platform/kernel/u-boot.git
2 years agoMerge https://source.denx.de/u-boot/custodians/u-boot-marvell
Tom Rini [Sun, 18 Sep 2022 12:27:23 +0000 (08:27 -0400)]
Merge https://source.denx.de/u-boot/custodians/u-boot-marvell

- mvebu: turris_omnia: Fix setting switch CONFIG pins on new board
  design (Marek)
- orion-timer: Use timer_conv_64() to fix timer wrap around (Stefan)

2 years agotimer: orion-timer: Use timer_conv_64() to fix timer wrap around
Stefan Roese [Thu, 15 Sep 2022 14:20:36 +0000 (16:20 +0200)]
timer: orion-timer: Use timer_conv_64() to fix timer wrap around

While testing on some Kirkwood platforms it was noticed that the timer
did not function correctly all the time. The driver did not correctly
handle 32bit timer value wrap arounds. Using the timer_conv_64()
conversion function fixes this issue.

Fixes: e9e73d78a8fb ("timer: add orion-timer support")
Suggested-by: Stefan Herbrechtsmeier <stefan.herbrechtsmeier-oss@weidmueller.com>
Signed-off-by: Stefan Roese <sr@denx.de>
Tested-by: Tony Dinh <mibodhi@gmail.com>
2 years agoarm: mvebu: turris_omnia: Fix setting switch CONFIG pins on new board design
Marek Behún [Tue, 13 Sep 2022 16:10:28 +0000 (18:10 +0200)]
arm: mvebu: turris_omnia: Fix setting switch CONFIG pins on new board design

It seems that waiting only 10 ms after releasing LAN switch from reset
is not enough for the strapping pins to latch the requested values.
P6_MODE[0] is latched to 0 instead of 1.

Increasing the delay to 50 ms fixes this issue.

Signed-off-by: Marek Behún <kabel@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
2 years agoMerge tag 'u-boot-stm32-20220915' of https://source.denx.de/u-boot/custodians/u-boot-stm
Tom Rini [Thu, 15 Sep 2022 16:17:49 +0000 (12:17 -0400)]
Merge tag 'u-boot-stm32-20220915' of https://source.denx.de/u-boot/custodians/u-boot-stm

- Fixes on STM32 I2C drivers
- Activate SCMI regulator for STM32MP15 defconfig, fix the usb start command
  for scmi device tree

2 years agoi2c: stm32: fix usage of rise/fall device tree properties
Jorge Ramirez-Ortiz [Mon, 12 Sep 2022 08:42:01 +0000 (10:42 +0200)]
i2c: stm32: fix usage of rise/fall device tree properties

These two device tree properties were not being applied.

Fixes: 1fd9eb68d6 ("i2c: stm32f7: move driver data of each instance in a privdata")
Signed-off-by: Jorge Ramirez-Ortiz <jorge@foundries.io>
Reviewed-by: Alain Volmat <alain.volmat@foss.st.com>
Reviewed-by: Patrice Chotard <patrice.chotard@foss.st.com>
Reviewed-by: Heiko Schocher <hs@denx.de>
Reviewed-by: Patrick Delaunay <patrick.delaunay@foss.st.com>
Tested-by: Patrick Delaunay <patrick.delaunay@foss.st.com>
2 years agoi2c: stm32: do not set the STOP condition on error
Alain Volmat [Mon, 12 Sep 2022 08:42:00 +0000 (10:42 +0200)]
i2c: stm32: do not set the STOP condition on error

Current function stm32_i2c_message_xfer is sending a STOP
whatever the result of the transaction is.  This can cause issues
such as making the bus busy since the controller itself is already
sending automatically a STOP when a NACK is generated.

Thanks to Jorge Ramirez-Ortiz for diagnosing and proposing a first
fix for this. [1]

[1] https://lore.kernel.org/u-boot/20220815145211.31342-2-jorge@foundries.io/

Reported-by: Jorge Ramirez-Ortiz, Foundries <jorge@foundries.io>
Signed-off-by: Jorge Ramirez-Ortiz <jorge@foundries.io>
Signed-off-by: Alain Volmat <alain.volmat@foss.st.com>
Reviewed-by: Patrice Chotard <patrice.chotard@foss.st.com>
Reviewed-by: Heiko Schocher <hs@denx.de>
Reviewed-by: Patrick Delaunay <patrick.delaunay@foss.st.com>
Tested-by: Patrick Delaunay <patrick.delaunay@foss.st.com>
2 years agoi2c: stm32: remove unused stop parameter in start & reload handling
Alain Volmat [Mon, 12 Sep 2022 08:41:59 +0000 (10:41 +0200)]
i2c: stm32: remove unused stop parameter in start & reload handling

Functions stm32_i2c_message_start and stm32_i2c_handle_reload
both get a stop boolean indicating if the transfer should end with
a STOP or not.  However no specific handling is needed in those
functions hence remove the parameter.

Signed-off-by: Alain Volmat <alain.volmat@foss.st.com>
Reviewed-by: Patrick Delaunay <patrick.delaunay@foss.st.com>
Reviewed-by: Heiko Schocher <hs@denx.de>
Reviewed-by: Patrice Chotard <patrice.chotard@foss.st.com>
2 years agoi2c: stm32: fix comment and remove unused AUTOEND bit
Alain Volmat [Mon, 12 Sep 2022 08:41:58 +0000 (10:41 +0200)]
i2c: stm32: fix comment and remove unused AUTOEND bit

Comment within stm32_i2c_message_start is misleading, indicating
that AUTOEND bit is setted while it is actually cleared.
Moreover, the bit is actually never setted so there is no need
to clear it hence get rid of this bit clear and the bit macro
as well.

Signed-off-by: Alain Volmat <alain.volmat@foss.st.com>
Reviewed-by: Patrick Delaunay <patrick.delaunay@foss.st.com>
Reviewed-by: Heiko Schocher <hs@denx.de>
Reviewed-by: Patrice Chotard <patrice.chotard@foss.st.com>
2 years agoi2c: stm32f7: fix clearing the control register
Jorge Ramirez-Ortiz [Mon, 15 Aug 2022 14:52:10 +0000 (16:52 +0200)]
i2c: stm32f7: fix clearing the control register

Bits should be set to 0, not 1.

Signed-off-by: Jorge Ramirez-Ortiz <jorge@foundries.io>
Reviewed-by: Patrice Chotard <patrice.chotard@foss.st.com>
Reviewed-by: Patrick Delaunay <patrick.delaunay@foss.st.com>
2 years agoconfigs: stm32mp15: activate DM_REGULATOR_SCMI
Patrick Delaunay [Wed, 7 Sep 2022 16:18:40 +0000 (18:18 +0200)]
configs: stm32mp15: activate DM_REGULATOR_SCMI

Activate the support of SCMI regulator to support the scmi_reg11,
scmi_reg18 and scmi_usb33 regulators present in the scmi device tree of
STMicroelectronics boards with stm32mp15-scmi.dtsi

Fixes: 6cccc8d396bf ("ARM: dts: stm32: add SCMI version of STM32 boards (DK1/DK2/ED1/EV1)")
Signed-off-by: Patrick Delaunay <patrick.delaunay@foss.st.com>
Reviewed-by: Patrice Chotard <patrice.chotard@foss.st.com>
2 years agoMerge https://source.denx.de/u-boot/custodians/u-boot-marvell
Tom Rini [Tue, 13 Sep 2022 12:19:42 +0000 (08:19 -0400)]
Merge https://source.denx.de/u-boot/custodians/u-boot-marvell

- mvebu: Armada 32bit: Cache setup fixes (Pali)
- cmd: mvebu/bubt: Misc enhancements (Pali)
- kirkwood: Add CONFIG_SUPPORT_PASSING_ATAGS (Tony)
- board: turris: Misc improvements (Pali)
- tools: kwboot: Change KWBOOT_MSG_RSP_TIMEO_AXP to 10ms (Stefan)
- tools: termios_linux.h: Fix compilation on non-glibc systems (Pali)

2 years agoarm: mvebu: Fix moving internal registers
Pali Rohár [Thu, 8 Sep 2022 14:06:54 +0000 (16:06 +0200)]
arm: mvebu: Fix moving internal registers

Commit 5bb2c550b11e ("arm: mvebu: Move internal registers in
arch_very_early_init() function") moved code from file cpu.c to lowlevel.c,
which moves Marvell internal registers from address INTREG_BASE_ADDR_REG to
SOC_REGS_PHY_BASE.

But the steps describing how to do it correctly were documented only in
older U-Boot versions and commit cefd764222ee ("arm: mvebu: Fix internal
register config on A38x") probably unintentionally removed important
details about MMU from code comments around.

Commit 5bb2c550b11e ("arm: mvebu: Move internal registers in
arch_very_early_init() function") implemented code movement according to
(now incomplete) comments which resulted in semi-broken code.

The result is that I-cache is currently disabled for all Armada 38x boards
and maybe there are some other (unreported / undetected) issues.

