The bootefi command creates a copy of the device-tree within the first
127 MiB of memory. This may lead to overwriting previously loaded binaries
(e.g. kernel, initrd).
Linux EFI stub itself copies U-Boot's copy of the device-tree. This means
there is not restriction for U-Boot to place the device-tree copy to any
address. (Restrictions existed for 32bit ARM before Linux commit
7a1be318f579 ("ARM: 9012/1: move device tree mapping out of linear region")
for legacy booting.
Reported-by: Alexandre Ghiti <alexghiti@rivosinc.com>
Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
Tested-by: Alexandre Ghiti <alexghiti@rivosinc.com>
Add an option to package a kernel into the debugging script used for
EFI.
The name of the kernel must be added to the script. By default it is
assumed that the kernel is built in the /tmp/kernel directory.
Signed-off-by: Simon Glass <sjg@chromium.org>
Add a command (for the app and payload) to display the tables provided
by EFI. Note that for the payload the tables should always be present, so
an error message is unnecessary and would bloat the code.
Signed-off-by: Simon Glass <sjg@chromium.org>
This code is used with EFI_LOADER but is also useful (with some
modifications) for the EFI app and payload. Move it into a shared
file.
Show the address of the table so it can be examined if needed. Also show
the table name as unknown if necessary. Our list of GUIDs is fairly
small.
Signed-off-by: Simon Glass <sjg@chromium.org>
These are currently only available when running with EFI_LOADER.
Expand this to include the app and payload, since it is useful to be
able to decode things there.
Signed-off-by: Simon Glass <sjg@chromium.org>
Provide information about the GUIDs supplied by QEMU, so far as it is
known.
These values are used in the 'efi table' command as well as the printf
format string %sU
Signed-off-by: Simon Glass <sjg@chromium.org>
These have been disabled due to the rudimentary support available. It is
a little better now, so enable these options.
Signed-off-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Heinrich Schuchardt <xypron.glpk@gmx.de>
When running the EFI app, we need to exit boot services before jumping
to Linux.
At some point it may be possible to jump to Linux and pass on the system
table, and:
* install the device-tree as configuration table
* use LoadImage() to load the kernel image (e.g. from memory)
* start the image with StartImage()
This should allow the Linux efistub to be used. For now, this is not
implemented.
Signed-off-by: Simon Glass <sjg@chromium.org>
Add the missing code to handle this. For a 64-bit kernel the entry
address is 0x200 bytes after the normal entry.
Rename the parameter to boot_linux_kernel() accordingly. Update the
comments to indicate that these are addresses, not pointers.
Signed-off-by: Simon Glass <sjg@chromium.org>
The U-Boot EFI app can run as a 64-bit program, so set the Kconfig
correctly in that case. Make sure it doesn't build SPL, since there is
no need to switch from 32 to 64 bit when running.
Signed-off-by: Simon Glass <sjg@chromium.org>
The incumbent function efi_alloc() is unused.
Replace dp_alloc() by a new function efi_alloc() that we can use more
widely.
Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
Reviewed-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
EFI device paths for block devices must be unique. If a non-unique device
path is discovered, probing of the block device fails.
Currently we use UsbClass() device path nodes. As multiple devices may
have the same vendor and product id these are non-unique. Instead we
should use Usb() device path nodes. They include the USB port on the
parent hub. Hence they are unique.
A USB storage device may contain multiple logical units. These can be
modeled as Ctrl() nodes.
Reported-by: Patrick Delaunay <patrick.delaunay@foss.st.com>
Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
* Add the definitions for Ctrl() device path nodes.
* Implement Ctrl() nodes in the device path to text protocol.
Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
EFI device paths for block devices must be unique. If a non-unique device
path is discovered, probing of the block device fails.
Currently we use UsbClass() device path nodes. As multiple devices may
have the same vendor and product id these are non-unique. Instead we
should use Usb() device path nodes. They include the USB port on the
parent hub. Hence they are unique.
A USB storage device may contain multiple logical units. These can be
modeled as Ctrl() nodes.
Reported-by: Patrick Delaunay <patrick.delaunay@foss.st.com>
Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
* Add the definitions for Ctrl() device path nodes.
* Implement Ctrl() nodes in the device path to text protocol.
Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
- Fixes for booting newer revs of the SoC in the Raspberry Pi 4
- Propagate some firmware DT properties to the loaded DT
- Update the Zero2W upstream DT name
Newer firmware can manage the SDCDIV clock divisor register, allowing
the divisor to scale with the core as necessary.
Leverage this ability if the firmware supports it.
