York Sun [Thu, 7 Dec 2017 21:16:07 +0000 (13:16 -0800)]
armv8: fix gd after relocation
Commit 21f4486faa5d ("armv8: update gd after relocate") sets x18
without checking the return value of spl_relocate_stack_gd().
Reviewed-by: Stephen Warren <swarren@nvidia.com> Signed-off-by: York Sun <york.sun@nxp.com> CC: Kever Yang <kever.yang@rock-chips.com> CC: Philipp Tomsich <philipp.tomsich@theobroma-systems.com>
Simon Glass [Mon, 4 Dec 2017 20:48:22 +0000 (13:48 -0700)]
usb: Correct use of debug()
With clang this gives a warning because hubsts appears to be used before
it is set, even if ultimately it is not used. Simplify the code to avoid
this problem.
Simon Glass [Mon, 4 Dec 2017 20:48:21 +0000 (13:48 -0700)]
mtdparts: Correct use of debug()
The debug() macro now evaluates its expression so does not need #ifdef
protection. In fact the current code causes a warning with the new log
implementation. Adjust the code to fix this.
Signed-off-by: Simon Glass <sjg@chromium.org> Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
Simon Glass [Mon, 4 Dec 2017 20:48:19 +0000 (13:48 -0700)]
Revert "sandbox: Drop special case console code for sandbox"
While sandbox works OK without the special-case code, it does result in
console output being stored in the pre-console buffer while sandbox starts
up. If there is a crash or a problem then there is no indication of what
is going on.
For ease of debugging it seems better to revert this change also.
Simon Glass [Mon, 4 Dec 2017 20:48:17 +0000 (13:48 -0700)]
Revert "sandbox: remove os_putc() and os_puts()"
While sandbox works OK without the special-case code, it does result in
console output being stored in the pre-console buffer while sandbox starts
up. If there is a crash or a problem then there is no indication of what
is going on.
For ease of debugging it seems better to revert this change.
Alan Ott [Wed, 29 Nov 2017 03:25:24 +0000 (22:25 -0500)]
i2c: at91_i2c: remove the .probe_chip function
The .probe_chip function is supposed to probe an i2c device on the bus to
determine whether a device is answering to a particular address.
at91_i2c_probe_chip() did not do anything resembling this and always
returned 0.
It looks as though at91_i2c_probe_chip() was intended to be a .probe
function for the controller, as it was copied-and-pasted to become
at91_i2c_probe() in 0bc8f640a4d7ed.
Removing the at91_i2c_probe_chip() function makes the higher layer
(i2c_probe_chip()) try a zero-length read transfer to test for the
presence of a device instead, which does work.
Signed-off-by: Alan Ott <alan@softiron.com> Acked-by: Wenyou Yang <wenyou.yang@microchip.com> Reviewed-by: Heiko Schocher <hs@denx.de>
Alan Ott [Wed, 29 Nov 2017 03:25:23 +0000 (22:25 -0500)]
i2c: at91_i2c: Wait for TXRDY after sending the first byte
The driver must wait for TXRDY after each byte is pushed into
the i2c FIFO before pushing the next byte. Previously this was
not done for the first byte, causing a race condition with zeros
sometimes being sent for the next byte (which is typically the
first actual data byte).
Signed-off-by: Alan Ott <alan@softiron.com> Reviewed-by: Heiko Schocher <hs@denx.de>
Philipp Tomsich [Mon, 4 Dec 2017 16:04:02 +0000 (17:04 +0100)]
tools: omapimage: fix corner-case in byteswap path
Since commit 2614a208471e ("common: command: tempory buffer should
have size of command line buf"), there have been consistent Travis CI
failures on my builds (interestingly not for Tom, even though building
the same commit id) due to a SEGV in building the byteswapped
omapimage:
arm: pcm051_rev3
make[2]: *** [MLO.byteswap] Error 139
^^^ error code for a SEGV
Turns out that the word-based byte-swapping loop in omapimage.c is to
blame. With the loop condition
while (swapped <= (sbuf->st_size / sizeof(uint32_t)))
there had been one-too-many iterations for all file sizes divisible by
the sizeof(uint32_t). I.e. we had 1 iteration for 0 bytes (and also 1
through 3 bytes) and 2 iterations at 4 bytes... clearly overshooting
on 0 and 4 bytes.