Reimplement it correctly. First flush all caches, then disable MMU and L2
cache and then move Marvell internal registers. There is no need to
explicitly disable I-cache.

After this change lzmadec command with lzma image of 0x7000000 bytes is
doing decompression just 5 seconds. Before this change it was 30 seconds.

To make lowlevel.S code more readable, extend asm/pl310.h header file to be
compatible with assembler and use macros from this file.

Fixes: 5bb2c550b11e ("arm: mvebu: Move internal registers in arch_very_early_init() function")
Signed-off-by: Pali Rohár <pali@kernel.org>
Signed-off-by: Marek Behún <kabel@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
2 years agoarm: mvebu: Enable L2 cache also on Armada 38x
Pali Rohár [Thu, 8 Sep 2022 14:06:53 +0000 (16:06 +0200)]
arm: mvebu: Enable L2 cache also on Armada 38x

For some unknown reason when L2 cache is disabled on Armada 385 then loadb,
loadx and loady commands do not work with higher baudrates than 115200
(they just abort transfer) and lzmadec command with lzma image of size
0x7000000 (maybe even smaller, we tested this one) is doing decompression
for more than 2 minutes. After enabling L2 cache decompression takes only
30s and loadb, loadx and loady are stable and working fine.

git bisect identified problematic commit 3308933d2fe9 ("arm: mvebu: Avoid
reading MVEBU_REG_PCIE_DEVID register too many times"). Before this commit
above issues were not present.

But investigation showed that above issue was possible to reproduce also by
reverting that commit and forcing compiler to do inline optimization of
mvebu_soc_family() function. Which seems that the root of this issue is in
caches and position of instruction of segments. So currently it is unknown
what is or was broken, but code movement, code inlining or other compiler
optimization triggered it.

Commit 3e5ce7ceeb94 ("arm: mvebu: Enable L2 cache on Armada XP") mentioned
that enabling L2 cache on Armada XP improved performance and that Armada
38x has L2 disabled (which is default state) and if needed it has to be
enabled in separate patch. As enabling L2 cache also improve performance
on Armada 38x, enable it.

Note that Aurora cache in no outer mode is available only on Armada XP,
hence it is not touched for Armada 38x code.

Fixes: 3308933d2fe9 ("arm: mvebu: Avoid reading MVEBU_REG_PCIE_DEVID register too many times")
Reported-by: Marek Behún <kabel@kernel.org>
Signed-off-by: Pali Rohár <pali@kernel.org>
Signed-off-by: Marek Behún <kabel@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
2 years agoarm: mvebu: lowlevel.S: Use CR_M from asm/system.h
Pali Rohár [Thu, 8 Sep 2022 14:06:52 +0000 (16:06 +0200)]
arm: mvebu: lowlevel.S: Use CR_M from asm/system.h

Replace magic constant 1 when disabling MMU by macro CR_M from include
header file asm/system.h.

Signed-off-by: Pali Rohár <pali@kernel.org>
Signed-off-by: Marek Behún <kabel@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
2 years agoarm: mvebu: Guard non-AXP code by checking for AXP
Pali Rohár [Thu, 8 Sep 2022 14:06:51 +0000 (16:06 +0200)]
arm: mvebu: Guard non-AXP code by checking for AXP

Commit c86d53fd88df ("arm: mvebu: Don't disable cache at startup on Armada
XP at all") introduced branch for non-AXP code which was guarded by A38X
condition. Fix this issue by checking for AXP platform, not by A38X.

Fixes: c86d53fd88df ("arm: mvebu: Don't disable cache at startup on Armada XP at all")
Signed-off-by: Pali Rohár <pali@kernel.org>
Signed-off-by: Marek Behún <kabel@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
2 years agoarm: mvebu: Fix function enable_caches
Pali Rohár [Thu, 8 Sep 2022 14:06:50 +0000 (16:06 +0200)]
arm: mvebu: Fix function enable_caches

Commit 3308933d2fe9 ("arm: mvebu: Avoid reading MVEBU_REG_PCIE_DEVID
register too many times") broke support for caches on all Armada SoCs.

Before that commit there was code:

    if (mvebu_soc_family() != MVEBU_SOC_A375) {
        dcache_enable();
    }

And after that commit there is code:

    if (IS_ENABLED(CONFIG_ARMADA_375)) {
        dcache_enable();
    }

Comment above this code says that d-cache should be disabled on Armada 375.
But new code inverted logic and broke Armada 375 and slowed down all other
Armada SoCs (including A38x).

Fix this issue by changing logic to:

    if (!IS_ENABLED(CONFIG_ARMADA_375)) {
        dcache_enable();
    }

Which matches behavior prior that commit.

Fixes: 3308933d2fe9 ("arm: mvebu: Avoid reading MVEBU_REG_PCIE_DEVID register too many times")
Signed-off-by: Pali Rohár <pali@kernel.org>
Signed-off-by: Marek Behún <kabel@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
2 years agoarm: mvebu: Mark constant data with const keyword
Pali Rohár [Fri, 9 Sep 2022 12:41:28 +0000 (14:41 +0200)]
arm: mvebu: Mark constant data with const keyword

Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
2 years agoarm: mvebu: turris_omnia: Allow to use second serial port
Pali Rohár [Fri, 9 Sep 2022 12:18:48 +0000 (14:18 +0200)]
arm: mvebu: turris_omnia: Allow to use second serial port

Turris Omnia has two serial ports. Both are already specified in device
tree file. But U-Boot by default does not allow to use more than one serial
port unless CONFIG_SERIAL_PROBE_ALL is not enabled.

After enabling CONFIG_SERIAL_PROBE_ALL, U-Boot see also second serial port
(but is inactive by default):

    => coninfo
    List of available devices:
    serial@12000 00000007 IO stdin stdout stderr
    serial@12100 00000007 IO

To allow simultaneously to use more input / output devices it is needed to
enable CONFIG_CONSOLE_MUX option.

With CONFIG_CONSOLE_MUX it is possible to call:

    => setenv stdout 'serial@12000,serial@12100'

And U-Boot output is then visible on both serial ports.

Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
2 years agotools: termios_linux.h: Fix compilation on non-glibc systems
Pali Rohár [Thu, 8 Sep 2022 14:59:36 +0000 (16:59 +0200)]
tools: termios_linux.h: Fix compilation on non-glibc systems

TCGETS2 is defined in header file asm/ioctls.h provided by linux kernel.
On glib systems it is automatically included by some other glibc include
header file and therefore TCGETS2 is present in termios_linux.h when
linux kernel provides it.

On non-glibc systems (e.g. musl) asm/ioctls.h is not automatically included
which results in the strange error that BOTHER is supported, TCGETS2 not
defined and struct termios does not provide c_ispeed member.

    tools/kwboot.c: In function 'kwboot_tty_change_baudrate':
    tools/kwboot.c:662:6: error: 'struct termios' has no member named 'c_ospeed'
      662 |   tio.c_ospeed = tio.c_ispeed = baudrate;
          |      ^

Fix this issue by explicitly including asm/ioctls.h file which provides
TCGETS2 macro (if supported on selected architecture) to not depending on
glibc auto-include behavior and because termios_linux.h requires it.

With this change it is possible compile kwboot with musl libc.

Reported-by: Michal Vasilek <michal.vasilek@nic.cz>
Signed-off-by: Pali Rohár <pali@kernel.org>
2 years agoarm: mvebu: turris_mox: Add support for distroboot $fdt_addr
Pali Rohár [Mon, 29 Aug 2022 13:44:48 +0000 (15:44 +0200)]
arm: mvebu: turris_mox: Add support for distroboot $fdt_addr

$fdt_addr is mandatory for systems which provides DTB in HW (e.g. ROM) and
wishes to pass that DTB to Linux.

Turris Mox contains DTB binary in SPI NOR memory at "dtb" partition which
starts at offset 0x7f0000 and is 0x10000 bytes long.

Armada 3700 CPU does not allow mapping SPI NOR memory into physical address
space like on other architectures and therefore set $fdt_addr variable to
memory range in RAM and loads this DTB binary from SPI NOR in misc_init_r()
function.

Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
2 years agoarm: mvebu: turris_{omnia, mox}: Reset bootdelay env for rescue
Pali Rohár [Sat, 27 Aug 2022 18:49:20 +0000 (20:49 +0200)]
arm: mvebu: turris_{omnia, mox}: Reset bootdelay env for rescue

When rescue mode was activated reset also bootdelay env variable to its
default value. This will ensure that reset button works and starts rescue
mode also in the case when user changed bootdelay env variable to -1 (which
has meaning to not start autoboot).

Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Marek Behún <kabel@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
2 years agoboard: turris: Initialize serial# env
Pali Rohár [Sat, 27 Aug 2022 18:06:30 +0000 (20:06 +0200)]
board: turris: Initialize serial# env

Store serial number from atsha cryptochip into the serial# env variable.
U-Boot automatically puts content of this variable into the root device
tree property serial-number when booting Linux kernel. Refactor turris
atsha code and from turris_atsha_otp_get_serial_number() function returns
directly string suitable for printing or storing into device tree. Because
during different boot stages is env storage read-only, it is not possible
to always store serial number into env storage. So introduce a new function
turris_atsha_otp_init_serial_number() which is called at later stage and
which ensures that serial number is correctly stored into env.

Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Marek Behún <kabel@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
2 years agoarm: mvebu: Espressobin: When emmc is not present disable it also in OF_LIVE
Pali Rohár [Sat, 27 Aug 2022 12:00:51 +0000 (14:00 +0200)]
arm: mvebu: Espressobin: When emmc is not present disable it also in OF_LIVE

Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
2 years agocmd: mvebu/bubt: Check for A38x/A37xx OTP secure bits and secure boot
Pali Rohár [Tue, 23 Aug 2022 12:52:24 +0000 (14:52 +0200)]
cmd: mvebu/bubt: Check for A38x/A37xx OTP secure bits and secure boot

For obvious reasons BootROMS rejects unsigned images when secure boot is
enabled in OTP secure bits. So check for OPT secure bits and do not allow
flashing unsigned images when secure boot is enabled. Access to OTP via
U-Boot fuse API is currently implemented only for A38x and A37xx SoCs.

Additionally Armada 3700 BootROM rejects signed trusted image when secure
boot is not enabled in OTP. So add also check for this case. On the other
hand Armada 38x BootROM acceps images with secure boot header when secure
boot is not enabled in OTP.

OTP secure bits may have burned also boot device source. Check it also and
reject flashing images to target storage which does not match OTP.

Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
2 years agocmd: mvebu/bubt: Check for A38x image data checksum
Pali Rohár [Tue, 23 Aug 2022 12:52:23 +0000 (14:52 +0200)]
cmd: mvebu/bubt: Check for A38x image data checksum

Currently for A38x image is checked only header checksum.
So check also for image data checksum to prevent flashing broken image.

Signed-off-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Stefan Roese <sr@denx.de>
2 years agotools: kwboot: Change KWBOOT_MSG_RSP_TIMEO_AXP to 10ms
Stefan Roese [Fri, 19 Aug 2022 07:43:59 +0000 (09:43 +0200)]
tools: kwboot: Change KWBOOT_MSG_RSP_TIMEO_AXP to 10ms

Testing on the theadorable Armada XP platform has shown, thaz using the
current value of 1000ms as response timeout does not result in reliable
booting via kwboot. Using 10ms seems to be much better. So let's change
this value to this 10ms instead.

Signed-off-by: Stefan Roese <sr@denx.de>
Acked-by: Pali Rohár <pali@kernel.org>
2 years agoarm: kirkwood: Add CONFIG_SUPPORT_PASSING_ATAGS
Tony Dinh [Wed, 17 Aug 2022 21:59:44 +0000 (14:59 -0700)]
arm: kirkwood: Add CONFIG_SUPPORT_PASSING_ATAGS

Add CONFIG_SUPPORT_PASSING_ATAGS and friends to support legacy
image method of booting. Debian and OpenWrt installer use uImage
with appended DTB for these Kirkwood boards.

Signed-off-by: Tony Dinh <mibodhi@gmail.com>
Reviewed-by: Stefan Roese <sr@denx.de>
2 years agoMerge tag 'dm-pull-12sep22' of https://source.denx.de/u-boot/custodians/u-boot-dm
Tom Rini [Mon, 12 Sep 2022 13:47:16 +0000 (09:47 -0400)]
Merge tag 'dm-pull-12sep22' of https://source.denx.de/u-boot/custodians/u-boot-dm

Binman VPL support (patch was lost)
Add board_rng_seed() as a temporary solution

2 years agobinman: Add VPL support
Simon Glass [Mon, 28 Feb 2022 14:16:54 +0000 (07:16 -0700)]
binman: Add VPL support

Add support for U-Boot's Verifying Program Loader phase.

Signed-off-by: Simon Glass <sjg@chromium.org>
2 years agofdt_support: add optional board_rng_seed() hook
Rasmus Villemoes [Mon, 22 Aug 2022 07:34:23 +0000 (09:34 +0200)]
fdt_support: add optional board_rng_seed() hook

A recurring theme on LKML is the boot process deadlocking due to some
process blocking waiting for random numbers, while the kernel's
Cryptographic Random Number Generator (crng) is not initalized yet,
but that very blocking means no activity happens that would generate
the entropy necessary to finalize seeding the crng.

This is not a problem on boards that have a good hwrng (when the
kernel is configured to trust it), whether in the CPU or in a TPM or
elsewhere. However, that's far from all boards out there. Moreover,
there are consumers in the kernel that try to obtain random numbers
very early, before the kernel has had any chance to initialize any
hwrng or other peripherals.

Allow a board to provide a board_rng_seed() function, which is
responsible for providing a value to be put into the rng-seed property
under the /chosen node.

The board code is responsible for how to actually obtain those
bytes.

- One possibility is for the board to load a seed "file" from
  somewhere (it need not be a file in a filesystem of course), and
  then ensure that that the same seed file does not get used on
  subsequent boots.

  * One way to do that is to delete the file, or otherwise mark it as
    invalid, then rely on userspace to create a new one, and living
    with the possibility of not finding a seed file during some boots.

  * Another is to use the scheme used by systemd-boot and create a new
    seed file immediately, but in a way that the seed passed to the
    kernel and the new (i.e. next) seed cannot be deduced from each
    other, see the explanation at
    https://lore.kernel.org/lkml/20190929090512.GB13049@gardel-login/
    and the current code at
    https://github.com/systemd/systemd/blob/main/src/boot/efi/random-seed.c

- The board may have an hwrng from which some bytes can be read; while
  the kernel can also do that, doing it in U-Boot and providing a seed
  ensures that even very early users in the kernel get good random
  numbers.

- If the board has a sensor of some sort (temperature, humidity, GPS,
  RTC, whatever), mixing in a reading of that doesn't hurt.

- etc. etc.

These can of course be combined.

The rng-seed property is mixed into the pool used by the linux
kernel's CRNG very early during boot. Whether it then actually
contributes towards the kernel considering the CRNG initialized
depends on whether the kernel has been configured with
CONFIG_RANDOM_TRUST_BOOTLOADER (nowadays overridable via the
random.trust_bootloader command line option). But that's for the BSP
developer to ultimately decide.

So, if the board needs to have all that logic, why not also just have
it do the actual population of /chosen/rng-seed in ft_board_setup(),
which is not that many extra lines of code?

I considered that, but decided handling this logically belongs in
fdt_chosen(). Also, apart from saving the board code from the few
lines of boilerplate, doing it in ft_board_setup() is too late for at
least some use cases. For example, I want to allow the board logic to
decide

  ok, let's pass back this buffer and use that as seed, but also let's
  set random.trust_bootloader=n so no entropy is credited.

This requires the rng-seed handling to happen before bootargs
handling. For example, during the very first boot, the board might not
have a proper seed file, but the board could still return (a hash of)
some CPU serial# or whatnot, so that at least no two boards ever get
the same seed - the kernel always mixes in the value passed in
rng-seed, but if it is not "trusted", the kernel would still go
through the same motions as it would if no rng-seed was passed before
considering its CRNG initialized. I.e., by returning that
unique-to-this-board value and setting random.trust_bootloader=n, the
board would be no worse off than if board_rng_seed() returned nothing
at all.

Signed-off-by: Rasmus Villemoes <rasmus.villemoes@prevas.dk>
2 years agoMerge tag 'efi-2022-10-rc5' of https://source.denx.de/u-boot/custodians/u-boot-efi
Tom Rini [Fri, 9 Sep 2022 18:10:51 +0000 (14:10 -0400)]
Merge tag 'efi-2022-10-rc5' of https://source.denx.de/u-boot/custodians/u-boot-efi

Pull request for efi-2022-10-rc5

Documentation:

* man-page for tftpput

UEFI:

* fix driver binding protocol for block IO devices
* don't delete invalid handles
* add a unit test for the EFI Conformance Profile Table

Other:

* correct short text for tftpboot

2 years agoefi_driver: don't bind internal block devices
Heinrich Schuchardt [Fri, 9 Sep 2022 06:57:58 +0000 (06:57 +0000)]
efi_driver: don't bind internal block devices

UEFI block devices can either mirror U-Boot's internal devices or be
provided by an EFI application like iPXE.

When ConnectController() is invoked for the EFI_BLOCK_IO_PROTOCOL
interface for such an application provided device we create a virtual
U-Boot block device of type "efi_blk".

Currently we do not call ConnectController() when handles for U-Boot's
internal block devices are created. If an EFI application calls
ConnectController() for a handle relating to an internal block device,
we erroneously create an extra "efi_blk" block device.

E.g. the UEFI shell has a command 'connect -r' which calls
ConnectController() for all handles with device path protocol.

In the Supported() method of our EFI_DRIVER_BINDING_PROTOCOL return
EFI_UNSUPPORTED when dealing with an U-Boot internal device.

