Tom Rini [Thu, 3 Jan 2019 13:39:44 +0000 (08:39 -0500)]
Merge tag 'for-master-20190103' of git://git.denx.de/u-boot-rockchip
Last-minute fixes for 2019.1:
- clamp DRAM size to below 32bit for 32bit targets to support 4GB
- fix copyright notice on some Rockchip-contributed files
- adjust vdd_log for the RK3399-Q7 to improve stability in some workloads
rockchip: rk3399: Add Kconfig option for full pinctrl driver
This patch adds a Kconfig option to enable the full pinctrl driver
for the RK3399. This flag needs to be enabed in order to get the
features of the full pinctrl driver compiled in (i.e. a .set_state()
callback).
Signed-off-by: Christoph Muellner <christoph.muellner@theobroma-systems.com> Reviewed-by: Philipp Tomsich <philipp.tomsich@theobroma-systems.com>
The current pinctrl driver for the RK3399 has a range of qulity issues.
E.g. it only implements the .set_state_simple() callback, it
does not parse the available pinctrl information from the DTS
(instead uses hardcoded values), is not flexible enough to cover
devices without 'interrupt' field in the DTS (e.g. PWM),
is not written generic enough to make code reusable among other
rockchip SoCs...
This patch addresses these issues by reimplementing the whole driver
from scratch using the .set_state() callback.
The new implementation covers all featurese of the old code
(i.e. it supports pinmuxing and pullup/pulldown configuration).
This patch has been tested on a RK3399-Q7 SoM (Puma).
Signed-off-by: Christoph Muellner <christoph.muellner@theobroma-systems.com> Reviewed-by: Philipp Tomsich <philipp.tomsich@theobroma-systems.com>
pinctrl_decode_pin_config_dm() is basically a feature-equivalent
implementation of pinctrl_decode_pin_config(), which operates
on struct udevice devices and uses the dev_read_*() API.
Signed-off-by: Christoph Muellner <christoph.muellner@theobroma-systems.com> Reviewed-by: Philipp Tomsich <philipp.tomsich@theobroma-systems.com>
Kever Yang [Fri, 28 Dec 2018 01:56:48 +0000 (09:56 +0800)]
rockchip: sdram-common: fix wrong size for 4GB in 32bit SoC
This is workaround for issue we can't get correct size for 4GB ram
in 32bit system and available before we really need ram space
out of 4GB, eg.enable ARM LAPE(rk3288 supports 8GB ram).
The size of 4GB is '0x1 00000000', and this value will be truncated
to 0 in 32bit system, and system can not get correct ram size.
Rockchip SoCs reserve a blob of space for peripheral near 4GB,
and we are now setting SDRAM_MAX_SIZE as max available space for
ram in 4GB, so we can use this directly to workaround the issue.
Signed-off-by: Kever Yang <kever.yang@rock-chips.com> Tested-By: Vagrant Cascadian <vagrant@debian.org> Reviewed-by: Philipp Tomsich <philipp.tomsich@theobroma-systems.com> Reviewed-by: Simon Glass <sjg@chromium.org> Tested-by: Simon Glass <sjg@chromium.org>
Current driver doesn't check if the destination pointer is NULL.
This cause the data from the FIFO to be stored inside the internal
SDRAM ( address 0 ).
The patch add simple check if the destination pointer is NULL.
Signed-off-by: Stefan Mavrodiev <stefan@olimex.com> Acked-by: Jagan Teki <jagan@openedev.com>
[jagan: fix commit message] Signed-off-by: Jagan Teki <jagan@amarulasolutions.com>
Lukasz Majewski [Mon, 3 Dec 2018 09:20:47 +0000 (10:20 +0100)]
nand: vybrid: Use calloc() instead of malloc() to allocate struct nfc
Without this change it is possible that Vybrid's NFC driver malloc() call
will obtain some memory used (and correctly free'd) by some previous
driver (in this case pinctrl for Vybrid).
As a result some fields of struct nfc - in out case mtd->_get_device - are
"pre initialized" with some random values.
On the latter stage of booting, when e.g. somebody calls 'mtdparts default'
the "data abort" is observed when __get_mtd_device() function is called.
The mtd->_get_device pointer is not NULL and wrong value is referenced.