Adapted from the following raspberrypi Linux kernel commit:
bcm2835-sdhost: Firmware manages the clock divisor
08532d242d
Signed-off-by: Vincent Fazio <vfazio@xes-inc.com>
Signed-off-by: Peter Robinson <pbrobinson@gmail.com>
In rpi-firmware 25e2b597ebfb2495eab4816a276758dcc6ea21f1,
the GET_CLOCK_RATE mailbox property was changed to return the last
value set by SET_CLOCK_RATE.
https://github.com/raspberrypi/firmware/issues/1619#issuecomment-917025502
Due to this change in firmware behavior, bcm2835_get_mmc_clock now
returns a clock rate of zero since we do not issue SET_CLOCK_RATE.
This results in degraded MMC performance.
SET_CLOCK_RATE fixes the clock to a specific value and disables scaling
so is not an ideal solution.
Instead, fallback to GET_MAX_CLOCK_RATE in bcm2835_get_mmc_clock if
GET_CLOCK_RATE returns zero.
Signed-off-by: Vincent Fazio <vfazio@xes-inc.com>
Signed-off-by: Peter Robinson <pbrobinson@gmail.com>
A38x BootROM has a bug which cause that BootROM loads data part of UART
image into RAM target address increased by one byte when source address
and header size stored in the image header are not same.
Workaround this bug by completely removing a gap between header and data
part of the UART image. Without gap, this BootROM bug is not triggered.
This gap can be present in SDIO or SATA image types which have aligned
start of the data part to the media sector size. With this workaround
kwboot should be able to convert and send SDIO or SATA images for UART
booting.
Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
Sending of very small images (smaller than 128 bytes = xmodem block size)
cause out-of-bound memory read access. Fix this issue by ensuring that
hdrsz when sending image is not larger than total size of the image.
Issue was introduced in commit f8017c3779 ("tools: kwboot: Fix sending
Kirkwood v0 images"). Special case when total image is smaller than header
size aligned to multiply of xmodem size is already handled since that
commit.
Fixes: f8017c3779 ("tools: kwboot: Fix sending Kirkwood v0 images")
Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
Commit 7665ed2fa0 ("tools: kwboot: Fix parsing UART image without data
checksum") added fixup code to insert place for data checksum if UART image
does not have it. Together with option -B (change baudrate), kwboot
calculates this checksum. Without option -B, it inserts only place for
checksum but does not calculate it.
This commit fix above logic and calculate data checksum also when kwboot is
used without -B option.
Fixes: 7665ed2fa0 ("tools: kwboot: Fix parsing UART image without data checksum")
Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
Ensure that UART aligned header size is always stored into kwbimage v1
header. It is needed for proper UART booting. Calculation of headersz field
was broken in commit d656f5a0ee ("tools: kwboot: Calculate real used
space in kwbimage header when calling kwboot_img_grow_hdr()") which
introduced optimization of kwboot_img_grow_hdr() function.
Fixes: d656f5a0ee ("tools: kwboot: Calculate real used space in kwbimage header when calling kwboot_img_grow_hdr()")
Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
Armada 385 BootROM ignores low 7 bits of headersz when parsing kwbimage
header of UART type, which effectively means that headersz is rounded down
to multiply of 128 bytes. For all other image types BootROM reads and use
all bits of headersz. Therefore fill into UART type of kwbimage v1 headersz
aligned to 128 bytes.
Fixes: 2b0980c240 ("tools: kwbimage: Fill the real header size into the main header")
Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
It's useful to enable Debug UART for future DM Serial regression tests
for Kirkwood boards.
Also, see background discussion in this thread:
https://lists.denx.de/pipermail/u-boot/2023-March/512010.html
Signed-off-by: Tony Dinh <mibodhi@gmail.com>
Reviewed-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
Same change as was done for mvebu in commit 5bb2c550b1 ("arm: mvebu: Move
internal registers in arch_very_early_init() function") but for kirkwood.
Signed-off-by: Pali Rohár <pali@kernel.org>
Tested-by: Tony Dinh <mibodhi@gmail.com>
Reviewed-by: Stefan Roese <sr@denx.de>
CONFIG_SF_DEFAULT_SPEED is used in SPL SPI to configure and probe the
flash device during DM SPI uclass probing process, if the
spi-max-frequency is not available in the DTB. Currently the max
frequency is not available, because of the probing mechanism in SPI
uclass has not been fully updated to DM.
The CONFIG_SF_DEFAULT_SPEED is set to 1Mhz if a board defconfig
does not specify it. This speed is too slow and result in a few
seconds delay while the u-boot image is loaded from flash. Based on a
survey of the device tree specifications for MVEBU boards, a sane default
value should be 10Mhz. The default of 10Mhz enables an almost
instantaneously loading of the u-boot image.
