Ye Li [Thu, 25 Mar 2021 09:30:36 +0000 (17:30 +0800)]
crypto: fsl: refactor for 32 bit version CAAM support on ARM64
Previous patch "MLK-18044-4: crypto: caam: Fix pointer size to 32bit
for i.MX8M" breaks the 64 bits CAAM.
Since i.MX CAAM are all 32 bits no matter the ARM arch (32 or 64),
to adapt and not break 64 bits CAAM support, add a new config
CONFIG_CAAM_64BIT and new relevant type "caam_dma_addr_t".
This config is default enabled when CONFIG_PHYS_64BIT is set except
for iMX8M.
Signed-off-by: Ye Li <ye.li@nxp.com> Reviewed-by: Horia Geantă <horia.geanta@nxp.com> Signed-off-by: Peng Fan <peng.fan@nxp.com>
Clement Faure [Thu, 25 Mar 2021 09:30:33 +0000 (17:30 +0800)]
imx8m: Add DEK blob encapsulation for imx8m
Add DEK blob encapsulation support for IMX8M through "dek_blob" command.
On ARMv8, u-boot runs in non-secure, thus cannot encapsulate a DEK blob
for encrypted boot.
The DEK blob is encapsulated by OP-TEE through a trusted application call.
U-boot sends and receives the DEK and the DEK blob binaries through OP-TEE
dynamic shared memory.
To enable the DEK blob encapsulation, add to the defconfig:
CONFIG_SECURE_BOOT=y
CONFIG_FAT_WRITE=y
CONFIG_CMD_DEKBLOB=y
Signed-off-by: Clement Faure <clement.faure@nxp.com> Reviewed-by: Ye Li <ye.li@nxp.com> Signed-off-by: Peng Fan <peng.fan@nxp.com>
imx: caam: new u-boot command to set PRIBLOB bitfield from CAAM SCFGR register to 0x3
It is highly recommended to set the PRIBLOB bitfield to 0x3 once your
encrypted boot image has booted up, this prevents the generation of new
blobs that can be used to decrypt an encrypted boot image. The PRIBLOB is
a sticky type bit and cannot be changed until the next power on reset.
Add the set_priblob_bitfield U-Boot command to prevent the generation of
new blobs.
Signed-off-by: Clement Le Marquis <clement.lemarquis@nxp.com> Acked-by: Ye Li <Ye.Li@nxp.com> Signed-off-by: Peng Fan <peng.fan@nxp.com>
Aymen Sghaier [Thu, 25 Mar 2021 09:30:29 +0000 (17:30 +0800)]
crypto: caam: Add secure memory vid 3 support
In i.MX8M platforms the secure memory block has a newer version
than those used in i.MX6/7 platforms, this patch update the driver
to use the correct registers offsets.
Signed-off-by: Aymen Sghaier <aymen.sghaier@nxp.com> Signed-off-by: Peng Fan <peng.fan@nxp.com>
Aymen Sghaier [Thu, 25 Mar 2021 09:30:28 +0000 (17:30 +0800)]
crypto: caam: Fix pointer size to 32bit for i.MX8M
The CAAM block used in i.MX8M is 32 bits address size but when the flag
PHYS_64BIT is enabled for armv8, the CAAM driver will try to use a
wrong pointer size.
This patch fixes this issue.
Signed-off-by: Aymen Sghaier <aymen.sghaier@nxp.com> Signed-off-by: Peng Fan <peng.fan@nxp.com>
Aymen Sghaier [Thu, 25 Mar 2021 09:30:26 +0000 (17:30 +0800)]
crypto: caam: Fix build warnings pointer casting
Enabling CAAM driver for i.MX8M platforms, a 64 bits architecture,
lead to casting warnings: from/to pointer to/from integer with
different size. This patch fix these warnings
Signed-off-by: Aymen Sghaier <aymen.sghaier@nxp.com> Signed-off-by: Peng Fan <peng.fan@nxp.com>
Breno Lima [Thu, 25 Mar 2021 09:30:21 +0000 (17:30 +0800)]
cmd: blob: Instantiate RNG before running CMD_BLOB
U-Boot can instantiate CAAM RNG if needed by crypto operations.
Call sec_init() prior running a blob operation to ensure
RNG is correctly instantiated.
Make sure CAAM clock is enabled and check if a job ring is
available for that operation.
