Marek Vasut [Wed, 4 Oct 2017 01:03:50 +0000 (03:03 +0200)]
ARM: rmobile: Fixup ULCB CPLD support after PFC rework
The ULCB CPLD support was not updated during the PFC table rework,
fix up the GPIO numbers until the CPLD support is rewritten to a
proper DM capable and DT probing driver.
York Sun [Thu, 28 Sep 2017 15:42:16 +0000 (08:42 -0700)]
armv8: ls1043ardb_sdcard: prepare falcon boot
Due to a conflict with recent Primary Protected Application (PPA),
PPA cannot be loaded for SPL stage, falcon boot is not enabled by
default. With compatible PPA image, to enable falcon boot, activate
these Kconfig options in defconfig
CONFIG_SPL_FIT=y
CONFIG_SPL_FSL_LS_PPA=y
CONFIG_SPL_GZIP=y
CONFIG_SPL_LOAD_FIT=y
CONFIG_SPL_OS_BOOT=y
CONFIG_SPL_OF_LIBFDT=y
Because environment variables are not avaiable during SPL stage for
SD boot, set "boot_os=y" as default.
York Sun [Thu, 28 Sep 2017 15:42:12 +0000 (08:42 -0700)]
armv8: fsl-layerscape: Avoid running dram_init_banksize again
gd->ram_size is reduced in this function to reserve secure memory.
Avoid running this function again to further reduce memory size.
This fixes issue for SPL boot with PPA image loaded in which case
secure memory is incorrectly allocated due to repeated calling.
York Sun [Thu, 28 Sep 2017 15:42:11 +0000 (08:42 -0700)]
cmd: spl: fix compiling error when CONFIG_CMD_SPL_WRITE_SIZE not defined
CONFIG_CMD_SPL_WRITE_SIZE is used for writing parameters to non-volatile
storage. So far it is only used for NAND. Fix compiling error when this
macro is not used for SD.
York Sun [Thu, 28 Sep 2017 15:42:10 +0000 (08:42 -0700)]
spl: fix assignment of board info to global data
Commit 15eb1d43bf47 ("spl: reorder the assignment of board info to
global data") intended to move assignment of board info earlier,
into board_init_r(). However, function preload_console_init() is
called either from spl_board_init() or from board_init_f(). For the
latter case, the board info assignment is much earlier than proposed
board_init_r(). Create a new function to fill gd->bd and call this
function when needed.
The MC boot sequence is contained in mc_env_boot. Update LS1088A
boards to use this function, and hook it to reset_phy so that it's
called late enough, after the ports have been initialized, for
proper DPC / DPL fixup.
Sumit Garg [Tue, 3 Oct 2017 21:50:49 +0000 (03:20 +0530)]
armv8: fsl-layerscape: Allocate Secure memory from first ddr region
This change is required due to trusted OS (OP-TEE) not being position
independent code, it requires compile time fixed base address.
To take care of this it is assumed that all layerscape armv8 platforms
has minimum 2G ddr in first region. So we can have fixed address
space (66 MB -> (2MB for PPA + 64MB for trusted OS)) allocated from
top of first 2G ddr region and compile trusted OS with this fixed
base address.
But one exception here is ls1012 where we have only 1G (rdb) or 512M
(frdm) ddr memory. For those we can have different fixed compile time
base addresses for trusted OS.
Gong Qianyu [Mon, 18 Sep 2017 08:59:28 +0000 (16:59 +0800)]
armv8: ls1046aqds: Fix NAND offset for Fman ucode and env
Fix a bug of 'commit 8104deb2d6b7 ("armv8: layerscape: Adjust memory
mapping for Flash/SD card on LS1046A")' as NAND block size is 256KB
on LS1046AQDS.
Rob Clark [Sun, 8 Oct 2017 15:33:08 +0000 (11:33 -0400)]
efi_loader: Fix disk dp's for pre-DM/legacy devices
This fixes an issue with OpenBSD's bootloader, and I think should also
fix a similar issue with grub2 on legacy devices. In the legacy case
we were creating disk objects for the partitions, but not also the
parent device.
efi_selftest: avoid dereferencing NULL in tpl test
The task priority levels test uses two events one passes the
notification counter as context. The other passes NULL.
Both use the same notification function. So we need to check
for NULL here.
efi_selftest: efi_st_memcmp return difference of bytes
If the memory regions are different efi_st_memcmp currently
returns the difference of the addresses. Insted the
difference of the first differing byte pair should be
returned.
This patch provides an EFI application to check the correct function
of the Simple Network Protocol implementation.
It sends a DHCP request and analyzes the DHCP offer.
Different error conditions including a 10s timeout are checked.
A successful execution will look like this:
=> bootefi nettest
Scanning disk ide.blk#0...
