Before this commit, usb_get_descriptor() failed for some flakey USB
devices. We hereby adopt the more robust linux implementation [1].
For instance, for the "Alcor Micro Corp. Flash Drive" (VID 0x058f, PID
0x6387), the following behavior occurs from time to time:
=> usb start
starting USB...
Bus xhci_pci: Register 10000840 NbrPorts 16
Starting the controller
USB XHCI 1.20
scanning bus xhci_pci for devices... usb_new_device: Cannot read configuration, skipping device 058f:6387
Signed-off-by: Philip Oberfichtner <pro@denx.de>
Reviewed-by: Marek Vasut <marex@denx.de>
[1] From a38297e3fb012 (Linux 6.9), see
https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/tree/drivers/usb/core/message.c?h=v6.9#n781
The dm_root_f field seems to be entirely write-only and hence
redundant, unless 'git grep' fails to find some access generated via
preprocessor token concatenation or similar.
Signed-off-by: Rasmus Villemoes <rasmus.villemoes@prevas.dk>
Reviewed-by: Simon Glass <sjg@chromium.org>
The dcache may not be enabled in SPL. Add a check to avoid trying to
use an undefined function.
Signed-off-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Tom Rini <trini@konsulko.com>
The default value CYCLIC_MAX_CPU_TIME_US was 1000, which is
a little bit too low for slower hardware and sandbox.
On my MIPS Boston FPGA board with interaptiv CPU, wdt_cyclic
can easily take 3200 us to run.
On azure pipeline sandbox_clang, wdt_cyclic some times goes
beyond 1300 us.
Raise default value to 5000, which is the value already taken
by octeon_nic32. This is still sufficent to maintain system
responsiveness.
Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com>
Signed-off-by: Stefan Roese <sr@denx.de>
common:
- spl: Introduce SoC specific init function
xilinx:
- Enable FF-A and NVMEM
- Rename spl_board_init() to spl_soc_init()
zynqmp:
- DT alignments
- Enable reset from SPL
- Enable USB3 for KD240
- Align multiboot register on Kria for proper reboot
- Allow multiboot environment write even in saved environment
- Move zynqmp commands from board/ to arch/
- Clean up xilinx_zynqmp.h
versal:
- Do not prioritize boot device if driver is not enabled
versal-net:
- Setup location for redundant variables in SPI
versal2:
- Add support for new SOC
mmc:
- Fix tap delay for SD on Versal NET
spi:
- Add SPI_NOR_OCTAL_READ flag for mx66uw2g345gx0 flash part
gpio:
- Cover MODEPIN firmware dependency
-----BEGIN PGP SIGNATURE-----
iF0EABECAB0WIQQbPNTMvXmYlBPRwx7KSWXLKUoMIQUCZnBfewAKCRDKSWXLKUoM
Ib/7AJ0cn3q7lMpxbSbRj33oftI8Vy8+owCeN35Nl8TN9l/DHbfusER4uVfnOVw=
=YMIB
-----END PGP SIGNATURE-----
Merge tag 'xilinx-for-v2024.10-rc1' of https://source.denx.de/u-boot/custodians/u-boot-microblaze into next
AMD/Xilinx changes for v2024.10-rc1
common:
- spl: Introduce SoC specific init function
xilinx:
- Enable FF-A and NVMEM
- Rename spl_board_init() to spl_soc_init()
zynqmp:
- DT alignments
- Enable reset from SPL
- Enable USB3 for KD240
- Align multiboot register on Kria for proper reboot
- Allow multiboot environment write even in saved environment
- Move zynqmp commands from board/ to arch/
- Clean up xilinx_zynqmp.h
versal:
- Do not prioritize boot device if driver is not enabled
versal-net:
- Setup location for redundant variables in SPI
versal2:
- Add support for new SOC
mmc:
- Fix tap delay for SD on Versal NET
spi:
- Add SPI_NOR_OCTAL_READ flag for mx66uw2g345gx0 flash part
gpio:
- Cover MODEPIN firmware dependency
Some architectures use spl_board_init() in their SoC specific
implementation. Board developers should be able to add board specific
implementation via spl_board_init(). Hence, introduce a spl_soc_init()
method which is called right before spl_board_init() for SoC
specific implementation.
