Marek Vasut [Fri, 19 Apr 2024 03:59:05 +0000 (05:59 +0200)]
ARM: stm32: Initialize TAMP_SMCR BKP..PROT fields on STM32MP15xx
In case of an OTP-CLOSED STM32MP15xx system, the CPU core 1 cannot be
released from endless loop in BootROM only by populating TAMP BKPxR 4
and 5 with magic and branch address and sending SGI0 interrupt from
core 0 to core 1 twice. TAMP_SMCR BKP..PROT fields must be initialized
as well to release the core 1 from endless loop during the second SGI0
handling on core 1. Initialize TAMP_SMCR to protect the first 32 backup
registers, the ones which contain the core 1 magic, branch address and
boot information.
This requirement seems to be undocumented, therefore it was necessary
to trace and analyze the STM32MP15xx BootROM using OpenOCD and objdump.
Ultimately, it turns out that a certain BootROM function reads out the
TAMP_SMCR register and tests whether the BKP..PROT fields are non-zero.
If they are zero, the BootROM code again waits for SGI0 using WFI, else
the execution moves forward until it reaches handoff to the TAMP BKPxR 5
branch address.
This fixes CPU core 1 release using U-Boot PSCI implementation on an
OTP-CLOSED system, i.e. system with fuse 0 bit 6 set.
Reviewed-by: Patrick Delaunay <patrick.delaunay@foss.st.com> Signed-off-by: Marek Vasut <marex@denx.de> Reviewed-by: Patrice Chotard <patrice.chotard@foss.st.com>
Marek Vasut [Sun, 14 Apr 2024 18:39:29 +0000 (20:39 +0200)]
ARM: stm32: Report OTP-CLOSED instead of rev.? on closed STM32MP15xx
SoC revision is only accessible via DBUMCU IDC register,
which requires BSEC.DENABLE DBGSWENABLE bit to be set to
make the register accessible, otherwise an access to the
register triggers bus fault. As BSEC.DBGSWENABLE is zero
in case of an OTP-CLOSED system, do NOT set DBGSWENABLE
bit as this might open a brief window for timing attacks.
Instead, report that this system is OTP-CLOSED and do not
report any SoC revision to avoid confusing users. Use an
SEC/C abbreviation to avoid growing SOC_NAME_SIZE .
Signed-off-by: Marek Vasut <marex@denx.de> Reviewed-by: Patrice Chotard <patrice.chotard@foss.st.com>
Marek Vasut [Sun, 14 Apr 2024 18:38:31 +0000 (20:38 +0200)]
ARM: stm32: Drop superfluous Makefile entry for ecdsa_romapi.o
The source file is in arch/arm/mach-stm32mp/ecdsa_romapi.c and not
in arch/arm/mach-stm32mp/stm32mp1/ecdsa_romapi.c . There are two
Makefile entries in each subdirectory. Drop the bogus one and keep
only the correct one, the one in arch/arm/mach-stm32mp/Makefile .
Signed-off-by: Marek Vasut <marex@denx.de> Reviewed-by: Patrice Chotard <patrice.chotard@foss.st.com>
Marek Vasut [Sun, 7 Apr 2024 20:21:07 +0000 (22:21 +0200)]
ARM: stm32: Jump to ep on successful resume in PSCI suspend code
In case the system has resumed successfully, the PSCI suspend resume
code has to jump to the 'ep' successful resume entry point code path,
otherwise the code has to jump to content of the LR register, which
points to failed resume code path.
To implement this distinction, rewrite LR register stored on stack
with 'ep' value in case of a successful resume, which is really in
every case unless some catastrophic failure occurred during suspend.
Without this change, Linux counts every resume as failed in
/sys/power/suspend_stats/fail
Signed-off-by: Marek Vasut <marex@denx.de> Reviewed-by: Patrice Chotard <patrice.chotard@foss.st.com>
Marek Vasut [Tue, 26 Mar 2024 12:07:32 +0000 (13:07 +0100)]
net: dwc_eth_qos: Add support for st, ext-phyclk property
The "st,ext-phyclk" property is a unification of "st,eth-clk-sel"
and "st,eth-ref-clk-sel" properties. All three properties define
ETH CK clock direction, however:
- "st,eth-clk-sel" selects clock direction for GMII/RGMII mode
- "st,eth-ref-clk-sel" selects clock direction for RMII mode
- "st,ext-phyclk" selects clock direction for all RMII/GMII/RGMII modes
The "st,ext-phyclk" is the preferrable property to use.
