2 * This file is part of UBIFS.
4 * Copyright (C) 2006-2008 Nokia Corporation.
6 * SPDX-License-Identifier: GPL-2.0+
8 * Authors: Adrian Hunter
9 * Artem Bityutskiy (Битюцкий Артём)
13 * This file contains journal replay code. It runs when the file-system is being
14 * mounted and requires no locking.
16 * The larger is the journal, the longer it takes to scan it, so the longer it
17 * takes to mount UBIFS. This is why the journal has limited size which may be
18 * changed depending on the system requirements. But a larger journal gives
19 * faster I/O speed because it writes the index less frequently. So this is a
20 * trade-off. Also, the journal is indexed by the in-memory index (TNC), so the
21 * larger is the journal, the more memory its index may consume.
25 #include <linux/compat.h>
26 #include <linux/err.h>
29 #include <linux/list_sort.h>
32 * struct replay_entry - replay list entry.
33 * @lnum: logical eraseblock number of the node
36 * @deletion: non-zero if this entry corresponds to a node deletion
37 * @sqnum: node sequence number
38 * @list: links the replay list
40 * @nm: directory entry name
41 * @old_size: truncation old size
42 * @new_size: truncation new size
44 * The replay process first scans all buds and builds the replay list, then
45 * sorts the replay list in nodes sequence number order, and then inserts all
46 * the replay entries to the TNC.
52 unsigned int deletion:1;
53 unsigned long long sqnum;
54 struct list_head list;
66 * struct bud_entry - entry in the list of buds to replay.
67 * @list: next bud in the list
68 * @bud: bud description object
69 * @sqnum: reference node sequence number
70 * @free: free bytes in the bud
71 * @dirty: dirty bytes in the bud
74 struct list_head list;
75 struct ubifs_bud *bud;
76 unsigned long long sqnum;
82 * set_bud_lprops - set free and dirty space used by a bud.
83 * @c: UBIFS file-system description object
84 * @b: bud entry which describes the bud
86 * This function makes sure the LEB properties of bud @b are set correctly
87 * after the replay. Returns zero in case of success and a negative error code
90 static int set_bud_lprops(struct ubifs_info *c, struct bud_entry *b)
92 const struct ubifs_lprops *lp;
97 lp = ubifs_lpt_lookup_dirty(c, b->bud->lnum);
104 if (b->bud->start == 0 && (lp->free != c->leb_size || lp->dirty != 0)) {
106 * The LEB was added to the journal with a starting offset of
107 * zero which means the LEB must have been empty. The LEB
108 * property values should be @lp->free == @c->leb_size and
109 * @lp->dirty == 0, but that is not the case. The reason is that
110 * the LEB had been garbage collected before it became the bud,
111 * and there was not commit inbetween. The garbage collector
112 * resets the free and dirty space without recording it
113 * anywhere except lprops, so if there was no commit then
114 * lprops does not have that information.
116 * We do not need to adjust free space because the scan has told
117 * us the exact value which is recorded in the replay entry as
120 * However we do need to subtract from the dirty space the
121 * amount of space that the garbage collector reclaimed, which
122 * is the whole LEB minus the amount of space that was free.
124 dbg_mnt("bud LEB %d was GC'd (%d free, %d dirty)", b->bud->lnum,
125 lp->free, lp->dirty);
126 dbg_gc("bud LEB %d was GC'd (%d free, %d dirty)", b->bud->lnum,
127 lp->free, lp->dirty);
128 dirty -= c->leb_size - lp->free;
130 * If the replay order was perfect the dirty space would now be
131 * zero. The order is not perfect because the journal heads
132 * race with each other. This is not a problem but is does mean
133 * that the dirty space may temporarily exceed c->leb_size
137 dbg_mnt("LEB %d lp: %d free %d dirty replay: %d free %d dirty",
138 b->bud->lnum, lp->free, lp->dirty, b->free,
141 lp = ubifs_change_lp(c, lp, b->free, dirty + b->dirty,
142 lp->flags | LPROPS_TAKEN, 0);
148 /* Make sure the journal head points to the latest bud */
149 err = ubifs_wbuf_seek_nolock(&c->jheads[b->bud->jhead].wbuf,
150 b->bud->lnum, c->leb_size - b->free);
153 ubifs_release_lprops(c);
158 * set_buds_lprops - set free and dirty space for all replayed buds.