This commit fixes the calculation of an up-rounded word-count and
makes sure to keep the zero-based loop-counter below the number of
words to be processed.
References: 2614a20 ("common: command: tempory buffer should have size of command line buf") Fixes: 79b9ebb ("omapimage: Add support for byteswapped SPI images") Signed-off-by: Philipp Tomsich <philipp.tomsich@theobroma-systems.com> Reviewed-by: Martin Elshuber <martin.elshuber@theobroma-systems.com>
Alexander Graf [Mon, 4 Dec 2017 15:33:51 +0000 (16:33 +0100)]
efi_stub: Use efi_uintn_t
Commit f5a2a93892f ("efi_loader: consistently use efi_uintn_t in boot
services") changed the internal EFI API header without adapting its existing
EFI stub users. Let's adapt the EFI stub as well.
Fixes: f5a2a93892f ("efi_loader: consistently use efi_uintn_t in boot services") Signed-off-by: Alexander Graf <agraf@suse.de> Reviewed-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
Peng Fan [Thu, 30 Nov 2017 02:07:07 +0000 (10:07 +0800)]
scripts: spl: fix typo
Typo fix: CONIFG->CONFIG
Signed-off-by: Peng Fan <peng.fan@nxp.com> Cc: Tom Rini <trini@konsulko.com> Cc: Simon Glass <sjg@chromium.org> Reviewed-by: Simon Glass <sjg@chromium.org>
Derald D. Woods [Wed, 29 Nov 2017 03:28:55 +0000 (03:28 +0000)]
omap3: evm: Explicitly use DISTRO_DEFAULTS features at startup
[primary] Check MMC 0:1 for /extlinux/extlinux.conf and boot
[fallback 1] Check MMC 0:1 zImage and run mmcbootz
[fallback 2] Check MMC 0:1 uImage and run mmcboot
[fallback 3] Check NAND partitions and run nandboot
If 'extlinux.conf' is not found on MMC 0, the previous boot behavior is
followed.
Signed-off-by: Derald D. Woods <woods.technical@gmail.com>
Neil Armstrong [Mon, 27 Nov 2017 09:35:46 +0000 (10:35 +0100)]
ARM: arch-meson: build memory banks using reported memory from registers
As discussed at [1], the Amlogic Meson GX SoCs can embed a BL31 firmware
and a secondary BL32 firmware.
Since mid-2017, the reserved memory address of the BL31 firmware was moved
and grown for security reasons.
But mainline U-Boot and Linux has the old address and size fixed.
These SoCs have a register interface to get the two firmware reserved
memory start and sizes.
This patch adds a dynamic reservation of the memory zones in the device tree bootmem
reserved memory zone used by the kernel in early boot.
To be complete, the memory zones are also added to the EFI reserved zones.
Depends on patchset "Add support for Amlogic GXL Based SBCs" at [2].
Changes since v1:
- switched the #if to if(IS_ENABLED()) to compile all code paths
- renamed function to meson_board_add_reserved_memory()
- added a mem.h header with comment
- updated all boards ft_board_setup()
Changes since RFC v2:
- reduced preprocessor load
- kept Odroid-C2 static memory mapping as exception
Changes since RFC v1:
- switch to fdt rsv mem table and efi reserve memory
- replaced in_le32 by readl()
Reviewed-by: Simon Glass <sjg@chromium.org> Signed-off-by: Neil Armstrong <narmstrong@baylibre.com>
[trini: Fix warning on khadas-vim over missing <asm/arch/mem.h> Signed-off-by: Tom Rini <trini@konsulko.com>
Andre Przywara [Mon, 27 Nov 2017 00:47:09 +0000 (00:47 +0000)]
armv8: shrink exception table code
In the moment our exception entry code needs 34 instructions, so we
can't use put it directly into the table entry, which offers "only"
32 instructions there. Right now we just put an unconditional branch
there, then use a macro to place the 34 instructions *per entry* after
that. That effectivly doubles the size of our exception table, which
is quite a waste, given that we use it mostly for debugging purposes.