Signed-off-by: Marek Vasut <marex@denx.de> Reviewed-by: Christophe ROULLIER <christophe.roullier@foss.st.com>
Marek Vasut [Tue, 26 Mar 2024 12:07:29 +0000 (13:07 +0100)]
net: dwc_eth_qos: Constify st, eth-* values parsed out of DT
Use const bool for the values parsed out of DT. Drop the duplicate
assignment of false into those bool variables, assign them directly
with the content parsed out of DT. Abbreviate the variable name too.
Reviewed-by: Patrice Chotard <patrice.chotard@foss.st.com> Signed-off-by: Marek Vasut <marex@denx.de>
Marek Vasut [Tue, 26 Mar 2024 12:07:25 +0000 (13:07 +0100)]
net: dwc_eth_qos: Scrub ifdeffery
Replace ifdef CONFIG_CLK with if (CONFIG_IS_ENABLED(CLK)) to improve code
build coverage. Some of the functions printed debug("%s: OK\n", __func__);
on exit with and without CLK enabled, some did not, make it consistent and
print nothing if CLK is disabled.
Marek Vasut [Tue, 26 Mar 2024 12:07:24 +0000 (13:07 +0100)]
net: dwc_eth_qos: Fold board_interface_eth_init into STM32 glue code
Move board_interface_eth_init() into eqos_probe_syscfg_stm32() in STM32
driver glue code. The eqos_probe_syscfg_stm32() parses STM32 specific DT
properties of this MAC and configures SYSCFG registers accordingly, there
is nothing board specific happening in this function, move it into generic
driver code instead. Drop the now unused duplicates from board files.
Reviewed-by: Patrice Chotard <patrice.chotard@foss.st.com> Signed-off-by: Marek Vasut <marex@denx.de>
Marek Vasut [Tue, 26 Mar 2024 12:07:23 +0000 (13:07 +0100)]
net: dwc_eth_qos: Rename eqos_stm32_config to eqos_stm32mp15_config
The current glue code is specific to STM32MP15xx, the upcoming STM32MP13xx
will introduce another entry specific to the STM32MP13xx. Rename the current
entry to eqos_stm32mp15_config in preparation for STM32MP13xx addition. No
functional change.
The FMC2 revision 2 supports security and isolation compliant with
the Resource Isolation Framework (RIF). From RIF point of view,
the FMC2 is composed of several independent resources, listed below,
which can be assigned to different security and compartment domains:
- 0: Common FMC_CFGR register.
- 1: EBI controller for Chip Select 1.
- 2: EBI controller for Chip Select 2.
- 3: EBI controller for Chip Select 3.
- 4: EBI controller for Chip Select 4.
- 5: NAND controller.
Add the support of the revision 2 of FMC2 IP.
- PCSCNTR register has been removed,
- CFGR register has been added,
- the bit used to enable the IP has moved from BCR1 to CFGR,
- the timeout for CEx deassertion has moved from PCSCNTR to BCRx,
- the continuous clock enable has moved from BCR1 to CFGR,
- the clk divide ratio has moved from BCR1 to CFGR.
The MP1 SoCs have only one signal to manage all the controllers (NWAIT).
The MP25 SOC has one RNB signal for the NAND controller and one NWAIT
signal for the memory controller.
Let's use a platform data structure for parameters that will differ
between MP1 and MP25.
Marek Vasut [Mon, 4 Mar 2024 18:25:43 +0000 (19:25 +0100)]
arm: stm32: Enable OHCI HCD support on STM32MP15xx DHSOM
The OHCI HCD is mandatory for USB 1.1 FS/LS device support, enable it.
This used to be enabled implicitly before, now that implicit dependency
disappeared and this got disabled. Enable it manually.
Signed-off-by: Marek Vasut <marex@denx.de> Reviewed-by: Patrice Chotard <patrice.chotard@foss.st.com>
stm32mp: cmd_stm32prog: add dependencies with USB_GADGET_DOWNLOAD
This patch avoids compilation issue when CONFIG_USB_GADGET is deactivated
in defconfig, with undefined reference to run_usb_dnl_gadget and to
g_dnl_set_product.