159 * @c: UBIFS file-system description object
161 * This function sets LEB properties for all replayed buds. Returns zero in
162 * case of success and a negative error code in case of failure.
164 static int set_buds_lprops(struct ubifs_info *c)
169 list_for_each_entry(b, &c->replay_buds, list) {
170 err = set_bud_lprops(c, b);
179 * trun_remove_range - apply a replay entry for a truncation to the TNC.
180 * @c: UBIFS file-system description object
181 * @r: replay entry of truncation
183 static int trun_remove_range(struct ubifs_info *c, struct replay_entry *r)
185 unsigned min_blk, max_blk;
186 union ubifs_key min_key, max_key;
189 min_blk = r->new_size / UBIFS_BLOCK_SIZE;
190 if (r->new_size & (UBIFS_BLOCK_SIZE - 1))
193 max_blk = r->old_size / UBIFS_BLOCK_SIZE;
194 if ((r->old_size & (UBIFS_BLOCK_SIZE - 1)) == 0)
197 ino = key_inum(c, &r->key);
199 data_key_init(c, &min_key, ino, min_blk);
200 data_key_init(c, &max_key, ino, max_blk);
202 return ubifs_tnc_remove_range(c, &min_key, &max_key);
206 * apply_replay_entry - apply a replay entry to the TNC.
207 * @c: UBIFS file-system description object
208 * @r: replay entry to apply
210 * Apply a replay entry to the TNC.
212 static int apply_replay_entry(struct ubifs_info *c, struct replay_entry *r)
216 dbg_mntk(&r->key, "LEB %d:%d len %d deletion %d sqnum %llu key ",
217 r->lnum, r->offs, r->len, r->deletion, r->sqnum);
219 /* Set c->replay_sqnum to help deal with dangling branches. */
220 c->replay_sqnum = r->sqnum;
222 if (is_hash_key(c, &r->key)) {
224 err = ubifs_tnc_remove_nm(c, &r->key, &r->nm);
226 err = ubifs_tnc_add_nm(c, &r->key, r->lnum, r->offs,
230 switch (key_type(c, &r->key)) {
233 ino_t inum = key_inum(c, &r->key);
235 err = ubifs_tnc_remove_ino(c, inum);
239 err = trun_remove_range(c, r);
242 err = ubifs_tnc_remove(c, &r->key);
246 err = ubifs_tnc_add(c, &r->key, r->lnum, r->offs,
251 if (c->need_recovery)
252 err = ubifs_recover_size_accum(c, &r->key, r->deletion,
260 * replay_entries_cmp - compare 2 replay entries.
261 * @priv: UBIFS file-system description object
262 * @a: first replay entry
263 * @a: second replay entry
265 * This is a comparios function for 'list_sort()' which compares 2 replay
266 * entries @a and @b by comparing their sequence numer. Returns %1 if @a has
267 * greater sequence number and %-1 otherwise.
269 static int replay_entries_cmp(void *priv, struct list_head *a,
272 struct replay_entry *ra, *rb;
278 ra = list_entry(a, struct replay_entry, list);
279 rb = list_entry(b, struct replay_entry, list);
280 ubifs_assert(ra->sqnum != rb->sqnum);
281 if (ra->sqnum > rb->sqnum)
287 * apply_replay_list - apply the replay list to the TNC.
288 * @c: UBIFS file-system description object
290 * Apply all entries in the replay list to the TNC. Returns zero in case of
291 * success and a negative error code in case of failure.
293 static int apply_replay_list(struct ubifs_info *c)
295 struct replay_entry *r;
298 list_sort(c, &c->replay_list, &replay_entries_cmp);
300 list_for_each_entry(r, &c->replay_list, list) {
303 err = apply_replay_entry(c, r);
312 * destroy_replay_list - destroy the replay.
313 * @c: UBIFS file-system description object
315 * Destroy the replay list.