Found 1 disks
WARNING: Invalid device tree, expect boot to fail
Network test
DHCP Discover
DHCP reply received from 192.168.76.2 (52:55:c0:a8:4c:02)
as broadcast message.
OK. The test was completed successfully.
The WaitForPacket event informs that a network package has been
received by the SimpleNetworkProtocol.
Signed-off-by: Heinrich Schuchardt <[email protected]>
[agraf: Move is_signaled = true line into efi_net_push()] Signed-off-by: Alexander Graf <[email protected]>
U-Boot does not implement all functions of the simple network
protocol. The unimplemented functions return either of
EFI_SUCCESS and EFI_INVALID_PARAMETER.
The UEFI spec foresees to return EFI_UNSUPPORTED in these cases.
The UEFI spec defines parameter index of WaitForEvent as UINTN*.
So we should use size_t here.
I deliberately do not use UINTN because I hold a following patch
that will eliminate UINTN because uppercase types to not match
the U-Boot coding style.
Rename counter to more illustrative names.
Update notification function description.
Simplify notification function.
Add comment for arbitrary non-zero value.
Document @return.
Use constants for return values of setup, execute, teardown.
efi_selftest: use efi_st_error for all error messages
All error messages in the selftests should use efi_st_error.
efi_st_error will print the file name and line number of the error.
Splitting message texts due to lines being over 80
characters is avoided. This resolves the issue reported
by Simon Glass in
https://lists.denx.de/pipermail/u-boot/2017-September/307387.html
Klaus Goger [Wed, 20 Sep 2017 11:50:41 +0000 (13:50 +0200)]
dm: ofnode: query correct property in livetree ofnode_get_addr_size
The livetree codepath of ofnode_get_addr_size always used the "reg"
property for of_get_property. Use the property parameter of the function
call instead and check the return value if the property exists.
Otherwise return FDT_ADDR_T_NONE.
This was discoverd while using SPI NOR with livetree.
spi_flash_decode_fdt checks for memory-map and will not fail with
livetree even if the property does not exist.
Simon Glass [Sun, 17 Sep 2017 22:54:53 +0000 (16:54 -0600)]
dm: gpio: Correct use of -ENODEV in drivers
In U-Boot -ENODEV means that there is no device. When there is a problem
with the device, drivers should return an error like -ENXIO or -EREMOTEIO.
When the device tree properties cannot be read correct , they should
return -EINVAL.
Update various GPIO drivers to follow this rule, to help with consistency
for future driver writers.
Simon Glass [Sun, 17 Sep 2017 22:54:52 +0000 (16:54 -0600)]
dm: gpio: Add a comment about not copying some drivers
These three drivers all use U_BOOT_DEVICE rather than device tree to
create devices, so have to do manual allocation of platform data. This is
not true for new platforms.
Add a more explicit comment so that people do not copy this approach with
new boards.
Simon Glass [Sun, 17 Sep 2017 22:54:51 +0000 (16:54 -0600)]
dm: gpio: vybrid_gpio: Correct driver's use of bind() method
It does not look like this driver needs to use a bind() method. It does
not manually create devices with device_bind() nor does it create devices
using U_BOOT_DEVICE(). It seems to only use device tree.
Therefore the manual allocation of platform data is not needed and is
confusing. Also platform data should be set up by the ofdata_to_platdata()
method, not bind().
Update the driver in case others use it as a model in future.
Patrice Chotard [Tue, 3 Oct 2017 13:47:59 +0000 (15:47 +0200)]
configs: stm32: update SRAM address for STM32H7 disco and eval
As suggested by Vikas Manocha, update embedded SRAM address
to use AXI SRAM available on D1 domain instead of using
AHB SRAM (D2 domain).
On some STM32H743 SoCs, D2 domain SRAM is accessible even if
SRAMxEN bit in AHB2ENR bits are not set whereas on others SoCs
version it's not accessible.
Bo Shen [Tue, 3 Oct 2017 05:47:59 +0000 (22:47 -0700)]
ARM: stm32f7: fix prescaler calculation of timer
As the timer 2 is on APB1 bus, the maximum of clock frequency of APB1 timer
clock is half of SYSCLK. Then to calculate the timer prescaler for timer 2
which need to be divided by 2.
Tuomas Tynkkynen [Sat, 30 Sep 2017 23:25:21 +0000 (02:25 +0300)]
fs/fat: Fix 'CACHE: Misaligned operation at range' warnings
The 'block' field of fat_itr needs to be properly aligned for DMA and
while it does have '__aligned(ARCH_DMA_MINALIGN)', the fat_itr structure
itself needs to be properly aligned as well.
While at it use malloc_cache_aligned() for the other aligned allocations
in the file as well.