Signed-off-by: Breno Lima <breno.lima@nxp.com> Reviewed-by: Ye Li <ye.li@nxp.com> Signed-off-by: Peng Fan <peng.fan@nxp.com>
Breno Lima [Thu, 25 Mar 2021 09:30:19 +0000 (17:30 +0800)]
mx6dq: hab: Fix chip version in hab.h code
Since commit 8891410c729b ("MLK-19848 mx6dq: Fix chip version issue for
rev1.3") it's not possible to call the HAB API functions on i.MX6DQ
SoC Rev 1.3:
The hab.h code is defining the HAB API base address according to the
old SoC revision number, thus failing when calling the HAB API
authenticate_image() function.
Fix this issue by using mx6dq rev 1.3 instead of mx6dq rev 1.5.
Signed-off-by: Breno Lima <breno.lima@nxp.com> Reviewed-by: Ye Li <ye.li@nxp.com> Signed-off-by: Peng Fan <peng.fan@nxp.com>
Ye Li [Thu, 25 Mar 2021 09:30:17 +0000 (17:30 +0800)]
iMX8M: Add support to enable CONFIG_IMX_HAB
Add some SOC level codes and build configurations to use HAB lib for
CONFIG_IMX_HAB (secure boot), like adding the SEC_CONFIG fuse, enable
fuse driver, CAAM clock function, and add CAAM secure RAM to MMU table.
The FSL_CAAM is temporally not enabled for iMX8M when CONFIG_IMX_HAB is set,
because we don't need the CAAM driver for SPL.
Signed-off-by: Ye Li <ye.li@nxp.com> Reviewed-by: Peng Fan <peng.fan@nxp.com> Signed-off-by: Peng Fan <peng.fan@nxp.com>
Breno Lima [Thu, 25 Mar 2021 09:30:14 +0000 (17:30 +0800)]
imx: hab: Fix build warnings in 32-bit targets
When building 32-bit targets with CONFIG_SECURE_BOOT and DEBUG enabled
the following warnings are displayed:
arch/arm/mach-imx/hab.c:840:41: warning: format '%lx' expects argument \
of type 'long unsigned int', but argument 3 has type 'uint32_t \
{aka unsigned int}' [-Wformat=]
printf("HAB check target 0x%08x-0x%08lx fail\n",
~~~~^
%08x
ddr_start, ddr_start + bytes);
arch/arm/mach-imx/hab.c:845:45: warning: format '%x' expects argument \
of type 'unsigned int', but argument 3 has type 'ulong \
{aka long unsigned int}' [-Wformat=]
printf("\nivt_offset = 0x%x, ivt addr = 0x%x\n", ivt_offset, ivt_addr);
~^
%lx
Fix warnings by providing the correct data type.
Reviewed-by: Ye Li <ye.li@nxp.com> Signed-off-by: Breno Lima <breno.lima@nxp.com> Signed-off-by: Peng Fan <peng.fan@nxp.com>
Breno Lima [Thu, 25 Mar 2021 09:30:13 +0000 (17:30 +0800)]
mx7ulp: hab: Add hab_status command for HABv4 M4 boot
When booting in low power or dual boot modes the M4 binary is
authenticated by the M4 ROM code.
Add an option in hab_status command so users can retrieve M4 HAB
failure and warning events.
=> hab_status m4
Secure boot disabled
HAB Configuration: 0xf0, HAB State: 0x66
No HAB Events Found!
Add command documentation in mx6_mx7_secure_boot.txt guide.
As HAB M4 API cannot be called from A7 core the code is parsing
the M4 HAB persistent memory region. The HAB persistent memory
stores HAB events, public keys and others HAB related information.
The HAB persistent memory region addresses and sizes can be found
in AN12263 "HABv4 RVT Guidelines and Recommendations".
Reviewed-by: Utkarsh Gupta <utkarsh.gupta@nxp.com> Reviewed-by: Ye Li <ye.li@nxp.com> Signed-off-by: Breno Lima <breno.lima@nxp.com> Signed-off-by: Peng Fan <peng.fan@nxp.com>
Breno Lima [Thu, 25 Mar 2021 09:30:10 +0000 (17:30 +0800)]
imx: hab: Enable hab.c to authenticate additional images in open configuration
Currently it's not possible to authenticate additional boot images in HAB
open configuration.