Note that this patch depends on this patch series (has been merged to
u-boot-marvell/next):
https://lists.denx.de/pipermail/u-boot/2023-March/511038.html
- RESEND: correct spelling of SF_DEFAULT_MODE
Signed-off-by: Tony Dinh <mibodhi@gmail.com>
Reviewed-by: Stefan Roese <sr@denx.de>
Here we implement usb_gadget_handle_interrupts() but did not include
<linux/usb/gadget.h> so did not have the declaration correct. Fix this
and add the missing include.
Signed-off-by: Tom Rini <trini@konsulko.com>
If we're building U-Boot with LTO, we don't want to use that for
examples as it's more work than required.
Signed-off-by: Tom Rini <trini@konsulko.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
We did not add a prototype for spl_mmc_get_uboot_raw_sector to
include/spl.h before, so add and document one now. Correct the incorrect
prototype in board/advantech/imx8mp_rsb3720a1/imx8mp_rsb3720a1.c and
ensure that we have spl.h where we define a non-weak
spl_mmc_get_uboot_raw_sector as well.
Signed-off-by: Tom Rini <trini@konsulko.com>
Some platforms were not including <cpu_func.h> which sets the prototype
for reset_cpu, and in turn had it set wrong. Correct these cases.
Signed-off-by: Tom Rini <trini@konsulko.com>
With clang-15 we now get reported that in the make_flame_tree function,
neither the missing_count nor depth variables are used, only
incremenete/decremented. Remove these.
Signed-off-by: Tom Rini <trini@konsulko.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
With clang-15, it is now reported that cont_img_count is unused. This is
true as the code will increment / reset this counter, but never
functionally use it. Remove it.
Signed-off-by: Tom Rini <trini@konsulko.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Fabio Estevam <festevam@denx.de>
Reviewed-by: Peng Fan <peng.fan@nxp.com>
With clang-15 we now will get warnings such as:
warning: a function declaration without a prototype is deprecated in all
versions of C [-Wstrict-prototypes]
And it is easy enough to address this warning here, even if we would
like to stay in sync more with upstream as it's a single location.
Signed-off-by: Tom Rini <trini@konsulko.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
With clang-15 we now will get warnings such as:
warning: a function declaration without a prototype is deprecated in all
versions of C [-Wstrict-prototypes]
And it is easy enough to address this warning here, as we aren't
concerned with re-syncing with an upstream.
Signed-off-by: Tom Rini <trini@konsulko.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
With clang-15 we now will get warnings such as:
warning: a function declaration without a prototype is deprecated in all
versions of C [-Wstrict-prototypes]
And it is easy enough to address this warning here, as we aren't
concerned with re-syncing with an upstream.
Signed-off-by: Tom Rini <trini@konsulko.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
We have a number of places in the code which use the following syntax:
void func(a, b, c)
int a; /* Does a */
something_t *b; /* Pointer to b */
int c; /* Does c */
{
...
}
Which while not what we document as our coding style, this is also code
which we have imported from other projects, and would like to re-sync
with in the future. While the biggest example of this is the zlib code,
there are other places as well. For now, we will silence this warning.
Signed-off-by: Tom Rini <trini@konsulko.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
- Assorted TI platform fixes, correct location of NXP boot format git
repository, don't try and mount partitions that are too small to be
ext4 as ext4, handle .bin files in .gitattributes, flush out panic
messages for sure, and correct console location on Arm total_compute.
Setting RX flow error handling will stall the channel until descriptors
are available to move RX data. Setting this bit causes issues when
tearing down ethernet DMA channel at the end of TFTP transfer as
unrelated network packets can cause teardown to stall indefinitely waiting
for driver to queue add more desc leading to channel hang with error
logs:
udma_stop_dev2mem TIMEOUT !
udma_stop_dev2mem: peer not stopped TIMEOUT !
udma_stop_dev2mem TIMEOUT !
Fix this by clearing rx_error_handling similar to how its done for UDMA
as part of udma_alloc_rchan_sci_req()
This fixes occasional TFTP Failures seen when downloading multiple files
one after the other on AM64/AM62 SoCs.
Fixes: 9a92851c33 ("dma: ti: k3-udma: Add BCDMA and PKTDMA support")
Reported-by: Nishanth Menon <nm@ti.com>
Signed-off-by: Vignesh Raghavendra <vigneshr@ti.com>
No need to mount a too small partition to handle a EXT4 file system.
This patch add a test on partition size before to read the
SUPERBLOCK_SIZE buffer and avoid error latter in fs_devread() function.
Signed-off-by: Patrick Delaunay <patrick.delaunay@foss.st.com>