Reported-by: Etienne Carriere <etienne.carriere@linaro.org>
Fixes: commit 05ef48a2484b ("efi_driver: EFI block driver")
Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
Reviewed-by: Etienne Carriere <etienne.carriere@linaro.org>
Tested-by: Etienne Carriere <etienne.carriere@linaro.org>
Reviewed-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
2 years agolib: efi_loader: don't delete invalid handles
Etienne Carriere [Wed, 7 Sep 2022 08:20:13 +0000 (10:20 +0200)]
lib: efi_loader: don't delete invalid handles

Change efi_delete_handle() to not free EFI handles twice.

This change tries to resolved an issue seen since U-Boot v2022.07
in which ExitBootService() attempts to release some EFI handles twice.

The issue was seen booting a EFI shell that invokes 'connect -r' and
then boots a Linux kernel. Execution of connect command makes EFI
subsystem to bind a block device for each root block devices EFI handles.
However these EFI device handles are already bound to a driver and we
can have 2 registered devices relating to the same EFI handler. On
ExitBootService(), the loop removing the devices makes these EFI handles
to be released twice which corrupts memory.

This patch prevents the memory release operation caused by the issue but
but does not resolve the underlying problem.

Signed-off-by: Etienne Carriere <etienne.carriere@linaro.org>
Add log message.
Reviewed-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
Reviewed-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
2 years agoefi_selftest: unit test for EFI Conformance Profile Table
Heinrich Schuchardt [Sat, 3 Sep 2022 13:58:19 +0000 (15:58 +0200)]
efi_selftest: unit test for EFI Conformance Profile Table

Add a new unit test to test the integrity of the
EFI Conformance Profile Table.

Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
Reviewed-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
2 years agoefi_selftest: export efi_st_get_config_table()
Heinrich Schuchardt [Sat, 3 Sep 2022 13:56:51 +0000 (15:56 +0200)]
efi_selftest: export efi_st_get_config_table()

We can use efi_st_get_config_table() in multiple unit tests.
Export the function.

Export system-table and boot-services.

Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
2 years agodoc: man-page for tftpput
Heinrich Schuchardt [Sat, 3 Sep 2022 11:31:04 +0000 (13:31 +0200)]
doc: man-page for tftpput

Provide a man-page for the tftpput command.

Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
2 years agocmd: correct short text for tftpboot
Heinrich Schuchardt [Sun, 4 Sep 2022 07:08:11 +0000 (09:08 +0200)]
cmd: correct short text for tftpboot

The command's name is a misnomer.
The command loads a file but does not run (boot) it.

Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
2 years agocmd: fix tftpput command
Heinrich Schuchardt [Sat, 3 Sep 2022 12:21:09 +0000 (12:21 +0000)]
cmd: fix tftpput command

Calling tftpput with less than 2 arguments must lead to a failure.

If tftpput is called with two arguments, these are the address and
the size of the file to be transferred.

Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
2 years agoMerge tag 'u-boot-stm32-20220907' of https://source.denx.de/u-boot/custodians/u-boot-stm
Tom Rini [Thu, 8 Sep 2022 12:33:41 +0000 (08:33 -0400)]
Merge tag 'u-boot-stm32-20220907' of https://source.denx.de/u-boot/custodians/u-boot-stm

- simplify the STM32MP15x package parsing code
- remove test on CONFIG_DM_REGULATOR in stm32mp1 board
  and enable CONFIG_DM_REGULATOR for stm32f769-disco
- handle ck_usbo_48m clock provided by USBPHYC to fix the command 'usb start'
  after alignment with Linux kernel v5.19 DT (clocks = <&usbphyc>)
- Fix SYS_HZ_CLOCK value for stih410-b2260 board
- Switch STMM32MP15x DHSOM to FMC2 EBI driver
- Remove hwlocks from pinctrl in STM32MP15x to avoid issue with kernel

2 years agoMerge tag 'fsl-qoriq-2022-9-7' of https://source.denx.de/u-boot/custodians/u-boot...
Tom Rini [Wed, 7 Sep 2022 12:39:12 +0000 (08:39 -0400)]
Merge tag 'fsl-qoriq-2022-9-7' of https://source.denx.de/u-boot/custodians/u-boot-fsl-qoriq

- Pali's patch not in my patchwork, got missed.
- Sean's patch pending for sometime, I just fix conflict when apply
  Sean's patch, so pick up.

2 years agoMerge tag 'dm-pull-7sep22' of https://source.denx.de/u-boot/custodians/u-boot-dm
Tom Rini [Wed, 7 Sep 2022 12:38:44 +0000 (08:38 -0400)]
Merge tag 'dm-pull-7sep22' of https://source.denx.de/u-boot/custodians/u-boot-dm

binman fixes for bintool support

2 years agonet: fm: Add support for FIT firmware
Sean Anderson [Wed, 7 Sep 2022 05:44:55 +0000 (13:44 +0800)]
net: fm: Add support for FIT firmware

Fman microcode is executable code (AFAICT) loaded into a
coprocessor. As such, if verified boot is enabled, it must be verified
like other executable code. However, this is not currently done.

This commit adds verified boot functionality by encapsulating the
microcode in a FIT, which can then be signed/verified as normal. By
default we allow fallback to unencapsulated firmware, but if
CONFIG_FIT_SIGNATURE is enabled, then we make it mandatory. Because
existing Layerscape do not use this config (instead enabling
CONFIG_CHAIN_OF_TRUST), this should not break any existing boards.

An example (mildly-abbreviated) its is provided below:

/ {
    #address-cells = <1>;

    images {
        firmware {
            data = /incbin/(/path/to/firmware);
            type = "firmware";
            arch = "arm64";
            compression = "none";
    signature {
                algo = "sha256,rsa2048";
                key-name-hint = "your key name";
            };
        };
    };

    configurations {
        default = "conf";
        conf {
            description = "Load FMAN microcode";
            fman = "firmware";
        };
    };
};

Reviewed-by: Simon Glass <sjg@chromium.org>
Signed-off-by: Sean Anderson <sean.anderson@seco.com>
Signed-off-by: Peng Fan <peng.fan@nxp.com>
2 years agonet: Convert fit verification to use fit_get_data_*
Sean Anderson [Tue, 16 Aug 2022 15:16:06 +0000 (11:16 -0400)]
net: Convert fit verification to use fit_get_data_*

Several ethernet drivers load firmware from FIT images. Convert them to
use the fit_get_data helpers.

Signed-off-by: Sean Anderson <sean.anderson@seco.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
Signed-off-by: Peng Fan <peng.fan@nxp.com>
2 years agocmd: fpga: Convert to use fit_get_data_node
Sean Anderson [Tue, 16 Aug 2022 15:16:05 +0000 (11:16 -0400)]
cmd: fpga: Convert to use fit_get_data_node

This converts the FIT loading process of the fpga command to use
fit_get_data_node.

Signed-off-by: Sean Anderson <sean.anderson@seco.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
Signed-off-by: Peng Fan <peng.fan@nxp.com>
2 years agoARMv8/sec_firmware: Convert to use fit_get_data_conf_prop
Sean Anderson [Tue, 16 Aug 2022 15:16:04 +0000 (11:16 -0400)]
ARMv8/sec_firmware: Convert to use fit_get_data_conf_prop

This reduces sec_firmware_get_data to a single call to
fit_get_data_conf_prop. I think sec_firmware_check_copy_loadable could also
be converted, but it does not map as straightforwardly, so I have left it
for a future cleanup.

Signed-off-by: Sean Anderson <sean.anderson@seco.com>
Signed-off-by: Peng Fan <peng.fan@nxp.com>
2 years agoimage: fit: Add some helpers for getting data
Sean Anderson [Tue, 16 Aug 2022 15:16:03 +0000 (11:16 -0400)]
image: fit: Add some helpers for getting data

Several different firmware users have repetitive code to extract the
firmware data from a FIT. Add some helper functions to reduce the amount
of repetition. fit_conf_get_prop_node (eventually) calls
fdt_check_node_offset_, so we can avoid an explicit if. In general, this
version avoids printing on error because the callers are typically
library functions, and because the FIT code generally has (debug)
prints of its own. One difference in these helpers is that they use
fit_image_get_data_and_size instead of fit_image_get_data, as the former
handles external data correctly.

Signed-off-by: Sean Anderson <sean.anderson@seco.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
Signed-off-by: Peng Fan <peng.fan@nxp.com>
2 years agoARMv8/sec_firmware: Remove SEC_FIRMWARE_FIT_CNF_NAME
Sean Anderson [Tue, 16 Aug 2022 15:16:02 +0000 (11:16 -0400)]
ARMv8/sec_firmware: Remove SEC_FIRMWARE_FIT_CNF_NAME

The config to use for FIT images can be better specified by enabling
CONFIG_MULTI_DTB_FIT and implementing board_fit_config_name_match.