The hab.c code is checking if the SEC_CONFIG[1] fuse is programmed prior
to calling the hab_authenticate_image() API function. Users cannot check
if their additional boot images has been correctly signed prior to closing
their device.
Enable hab.c to authenticate additional boot images in open mode so HAB
events can be retrieved through get_hab_status() function.
Signed-off-by: Breno Lima <breno.lima@nxp.com> Reviewed-by: Ye Li <ye.li@nxp.com> Signed-off-by: Peng Fan <peng.fan@nxp.com>
Ye Li [Thu, 25 Mar 2021 09:30:09 +0000 (17:30 +0800)]
hab: Change calling to ROM API failsafe
Modify to use hab_rvt_failsafe function for failsafe ROM API, not
directly call its ROM address. This function will wrap the sip call for iMX8M
platforms.
Signed-off-by: Ye Li <ye.li@nxp.com> Signed-off-by: Peng Fan <peng.fan@nxp.com>
Utkarsh Gupta [Thu, 25 Mar 2021 09:30:08 +0000 (17:30 +0800)]
imx: HAB: Validate IVT before authenticating image
Calling csf_is_valid() with an un-signed image may lead to data abort
as the CSF pointer could be pointing to a garbage address when accessed
in HAB_HDR_LEN(*(const struct hab_hdr *)(ulong)ivt_initial->csf).
To avoid such errors during authentication process, validate IVT structure
by calling validate_ivt function which checks the following values in an IVT:
Peng Fan [Thu, 25 Mar 2021 09:30:07 +0000 (17:30 +0800)]
imx: HAB: Update hab codes to support ARM64 and i.MX8M
There are some changes to support ARM64 i.MX8M platform in this patches:
1. The hab_rvt base and function vectors are different as i.MX6/7
2. Need to bypass an workaround for i.MX6 to fix problem in MMU.
3. The x18 register needed save & restore before calling any HAB API. According
to ARM procedure call spec, the x18 is caller saved when it is used as
temporary register. So calling HAB API may scratch this register, and
cause crash once accessing the gd pointer.
On ARMv7, the r9 is callee saved when it is used as variable register. So
no need to save & restore it.
4. Add SEC_CONFIG fuse for iMX8M
When current EL is not EL3, the direct calling to HAB will fail because
CAAM/SNVS can't initialize at non-secure mode. In this case, we use
SIP call to run the HAB in ATF.
Signed-off-by: Ye Li <ye.li@nxp.com> Signed-off-by: Peng Fan <peng.fan@nxp.com>
Breno Lima [Thu, 25 Mar 2021 09:30:04 +0000 (17:30 +0800)]
imx: Ensure CAAM clock is enabled prior getting out_jr_size
Prior calling sec_in32() we have to ensure CAAM clock is enabled, the
function sec_in32() is reading CAAM registers and if CAAM clock is disabled
the system will hang.
Signed-off-by: Breno Lima <breno.lima@nxp.com> Reviewed-by: Ye Li <ye.li@nxp.com> Signed-off-by: Peng Fan <peng.fan@nxp.com>
Breno Lima [Thu, 25 Mar 2021 09:30:03 +0000 (17:30 +0800)]
imx: Avoid hardcoded output ring size register offset (ORSR)
The CAAM output ring size register offset is currently defined in fsl_sec.h
as FSL_CAAM_ORSR_JRa_OFFSET, use this definition to avoid hardcoded value in
i.MX common code.
Signed-off-by: Breno Lima <breno.lima@nxp.com> Reviewed-by: Ye Li <ye.li@nxp.com> Signed-off-by: Peng Fan <peng.fan@nxp.com>
Breno Lima [Thu, 25 Mar 2021 09:30:02 +0000 (17:30 +0800)]
imx: imx7 Support for Manufacturing Protection
This code was originally developed by Raul Cardenas <raul.casas@nxp.com>
and modified to be applied in U-Boot imx_v2017.03.
More information about the initial submission can be seen
in the link below:
https://lists.denx.de/pipermail/u-boot/2016-February/245273.html
i.MX7D has an a protection feature for Manufacturing process.
This feature uses asymmetric encryption to sign and verify
authenticated software handled between parties. This command
enables the use of such feature.
The private key is unique and generated once per device.
