex = &inflate_mutex,
+#endif
        .name = "zlib",
        .capi_name = "deflate",
        .decompress = gzip_decompress,
@@ -66,6 +77,82 @@ static struct ubifs_compressor zlib_compr = {
 /* All UBIFS compressors */
 struct ubifs_compressor *ubifs_compressors[UBIFS_COMPR_TYPES_CNT];
 
+
+#ifdef __UBOOT__
+/* from mm/util.c */
+
+/**
+ * kmemdup - duplicate region of memory
+ *
+ * @src: memory region to duplicate
+ * @len: memory region length
+ * @gfp: GFP mask to use
+ */
+void *kmemdup(const void *src, size_t len, gfp_t gfp)
+{
+       void *p;
+
+       p = kmalloc(len, gfp);
+       if (p)
+               memcpy(p, src, len);
+       return p;
+}
+
+struct crypto_comp {
+       int compressor;
+};
+
+static inline struct crypto_comp *crypto_alloc_comp(const char *alg_name,
+                                               u32 type, u32 mask)
+{
+       struct ubifs_compressor *comp;
+       struct crypto_comp *ptr;
+       int i = 0;
+
+       ptr = malloc(sizeof(struct crypto_comp));
+       while (i < UBIFS_COMPR_TYPES_CNT) {
+               comp = ubifs_compressors[i];
+               if (!comp) {
+                       i++;
+                       continue;
+               }
+               if (strncmp(alg_name, comp->capi_name, strlen(alg_name)) == 0) {
+                       ptr->compressor = i;
+                       return ptr;
+               }
+               i++;
+       }
+       if (i >= UBIFS_COMPR_TYPES_CNT) {
+               ubifs_err("invalid compression type %s", alg_name);
+               free (ptr);
+               return NULL;
+       }
+       return ptr;
+}
+static inline int crypto_comp_decompress(struct crypto_comp *tfm,
+                               const u8 *src, unsigned int slen,
+                               u8 *dst, unsigned int *dlen)
+{
+       struct ubifs_compressor *compr = ubifs_compressors[tfm->compressor];
+       int err;
+
+       if (compr->compr_type == UBIFS_COMPR_NONE) {
+               memcpy(dst, src, slen);
+               *dlen = slen;
+               return 0;
+       }
+
+       err = compr->decompress(src, slen, dst, (size_t *)dlen);
+       if (err)
+               ubifs_err("cannot decompress %d bytes, compressor %s, "
+                         "error %d", slen, compr->name, err);
+
+       return err;
+
+       return 0;
+}
+#endif
+
 /**
  * ubifs_decompress - decompress data.
  * @in_buf: data to decompress
@@ -102,10 +189,15 @@ int ubifs_decompress(const void *in_buf, int in_len, void *out_buf,
                return 0;
        }
 
-       err = compr->decompress(in_buf, in_len, out_buf, (size_t *)out_len);
+       if (compr->decomp_mutex)
+               mutex_lock(compr->decomp_mutex);
+       err = crypto_comp_decompress(compr->cc, in_buf, in_len, out_buf,
+                                    (unsigned int *)out_len);
+       if (compr->decomp_mutex)
+               mutex_unlock(compr->decomp_mutex);
        if (err)
-               ubifs_err("cannot decompress %d bytes, compressor %s, "
-                         "error %d", in_len, compr->name, err);
+               ubifs_err("cannot decompress %d bytes, compressor %s, error %d",
+                         in_len, compr->name, err);
 
        return err;
 }
@@ -127,6 +219,15 @@ static int __init compr_init(struct ubifs_compressor *compr)
        ubifs_compressors[compr->compr_type]->decompress += gd->reloc_off;
 #endif
 
+       if (compr->capi_name) {
+               compr->cc = crypto_alloc_comp(compr->capi_name, 0, 0);
+               if (IS_ERR(compr->cc)) {
+                       ubifs_err("cannot initialize compressor %s, error %ld",
+                                 compr->name, PTR_ERR(compr->cc));
+                       return PTR_ERR(compr->cc);
+               }
+       }
+
        return 0;
 }
 
@@ -188,7 +289,9 @@ static int filldir(struct ubifs_info *c, const char *name, int namlen,
        }
        ctime_r((time_t *)&inode->i_mtime, filetime);
        printf("%9lld  %24.24s  ", inode->i_size, filetime);
+#ifndef __UBOOT__
        ubifs_iput(inode);
+#endif
 
        printf("%s\n", name);
 
@@ -562,7 +665,7 @@ static int read_block(struct inode *inode, void *addr, unsigned int block,
 dump:
        ubifs_err("bad data node (block %u, inode %lu)",
                  block, inode->i_ino);
-       dbg_dump_node(c, dn);
+       ubifs_dump_node(c, dn);
        return -EINVAL;
 }
 
index 221320157254d5a24459a977c6e384c185b3131b..acc6a404dde477a0b161aa88ad3c89a74ca7db73 100644 (file)
@@ -6,18 +6,7 @@
  * (C) Copyright 2008-2009
  * Stefan Roese, DENX Software Engineering, sr@denx.de.
  *
- * This program is free software; you can redistribute it and/or modify it
- * under the terms of the GNU General Public License version 2 as published by
- * the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful, but WITHOUT
- * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
- * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
- * more details.
- *
- * You should have received a copy of the GNU General Public License along with
- * this program; if not, write to the Free Software Foundation, Inc., 51
- * Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
+ * SPDX-License-Identifier:    GPL-2.0+
  *
  * Authors: Artem Bityutskiy (Битюцкий Артём)
  *          Adrian Hunter
 #ifndef __UBIFS_H__
 #define __UBIFS_H__
 
-#if 0  /* Enable for debugging output */
-#define CONFIG_UBIFS_FS_DEBUG
-#define CONFIG_UBIFS_FS_DEBUG_MSG_LVL  3
-#endif
-
+#define __UBOOT__
+#ifndef __UBOOT__
+#include <asm/div64.h>
+#include <linux/statfs.h>
+#include <linux/fs.h>
+#include <linux/err.h>
+#include <linux/sched.h>
+#include <linux/slab.h>
+#include <linux/vmalloc.h>
+#include <linux/spinlock.h>
+#include <linux/mutex.h>
+#include <linux/rwsem.h>
+#include <linux/mtd/ubi.h>
+#include <linux/pagemap.h>
+#include <linux/backing-dev.h>
+#include "ubifs-media.h"
+#else
 #include <ubi_uboot.h>
+
 #include <linux/ctype.h>
 #include <linux/time.h>
 #include <linux/math64.h>
@@ -70,13 +72,26 @@ void iput(struct inode *inode);
 #define atomic_long_dec(a)
 #define        atomic_long_sub(a, b)
 
+typedef unsigned long atomic_long_t;
+
 /* linux/include/time.h */
+#define NSEC_PER_SEC   1000000000L
+#define get_seconds()  0
+#define CURRENT_TIME_SEC       ((struct timespec) { get_seconds(), 0 })
 
 struct timespec {
        time_t  tv_sec;         /* seconds */
        long    tv_nsec;        /* nanoseconds */
 };
 
+static struct timespec current_fs_time(struct super_block *sb)
+{
+       struct timespec now;
+       now.tv_sec = 0;
+       now.tv_nsec = 0;
+       return now;
+};
+
 /* linux/include/dcache.h */
 
 /*
@@ -89,111 +104,245 @@ struct timespec {
 struct qstr {
        unsigned int hash;
        unsigned int len;
+#ifndef __UBOOT__
        const char *name;
+#else
+       char *name;
+#endif
+};
+
+/* include/linux/fs.h */
+
+/* Possible states of 'frozen' field */
+enum {
+       SB_UNFROZEN = 0,                /* FS is unfrozen */
+       SB_FREEZE_WRITE = 1,            /* Writes, dir ops, ioctls frozen */
+       SB_FREEZE_PAGEFAULT = 2,        /* Page faults stopped as well */
+       SB_FREEZE_FS = 3,               /* For internal FS use (e.g. to stop
+                                        * internal threads if needed) */
+       SB_FREEZE_COMPLETE = 4,         /* ->freeze_fs finished successfully */
 };
 
+#define SB_FREEZE_LEVELS (SB_FREEZE_COMPLETE - 1)
+
+struct sb_writers {
+#ifndef __UBOOT__
+       /* Counters for counting writers at each level */
+       struct percpu_counter   counter[SB_FREEZE_LEVELS];
+#endif
+       wait_queue_head_t       wait;           /* queue for waiting for
+                                                  writers / faults to finish */
+       int                     frozen;         /* Is sb frozen? */
+       wait_queue_head_t       wait_unfrozen;  /* queue for waiting for
+                                                  sb to be thawed */
+#ifdef CONFIG_DEBUG_LOCK_ALLOC
+       struct lockdep_map      lock_map[SB_FREEZE_LEVELS];
+#endif
+};
+
+struct address_space {
+       struct inode            *host;          /* owner: inode, block_device */
+#ifndef __UBOOT__
+       struct radix_tree_root  page_tree;      /* radix tree of all pages */
+#endif
+       spinlock_t              tree_lock;      /* and lock protecting it */
+       unsigned int            i_mmap_writable;/* count VM_SHARED mappings */
+       struct rb_root          i_mmap;         /* tree of private and shared mappings */
+       struct list_head        i_mmap_nonlinear;/*list VM_NONLINEAR mappings */
+       struct mutex            i_mmap_mutex;   /* protect tree, count, list */
+       /* Protected by tree_lock together with the radix tree */
+       unsigned long           nrpages;        /* number of total pages */
+       pgoff_t                 writeback_index;/* writeback starts here */
+       const struct address_space_operations *a_ops;   /* methods */
+       unsigned long           flags;          /* error bits/gfp mask */
+#ifndef __UBOOT__
+       struct backing_dev_info *backing_dev_info; /* device readahead, etc */
+#endif
+       spinlock_t              private_lock;   /* for use by the address_space */
+       struct list_head        private_list;   /* ditto */
+       void                    *private_data;  /* ditto */
+} __attribute__((aligned(sizeof(long))));
+
+/*
+ * Keep mostly read-only and often accessed (especially for
+ * the RCU path lookup and 'stat' data) fields at the beginning
+ * of the 'struct inode'
+ */
 struct inode {
-       struct hlist_node       i_hash;
-       struct list_head        i_list;
-       struct list_head        i_sb_list;
-       struct list_head        i_dentry;
+       umode_t                 i_mode;
+       unsigned short          i_opflags;
+       kuid_t                  i_uid;
+       kgid_t                  i_gid;
+       unsigned int            i_flags;
+
+#ifdef CONFIG_FS_POSIX_ACL
+       struct posix_acl        *i_acl;
+       struct posix_acl        *i_default_acl;
+#endif
+
+       const struct inode_operations   *i_op;
+       struct super_block      *i_sb;
+       struct address_space    *i_mapping;
+
+#ifdef CONFIG_SECURITY
+       void                    *i_security;
+#endif
+
+       /* Stat data, not accessed from path walking */
        unsigned long           i_ino;
-       unsigned int            i_nlink;
-       uid_t                   i_uid;
-       gid_t                   i_gid;
+       /*
+        * Filesystems may only read i_nlink directly.  They shall use the
+        * following functions for modification:
+        *
+        *    (set|clear|inc|drop)_nlink
+        *    inode_(inc|dec)_link_count
+        */
+       union {
+               const unsigned int i_nlink;
+               unsigned int __i_nlink;
+       };
        dev_t                   i_rdev;
-       u64                     i_version;
        loff_t                  i_size;
-#ifdef __NEED_I_SIZE_ORDERED
-       seqcount_t              i_size_seqcount;
-#endif
        struct timespec         i_atime;
        struct timespec         i_mtime;
        struct timespec         i_ctime;
-       unsigned int            i_blkbits;
-       unsigned short          i_bytes;
-       umode_t                 i_mode;
        spinlock_t              i_lock; /* i_blocks, i_bytes, maybe i_size */
+       unsigned short          i_bytes;
+       unsigned int            i_blkbits;
+       blkcnt_t                i_blocks;
+
+#ifdef __NEED_I_SIZE_ORDERED
+       seqcount_t              i_size_seqcount;
+#endif
+
+       /* Misc */
+       unsigned long           i_state;
        struct mutex            i_mutex;
-       struct rw_semaphore     i_alloc_sem;
-       const struct inode_operations   *i_op;
+
+       unsigned long           dirtied_when;   /* jiffies of first dirtying */
+
+       struct hlist_node       i_hash;
+       struct list_head        i_wb_list;      /* backing dev IO list */
+       struct list_head        i_lru;          /* inode LRU list */
+       struct list_head        i_sb_list;
+       union {
+               struct hlist_head       i_dentry;
+               struct rcu_head         i_rcu;
+       };
+       u64                     i_version;
+       atomic_t                i_count;
+       atomic_t                i_dio_count;
+       atomic_t                i_writecount;
        const struct file_operations    *i_fop; /* former ->i_op->default_file_ops */
-       struct super_block      *i_sb;
        struct file_lock        *i_flock;
+       struct address_space    i_data;
 #ifdef CONFIG_QUOTA
        struct dquot            *i_dquot[MAXQUOTAS];
 #endif
        struct list_head        i_devices;
-       int                     i_cindex;
+       union {
+               struct pipe_inode_info  *i_pipe;
+               struct block_device     *i_bdev;
+               struct cdev             *i_cdev;
+       };
 
        __u32                   i_generation;
 
-#ifdef CONFIG_DNOTIFY
-       unsigned long           i_dnotify_mask; /* Directory notify events */
-       struct dnotify_struct   *i_dnotify; /* for directory notifications */
+#ifdef CONFIG_FSNOTIFY
+       __u32                   i_fsnotify_mask; /* all events this inode cares about */
+       struct hlist_head       i_fsnotify_marks;
 #endif
 
-#ifdef CONFIG_INOTIFY
-       struct list_head        inotify_watches; /* watches on this inode */
-       struct mutex            inotify_mutex;  /* protects the watches list */
+#ifdef CONFIG_IMA
+       atomic_t                i_readcount; /* struct files open RO */
 #endif
+       void                    *i_private; /* fs or device private pointer */
+};
 
-       unsigned long           i_state;
-       unsigned long           dirtied_when;   /* jiffies of first dirtying */
-
-       unsigned int            i_flags;
-
-#ifdef CONFIG_SECURITY
-       void                    *i_security;
+struct super_operations {
+       struct inode *(*alloc_inode)(struct super_block *sb);
+       void (*destroy_inode)(struct inode *);
+
+       void (*dirty_inode) (struct inode *, int flags);
+       int (*write_inode) (struct inode *, struct writeback_control *wbc);
+       int (*drop_inode) (struct inode *);
+       void (*evict_inode) (struct inode *);
+       void (*put_super) (struct super_block *);
+       int (*sync_fs)(struct super_block *sb, int wait);
+       int (*freeze_fs) (struct super_block *);
+       int (*unfreeze_fs) (struct super_block *);
+#ifndef __UBOOT__
+       int (*statfs) (struct dentry *, struct kstatfs *);
 #endif
-       void                    *i_private; /* fs or device private pointer */
+       int (*remount_fs) (struct super_block *, int *, char *);
+       void (*umount_begin) (struct super_block *);
+
+#ifndef __UBOOT__
+       int (*show_options)(struct seq_file *, struct dentry *);
+       int (*show_devname)(struct seq_file *, struct dentry *);
+       int (*show_path)(struct seq_file *, struct dentry *);
+       int (*show_stats)(struct seq_file *, struct dentry *);
+#endif
+#ifdef CONFIG_QUOTA
+       ssize_t (*quota_read)(struct super_block *, int, char *, size_t, loff_t);
+       ssize_t (*quota_write)(struct super_block *, int, const char *, size_t, loff_t);
+#endif
+       int (*bdev_try_to_free_page)(struct super_block*, struct page*, gfp_t);
+       long (*nr_cached_objects)(struct super_block *, int);
+       long (*free_cached_objects)(struct super_block *, long, int);
 };
 
 struct super_block {
        struct list_head        s_list;         /* Keep this first */
        dev_t                   s_dev;          /* search index; _not_ kdev_t */
-       unsigned long           s_blocksize;
        unsigned char           s_blocksize_bits;
-       unsigned char           s_dirt;
-       unsigned long long      s_maxbytes;     /* Max file size */
+       unsigned long           s_blocksize;
+       loff_t                  s_maxbytes;     /* Max file size */
        struct file_system_type *s_type;
        const struct super_operations   *s_op;
-       struct dquot_operations *dq_op;
-       struct quotactl_ops     *s_qcop;
+       const struct dquot_operations   *dq_op;
+       const struct quotactl_ops       *s_qcop;
        const struct export_operations *s_export_op;
        unsigned long           s_flags;
        unsigned long           s_magic;
        struct dentry           *s_root;
        struct rw_semaphore     s_umount;
-       struct mutex            s_lock;
        int                     s_count;
-       int                     s_syncing;
-       int                     s_need_sync_fs;
+       atomic_t                s_active;
 #ifdef CONFIG_SECURITY
        void                    *s_security;
 #endif
-       struct xattr_handler    **s_xattr;
+       const struct xattr_handler **s_xattr;
 
        struct list_head        s_inodes;       /* all inodes */
-       struct list_head        s_dirty;        /* dirty inodes */
-       struct list_head        s_io;           /* parked for writeback */
-       struct list_head        s_more_io;      /* parked for more writeback */
-       struct hlist_head       s_anon;         /* anonymous dentries for (nfs) exporting */
-       struct list_head        s_files;
-       /* s_dentry_lru and s_nr_dentry_unused are protected by dcache_lock */
-       struct list_head        s_dentry_lru;   /* unused dentry lru */
-       int                     s_nr_dentry_unused;     /* # of dentry on lru */
-
+#ifndef __UBOOT__
+       struct hlist_bl_head    s_anon;         /* anonymous dentries for (nfs) exporting */
+#endif
+       struct list_head        s_mounts;       /* list of mounts; _not_ for fs use */
        struct block_device     *s_bdev;
+#ifndef __UBOOT__
+       struct backing_dev_info *s_bdi;
+#endif
        struct mtd_info         *s_mtd;
-       struct list_head        s_instances;
+       struct hlist_node       s_instances;
+#ifndef __UBOOT__
+       struct quota_info       s_dquot;        /* Diskquota specific options */
+#endif
 
-       int                     s_frozen;
-       wait_queue_head_t       s_wait_unfrozen;
+       struct sb_writers       s_writers;
 
        char s_id[32];                          /* Informational name */
+       u8 s_uuid[16];                          /* UUID */
 
        void                    *s_fs_info;     /* Filesystem private info */
+       unsigned int            s_max_links;
+#ifndef __UBOOT__
+       fmode_t                 s_mode;
+#endif
+
+       /* Granularity of c/m/atime in ns.
+          Cannot be worse than a second */
+       u32                s_time_gran;
 
        /*
         * The next field is for VFS *only*. No filesystems have any business
@@ -201,66 +350,83 @@ struct super_block {
         */
        struct mutex s_vfs_rename_mutex;        /* Kludge */
 
-       /* Granularity of c/m/atime in ns.
-          Cannot be worse than a second */
-       u32                s_time_gran;
-
        /*
         * Filesystem subtype.  If non-empty the filesystem type field
         * in /proc/mounts will be "type.subtype"
         */
        char *s_subtype;
 