Signed-off-by: Patrick Delaunay <patrick.delaunay@foss.st.com> Reviewed-by: Igor Opaniuk <igor.opaniuk@foundries.io> Reviewed-by: Patrice Chotard <patrice.chotard@foss.st.com>
This prints the MMC device being read similar to how we print the MMC
device we write to when e.g. calling saveenv.
One of the side effects is that the boot log now shows from which MMC
device the env was loaded:
Loading Environment from MMC... Reading from MMC(1)... OK
This is useful to identify which MMC device the environment was loaded
from for boards where there are more than one (e.g. eMMC and SD card)
without adding some debug messages manually.
Sadly, there's no way to know which of the default or redundant
environment is being read from env_mmc_load before env_import_redund is
called so it is printing a bit later (and possibly after error/warning
messages).
Michal Simek [Mon, 15 Apr 2024 08:20:05 +0000 (10:20 +0200)]
Kconfig: Remove all default n options
default n doesn't need to be specified. It is default option anyway.
Similar changes have been done by commit 18370f14975c ("Kconfig: Remove all
default n/no options").
Marek Vasut [Sun, 14 Apr 2024 18:37:20 +0000 (20:37 +0200)]
boot: fdt: Turn all addresses and sizes into u64
In case of systems where DRAM bank ends at the edge of 32bit boundary,
start + size calculations would overflow. This happens on STM32MP15xx
with 1 DRAM bank starting at 0xc0000000 and 1 GiB of DRAM. This is a
usual 32bit system DRAM size overflow, fix it by doing all DRAM size
and offset calculations using u64 types. This also covers a case where
a 32bit PAE system might be able to address up to 36bits of DRAM.
Fixes: a4df06e41fa2 ("boot: fdt: Change type of env_get_bootm_low() to phys_addr_t") Signed-off-by: Marek Vasut <marex@denx.de>
This patch adds DEFLATE compression algorithm support. It's a good choice
to trade off between compression ratios and performance compared to LZ4.
Alternatively, DEFLATE could be used for some specific files since EROFS
supports multiple compression algorithms in one image.
Charles Hardin [Fri, 12 Apr 2024 20:45:33 +0000 (13:45 -0700)]
net: add support to parse the NIS domain for the dhcp options
There is code in the bootp parsing for NIS domain and add the
same support for the dhcp options as well. This allows the same
usage of the data when the dhcp command is used in the boot
command.
Signed-off-by: Charles Hardin <ckhardin@gmail.com>
input: button_kbd: gracefully handle buttons that fail probe
If a button device fails to probe, it will still be added to the uclass
device list, and therefore will still be iterated over in
button_read_keys() resulting in a UAF on the buttons private data.
Resolve this by unbinding button devices that aren't active after
probing, and print a warning so it's clear that the button is broken.
Tom Rini [Thu, 18 Apr 2024 18:13:40 +0000 (12:13 -0600)]
Merge tag 'tpm-master-18042024' of https://source.denx.de/u-boot/custodians/u-boot-tpm
Igor says:
"The problem initially was in the TEE sandbox driver implementation
(drivers/tee/sandbox.c) and it's limitations, which doesn't
permit to have multiple simultaneous sessions with different TAs.
This is what actually happened in this CI run [1], firstly "optee_rpmb"
cmd was executed (and after execution we had one session open), and
then "scp03", which also makes calls to OP-TEE, however it fails
in sandbox_tee_open_session() because of this check:
if (state->ta) {
printf("A session is already open\n");
return -EBUSY;
}
I had two ways in mind to address that:
1. Close a session on each optee_rpmb cmd invocation.
I don't see any reason to keep this session open, as obviously
there is no other mechanism (tbh, I don't know if DM calls ".remove" for active
devices) to close it automatically before handing over control to
Linux kernel. As a result we might end up with some orphaned sessions
registered in OP-TEE OS core (obvious resource leak).
2. Extend TEE sandbox driver, add support for multiple
simultaneous sessions just to handle the case.
I've chosen the first approach, as IMO it was "kill two birds with one stone",
I could address resource leak in OP-TEE and bypass limitations of
TEE sandbox driver."
Tom Rini [Thu, 18 Apr 2024 14:29:35 +0000 (08:29 -0600)]
Revert "Merge patch series "pxe: Allow extlinux booting without CMDLINE enabled""
As reported by Jonas Karlman this series breaks booting on some AArch64
platforms with common use cases. For now the best path forward is to
revert the series.
Unlike CMPC885 and CMPCPRO boards, the FPGA of MCR3000 board doesn't
load code automatically but needs to be loaded by software through SPI.