And it is stored in secure memory and only accessible by CAAM.
Therefore, the public key generation and signature functions
are the only functions available for the user.
The manufacturing-protection authentication process can be used to
authenticate the chip to the OEM's server.
Command usage:
Print the public key for the device.
- mfgprot pubk
Generates Signature over given data.
- mfgprot sign <data_address> <data_size>
Signed-off-by: Raul Ulises Cardenas <raul.casas@nxp.com> Signed-off-by: Breno Lima <breno.lima@nxp.com> Reviewed-by: Fabio Estevam <fabio.estevam@nxp.com> Reviewed-by: Ye Li <ye.li@nxp.com> Signed-off-by: Peng Fan <peng.fan@nxp.com>
Ye Li [Fri, 19 Mar 2021 07:57:14 +0000 (15:57 +0800)]
imx8m: ddr: Disable CA VREF Training for LPDDR4
Users reported LPDDR4 MR12 value is set to 0 during PHY training,
not the value from FSP timing structure, which cause compliance test failed.
The root cause is the CATrainOpt[0] is set to 1 in 2D FSP timing
but not set in 1D. According to PHY training application node,
to enable the feature both 1D and 2D need set this field to 1,
otherwise the training result will be incorrect.
The PHY training doc also recommends to set CATrainOpt[0] to 0 to use
MR12 value from message block (FSP structure). So update the LPDDR4
scripts of all mscale to clear CATrainOpt[0].
Signed-off-by: Ye Li <ye.li@nxp.com> Reviewed-by: Jacky Bai <ping.bai@nxp.com> Signed-off-by: Peng Fan <peng.fan@nxp.com>
Ye Li [Fri, 19 Mar 2021 07:57:12 +0000 (15:57 +0800)]
imx8m: Update thermal and PMU kernel nodes for dual/single cores
For dual core and single core iMX8M parts, the thermal node and PMU node
in kernel DTB also needs update to remove the refers to deleted core nodes.
Otherwise both driver will fail to work.
Signed-off-by: Ye Li <ye.li@nxp.com> Reviewed-by: Peng Fan <peng.fan@nxp.com> Signed-off-by: Peng Fan <peng.fan@nxp.com>
Peng Fan [Fri, 19 Mar 2021 07:57:04 +0000 (15:57 +0800)]
imx8mp_evk: Increase VDD_ARM to 0.95v Overdrive voltage
There is a frequency/timing limitation for SOC and ARM, if SOC is OD
voltage/OD freq, then ARM can't run at ND voltage/1.2Ghz, it may have
timing risk from SOC to ARM.
Current VDD_SOC is set to 0.95v OD voltage in SPL, and kernel will
increase bus clocks to OD frequency before it increases ARM voltage.
So to conform to the limitation, we'd better increases VDD_ARM to OD
voltage in SPL.
Signed-off-by: Ye Li <ye.li@nxp.com> Signed-off-by: Peng Fan <peng.fan@nxp.com>
Ye Li [Fri, 19 Mar 2021 07:57:00 +0000 (15:57 +0800)]
imx8mp_evk: Update LPDDR4 timing for new FW 202006
After switching to new LPDDR4 firmware 202006 version, have to
update the LPDDR4 timing accordingly from RPA tool.
Signed-off-by: Ye Li <ye.li@nxp.com> Tested-by: Sherry Sun <sherry.sun@nxp.com> Tested-by: Jacky Bai <ping.bai@nxp.com> Reviewed-by: Jacky Bai <ping.bai@nxp.com> Signed-off-by: Peng Fan <peng.fan@nxp.com>
Tom Rini [Mon, 15 Mar 2021 17:19:01 +0000 (13:19 -0400)]
tools: Integrate the Dockerfile used for CI
Integrate the Dockerfile from
https://source.denx.de/u-boot/gitlab-ci-runner.git as of
commit bc6130d572f1 ("Dockerfile: Remove high UID/GID") and introduce a
short rST on how to build the container.
Cc: Heinrich Schuchardt <xypron.glpk@gmx.de> Signed-off-by: Tom Rini <trini@konsulko.com>
Tom Rini [Fri, 26 Feb 2021 12:52:28 +0000 (07:52 -0500)]
pytest: Lower pygit2 requirement
The latest versions of pygit2 are not available in practically any
distribution at this time. Furthermore, we don't need the latest in
order to run all of our testsuites. Reduce this version requirement to
something older that meets our needs while still supporting running our
tests on older hosts (and so, test labs).