317 static void destroy_replay_list(struct ubifs_info *c)
319 struct replay_entry *r, *tmp;
321 list_for_each_entry_safe(r, tmp, &c->replay_list, list) {
322 if (is_hash_key(c, &r->key))
330 * insert_node - insert a node to the replay list
331 * @c: UBIFS file-system description object
332 * @lnum: node logical eraseblock number
336 * @sqnum: sequence number
337 * @deletion: non-zero if this is a deletion
338 * @used: number of bytes in use in a LEB
339 * @old_size: truncation old size
340 * @new_size: truncation new size
342 * This function inserts a scanned non-direntry node to the replay list. The
343 * replay list contains @struct replay_entry elements, and we sort this list in
344 * sequence number order before applying it. The replay list is applied at the
345 * very end of the replay process. Since the list is sorted in sequence number
346 * order, the older modifications are applied first. This function returns zero
347 * in case of success and a negative error code in case of failure.
349 static int insert_node(struct ubifs_info *c, int lnum, int offs, int len,
350 union ubifs_key *key, unsigned long long sqnum,
351 int deletion, int *used, loff_t old_size,
354 struct replay_entry *r;
356 dbg_mntk(key, "add LEB %d:%d, key ", lnum, offs);
358 if (key_inum(c, key) >= c->highest_inum)
359 c->highest_inum = key_inum(c, key);
361 r = kzalloc(sizeof(struct replay_entry), GFP_KERNEL);
366 *used += ALIGN(len, 8);
370 r->deletion = !!deletion;
372 key_copy(c, key, &r->key);
373 r->old_size = old_size;
374 r->new_size = new_size;
376 list_add_tail(&r->list, &c->replay_list);
381 * insert_dent - insert a directory entry node into the replay list.
382 * @c: UBIFS file-system description object
383 * @lnum: node logical eraseblock number
387 * @name: directory entry name
388 * @nlen: directory entry name length
389 * @sqnum: sequence number
390 * @deletion: non-zero if this is a deletion
391 * @used: number of bytes in use in a LEB
393 * This function inserts a scanned directory entry node or an extended
394 * attribute entry to the replay list. Returns zero in case of success and a
395 * negative error code in case of failure.
397 static int insert_dent(struct ubifs_info *c, int lnum, int offs, int len,
398 union ubifs_key *key, const char *name, int nlen,
399 unsigned long long sqnum, int deletion, int *used)
401 struct replay_entry *r;
404 dbg_mntk(key, "add LEB %d:%d, key ", lnum, offs);
405 if (key_inum(c, key) >= c->highest_inum)
406 c->highest_inum = key_inum(c, key);
408 r = kzalloc(sizeof(struct replay_entry), GFP_KERNEL);
412 nbuf = kmalloc(nlen + 1, GFP_KERNEL);
419 *used += ALIGN(len, 8);
423 r->deletion = !!deletion;
425 key_copy(c, key, &r->key);
427 memcpy(nbuf, name, nlen);
431 list_add_tail(&r->list, &c->replay_list);
436 * ubifs_validate_entry - validate directory or extended attribute entry node.
437 * @c: UBIFS file-system description object
438 * @dent: the node to validate
440 * This function validates directory or extended attribute entry node @dent.
441 * Returns zero if the node is all right and a %-EINVAL if not.
443 int ubifs_validate_entry(struct ubifs_info *c,
444 const struct ubifs_dent_node *dent)
446 int key_type = key_type_flash(c, dent->key);
447 int nlen = le16_to_cpu(dent->nlen);
449 if (le32_to_cpu(dent->ch.len) != nlen + UBIFS_DENT_NODE_SZ + 1 ||
450 dent->type >= UBIFS_ITYPES_CNT ||
451 nlen > UBIFS_MAX_NLEN || dent->name[nlen] != 0 ||
452 strnlen(dent->name, nlen) != nlen ||
453 le64_to_cpu(dent->inum) > MAX_INUM) {
454 ubifs_err(c, "bad %s node", key_type == UBIFS_DENT_KEY ?
455 "directory entry" : "extended attribute entry");
459 if (key_type != UBIFS_DENT_KEY && key_type != UBIFS_XENT_KEY) {
460 ubifs_err(c, "bad key type %d", key_type);
468 * is_last_bud - check if the bud is the last in the journal head.
469 * @c: UBIFS file-system description object
470 * @bud: bud description object
472 * This function checks if bud @bud is the last bud in its journal head. This
473 * information is then used by 'replay_bud()' to decide whether the bud can
474 * have corruptions or not. Indeed, only last buds can be corrupted by power
475 * cuts. Returns %1 if this is the last bud, and %0 if not.