Fixes: 2460098cffacd1 ("fs/fat: Reduce stack usage") Signed-off-by: Tuomas Tynkkynen <[email protected]> Reviewed-by: Tom Rini <[email protected]>
Tom Rini [Fri, 29 Sep 2017 18:32:44 +0000 (14:32 -0400)]
cmd: Make CMD_LZMA / CMD_UNZIP default y if CMD_BOOTI
In the Linux Kernel on ARM64, the Image.COMPRESSION file is not
self-extracting in the way that x86 and ARM images are. So when
CMD_BOOTI is enabled we should also default to enabling CMD_UNZIP and
CONFIG_LZMA in order for the user to be able to decompress many of the
common compressions that will be done to an Image file.
Sam Protsenko [Thu, 28 Sep 2017 19:33:45 +0000 (12:33 -0700)]
scripts: Move Kconfig contents to cmd/Kconfig
On case-insensitive file systems we have collision between
scripts/kconfig/ directory and scripts/Kconfig file. This patch moves
scripts/Kcofnig contents to cmd/Kconfig to fix that.
serial: stm32x7: remove useless CONFIG_CLK and OF_CONTROL flag
This driver is currently used by STM32F7 and STM32H7 SoCs.
As CONFIG_CLK and OF_CONTROL flags are set by default for these
2 SoCs, this flag becomes useless in this driver, so remove it.
relocate_code() allocates 32 bytes stack frame but only 16 bytes are
freed before return. it will cause errors to possible previous frames
and doesn't make relocate_code() look like a function.
ARMv8: get new GD address from gd->new_gd directly
the new GD address is calculated via board data BD currently
it require the new GD area locates below BD tightly, so a strict
constraint is imposed on memory layout which maybe make special
platform unpleasant.
fix it by getting new GD address from gd->new_gd directly.
Tom Rini [Sat, 7 Oct 2017 15:27:59 +0000 (11:27 -0400)]
cmd/gpt.c, cmd/nvedit.c, tools/fit_image.c: Rework recent fixes for Coverity
The recent changes to these files did not completely fix the previous
issues, or introduced different (minor) issues. In cmd/gpt.c we need to
dereference str_disk_guid to be sure that malloc worked. In
cmd/nvedit.c we need to be careful that we can also fit in that leading
space when adding to the string. And in tools/fit_image.c we need to
re-work the error handling slightly in fit_import_data() so that we only
call munmap() once. We have two error paths here, one where we have an
fd to close and one where we do not. Adjust labels to match this.
Andy Shevchenko [Tue, 3 Oct 2017 11:55:08 +0000 (14:55 +0300)]
x86: edison: Bring minimal ACPI support to the board
This board is based on Intel Tangier SoC (Intel Merrifield platform)
and may utilize ACPI powerfulness.
Bring minimum support by appending initial DSDT table for it.
Note, the addresses for generated tables are carefully chosen to avoid
any conflicts with existing shadowed BIOS data. The user have somewhat
like ~31 kB available for compiled ACPI tables that ought to be enough.
Andy Shevchenko [Tue, 3 Oct 2017 11:55:07 +0000 (14:55 +0300)]
x86: tangier: Enable ACPI support for Intel Tangier
Intel Tangier SoC is a part of Intel Merrifield platform which doesn't
utilize ACPI by default. Here is an attempt to unleash ACPI flexibility
power on Intel Merrifield based platforms.
The change brings minimum support of the devices that found on
Intel Merrifield based end user device.
Tom Rini [Wed, 27 Sep 2017 02:14:44 +0000 (22:14 -0400)]
tools/fit_image.c: Update some return code paths
Coverity has found some problems with the return paths in parts of this
code. We have a case where we were going to the wrong part of the
unwind (open() failed so we cannot close the fd), a case where we were
only free()ing our buf on the error path and finally a case where we did
not munmap in the failure path.
Tom Rini [Wed, 27 Sep 2017 00:44:32 +0000 (20:44 -0400)]
cmd/pxe.c: Rework initrd and bootargs handling slightly
For the initrd portion of handling our bootm arguments we do not have a
sufficiently long enough buffer for some improbable 64bit cases. Expand
this buffer to allow for a 64bit address and almost 256MB initrd to be
used. Make use of strncpy/strncat when constructing the values here
since we know what the worst case valid values are, length wise.
Similarly for bootargs themselves, we need to make use of strlen/sizeof
and strncpy/strncat to ensure that we don't overflow bootargs itself.
Tom Rini [Tue, 26 Sep 2017 23:37:11 +0000 (19:37 -0400)]
cmd/nvedit.c: Update input handling to cover overflow cases
When we have multiple messages provided, we need to be sure that we do
not exceed the length of our 'message' buffer. In the for loop, make
sure that pos is not larger than message. Only copy in at most however
much of the message buffer remains. Finally, if we have not reached the
end of the message buffer, put in a space and NULL, and if we have,
ensure the buffer is now NULL termined.