+#ifndef __UBOOT__
        /*
         * Saved mount options for lazy filesystems using
         * generic_show_options()
         */
-       char *s_options;
+       char __rcu *s_options;
+#endif
+       const struct dentry_operations *s_d_op; /* default d_op for dentries */
+
+       /*
+        * Saved pool identifier for cleancache (-1 means none)
+        */
+       int cleancache_poolid;
+
+#ifndef __UBOOT__
+       struct shrinker s_shrink;       /* per-sb shrinker handle */
+#endif
+
+       /* Number of inodes with nlink == 0 but still referenced */
+       atomic_long_t s_remove_count;
+
+       /* Being remounted read-only */
+       int s_readonly_remount;
+
+       /* AIO completions deferred from interrupt context */
+       struct workqueue_struct *s_dio_done_wq;
+
+#ifndef __UBOOT__
+       /*
+        * Keep the lru lists last in the structure so they always sit on their
+        * own individual cachelines.
+        */
+       struct list_lru         s_dentry_lru ____cacheline_aligned_in_smp;
+       struct list_lru         s_inode_lru ____cacheline_aligned_in_smp;
+#endif
+       struct rcu_head         rcu;
 };
 
 struct file_system_type {
        const char *name;
        int fs_flags;
-       int (*get_sb) (struct file_system_type *, int,
-                      const char *, void *, struct vfsmount *);
+#define FS_REQUIRES_DEV                1 
+#define FS_BINARY_MOUNTDATA    2
+#define FS_HAS_SUBTYPE         4
+#define FS_USERNS_MOUNT                8       /* Can be mounted by userns root */
+#define FS_USERNS_DEV_MOUNT    16 /* A userns mount does not imply MNT_NODEV */
+#define FS_RENAME_DOES_D_MOVE  32768   /* FS will handle d_move() during rename() internally. */
+       struct dentry *(*mount) (struct file_system_type *, int,
+                      const char *, void *);
        void (*kill_sb) (struct super_block *);
        struct module *owner;
        struct file_system_type * next;
-       struct list_head fs_supers;
+       struct hlist_head fs_supers;
+
+#ifndef __UBOOT__
+       struct lock_class_key s_lock_key;
+       struct lock_class_key s_umount_key;
+       struct lock_class_key s_vfs_rename_key;
+       struct lock_class_key s_writers_key[SB_FREEZE_LEVELS];
+
+       struct lock_class_key i_lock_key;
+       struct lock_class_key i_mutex_key;
+       struct lock_class_key i_mutex_dir_key;
+#endif
 };
 
+/* include/linux/mount.h */
 struct vfsmount {
-       struct list_head mnt_hash;
-       struct vfsmount *mnt_parent;    /* fs we are mounted on */
-       struct dentry *mnt_mountpoint;  /* dentry of mountpoint */
        struct dentry *mnt_root;        /* root of the mounted tree */
        struct super_block *mnt_sb;     /* pointer to superblock */
-       struct list_head mnt_mounts;    /* list of children, anchored here */
-       struct list_head mnt_child;     /* and going through their mnt_child */
        int mnt_flags;
-       /* 4 bytes hole on 64bits arches */
-       const char *mnt_devname;        /* Name of device e.g. /dev/dsk/hda1 */
-       struct list_head mnt_list;
-       struct list_head mnt_expire;    /* link in fs-specific expiry list */
-       struct list_head mnt_share;     /* circular list of shared mounts */
-       struct list_head mnt_slave_list;/* list of slave mounts */
-       struct list_head mnt_slave;     /* slave list entry */
-       struct vfsmount *mnt_master;    /* slave is on master->mnt_slave_list */
-       struct mnt_namespace *mnt_ns;   /* containing namespace */
-       int mnt_id;                     /* mount identifier */
-       int mnt_group_id;               /* peer group identifier */
-       /*
-        * We put mnt_count & mnt_expiry_mark at the end of struct vfsmount
-        * to let these frequently modified fields in a separate cache line
-        * (so that reads of mnt_flags wont ping-pong on SMP machines)
-        */
-       int mnt_expiry_mark;            /* true if marked for expiry */
-       int mnt_pinned;
-       int mnt_ghosts;
-       /*
-        * This value is not stable unless all of the mnt_writers[] spinlocks
-        * are held, and all mnt_writer[]s on this mount have 0 as their ->count
-        */
 };
 
 struct path {
@@ -451,32 +617,35 @@ static inline ino_t parent_ino(struct dentry *dentry)
 
 /* debug.c */
 
-#define DEFINE_SPINLOCK(...)
 #define module_param_named(...)
 
 /* misc.h */
 #define mutex_lock_nested(...)
 #define mutex_unlock_nested(...)
 #define mutex_is_locked(...)   0
+#endif
 
 /* Version of this UBIFS implementation */
 #define UBIFS_VERSION 1
 
 /* Normal UBIFS messages */
-#ifdef CONFIG_UBIFS_SILENCE_MSG
-#define ubifs_msg(fmt, ...)
-#else
-#define ubifs_msg(fmt, ...) \
-               printk(KERN_NOTICE "UBIFS: " fmt "\n", ##__VA_ARGS__)
-#endif
+#define ubifs_msg(fmt, ...) pr_notice("UBIFS: " fmt "\n", ##__VA_ARGS__)
 /* UBIFS error messages */
-#define ubifs_err(fmt, ...)                                                  \
-       printk(KERN_ERR "UBIFS error (pid %d): %s: " fmt "\n", 0, \
+#ifndef __UBOOT__
+#define ubifs_err(fmt, ...)                                         \
+       pr_err("UBIFS error (pid %d): %s: " fmt "\n", current->pid, \
               __func__, ##__VA_ARGS__)
 /* UBIFS warning messages */
-#define ubifs_warn(fmt, ...)                                         \
-       printk(KERN_WARNING "UBIFS warning (pid %d): %s: " fmt "\n", \
-              0, __func__, ##__VA_ARGS__)
+#define ubifs_warn(fmt, ...)                                        \
+       pr_warn("UBIFS warning (pid %d): %s: " fmt "\n",            \
+               current->pid, __func__, ##__VA_ARGS__)
+#else
+#define ubifs_err(fmt, ...)                                         \
+       pr_err("UBIFS error: %s: " fmt "\n", __func__, ##__VA_ARGS__)
+/* UBIFS warning messages */
+#define ubifs_warn(fmt, ...)                                        \
+       pr_warn("UBIFS warning: %s: " fmt "\n", __func__, ##__VA_ARGS__)
+#endif
 
 /* UBIFS file system VFS magic number */
 #define UBIFS_SUPER_MAGIC 0x24051905
@@ -509,9 +678,6 @@ static inline ino_t parent_ino(struct dentry *dentry)
 #define INUM_WARN_WATERMARK 0xFFF00000
 #define INUM_WATERMARK      0xFFFFFF00
 
-/* Largest key size supported in this implementation */
-#define CUR_MAX_KEY_LEN UBIFS_SK_LEN
-
 /* Maximum number of entries in each LPT (LEB category) heap */
 #define LPT_HEAP_SZ 256
 
@@ -521,8 +687,9 @@ static inline ino_t parent_ino(struct dentry *dentry)
  */
 #define BGT_NAME_PATTERN "ubifs_bgt%d_%d"
 
-/* Default write-buffer synchronization timeout (5 secs) */
-#define DEFAULT_WBUF_TIMEOUT (5 * HZ)
+/* Write-buffer synchronization timeout interval in seconds */
+#define WBUF_TIMEOUT_SOFTLIMIT 3
+#define WBUF_TIMEOUT_HARDLIMIT 5
 
 /* Maximum possible inode number (only 32-bit inodes are supported now) */
 #define MAX_INUM 0xFFFFFFFF
@@ -530,12 +697,10 @@ static inline ino_t parent_ino(struct dentry *dentry)
 /* Number of non-data journal heads */
 #define NONDATA_JHEADS_CNT 2
 
-/* Garbage collector head */
-#define GCHD   0
-/* Base journal head number */
-#define BASEHD 1
-/* First "general purpose" journal head */
-#define DATAHD 2
+/* Shorter names for journal head numbers for internal usage */
+#define GCHD   UBIFS_GC_HEAD
+#define BASEHD UBIFS_BASE_HEAD
+#define DATAHD UBIFS_DATA_HEAD
 
 /* 'No change' value for 'ubifs_change_lp()' */
 #define LPROPS_NC 0x80000001
@@ -545,8 +710,12 @@ static inline ino_t parent_ino(struct dentry *dentry)
  * in TNC. However, when replaying, it is handy to introduce fake "truncation"
  * keys for truncation nodes because the code becomes simpler. So we define
  * %UBIFS_TRUN_KEY type.
+ *
+ * But otherwise, out of the journal reply scope, the truncation keys are
+ * invalid.
  */
-#define UBIFS_TRUN_KEY UBIFS_KEY_TYPES_CNT
+#define UBIFS_TRUN_KEY    UBIFS_KEY_TYPES_CNT
+#define UBIFS_INVALID_KEY UBIFS_KEY_TYPES_CNT
 
 /*
  * How much a directory entry/extended attribute entry adds to the parent/host
@@ -573,6 +742,12 @@ static inline ino_t parent_ino(struct dentry *dentry)
  */
 #define WORST_COMPR_FACTOR 2
 
+/*
+ * How much memory is needed for a buffer where we comress a data node.
+ */
+#define COMPRESSED_DATA_NODE_BUF_SZ \
+       (UBIFS_DATA_NODE_SZ + UBIFS_BLOCK_SIZE * WORST_COMPR_FACTOR)
+
 /* Maximum expected tree height for use by bottom_up_buf */
 #define BOTTOM_UP_HEIGHT 64
 
@@ -646,14 +821,14 @@ enum {
  * LPT cnode flag bits.
  *
  * DIRTY_CNODE: cnode is dirty
- * COW_CNODE: cnode is being committed and must be copied before writing
  * OBSOLETE_CNODE: cnode is being committed and has been copied (or deleted),
- * so it can (and must) be freed when the commit is finished
+ *                 so it can (and must) be freed when the commit is finished
+ * COW_CNODE: cnode is being committed and must be copied before writing
  */
 enum {
        DIRTY_CNODE    = 0,
-       COW_CNODE      = 1,
-       OBSOLETE_CNODE = 2,
+       OBSOLETE_CNODE = 1,
+       COW_CNODE      = 2,
 };
 
 /*
@@ -693,10 +868,10 @@ struct ubifs_old_idx {
 
 /* The below union makes it easier to deal with keys */
 union ubifs_key {
-       uint8_t u8[CUR_MAX_KEY_LEN];
-       uint32_t u32[CUR_MAX_KEY_LEN/4];
-       uint64_t u64[CUR_MAX_KEY_LEN/8];
-       __le32 j32[CUR_MAX_KEY_LEN/4];
+       uint8_t u8[UBIFS_SK_LEN];
+       uint32_t u32[UBIFS_SK_LEN/4];
+       uint64_t u64[UBIFS_SK_LEN/8];
+       __le32 j32[UBIFS_SK_LEN/4];
 };
 
 /**
@@ -805,9 +980,9 @@ struct ubifs_gced_idx_leb {
  * The @ui_size is a "shadow" variable for @inode->i_size and UBIFS uses
  * @ui_size instead of @inode->i_size. The reason for this is that UBIFS cannot
  * make sure @inode->i_size is always changed under @ui_mutex, because it
- * cannot call 'vmtruncate()' with @ui_mutex locked, because it would deadlock
- * with 'ubifs_writepage()' (see file.c). All the other inode fields are
- * changed under @ui_mutex, so they do not need "shadow" fields. Note, one
+ * cannot call 'truncate_setsize()' with @ui_mutex locked, because it would
+ * deadlock with 'ubifs_writepage()' (see file.c). All the other inode fields
+ * are changed under @ui_mutex, so they do not need "shadow" fields. Note, one
  * could consider to rework locking and base it on "shadow" fields.
  */
 struct ubifs_inode {
@@ -1068,17 +1243,19 @@ typedef int (*ubifs_lpt_scan_callback)(struct ubifs_info *c,
  * @offs: write-buffer offset in this logical eraseblock
  * @avail: number of bytes available in the write-buffer
  * @used:  number of used bytes in the write-buffer
- * @dtype: type of data stored in this LEB (%UBI_LONGTERM, %UBI_SHORTTERM,
- * %UBI_UNKNOWN)
+ * @size: write-buffer size (in [@c->min_io_size, @c->max_write_size] range)
  * @jhead: journal head the mutex belongs to (note, needed only to shut lockdep
  *         up by 'mutex_lock_nested()).
  * @sync_callback: write-buffer synchronization callback
  * @io_mutex: serializes write-buffer I/O
  * @lock: serializes @buf, @lnum, @offs, @avail, @used, @next_ino and @inodes
  *        fields
+ * @softlimit: soft write-buffer timeout interval
+ * @delta: hard and soft timeouts delta (the timer expire inteval is @softlimit
+ *         and @softlimit + @delta)
  * @timer: write-buffer timer
- * @timeout: timer expire interval in jiffies
- * @need_sync: it is set if its timer expired and needs sync
+ * @no_timer: non-zero if this write-buffer does not have a timer
+ * @need_sync: non-zero if the timer expired and the wbuf needs sync'ing
  * @next_ino: points to the next position of the following inode number
  * @inodes: stores the inode numbers of the nodes which are in wbuf
  *
@@ -1099,13 +1276,16 @@ struct ubifs_wbuf {
        int offs;
        int avail;
        int used;
-       int dtype;
+       int size;
        int jhead;
        int (*sync_callback)(struct ubifs_info *c, int lnum, int free, int pad);
        struct mutex io_mutex;
        spinlock_t lock;
-       int timeout;
-       int need_sync;
+//     ktime_t softlimit;
+//     unsigned long long delta;
+//     struct hrtimer timer;
+       unsigned int no_timer:1;
+       unsigned int need_sync:1;
        int next_ino;
        ino_t *inodes;
 };
@@ -1130,12 +1310,14 @@ struct ubifs_bud {
  * struct ubifs_jhead - journal head.
  * @wbuf: head's write-buffer
  * @buds_list: list of bud LEBs belonging to this journal head
+ * @grouped: non-zero if UBIFS groups nodes when writing to this journal head
  *
  * Note, the @buds list is protected by the @c->buds_lock.
  */
 struct ubifs_jhead {
        struct ubifs_wbuf wbuf;
        struct list_head buds_list;
+       unsigned int grouped:1;
 };
 
 /**
@@ -1171,6 +1353,9 @@ struct ubifs_zbranch {
  * @offs: offset of the corresponding indexing node
  * @len: length  of the corresponding indexing node
  * @zbranch: array of znode branches (@c->fanout elements)
+ *
+ * Note! The @lnum, @offs, and @len fields are not really needed - we have them
+ * only for internal consistency check. They could be removed to save some RAM.
  */
 struct ubifs_znode {
        struct ubifs_znode *parent;
@@ -1181,9 +1366,9 @@ struct ubifs_znode {
        int child_cnt;
        int iip;
        int alt;
-#ifdef CONFIG_UBIFS_FS_DEBUG
-       int lnum, offs, len;
-#endif
+       int lnum;
+       int offs;
+       int len;
        struct ubifs_zbranch zbranch[];
 };
 
@@ -1236,10 +1421,15 @@ struct ubifs_node_range {
  */
 struct ubifs_compressor {
        int compr_type;
-       char *name;
-       char *capi_name;
+       struct crypto_comp *cc;
+       struct mutex *comp_mutex;
+       struct mutex *decomp_mutex;
+       const char *name;
+       const char *capi_name;
+#ifdef __UBOOT__
        int (*decompress)(const unsigned char *in, size_t in_len,
                          unsigned char *out, size_t *out_len);
+#endif
 };
 
 /**
@@ -1313,6 +1503,8 @@ struct ubifs_budget_req {
  * @dnext: next orphan to delete
  * @inum: inode number
  * @new: %1 => added since the last commit, otherwise %0
+ * @cmt: %1 => commit pending, otherwise %0
+ * @del: %1 => delete pending, otherwise %0
  */
 struct ubifs_orphan {
        struct rb_node rb;
@@ -1321,7 +1513,9 @@ struct ubifs_orphan {
        struct ubifs_orphan *cnext;
        struct ubifs_orphan *dnext;
        ino_t inum;
-       int new;
+       unsigned new:1;
+       unsigned cmt:1;
+       unsigned del:1;
 };
 
 /**
@@ -1344,6 +1538,40 @@ struct ubifs_mount_opts {
        unsigned int compr_type:2;
 };
 
+/**
+ * struct ubifs_budg_info - UBIFS budgeting information.
+ * @idx_growth: amount of bytes budgeted for index growth
+ * @data_growth: amount of bytes budgeted for cached data
+ * @dd_growth: amount of bytes budgeted for cached data that will make
+ *             other data dirty
+ * @uncommitted_idx: amount of bytes were budgeted for growth of the index, but
+ *                   which still have to be taken into account because the index
+ *                   has not been committed so far
+ * @old_idx_sz: size of index on flash
+ * @min_idx_lebs: minimum number of LEBs required for the index
+ * @nospace: non-zero if the file-system does not have flash space (used as
+ *           optimization)
+ * @nospace_rp: the same as @nospace, but additionally means that even reserved
+ *              pool is full
+ * @page_budget: budget for a page (constant, nenver changed after mount)
+ * @inode_budget: budget for an inode (constant, nenver changed after mount)
+ * @dent_budget: budget for a directory entry (constant, nenver changed after
+ *               mount)
+ */
+struct ubifs_budg_info {
+       long long idx_growth;
+       long long data_growth;
+       long long dd_growth;
+       long long uncommitted_idx;
+       unsigned long long old_idx_sz;
+       int min_idx_lebs;
+       unsigned int nospace:1;
+       unsigned int nospace_rp:1;
+       int page_budget;
+       int inode_budget;
+       int dent_budget;
+};
+
 struct ubifs_debug_info;
 