Since the register saving part is actually identical, let's just convert
that macro into a function, and "bl" into it directly from the exception
slot, of course after having saved at least the original LR.
This saves us about 950 bytes of code, which is quite a relief for some
tight SPLs, in particular the 64-bit Allwinner ones.
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Simon Glass [Sat, 25 Nov 2017 18:57:33 +0000 (11:57 -0700)]
test: compression: Convert to unit test framework
Adjust this test to use the unit test framework. Drop the two existing
commands for running the tests and replace them with a single
'ut compression' command, with sub-commands.
Signed-off-by: Simon Glass <sjg@chromium.org>
[trini: Continue to have ret = run_test_internal(...) in run_test so ret
is always initialized] Signed-off-by: Tom Rini <trini@konsulko.com>
We are planning to share more code between different NAND based
devices (SPI NAND, OneNAND and raw NANDs), but before doing that
we need to move the existing include/linux/mtd/nand.h file into
include/linux/mtd/rawnand.h so we can later create a nand.h header
containing all common structure and function prototypes.
York Sun [Sat, 18 Nov 2017 19:09:08 +0000 (11:09 -0800)]
mtd: cfi: Fix checking status register feature
Commit 72443c7f7d21 ("mtd: cfi: Add support for status register
polling") added a feature check to determine if status register
is available for certain flash chips. The "lower software bits"
register used to determine this feature is not backward compati-
ble. Older flash chips without this feature has reserved value
0xff. Instead of checking "lower software bits" register, use
CFI primary vendor-specific extended query. Since CFI version
1.4, software features can be read from offset 0x53 according to
document AN201168 from Cypress.
Signed-off-by: York Sun <york.sun@nxp.com> CC: Marek Vasut <marek.vasut@gmail.com> Tested-by: Marek Vasut <marek.vasut@gmail.com> Signed-off-by: Stefan Roese <sr@denx.de>
Simon Glass [Sat, 25 Nov 2017 18:57:32 +0000 (11:57 -0700)]
test/py: Allow any unit test suite to be found
The u-boot.sym file is scanned to find unit test suites for execution. At
present it only finds those whose names start with 'dm' or 'env'. This
code is buried in the bowels of the test code so when adding a new suite
it is not easy to discover why it is ignored by the test framework.
There seems to be no need to make this restriction. Drop it.
Signed-off-by: Simon Glass <sjg@chromium.org> Acked-by: Stephen Warren <swarren@nvidia.com>
Simon Glass [Sat, 25 Nov 2017 18:57:31 +0000 (11:57 -0700)]
test: compression: Put test variables in a struct
At present the test setup is somewhat mixed with the test itself. But if
the test setup fails (which it should not) then the test is actually
invalid. Put all the test buffers and sizes in a struct and separate out
the core code into a function.
This will make it easier to move the code to use the unit test framework.
Simon Glass [Sat, 25 Nov 2017 18:57:29 +0000 (11:57 -0700)]
test: Add a command function for test execution
The logic to either iterate through a list of tests or pick a named test
is common to at lest two test suits. Move this logic into a new function
and call it from the environment tests.
Peter Robinson [Thu, 23 Nov 2017 13:14:17 +0000 (13:14 +0000)]
omap2: set the ethaddr as well as the usbethaddr env var to ensure static MAC
The kernel gets the ethernet MAC from the ethaddr variable, the omap boards for
devices with USB based eth adapters just set the usbethaddr which doesn't appear
to get passed to the kernel. The same Raspberry Pi code sets both ethaddr and
usbethaddr so lets do that so linux (tested 4.13 and 4.14) get a static rather
than a random MAC address, while not regressing users of usbethaddr.
Signed-off-by: Peter Robinson <pbrobinson@gmail.com>
Alexander Graf [Fri, 1 Dec 2017 21:09:50 +0000 (22:09 +0100)]
efi_loader helloworld.efi: Fix building with -Os
Depending on your compiler, when compiling the hello world efi binary
with -Os, gcc might think it's a smart idea to replace common patterns
such as memory copies with explicit calls to memcpy().