477 static int is_last_bud(struct ubifs_info *c, struct ubifs_bud *bud)
479 struct ubifs_jhead *jh = &c->jheads[bud->jhead];
480 struct ubifs_bud *next;
484 if (list_is_last(&bud->list, &jh->buds_list))
488 * The following is a quirk to make sure we work correctly with UBIFS
489 * images used with older UBIFS.
491 * Normally, the last bud will be the last in the journal head's list
492 * of bud. However, there is one exception if the UBIFS image belongs
493 * to older UBIFS. This is fairly unlikely: one would need to use old
494 * UBIFS, then have a power cut exactly at the right point, and then
495 * try to mount this image with new UBIFS.
497 * The exception is: it is possible to have 2 buds A and B, A goes
498 * before B, and B is the last, bud B is contains no data, and bud A is
499 * corrupted at the end. The reason is that in older versions when the
500 * journal code switched the next bud (from A to B), it first added a
501 * log reference node for the new bud (B), and only after this it
502 * synchronized the write-buffer of current bud (A). But later this was
503 * changed and UBIFS started to always synchronize the write-buffer of
504 * the bud (A) before writing the log reference for the new bud (B).
506 * But because older UBIFS always synchronized A's write-buffer before
507 * writing to B, we can recognize this exceptional situation but
508 * checking the contents of bud B - if it is empty, then A can be
509 * treated as the last and we can recover it.
511 * TODO: remove this piece of code in a couple of years (today it is
514 next = list_entry(bud->list.next, struct ubifs_bud, list);
515 if (!list_is_last(&next->list, &jh->buds_list))
518 err = ubifs_leb_read(c, next->lnum, (char *)&data, next->start, 4, 1);
522 return data == 0xFFFFFFFF;
526 * replay_bud - replay a bud logical eraseblock.
527 * @c: UBIFS file-system description object
528 * @b: bud entry which describes the bud
530 * This function replays bud @bud, recovers it if needed, and adds all nodes
531 * from this bud to the replay list. Returns zero in case of success and a
532 * negative error code in case of failure.
534 static int replay_bud(struct ubifs_info *c, struct bud_entry *b)
536 int is_last = is_last_bud(c, b->bud);
537 int err = 0, used = 0, lnum = b->bud->lnum, offs = b->bud->start;
538 struct ubifs_scan_leb *sleb;
539 struct ubifs_scan_node *snod;
541 dbg_mnt("replay bud LEB %d, head %d, offs %d, is_last %d",
542 lnum, b->bud->jhead, offs, is_last);
544 if (c->need_recovery && is_last)
546 * Recover only last LEBs in the journal heads, because power
547 * cuts may cause corruptions only in these LEBs, because only
548 * these LEBs could possibly be written to at the power cut
551 sleb = ubifs_recover_leb(c, lnum, offs, c->sbuf, b->bud->jhead);
553 sleb = ubifs_scan(c, lnum, offs, c->sbuf, 0);
555 return PTR_ERR(sleb);
558 * The bud does not have to start from offset zero - the beginning of
559 * the 'lnum' LEB may contain previously committed data. One of the
560 * things we have to do in replay is to correctly update lprops with
561 * newer information about this LEB.
563 * At this point lprops thinks that this LEB has 'c->leb_size - offs'
564 * bytes of free space because it only contain information about
567 * But we know that real amount of free space is 'c->leb_size -
568 * sleb->endpt', and the space in the 'lnum' LEB between 'offs' and
569 * 'sleb->endpt' is used by bud data. We have to correctly calculate
570 * how much of these data are dirty and update lprops with this
573 * The dirt in that LEB region is comprised of padding nodes, deletion
574 * nodes, truncation nodes and nodes which are obsoleted by subsequent
575 * nodes in this LEB. So instead of calculating clean space, we
576 * calculate used space ('used' variable).