 /**
@@ -1387,6 +1615,7 @@ struct ubifs_debug_info;
  * @cmt_wq: wait queue to sleep on if the log is full and a commit is running
  *
  * @big_lpt: flag that LPT is too big to write whole during commit
+ * @space_fixup: flag indicating that free space in LEBs needs to be cleaned up
  * @no_chk_data_crc: do not check CRCs when reading data nodes (except during
  *                   recovery)
  * @bulk_read: enable bulk-reads
@@ -1418,6 +1647,11 @@ struct ubifs_debug_info;
  * @bu_mutex: protects the pre-allocated bulk-read buffer and @c->bu
  * @bu: pre-allocated bulk-read information
  *
+ * @write_reserve_mutex: protects @write_reserve_buf
+ * @write_reserve_buf: on the write path we allocate memory, which might
+ *                     sometimes be unavailable, in which case we use this
+ *                     write reserve buffer
+ *
  * @log_lebs: number of logical eraseblocks in the log
  * @log_bytes: log size in bytes
  * @log_last: last LEB of the log
@@ -1439,43 +1673,34 @@ struct ubifs_debug_info;
  *
  * @min_io_size: minimal input/output unit size
  * @min_io_shift: number of bits in @min_io_size minus one
+ * @max_write_size: maximum amount of bytes the underlying flash can write at a
+ *                  time (MTD write buffer size)
+ * @max_write_shift: number of bits in @max_write_size minus one
  * @leb_size: logical eraseblock size in bytes
+ * @leb_start: starting offset of logical eraseblocks within physical
+ *             eraseblocks
  * @half_leb_size: half LEB size
+ * @idx_leb_size: how many bytes of an LEB are effectively available when it is
+ *                used to store indexing nodes (@leb_size - @max_idx_node_sz)
  * @leb_cnt: count of logical eraseblocks
  * @max_leb_cnt: maximum count of logical eraseblocks
  * @old_leb_cnt: count of logical eraseblocks before re-size
  * @ro_media: the underlying UBI volume is read-only
+ * @ro_mount: the file-system was mounted as read-only
+ * @ro_error: UBIFS switched to R/O mode because an error happened
  *
  * @dirty_pg_cnt: number of dirty pages (not used)
  * @dirty_zn_cnt: number of dirty znodes
  * @clean_zn_cnt: number of clean znodes
  *
- * @budg_idx_growth: amount of bytes budgeted for index growth
- * @budg_data_growth: amount of bytes budgeted for cached data
- * @budg_dd_growth: amount of bytes budgeted for cached data that will make
- *                  other data dirty
- * @budg_uncommitted_idx: amount of bytes were budgeted for growth of the index,
- *                        but which still have to be taken into account because
- *                        the index has not been committed so far
- * @space_lock: protects @budg_idx_growth, @budg_data_growth, @budg_dd_growth,
- *              @budg_uncommited_idx, @min_idx_lebs, @old_idx_sz, @lst,
- *              @nospace, and @nospace_rp;
- * @min_idx_lebs: minimum number of LEBs required for the index
- * @old_idx_sz: size of index on flash
+ * @space_lock: protects @bi and @lst
+ * @lst: lprops statistics
+ * @bi: budgeting information
  * @calc_idx_sz: temporary variable which is used to calculate new index size
  *               (contains accurate new index size at end of TNC commit start)
- * @lst: lprops statistics
- * @nospace: non-zero if the file-system does not have flash space (used as
- *           optimization)
- * @nospace_rp: the same as @nospace, but additionally means that even reserved
- *              pool is full
- *
- * @page_budget: budget for a page
- * @inode_budget: budget for an inode
- * @dent_budget: budget for a directory entry
  *
  * @ref_node_alsz: size of the LEB reference node aligned to the min. flash
- * I/O unit
+ *                 I/O unit
  * @mst_node_alsz: master node aligned size
  * @min_idx_node_sz: minimum indexing node aligned on 8-bytes boundary
  * @max_idx_node_sz: maximum indexing node aligned on 8-bytes boundary
@@ -1558,9 +1783,11 @@ struct ubifs_debug_info;
  *             previous commit start
  * @uncat_list: list of un-categorized LEBs
  * @empty_list: list of empty LEBs
- * @freeable_list: list of freeable non-index LEBs (free + dirty == leb_size)
- * @frdi_idx_list: list of freeable index LEBs (free + dirty == leb_size)
+ * @freeable_list: list of freeable non-index LEBs (free + dirty == @leb_size)
+ * @frdi_idx_list: list of freeable index LEBs (free + dirty == @leb_size)
  * @freeable_cnt: number of freeable LEBs in @freeable_list
+ * @in_a_category_cnt: count of lprops which are in a certain category, which
+ *                     basically meants that they were loaded from the flash
  *
  * @ltab_lnum: LEB number of LPT's own lprops table
  * @ltab_offs: offset of LPT's own lprops table
@@ -1577,25 +1804,29 @@ struct ubifs_debug_info;
  * @rp_uid: reserved pool user ID
  * @rp_gid: reserved pool group ID
  *
- * @empty: if the UBI device is empty
- * @replay_tree: temporary tree used during journal replay
+ * @empty: %1 if the UBI device is empty
+ * @need_recovery: %1 if the file-system needs recovery
+ * @replaying: %1 during journal replay
+ * @mounting: %1 while mounting
+ * @remounting_rw: %1 while re-mounting from R/O mode to R/W mode
  * @replay_list: temporary list used during journal replay
  * @replay_buds: list of buds to replay
  * @cs_sqnum: sequence number of first node in the log (commit start node)
  * @replay_sqnum: sequence number of node currently being replayed
- * @need_recovery: file-system needs recovery
- * @replaying: set to %1 during journal replay
- * @unclean_leb_list: LEBs to recover when mounting ro to rw
- * @rcvrd_mst_node: recovered master node to write when mounting ro to rw
+ * @unclean_leb_list: LEBs to recover when re-mounting R/O mounted FS to R/W
+ *                    mode
+ * @rcvrd_mst_node: recovered master node to write when re-mounting R/O mounted
+ *                  FS to R/W mode
  * @size_tree: inode size information for recovery
- * @remounting_rw: set while remounting from ro to rw (sb flags have MS_RDONLY)
- * @always_chk_crc: always check CRCs (while mounting and remounting rw)
  * @mount_opts: UBIFS-specific mount options
  *
  * @dbg: debugging-related information
  */
 struct ubifs_info {
        struct super_block *vfs_sb;
+#ifndef __UBOOT__
+       struct backing_dev_info bdi;
+#endif
 
        ino_t highest_inum;
        unsigned long long max_sqnum;
@@ -1628,6 +1859,7 @@ struct ubifs_info {
        wait_queue_head_t cmt_wq;
 
        unsigned int big_lpt:1;
+       unsigned int space_fixup:1;
        unsigned int no_chk_data_crc:1;
        unsigned int bulk_read:1;
        unsigned int default_compr:2;
@@ -1657,6 +1889,9 @@ struct ubifs_info {
        struct mutex bu_mutex;
        struct bu_info bu;
 
+       struct mutex write_reserve_mutex;
+       void *write_reserve_buf;
+
        int log_lebs;
        long long log_bytes;
        int log_last;
@@ -1678,28 +1913,27 @@ struct ubifs_info {
 
        int min_io_size;
        int min_io_shift;
+       int max_write_size;
+       int max_write_shift;
        int leb_size;
+       int leb_start;
        int half_leb_size;
+       int idx_leb_size;
        int leb_cnt;
        int max_leb_cnt;
        int old_leb_cnt;
-       int ro_media;
+       unsigned int ro_media:1;
+       unsigned int ro_mount:1;
+       unsigned int ro_error:1;
+
+       atomic_long_t dirty_pg_cnt;
+       atomic_long_t dirty_zn_cnt;
+       atomic_long_t clean_zn_cnt;
 
-       long long budg_idx_growth;
-       long long budg_data_growth;
-       long long budg_dd_growth;
-       long long budg_uncommitted_idx;
        spinlock_t space_lock;
-       int min_idx_lebs;
-       unsigned long long old_idx_sz;
-       unsigned long long calc_idx_sz;
        struct ubifs_lp_stats lst;
-       unsigned int nospace:1;
-       unsigned int nospace_rp:1;
-
-       int page_budget;
-       int inode_budget;
-       int dent_budget;
+       struct ubifs_budg_info bi;
+       unsigned long long calc_idx_sz;
 
        int ref_node_alsz;
        int mst_node_alsz;
@@ -1785,6 +2019,7 @@ struct ubifs_info {
        struct list_head freeable_list;
        struct list_head frdi_idx_list;
        int freeable_cnt;
+       int in_a_category_cnt;
 
        int ltab_lnum;
        int ltab_offs;
@@ -1798,31 +2033,32 @@ struct ubifs_info {
 
        long long rp_size;
        long long report_rp_size;
-       uid_t rp_uid;
-       gid_t rp_gid;
+       kuid_t rp_uid;
+       kgid_t rp_gid;
 
        /* The below fields are used only during mounting and re-mounting */
-       int empty;
-       struct rb_root replay_tree;
+       unsigned int empty:1;
+       unsigned int need_recovery:1;
+       unsigned int replaying:1;
+       unsigned int mounting:1;
+       unsigned int remounting_rw:1;
        struct list_head replay_list;
        struct list_head replay_buds;
        unsigned long long cs_sqnum;
        unsigned long long replay_sqnum;
-       int need_recovery;
-       int replaying;
        struct list_head unclean_leb_list;
        struct ubifs_mst_node *rcvrd_mst_node;
        struct rb_root size_tree;
-       int remounting_rw;
-       int always_chk_crc;
        struct ubifs_mount_opts mount_opts;
 
-#ifdef CONFIG_UBIFS_FS_DEBUG
+#ifndef __UBOOT__
        struct ubifs_debug_info *dbg;
 #endif
 };
 
+extern struct list_head ubifs_infos;
 extern spinlock_t ubifs_infos_lock;
+extern atomic_long_t ubifs_clean_zn_cnt;
 extern struct kmem_cache *ubifs_inode_slab;
 extern const struct super_operations ubifs_super_operations;
 extern const struct address_space_operations ubifs_file_address_operations;
@@ -1836,16 +2072,23 @@ extern struct ubifs_compressor *ubifs_compressors[UBIFS_COMPR_TYPES_CNT];
 
 /* io.c */
 void ubifs_ro_mode(struct ubifs_info *c, int err);
+int ubifs_leb_read(const struct ubifs_info *c, int lnum, void *buf, int offs,
+                  int len, int even_ebadmsg);
+int ubifs_leb_write(struct ubifs_info *c, int lnum, const void *buf, int offs,
+                   int len);
+int ubifs_leb_change(struct ubifs_info *c, int lnum, const void *buf, int len);
+int ubifs_leb_unmap(struct ubifs_info *c, int lnum);
+int ubifs_leb_map(struct ubifs_info *c, int lnum);
+int ubifs_is_mapped(const struct ubifs_info *c, int lnum);
 int ubifs_wbuf_write_nolock(struct ubifs_wbuf *wbuf, void *buf, int len);
-int ubifs_wbuf_seek_nolock(struct ubifs_wbuf *wbuf, int lnum, int offs,
-                          int dtype);
+int ubifs_wbuf_seek_nolock(struct ubifs_wbuf *wbuf, int lnum, int offs);
 int ubifs_wbuf_init(struct ubifs_info *c, struct ubifs_wbuf *wbuf);
 int ubifs_read_node(const struct ubifs_info *c, void *buf, int type, int len,
                    int lnum, int offs);
 int ubifs_read_node_wbuf(struct ubifs_wbuf *wbuf, void *buf, int type, int len,
                         int lnum, int offs);
 int ubifs_write_node(struct ubifs_info *c, void *node, int len, int lnum,
-                    int offs, int dtype);
+                    int offs);
 int ubifs_check_node(const struct ubifs_info *c, const void *buf, int lnum,
                     int offs, int quiet, int must_chk_crc);
 void ubifs_prepare_node(struct ubifs_info *c, void *buf, int len, int pad);
@@ -1859,7 +2102,7 @@ int ubifs_sync_wbufs_by_inode(struct ubifs_info *c, struct inode *inode);
 
 /* scan.c */
 struct ubifs_scan_leb *ubifs_scan(const struct ubifs_info *c, int lnum,
-                                 int offs, void *sbuf);
+                                 int offs, void *sbuf, int quiet);
 void ubifs_scan_destroy(struct ubifs_scan_leb *sleb);
 int ubifs_scan_a_node(const struct ubifs_info *c, void *buf, int len, int lnum,
                      int offs, int quiet);
@@ -1921,7 +2164,7 @@ long long ubifs_reported_space(const struct ubifs_info *c, long long free);
 long long ubifs_calc_available(const struct ubifs_info *c, int min_idx_lebs);
 
 /* find.c */
-int ubifs_find_free_space(struct ubifs_info *c, int min_space, int *free,
+int ubifs_find_free_space(struct ubifs_info *c, int min_space, int *offs,
                          int squeeze);
 int ubifs_find_free_leb_for_idx(struct ubifs_info *c);
 int ubifs_find_dirty_leb(struct ubifs_info *c, struct ubifs_lprops *ret_lp,
@@ -1983,8 +2226,13 @@ int ubifs_tnc_read_node(struct ubifs_info *c, struct ubifs_zbranch *zbr,
 int ubifs_tnc_start_commit(struct ubifs_info *c, struct ubifs_zbranch *zroot);
 int ubifs_tnc_end_commit(struct ubifs_info *c);
 
+#ifndef __UBOOT__
 /* shrinker.c */
-int ubifs_shrinker(int nr_to_scan, gfp_t gfp_mask);
+unsigned long ubifs_shrink_scan(struct shrinker *shrink,
+                               struct shrink_control *sc);
+unsigned long ubifs_shrink_count(struct shrinker *shrink,
+                                struct shrink_control *sc);
+#endif
 
 /* commit.c */
 int ubifs_bg_thread(void *info);
@@ -2003,6 +2251,7 @@ int ubifs_write_master(struct ubifs_info *c);
 int ubifs_read_superblock(struct ubifs_info *c);
 struct ubifs_sb_node *ubifs_read_sb_node(struct ubifs_info *c);
 int ubifs_write_sb_node(struct ubifs_info *c, struct ubifs_sb_node *sup);
+int ubifs_fixup_free_space(struct ubifs_info *c);
 
 /* replay.c */
 int ubifs_validate_entry(struct ubifs_info *c,
@@ -2084,14 +2333,15 @@ const struct ubifs_lprops *ubifs_fast_find_free(struct ubifs_info *c);
 const struct ubifs_lprops *ubifs_fast_find_empty(struct ubifs_info *c);
 const struct ubifs_lprops *ubifs_fast_find_freeable(struct ubifs_info *c);
 const struct ubifs_lprops *ubifs_fast_find_frdi_idx(struct ubifs_info *c);
+int ubifs_calc_dark(const struct ubifs_info *c, int spc);
 
 /* file.c */
-int ubifs_fsync(struct file *file, struct dentry *dentry, int datasync);
+int ubifs_fsync(struct file *file, loff_t start, loff_t end, int datasync);
 int ubifs_setattr(struct dentry *dentry, struct iattr *attr);
 
 /* dir.c */
 struct inode *ubifs_new_inode(struct ubifs_info *c, const struct inode *dir,
-                             int mode);
+                             umode_t mode);
 int ubifs_getattr(struct vfsmount *mnt, struct dentry *dentry,
                  struct kstat *stat);
 
@@ -2111,11 +2361,11 @@ int ubifs_iput(struct inode *inode);
 int ubifs_recover_master_node(struct ubifs_info *c);
 int ubifs_write_rcvrd_mst_node(struct ubifs_info *c);
 struct ubifs_scan_leb *ubifs_recover_leb(struct ubifs_info *c, int lnum,
-                                        int offs, void *sbuf, int grouped);
+                                        int offs, void *sbuf, int jhead);
 struct ubifs_scan_leb *ubifs_recover_log_leb(struct ubifs_info *c, int lnum,
                                             int offs, void *sbuf);
-int ubifs_recover_inl_heads(const struct ubifs_info *c, void *sbuf);
-int ubifs_clean_lebs(const struct ubifs_info *c, void *sbuf);
+int ubifs_recover_inl_heads(struct ubifs_info *c, void *sbuf);
+int ubifs_clean_lebs(struct ubifs_info *c, void *sbuf);
 int ubifs_rcvry_gc_commit(struct ubifs_info *c);
 int ubifs_recover_size_accum(struct ubifs_info *c, union ubifs_key *key,
                             int deletion, loff_t new_size);
@@ -2131,24 +2381,22 @@ long ubifs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg);
 
 /* compressor.c */
 int __init ubifs_compressors_init(void);
-void __exit ubifs_compressors_exit(void);
+void ubifs_compressors_exit(void);
 void ubifs_compress(const void *in_buf, int in_len, void *out_buf, int *out_len,
                    int *compr_type);
 int ubifs_decompress(const void *buf, int len, void *out, int *out_len,
                     int compr_type);
 
+#include "debug.h"
+#include "misc.h"
+#include "key.h"
+
+#ifdef __UBOOT__
 /* these are used in cmd_ubifs.c */
 int ubifs_init(void);
-int ubifs_mount(char *vol_name);
+int uboot_ubifs_mount(char *vol_name);
 void ubifs_umount(struct ubifs_info *c);
 int ubifs_ls(char *dir_name);
 int ubifs_load(char *filename, u32 addr, u32 size);
-
-#include "debug.h"
-#include "misc.h"
-#include "key.h"
-
-/* todo: Move these to a common U-Boot header */
-int lzo1x_decompress_safe(const unsigned char *in, size_t in_len,
-                         unsigned char *out, size_t *out_len);
+#endif
 #endif /* !__UBIFS_H__ */
index 25a3d3a3d1529629ad52daf433c9528bdaea9f36..be81d3824af5cc3e143075bc55698145edcbde2e 100644 (file)
@@ -4,13 +4,14 @@
  *  NAND family Bad Block Management (BBM) header file
  *    - Bad Block Table (BBT) implementation
  *
- *  Copyright (c) 2005-2007 Samsung Electronics
+ *  Copyright © 2005 Samsung Electronics
  *  Kyungmin Park <kyungmin.park@samsung.com>
  *
- *  Copyright (c) 2000-2005
+ *  Copyright © 2000-2005
  *  Thomas Gleixner <tglx@linuxtronix.de>
  *
  * SPDX-License-Identifier:    GPL-2.0+
+ *
  */
 #ifndef __LINUX_MTD_BBM_H
 #define __LINUX_MTD_BBM_H
 
 /**
  * struct nand_bbt_descr - bad block table descriptor
- * @param options      options for this descriptor
- * @param pages                the page(s) where we find the bbt, used with
- *                     option BBT_ABSPAGE when bbt is searched,
- *                     then we store the found bbts pages here.
- *                     Its an array and supports up to 8 chips now
- * @param offs         offset of the pattern in the oob area of the page
- * @param veroffs      offset of the bbt version counter in the oob are of the page
- * @param version      version read from the bbt page during scan
- * @param len          length of the pattern, if 0 no pattern check is performed
- * @param maxblocks    maximum number of blocks to search for a bbt. This number of
- *                     blocks is reserved at the end of the device
- *                     where the tables are written.
- * @param reserved_block_code  if non-0, this pattern denotes a reserved
- *                     (rather than bad) block in the stored bbt
- * @param pattern      pattern to identify bad block table or factory marked
- *                     good / bad blocks, can be NULL, if len = 0
+ * @options:   options for this descriptor
+ * @pages:     the page(s) where we find the bbt, used with option BBT_ABSPAGE
+ *             when bbt is searched, then we store the found bbts pages here.
+ *             Its an array and supports up to 8 chips now
+ * @offs:      offset of the pattern in the oob area of the page
+ * @veroffs:   offset of the bbt version counter in the oob are of the page
+ * @version:   version read from the bbt page during scan
+ * @len:       length of the pattern, if 0 no pattern check is performed
+ * @maxblocks: maximum number of blocks to search for a bbt. This number of
+ *             blocks is reserved at the end of the device where the tables are
+ *             written.
+ * @reserved_block_code: if non-0, this pattern denotes a reserved (rather than
+ *              bad) block in the stored bbt
+ * @pattern:   pattern to identify bad block table or factory marked good /
+ *             bad blocks, can be NULL, if len = 0
  *
  * Descriptor for the bad block table marker and the descriptor for the
  * pattern which identifies good and bad blocks. The assumption is made
@@ -81,10 +81,6 @@ struct nand_bbt_descr {
  * with NAND_BBT_CREATE.
  */
 #define NAND_BBT_CREATE_EMPTY  0x00000400
-/* Search good / bad pattern through all pages of a block */
-#define NAND_BBT_SCANALLPAGES  0x00000800
-/* Scan block empty during good / bad block scan */
-#define NAND_BBT_SCANEMPTY     0x00001000
 /* Write bbt if neccecary */
 #define NAND_BBT_WRITE         0x00002000
 /* Read and write back block contents when writing bbt */
@@ -122,22 +118,27 @@ struct nand_bbt_descr {
 /*
  * Constants for oob configuration
  */
-#define ONENAND_BADBLOCK_POS   0
+#define NAND_SMALL_BADBLOCK_POS                5
+#define NAND_LARGE_BADBLOCK_POS                0
+#define ONENAND_BADBLOCK_POS           0
 
 /*
  * Bad block scanning errors
  */
-#define ONENAND_BBT_READ_ERROR          1
-#define ONENAND_BBT_READ_ECC_ERROR      2
-#define ONENAND_BBT_READ_FATAL_ERROR    4
+#define ONENAND_BBT_READ_ERROR         1
+#define ONENAND_BBT_READ_ECC_ERROR     2
+#define ONENAND_BBT_READ_FATAL_ERROR   4
 
 /**
- * struct bbt_info - [GENERIC] Bad Block Table data structure
- * @param bbt_erase_shift      [INTERN] number of address bits in a bbt entry
- * @param badblockpos          [INTERN] position of the bad block marker in the oob area
- * @param bbt                  [INTERN] bad block table pointer
- * @param badblock_pattern     [REPLACEABLE] bad block scan pattern used for initial bad block scan
- * @param priv                 [OPTIONAL] pointer to private bbm date
+ * struct bbm_info - [GENERIC] Bad Block Table data structure
+ * @bbt_erase_shift:   [INTERN] number of address bits in a bbt entry
+ * @badblockpos:       [INTERN] position of the bad block marker in the oob area
+ * @options:           options for this descriptor
+ * @bbt:               [INTERN] bad block table pointer
+ * @isbad_bbt:         function to determine if a block is bad
+ * @badblock_pattern:  [REPLACEABLE] bad block scan pattern used for
+ *                     initial bad block scan
+ * @priv:              [OPTIONAL] pointer to private bbm date
  */
 struct bbm_info {
        int bbt_erase_shift;
@@ -146,7 +147,7 @@ struct bbm_info {
 
        uint8_t *bbt;
 