Signed-off-by: Lukasz Majewski <lukma@denx.de> Reviewed-by: Stefan Agner <stefan.agner@toradex.com>
Harald Seiler [Fri, 7 Dec 2018 09:19:36 +0000 (10:19 +0100)]
pcm058: fix NAND flash not using badblock table
Currently, U-Boot ignores the BBT stored in the last 4 blocks of NAND
flash because the NAND_BBT_USE_FLASH flag is not set. This leads to
two issues:
* U-Boot silently uses a memory-only BBT which is initialized with all
blocks marked as good. This means, actual bad blocks are marked good
and U-Boot might try writing to or reading from them.
* The BBT in flash, which will be created once Linux boots up, is not
off limits for a driver ontop, like UBI. While it does not seem to
consistently produce an error, sometimes UBI will fail to attach
because the BBT blocks obviously don't contain valid UBI data.
To fix this, this patch sets the CONFIG_SYS_NAND_USE_FLASH_BBT option,
which is used in ./drivers/mtd/nand/raw/mxs_nand.c to decide whether
a BBT in flash is used.
CPU: Freescale i.MX8MQ rev2.0 at 1000 MHz
Reset cause: POR
Model: Freescale i.MX8MQ EVK
DRAM: 3 GiB
MMC: FSL_SDHC: 0, FSL_SDHC: 1
In: serial
Out: serial
Err: serial
Net:
Warning: ethernet@30be0000 using MAC address from ROM
eth0: ethernet@30be0000
Hit any key to stop autoboot: 0
"
Peng Fan [Tue, 20 Nov 2018 10:19:46 +0000 (10:19 +0000)]
imx: imx8mq: build flash.bin
Build flash.bin for i.MX8MQ, it will include signed hdmi firmware,
spl, ddr firmware, fit image(bl31.bin, u-boot-nodtb.bin, dtb).
Burn it to 33KB offset of SD card.
Ye Li [Tue, 20 Nov 2018 10:19:15 +0000 (10:19 +0000)]
imx8m: Enable CONFIG_SPL_FIT_IMAGE_TINY for iMX8M
If we don't define CONFIG_SPL_FIT_IMAGE_TINY, when loading images from FIT,
the SPL will record all loadables' info to u-boot's FDT. This causes
problem when HAB is enabled, because FDT's content is modified before
we authenticate it.
Signed-off-by: Ye Li <ye.li@nxp.com> Signed-off-by: Peng Fan <peng.fan@nxp.com>
imx: hab: extend hab_auth_img to calculate ivt_offset
Current implementation of hab_auth_img command needs ivt_offset to
authenticate the image. But ivt header is placed at the end of image
date after padding.
This leaves the usage of hab_auth_img command to fixed size or static
offset for ivt header. New function "get_image_ivt_offset" is introduced
to find the ivt offset during runtime. The case conditional check in this
function is same as boot_get_kernel in common/bootm.c
With this variable length image e.g. FIT image with any random size can
have IVT at the end and ivt_offset option can be left optional
Can be used as "hab_auth_img $loadaddr $filesize" from u-boot script
Signed-off-by: Parthiban Nallathambi <pn@denx.de> Reviewed-by: Breno Lima <breno.lima@nxp.com>
Ye Li [Sat, 17 Nov 2018 09:10:25 +0000 (09:10 +0000)]
SPL: Add HAB image authentication to FIT
Introduce two board level callback functions to FIT image loading process, and
a SPL_FIT_FOUND flag to differentiate FIT image or RAW image.
Implement functions in imx common SPL codes to call HAB funtion
to authenticate the FIT image. Generally, we have to sign multiple regions
in FIT image:
1. Sign FIT FDT data (configuration)
2. Sign FIT external data (Sub-images)
Because the CSF supports to sign multiple memory blocks, so that we can use one
signature to cover all regions in FIT image and only authenticate once.
The authentication should be done after the entire FIT image is loaded into
memory including all sub-images.
We use "-p" option to generate FIT image to reserve a space for FIT IVT
and FIT CSF, also this help to fix the offset of the external data (u-boot-nodtb.bin,
ATF, u-boot DTB).