While that sounds great at first, we don't have any memcpy() available
in our helloworld build target. So let's indicate to gcc that we really
do want to have the code be built as freestanding.
Fixes: bbf75dd9 ("efi_loader: output load options in helloworld") Signed-off-by: Alexander Graf <agraf@suse.de>
Alexander Graf [Fri, 1 Dec 2017 15:10:33 +0000 (16:10 +0100)]
efi_loader: Fix partition offsets
Commit 884bcf6f65 (efi_loader: use proper device-paths for partitions) tried
to introduce the el torito scheme to all partition table types: Spawn
individual disk objects for each partition on a disk.
Unfortunately, that code ended up creating partitions with offset=0 which meant
that anyone accessing these objects gets data from the raw block device instead
of the partition.
Furthermore, all the el torito logic to spawn devices for partitions was
duplicated. So let's merge the two code paths and give partition disk objects
good offsets to work from, so that payloads can actually make use of them.
Fixes: 884bcf6f65 (efi_loader: use proper device-paths for partitions) Reported-by: Yousaf Kaukab <yousaf.kaukab@suse.com> Signed-off-by: Alexander Graf <agraf@suse.de>
efi_loader: helper function to add EFI object to list
To avoid duplicate coding provide a helper function that
initializes an EFI object and adds it to the EFI object
list.
efi_exit() is the only place where we dereference a handle
to obtain a protocol interface. Add a comment to the function.
Suggested-by: Alexander Graf <agraf@suse.de> Reviewed-by: Simon Glass <sjg@chromium.org> Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de> Signed-off-by: Alexander Graf <agraf@suse.de>
We need to implement to different functions for the
EFI_DEVICE_PATH_TO_TEXT_PROTOCOL:
ConvertDeviceNodeToText
ConvertDevicePathToText
A recent patch screwed up efi_convert_device_node_to_text
to expect a device path and not a node.
The patch makes both service functions work again.
efi_convert_device_node_to_text is renamed to
efi_convert_single_device_node_to_text and
efi_convert_device_node_to_text_ext is renamed to
efi_convert_device_node_to_text to avoid future
confusion.
A test of ConvertDeviceNodeToText will be provided in
a follow-up patch.
Fixes: adae4313cdd efi_loader: flesh out device-path to text Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de> Signed-off-by: Alexander Graf <agraf@suse.de>
This used to work in v2017.09. Testing has shown, that this bug was
introduced with patch 95c5553e [efi_loader: refactor boot device and
loaded_image handling].
This patch now checks if a valid "desc" is returned from blk_get_dev()
and only continues when "desc" is available. Resulting in this cmd
output (again):
=> load usb 0:1 03000000 abc
** Bad device usb 0 **
Signed-off-by: Stefan Roese <sr@denx.de> Reviewed-by: Heinrich Schuchardt <xypron.glpk@gmx.de> Signed-off-by: Alexander Graf <agraf@suse.de>
Jonathan Gray [Wed, 22 Nov 2017 03:18:59 +0000 (14:18 +1100)]
efi_loader: initialise partition_signature memory
Zero partition_signature in the efi_device_path_hard_drive_path
structure when signature_type is 0 (no signature) as required by the
UEFI specification.
This is required so that efi_dp_match() will work as expected
when doing memcmp() comparisons. Previously uninitialised memory
would cause it not match nodes when it should have when the signature
type was not GUID.
Corrects a problem where the loaded image protocol would not return a
device path with MEDIA_DEVICE causing the OpenBSD bootloader to fail
on rpi_3 and other targets.
v2: Also handle signature_type 1 (MBR) as described in the specification
Signed-off-by: Jonathan Gray <jsg@jsg.id.au> Tested-by: Artturi Alm <artturi.alm@gmail.com> Signed-off-by: Alexander Graf <agraf@suse.de>
efi_loader: helper functions for protocol management
This patch provides helper functions to manage protocols.
efi_search_protocol - find a protocol on a handle
efi_add_protocol - install a protocol on a handle
efi_remove_protocol - remove a protocol from a handle
efi_remove_all_protocols - remove all protocols from a handle
Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de> Signed-off-by: Alexander Graf <agraf@suse.de>