-       int (*isbad_bbt) (struct mtd_info * mtd, loff_t ofs, int allowbbt);
+       int (*isbad_bbt)(struct mtd_info *mtd, loff_t ofs, int allowbbt);
 
        /* TODO Add more NAND specific fileds */
        struct nand_bbt_descr *badblock_pattern;
@@ -155,7 +156,7 @@ struct bbm_info {
 };
 
 /* OneNAND BBT interface */
-extern int onenand_scan_bbt (struct mtd_info *mtd, struct nand_bbt_descr *bd);
-extern int onenand_default_bbt (struct mtd_info *mtd);
+extern int onenand_scan_bbt(struct mtd_info *mtd, struct nand_bbt_descr *bd);
+extern int onenand_default_bbt(struct mtd_info *mtd);
 
-#endif                         /* __LINUX_MTD_BBM_H */
+#endif /* __LINUX_MTD_BBM_H */
index c92b4ddc9be01ac3c1c4a08a0982024c524bb79d..195a4a5426b6b928b40bbe62dd3fafae708a0fd4 100644 (file)
 struct mtd_info *mtd_concat_create(
     struct mtd_info *subdev[],  /* subdevices to concatenate */
     int num_devs,               /* number of subdevices      */
+#ifndef __UBOOT__
     const char *name);          /* name for the new device   */
+#else
+    char *name);          /* name for the new device   */
+#endif
 
 void mtd_concat_destroy(struct mtd_info *mtd);
 
diff --git a/include/linux/mtd/flashchip.h b/include/linux/mtd/flashchip.h
new file mode 100644 (file)
index 0000000..7028ee1
--- /dev/null
@@ -0,0 +1,105 @@
+/*
+ * Copyright © 2000      Red Hat UK Limited
+ * Copyright © 2000-2010 David Woodhouse <dwmw2@infradead.org>
+ *
+ * SPDX-License-Identifier:    GPL-2.0+
+ *
+ */
+
+#ifndef __MTD_FLASHCHIP_H__
+#define __MTD_FLASHCHIP_H__
+
+#define __UBOOT__
+#ifndef __UBOOT__
+/* For spinlocks. sched.h includes spinlock.h from whichever directory it
+ * happens to be in - so we don't have to care whether we're on 2.2, which
+ * has asm/spinlock.h, or 2.4, which has linux/spinlock.h
+ */
+#include <linux/sched.h>
+#include <linux/mutex.h>
+#endif
+
+typedef enum {
+       FL_READY,
+       FL_STATUS,
+       FL_CFI_QUERY,
+       FL_JEDEC_QUERY,
+       FL_ERASING,
+       FL_ERASE_SUSPENDING,
+       FL_ERASE_SUSPENDED,
+       FL_WRITING,
+       FL_WRITING_TO_BUFFER,
+       FL_OTP_WRITE,
+       FL_WRITE_SUSPENDING,
+       FL_WRITE_SUSPENDED,
+       FL_PM_SUSPENDED,
+       FL_SYNCING,
+       FL_UNLOADING,
+       FL_LOCKING,
+       FL_UNLOCKING,
+       FL_POINT,
+       FL_XIP_WHILE_ERASING,
+       FL_XIP_WHILE_WRITING,
+       FL_SHUTDOWN,
+       /* These 2 come from nand_state_t, which has been unified here */
+       FL_READING,
+       FL_CACHEDPRG,
+       /* These 4 come from onenand_state_t, which has been unified here */
+       FL_RESETING,
+       FL_OTPING,
+       FL_PREPARING_ERASE,
+       FL_VERIFYING_ERASE,
+
+       FL_UNKNOWN
+} flstate_t;
+
+
+
+/* NOTE: confusingly, this can be used to refer to more than one chip at a time,
+   if they're interleaved.  This can even refer to individual partitions on
+   the same physical chip when present. */
+
+struct flchip {
+       unsigned long start; /* Offset within the map */
+       //      unsigned long len;
+       /* We omit len for now, because when we group them together
+          we insist that they're all of the same size, and the chip size
+          is held in the next level up. If we get more versatile later,
+          it'll make it a damn sight harder to find which chip we want from
+          a given offset, and we'll want to add the per-chip length field
+          back in.
+       */
+       int ref_point_counter;
+       flstate_t state;
+       flstate_t oldstate;
+
+       unsigned int write_suspended:1;
+       unsigned int erase_suspended:1;
+       unsigned long in_progress_block_addr;
+
+       struct mutex mutex;
+#ifndef __UBOOT__
+       wait_queue_head_t wq; /* Wait on here when we're waiting for the chip
+                            to be ready */
+#endif
+       int word_write_time;
+       int buffer_write_time;
+       int erase_time;
+
+       int word_write_time_max;
+       int buffer_write_time_max;
+       int erase_time_max;
+
+       void *priv;
+};
+
+/* This is used to handle contention on write/erase operations
+   between partitions of the same physical chip. */
+struct flchip_shared {
+       struct mutex lock;
+       struct flchip *writing;
+       struct flchip *erasing;
+};
+
+
+#endif /* __MTD_FLASHCHIP_H__ */
index a65b6815515f2e087baa4ad41b860e7b9ee38286..b7b47570ea57e49075701566942cd1057bbdab8e 100644 (file)
@@ -1,48 +1,45 @@
 /*
- * Copyright (C) 1999-2003 David Woodhouse <dwmw2@infradead.org> et al.
+ * Copyright © 1999-2010 David Woodhouse <dwmw2@infradead.org> et al.
  *
  * Released under GPL
+ *
  */
 
 #ifndef __MTD_MTD_H__
 #define __MTD_MTD_H__
 
+#define __UBOOT__
+#ifndef __UBOOT__
 #include <linux/types.h>
-#include <div64.h>
+#include <linux/uio.h>
+#include <linux/notifier.h>
+#include <linux/device.h>
+
+#include <mtd/mtd-abi.h>
+
+#include <asm/div64.h>
+#else
+#include <linux/compat.h>
 #include <mtd/mtd-abi.h>
 #include <asm/errno.h>
+#include <div64.h>
 
-#define MTD_CHAR_MAJOR 90
-#define MTD_BLOCK_MAJOR 31
 #define MAX_MTD_DEVICES 32
+#endif
 
 #define MTD_ERASE_PENDING      0x01
 #define MTD_ERASING            0x02
 #define MTD_ERASE_SUSPEND      0x04
-#define MTD_ERASE_DONE          0x08
-#define MTD_ERASE_FAILED        0x10
+#define MTD_ERASE_DONE         0x08
+#define MTD_ERASE_FAILED       0x10
 
-#define MTD_FAIL_ADDR_UNKNOWN  -1LL
+#define MTD_FAIL_ADDR_UNKNOWN -1LL
 
 /*
- * Enumeration for NAND/OneNAND flash chip state
+ * If the erase fails, fail_addr might indicate exactly which block failed. If
+ * fail_addr = MTD_FAIL_ADDR_UNKNOWN, the failure was not at the device level
+ * or was not specific to any particular block.
  */
-enum {
-       FL_READY,
-       FL_READING,
-       FL_WRITING,
-       FL_ERASING,
-       FL_SYNCING,
-       FL_CACHEDPRG,
-       FL_RESETING,
-       FL_UNLOCKING,
-       FL_LOCKING,
-       FL_PM_SUSPENDED,
-};
-
-/* If the erase fails, fail_addr might indicate exactly which block failed.  If
-   fail_addr = MTD_FAIL_ADDR_UNKNOWN, the failure was not at the device level or was not
-   specific to any particular block. */
 struct erase_info {
        struct mtd_info *mtd;
        uint64_t addr;
@@ -50,8 +47,8 @@ struct erase_info {
        uint64_t fail_addr;
        u_long time;
        u_long retries;
-       u_int dev;
-       u_int cell;
+       unsigned dev;
+       unsigned cell;
        void (*callback) (struct erase_info *self);
        u_long priv;
        u_char state;
@@ -60,9 +57,9 @@ struct erase_info {
 };
 
 struct mtd_erase_region_info {
-       uint64_t offset;                        /* At which this region starts, from the beginning of the MTD */
-       u_int32_t erasesize;            /* For this region */
-       u_int32_t numblocks;            /* Number of blocks of erasesize in this region */
+       uint64_t offset;                /* At which this region starts, from the beginning of the MTD */
+       uint32_t erasesize;             /* For this region */
+       uint32_t numblocks;             /* Number of blocks of erasesize in this region */
        unsigned long *lockmap;         /* If keeping bitmap of locks */
 };
 
@@ -81,7 +78,7 @@ struct mtd_erase_region_info {
  * @datbuf:    data buffer - if NULL only oob data are read/written
  * @oobbuf:    oob data buffer
  *
- * Note, it is allowed to read more then one OOB area at one go, but not write.
+ * Note, it is allowed to read more than one OOB area at one go, but not write.
  * The interface assumes that the OOB write requests program only one page's
  * OOB area.
  */
@@ -109,26 +106,30 @@ struct mtd_oob_ops {
 #endif
 
 /*
- * ECC layout control structure. Exported to userspace for
- * diagnosis and to allow creation of raw images
+ * Internal ECC layout control structure. For historical reasons, there is a
+ * similar, smaller struct nand_ecclayout_user (in mtd-abi.h) that is retained
+ * for export to user-space via the ECCGETLAYOUT ioctl.
+ * nand_ecclayout should be expandable in the future simply by the above macros.
  */
 struct nand_ecclayout {
-       uint32_t eccbytes;
-       uint32_t eccpos[MTD_MAX_ECCPOS_ENTRIES_LARGE];
-       uint32_t oobavail;
+       __u32 eccbytes;
+       __u32 eccpos[MTD_MAX_ECCPOS_ENTRIES_LARGE];
+       __u32 oobavail;
        struct nand_oobfree oobfree[MTD_MAX_OOBFREE_ENTRIES_LARGE];
 };
 
+struct module; /* only needed for owner field in mtd_info */
+
 struct mtd_info {
        u_char type;
-       u_int32_t flags;
-       uint64_t size;   /* Total size of the MTD */
+       uint32_t flags;
+       uint64_t size;   // Total size of the MTD
 
        /* "Major" erase size for the device. Naïve users may take this
         * to be the only erase size available, or may use the more detailed
         * information below if they desire
         */
-       u_int32_t erasesize;
+       uint32_t erasesize;
        /* Minimal writable flash unit size. In case of NOR flash it is 1 (even
         * though individual bits can be cleared), in case of NAND flash it is
         * one NAND page (or half, or one-fourths of it), in case of ECC-ed NOR
@@ -136,10 +137,31 @@ struct mtd_info {
         * Any driver registering a struct mtd_info must ensure a writesize of
         * 1 or larger.
         */
-       u_int32_t writesize;
+       uint32_t writesize;
+
+       /*
+        * Size of the write buffer used by the MTD. MTD devices having a write
+        * buffer can write multiple writesize chunks at a time. E.g. while
+        * writing 4 * writesize bytes to a device with 2 * writesize bytes
+        * buffer the MTD driver can (but doesn't have to) do 2 writesize
+        * operations, but not 4. Currently, all NANDs have writebufsize
+        * equivalent to writesize (NAND page size). Some NOR flashes do have
+        * writebufsize greater than writesize.
+        */
+       uint32_t writebufsize;
 
-       u_int32_t oobsize;   /* Amount of OOB data per block (e.g. 16) */
-       u_int32_t oobavail;  /* Available OOB bytes per block */
+       uint32_t oobsize;   // Amount of OOB data per block (e.g. 16)
+       uint32_t oobavail;  // Available OOB bytes per block
+
+       /*
+        * If erasesize is a power of 2 then the shift is stored in
+        * erasesize_shift otherwise erasesize_shift is zero. Ditto writesize.
+        */
+       unsigned int erasesize_shift;
+       unsigned int writesize_shift;
+       /* Masks based on erasesize_shift and writesize_shift */
+       unsigned int erasesize_mask;
+       unsigned int writesize_mask;
 
        /*
         * read ops return -EUCLEAN if max number of bitflips corrected on any
@@ -150,13 +172,20 @@ struct mtd_info {
         */
        unsigned int bitflip_threshold;
 
-       /* Kernel-only stuff starts here. */
+       // Kernel-only stuff starts here.
+#ifndef __UBOOT__
        const char *name;
+#else
+       char *name;
+#endif
        int index;
 
        /* ECC layout structure pointer - read only! */
        struct nand_ecclayout *ecclayout;
 
+       /* the ecc step size. */
+       unsigned int ecc_step_size;
+
        /* max number of correctible bit errors per ecc step */
        unsigned int ecc_strength;
 
@@ -171,44 +200,51 @@ struct mtd_info {
         * wrappers instead.
         */
        int (*_erase) (struct mtd_info *mtd, struct erase_info *instr);
+#ifndef __UBOOT__
        int (*_point) (struct mtd_info *mtd, loff_t from, size_t len,
-                       size_t *retlen, void **virt, phys_addr_t *phys);
-       void (*_unpoint) (struct mtd_info *mtd, loff_t from, size_t len);
+                      size_t *retlen, void **virt, resource_size_t *phys);
+       int (*_unpoint) (struct mtd_info *mtd, loff_t from, size_t len);
+#endif
+       unsigned long (*_get_unmapped_area) (struct mtd_info *mtd,
+                                            unsigned long len,
+                                            unsigned long offset,
+                                            unsigned long flags);
        int (*_read) (struct mtd_info *mtd, loff_t from, size_t len,
-                    size_t *retlen, u_char *buf);
+                     size_t *retlen, u_char *buf);
        int (*_write) (struct mtd_info *mtd, loff_t to, size_t len,
-                     size_t *retlen, const u_char *buf);
-
-       /* In blackbox flight recorder like scenarios we want to make successful
-          writes in interrupt context. panic_write() is only intended to be
-          called when its known the kernel is about to panic and we need the
-          write to succeed. Since the kernel is not going to be running for much
-          longer, this function can break locks and delay to ensure the write
-          succeeds (but not sleep). */
-
-       int (*_panic_write) (struct mtd_info *mtd, loff_t to, size_t len, size_t *retlen, const u_char *buf);
-
+                      size_t *retlen, const u_char *buf);
+       int (*_panic_write) (struct mtd_info *mtd, loff_t to, size_t len,
+                            size_t *retlen, const u_char *buf);
        int (*_read_oob) (struct mtd_info *mtd, loff_t from,
-                        struct mtd_oob_ops *ops);
+                         struct mtd_oob_ops *ops);
        int (*_write_oob) (struct mtd_info *mtd, loff_t to,
-                        struct mtd_oob_ops *ops);
+                          struct mtd_oob_ops *ops);
        int (*_get_fact_prot_info) (struct mtd_info *mtd, struct otp_info *buf,
-                                  size_t len);
+                                   size_t len);
        int (*_read_fact_prot_reg) (struct mtd_info *mtd, loff_t from,
-                                  size_t len, size_t *retlen, u_char *buf);
+                                   size_t len, size_t *retlen, u_char *buf);
        int (*_get_user_prot_info) (struct mtd_info *mtd, struct otp_info *buf,
-                                  size_t len);
+                                   size_t len);
        int (*_read_user_prot_reg) (struct mtd_info *mtd, loff_t from,
-                                  size_t len, size_t *retlen, u_char *buf);
-       int (*_write_user_prot_reg) (struct mtd_info *mtd, loff_t to, size_t len,
-                                   size_t *retlen, u_char *buf);
+                                   size_t len, size_t *retlen, u_char *buf);
+       int (*_write_user_prot_reg) (struct mtd_info *mtd, loff_t to,
+                                    size_t len, size_t *retlen, u_char *buf);
        int (*_lock_user_prot_reg) (struct mtd_info *mtd, loff_t from,
-                                  size_t len);
+                                   size_t len);
+#ifndef __UBOOT__
+       int (*_writev) (struct mtd_info *mtd, const struct kvec *vecs,
+                       unsigned long count, loff_t to, size_t *retlen);
+#endif
        void (*_sync) (struct mtd_info *mtd);
        int (*_lock) (struct mtd_info *mtd, loff_t ofs, uint64_t len);
        int (*_unlock) (struct mtd_info *mtd, loff_t ofs, uint64_t len);
+       int (*_is_locked) (struct mtd_info *mtd, loff_t ofs, uint64_t len);
        int (*_block_isbad) (struct mtd_info *mtd, loff_t ofs);
        int (*_block_markbad) (struct mtd_info *mtd, loff_t ofs);
+#ifndef __UBOOT__
+       int (*_suspend) (struct mtd_info *mtd);
+       void (*_resume) (struct mtd_info *mtd);
+#endif
        /*
         * If the driver is something smart, like UBI, it may need to maintain
         * its own reference counting. The below functions are only for driver.
@@ -216,16 +252,12 @@ struct mtd_info {
        int (*_get_device) (struct mtd_info *mtd);
        void (*_put_device) (struct mtd_info *mtd);
 
-/* XXX U-BOOT XXX */
-#if 0
-       /* kvec-based read/write methods.
-          NB: The 'count' parameter is the number of _vectors_, each of
-          which contains an (ofs, len) tuple.
-       */
-       int (*writev) (struct mtd_info *mtd, const struct kvec *vecs, unsigned long count, loff_t to, size_t *retlen);
-#endif
-/* XXX U-BOOT XXX */
-#if 0
+#ifndef __UBOOT__
+       /* Backing device capabilities for this device
+        * - provides mmap capabilities
+        */
+       struct backing_dev_info *backing_dev_info;
+
        struct notifier_block reboot_notifier;  /* default mode before reboot */
 #endif
 
@@ -237,10 +269,20 @@ struct mtd_info {
        void *priv;
 
        struct module *owner;
+#ifndef __UBOOT__
+       struct device dev;
+#endif
        int usecount;
 };
 
 int mtd_erase(struct mtd_info *mtd, struct erase_info *instr);
+#ifndef __UBOOT__
+int mtd_point(struct mtd_info *mtd, loff_t from, size_t len, size_t *retlen,
+             void **virt, resource_size_t *phys);
+int mtd_unpoint(struct mtd_info *mtd, loff_t from, size_t len);
+#endif
+unsigned long mtd_get_unmapped_area(struct mtd_info *mtd, unsigned long len,
+                                   unsigned long offset, unsigned long flags);
 int mtd_read(struct mtd_info *mtd, loff_t from, size_t len, size_t *retlen,
             u_char *buf);
 int mtd_write(struct mtd_info *mtd, loff_t to, size_t len, size_t *retlen,
@@ -273,8 +315,7 @@ int mtd_write_user_prot_reg(struct mtd_info *mtd, loff_t to, size_t len,
                            size_t *retlen, u_char *buf);
 int mtd_lock_user_prot_reg(struct mtd_info *mtd, loff_t from, size_t len);
 
-/* XXX U-BOOT XXX */
-#if 0
+#ifndef __UBOOT__
 int mtd_writev(struct mtd_info *mtd, const struct kvec *vecs,
               unsigned long count, loff_t to, size_t *retlen);
 #endif
@@ -291,22 +332,59 @@ int mtd_is_locked(struct mtd_info *mtd, loff_t ofs, uint64_t len);
 int mtd_block_isbad(struct mtd_info *mtd, loff_t ofs);
 int mtd_block_markbad(struct mtd_info *mtd, loff_t ofs);
 