The signed FIT image layout is as below:
--------------------------------------------------
| | | | | | | |
| FIT | FIT | FIT | | U-BOOT | ATF | U-BOOT |
| FDT | IVT | CSF | | nodtb.bin | | DTB |
| | | | | | | |
--------------------------------------------------
Signed-off-by: Ye Li <ye.li@nxp.com> Reviewed-by: Peng Fan <peng.fan@nxp.com> Reviewed-by: Tom Rini <trini@konsulko.com> Signed-off-by: Peng Fan <peng.fan@nxp.com>
CONFIG_CMD_BOOTEFI is enabled by Kconfig default, but rarely
used on this board/platform.
So let's disable it for the boards default config.
This also saves around 16 KiB in the final u-boot.sb.
Signed-off-by: Michael Heimpold <mhei@heimpold.de>
Michael Heimpold [Mon, 29 Oct 2018 19:21:51 +0000 (20:21 +0100)]
configs: mx23_olinuxino_defconfig: fix status led definition
While migrating individual status led usages to Kconfig stuff,
a (random) value was introduced for this board which does not
work but produces the following error message during boot:
__led_init: failed requesting GPIO59!
Since Kconfig does not seem to accept a define as this point,
but the mxs gpio driver requires not only a simple integer value,
we need to use the plain value of MX23_PAD_SSP1_DETECT__GPIO_2_1.
Signed-off-by: Michael Heimpold <mhei@heimpold.de> Fixes: 2d8d190c8394 ("status_led: Kconfig migration")
Martin Fuzzey [Wed, 24 Oct 2018 08:21:18 +0000 (10:21 +0200)]
w1: Add driver for i.MX bus master controller
Two variants of controllers are supported:
V1 (bitwise only) found in
i.MX21, i.MX27, i.MX31, i.MX51
V2 (byte operations) found in
i.MX25, i.MX35, i.MX50, i.MX53
Only tested on i.MX53 hardware but in both modes
(by modifying the device tree).
Signed-off-by: Martin Fuzzey <martin.fuzzey@flowbird.group>
Xiaoliang Yang [Thu, 18 Oct 2018 10:27:46 +0000 (18:27 +0800)]
watchdog: imx: add config to disable wdog reset
Add Kconfig option WATCHDOG_RESET_DISABLE to disable watchdog reset
in imx_watchdog driver, so that the watchdog will not be fed in
u-boot if CONFIG_WATCHDOG_RESET_DISABLE is enabled.
Signed-off-by: Xiaoliang Yang <xiaoliang.yang_1@nxp.com>
Xiaoliang Yang [Thu, 18 Oct 2018 10:27:45 +0000 (18:27 +0800)]
watchdog: driver support for fsl-lsch2
Support watchdog driver for fsl-lsch2. It's disabled in default.
If you want to use it, please enable CONFIG_IMX_WATCHDOG.
Define CONFIG_WATCHDOG_TIMEOUT_MSECS to set watchdog timeout.
Signed-off-by: Xiaoliang Yang <xiaoliang.yang_1@nxp.com>
Denis Zalevskiy [Wed, 17 Oct 2018 08:33:30 +0000 (10:33 +0200)]
board: ge: Move VPD reading to the vpd_reader
Merge functionality duplicated in bx50v3 and mx53ppd: the logic
is the same except that process_vpd is called at different phases.
Also read_vpd could end up in error, so there is no VPD data in this
case - it shouldn't be processed.
Bin Meng [Sat, 22 Dec 2018 09:55:49 +0000 (01:55 -0800)]
fs: cbfs: Fix out of bound access during CBFS walking through
The call to file_cbfs_fill_cache() is given with the parameter
'start' pointing to the offset by the CBFS base address, but
with the parameter 'size' that equals to the whole CBFS size.
During CBFS walking through, it checks files one by one and
after it pass over the end of the CBFS which is 4GiB boundary
it tries to check files from address 0 and so on, until the
overall size the codes checked hits to the given 'size'.
Fix this by passing 'start' pointing to the CBFS base address.
Signed-off-by: Bin Meng <bmeng.cn@gmail.com> Reviewed-by: Simon Glass <sjg@chromium.org>
cbfs_fileheader.len indicates the content size of the file in the
cbfs, and it has nothing to do with cbfs_fileheader.offset which
is the starting address of the file in the cbfs.
Remove such check in file_cbfs_next_file(). Before this change
'cbfsinit' failed with 'Bad CBFS file'. After this change all cbfs
commands are working as expected.