Reported-by: Tom Warren <twarren@nvidia.com> Signed-off-by: Tom Rini <trini@konsulko.com>
Tom Rini [Fri, 26 Feb 2021 12:52:27 +0000 (07:52 -0500)]
GitLab: Remove "tags" stanzas
Given the structure of our current GitLab tests, we don't make real use
of the tags. Furthermore, these tags prevent the automatic usage of the
default GitLab runners. Remove these tags.
Reported-by: Roger Meier <r.meier@siemens.com> Signed-off-by: Tom Rini <trini@konsulko.com>
Tom Rini [Mon, 15 Feb 2021 15:52:19 +0000 (10:52 -0500)]
Azure: Rework SH / Renesas job
Now that there is a single SuperH platform, rework the Azure job
slightly. Azure build time limits mean that we need to split the world
build up still. Make a single build job for the single Renesas SuperH
platform as well as all of the ARM platforms from Renesas.
Cc: Marek Vasut <marek.vasut+renesas@gmail.com> Signed-off-by: Tom Rini <trini@konsulko.com>
JFFS2: fix jffs2 summary datacrc status uninitialized
The function jffs2_1pass_read_inode() was discarding the summary
inodes and dirent because the value in datacrc flag wasn't
initialized in function jffs2_sum_process_sum_data().
This fix initializes the status of all summary records to indicate
that the CRC needs to be verified when they are loaded.
Before this fix, the behaviors produced by the undefined value of
datacrc was:
- Summary's registries were discarded when 'b->datacrc' is equal
as 'CRC_BAD'.
- Summary's registries were not checked when b->datacrc differs of
'CRC_BAD' and 'CRC_UNKNOWN'
So, almost all of the time the crc just isn't checked, and in some
cases the registries are discarded.
Signed-off-by: Wagner Popov dos Santos <wpopov@gmail.com>
Fixes address violation in functions read_nand_cached() and
read_onenand_cached(). This happens because these functions
try to read a fixed amount
of data even when the offset+length
is above the nand's limit.
Signed-off-by: Wagner Popov dos Santos <wpopov@gmail.com>
Tom Rini [Wed, 7 Apr 2021 02:42:55 +0000 (22:42 -0400)]
Merge tag 'mmc-2021-4-6' of https://source.denx.de/u-boot/custodians/u-boot-mmc
Update hwpartition usage
Check bootbus's arguments
workaround for erratum A-011334 for fsl_esdhc driver
add pulse width detection workaround for fsl_esdhc driver
Use alias num before checking mmc index when creating device
Stefan Roese [Tue, 6 Apr 2021 10:10:18 +0000 (12:10 +0200)]
usb: xhci: Make debug output better readable and checkpatch clean
This change makes debugging a bit easier as the output is better
readable with the added space. The explicit le16_to_cpu() is not
needed in the output. Also this patch moves the strings into one line
to make the patch checkpatch clean.
Signed-off-by: Stefan Roese <sr@denx.de> Cc: Aaron Williams <awilliams@marvell.com> Cc: Chandrakala Chavva <cchavva@marvell.com> Cc: Bin Meng <bmeng.cn@gmail.com> Cc: Marek Vasut <marex@denx.de>
Aaron Williams [Tue, 6 Apr 2021 10:10:17 +0000 (12:10 +0200)]
usb: xhci: Add missing xhci_readl()
Accessing the xHCI controller registers should be done via the
xhci_readl/writel functions. This patch adds this to a few missing
places.
Signed-off-by: Aaron Williams <awilliams@marvell.com> Signed-off-by: Stefan Roese <sr@denx.de> Cc: Chandrakala Chavva <cchavva@marvell.com> Cc: Bin Meng <bmeng.cn@gmail.com> Cc: Marek Vasut <marex@denx.de>
Aaron Williams [Tue, 6 Apr 2021 10:10:16 +0000 (12:10 +0200)]
usb: hub: Fix usb_get_port_status() for big-endian platforms
Add missing endianness conversions to usb_get_port_status(). This
(amongst others) is necessary to enable the use of USB 3 hubs on
big-endian platforms like MIPS Octeon.