+#ifndef __UBOOT__
+static inline int mtd_suspend(struct mtd_info *mtd)
+{
+       return mtd->_suspend ? mtd->_suspend(mtd) : 0;
+}
+
+static inline void mtd_resume(struct mtd_info *mtd)
+{
+       if (mtd->_resume)
+               mtd->_resume(mtd);
+}
+#endif
+
 static inline uint32_t mtd_div_by_eb(uint64_t sz, struct mtd_info *mtd)
 {
+       if (mtd->erasesize_shift)
+               return sz >> mtd->erasesize_shift;
        do_div(sz, mtd->erasesize);
        return sz;
 }
 
 static inline uint32_t mtd_mod_by_eb(uint64_t sz, struct mtd_info *mtd)
 {
+       if (mtd->erasesize_shift)
+               return sz & mtd->erasesize_mask;
        return do_div(sz, mtd->erasesize);
 }
 
+static inline uint32_t mtd_div_by_ws(uint64_t sz, struct mtd_info *mtd)
+{
+       if (mtd->writesize_shift)
+               return sz >> mtd->writesize_shift;
+       do_div(sz, mtd->writesize);
+       return sz;
+}
+
+static inline uint32_t mtd_mod_by_ws(uint64_t sz, struct mtd_info *mtd)
+{
+       if (mtd->writesize_shift)
+               return sz & mtd->writesize_mask;
+       return do_div(sz, mtd->writesize);
+}
+
 static inline int mtd_has_oob(const struct mtd_info *mtd)
 {
        return mtd->_read_oob && mtd->_write_oob;
 }
 
+static inline int mtd_type_is_nand(const struct mtd_info *mtd)
+{
+       return mtd->type == MTD_NANDFLASH || mtd->type == MTD_MLCNANDFLASH;
+}
+
 static inline int mtd_can_have_bb(const struct mtd_info *mtd)
 {
        return !!mtd->_block_isbad;
@@ -314,27 +392,36 @@ static inline int mtd_can_have_bb(const struct mtd_info *mtd)
 
        /* Kernel-side ioctl definitions */
 
-extern int add_mtd_device(struct mtd_info *mtd);
-extern int del_mtd_device (struct mtd_info *mtd);
-
+struct mtd_partition;
+struct mtd_part_parser_data;
+
+extern int mtd_device_parse_register(struct mtd_info *mtd,
+                                    const char * const *part_probe_types,
+                                    struct mtd_part_parser_data *parser_data,
+                                    const struct mtd_partition *defparts,
+                                    int defnr_parts);
+#define mtd_device_register(master, parts, nr_parts)   \
+       mtd_device_parse_register(master, NULL, NULL, parts, nr_parts)
+extern int mtd_device_unregister(struct mtd_info *master);
 extern struct mtd_info *get_mtd_device(struct mtd_info *mtd, int num);
+extern int __get_mtd_device(struct mtd_info *mtd);
+extern void __put_mtd_device(struct mtd_info *mtd);
 extern struct mtd_info *get_mtd_device_nm(const char *name);
-
 extern void put_mtd_device(struct mtd_info *mtd);
-extern void mtd_get_len_incl_bad(struct mtd_info *mtd, uint64_t offset,
-                                const uint64_t length, uint64_t *len_incl_bad,
-                                int *truncated);
-/* XXX U-BOOT XXX */
-#if 0
+
+
+#ifndef __UBOOT__
 struct mtd_notifier {
        void (*add)(struct mtd_info *mtd);
        void (*remove)(struct mtd_info *mtd);
        struct list_head list;
 };
 
+
 extern void register_mtd_user (struct mtd_notifier *new);
 extern int unregister_mtd_user (struct mtd_notifier *old);
 #endif
+void *mtd_kmalloc_up_to(const struct mtd_info *mtd, size_t *size);
 
 #ifdef CONFIG_MTD_PARTITIONS
 void mtd_erase_callback(struct erase_info *instr);
@@ -346,6 +433,7 @@ static inline void mtd_erase_callback(struct erase_info *instr)
 }
 #endif
 
+#ifdef __UBOOT__
 /*
  * Debugging macro and defines
  */
@@ -372,7 +460,11 @@ static inline void mtd_erase_callback(struct erase_info *instr)
 #define pr_info(args...)       MTDDEBUG(MTD_DEBUG_LEVEL0, args)
 #define pr_warn(args...)       MTDDEBUG(MTD_DEBUG_LEVEL0, args)
 #define pr_err(args...)                MTDDEBUG(MTD_DEBUG_LEVEL0, args)
-
+#define pr_crit(args...)       MTDDEBUG(MTD_DEBUG_LEVEL0, args)
+#define pr_cont(args...)       MTDDEBUG(MTD_DEBUG_LEVEL0, args)
+#define pr_notice(args...)     MTDDEBUG(MTD_DEBUG_LEVEL0, args)
+#endif
 static inline int mtd_is_bitflip(int err) {
        return err == -EUCLEAN;
 }
@@ -385,4 +477,10 @@ static inline int mtd_is_bitflip_or_eccerr(int err) {
        return mtd_is_bitflip(err) || mtd_is_eccerr(err);
 }
 
+#ifdef __UBOOT__
+/* drivers/mtd/mtdcore.h */
+int add_mtd_device(struct mtd_info *mtd);
+int add_mtd_partitions(struct mtd_info *, const struct mtd_partition *, int);
+int del_mtd_partitions(struct mtd_info *);
+#endif
 #endif /* __MTD_MTD_H__ */
index 991bd8e63ed0dccf945062b5085065004e667abe..489c703b9b2cf3c3703877cfa1347e9eede453ee 100644 (file)
@@ -5,9 +5,7 @@
  *                        Steven J. Hill <sjhill@realitydiluted.com>
  *                       Thomas Gleixner <tglx@linutronix.de>
  *
- * This program is free software; you can redistribute it and/or modify
- * it under the terms of the GNU General Public License version 2 as
- * published by the Free Software Foundation.
+ * SPDX-License-Identifier:    GPL-2.0+
  *
  * Info:
  *     Contains standard defines and IDs for NAND flash devices
 #ifndef __LINUX_MTD_NAND_H
 #define __LINUX_MTD_NAND_H
 
+#define __UBOOT__
+#ifndef __UBOOT__
+#include <linux/wait.h>
+#include <linux/spinlock.h>
+#include <linux/mtd/mtd.h>
+#include <linux/mtd/flashchip.h>
+#include <linux/mtd/bbm.h>
+#else
 #include "config.h"
 
 #include "linux/compat.h"
 #include "linux/mtd/mtd.h"
+#include "linux/mtd/flashchip.h"
 #include "linux/mtd/bbm.h"
-
+#endif
 
 struct mtd_info;
 struct nand_flash_dev;
 /* Scan and identify a NAND device */
-extern int nand_scan (struct mtd_info *mtd, int max_chips);
-/* Separate phases of nand_scan(), allowing board driver to intervene
- * and override command or ECC setup according to flash type */
+extern int nand_scan(struct mtd_info *mtd, int max_chips);
+/*
+ * Separate phases of nand_scan(), allowing board driver to intervene
+ * and override command or ECC setup according to flash type.
+ */
 extern int nand_scan_ident(struct mtd_info *mtd, int max_chips,
-                          const struct nand_flash_dev *table);
+                          struct nand_flash_dev *table);
 extern int nand_scan_tail(struct mtd_info *mtd);
 
 /* Free resources held by the NAND device */
@@ -41,12 +50,23 @@ extern void nand_release(struct mtd_info *mtd);
 /* Internal helper for board drivers which need to override command function */
 extern void nand_wait_ready(struct mtd_info *mtd);
 
+#ifndef __UBOOT__
+/* locks all blocks present in the device */
+extern int nand_lock(struct mtd_info *mtd, loff_t ofs, uint64_t len);
+
+/* unlocks specified locked blocks */
+extern int nand_unlock(struct mtd_info *mtd, loff_t ofs, uint64_t len);
+
+/* The maximum number of NAND chips in an array */
+#define NAND_MAX_CHIPS         8
+#endif
+
 /*
  * This constant declares the max. oobsize / page, which
  * is supported now. If you add a chip with bigger oobsize/page
  * adjust this accordingly.
  */
-#define NAND_MAX_OOBSIZE       640
+#define NAND_MAX_OOBSIZE       744
 #define NAND_MAX_PAGESIZE      8192
 
 /*
@@ -76,7 +96,6 @@ extern void nand_wait_ready(struct mtd_info *mtd);
 #define NAND_CMD_READOOB       0x50
 #define NAND_CMD_ERASE1                0x60
 #define NAND_CMD_STATUS                0x70
-#define NAND_CMD_STATUS_MULTI  0x71
 #define NAND_CMD_SEQIN         0x80
 #define NAND_CMD_RNDIN         0x85
 #define NAND_CMD_READID                0x90
@@ -87,10 +106,8 @@ extern void nand_wait_ready(struct mtd_info *mtd);
 #define NAND_CMD_RESET         0xff
 
 #define NAND_CMD_LOCK          0x2a
-#define NAND_CMD_LOCK_TIGHT    0x2c
 #define NAND_CMD_UNLOCK1       0x23
 #define NAND_CMD_UNLOCK2       0x24
-#define NAND_CMD_LOCK_STATUS   0x7a
 
 /* Extended commands for large page devices */
 #define NAND_CMD_READSTART     0x30
@@ -164,21 +181,12 @@ typedef enum {
 /* Chip has copy back function */
 #define NAND_COPYBACK          0x00000010
 /*
- * AND Chip which has 4 banks and a confusing page / block
- * assignment. See Renesas datasheet for further information.
+ * Chip requires ready check on read (for auto-incremented sequential read).
+ * True only for small page devices; large page devices do not support
+ * autoincrement.
  */
-#define NAND_IS_AND            0x00000020
-/*
- * Chip has a array of 4 pages which can be read without
- * additional ready /busy waits.
- */
-#define NAND_4PAGE_ARRAY       0x00000040
-/*
- * Chip requires that BBT is periodically rewritten to prevent
- * bits from adjacent blocks from 'leaking' in altering data.
- * This happens with the Renesas AG-AND chips, possibly others.
- */
-#define BBT_AUTO_REFRESH       0x00000080
+#define NAND_NEED_READRDY      0x00000100
+
 /* Chip does not allow subpage writes */
 #define NAND_NO_SUBPAGE_WRITE  0x00000200
 
@@ -189,16 +197,13 @@ typedef enum {
 #define NAND_ROM               0x00000800
 
 /* Device supports subpage reads */
-#define NAND_SUBPAGE_READ       0x00001000
+#define NAND_SUBPAGE_READ      0x00001000
 
 /* Options valid for Samsung large page devices */
-#define NAND_SAMSUNG_LP_OPTIONS \
-       (NAND_NO_PADDING | NAND_CACHEPRG | NAND_COPYBACK)
+#define NAND_SAMSUNG_LP_OPTIONS NAND_CACHEPRG
 
 /* Macros to identify the above */
-#define NAND_MUST_PAD(chip) (!(chip->options & NAND_NO_PADDING))
 #define NAND_HAS_CACHEPROG(chip) ((chip->options & NAND_CACHEPRG))
-#define NAND_HAS_COPYBACK(chip) ((chip->options & NAND_COPYBACK))
 #define NAND_HAS_SUBPAGE_READ(chip) ((chip->options & NAND_SUBPAGE_READ))
 
 /* Non chip related options */
@@ -211,6 +216,13 @@ typedef enum {
 #define NAND_OWN_BUFFERS       0x00020000
 /* Chip may not exist, so silence any errors in scan */
 #define NAND_SCAN_SILENT_NODEV 0x00040000
+/*
+ * Autodetect nand buswidth with readid/onfi.
+ * This suppose the driver will configure the hardware in 8 bits mode
+ * when calling nand_scan_ident, and update its configuration
+ * before calling nand_scan_tail.
+ */
+#define NAND_BUSWIDTH_AUTO      0x00080000
 
 /* Options set by nand scan */
 /* bbt has already been read */
@@ -221,10 +233,15 @@ typedef enum {
 /* Cell info constants */
 #define NAND_CI_CHIPNR_MSK     0x03
 #define NAND_CI_CELLTYPE_MSK   0x0C
+#define NAND_CI_CELLTYPE_SHIFT 2
 
 /* Keep gcc happy */
 struct nand_chip;
 
+/* ONFI features */
+#define ONFI_FEATURE_16_BIT_BUS                (1 << 0)
+#define ONFI_FEATURE_EXT_PARAM_PAGE    (1 << 7)
+
 /* ONFI timing mode, used in both asynchronous and synchronous mode */
 #define ONFI_TIMING_MODE_0             (1 << 0)
 #define ONFI_TIMING_MODE_1             (1 << 1)
@@ -237,9 +254,15 @@ struct nand_chip;
 /* ONFI feature address */
 #define ONFI_FEATURE_ADDR_TIMING_MODE  0x1
 
+/* Vendor-specific feature address (Micron) */
+#define ONFI_FEATURE_ADDR_READ_RETRY   0x89
+
 /* ONFI subfeature parameters length */
 #define ONFI_SUBFEATURE_PARAM_LEN      4
 
+/* ONFI optional commands SET/GET FEATURES supported? */
+#define ONFI_OPT_CMD_SET_GET_FEATURES  (1 << 2)
+
 struct nand_onfi_params {
        /* rev info and features block */
        /* 'O' 'N' 'F' 'I'  */
@@ -247,7 +270,10 @@ struct nand_onfi_params {
        __le16 revision;
        __le16 features;
        __le16 opt_cmd;
-       u8 reserved[22];
+       u8 reserved0[2];
+       __le16 ext_param_page_length; /* since ONFI 2.1 */
+       u8 num_of_param_pages;        /* since ONFI 2.1 */
+       u8 reserved1[17];
 
        /* manufacturer information block */
        char manufacturer[12];
@@ -291,19 +317,74 @@ struct nand_onfi_params {
        __le16 io_pin_capacitance_typ;
        __le16 input_pin_capacitance_typ;
        u8 input_pin_capacitance_max;
-       u8 driver_strenght_support;
+       u8 driver_strength_support;
        __le16 t_int_r;
        __le16 t_ald;
        u8 reserved4[7];
 
        /* vendor */
-       u8 reserved5[90];
+       __le16 vendor_revision;
+       u8 vendor[88];
 
        __le16 crc;
-} __attribute__((packed));
+} __packed;
 
 #define ONFI_CRC_BASE  0x4F4E
 
+/* Extended ECC information Block Definition (since ONFI 2.1) */
+struct onfi_ext_ecc_info {
+       u8 ecc_bits;
+       u8 codeword_size;
+       __le16 bb_per_lun;
+       __le16 block_endurance;
+       u8 reserved[2];
+} __packed;
+
+#define ONFI_SECTION_TYPE_0    0       /* Unused section. */
+#define ONFI_SECTION_TYPE_1    1       /* for additional sections. */
+#define ONFI_SECTION_TYPE_2    2       /* for ECC information. */
+struct onfi_ext_section {
+       u8 type;
+       u8 length;
+} __packed;
+
+#define ONFI_EXT_SECTION_MAX 8
+
+/* Extended Parameter Page Definition (since ONFI 2.1) */
+struct onfi_ext_param_page {
+       __le16 crc;
+       u8 sig[4];             /* 'E' 'P' 'P' 'S' */
+       u8 reserved0[10];
+       struct onfi_ext_section sections[ONFI_EXT_SECTION_MAX];
+
+       /*
+        * The actual size of the Extended Parameter Page is in
+        * @ext_param_page_length of nand_onfi_params{}.
+        * The following are the variable length sections.
+        * So we do not add any fields below. Please see the ONFI spec.
+        */
+} __packed;
+
+struct nand_onfi_vendor_micron {
+       u8 two_plane_read;
+       u8 read_cache;
+       u8 read_unique_id;
+       u8 dq_imped;
+       u8 dq_imped_num_settings;
+       u8 dq_imped_feat_addr;
+       u8 rb_pulldown_strength;
+       u8 rb_pulldown_strength_feat_addr;
+       u8 rb_pulldown_strength_num_settings;
+       u8 otp_mode;
+       u8 otp_page_start;
+       u8 otp_data_prot_addr;
+       u8 otp_num_pages;
+       u8 otp_feat_addr;
+       u8 read_retry_options;
+       u8 reserved[72];
+       u8 param_revision;
+} __packed;
+
 /**
  * struct nand_hw_control - Control structure for hardware controller (e.g ECC generator) shared among independent devices
  * @lock:               protection lock
@@ -313,12 +394,11 @@ struct nand_onfi_params {
  *                     when a hw controller is available.
  */
 struct nand_hw_control {
-/* XXX U-BOOT XXX */
-#if 0
-       spinlock_t       lock;
+       spinlock_t lock;
+       struct nand_chip *active;
+#ifndef __UBOOT__
        wait_queue_head_t wq;
 #endif
-       struct nand_chip *active;
 };
 
 /**
@@ -344,6 +424,7 @@ struct nand_hw_control {
  *             any single ECC step, 0 if bitflips uncorrectable, -EIO hw error
  * @read_subpage:      function to read parts of the page covered by ECC;
  *                     returns same as read_page()
+ * @write_subpage:     function to write parts of the page covered by ECC.
  * @write_page:        function to write a page according to the ECC generator
  *             requirements.
  * @write_oob_raw:     function to write chip OOB data without ECC
@@ -375,6 +456,9 @@ struct nand_ecc_ctrl {
                        uint8_t *buf, int oob_required, int page);
        int (*read_subpage)(struct mtd_info *mtd, struct nand_chip *chip,
                        uint32_t offs, uint32_t len, uint8_t *buf);
+       int (*write_subpage)(struct mtd_info *mtd, struct nand_chip *chip,
+                       uint32_t offset, uint32_t data_len,
+                       const uint8_t *data_buf, int oob_required);
        int (*write_page)(struct mtd_info *mtd, struct nand_chip *chip,
                        const uint8_t *buf, int oob_required);
        int (*write_oob_raw)(struct mtd_info *mtd, struct nand_chip *chip,
@@ -396,10 +480,16 @@ struct nand_ecc_ctrl {
  * consecutive order.
  */
 struct nand_buffers {
+#ifndef __UBOOT__
+       uint8_t *ecccalc;
+       uint8_t *ecccode;
+       uint8_t *databuf;
+#else
        uint8_t ecccalc[ALIGN(NAND_MAX_OOBSIZE, ARCH_DMA_MINALIGN)];
        uint8_t ecccode[ALIGN(NAND_MAX_OOBSIZE, ARCH_DMA_MINALIGN)];
        uint8_t databuf[ALIGN(NAND_MAX_PAGESIZE + NAND_MAX_OOBSIZE,
                              ARCH_DMA_MINALIGN)];
+#endif
 };
 