Signed-off-by: Lukas Funke <lukas.funke@weidmueller.com>
Reviewed-by: Devarsh Thakkar <devarsht@ti.com>
Link: https://lore.kernel.org/r/20240327121153.2455126-2-lukas.funke-oss@weidmueller.com
Signed-off-by: Michal Simek <michal.simek@amd.com>
There are of course not a whole lot of examples in-tree yet, but
before they appear, let's make this API change: Instead of separately
allocating a 'struct cyclic_info', make the users embed such an
instance in their own structure, and make the convention that the
callback simply receives the 'struct cyclic_info *', from which the
clients can get their own data using the container_of() macro.
This has a number of advantages.
First, it means cyclic_register() simply cannot fail, simplifying the
code. The necessary storage will simply be allocated automatically
when the client's own structure is allocated (often via
uclass_priv_auto or similar).
Second, code for which CONFIG_CYCLIC is just an option can more easily
be written without #ifdefs, if we just provide an empty struct
cyclic_info {}. For example, the nested CONFIG_IS_ENABLED()s in
https://lore.kernel.org/u-boot/20240316201416.211480-1-marek.vasut+renesas@mailbox.org/
are mostly due to the existence of the 'struct cyclic_info *' member
being guarded by #ifdef CONFIG_CYCLIC.
And we do probably want to avoid the extra memory overhead of that
member when !CONFIG_CYCLIC. But that is automatic if, instead of a
'struct cyclic_info *', one simply embeds a 'struct cyclic_info',
which will have size 0 when !CONFIG_CYCLIC. Also, the no-op
cyclic_register() function can just unconditionally be called, and the
compiler will see that (1) the callback is referenced, so not emit a
warning for a maybe-unused function and (2) see that it can actually
never be reached, so not emit any code for it.
Reviewed-by: Stefan Roese <sr@denx.de>
Signed-off-by: Rasmus Villemoes <rasmus.villemoes@prevas.dk>
We are not checking the return value of strdup(), nor
freeing the string in cyclic_unregister().
However, all current users either pass a string literal or the
dev->name of the client device. So in all cases the name string will
live at least as long as the cyclic_info is registered, so just make
that a requirement.
Reviewed-by: Stefan Roese <sr@denx.de>
Signed-off-by: Rasmus Villemoes <rasmus.villemoes@prevas.dk>
As this is an SPL related driver, and in SPL enabling SPL_BLK is
optional, make this depend on the correct symbol.
Signed-off-by: Tom Rini <trini@konsulko.com>
All ARCH_K3 platforms need about of 2MB of malloc space post
reallocation. Since, this space is allocated from SDRAM, provide a
generous 2MB space by default.
Platforms requiring more than 2MB can override in defconfig as needed.
Signed-off-by: Vignesh Raghavendra <vigneshr@ti.com>
Reviewed-by: Francesco Dolcini <francesco.dolcini@toradex.com>
Marek Behún <kabel@kernel.org> says:
This series contains improvements for the 'eeprom' command:
- refactors
- fixes
- improvements
- ability to use driver model EEPROMs (uclass UCLASS_I2C_EEPROM)
- more flexible EEPROM layout support
It should not cause any behavior change for any existing board.
This series is a dependency for some DDR issue fixes for Turris Omnia.
I will be sending that one separately.
github PR link (with CI):
https://github.com/u-boot/u-boot/pull/540
- there is a failure for
test.py for sandbox sandbox_clang
but it seems unrelated to these changes
The byte variable is of type unsigned char, it is never less than zero.
The error case is handled by *endptr, so drop the comparison altogether.
Signed-off-by: Marek Behún <kabel@kernel.org>
The __eeprom_field_update_bin() function is expected to parse a hex
string into bytes (potentially in reverse order), but the
simple_strtoul() function is given 0 as base. This does not work since
the string does not contain '0x' prefix. Add explicit base 16.
Signed-off-by: Marek Behún <kabel@kernel.org>
Split the eeprom layout field finding code from the
eeprom_layout_update_field() function in order to make it usable in
alternative implementations of update method.