579 list_for_each_entry(snod, &sleb->nodes, list) {
584 if (snod->sqnum >= SQNUM_WATERMARK) {
585 ubifs_err(c, "file system's life ended");
589 if (snod->sqnum > c->max_sqnum)
590 c->max_sqnum = snod->sqnum;
592 switch (snod->type) {
595 struct ubifs_ino_node *ino = snod->node;
596 loff_t new_size = le64_to_cpu(ino->size);
598 if (le32_to_cpu(ino->nlink) == 0)
600 err = insert_node(c, lnum, snod->offs, snod->len,
601 &snod->key, snod->sqnum, deletion,
605 case UBIFS_DATA_NODE:
607 struct ubifs_data_node *dn = snod->node;
608 loff_t new_size = le32_to_cpu(dn->size) +
609 key_block(c, &snod->key) *
612 err = insert_node(c, lnum, snod->offs, snod->len,
613 &snod->key, snod->sqnum, deletion,
617 case UBIFS_DENT_NODE:
618 case UBIFS_XENT_NODE:
620 struct ubifs_dent_node *dent = snod->node;
622 err = ubifs_validate_entry(c, dent);
626 err = insert_dent(c, lnum, snod->offs, snod->len,
627 &snod->key, dent->name,
628 le16_to_cpu(dent->nlen), snod->sqnum,
629 !le64_to_cpu(dent->inum), &used);
632 case UBIFS_TRUN_NODE:
634 struct ubifs_trun_node *trun = snod->node;
635 loff_t old_size = le64_to_cpu(trun->old_size);
636 loff_t new_size = le64_to_cpu(trun->new_size);
639 /* Validate truncation node */
640 if (old_size < 0 || old_size > c->max_inode_sz ||
641 new_size < 0 || new_size > c->max_inode_sz ||
642 old_size <= new_size) {
643 ubifs_err(c, "bad truncation node");
648 * Create a fake truncation key just to use the same
649 * functions which expect nodes to have keys.
651 trun_key_init(c, &key, le32_to_cpu(trun->inum));
652 err = insert_node(c, lnum, snod->offs, snod->len,
653 &key, snod->sqnum, 1, &used,
658 ubifs_err(c, "unexpected node type %d in bud LEB %d:%d",
659 snod->type, lnum, snod->offs);
667 ubifs_assert(ubifs_search_bud(c, lnum));
668 ubifs_assert(sleb->endpt - offs >= used);
669 ubifs_assert(sleb->endpt % c->min_io_size == 0);
671 b->dirty = sleb->endpt - offs - used;
672 b->free = c->leb_size - sleb->endpt;
673 dbg_mnt("bud LEB %d replied: dirty %d, free %d",
674 lnum, b->dirty, b->free);
677 ubifs_scan_destroy(sleb);
681 ubifs_err(c, "bad node is at LEB %d:%d", lnum, snod->offs);
682 ubifs_dump_node(c, snod->node);
683 ubifs_scan_destroy(sleb);
688 * replay_buds - replay all buds.
689 * @c: UBIFS file-system description object
691 * This function returns zero in case of success and a negative error code in
694 static int replay_buds(struct ubifs_info *c)
698 unsigned long long prev_sqnum = 0;
700 list_for_each_entry(b, &c->replay_buds, list) {
701 err = replay_bud(c, b);
705 ubifs_assert(b->sqnum > prev_sqnum);
706 prev_sqnum = b->sqnum;
713 * destroy_bud_list - destroy the list of buds to replay.
714 * @c: UBIFS file-system description object
716 static void destroy_bud_list(struct ubifs_info *c)
720 while (!list_empty(&c->replay_buds)) {
721 b = list_entry(c->replay_buds.next, struct bud_entry, list);
728 * add_replay_bud - add a bud to the list of buds to replay.
729 * @c: UBIFS file-system description object
730 * @lnum: bud logical eraseblock number to replay
731 * @offs: bud start offset
732 * @jhead: journal head to which this bud belongs
733 * @sqnum: reference node sequence number
735 * This function returns zero in case of success and a negative error code in
738 static int add_replay_bud(struct ubifs_info *c, int lnum, int offs, int jhead,
739 unsigned long long sqnum)
741 struct ubifs_bud *bud;
744 dbg_mnt("add replay bud LEB %d:%d, head %d", lnum, offs, jhead);
746 bud = kmalloc(sizeof(struct ubifs_bud), GFP_KERNEL);
750 b = kmalloc(sizeof(struct bud_entry), GFP_KERNEL);
759 ubifs_add_bud(c, bud);
763 list_add_tail(&b->list, &c->replay_buds);
769 * validate_ref - validate a reference node.