 /**
@@ -410,13 +500,13 @@ struct nand_buffers {
  *                     flash device.
  * @read_byte:         [REPLACEABLE] read one byte from the chip
  * @read_word:         [REPLACEABLE] read one word from the chip
+ * @write_byte:                [REPLACEABLE] write a single byte to the chip on the
+ *                     low 8 I/O lines
  * @write_buf:         [REPLACEABLE] write data from the buffer to the chip
  * @read_buf:          [REPLACEABLE] read data from the chip into the buffer
- * @verify_buf:                [REPLACEABLE] verify buffer contents against the chip
- *                     data.
  * @select_chip:       [REPLACEABLE] select chip nr
- * @block_bad:         [REPLACEABLE] check, if the block is bad
- * @block_markbad:     [REPLACEABLE] mark the block bad
+ * @block_bad:         [REPLACEABLE] check if a block is bad, using OOB markers
+ * @block_markbad:     [REPLACEABLE] mark a block bad
  * @cmd_ctrl:          [BOARDSPECIFIC] hardwarespecific function for controlling
  *                     ALE/CLE/nCE. Also used to write command and address
  * @init_size:         [BOARDSPECIFIC] hardwarespecific function for setting
@@ -431,6 +521,8 @@ struct nand_buffers {
  *                     commands to the chip.
  * @waitfunc:          [REPLACEABLE] hardwarespecific function for wait on
  *                     ready.
+ * @setup_read_retry:  [FLASHSPECIFIC] flash (vendor) specific function for
+ *                     setting the read-retry mode. Mostly needed for MLC NAND.
  * @ecc:               [BOARDSPECIFIC] ECC control structure
  * @buffers:           buffer structure for read/write
  * @hwcontrol:         platform-specific hardware control structure
@@ -458,7 +550,13 @@ struct nand_buffers {
  * @badblockbits:      [INTERN] minimum number of set bits in a good block's
  *                     bad block marker position; i.e., BBM == 11110111b is
  *                     not bad when badblockbits == 7
- * @cellinfo:          [INTERN] MLC/multichip data from chip ident
+ * @bits_per_cell:     [INTERN] number of bits per cell. i.e., 1 means SLC.
+ * @ecc_strength_ds:   [INTERN] ECC correctability from the datasheet.
+ *                     Minimum amount of bit errors per @ecc_step_ds guaranteed
+ *                     to be correctable. If unknown, set to zero.
+ * @ecc_step_ds:       [INTERN] ECC step required by the @ecc_strength_ds,
+ *                      also from the datasheet. It is the recommended ECC step
+ *                     size, if known; if unknown, set to zero.
  * @numchips:          [INTERN] number of physical chips
  * @chipsize:          [INTERN] the size of one chip for multichip arrays
  * @pagemask:          [INTERN] page number mask = number of (pages / chip) - 1
@@ -471,9 +569,9 @@ struct nand_buffers {
  *                     non 0 if ONFI supported.
  * @onfi_params:       [INTERN] holds the ONFI page parameter when ONFI is
  *                     supported, 0 otherwise.
- * @onfi_set_features  [REPLACEABLE] set the features for ONFI nand
- * @onfi_get_features  [REPLACEABLE] get the features for ONFI nand
- * @ecclayout:         [REPLACEABLE] the default ECC placement scheme
+ * @read_retries:      [INTERN] the number of read retry modes supported
+ * @onfi_set_features: [REPLACEABLE] set the features for ONFI nand
+ * @onfi_get_features: [REPLACEABLE] get the features for ONFI nand
  * @bbt:               [INTERN] bad block table pointer
  * @bbt_td:            [REPLACEABLE] bad block table descriptor for flash
  *                     lookup.
@@ -496,9 +594,14 @@ struct nand_chip {
 
        uint8_t (*read_byte)(struct mtd_info *mtd);
        u16 (*read_word)(struct mtd_info *mtd);
+       void (*write_byte)(struct mtd_info *mtd, uint8_t byte);
        void (*write_buf)(struct mtd_info *mtd, const uint8_t *buf, int len);
        void (*read_buf)(struct mtd_info *mtd, uint8_t *buf, int len);
-       int (*verify_buf)(struct mtd_info *mtd, const uint8_t *buf, int len);
+#ifdef __UBOOT__
+#if defined(CONFIG_MTD_NAND_VERIFY_WRITE)
+        int (*verify_buf)(struct mtd_info *mtd, const uint8_t *buf, int len);
+#endif
+#endif
        void (*select_chip)(struct mtd_info *mtd, int chip);
        int (*block_bad)(struct mtd_info *mtd, loff_t ofs, int getchip);
        int (*block_markbad)(struct mtd_info *mtd, loff_t ofs);
@@ -514,12 +617,13 @@ struct nand_chip {
        int (*errstat)(struct mtd_info *mtd, struct nand_chip *this, int state,
                        int status, int page);
        int (*write_page)(struct mtd_info *mtd, struct nand_chip *chip,
-                       const uint8_t *buf, int oob_required, int page,
-                       int cached, int raw);
+                       uint32_t offset, int data_len, const uint8_t *buf,
+                       int oob_required, int page, int cached, int raw);
        int (*onfi_set_features)(struct mtd_info *mtd, struct nand_chip *chip,
                        int feature_addr, uint8_t *subfeature_para);
        int (*onfi_get_features)(struct mtd_info *mtd, struct nand_chip *chip,
                        int feature_addr, uint8_t *subfeature_para);
+       int (*setup_read_retry)(struct mtd_info *mtd, int retry_mode);
 
        int chip_delay;
        unsigned int options;
@@ -535,20 +639,26 @@ struct nand_chip {
        int pagebuf;
        unsigned int pagebuf_bitflips;
        int subpagesize;
-       uint8_t cellinfo;
+       uint8_t bits_per_cell;
+       uint16_t ecc_strength_ds;
+       uint16_t ecc_step_ds;
        int badblockpos;
        int badblockbits;
 
        int onfi_version;
 #ifdef CONFIG_SYS_NAND_ONFI_DETECTION
-       struct nand_onfi_params onfi_params;
+       struct nand_onfi_params onfi_params;
 #endif
 
-       int state;
+       int read_retries;
+
+       flstate_t state;
 
        uint8_t *oob_poi;
        struct nand_hw_control *controller;
+#ifdef __UBOOT__
        struct nand_ecclayout *ecclayout;
+#endif
 
        struct nand_ecc_ctrl ecc;
        struct nand_buffers *buffers;
@@ -577,26 +687,83 @@ struct nand_chip {
 #define NAND_MFR_AMD           0x01
 #define NAND_MFR_MACRONIX      0xc2
 #define NAND_MFR_EON           0x92
+#define NAND_MFR_SANDISK       0x45
+#define NAND_MFR_INTEL         0x89
+
+/* The maximum expected count of bytes in the NAND ID sequence */
+#define NAND_MAX_ID_LEN 8
+
+/*
+ * A helper for defining older NAND chips where the second ID byte fully
+ * defined the chip, including the geometry (chip size, eraseblock size, page
+ * size). All these chips have 512 bytes NAND page size.
+ */
+#define LEGACY_ID_NAND(nm, devid, chipsz, erasesz, opts)          \
+       { .name = (nm), {{ .dev_id = (devid) }}, .pagesize = 512, \
+         .chipsize = (chipsz), .erasesize = (erasesz), .options = (opts) }
+
+/*
+ * A helper for defining newer chips which report their page size and
+ * eraseblock size via the extended ID bytes.
+ *
+ * The real difference between LEGACY_ID_NAND and EXTENDED_ID_NAND is that with
+ * EXTENDED_ID_NAND, manufacturers overloaded the same device ID so that the
+ * device ID now only represented a particular total chip size (and voltage,
+ * buswidth), and the page size, eraseblock size, and OOB size could vary while
+ * using the same device ID.
+ */
+#define EXTENDED_ID_NAND(nm, devid, chipsz, opts)                      \
+       { .name = (nm), {{ .dev_id = (devid) }}, .chipsize = (chipsz), \
+         .options = (opts) }
+
+#define NAND_ECC_INFO(_strength, _step)        \
+                       { .strength_ds = (_strength), .step_ds = (_step) }
+#define NAND_ECC_STRENGTH(type)                ((type)->ecc.strength_ds)
+#define NAND_ECC_STEP(type)            ((type)->ecc.step_ds)
 
 /**
  * struct nand_flash_dev - NAND Flash Device ID Structure
- * @name:      Identify the device type
- * @id:                device ID code
- * @pagesize:  Pagesize in bytes. Either 256 or 512 or 0
- *             If the pagesize is 0, then the real pagesize
- *             and the eraseize are determined from the
- *             extended id bytes in the chip
- * @erasesize: Size of an erase block in the flash device.
- * @chipsize:  Total chipsize in Mega Bytes
- * @options:   Bitfield to store chip relevant options
+ * @name: a human-readable name of the NAND chip
+ * @dev_id: the device ID (the second byte of the full chip ID array)
+ * @mfr_id: manufecturer ID part of the full chip ID array (refers the same
+ *          memory address as @id[0])
+ * @dev_id: device ID part of the full chip ID array (refers the same memory
+ *          address as @id[1])
+ * @id: full device ID array
+ * @pagesize: size of the NAND page in bytes; if 0, then the real page size (as
+ *            well as the eraseblock size) is determined from the extended NAND
+ *            chip ID array)
+ * @chipsize: total chip size in MiB
+ * @erasesize: eraseblock size in bytes (determined from the extended ID if 0)
+ * @options: stores various chip bit options
+ * @id_len: The valid length of the @id.
+ * @oobsize: OOB size
+ * @ecc.strength_ds: The ECC correctability from the datasheet, same as the
+ *                   @ecc_strength_ds in nand_chip{}.
+ * @ecc.step_ds: The ECC step required by the @ecc.strength_ds, same as the
+ *               @ecc_step_ds in nand_chip{}, also from the datasheet.
+ *               For example, the "4bit ECC for each 512Byte" can be set with
+ *               NAND_ECC_INFO(4, 512).
  */
 struct nand_flash_dev {
        char *name;
-       int id;
-       unsigned long pagesize;
-       unsigned long chipsize;
-       unsigned long erasesize;
-       unsigned long options;
+       union {
+               struct {
+                       uint8_t mfr_id;
+                       uint8_t dev_id;
+               };
+               uint8_t id[NAND_MAX_ID_LEN];
+       };
+       unsigned int pagesize;
+       unsigned int chipsize;
+       unsigned int erasesize;
+       unsigned int options;
+       uint16_t id_len;
+       uint16_t oobsize;
+       struct {
+               uint16_t strength_ds;
+               uint16_t step_ds;
+       } ecc;
 };
 
 /**
@@ -609,23 +776,25 @@ struct nand_manufacturers {
        char *name;
 };
 
-extern const struct nand_flash_dev nand_flash_ids[];
-extern const struct nand_manufacturers nand_manuf_ids[];
+extern struct nand_flash_dev nand_flash_ids[];
+extern struct nand_manufacturers nand_manuf_ids[];
 
 extern int nand_scan_bbt(struct mtd_info *mtd, struct nand_bbt_descr *bd);
-extern int nand_update_bbt(struct mtd_info *mtd, loff_t offs);
 extern int nand_default_bbt(struct mtd_info *mtd);
+extern int nand_markbad_bbt(struct mtd_info *mtd, loff_t offs);
 extern int nand_isbad_bbt(struct mtd_info *mtd, loff_t offs, int allowbbt);
 extern int nand_erase_nand(struct mtd_info *mtd, struct erase_info *instr,
                           int allowbbt);
 extern int nand_do_read(struct mtd_info *mtd, loff_t from, size_t len,
                        size_t *retlen, uint8_t *buf);
 
+#ifdef __UBOOT__
 /*
 * Constants for oob configuration
 */
 #define NAND_SMALL_BADBLOCK_POS                5
 #define NAND_LARGE_BADBLOCK_POS                0
+#endif
 
 /**
  * struct platform_nand_chip - chip level device structure
@@ -656,20 +825,29 @@ struct platform_device;
 
 /**
  * struct platform_nand_ctrl - controller level device structure
+ * @probe:             platform specific function to probe/setup hardware
+ * @remove:            platform specific function to remove/teardown hardware
  * @hwcontrol:         platform specific hardware control structure
  * @dev_ready:         platform specific function to read ready/busy pin
  * @select_chip:       platform specific chip select function
  * @cmd_ctrl:          platform specific function for controlling
  *                     ALE/CLE/nCE. Also used to write command and address
+ * @write_buf:         platform specific function for write buffer
+ * @read_buf:          platform specific function for read buffer
+ * @read_byte:         platform specific function to read one byte from chip
  * @priv:              private data to transport driver specific settings
  *
  * All fields are optional and depend on the hardware driver requirements
  */
 struct platform_nand_ctrl {
+       int (*probe)(struct platform_device *pdev);
+       void (*remove)(struct platform_device *pdev);
        void (*hwcontrol)(struct mtd_info *mtd, int cmd);
        int (*dev_ready)(struct mtd_info *mtd);
        void (*select_chip)(struct mtd_info *mtd, int chip);
        void (*cmd_ctrl)(struct mtd_info *mtd, int dat, unsigned int ctrl);
+       void (*write_buf)(struct mtd_info *mtd, const uint8_t *buf, int len);
+       void (*read_buf)(struct mtd_info *mtd, uint8_t *buf, int len);
        unsigned char (*read_byte)(struct mtd_info *mtd);
        void *priv;
 };
@@ -693,16 +871,14 @@ struct platform_nand_chip *get_platform_nandchip(struct mtd_info *mtd)
        return chip->priv;
 }
 
-/* Standard NAND functions from nand_base.c */
-void nand_write_buf(struct mtd_info *mtd, const uint8_t *buf, int len);
-void nand_write_buf16(struct mtd_info *mtd, const uint8_t *buf, int len);
-void nand_read_buf(struct mtd_info *mtd, uint8_t *buf, int len);
-void nand_read_buf16(struct mtd_info *mtd, uint8_t *buf, int len);
-uint8_t nand_read_byte(struct mtd_info *mtd);
+#ifdef CONFIG_SYS_NAND_ONFI_DETECTION
+/* return the supported features. */
+static inline int onfi_feature(struct nand_chip *chip)
+{
+       return chip->onfi_version ? le16_to_cpu(chip->onfi_params.features) : 0;
+}
 
 /* return the supported asynchronous timing mode. */
-
-#ifdef CONFIG_SYS_NAND_ONFI_DETECTION
 static inline int onfi_get_async_timing_mode(struct nand_chip *chip)
 {
        if (!chip->onfi_version)
@@ -719,6 +895,16 @@ static inline int onfi_get_sync_timing_mode(struct nand_chip *chip)
 }
 #endif
 
+/*
+ * Check if it is a SLC nand.
+ * The !nand_is_slc() can be used to check the MLC/TLC nand chips.
+ * We do not distinguish the MLC and TLC now.
+ */
+static inline bool nand_is_slc(struct nand_chip *chip)
+{
+       return chip->bits_per_cell == 1;
+}
+
 /**
  * Check if the opcode's address should be sent only on the lower 8 bits
  * @command: opcode to check
@@ -737,5 +923,12 @@ static inline int nand_opcode_8bits(unsigned int command)
        return 0;
 }
 
-
+#ifdef __UBOOT__
+/* Standard NAND functions from nand_base.c */
+void nand_write_buf(struct mtd_info *mtd, const uint8_t *buf, int len);
+void nand_write_buf16(struct mtd_info *mtd, const uint8_t *buf, int len);
+void nand_read_buf(struct mtd_info *mtd, uint8_t *buf, int len);
+void nand_read_buf16(struct mtd_info *mtd, uint8_t *buf, int len);
+uint8_t nand_read_byte(struct mtd_info *mtd);
+#endif
 #endif /* __LINUX_MTD_NAND_H */
index d1d9a96d583f89fccfb7f1cfc5c2c55369467385..ce0e8dbee4eb30e3585ab8722e9e5e070ea7532d 100644 (file)
@@ -1,11 +1,9 @@
 /*
  * MTD partitioning layer definitions
  *
- * (C) 2000 Nicolas Pitre <nico@cam.org>
+ * (C) 2000 Nicolas Pitre <nico@fluxnic.net>
  *
  * This code is GPL
- *
- * $Id: partitions.h,v 1.17 2005/11/07 11:14:55 gleixner Exp $
  */
 
 #ifndef MTD_PARTITIONS_H
@@ -18,7 +16,7 @@
  * Partition definition structure:
  *
  * An array of struct partition is passed along with a MTD object to
- * add_mtd_partitions() to create them.
+ * mtd_device_register() to create them.
  *
  * For each partition, these fields are available:
  * name: string that will be used to label the partition's MTD device.
@@ -26,7 +24,9 @@
  *     will extend to the end of the master MTD device.
  * offset: absolute starting position within the master MTD device; if
  *     defined as MTDPART_OFS_APPEND, the partition will start where the
- *     previous one ended; if MTDPART_OFS_NXTBLK, at the next erase block.
+ *     previous one ended; if MTDPART_OFS_NXTBLK, at the next erase block;
+ *     if MTDPART_OFS_RETAIN, consume as much as possible, leaving size
+ *     after the end of partition.
  * mask_flags: contains flags that have to be masked (removed) from the
  *     master MTD flag set for the corresponding MTD partition.
  *     For example, to force a read-only partition, simply adding
  */
 
 struct mtd_partition {
-       char *name;                     /* identifier string */
+       const char *name;               /* identifier string */
        uint64_t size;                  /* partition size */
        uint64_t offset;                /* offset within the master MTD space */
-       u_int32_t mask_flags;           /* master MTD flags to mask out for this partition */
-       struct nand_ecclayout *ecclayout;       /* out of band layout for this partition (NAND only)*/
-       struct mtd_info **mtdp;         /* pointer to store the MTD object */
+       uint32_t mask_flags;            /* master MTD flags to mask out for this partition */
+       struct nand_ecclayout *ecclayout;       /* out of band layout for this partition (NAND only) */
 };
 
+#define MTDPART_OFS_RETAIN     (-3)
 #define MTDPART_OFS_NXTBLK     (-2)
 #define MTDPART_OFS_APPEND     (-1)
 #define MTDPART_SIZ_FULL       (0)
 
 
-int add_mtd_partitions(struct mtd_info *, const struct mtd_partition *, int);
-int del_mtd_partitions(struct mtd_info *);
+struct mtd_info;
+struct device_node;
+
+#ifndef __UBOOT__
+/**
+ * struct mtd_part_parser_data - used to pass data to MTD partition parsers.
+ * @origin: for RedBoot, start address of MTD device
+ * @of_node: for OF parsers, device node containing partitioning information
+ */
+struct mtd_part_parser_data {
+       unsigned long origin;
+       struct device_node *of_node;
+};
+
 
-#if 0
 /*
  * Functions dealing with the various ways of partitioning the space
  */
@@ -62,23 +73,18 @@ struct mtd_part_parser {
        struct list_head list;
        struct module *owner;
        const char *name;
-       int (*parse_fn)(struct mtd_info *, struct mtd_partition **, unsigned long);
+       int (*parse_fn)(struct mtd_info *, struct mtd_partition **,
+                       struct mtd_part_parser_data *);
 };
 
-extern int register_mtd_parser(struct mtd_part_parser *parser);
-extern int deregister_mtd_parser(struct mtd_part_parser *parser);
-extern int parse_mtd_partitions(struct mtd_info *master, const char **types,
-                               struct mtd_partition **pparts, unsigned long origin);
-
-#define put_partition_parser(p) do { module_put((p)->owner); } while(0)
-
-struct device;
-struct device_node;
-
-int __devinit of_mtd_parse_partitions(struct device *dev,
-                                     struct mtd_info *mtd,
-                                     struct device_node *node,
-                                     struct mtd_partition **pparts);
+extern void register_mtd_parser(struct mtd_part_parser *parser);
+extern void deregister_mtd_parser(struct mtd_part_parser *parser);
 #endif
 
+int mtd_is_partition(const struct mtd_info *mtd);
+int mtd_add_partition(struct mtd_info *master, const char *name,
+                     long long offset, long long length);
+int mtd_del_partition(struct mtd_info *master, int partno);
+uint64_t mtd_get_device_size(const struct mtd_info *mtd);
+
 #endif
index 4755770c547b44d660fadff3fc3af026fcb3333d..d9e58aedf6398487b705eed3ea653b80dd4b1b43 100644 (file)
@@ -9,9 +9,15 @@
 #ifndef __LINUX_UBI_H__
 #define __LINUX_UBI_H__
 
-/* #include <asm/ioctl.h> */
 #include <linux/types.h>
+#define __UBOOT__
+#ifndef __UBOOT__
+#include <linux/ioctl.h>
 #include <mtd/ubi-user.h>
+#endif
+
+/* All voumes/LEBs */
+#define UBI_ALL -1
 