Signed-off-by: Christian Gmeiner <christian.gmeiner@gmail.com>
[bmeng: keep the necessary header sanity check] Signed-off-by: Bin Meng <bmeng.cn@gmail.com> Reviewed-by: Simon Glass <sjg@chromium.org>
Revert "sunxi: board: Print error after power initialization fails"
Commit a8011eb84dfa("sunxi: board: Print error after power initialization
fails") moved the DRAM init after the increase of the CPU clock
frequency. This lead to various DRAM initialisation failures on some
boards (hangs or wrong size reported, on a NanoPi Duo2 and OrangePi
Zero, for instance). Lowering the CPU frequency significantly (for instance
to 408 MHz) seems to work around the problem, so this points to some timing
issues in the DRAM code.
Debugging this sounds like a larger job, so let's just revert this patch
to bring back those boards.
Beside this probably unintended change the patch just moved the error
message around, so reverting this is not a real loss.
Tested-By: Priit Laes <plaes@plaes.org> Signed-off-by: Karl Palsson <karlp@tweak.net.au> Signed-off-by: Andre Przywara <andre.przywara@arm.com> Reviewed-by: Jagan Teki <jagan@openedev.com>
Tom Rini [Sat, 29 Dec 2018 03:00:24 +0000 (22:00 -0500)]
Merge branch 'master' of git://git.denx.de/u-boot-uniphier
Please pull UniPhier updates v2019.01
- import DIV_ROUND_CLOSEST_ULL macro from Linux
- import improvement and fix of Denali NAND driver from Linux
- add NAND 200MHz clock to clk driver
- allow CONFIG_BOOTCOMMAND to run custom boot command/script
- sync DT with Linux 4.20
This commit improves the ->setup_data_interface() hook.
The denali_setup_data_interface() needs the frequency of clk_x
and the ratio of clk_x / clk.
The latter is currently hardcoded in the driver, like this:
#define DENALI_CLK_X_MULT 6
The IP datasheet requires that clk_x / clk be 4, 5, or 6. I just
chose 6 because it is the most defensive value, but it is not optimal.
By getting the clock rate of both "clk" and "clk_x", the driver can
compute the timing values more precisely.
To not break the existing platforms, the fallback value, 50 MHz is
provided. It is true for all upstreamed platforms.
Currently, denali_dt.c requires a single anonymous clock, but
the Denali User's Guide requires three clocks for this IP:
- clk: controller core clock
- clk_x: bus interface clock
- ecc_clk: clock at which ECC circuitry is run
This commit supports these named clocks to represent the real hardware.
For the backward compatibility, the driver still accepts a single clock
just as before. The clk_x_rate is taken from the clock driver again if
the named clock "clk_x" is available. This will happen only for future
DT, hence the existing DT files are not affected.
Masahiro Yamada [Wed, 19 Dec 2018 11:03:14 +0000 (20:03 +0900)]
ARM: uniphier: allow to source boot script before distro-boot
Some users might need additional setups before booting the kernel.
If there is found a file 'boot.scr', run it before invoking the
distro boot command.
Lokesh Vutla [Mon, 24 Dec 2018 11:07:41 +0000 (16:37 +0530)]
power: regulator: Return success on attempt to disable an always-on regulator
commit 4f86a724e82c0 ("power: regulator: denied disable on always-on
regulator") throws an error when requested to disable an always-on
regulator. It is right that an always-on regulator should not be
attempted to be disabled. But at the same time regulator framework
should not return an error when such request is received. Instead
it should just return success without attempting to disable the
specified regulator. This is because the requesting driver will
not have the idea if the regulator is always-on or not. The
requesting driver will always try to enable/disable regulator as
per the required flow. So it is upto regulator framework to not
break such scenarios.
Michal Simek [Wed, 19 Dec 2018 15:57:38 +0000 (16:57 +0100)]
phy: Fix u-boot coruption when fixed-phy is used
When fixed-link phy is used subnode offset is used as phy address. This
number is bigger then space allocated for bus structure (allocated via
mdio_alloc).
bus->phymap[] array has PHY_MAX_ADDR size (32).
That's why writing bus->phymap[addr] where addr is < 0 or > PHY_MAX_ADDR
is causing write to memory which can caused full U-Boot crash.
The patch is checking if address is in correct range.