Signed-off-by: Sean Anderson <sean.anderson@seco.com>
Signed-off-by: Peng Fan <peng.fan@nxp.com>
2 years agoboard: freescale: p1_p2_rdb_pc: Calculate offsets for eSDHC boot sector
Pali Rohár [Mon, 1 Aug 2022 12:50:12 +0000 (14:50 +0200)]
board: freescale: p1_p2_rdb_pc: Calculate offsets for eSDHC boot sector

Correctly calculate offsets between SPL and proper U-Boot when new config
option CONFIG_FSL_PREPBL_ESDHC_BOOT_SECTOR for generating eSDHC boot sector
is enabled. Otherwise SPL would not be able to boot proper U-Boot.

Signed-off-by: Pali Rohár <pali@kernel.org>
Signed-off-by: Peng Fan <peng.fan@nxp.com>
2 years agoboard: freescale: p1_p2_rdb_pc: Delete watchdog max6370 node in load_default mode
Pali Rohár [Mon, 1 Aug 2022 13:35:43 +0000 (15:35 +0200)]
board: freescale: p1_p2_rdb_pc: Delete watchdog max6370 node in load_default mode

CPLD in load_default mode ignores watchdog reset signal. It does not reset
board when watchdog triggers reset signal.

Detect load_default mode by GPIO7 - LOAD_DEFAULT_N and delete watchdog
max6370 node from device to prevent registering driver for non-working
watchdog.

Signed-off-by: Pali Rohár <pali@kernel.org>
Signed-off-by: Peng Fan <peng.fan@nxp.com>
2 years agoboard: freescale: p1_p2_rdb_pc: Add env commands norlowerboot, norupperboot, sd2boot...
Pali Rohár [Mon, 25 Apr 2022 14:50:43 +0000 (16:50 +0200)]
board: freescale: p1_p2_rdb_pc: Add env commands norlowerboot, norupperboot, sd2boot and defboot

All *boot env commands overrides default boot source location via i2c.
After board reset without power off, BootROM then starts booting U-Boot
from this specified location instead of the default one.

Add new env command defboot which reverts boot location to the default
value, which in most cases is configurable by HW DIP switches.

And add new env commands norlowerboot, norupperboot, sd2boot to boot from
other locations. norlowerboot would instruct BootROM to boot from lower NOR
bank, norupperboot from upper NOR bank and sd2boot from SD card with
alternative configuration.

Signed-off-by: Pali Rohár <pali@kernel.org>
Signed-off-by: Peng Fan <peng.fan@nxp.com>
2 years agoARM: dts: stm32mp15: remove hwlocks from pinctrl
Etienne Carriere [Mon, 5 Sep 2022 09:15:28 +0000 (11:15 +0200)]
ARM: dts: stm32mp15: remove hwlocks from pinctrl

Removes hwlocks properties from stm32mp151 pinctrl node. These locks
could be used for other purpose, depending on board and software
configuration hence do not enforce their use to protect pinctrl
devices.

This patch is an alignment with Linux device tree with v6.0 as the
hwsem support wasn’t yet added in pincontrol in kernel. It avoids
issues when the Linux kernel is started with the U-Boot device tree.

Cc: Patrice Chotard <patrice.chotard@foss.st.com>
Cc: Patrick Delaunay <patrick.delaunay@foss.st.com>
Signed-off-by: Etienne Carriere <etienne.carriere@linaro.org>
Reviewed-by: Patrick Delaunay <patrick.delaunay@foss.st.com>
2 years agoARM: stm32: Switch DHSOM to FMC2 EBI driver
Marek Vasut [Tue, 23 Aug 2022 17:27:08 +0000 (19:27 +0200)]
ARM: stm32: Switch DHSOM to FMC2 EBI driver

Perform long overdue conversion of ad-hoc FMC2 EBI bus initialization
to upstream FMC2 EBI driver. No functional change.

Signed-off-by: Marek Vasut <marex@denx.de>
Cc: Patrice Chotard <patrice.chotard@foss.st.com>
Cc: Patrick Delaunay <patrick.delaunay@foss.st.com>
Reviewed-by: Patrice Chotard <patrice.chotard@foss.st.com>
Reviewed-by: Patrick Delaunay <patrick.delaunay@foss.st.com>
2 years agoMerge https://source.denx.de/u-boot/custodians/u-boot-riscv
Tom Rini [Tue, 6 Sep 2022 13:01:39 +0000 (09:01 -0400)]
Merge https://source.denx.de/u-boot/custodians/u-boot-riscv

2 years agoMerge tag 'fsl-qoriq-2022-9-6' of https://source.denx.de/u-boot/custodians/u-boot...
Tom Rini [Tue, 6 Sep 2022 12:59:51 +0000 (08:59 -0400)]
Merge tag 'fsl-qoriq-2022-9-6' of https://source.denx.de/u-boot/custodians/u-boot-fsl-qoriq

Reset fixes for p1_p2_rdb_pc
Fix use after free issue fix in fsl_enetc.c
Fix for fsl ddr: make bank_addr_bits reflect actual bits
sl28 board update

2 years agoconfigs: stih410-b2260: Fix SYS_HZ_CLOCK value
Patrice Chotard [Thu, 25 Aug 2022 07:14:57 +0000 (09:14 +0200)]
configs: stih410-b2260: Fix SYS_HZ_CLOCK value

SYS_HZ_CLOCK was wrongly set to 1GHz whereas it's set to 750MHz
by default by bootrom.

Signed-off-by: Patrice Chotard <patrice.chotard@foss.st.com>
Reviewed-by: Grzegorz Szymaszek <gszymaszek@short.pl>
2 years agoconfigs: stm32: Enable CONFIG_DM_REGULATOR for stm32f769-disco
Patrice Chotard [Wed, 24 Aug 2022 13:44:40 +0000 (15:44 +0200)]
configs: stm32: Enable CONFIG_DM_REGULATOR for stm32f769-disco

Since commit 5bc6f8c2a97e("video: stm32: remove test on CONFIG_DM_REGULATOR")
backlight was broken with the following message at boot:
stm32-display-dsi dsi@40016c00: Warning: cannot get phy dsi supply
stm32_display display-controller@40016800: panel panel enable backlight error -38

DM_REGULATOR flag must be enabled to fix this issue

Signed-off-by: Patrice Chotard <patrice.chotard@foss.st.com>
Reviewed-by: Patrick Delaunay <patrick.delaunay@foss.st.com>
2 years agoARM: dts: stm32: Fix display-timings settings for stm32f746-disco
Patrice Chotard [Wed, 24 Aug 2022 13:42:37 +0000 (15:42 +0200)]
ARM: dts: stm32: Fix display-timings settings for stm32f746-disco

Since commit ef4ce6df3289 "video: stm32: stm32_ltdc: fix data enable polarity"
The panel display output wasn't functional anymore.
Device tree display-timings de-active property value must be updated
to 1.

Signed-off-by: Patrice Chotard <patrice.chotard@foss.st.com>
Reviewed-by: Patrick Delaunay <patrick.delaunay@foss.st.com>
2 years agoclk: stm32mp: handle ck_usbo_48m clock provided by USBPHYC
Patrick Delaunay [Tue, 26 Apr 2022 12:37:49 +0000 (14:37 +0200)]
clk: stm32mp: handle ck_usbo_48m clock provided by USBPHYC

Handle the input clock of RCC USB_PHY_48, provided by USBPHYC
and named "ck_usbo_48m".

Signed-off-by: Patrick Delaunay <patrick.delaunay@foss.st.com>
Reviewed-by: Patrice Chotard <patrice.chotard@foss.st.com>
2 years agophy: stm32-usbphyc: usbphyc is a clock provider of ck_usbo_48m clock
Patrick Delaunay [Tue, 26 Apr 2022 12:37:48 +0000 (14:37 +0200)]
phy: stm32-usbphyc: usbphyc is a clock provider of ck_usbo_48m clock

ck_usbo_48m is generated by usbphyc PLL and used by OTG controller
for Full-Speed use cases with dedicated Full-Speed transceiver.

ck_usbo_48m is available as soon as the PLL is enabled.

Signed-off-by: Patrick Delaunay <patrick.delaunay@foss.st.com>
Reviewed-by: Patrice Chotard <patrice.chotard@foss.st.com>
2 years agophy: stm32-usbphyc: add counter of PLL consumer
Patrick Delaunay [Tue, 26 Apr 2022 12:37:47 +0000 (14:37 +0200)]
phy: stm32-usbphyc: add counter of PLL consumer

Add the counter of the PLL user n_pll_cons managed by the 2 functions
stm32_usbphyc_pll_enable / stm32_usbphyc_pll_disable.

This counter allow to remove the function stm32_usbphyc_is_init
and it is a preliminary step for ck_usbo_48m introduction.