 /*
  * enum ubi_open_mode - UBI volume open mode constants.
@@ -33,13 +39,13 @@ enum {
  * @size: how many physical eraseblocks are reserved for this volume
  * @used_bytes: how many bytes of data this volume contains
  * @used_ebs: how many physical eraseblocks of this volume actually contain any
- * data
+ *            data
  * @vol_type: volume type (%UBI_DYNAMIC_VOLUME or %UBI_STATIC_VOLUME)
  * @corrupted: non-zero if the volume is corrupted (static volumes only)
  * @upd_marker: non-zero if the volume has update marker set
  * @alignment: volume alignment
  * @usable_leb_size: how many bytes are available in logical eraseblocks of
- * this volume
+ *                   this volume
  * @name_len: volume name length
  * @name: volume name
  * @cdev: UBI volume character device major and minor numbers
@@ -75,7 +81,7 @@ enum {
  * physical eraseblock size and on how much bytes UBI headers consume. But
  * because of the volume alignment (@alignment), the usable size of logical
  * eraseblocks if a volume may be less. The following equation is true:
- *     @usable_leb_size = LEB size - (LEB size mod @alignment),
+ *     @usable_leb_size = LEB size - (LEB size mod @alignment),
  * where LEB size is the logical eraseblock size defined by the UBI device.
  *
  * The alignment is multiple to the minimal flash input/output unit size or %1
@@ -104,20 +110,79 @@ struct ubi_volume_info {
  * struct ubi_device_info - UBI device description data structure.
  * @ubi_num: ubi device number
  * @leb_size: logical eraseblock size on this UBI device
+ * @leb_start: starting offset of logical eraseblocks within physical
+ *             eraseblocks
  * @min_io_size: minimal I/O unit size
+ * @max_write_size: maximum amount of bytes the underlying flash can write at a
+ *                  time (MTD write buffer size)
  * @ro_mode: if this device is in read-only mode
  * @cdev: UBI character device major and minor numbers
  *
  * Note, @leb_size is the logical eraseblock size offered by the UBI device.
  * Volumes of this UBI device may have smaller logical eraseblock size if their
  * alignment is not equivalent to %1.
+ *
+ * The @max_write_size field describes flash write maximum write unit. For
+ * example, NOR flash allows for changing individual bytes, so @min_io_size is
+ * %1. However, it does not mean than NOR flash has to write data byte-by-byte.
+ * Instead, CFI NOR flashes have a write-buffer of, e.g., 64 bytes, and when
+ * writing large chunks of data, they write 64-bytes at a time. Obviously, this
+ * improves write throughput.
+ *
+ * Also, the MTD device may have N interleaved (striped) flash chips
+ * underneath, in which case @min_io_size can be physical min. I/O size of
+ * single flash chip, while @max_write_size can be N * @min_io_size.
+ *
+ * The @max_write_size field is always greater or equivalent to @min_io_size.
+ * E.g., some NOR flashes may have (@min_io_size = 1, @max_write_size = 64). In
+ * contrast, NAND flashes usually have @min_io_size = @max_write_size = NAND
+ * page size.
  */
 struct ubi_device_info {
        int ubi_num;
        int leb_size;
+       int leb_start;
        int min_io_size;
+       int max_write_size;
        int ro_mode;
+#ifndef __UBOOT__
        dev_t cdev;
+#endif
+};
+
+/*
+ * Volume notification types.
+ * @UBI_VOLUME_ADDED: a volume has been added (an UBI device was attached or a
+ *                    volume was created)
+ * @UBI_VOLUME_REMOVED: a volume has been removed (an UBI device was detached
+ *                     or a volume was removed)
+ * @UBI_VOLUME_RESIZED: a volume has been re-sized
+ * @UBI_VOLUME_RENAMED: a volume has been re-named
+ * @UBI_VOLUME_UPDATED: data has been written to a volume
+ *
+ * These constants define which type of event has happened when a volume
+ * notification function is invoked.
+ */
+enum {
+       UBI_VOLUME_ADDED,
+       UBI_VOLUME_REMOVED,
+       UBI_VOLUME_RESIZED,
+       UBI_VOLUME_RENAMED,
+       UBI_VOLUME_UPDATED,
+};
+
+/*
+ * struct ubi_notification - UBI notification description structure.
+ * @di: UBI device description object
+ * @vi: UBI volume description object
+ *
+ * UBI notifiers are called with a pointer to an object of this type. The
+ * object describes the notification. Namely, it provides a description of the
+ * UBI device and UBI volume the notification informs about.
+ */
+struct ubi_notification {
+       struct ubi_device_info di;
+       struct ubi_volume_info vi;
 };
 
 /* UBI descriptor given to users when they open UBI volumes */
@@ -129,17 +194,37 @@ void ubi_get_volume_info(struct ubi_volume_desc *desc,
 struct ubi_volume_desc *ubi_open_volume(int ubi_num, int vol_id, int mode);
 struct ubi_volume_desc *ubi_open_volume_nm(int ubi_num, const char *name,
                                           int mode);
+struct ubi_volume_desc *ubi_open_volume_path(const char *pathname, int mode);
+
+#ifndef __UBOOT__
+typedef        int (*notifier_fn_t)(void *nb,
+                       unsigned long action, void *data);
+
+struct notifier_block {
+       notifier_fn_t notifier_call;
+       struct notifier_block *next;
+       void *next;
+       int priority;
+};
+
+int ubi_register_volume_notifier(struct notifier_block *nb,
+                                int ignore_existing);
+int ubi_unregister_volume_notifier(struct notifier_block *nb);
+#endif
+
 void ubi_close_volume(struct ubi_volume_desc *desc);
 int ubi_leb_read(struct ubi_volume_desc *desc, int lnum, char *buf, int offset,
                 int len, int check);
 int ubi_leb_write(struct ubi_volume_desc *desc, int lnum, const void *buf,
-                 int offset, int len, int dtype);
+                 int offset, int len);
 int ubi_leb_change(struct ubi_volume_desc *desc, int lnum, const void *buf,
-                  int len, int dtype);
+                  int len);
 int ubi_leb_erase(struct ubi_volume_desc *desc, int lnum);
 int ubi_leb_unmap(struct ubi_volume_desc *desc, int lnum);
-int ubi_leb_map(struct ubi_volume_desc *desc, int lnum, int dtype);
+int ubi_leb_map(struct ubi_volume_desc *desc, int lnum);
 int ubi_is_mapped(struct ubi_volume_desc *desc, int lnum);
+int ubi_sync(int ubi_num);
+int ubi_flush(int ubi_num, int vol_id, int lnum);
 
 /*
  * This function is the same as the 'ubi_leb_read()' function, but it does not
@@ -150,25 +235,4 @@ static inline int ubi_read(struct ubi_volume_desc *desc, int lnum, char *buf,
 {
        return ubi_leb_read(desc, lnum, buf, offset, len, 0);
 }
-
-/*
- * This function is the same as the 'ubi_leb_write()' functions, but it does
- * not have the data type argument.
- */
-static inline int ubi_write(struct ubi_volume_desc *desc, int lnum,
-                           const void *buf, int offset, int len)
-{
-       return ubi_leb_write(desc, lnum, buf, offset, len, UBI_UNKNOWN);
-}
-
-/*
- * This function is the same as the 'ubi_leb_change()' functions, but it does
- * not have the data type argument.
- */
-static inline int ubi_change(struct ubi_volume_desc *desc, int lnum,
-                                   const void *buf, int len)
-{
-       return ubi_leb_change(desc, lnum, buf, len, UBI_UNKNOWN);
-}
-
 #endif /* !__LINUX_UBI_H__ */
index ac3c29876048de8547e2b0fd514fe54085bcb2d3..b9f4bcb154bf5d073ce80286eb5b315cdf191760 100644 (file)
@@ -1,30 +1,44 @@
 /*
- * $Id: mtd-abi.h,v 1.13 2005/11/07 11:14:56 gleixner Exp $
+ * Copyright © 1999-2010 David Woodhouse <dwmw2@infradead.org> et al.
+ *
+ * SPDX-License-Identifier:    GPL-2.0+
  *
- * Portions of MTD ABI definition which are shared by kernel and user space
  */
 
 #ifndef __MTD_ABI_H__
 #define __MTD_ABI_H__
 
-#if 1
+#define __UBOOT__
+#ifdef __UBOOT__
 #include <linux/compat.h>
 #endif
 
 #include <linux/compiler.h>
 
 struct erase_info_user {
-       uint32_t start;
-       uint32_t length;
+       __u32 start;
+       __u32 length;
+};
+
+struct erase_info_user64 {
+       __u64 start;
+       __u64 length;
 };
 
 struct mtd_oob_buf {
-       uint32_t start;
-       uint32_t length;
+       __u32 start;
+       __u32 length;
        unsigned char __user *ptr;
 };
 
-/*
+struct mtd_oob_buf64 {
+       __u64 start;
+       __u32 pad;
+       __u32 length;
+       __u64 usr_ptr;
+};
+
+/**
  * MTD operation modes
  *
  * @MTD_OPS_PLACE_OOB: OOB data are placed at the given offset (default)
@@ -43,18 +57,45 @@ enum {
        MTD_OPS_RAW = 2,
 };
 
+/**
+ * struct mtd_write_req - data structure for requesting a write operation
+ *
+ * @start:     start address
+ * @len:       length of data buffer
+ * @ooblen:    length of OOB buffer
+ * @usr_data:  user-provided data buffer
+ * @usr_oob:   user-provided OOB buffer
+ * @mode:      MTD mode (see "MTD operation modes")
+ * @padding:   reserved, must be set to 0
+ *
+ * This structure supports ioctl(MEMWRITE) operations, allowing data and/or OOB
+ * writes in various modes. To write to OOB-only, set @usr_data == NULL, and to
+ * write data-only, set @usr_oob == NULL. However, setting both @usr_data and
+ * @usr_oob to NULL is not allowed.
+ */
+struct mtd_write_req {
+       __u64 start;
+       __u64 len;
+       __u64 ooblen;
+       __u64 usr_data;
+       __u64 usr_oob;
+       __u8 mode;
+       __u8 padding[7];
+};
+
 #define MTD_ABSENT             0
 #define MTD_RAM                        1
 #define MTD_ROM                        2
 #define MTD_NORFLASH           3
-#define MTD_NANDFLASH          4
+#define MTD_NANDFLASH          4       /* SLC NAND */
 #define MTD_DATAFLASH          6
 #define MTD_UBIVOLUME          7
+#define MTD_MLCNANDFLASH       8       /* MLC NAND (including TLC) */
 
 #define MTD_WRITEABLE          0x400   /* Device is writeable */
 #define MTD_BIT_WRITEABLE      0x800   /* Single bits can be flipped */
 #define MTD_NO_ERASE           0x1000  /* No erase necessary */
-#define MTD_STUPID_LOCK                0x2000  /* Always locked after reset */
+#define MTD_POWERUP_LOCK       0x2000  /* Always locked after reset */
 
 /* Some common devices / combinations of capabilities */
 #define MTD_CAP_ROM            0
@@ -62,12 +103,12 @@ enum {
 #define MTD_CAP_NORFLASH       (MTD_WRITEABLE | MTD_BIT_WRITEABLE)
 #define MTD_CAP_NANDFLASH      (MTD_WRITEABLE)
 
-/* ECC byte placement */
-#define MTD_NANDECC_OFF                0       /* Switch off ECC (Not recommended) */
-#define MTD_NANDECC_PLACE      1       /* Use the given placement in the structure (YAFFS1 legacy mode) */
-#define MTD_NANDECC_AUTOPLACE  2       /* Use the default placement scheme */
-#define MTD_NANDECC_PLACEONLY  3       /* Use the given placement in the structure (Do not store ecc result on read) */
-#define MTD_NANDECC_AUTOPL_USR 4       /* Use the given autoplacement scheme rather than using the default */
+/* Obsolete ECC byte placement modes (used with obsolete MEMGETOOBSEL) */
+#define MTD_NANDECC_OFF                0       // Switch off ECC (Not recommended)
+#define MTD_NANDECC_PLACE      1       // Use the given placement in the structure (YAFFS1 legacy mode)
+#define MTD_NANDECC_AUTOPLACE  2       // Use the default placement scheme
+#define MTD_NANDECC_PLACEONLY  3       // Use the given placement in the structure (Do not store ecc result on read)
+#define MTD_NANDECC_AUTOPL_USR         4       // Use the given autoplacement scheme rather than using the default
 
 /* OTP mode selection */
 #define MTD_OTP_OFF            0
@@ -75,32 +116,35 @@ enum {
 #define MTD_OTP_USER           2
 
 struct mtd_info_user {
-       uint8_t type;
-       uint32_t flags;
-       uint32_t size;                  /* Total size of the MTD */
-       uint32_t erasesize;
-       uint32_t writesize;
-       uint32_t oobsize;               /* Amount of OOB data per block (e.g. 16) */
-       /* The below two fields are obsolete and broken, do not use them
-        * (TODO: remove at some point) */
-       uint32_t ecctype;
-       uint32_t eccsize;
+       __u8 type;
+       __u32 flags;
+       __u32 size;     /* Total size of the MTD */
+       __u32 erasesize;
+       __u32 writesize;
+       __u32 oobsize;  /* Amount of OOB data per block (e.g. 16) */
+       __u64 padding;  /* Old obsolete field; do not use */
 };
 
 struct region_info_user {
-       uint32_t offset;                /* At which this region starts,
-                                        * from the beginning of the MTD */
-       uint32_t erasesize;             /* For this region */
-       uint32_t numblocks;             /* Number of blocks in this region */
-       uint32_t regionindex;
+       __u32 offset;           /* At which this region starts,
+                                * from the beginning of the MTD */
+       __u32 erasesize;        /* For this region */
+       __u32 numblocks;        /* Number of blocks in this region */
+       __u32 regionindex;
 };
 
 struct otp_info {
-       uint32_t start;
-       uint32_t length;
-       uint32_t locked;
+       __u32 start;
+       __u32 length;
+       __u32 locked;
 };
 
+/*
+ * Note, the following ioctl existed in the past and was removed:
+ * #define MEMSETOOBSEL           _IOW('M', 9, struct nand_oobinfo)
+ * Try to avoid adding a new ioctl with the same ioctl number.
+ */
+
 /* Get basic MTD characteristics info (better to use sysfs) */
 #define MEMGETINFO             _IOR('M', 1, struct mtd_info_user)
 /* Erase segment of MTD */
@@ -118,12 +162,11 @@ struct otp_info {
 /* Get information about the erase region for a specific index */
 #define MEMGETREGIONINFO       _IOWR('M', 8, struct region_info_user)
 /* Get info about OOB modes (e.g., RAW, PLACE, AUTO) - legacy interface */
-#define MEMSETOOBSEL           _IOW('M', 9, struct nand_oobinfo)
 #define MEMGETOOBSEL           _IOR('M', 10, struct nand_oobinfo)
 /* Check if an eraseblock is bad */
-#define MEMGETBADBLOCK         _IOW('M', 11, loff_t)
+#define MEMGETBADBLOCK         _IOW('M', 11, __kernel_loff_t)
 /* Mark an eraseblock as bad */
-#define MEMSETBADBLOCK         _IOW('M', 12, loff_t)
+#define MEMSETBADBLOCK         _IOW('M', 12, __kernel_loff_t)
 /* Set OTP (One-Time Programmable) mode (factory vs. user) */
 #define OTPSELECT              _IOR('M', 13, int)
 /* Get number of OTP (One-Time Programmable) regions */
@@ -133,26 +176,57 @@ struct otp_info {
 /* Lock a given range of user data (must be in mode %MTD_FILE_MODE_OTP_USER) */
 #define OTPLOCK                        _IOR('M', 16, struct otp_info)
 /* Get ECC layout (deprecated) */
-#define ECCGETLAYOUT           _IOR('M', 17, struct nand_ecclayout)
+#define ECCGETLAYOUT           _IOR('M', 17, struct nand_ecclayout_user)
 /* Get statistics about corrected/uncorrected errors */
 #define ECCGETSTATS            _IOR('M', 18, struct mtd_ecc_stats)
 /* Set MTD mode on a per-file-descriptor basis (see "MTD file modes") */
 #define MTDFILEMODE            _IO('M', 19)
+/* Erase segment of MTD (supports 64-bit address) */
+#define MEMERASE64             _IOW('M', 20, struct erase_info_user64)
+/* Write data to OOB (64-bit version) */
+#define MEMWRITEOOB64          _IOWR('M', 21, struct mtd_oob_buf64)
+/* Read data from OOB (64-bit version) */
+#define MEMREADOOB64           _IOWR('M', 22, struct mtd_oob_buf64)
+/* Check if chip is locked (for MTD that supports it) */
+#define MEMISLOCKED            _IOR('M', 23, struct erase_info_user)
+/*
+ * Most generic write interface; can write in-band and/or out-of-band in various
+ * modes (see "struct mtd_write_req"). This ioctl is not supported for flashes
+ * without OOB, e.g., NOR flash.
+ */
+#define MEMWRITE               _IOWR('M', 24, struct mtd_write_req)
 
 /*
  * Obsolete legacy interface. Keep it in order not to break userspace
  * interfaces
  */
 struct nand_oobinfo {
-       uint32_t useecc;
-       uint32_t eccbytes;
-       uint32_t oobfree[8][2];
-       uint32_t eccpos[48];
+       __u32 useecc;
+       __u32 eccbytes;
+       __u32 oobfree[8][2];
+       __u32 eccpos[32];
 };
 
 struct nand_oobfree {
-       uint32_t offset;
-       uint32_t length;
+       __u32 offset;
+       __u32 length;
+};
+
+#define MTD_MAX_OOBFREE_ENTRIES        8
+#define MTD_MAX_ECCPOS_ENTRIES 64
+/*
+ * OBSOLETE: ECC layout control structure. Exported to user-space via ioctl
+ * ECCGETLAYOUT for backwards compatbility and should not be mistaken as a
+ * complete set of ECC information. The ioctl truncates the larger internal
+ * structure to retain binary compatibility with the static declaration of the
+ * ioctl. Note that the "MTD_MAX_..._ENTRIES" macros represent the max size of
+ * the user struct, not the MAX size of the internal struct nand_ecclayout.
+ */
+struct nand_ecclayout_user {
+       __u32 eccbytes;
+       __u32 eccpos[MTD_MAX_ECCPOS_ENTRIES];
+       __u32 oobavail;
+       struct nand_oobfree oobfree[MTD_MAX_OOBFREE_ENTRIES];
 };
 
 /**
@@ -164,10 +238,10 @@ struct nand_oobfree {
  * @bbtblocks: number of blocks reserved for bad block tables
  */
 struct mtd_ecc_stats {
-       uint32_t corrected;
-       uint32_t failed;
-       uint32_t badblocks;
-       uint32_t bbtblocks;
+       __u32 corrected;
+       __u32 failed;
+       __u32 badblocks;
+       __u32 bbtblocks;
 };
 
 /*
@@ -188,10 +262,15 @@ struct mtd_ecc_stats {
  * used out of necessity (e.g., `write()', ioctl(MEMWRITEOOB64)).
  */
 enum mtd_file_modes {
-       MTD_MODE_NORMAL = MTD_OTP_OFF,
-       MTD_MODE_OTP_FACTORY = MTD_OTP_FACTORY,
-       MTD_MODE_OTP_USER = MTD_OTP_USER,
-       MTD_MODE_RAW,
+       MTD_FILE_MODE_NORMAL = MTD_OTP_OFF,
+       MTD_FILE_MODE_OTP_FACTORY = MTD_OTP_FACTORY,
+       MTD_FILE_MODE_OTP_USER = MTD_OTP_USER,
+       MTD_FILE_MODE_RAW,
 };
 
+static inline int mtd_type_is_nand_user(const struct mtd_info_user *mtd)
+{
+       return mtd->type == MTD_NANDFLASH || mtd->type == MTD_MLCNANDFLASH;
+}
+
 #endif /* __MTD_ABI_H__ */
index 1ccc06ea68f727b77c9e7c470f3b8858e39f4ccc..c93914ad25ca55ce5d0764889dff7517d9877a68 100644 (file)
@@ -1,7 +1,7 @@
 /*
- * Copyright (c) International Business Machines Corp., 2006
+ * Copyright © International Business Machines Corp., 2006
  *
- * SPDX-License-Identifier:    GPL-2.0+
+ * SPDX-License-Identifier:    GPL-2.0+
  *
  * Author: Artem Bityutskiy (Битюцкий Артём)
  */
@@ -9,6 +9,8 @@
 #ifndef __UBI_USER_H__
 #define __UBI_USER_H__
 