Signed-off-by: Michal Simek <michal.simek@xilinx.com>
Lokesh Vutla [Wed, 19 Dec 2018 07:23:31 +0000 (12:53 +0530)]
arm: K3: Fix usage of CONFIG_SYS_K3_KEY
For signing the tiboot3.bin image, an optional KEY file can be passed
using CONFIG_SYS_K3_KEY. Right now, Makefile scripts directly takes
the config value and uses it for signing. This is okay if the build
directory is a sub-directory of source tree, otherwise it fails.
Fix it by using the path relative to the source tree.
dev_stor_get() is not able to find the next available device in the current
storage group when the previous enumerated device belongs to a different
group or class (e.g. network).
The root cause is the device group iterator not being reset after an
unsuccessful search for the last returned device so that the following
search for the next available device will start from beginning.
The issue has been identified by loading and booting GRUB in a QEMU
vexpress-a9 environment.
Lukasz Majewski [Tue, 18 Dec 2018 11:30:50 +0000 (12:30 +0100)]
dm: pinctrl: Prevent (re-)configuring pins when already done before relocation
This commit prevents from re-configuring pins if those were configured
before relocation.
Some pins - like UART or DDR must be setup before relocation
(as they have 'u-boot,dm-pre-reloc' property set in DTS). Without this
change, those pins are re-configured after relocation (pre_reloc_only = 0,
so we do not "continue").
Such behavior may be a problem for DDR PAD configuration, as they might
be already leveled/tuned with original setup).
Adam Ford [Fri, 14 Dec 2018 22:28:30 +0000 (16:28 -0600)]
ARM: mach-omap2: omap3: Fix GPIO clocking in SPL
OMAP3_GPIO_x is needed to enable each GPIO bank on the OMAP3
boards. At one point, the #ifdef's were replaced with
if CONFIG_IS_ENABLED but this won't work for people who need
OMAP3_GPIO_x in SPL since the SPL prefix for this option isn't
used in Kconfig. This patch moves the check to #if defined and
also makes Kconfig select the banks if CMD_GPIO is used which
makes the checks in the code less cumbersome.
Fixes: bd8a9c14c91c ("arm: mach-omap2/omap3/clock.c: Enable
all GPIO with CMD_GPIO")
Reported-by: Liam O'Shaughnessy <liam.o.shaughnessy@gumstix.com> Signed-off-by: Adam Ford <aford173@gmail.com>
[trini: Migrate omap3_igep00x0.h] Signed-off-by: Tom Rini <trini@konsulko.com>
Alexander Graf [Tue, 11 Dec 2018 09:00:42 +0000 (10:00 +0100)]
efi_loader: Make RTS relocation more robust
While changing the RTS alignment to 64KB in commit 7a82c3051c8f
("efi_loader: Align runtime section to 64kb") the relocation code
started to break.
The reason for that is that we didn't actually look at the real
relocation data. We merely took the RUNTIME_CODE section as a
hint and started to relocate based on self calculated data from
that point on. That calculation was now out of sync though.
To ensure we're not running into such a situation again, this patch
makes the runtime relocation code a bit more robust. We can just
trust the phys/virt hints from the payload. We also should check that
we really only have a single section, as the code doesn't handle
multiple code relocations yet.
Fixes: 7a82c3051c8f ("efi_loader: Align runtime section to 64kb") Reported-by: Heinrich Schuchardt <xypron.glpk@gmx.de> Reported-by: Loic Devulder <ldevulder@suse.de> Signed-off-by: Alexander Graf <agraf@suse.de> Reviewed-by: Heinrich Schuchardt <xypron.glpk@gmx.de> Tested-by: Loic Devulder <ldevulder@suse.de> Tested-by: Jonathan Gray <jsg@jsg.id.au> Signed-off-by: Alexander Graf <agraf@suse.de>
lib: crc32: mark static variable as __efi_runtime_data
In commit 483dbab9f931 ("lib: crc32: mark function crc32() as
__efi_runtime") one local static variable was missed. It did not end up in
the __efi_runtime_data section as it should.
If CONFIG_DYNAMIC_CRC_TABLE=y a data abort execption may occur when the
UEFI payload calls the SetVirtualAddressMap() runtime service.
Reported-by: Dominik Adamski <dominik.adamski@northerntech.community> Fixes: 483dbab9f931 ("lib: crc32: mark function crc32() as __efi_runtime") Signed-off-by: Heinrich Schuchardt <xypron.glpk@gmx.de> Signed-off-by: Alexander Graf <agraf@suse.de>
projects / u-boot.git / log