Signed-off-by: Patrick Delaunay <patrick.delaunay@foss.st.com>
Reviewed-by: Patrice Chotard <patrice.chotard@foss.st.com>
2 years agoboard: stm32mp1: remove test on CONFIG_DM_REGULATOR
Patrick Delaunay [Mon, 20 Jun 2022 10:36:10 +0000 (12:36 +0200)]
board: stm32mp1: remove test on CONFIG_DM_REGULATOR

The tests on CONFIG_DM_REGULATOR, added to avoid compilation issues, can
now be removed, they are no more needed since the commit 16cc5ad0b439
("power: regulator: add dummy helper").

Signed-off-by: Patrick Delaunay <patrick.delaunay@foss.st.com>
Reviewed-by: Patrice Chotard <patrice.chotard@foss.st.com>
2 years agostm32mp: simplify the STM32MP15x package parsing code
Patrick Delaunay [Mon, 20 Jun 2022 07:50:01 +0000 (09:50 +0200)]
stm32mp: simplify the STM32MP15x package parsing code

Simplify the package parsing code for STM32MP15X as package can be
affected with get_cpu_package() result.

Signed-off-by: Patrick Delaunay <patrick.delaunay@foss.st.com>
Reviewed-by: Patrice Chotard <patrice.chotard@foss.st.com>
2 years agoboard: freescale: p1_p2_rdb_pc: Turn off watchdog before reset
Pali Rohár [Mon, 1 Aug 2022 13:31:46 +0000 (15:31 +0200)]
board: freescale: p1_p2_rdb_pc: Turn off watchdog before reset

P1/P2 RDB boards have external max6370 watchdog connected to CPLD and this
watchdog is not deactivated on board reset. So if it is active during board
reset, it can trigger another reset when CPU is booting U-Boot. To prevent
possible infinite reset loop caused by external watchdog, turn it off
before reset.

Do it via a new board_reset_prepare() callback which is called from
do_reset() function before any reset sequence.

Signed-off-by: Pali Rohár <pali@kernel.org>
Signed-off-by: Peng Fan <peng.fan@nxp.com>
2 years agoboard: freescale: p1_p2_rdb_pc: Avoid usage of CPLD's system reset register
Pali Rohár [Mon, 1 Aug 2022 13:31:45 +0000 (15:31 +0200)]
board: freescale: p1_p2_rdb_pc: Avoid usage of CPLD's system reset register

CPLD's system reset register is buggy and requires workaround in U-Boot.
So use this kind of board reset only when there is no other reset option.

Introduce a new board_reset_last() callback which is last-stage
board-specific reset and implement CPLD's system reset in this new
board_reset_last() callback instead of board_reset() callback.

Fixes: 20fb58fc5a1c ("board: freescale: p1_p2_rdb_pc: Implement board_reset()")
Signed-off-by: Pali Rohár <pali@kernel.org>
Signed-off-by: Peng Fan <peng.fan@nxp.com>
2 years agoboard: freescale: p1_p2_rdb_pc: Add workaround for non-working watchdog
Pali Rohár [Mon, 1 Aug 2022 13:31:44 +0000 (15:31 +0200)]
board: freescale: p1_p2_rdb_pc: Add workaround for non-working watchdog

If watchdog timer was already set to non-disabled value then it means that
watchdog timer was already activated, has already expired and caused CPU
reset. If this happened then due to CPLD firmware bug, writing to wd_cfg
register has no effect and therefore it is not possible to reactivate
watchdog timer again. Watchdog starts working again after CPU reset via
non-watchdog method.

Implement this workaround (reset CPU when it was reset by watchdog) to make
watchdog usable again. Watchdog timer logic on these P1/P2 RDB boards is
connected to CPLD, not to SoC itself.

Note that reset does not occur immediately after calling do_reset(), but
after few ms later as real reset is done by CPLD. So it is normal that
function do_reset() returns. Therefore hangs after calling do_reset() to
prevent CPU execution of the rest U-Boot code.

Signed-off-by: Pali Rohár <pali@kernel.org>
Signed-off-by: Peng Fan <peng.fan@nxp.com>
2 years agoboard: freescale: p1_p2_rdb_pc: Add workaround for board reset reboot loop
Pali Rohár [Mon, 1 Aug 2022 13:31:43 +0000 (15:31 +0200)]
board: freescale: p1_p2_rdb_pc: Add workaround for board reset reboot loop

CPLD's system reset register on P1/P2 RDB boards is not autocleared after
flipping it. If this register is set to one in 100ms after reset starts
then CPLD triggers another CPU reset.

This means that trying to reset board via CPLD system reset register cause
reboot loop. To prevent this reboot loop, the only workaround is to try to
clear CPLD's system reset register as early as possible. U-Boot is already
doing it in its board_early_init_f() function, which seems to be enough as
register is cleared prior CPLD triggers another reset.

But board_early_init_f() is not called from SPL and therefore usage of SPL
can cause reboot loop.

To prevent reboot loop when using SPL, call board_early_init_f() function
in SPL too. For accessing CPLD memory space it is needed to have CPLD entry
in TLB.

With this change it is possible to trigger board reset via CPLD's system
reset register on P2020 RDB board.

Signed-off-by: Pali Rohár <pali@kernel.org>
Signed-off-by: Peng Fan <peng.fan@nxp.com>
2 years agoRISC-V: enable CONFIG_SYSRESET_SBI by default
Heinrich Schuchardt [Mon, 5 Sep 2022 14:40:49 +0000 (16:40 +0200)]
RISC-V: enable CONFIG_SYSRESET_SBI by default

System reset via the SRST extension in the SBI should be the default.
The driver checks if the extension is available when probing.
So there is no risk in enabling it.

Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
Reviewed-by: Leo Yu-Chi Liang <ycliang@andestech.com>
2 years agocmd/sbi: format KVM version
Heinrich Schuchardt [Sun, 14 Aug 2022 19:57:14 +0000 (21:57 +0200)]
cmd/sbi: format KVM version

Format the KVM implementation number in a human readable form.

With the patch output of the sbi command for Linux 5.19.1 looks like:

    => sbi
    SBI 0.3
    KVM 5.19.1
    Machine:
      Vendor ID 0
      Architecture ID 7005c
      Implementation ID 7005c
    Extensions:
      SBI Base Functionality
      Timer Extension
      IPI Extension
      RFENCE Extension
      Hart State Management Extension
      System Reset Extension

Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
Reviewed-by: Leo Yu-Chi Liang <ycliang@andestech.com>
2 years agoriscv: dts: sifive: Synchronize FU740 and Unmatched DT
Icenowy Zheng [Thu, 25 Aug 2022 08:11:19 +0000 (16:11 +0800)]
riscv: dts: sifive: Synchronize FU740 and Unmatched DT

These DT files are synchronized from Linux 5.19.

Signed-off-by: Icenowy Zheng <uwu@icenowy.me>
Reviewed-by: Leo Yu-Chi Liang <ycliang@andestech.com>
2 years agodt-bindings: clock: sifive: sync FU740 PRCI clock binding header
Icenowy Zheng [Thu, 25 Aug 2022 08:11:18 +0000 (16:11 +0800)]
dt-bindings: clock: sifive: sync FU740 PRCI clock binding header

This commit sychronizes the header file for FU740 PRCI clocks with the
one from Linux 5.19.

The constant values are the same, but all constant names are changed
(most are just prefixed with FU740_).

Signed-off-by: Icenowy Zheng <uwu@icenowy.me>
Reviewed-by: Leo Yu-Chi Liang <ycliang@andestech.com>
2 years agonet: enetc: Fix use after free issue in fsl_enetc.c
Siarhei Yasinski [Wed, 31 Aug 2022 10:57:37 +0000 (10:57 +0000)]
net: enetc: Fix use after free issue in fsl_enetc.c

If ethernet connected to SFP, like this:

&enetc_port0 {
            phy-connection-type = "sgmii";
            sfp = <&sfp0>;
            managed = "in-band-status";
            status = "okay";
};

Then enetc_config_phy returns -ENODEV and the memory containing the mdio interface is freed.
It's better to unregister and free mdio resources.

Signed-off-by: Siarhei Yasinski <siarhei.yasinski@sintecs.eu>
Reviewed-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Ramon Fried <rfried.dev@gmail.com>
Signed-off-by: Peng Fan <peng.fan@nxp.com>
2 years agoddr: fsl: Make bank_addr_bits reflect actual bits
Sean Anderson [Tue, 30 Aug 2022 21:01:07 +0000 (17:01 -0400)]
ddr: fsl: Make bank_addr_bits reflect actual bits

In both the Freescale DDR controller and the SPD spec, bank address bits
are stored as the number of bank address bits minus 2. For example, if a
chip had 8 banks (3 total bank address bits), the value of
bank_addr_bits would be 1. This is rather surprising for users
configuring their memory manually, since they can't set bank_addr_bits
to the actual number of bank address bits. Rectify this.

There is at least one example of this kind of mistake already, in
board/freescale/t102xrdb/ddr.c. The documented MT40A512M8HX has two bank
address bits, but bank_addr_bits was set to 2, implying 4 bank address
bits. Such a value is reserved in BA_BITS_CS, but I suspect the
controller simply ignores the top bit, making this kind of mistake
harmless, if misleading.