Signed-off-by: Marek Behún <kabel@kernel.org>
Assign the default eeprom layout parameter .data_size and methods
.print() and .update() before calling eeprom_layout_assign() in
eeprom_layout_setup().
This allows eeprom_layout_assign() to overwrite these if needed.
Signed-off-by: Marek Behún <kabel@kernel.org>
Move snprintf to stdio.h since it is needed by exteranl libraries.
Signed-off-by: Raymond Mao <raymond.mao@linaro.org>
Reviewed-by: Tom Rini <trini@konsulko.com>
Reviewed-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
When bringing in the series 'arm: dts: am62-beagleplay: Fix Beagleplay
Ethernet"' I failed to notice that b4 noticed it was based on next and
so took that as the base commit and merged that part of next to master.
This reverts commit c8ffd1356d, reversing
changes made to 2ee6f3a5f7.
Reported-by: Jonas Karlman <jonas@kwiboo.se>
Signed-off-by: Tom Rini <trini@konsulko.com>
This follows the example of RISC-V where <asm/global_data.h> includes
<asm/u-boot.h> directly as "gd" includes a reference to bd_info already
and so the first must include the second anyhow. We then remove
<asm/u-boot.h> from all of the places which include references to "gd"
an so have <asm/global_data.h> already.
Signed-off-by: Tom Rini <trini@konsulko.com>
Eugene Uriev <eugeneuriev@gmail.com> says:
There was no "mcheck" for U-Boot before.
Since U-Boot has only 1 thread, and normally makes 4000+ - 6000+
mallocs, it's better to use havier canaries to protect heap-chunks.
My variant uses 2x8 = 16byte-long protector. And
the multiplier could be changed to tune speed/protection tradeoff.
This protects not only against memset()-s, but against "near" wild
pointers too, and makes more probable to catch "distant" ones.
The core file of the set is included into the C-file, not complied
separately in order to enable (potential) coexisting of
mcheck-protectors, e.g. malloc_simple(.) and dlmalloc simultaneously.
My tests were for ARM SoC, 64bit, so the patch is
aware of alignment.
Primary this patch is for using by developers: to verify,
if a change doesn't break the heap integrity.
By default the mcheck is disabled and wouldn't affect the boot.
I used pedantic mode, canary=16byte, registry-size=6608.
For my system the overhead was 230ms.
The using of pre-reloc/malloc_simple heap is
too hard to follow after the relocation.
So lets drop it from the pedantic registry
and switch to dlmalloc, when moved.
The offset is ignored, but kept in the API
for the probable case, when that early
heap is relocated too.
Signed-off-by: Eugene Uriev <eugeneuriev@gmail.com>
The pedantic mode is run-time contolled, so
appropriate registry take place everytime.
Maybe it's worth to use compile-time control only.
So, the registry could be optimized out by an #ifdef.
Signed-off-by: Eugene Uriev <eugeneuriev@gmail.com>
This changes are probable worth to be generalized in a
separate .h-file so, making it able to cover
libc-mallocs and others, without too much copy-paste.
But the malloc<=>mALLOc substitutions interfere
with an elegant way to do this.
Signed-off-by: Eugene Uriev <eugeneuriev@gmail.com>
These fast helpers sometimes breach mem-chunk boundaries.
Thus they trigger mcheck alarm. Standard ones are accurate though.
Signed-off-by: Eugene Uriev <eugeneuriev@gmail.com>
We may enter the command line interface in a state where on the remote
console the cursor is not shown. Send an escape sequence to enable it.
Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
Those keyboards do not return the current device state. Polling will
timeout unless there are key presses. This is not a problem during
operation but the initial device state query during probing will fail.
Skip this step in usb_kbd_probe_dev() to make these devices useable.
Not all Apple keyboards behave like this. A keyboard with USB
vendor/product ID 05ac:0221 is reported to work with the current code.
Unfortunately some Keychron keyboards "re-use" Apple's vendor ID and
show the same behavior (Keychron C2, 05ac:024f for example).