770 * @c: UBIFS file-system description object
771 * @ref: the reference node to validate
772 * @ref_lnum: LEB number of the reference node
773 * @ref_offs: reference node offset
775 * This function returns %1 if a bud reference already exists for the LEB. %0 is
776 * returned if the reference node is new, otherwise %-EINVAL is returned if
779 static int validate_ref(struct ubifs_info *c, const struct ubifs_ref_node *ref)
781 struct ubifs_bud *bud;
782 int lnum = le32_to_cpu(ref->lnum);
783 unsigned int offs = le32_to_cpu(ref->offs);
784 unsigned int jhead = le32_to_cpu(ref->jhead);
787 * ref->offs may point to the end of LEB when the journal head points
788 * to the end of LEB and we write reference node for it during commit.
789 * So this is why we require 'offs > c->leb_size'.
791 if (jhead >= c->jhead_cnt || lnum >= c->leb_cnt ||
792 lnum < c->main_first || offs > c->leb_size ||
793 offs & (c->min_io_size - 1))
796 /* Make sure we have not already looked at this bud */
797 bud = ubifs_search_bud(c, lnum);
799 if (bud->jhead == jhead && bud->start <= offs)
801 ubifs_err(c, "bud at LEB %d:%d was already referred", lnum, offs);
809 * replay_log_leb - replay a log logical eraseblock.
810 * @c: UBIFS file-system description object
811 * @lnum: log logical eraseblock to replay
812 * @offs: offset to start replaying from
815 * This function replays a log LEB and returns zero in case of success, %1 if
816 * this is the last LEB in the log, and a negative error code in case of
819 static int replay_log_leb(struct ubifs_info *c, int lnum, int offs, void *sbuf)
822 struct ubifs_scan_leb *sleb;
823 struct ubifs_scan_node *snod;
824 const struct ubifs_cs_node *node;
826 dbg_mnt("replay log LEB %d:%d", lnum, offs);
827 sleb = ubifs_scan(c, lnum, offs, sbuf, c->need_recovery);
829 if (PTR_ERR(sleb) != -EUCLEAN || !c->need_recovery)
830 return PTR_ERR(sleb);
832 * Note, the below function will recover this log LEB only if
833 * it is the last, because unclean reboots can possibly corrupt
834 * only the tail of the log.
836 sleb = ubifs_recover_log_leb(c, lnum, offs, sbuf);
838 return PTR_ERR(sleb);
841 if (sleb->nodes_cnt == 0) {
847 snod = list_entry(sleb->nodes.next, struct ubifs_scan_node, list);
848 if (c->cs_sqnum == 0) {
850 * This is the first log LEB we are looking at, make sure that
851 * the first node is a commit start node. Also record its
852 * sequence number so that UBIFS can determine where the log
853 * ends, because all nodes which were have higher sequence
856 if (snod->type != UBIFS_CS_NODE) {
857 ubifs_err(c, "first log node at LEB %d:%d is not CS node",
861 if (le64_to_cpu(node->cmt_no) != c->cmt_no) {
862 ubifs_err(c, "first CS node at LEB %d:%d has wrong commit number %llu expected %llu",
864 (unsigned long long)le64_to_cpu(node->cmt_no),
869 c->cs_sqnum = le64_to_cpu(node->ch.sqnum);
870 dbg_mnt("commit start sqnum %llu", c->cs_sqnum);
873 if (snod->sqnum < c->cs_sqnum) {
875 * This means that we reached end of log and now
876 * look to the older log data, which was already
877 * committed but the eraseblock was not erased (UBIFS
878 * only un-maps it). So this basically means we have to
879 * exit with "end of log" code.