Fixes: e8a7f1c32b5 ("powerpc/t1023rdb: Add T1023 RDB board support")
Signed-off-by: Sean Anderson <sean.anderson@seco.com>
Signed-off-by: Peng Fan <peng.fan@nxp.com>
2 years agoddr: fsl: Reduce the size of interactive options
Sean Anderson [Tue, 30 Aug 2022 20:54:39 +0000 (16:54 -0400)]
ddr: fsl: Reduce the size of interactive options

The interactive mode uses large several tables of options which can be
configured. However, much of the contents of these tables are
repetetive. For example, no struct is larger than half a kilobyte, so
the offset only takes up 9 bits. Similarly, the size is only ever 4 or
8, and printhex is a boolean. Reduce the size of these fields. This
reduces the size of the options tables by around 10 KiB. However, the
largest contributor to the size of the options tables is the use of a
pointer for the strings. A better approach would be to use a separate
array of strings, and store an integer index in the options tables.
However, this would require a large re-architecting of this file.

Signed-off-by: Sean Anderson <sean.anderson@seco.com>
Signed-off-by: Peng Fan <peng.fan@nxp.com>
2 years agoboard: sl28: remove COUNTER_FREQUENCY_REAL
Michael Walle [Tue, 23 Aug 2022 09:30:18 +0000 (11:30 +0200)]
board: sl28: remove COUNTER_FREQUENCY_REAL

The frequency of the system counter is static which is given by the
COUNTER_FREQUENCY option. Remove COUNTER_FREQUENCY_REAL.

Signed-off-by: Michael Walle <michael@walle.cc>
Signed-off-by: Peng Fan <peng.fan@nxp.com>
2 years agoboard: sl28: support dynamic prompts
Michael Walle [Tue, 23 Aug 2022 09:30:17 +0000 (11:30 +0200)]
board: sl28: support dynamic prompts

Depending on the boot source, set different CLI prompts. This will help
the user to figure out in which mode the bootloader was started. There
are two special modes: failsafe and SDHC boot.

Signed-off-by: Michael Walle <michael@walle.cc>
Signed-off-by: Peng Fan <peng.fan@nxp.com>
2 years agoboard: sl28: add user friendly names for the boot sources
Michael Walle [Tue, 23 Aug 2022 09:30:16 +0000 (11:30 +0200)]
board: sl28: add user friendly names for the boot sources

During startup the SPL will print where the u-boot proper is read from.
Instead of using the default names, provide more user friendly names.

Signed-off-by: Michael Walle <michael@walle.cc>
Signed-off-by: Peng Fan <peng.fan@nxp.com>
2 years agoboard: sl28: implement additional bootsources
Michael Walle [Tue, 23 Aug 2022 09:30:15 +0000 (11:30 +0200)]
board: sl28: implement additional bootsources

The board is able to boot from the following source:
 - user-updateble SPI flash
 - write-protected part of the same SPI flash
 - eMMC
 - SD card

Implement the needed function hooks to support all of these boot
sources.

Signed-off-by: Michael Walle <michael@walle.cc>
Signed-off-by: Peng Fan <peng.fan@nxp.com>
2 years agoarmv8: layerscape: spl: mark OCRAM as non-secure
Michael Walle [Tue, 23 Aug 2022 09:30:14 +0000 (11:30 +0200)]
armv8: layerscape: spl: mark OCRAM as non-secure

By default the OCRAM is marked as secure. While the SPL runs in EL3 and
thus can access it, DMA devices cannot. Mark the whole OCRAM as
non-secure.
This will fix MMC and SD card boot on LS1028A when using SPL instead of
TF-A.

Signed-off-by: Michael Walle <michael@walle.cc>
Signed-off-by: Peng Fan <peng.fan@nxp.com>
2 years agoPrepare v2022.10-rc4
Tom Rini [Tue, 6 Sep 2022 00:32:56 +0000 (20:32 -0400)]
Prepare v2022.10-rc4

Signed-off-by: Tom Rini <trini@konsulko.com>
2 years agoconfigs: Resync with savedefconfig
Tom Rini [Tue, 6 Sep 2022 00:32:14 +0000 (20:32 -0400)]
configs: Resync with savedefconfig

Rsync all defconfig files using moveconfig.py

Signed-off-by: Tom Rini <trini@konsulko.com>
2 years agoriscv: dts: Sync important Unmatched pmic and qspi0 changes from Linux
Jessica Clarke [Fri, 12 Aug 2022 17:50:03 +0000 (18:50 +0100)]
riscv: dts: Sync important Unmatched pmic and qspi0 changes from Linux

This adds the onkey, RTC and watchdog children to the DA9063 PMIC node,
fixes the compatible for qspi0's flash node to match the official DT
schema (it being an is25wp256 is discoverable, hence jedec,spi-nor is
the only compatible that should be present) and exposes the card detect
GPIO.

Note that the device trees still diverge in some places (including
important things like the PCIe controller's clock name) and should be
cleaned up so that a common device tree is used in both projects rather
than having different bindings. This patch does not attempt to do that,
merely expose important functionality present in Linux's that is not in
U-Boot's so that it can be used without the OS providing its own bundled
copy.

Signed-off-by: Jessica Clarke <jrtc27@jrtc27.com>
Reviewed-by: Leo Yu-Chi Liang <ycliang@andestech.com>
2 years agoMerge tag 'u-boot-rockchip-20220905' of https://source.denx.de/u-boot/custodians...
Tom Rini [Mon, 5 Sep 2022 02:35:40 +0000 (22:35 -0400)]
Merge tag 'u-boot-rockchip-20220905' of https://source.denx.de/u-boot/custodians/u-boot-rockchip

- migrate to use binman for U-Boot image generate on rockchip platform;
- Some fixes for rk3399 and rk3308;

2 years agoclk: rockchip: rk3399: Fix Unknown clock 77 on mmc@fe310000
Michal Suchanek [Sun, 21 Aug 2022 07:17:24 +0000 (09:17 +0200)]
clk: rockchip: rk3399: Fix Unknown clock 77 on mmc@fe310000

Adding some debug prints I can see:

MMC:   mmc@fe320000: Got clock clock-controller@ff760000 76
mmc@fe310000: Got clock clock-controller@ff760000 77
Unknown clock 77
rockchip_dwmmc_get_mmc_clk: err=-2
mmc@fe310000: 3, mmc@fe320000: 1, mmc@fe330000: 0

According to kernel code the SDIO clock is identical to SDMMC clock
except for the con 16->15 change.

Add support for the clock to avoid the error.

Signed-off-by: Michal Suchanek <msuchanek@suse.de>
Reviewed-by: Kever Yang <kever.yang@rock-chips.com>
2 years agorockchip: add u-boot-rockchip-spi.bin image for booting from SPI-NOR flash
Quentin Schulz [Fri, 2 Sep 2022 13:10:55 +0000 (15:10 +0200)]
rockchip: add u-boot-rockchip-spi.bin image for booting from SPI-NOR flash

This new image is similar to u-boot-rockchip.bin except that it's
destined to be flashed on SPI-NOR flashes.

Cc: Quentin Schulz <foss+uboot@0leil.net>
Signed-off-by: Quentin Schulz <quentin.schulz@theobroma-systems.com>
Reviewed-by: Kever Yang <kever.yang@rock-chips.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
2 years agorockchip: allow to build SPI images even without HAS_ROM option
Quentin Schulz [Fri, 2 Sep 2022 13:10:54 +0000 (15:10 +0200)]
rockchip: allow to build SPI images even without HAS_ROM option

This prepares for the creation of a u-boot-rockchip-spi.bin image
similar to u-boot-rockchip.bin to the exception it's destined for
SPI-NOR flashes instead of MMC storage medium.

Cc: Quentin Schulz <foss+uboot@0leil.net>
Signed-off-by: Quentin Schulz <quentin.schulz@theobroma-systems.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Kever Yang <kever.yang@rock-chips.com>
2 years agorockchip: simplify binman image dependencies addition to INPUTS
Quentin Schulz [Fri, 2 Sep 2022 13:10:53 +0000 (15:10 +0200)]
rockchip: simplify binman image dependencies addition to INPUTS

By factoring SPL check in the first condition, this makes the checks a
bit less convoluted and more readable.

Cc: Quentin Schulz <foss+uboot@0leil.net>
Signed-off-by: Quentin Schulz <quentin.schulz@theobroma-systems.com>
Reviewed-by: Kever Yang <kever.yang@rock-chips.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
2 years agorockchip: generate u-boot-rockchip.bin with binman for ARM64 boards
Quentin Schulz [Fri, 2 Sep 2022 13:10:52 +0000 (15:10 +0200)]
rockchip: generate u-boot-rockchip.bin with binman for ARM64 boards

This allows to build u-boot-rockchip.bin binary with binman for Rockchip
ARM64 boards instead of the legacy Makefile way.