Signed-off-by: Aaron Williams <awilliams@marvell.com> Signed-off-by: Stefan Roese <sr@denx.de> Cc: Chandrakala Chavva <cchavva@marvell.com> Cc: Bin Meng <bmeng.cn@gmail.com> Cc: Marek Vasut <marex@denx.de>
Jaehoon Chung [Fri, 26 Feb 2021 09:38:20 +0000 (18:38 +0900)]
cmd: mmc: check whether bootbus's arguments is valid or not
According to Specification, each bit have valid value.
But it doesn't check whether arguments is valid or not.
It has potential bug with arguments passed by wrong value.
Jaehoon Chung [Fri, 26 Feb 2021 07:07:18 +0000 (16:07 +0900)]
cmd: mmc: modify more readable about hwpartition usage
Modified more readable about hwpartition usage.
Because it's difficult to understand how to use its command.
The arguments didn't optional.
mmc hwpartition needs to pass <USER> <GP> <MODE> as arguments.
Description about each arguments what is required is the below:
USER - <user> <enh> <start> <cnt> <wrrel> <{on|off}>
GP - <{gp1|gp2|gp3|gp4}> <cnt> <enh> <wrrel> <{on|off}>
MODE - <{check|set|complete}>
Neil Armstrong [Wed, 24 Feb 2021 19:33:56 +0000 (20:33 +0100)]
net: designware: add Amlogic Meson8b & later glue driver
This adds a proper glue driver for the Designware DWMAC ethernet MAC IP
found in the Amlogic Meson8, GXBB, GXL, GXM, G12A, G12B & SM1 SoCs.
This is aimed to replace the static ethernet link setup found on the board
init code for the Amlogic SoC based boards.
Tested on a libretech-cc (S905x Internal RMII 10/100 PHY) and Khadas VIM3 (A113d
with external 10/100/1000 RGMII PHY) to cover the most extreme setups.
Signed-off-by: Neil Armstrong <narmstrong@baylibre.com>
Neil Armstrong [Wed, 24 Feb 2021 14:02:39 +0000 (15:02 +0100)]
net: designware: add DM_MDIO support
Add support for DM_MDIO to connect to PHY and expose a MDIO device for the
internal MDIO bus in order to dynamically connect to MDIO PHYs with DT
with eventual MDIO muxes in between.
Signed-off-by: Neil Armstrong <narmstrong@baylibre.com> Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
Pali Rohár [Tue, 9 Mar 2021 13:26:57 +0000 (14:26 +0100)]
arm: mvebu: Espressobin: Enable watchdog support but do not start it
Espressobin had disabled watchdog support (CONFIG_WDT) because older stable
Linux kernel versions (which are used by current stable OpenWRT and Debian
versions) do not have support for Armada 3700 watchdog driver. Therefore
they are not able to periodically kick watchdog so Espressobin enter into
boot loop.
This change enable CONFIG_WDT, CONFIG_WDT_ARMADA_37XX and CONFIG_CMD_WDT
options which add support for U-Boot 'wdt' command. And unset new
CONFIG_WATCHDOG_AUTOSTART option which cause that watchdog is not
automatically started by U-Boot during init phase, like when CONFIG_WDT
option is not set at all.
So with this change, U-Boot on Espressobin would have working 'wdt' command
which can be used from boot scripts (e.g. for enabling watchdog prior new
Linux booting kernel). But default behavior of watchdog status stays
unchanged, U-Boot does not start watchdog on Espressobin during its init
phase.
Signed-off-by: Pali Rohár <pali@kernel.org> Reviewed-by: Stefan Roese <sr@denx.de>
Pali Rohár [Tue, 9 Mar 2021 13:26:56 +0000 (14:26 +0100)]
watchdog: Allow to use CONFIG_WDT without starting watchdog
In some cases it is useful to compile support for U-Boot command 'wdt'
without starting HW watchdog in early U-Boot phase. For example when the
user want to start the watchdog only on demand by some boot script.
This change adds a new compile option WATCHDOG_AUTOSTART to control whether
U-Boot should automatically start the watchdog during init phase or not.
This option is enabled by default as it was the default behavior prior
introducing this new change. When compiling U-Boot users can decide to turn
this option off.
Signed-off-by: Pali Rohár <pali@kernel.org> Reviewed-by: Stefan Roese <sr@denx.de>