+#include <linux/types.h>
+
 /*
  * UBI device creation (the same as MTD device attachment)
  * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
  * UBI volume creation
  * ~~~~~~~~~~~~~~~~~~~
  *
- * UBI volumes are created via the %UBI_IOCMKVOL IOCTL command of UBI character
+ * UBI volumes are created via the %UBI_IOCMKVOL ioctl command of UBI character
  * device. A &struct ubi_mkvol_req object has to be properly filled and a
- * pointer to it has to be passed to the IOCTL.
+ * pointer to it has to be passed to the ioctl.
  *
  * UBI volume deletion
  * ~~~~~~~~~~~~~~~~~~~
  *
- * To delete a volume, the %UBI_IOCRMVOL IOCTL command of the UBI character
+ * To delete a volume, the %UBI_IOCRMVOL ioctl command of the UBI character
  * device should be used. A pointer to the 32-bit volume ID hast to be passed
- * to the IOCTL.
+ * to the ioctl.
  *
  * UBI volume re-size
  * ~~~~~~~~~~~~~~~~~~
  *
- * To re-size a volume, the %UBI_IOCRSVOL IOCTL command of the UBI character
+ * To re-size a volume, the %UBI_IOCRSVOL ioctl command of the UBI character
  * device should be used. A &struct ubi_rsvol_req object has to be properly
- * filled and a pointer to it has to be passed to the IOCTL.
+ * filled and a pointer to it has to be passed to the ioctl.
+ *
+ * UBI volumes re-name
+ * ~~~~~~~~~~~~~~~~~~~
+ *
+ * To re-name several volumes atomically at one go, the %UBI_IOCRNVOL command
+ * of the UBI character device should be used. A &struct ubi_rnvol_req object
+ * has to be properly filled and a pointer to it has to be passed to the ioctl.
  *
  * UBI volume update
  * ~~~~~~~~~~~~~~~~~
  *
- * Volume update should be done via the %UBI_IOCVOLUP IOCTL command of the
+ * Volume update should be done via the %UBI_IOCVOLUP ioctl command of the
  * corresponding UBI volume character device. A pointer to a 64-bit update
- * size should be passed to the IOCTL. After this, UBI expects user to write
+ * size should be passed to the ioctl. After this, UBI expects user to write
  * this number of bytes to the volume character device. The update is finished
  * when the claimed number of bytes is passed. So, the volume update sequence
  * is something like:
  * write(fd, buf, image_size);
  * close(fd);
  *
- * Atomic eraseblock change
+ * Logical eraseblock erase
+ * ~~~~~~~~~~~~~~~~~~~~~~~~
+ *
+ * To erase a logical eraseblock, the %UBI_IOCEBER ioctl command of the
+ * corresponding UBI volume character device should be used. This command
+ * unmaps the requested logical eraseblock, makes sure the corresponding
+ * physical eraseblock is successfully erased, and returns.
+ *
+ * Atomic logical eraseblock change
+ * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+ *
+ * Atomic logical eraseblock change operation is called using the %UBI_IOCEBCH
+ * ioctl command of the corresponding UBI volume character device. A pointer to
+ * a &struct ubi_leb_change_req object has to be passed to the ioctl. Then the
+ * user is expected to write the requested amount of bytes (similarly to what
+ * should be done in case of the "volume update" ioctl).
+ *
+ * Logical eraseblock map
+ * ~~~~~~~~~~~~~~~~~~~~~
+ *
+ * To map a logical eraseblock to a physical eraseblock, the %UBI_IOCEBMAP
+ * ioctl command should be used. A pointer to a &struct ubi_map_req object is
+ * expected to be passed. The ioctl maps the requested logical eraseblock to
+ * a physical eraseblock and returns.  Only non-mapped logical eraseblocks can
+ * be mapped. If the logical eraseblock specified in the request is already
+ * mapped to a physical eraseblock, the ioctl fails and returns error.
+ *
+ * Logical eraseblock unmap
  * ~~~~~~~~~~~~~~~~~~~~~~~~
  *
- * Atomic eraseblock change operation is done via the %UBI_IOCEBCH IOCTL
- * command of the corresponding UBI volume character device. A pointer to
- * &struct ubi_leb_change_req has to be passed to the IOCTL. Then the user is
- * expected to write the requested amount of bytes. This is similar to the
- * "volume update" IOCTL.
+ * To unmap a logical eraseblock to a physical eraseblock, the %UBI_IOCEBUNMAP
+ * ioctl command should be used. The ioctl unmaps the logical eraseblocks,
+ * schedules corresponding physical eraseblock for erasure, and returns. Unlike
+ * the "LEB erase" command, it does not wait for the physical eraseblock being
+ * erased. Note, the side effect of this is that if an unclean reboot happens
+ * after the unmap ioctl returns, you may find the LEB mapped again to the same
+ * physical eraseblock after the UBI is run again.
+ *
+ * Check if logical eraseblock is mapped
+ * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+ *
+ * To check if a logical eraseblock is mapped to a physical eraseblock, the
+ * %UBI_IOCEBISMAP ioctl command should be used. It returns %0 if the LEB is
+ * not mapped, and %1 if it is mapped.
+ *
+ * Set an UBI volume property
+ * ~~~~~~~~~~~~~~~~~~~~~~~~~
+ *
+ * To set an UBI volume property the %UBI_IOCSETPROP ioctl command should be
+ * used. A pointer to a &struct ubi_set_vol_prop_req object is expected to be
+ * passed. The object describes which property should be set, and to which value
+ * it should be set.
  */
 
 /*
 /* Maximum volume name length */
 #define UBI_MAX_VOLUME_NAME 127
 
-/* IOCTL commands of UBI character devices */
+/* ioctl commands of UBI character devices */
 
 #define UBI_IOC_MAGIC 'o'
 
 /* Create an UBI volume */
 #define UBI_IOCMKVOL _IOW(UBI_IOC_MAGIC, 0, struct ubi_mkvol_req)
 /* Remove an UBI volume */
-#define UBI_IOCRMVOL _IOW(UBI_IOC_MAGIC, 1, int32_t)
+#define UBI_IOCRMVOL _IOW(UBI_IOC_MAGIC, 1, __s32)
 /* Re-size an UBI volume */
 #define UBI_IOCRSVOL _IOW(UBI_IOC_MAGIC, 2, struct ubi_rsvol_req)
+/* Re-name volumes */
+#define UBI_IOCRNVOL _IOW(UBI_IOC_MAGIC, 3, struct ubi_rnvol_req)
 
-/* IOCTL commands of the UBI control character device */
+/* ioctl commands of the UBI control character device */
 
 #define UBI_CTRL_IOC_MAGIC 'o'
 
 /* Attach an MTD device */
 #define UBI_IOCATT _IOW(UBI_CTRL_IOC_MAGIC, 64, struct ubi_attach_req)
 /* Detach an MTD device */
-#define UBI_IOCDET _IOW(UBI_CTRL_IOC_MAGIC, 65, int32_t)
+#define UBI_IOCDET _IOW(UBI_CTRL_IOC_MAGIC, 65, __s32)
 
-/* IOCTL commands of UBI volume character devices */
+/* ioctl commands of UBI volume character devices */
 
 #define UBI_VOL_IOC_MAGIC 'O'
 
-/* Start UBI volume update */
-#define UBI_IOCVOLUP _IOW(UBI_VOL_IOC_MAGIC, 0, int64_t)
-/* An eraseblock erasure command, used for debugging, disabled by default */
-#define UBI_IOCEBER _IOW(UBI_VOL_IOC_MAGIC, 1, int32_t)
-/* An atomic eraseblock change command */
-#define UBI_IOCEBCH _IOW(UBI_VOL_IOC_MAGIC, 2, int32_t)
+/* Start UBI volume update
+ * Note: This actually takes a pointer (__s64*), but we can't change
+ *       that without breaking the ABI on 32bit systems
+ */
+#define UBI_IOCVOLUP _IOW(UBI_VOL_IOC_MAGIC, 0, __s64)
+/* LEB erasure command, used for debugging, disabled by default */
+#define UBI_IOCEBER _IOW(UBI_VOL_IOC_MAGIC, 1, __s32)
+/* Atomic LEB change command */
+#define UBI_IOCEBCH _IOW(UBI_VOL_IOC_MAGIC, 2, __s32)
+/* Map LEB command */
+#define UBI_IOCEBMAP _IOW(UBI_VOL_IOC_MAGIC, 3, struct ubi_map_req)
+/* Unmap LEB command */
+#define UBI_IOCEBUNMAP _IOW(UBI_VOL_IOC_MAGIC, 4, __s32)
+/* Check if LEB is mapped command */
+#define UBI_IOCEBISMAP _IOR(UBI_VOL_IOC_MAGIC, 5, __s32)
+/* Set an UBI volume property */
+#define UBI_IOCSETVOLPROP _IOW(UBI_VOL_IOC_MAGIC, 6, \
+                              struct ubi_set_vol_prop_req)
 
 /* Maximum MTD device name length supported by UBI */
 #define MAX_UBI_MTD_NAME_LEN 127
 
-/*
- * UBI data type hint constants.
- *
- * UBI_LONGTERM: long-term data
- * UBI_SHORTTERM: short-term data
- * UBI_UNKNOWN: data persistence is unknown
- *
- * These constants are used when data is written to UBI volumes in order to
- * help the UBI wear-leveling unit to find more appropriate physical
- * eraseblocks.
- */
-enum {
-       UBI_LONGTERM  = 1,
-       UBI_SHORTTERM = 2,
-       UBI_UNKNOWN   = 3,
-};
+/* Maximum amount of UBI volumes that can be re-named at one go */
+#define UBI_MAX_RNVOL 32
 
 /*
  * UBI volume type constants.
@@ -144,11 +197,23 @@ enum {
        UBI_STATIC_VOLUME  = 4,
 };
 
+/*
+ * UBI set volume property ioctl constants.
+ *
+ * @UBI_VOL_PROP_DIRECT_WRITE: allow (any non-zero value) or disallow (value 0)
+ *                             user to directly write and erase individual
+ *                             eraseblocks on dynamic volumes
+ */
+enum {
+       UBI_VOL_PROP_DIRECT_WRITE = 1,
+};
+
 /**
  * struct ubi_attach_req - attach MTD device request.
  * @ubi_num: UBI device number to create
  * @mtd_num: MTD device number to attach
  * @vid_hdr_offset: VID header offset (use defaults if %0)
+ * @max_beb_per1024: maximum expected number of bad PEB per 1024 PEBs
  * @padding: reserved for future, not used, has to be zeroed
  *
  * This data structure is used to specify MTD device UBI has to attach and the
@@ -164,20 +229,33 @@ enum {
  * it will be 512 in case of a 2KiB page NAND flash with 4 512-byte sub-pages.
  *
  * But in rare cases, if this optimizes things, the VID header may be placed to
- * a different offset. For example, the boot-loader might do things faster if the
- * VID header sits at the end of the first 2KiB NAND page with 4 sub-pages. As
- * the boot-loader would not normally need to read EC headers (unless it needs
- * UBI in RW mode), it might be faster to calculate ECC. This is weird example,
- * but it real-life example. So, in this example, @vid_hdr_offer would be
- * 2KiB-64 bytes = 1984. Note, that this position is not even 512-bytes
- * aligned, which is OK, as UBI is clever enough to realize this is 4th sub-page
- * of the first page and add needed padding.
+ * a different offset. For example, the boot-loader might do things faster if
+ * the VID header sits at the end of the first 2KiB NAND page with 4 sub-pages.
+ * As the boot-loader would not normally need to read EC headers (unless it
+ * needs UBI in RW mode), it might be faster to calculate ECC. This is weird
+ * example, but it real-life example. So, in this example, @vid_hdr_offer would
+ * be 2KiB-64 bytes = 1984. Note, that this position is not even 512-bytes
+ * aligned, which is OK, as UBI is clever enough to realize this is 4th
+ * sub-page of the first page and add needed padding.
+ *
+ * The @max_beb_per1024 is the maximum amount of bad PEBs UBI expects on the
+ * UBI device per 1024 eraseblocks.  This value is often given in an other form
+ * in the NAND datasheet (min NVB i.e. minimal number of valid blocks). The
+ * maximum expected bad eraseblocks per 1024 is then:
+ *    1024 * (1 - MinNVB / MaxNVB)
+ * Which gives 20 for most NAND devices.  This limit is used in order to derive
+ * amount of eraseblock UBI reserves for handling new bad blocks. If the device
+ * has more bad eraseblocks than this limit, UBI does not reserve any physical
+ * eraseblocks for new bad eraseblocks, but attempts to use available
+ * eraseblocks (if any). The accepted range is 0-768. If 0 is given, the
+ * default kernel value of %CONFIG_MTD_UBI_BEB_LIMIT will be used.
  */
 struct ubi_attach_req {
-       int32_t ubi_num;
-       int32_t mtd_num;
-       int32_t vid_hdr_offset;
-       uint8_t padding[12];
+       __s32 ubi_num;
+       __s32 mtd_num;
+       __s32 vid_hdr_offset;
+       __s16 max_beb_per1024;
+       __s8 padding[10];
 };
 
 /**
@@ -212,15 +290,15 @@ struct ubi_attach_req {
  * BLOBs, without caring about how to properly align them.
  */
 struct ubi_mkvol_req {
-       int32_t vol_id;
-       int32_t alignment;
-       int64_t bytes;
-       int8_t vol_type;
-       int8_t padding1;
-       int16_t name_len;
-       int8_t padding2[4];
+       __s32 vol_id;
+       __s32 alignment;
+       __s64 bytes;
+       __s8 vol_type;
+       __s8 padding1;
+       __s16 name_len;
+       __s8 padding2[4];
        char name[UBI_MAX_VOLUME_NAME + 1];
-} __attribute__ ((packed));
+} __packed;
 
 /**
  * struct ubi_rsvol_req - a data structure used in volume re-size requests.
@@ -229,28 +307,105 @@ struct ubi_mkvol_req {
  *
  * Re-sizing is possible for both dynamic and static volumes. But while dynamic
  * volumes may be re-sized arbitrarily, static volumes cannot be made to be
- * smaller then the number of bytes they bear. To arbitrarily shrink a static
+ * smaller than the number of bytes they bear. To arbitrarily shrink a static
  * volume, it must be wiped out first (by means of volume update operation with
  * zero number of bytes).
  */
 struct ubi_rsvol_req {
-       int64_t bytes;
-       int32_t vol_id;
-} __attribute__ ((packed));
+       __s64 bytes;
+       __s32 vol_id;
+} __packed;
 
 /**
- * struct ubi_leb_change_req - a data structure used in atomic logical
- *                             eraseblock change requests.
+ * struct ubi_rnvol_req - volumes re-name request.
+ * @count: count of volumes to re-name
+ * @padding1:  reserved for future, not used, has to be zeroed
+ * @vol_id: ID of the volume to re-name
+ * @name_len: name length
+ * @padding2:  reserved for future, not used, has to be zeroed
+ * @name: new volume name
+ *
+ * UBI allows to re-name up to %32 volumes at one go. The count of volumes to
+ * re-name is specified in the @count field. The ID of the volumes to re-name
+ * and the new names are specified in the @vol_id and @name fields.
+ *
+ * The UBI volume re-name operation is atomic, which means that should power cut
+ * happen, the volumes will have either old name or new name. So the possible
+ * use-cases of this command is atomic upgrade. Indeed, to upgrade, say, volumes
+ * A and B one may create temporary volumes %A1 and %B1 with the new contents,
+ * then atomically re-name A1->A and B1->B, in which case old %A and %B will
+ * be removed.
+ *
+ * If it is not desirable to remove old A and B, the re-name request has to
+ * contain 4 entries: A1->A, A->A1, B1->B, B->B1, in which case old A1 and B1
+ * become A and B, and old A and B will become A1 and B1.
+ *
+ * It is also OK to request: A1->A, A1->X, B1->B, B->Y, in which case old A1
+ * and B1 become A and B, and old A and B become X and Y.
+ *
+ * In other words, in case of re-naming into an existing volume name, the
+ * existing volume is removed, unless it is re-named as well at the same
+ * re-name request.
+ */
+struct ubi_rnvol_req {
+       __s32 count;
+       __s8 padding1[12];
+       struct {
+               __s32 vol_id;
+               __s16 name_len;
+               __s8  padding2[2];
+               char    name[UBI_MAX_VOLUME_NAME + 1];
+       } ents[UBI_MAX_RNVOL];
+} __packed;
+
+/**
+ * struct ubi_leb_change_req - a data structure used in atomic LEB change
+ *                             requests.
  * @lnum: logical eraseblock number to change
  * @bytes: how many bytes will be written to the logical eraseblock
- * @dtype: data type (%UBI_LONGTERM, %UBI_SHORTTERM, %UBI_UNKNOWN)
+ * @dtype: pass "3" for better compatibility with old kernels
  * @padding: reserved for future, not used, has to be zeroed
+ *
+ * The @dtype field used to inform UBI about what kind of data will be written
+ * to the LEB: long term (value 1), short term (value 2), unknown (value 3).
+ * UBI tried to pick a PEB with lower erase counter for short term data and a
+ * PEB with higher erase counter for long term data. But this was not really
+ * used because users usually do not know this and could easily mislead UBI. We
+ * removed this feature in May 2012. UBI currently just ignores the @dtype
+ * field. But for better compatibility with older kernels it is recommended to
+ * set @dtype to 3 (unknown).
  */
 struct ubi_leb_change_req {
-       int32_t lnum;
-       int32_t bytes;
-       uint8_t dtype;
-       uint8_t padding[7];
-} __attribute__ ((packed));
+       __s32 lnum;
+       __s32 bytes;
+       __s8  dtype; /* obsolete, do not use! */
+       __s8  padding[7];
+} __packed;
+
+/**
+ * struct ubi_map_req - a data structure used in map LEB requests.
+ * @dtype: pass "3" for better compatibility with old kernels
+ * @lnum: logical eraseblock number to unmap
+ * @padding: reserved for future, not used, has to be zeroed
+ */
+struct ubi_map_req {
+       __s32 lnum;
+       __s8  dtype; /* obsolete, do not use! */
+       __s8  padding[3];
+} __packed;
+
+
+/**
+ * struct ubi_set_vol_prop_req - a data structure used to set an UBI volume
+ *                               property.
+ * @property: property to set (%UBI_VOL_PROP_DIRECT_WRITE)
+ * @padding: reserved for future, not used, has to be zeroed
+ * @value: value to set
+ */
+struct ubi_set_vol_prop_req {
+       __u8  property;
+       __u8  padding[7];
+       __u64 value;
+}  __packed;
 
 #endif /* __UBI_USER_H__ */
index a25e9d9ef3bc8fb10acfdb94e3b54f6299800f52..29fb166934c8d5d71e09d080665ed555910fe7da 100644 (file)
 #include <linux/compat.h>
 
 /* common */
-#define spin_lock_init(...)
-#define spin_lock(...)
-#define spin_lock_irqsave(lock, flags) do { debug("%lu\n", flags); } while (0)
-#define spin_unlock(...)
-#define spin_unlock_irqrestore(lock, flags) do {flags = 0; } while (0)
-#define disable_irq(...)
-#define enable_irq(...)
-
-#define mutex_init(...)
-#define mutex_lock(...)
-#define mutex_unlock(...)
-
-#define GFP_KERNEL     0
-
-#define IRQ_HANDLED    1
-
 #define ENOTSUPP       524     /* Operation is not supported */
 
 #define BITS_PER_BYTE                          8