Until now it was loaded later by Linux, but we'd like U-boot to have
access to some information that require the FPGA, like board address
in racks.
So, implemented the load of FPGA in U-boot.
Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
---
To avoid spamming your email boxes, the code isn't included in
the emailed patch but will be present in the PULL request
Instead of cross using cross-directory makefile directives,
add a Makefile in board/cssi/common/ directory in order to
benefit from HAVE_VENDOR_COMMON_LIB logic.
spi: mpc8xx: Use 16 bit mode for large transfers with even size
On CPM, the RISC core is a lot more efficiant when doing transfers
in 16-bits chunks than in 8-bits chunks, but unfortunately the
words need to be byte swapped.
So, for large tranfers with an even size, allocate a temporary
buffer and byte-swap data before and after transfer.
This change allows setting higher speed for transfer. For instance
on an MPC 8xx (CPM1 comms RISC processor), the documentation tells
that transfer in byte mode at 1 kbit/s uses 0.200% of CPM load
at 25 MHz while a word transfer at the same speed uses 0.032%
of CPM load. This means the speed can be 6 times higher in
word mode for the same CPM load.
For small transfers, the load reduction is not worth the CPU load
required to allocate the temporary buffer, so do it only when data
size is over 64 bytes.
board: cssi: Add support for SPI bus on MCR3000 board
MCR3000 board has some components tied to the SPI bus, like the Texas
Instruments LM74 temperature sensor.
Add support for SPI bus. The SPI chipselects are a bit special in the
way that they are driven by 3 bits in a register of the board's CPLD
where the value writen in those bits exclusively activates one of the
7 possible chipselects and value 0 sets all chipselets to inactive.
So add a special GPIO driver that simulates GPIOs for those chipselect.
Hugo Dubois [Tue, 5 Dec 2023 16:10:44 +0000 (17:10 +0100)]
board: cssi: Properly initialise MAC address for fibre on CMPC885 board
CMPC885 board can be pluged on a mother board with fibre interface, so
fibre interface MAC address must be initialised for that case.
Signed-off-by: Hugo Dubois <hugo.dubois.ext@csgroup.eu> Reviewed-by: CASAUBON Jean Michel <jean-michel.casaubon@csgroup.eu> Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Hugo Dubois [Fri, 1 Dec 2023 14:56:12 +0000 (15:56 +0100)]
board: cssi: Initialise port F on MIAE
When equipped with the SRSA audio board, MIAE equipment
has an additional port called port F.
Initialise that port just like other ports of the board, so
that it is already configured when starting Linux kernel.
Signed-off-by: Hugo Dubois <hugo.dubois.ext@csgroup.eu> Reviewed-by: CASAUBON Jean Michel <jean-michel.casaubon@csgroup.eu> Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Fix it by removing cell-index and cell-size which is unused and add
reg property. Also fix node name to be in line with reg value.
Also add missing compatible for eeprom node.
Igor Opaniuk [Thu, 4 Apr 2024 13:19:51 +0000 (15:19 +0200)]
tee: remove common.h inclusion
The usage of the common.h include file is deprecated [1], and has already
been removed from several files.
Get rid of all inclusions in the "drivers/tee" directory, and replace it
with required include files directly where needed.
[1] doc/develop/codingstyle.rst
Reviewed-by: Ilias Apalodimas <ilias.apalodimas@linaro.org> Signed-off-by: Igor Opaniuk <igor.opaniuk@gmail.com> Reviewed-by: Sam Protsenko <semen.protsenko@linaro.org> Signed-off-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
Igor Opaniuk [Thu, 4 Apr 2024 13:19:48 +0000 (15:19 +0200)]
cmd: optee_rpmb: close tee session
Close tee session after each optee_rpmb invocation, as there is no
reason to keep it open, considering the absence of any available mechanism
to clean up all open sessions automatically before handing over control
to the Linux kernel. Without proper clean-up we might end up with orphaned
sessions registered in OP-TEE OS core (obvious resource leak).
Signed-off-by: Igor Opaniuk <igor.opaniuk@gmail.com> Signed-off-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
Marek Vasut [Mon, 25 Mar 2024 08:17:54 +0000 (09:17 +0100)]
ARM: dts: renesas: Switch to using upstream DT
Enable OF_UPSTREAM to use upstream DT and add renesas/ prefix to the
DEFAULT_DEVICE_TREE and OF_LIST. And thereby directly build DTB from
dts/upstream/src/ including *-u-boot.dtsi files from arch/$(ARCH)/dts/
directory.