Reviewed-by: Marek Vasut <marex@denx.de>
Reviewed-by: Neal Gompa <neal@gompa.dev>
Signed-off-by: Janne Grunau <j@jannau.net>
Add the environment variable "usb_ignorelist" to prevent USB devices
listed in it from being bound to drivers. This allows to ignore devices
which are undesirable or trigger bugs in u-boot's USB stack.
Devices emulating keyboards are one example of undesirable devices as
u-boot currently supports only a single USB keyboard device. Most
commonly, people run into this with Yubikeys, so let's ignore those in
the default environment.
Based on previous USB keyboard specific patches for the same purpose.
Link: https://lore.kernel.org/u-boot/7ab604fb-0fec-4f5e-8708-7a3a7e2cb568@denx.de/
Reviewed-by: Neal Gompa <neal@gompa.dev>
Reviewed-by: Marek Vasut <marex@denx.de>
Signed-off-by: Janne Grunau <j@jannau.net>
When Falcon Mode is enabled on RISC-V, use CONFIG_VAL
to check PAYLOAD_ARGS_ADDR, not CONFIG_IS_ENABLED.
Fixes: 10c4ab898c ("spl: riscv: falcon: move fdt blob to specified address")
Signed-off-by: Randolph <randolph@andestech.com>
Tested-by: Leo Yu-Chi Liang <ycliang@andestech.com>
Add a test for the button_cmd feature. This validates that commands can
be mapped to two buttons, that the correct command runs based on which
button is pressed, that only 1 command is run, and that no command runs
if button_cmd_0_name is wrong or unset.
Additionally, fix a potential uninitialised variable use caught by these
tests, the btn variable in get_button_cmd() is assumed to be null if
button_get_by_label() fails, but it's actually used uninitialised in
that case.
CONFIG_BUTTON is now enabled automatically and was removed when running
save_defconfig.
Fixes: e761035b64 ("boot: add support for button commands")
Signed-off-by: Caleb Connolly <caleb.connolly@linaro.org>
There could be two memory allocation scheme in SPL phase.
Explicitly print the corresponding error message.
Signed-off-by: Leo Yu-Chi Liang <ycliang@andestech.com>
This commit allows users to choose the appropriate memory
allocation method between static allocated and dynamically
calloc. The previous static-array way will not obviously
contribute to the final binary size since it is uninitialized,
and might have better performance than the dynamical one.
Now we provide the users with both the two options.
Signed-off-by: Hanyuan Zhao <hanyuan-z@qq.com>
This commit simply modifies the history initialize function,
replacing the return value by panic with reasons. The calling
chains of hist_init don't have steps explicitly throwing or
dealing with the ENOMEM error, and once the init fails, the
whole system is died. Using panic here to provide error
information instead.
Signed-off-by: Hanyuan Zhao <hanyuan-z@qq.com>
Drop old implementation and use hash_command() instead, as
how it's currently done for crc32 and sha1sum cmds.
Test:
=> md5sum 0x60000000 0x200
md5 for 60000000 ... 600001ff ==> e6bbbe95f5b41996f4a9b9af7bbd4050
Signed-off-by: Igor Opaniuk <igor.opaniuk@gmail.com>
Return value of a function 'malloc_cache_aligned'
is dereferenced at autoboot.c:207 without checking for NULL,
but it is usually checked for this function.
Found by RASU JSC.
Signed-off-by: Maks Mishin <maks.mishinFZ@gmail.com>
When `gd` is relocated during `spl_relocate_stack_gd()` the
doubly-linked circular list in the `log_head` member is broken.
The last element of the list should point back to the initial
`list_head`, but as the initial `list_head` is moved the pointer becomes
stale. As a result the loop in `log_dispatch` would never finish.
Signed-off-by: Thomas Weißschuh <thomas.weissschuh@linutronix.de>
ARCH_NEXELL: Change calculation of monitor length (gd->mon_len) to fix
relocation at boards with s5p4418-SoC (ARCH_NEXELL). At s5p4418, _start
is after the header (NSIH). Therefore the monitor length has to be
calculated using __image_copy_start instead of _start in order the
whole monitor code is relocated.
Signed-off-by: Stefan Bosch <stefan_b@posteo.net>