Reported-by: Coverity (CID: 165116) Signed-off-by: Tom Rini <[email protected]>
Tom Rini [Tue, 26 Sep 2017 16:43:12 +0000 (12:43 -0400)]
test/overlay: Fix various malloc/free leaks
With the overlay tests now being built in sandbox Coverity has found a
number of issues in the tests. In short, if malloc ever failed we would
leak the previous mallocs, so we need to do the usual goto pattern to
free each in turn. Finally, we always looked at the free()d location to
see how many tests had failed for the return code.
Create a common exit for most of the error handling code in
do_rename_gpt_parts. Delete the list elements in disk_partitions
before calling INIT_LIST_HEAD from get_gpt_info() a second time.
The SIZEOF_MISMATCH error is not addressed, since that problem was
already fixed by "GPT: incomplete initialization in
allocate_disk_part".
Tuomas Tynkkynen [Mon, 25 Sep 2017 19:06:34 +0000 (22:06 +0300)]
fs-test: Add test for a filename using '..' to go back to the root
The previous commit fixed a problem in FAT code where going back to the
root directory using '..' wouldn't work correctly on FAT12 or FAT16.
Add a test to exercise this case (which was once fixed in commit 18a10d46f26 "fat: handle paths that include ../" but reintroduced due to
the directory iterator refactoring).
This test only very barely catches the problem - without the fix the
size command still gives valid output but the additional spurious
"Invalid FAT entry" error message makes it not get caught in the
'egrep -A3 ' output. I tried to make a proper test that grows the root
directory to two clusters lots of with dummy files but that causes the
write tests to crash the sandbox totally...
Tuomas Tynkkynen [Mon, 25 Sep 2017 19:06:33 +0000 (22:06 +0300)]
fs/fat: Fix pathnames using '..' that lead to the root directory
If we end up back in the root directory via a '..' directory entry, set
itr->is_root accordingly. Failing to do that gives spews like
"Invalid FAT entry" and being unable to access directory entries located
past the first cluster of the root directory.
Fixes: 8eafae209c35 ("fat/fs: convert to directory iterators") Reviewed-by: Tom Rini <[email protected]> Signed-off-by: Tuomas Tynkkynen <[email protected]>
Tuomas Tynkkynen [Mon, 25 Sep 2017 19:06:32 +0000 (22:06 +0300)]
fs-test: Add FAT16 support
Currently we can only test FAT32 which is the default FAT version that
mkfs.vfat creates by default. Instead make it explicitly create either a
FAT16 or a FAT32 volume. This allows us to exercise more code, for
instance the root directory handling is done differently in FAT32 than
the older FATs.
Adding FAT12 support is a much bigger job since the test creates a 2.5GB
file and the FAT12 maximum partition size is way smaller than that.
Tuomas Tynkkynen [Mon, 25 Sep 2017 19:06:31 +0000 (22:06 +0300)]
fs/ext4: Fix group descriptor checksum calculation
The current code doesn't compute the group descriptor checksum correctly
for the filesystems that e2fsprogs 1.43.4 creates (they have
'Group descriptor size: 64' as reported by tune2fs). Extend the checksum
calculation to be done as ext4_group_desc_csum() does in Linux.
This fixes these errors in dmesg from running fs-test.sh and makes it
succeed again:
[1671902.620699] EXT4-fs (loop1): ext4_check_descriptors: Checksum for group 0 failed (35782!=10965)
[1671902.620706] EXT4-fs (loop1): group descriptors corrupted!
Felix Brack [Fri, 22 Sep 2017 12:27:28 +0000 (14:27 +0200)]
doc: Fix statements on case sensitivity for regulator prefix
The regulator bindings state that regulator prefixes are allowd to be
in upper or lower case. However pmic_bind_children from pmic_uclass uses
strncmp to compare DT node name against prefix. This comparison is case
sensitive hence the regulator driver prefix case matters.
Sam Protsenko [Thu, 21 Sep 2017 22:51:59 +0000 (01:51 +0300)]
omap: Fix warning when looking for userdata part
When eMMC was formattaed for Linux partition table, "userdata" partition
is missing. In this case, part_get_info_by_name() iterates over all
registered drivers (which are PART_TYPE_EFI, PART_TYPE_DOS and
PART_TYPE_ISO). And when it comes to PART_TYPE_ISO (which has empty
partition table), we can see next warning in U-Boot output:
** First descriptor is NOT a primary desc on 1:1 **
This patch switches to part_get_info_by_name_type() API in order to
check only EFI partitions for "userdata" partitions. This eliminates
mentioned warning.
projects / u-boot.git / log