The configuration update has been done using the following script:
```
sed -i '/^CONFIG_DEFAULT_DEVICE_TREE/ s@="@&renesas/@' `git grep -li renesas configs`
sed -i '/^CONFIG_OF_LIST/ s@r8a@renesas/&@g' `git grep -li renesas configs`
```
The RZN1 Snarc board does not seem to have a matching Linux kernel
DT counterpart, this is currently not switched to OF upstream.
Signed-off-by: Marek Vasut <marek.vasut+renesas@mailbox.org> Reviewed-by: Sumit Garg <sumit.garg@linaro.org>
Marek Vasut [Mon, 18 Mar 2024 14:59:38 +0000 (15:59 +0100)]
ARM: dts: renesas: Switch to using upstream DT on 64bit Renesas SoCs
Enable OF_UPSTREAM to use upstream DT and add renesas/ prefix to the
DEFAULT_DEVICE_TREE and OF_LIST. And thereby directly build DTB from
dts/upstream/src/ including *-u-boot.dtsi files from arch/$(ARCH)/dts/
directory.
The configuration update has been done using the following script:
```
sed -i '/^CONFIG_DEFAULT_DEVICE_TREE/ s@="@&renesas/@' `git grep -li renesas configs`
sed -i '/^CONFIG_OF_LIST/ s@r8a@renesas/&@g' `git grep -li renesas configs`
```
There are SoCs which are not included in this patch. The 32bit SoCs
require further infrastructure work. R8A779H0 is coming during the
next upstream DT synchronization cycle as it is not included in
current upstream DTs yet.
Signed-off-by: Marek Vasut <marek.vasut+renesas@mailbox.org> Reviewed-by: Adam Ford <aford173@gmail.com> # Beacon boards Reviewed-by: Paul Barker <paul.barker.ct@bp.renesas.com> Reviewed-by: Sumit Garg <sumit.garg@linaro.org>
Marek Vasut [Mon, 18 Mar 2024 14:59:37 +0000 (15:59 +0100)]
ARM: dts: renesas: Replace R8A77950 with R8A77951
R-Car R8A77950 H3 ES1.* was only available to an internal development
group and needed a lot of quirks and workarounds. Public users only
have R-Car R8A77950 H3 ES2 and newer. Switch existing systems to use
R8A77951 device trees.
Signed-off-by: Marek Vasut <marek.vasut+renesas@mailbox.org>
Marek Vasut [Sun, 17 Mar 2024 06:23:38 +0000 (07:23 +0100)]
ARM: dts: renesas: Stop using the -u-boot DTs for build
The U-Boot build system can automatically paste -u-boot.dtsi at the
end of matching .dts during build. Stop emulating this behavior and
rename the -u-boot.dts files to -u-boot.dtsi, drop "#include...dts"
from those new u-boot.dtsi files, and update board configuration
accordingly.
The rename, '#include...dts` scrubbing and configuration update has
been done using the following script:
```
$ find . -name r[78]\*-u-boot.dts | sort -u | while read line ; do \
git mv ${line%-u-boot.dts}-u-boot.dts ${line%-u-boot.dts}-u-boot.dtsi ; \
done
$ sed -i '/^#include.*dts"/ d' `find . -name r[78]\*-u-boot.dtsi`
$ sed -i 's@-u-boot@@g' `git grep -li renesas configs`
```
The Salvator-X and ULCB board files have been updated manually.
Signed-off-by: Marek Vasut <marek.vasut+renesas@mailbox.org> Acked-by: Adam Ford <aford173@gmail.com>
sandbox: don't call os_close with invalid file descriptor
If open() fails it returns -1. Calling close() with this value
makes no sense. Return -EIO instead.
Addresses-Coverity-ID: 185828 Improper use of negative value Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com> Fixes: 566bf3a8698 ("sandbox: Add a function to read a host file") Reviewed-by: Sean Anderson <seanga2@gmail.com>
Andrew Davis [Wed, 10 Apr 2024 18:38:34 +0000 (13:38 -0500)]
arm: mach-k3: security: Lower verbosity of cert message for GP
When we find a certificate on an image to be booted on a GP device we
print out a message explaining that the certificate is being skipped.
This message is rather long and is printed for every image. Shorten
the message and make the long version into a debug message.