885 /* Make sure the first node sits at offset zero of the LEB */
886 if (snod->offs != 0) {
887 ubifs_err(c, "first node is not at zero offset");
891 list_for_each_entry(snod, &sleb->nodes, list) {
894 if (snod->sqnum >= SQNUM_WATERMARK) {
895 ubifs_err(c, "file system's life ended");
899 if (snod->sqnum < c->cs_sqnum) {
900 ubifs_err(c, "bad sqnum %llu, commit sqnum %llu",
901 snod->sqnum, c->cs_sqnum);
905 if (snod->sqnum > c->max_sqnum)
906 c->max_sqnum = snod->sqnum;
908 switch (snod->type) {
909 case UBIFS_REF_NODE: {
910 const struct ubifs_ref_node *ref = snod->node;
912 err = validate_ref(c, ref);
914 break; /* Already have this bud */
918 err = add_replay_bud(c, le32_to_cpu(ref->lnum),
919 le32_to_cpu(ref->offs),
920 le32_to_cpu(ref->jhead),
928 /* Make sure it sits at the beginning of LEB */
929 if (snod->offs != 0) {
930 ubifs_err(c, "unexpected node in log");
935 ubifs_err(c, "unexpected node in log");
940 if (sleb->endpt || c->lhead_offs >= c->leb_size) {
941 c->lhead_lnum = lnum;
942 c->lhead_offs = sleb->endpt;
947 ubifs_scan_destroy(sleb);
951 ubifs_err(c, "log error detected while replaying the log at LEB %d:%d",
952 lnum, offs + snod->offs);
953 ubifs_dump_node(c, snod->node);
954 ubifs_scan_destroy(sleb);
959 * take_ihead - update the status of the index head in lprops to 'taken'.
960 * @c: UBIFS file-system description object
962 * This function returns the amount of free space in the index head LEB or a
963 * negative error code.
965 static int take_ihead(struct ubifs_info *c)
967 const struct ubifs_lprops *lp;
972 lp = ubifs_lpt_lookup_dirty(c, c->ihead_lnum);
980 lp = ubifs_change_lp(c, lp, LPROPS_NC, LPROPS_NC,
981 lp->flags | LPROPS_TAKEN, 0);
989 ubifs_release_lprops(c);
994 * ubifs_replay_journal - replay journal.
995 * @c: UBIFS file-system description object
997 * This function scans the journal, replays and cleans it up. It makes sure all
998 * memory data structures related to uncommitted journal are built (dirty TNC
999 * tree, tree of buds, modified lprops, etc).
1001 int ubifs_replay_journal(struct ubifs_info *c)
1003 int err, lnum, free;
1005 BUILD_BUG_ON(UBIFS_TRUN_KEY > 5);
1007 /* Update the status of the index head in lprops to 'taken' */
1008 free = take_ihead(c);
1010 return free; /* Error code */
1012 if (c->ihead_offs != c->leb_size - free) {
1013 ubifs_err(c, "bad index head LEB %d:%d", c->ihead_lnum,
1018 dbg_mnt("start replaying the journal");
1020 lnum = c->ltail_lnum = c->lhead_lnum;
1023 err = replay_log_leb(c, lnum, 0, c->sbuf);
1025 if (lnum != c->lhead_lnum)
1026 /* We hit the end of the log */
1030 * The head of the log must always start with the
1031 * "commit start" node on a properly formatted UBIFS.
1032 * But we found no nodes at all, which means that
1033 * someting went wrong and we cannot proceed mounting
1036 ubifs_err(c, "no UBIFS nodes found at the log head LEB %d:%d, possibly corrupted",
1042 lnum = ubifs_next_log_lnum(c, lnum);
1043 } while (lnum != c->ltail_lnum);
1045 err = replay_buds(c);
1049 err = apply_replay_list(c);
1053 err = set_buds_lprops(c);
1058 * UBIFS budgeting calculations use @c->bi.uncommitted_idx variable
1059 * to roughly estimate index growth. Things like @c->bi.min_idx_lebs
1060 * depend on it. This means we have to initialize it to make sure
1061 * budgeting works properly.
1063 c->bi.uncommitted_idx = atomic_long_read(&c->dirty_zn_cnt);
1064 c->bi.uncommitted_idx *= c->max_idx_node_sz;
1066 ubifs_assert(c->bud_bytes <= c->max_bud_bytes || c->need_recovery);
1067 dbg_mnt("finished, log head LEB %d:%d, max_sqnum %llu, highest_inum %lu",
1068 c->lhead_lnum, c->lhead_offs, c->max_sqnum,
1069 (unsigned long)c->highest_inum);
1071 destroy_replay_list(c);
1072 destroy_bud_list(c);