Signed-off-by: Andrew Davis <afd@ti.com> Tested-by: Jonathan Humphreys <j-humphreys@ti.com>
configs: am64x_evm_*_defconfig: Increase offsets for eMMC raw boot
EMMC boot can fail due to the size of R5 SPL image growing beyond the
500KB of memory allocated in eMMC. Update offsets for eMMMC raw boot
to load each binary from the correct address in eMMC according to the
following eMMC layout:
This is a trivial but significant optimization:
mkimage took >200ms (and 49489 writes (of which 49456 512)),
now it takes 110ms (and 419 writes (of which 386 64k)).
sendfile is much more appropriate for this and is done in one syscall,
but doesn't bring any significant speedups over 64k r/w
at the 13M size ranges, so there's no need to introduce
#if __linux__
while((size = sendfile(fd_dst, fd_src, NULL, 128 * 1024 * 1024)) > 0)
;
if(size != -1) {
ret = 0;
goto out;
}
#endif
Makefile.lib: find capsule ESL dtsi file with CONFIG_OF_UPSTREAM
When CONFIG_OF_UPSTREAM is enabled, DTS files are in SOC subdirectories (vs the
top level dts directory), but when CONFIG_EFI_CAPSULE_AUTHENTICATE is enabled,
the dynamically created dtsi file containing the capsule ESL DT node is in the
parent directory. This results in a build failure because the #include inserted
in the DTS file is local to the current directory. Update Makefile to have the
DT preprocessing of #includes search in the parent (dts top level) directory
too.
Signed-off-by: Jonathan Humphreys <j-humphreys@ti.com>
This patch series is needed to get U-Boot to boot from a ZFS filesystem
on an aarch64 computer. Some of the patches are not architecture specific
and would be needed to boot ZFS on other platforms as well. The ZFS
support in U-Boot hasn't been substantively touched in several years and
to me it seems like it must have been broken for a long time on all
platforms, but I have only tested on aarch64.
Since there doesn't seem to be a mantainer for this area who I can cc,
I'm hoping these patches get seen and pulled in by a general U-Boot
maintainer.
[trini: Per Igor's comment and Phaedrus agreement, dropped his Tested-by
on the patches themselves]
Without this patch, the while loop being modified goes on infinitely,
but with the patch I am able to boot linux on zfs on a jetson tx2 nx.
It seems like this code was never tested because the logic is clearly
wrong. The function do_div(a,b) does a division that modifies the first
parameter to have a = a / b, and returns the remainder of the division.
So clearly in the usual case when file->offset = 0, the line
"blkid = do_div(blkid, blksz);" just results in blkid being set to zero
on every iteration of the loop, rather than being incremented as blocks
are read. Hence the zeroth block is read over and over and this becomes
an infinite loop.
So instead capture the remainder of the division in a "blkoff" variable,
and use that to properly calculate the memory address to move from in
memmove() below.
For example, if file->offset were 1337, on the first iteration of the
loop blkid would be 0 and blkoff would be 1337. If the blksz is 131072
(as it was for me), that amount of data would be copied into
data->file_buf. movesize would be 131072 - 1337 = 129735 so 129735 bytes
would be moved into buf. On the second iteration of the loop (assuming
there is one), red would be 129735, blkid would be 1, blkoff would be 0,
and 131072 bytes would be copied into buf. And so on...
As evidenced by how other filesystems handle it, a return value of 0
from fs_devread() means failure; nonzero means success. The opposite
assumption was being made in zfs.c for the use of zfs_devread() so fix
the confusion by making zfs_devread() return 0 on success.
It probably doesn't make sense to change the handling of zfs_devread()
in zfs.c instead, because as it is it matches the semantics of the other
functions there.
This happens when be64_to_cpu() is called on a value that exists at a
memory address that's 4 byte aligned but not 8 byte aligned (e.g. an
address ending in 04). The call stack where that happens is:
check_pool_label() ->
zfs_nvlist_lookup_uint64(vdevnvlist, ZPOOL_CONFIG_ASHIFT,...) ->
be64_to_cpu()
This code was hitting the error code path whenever malloc() succeeded
rather than when it failed, so presumably this part of the code hasn't
been tested. I had to apply this fix (and others) to get U-Boot to boot
from ZFS on an Nvidia Jetson TX2 NX SoM (an aarch64 computer).
projects / u-boot.git / log