Cc: Quentin Schulz <foss+uboot@0leil.net>
Signed-off-by: Quentin Schulz <quentin.schulz@theobroma-systems.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
2 years agorockchip: generate idbloader.img content for u-boot-rockchip.bin with binman for ARM
Quentin Schulz [Fri, 2 Sep 2022 13:10:51 +0000 (15:10 +0200)]
rockchip: generate idbloader.img content for u-boot-rockchip.bin with binman for ARM

idbloader.img content - currently created by way of Makefile - can be
created by binman directly.

So let's do that for Rockchip ARM platforms.

Cc: Quentin Schulz <foss+uboot@0leil.net>
Signed-off-by: Quentin Schulz <quentin.schulz@theobroma-systems.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Kever Yang <kever.yang@rock-chips.com>
2 years agorockchip: remove binman temporary files when cleaning
Quentin Schulz [Fri, 2 Sep 2022 13:10:50 +0000 (15:10 +0200)]
rockchip: remove binman temporary files when cleaning

Binman mkimage entry generates temporary files so let's remove them
when calling `make clean`.

Fixes: 9b312e26fc77 ("rockchip: Enable building a SPI ROM image on jerry")
Cc: Quentin Schulz <foss+uboot@0leil.net>
Reported-by: Johan Jonker <jbx6244@gmail.com>
Signed-off-by: Quentin Schulz <quentin.schulz@theobroma-systems.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
Reviewed-by: Kever Yang <kever.yang@rock-chips.com>
2 years agobinman: allow user-defined filenames for mkimage entry
Quentin Schulz [Fri, 2 Sep 2022 13:10:49 +0000 (15:10 +0200)]
binman: allow user-defined filenames for mkimage entry

mkimage entry currently creates a file whose name is derived from the
section name containing said entry.

Let's allow the user to define a filename for the mkimage-generated
binary by using the 'filename' DT property.

Cc: Quentin Schulz <foss+uboot@0leil.net>
Signed-off-by: Quentin Schulz <quentin.schulz@theobroma-systems.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
2 years agobinman: add support for skipping file concatenation for mkimage
Quentin Schulz [Fri, 2 Sep 2022 13:10:48 +0000 (15:10 +0200)]
binman: add support for skipping file concatenation for mkimage

Some image types handled by mkimage require the datafiles to be passed
independently (-d data1:data2) for specific handling of each. A
concatenation of datafiles prior to passing them to mkimage wouldn't
work.

That is the case for rkspi for example which requires page alignment
and only writing 2KB every 4KB.

This adds the ability to tell binman to pass the datafiles without
prior concatenation to mkimage, by adding the multiple-data-files
boolean property to the mkimage node.

Cc: Quentin Schulz <foss+uboot@0leil.net>
Signed-off-by: Quentin Schulz <quentin.schulz@theobroma-systems.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
2 years agorockchip: rk3399: sync spl_boot_devices_tbl and boot_devices node paths
Quentin Schulz [Fri, 15 Jul 2022 15:15:52 +0000 (17:15 +0200)]
rockchip: rk3399: sync spl_boot_devices_tbl and boot_devices node paths

While technically not a bug, let's have some consistency in paths
returned by u-boot,spl-boot-order look-up and the one saved in
u-boot,spl-boot-device by syncing spl_boot_devices_tbl and boot_devices
node paths.

Cc: Quentin Schulz <foss+uboot@0leil.net>
Signed-off-by: Quentin Schulz <quentin.schulz@theobroma-systems.com>
Tested-by: Xavier Drudis Ferran <xdrudis@tinet.cat>
Reviewed-by: Xavier Drudis Ferran <xdrudis@tinet.cat>
Reviewed-by: Kever Yang <kever.yang@rock-chips.com>
2 years agorockchip: rk3399: fix incorrect boot-device in u-boot, spl-boot-device
Quentin Schulz [Fri, 15 Jul 2022 15:15:51 +0000 (17:15 +0200)]
rockchip: rk3399: fix incorrect boot-device in u-boot, spl-boot-device

On RK3399, mmc0 is eMMC and mmc1 is SD card, c.f. console:
MMC:   mmc@fe320000: 1, mmc@fe330000: 0

In arch/arm/mach-rockchip/spl-boot-order.c:board_boot_order, the
boot_device (BOOT_DEVICE_*) value is gotten from spl_node_to_boot_device
function. Said function returns BOOT_DEVICE_MMC1 for mmc0 (eMMC) and
BOOT_DEVICE_MMC2 for mmc1 (SD card).

Since the SD card controller is at mmc@fe320000, it should be associated
with BOOT_DEVICE_MMC2 and not BOOT_DEVICE_MMC1. Same applies to eMMC.

Let's fix that by swapping the two BOOT_DEVICEs.

Cc: Quentin Schulz <foss+uboot@0leil.net>
Signed-off-by: Quentin Schulz <quentin.schulz@theobroma-systems.com>
Tested-by: Xavier Drudis Ferran <xdrudis@tinet.cat>
2 years agorockchip: rk3308: fix same-as-spl boot order
John Keeping [Thu, 14 Jul 2022 14:18:37 +0000 (15:18 +0100)]
rockchip: rk3308: fix same-as-spl boot order

Rockchip SoCs need the boot_devices array defined in order to map the
bootloader's value to a U-Boot device.  Implement this for rk3308.

Signed-off-by: John Keeping <john@metanate.com>
Reviewed-by: Philipp Tomsich <philipp.tomsich@vrull.eu>
Reviewed-by: Kever Yang <kever.yang@rock-chips.com>
2 years agorockchip: rk3308: fix rockchip_dnl_key_pressed() on roc-cc
John Keeping [Thu, 14 Jul 2022 14:09:12 +0000 (15:09 +0100)]
rockchip: rk3308: fix rockchip_dnl_key_pressed() on roc-cc

Commit 6aa4fe3912 ("dm: core: Rename and fix uclass_get_by_name_len()")
changed uclass_get_device_by_name() to an exact match when previously it
behaved as a prefix match.

The roc-cc code relied on this prefix match by only specifying part of
the device name.  Fix this by using the full name including the address.

Signed-off-by: John Keeping <john@metanate.com>
Reviewed-by: Kever Yang <kever.yang@rock-chips.com>
2 years agorockchip: rk3399: boot_devices: fix eMMC node name
Quentin Schulz [Mon, 11 Jul 2022 14:15:33 +0000 (16:15 +0200)]
rockchip: rk3399: boot_devices: fix eMMC node name

When idbloader.img is flashed on the eMMC, the SPL still tries to load
from SPI-NOR first.

This is due to an incorrect look-up in the Device Tree. Since commit
822556a93459 ("arm: dts: sync the Rockhip 3399 SoCs from Linux"), the
node name (but not label) changed from sdhci@fe330000 to mmc@fe330000
meaning U-Boot SPL is not looking for the correct node name anymore and
fails to find the "same-as-spl" node when eMMC is the medium from which
the SPL booted.

Fixes: 822556a93459 ("arm: dts: sync the Rockhip 3399 SoCs from Linux")
Cc: Quentin Schulz <foss+uboot@0leil.net>
Signed-off-by: Quentin Schulz <quentin.schulz@theobroma-systems.com>
Reviewed-by: Xavier Drudis Ferran <xdrudis@tinet.cat>
Reviewed-by: Artem Lapkin <email2tema@gmail.com>
Tested-by: Artem Lapkin <email2tema@gmail.com>
Tested-by: Lapkin Artem <email2tema@gmail.com>
Tested-by: Lapkin Artem <email2tema@gmail.com>
Reviewed-by: Kever Yang <kever.yang@rock-chips.com>
2 years agoram: rk3399: Conduct memory training at 400MHz
Lee Jones [Thu, 11 Aug 2022 07:58:48 +0000 (08:58 +0100)]
ram: rk3399: Conduct memory training at 400MHz

Currently the default initialisation frequency is 50MHz.  Although
this does appear to be suitable for some LPDDR4 RAM chips, training at
this low frequency has been seen to cause Column errors, leading to
Capacity check errors on others.

Here we force RAM initialisation to happen at 400MHz before ramping up
to the final value running value of 800MHz after everything has been
successfully configured.

Link: https://lore.kernel.org/u-boot/Yo4v3jUeHXTovjOH@google.com/
Suggested-by: YouMin Chen <cym@rock-chips.com>
Signed-off-by: Lee Jones <lee@kernel.org>
Tested-by: Xavier Drudis Ferran <xdrudis@tinet.cat>
Reviewed-by: Kever Yang <kever.yang@rock-chips.com>
Tested-by: Michal Suchánek <msuchanek@suse.de>
2 years agoram: rk3399: Fix faulty frequency change reports
Lee Jones [Thu, 11 Aug 2022 07:58:47 +0000 (08:58 +0100)]
ram: rk3399: Fix faulty frequency change reports

Frequency changes to 400MHz are presently reported as:

  lpddr4_set_rate_0: change freq to 400000000 mhz 0, 1

This is obviously wrong by 6 orders of magnitude.

Ensure frequency changes are reported accurately.

Signed-off-by: Lee Jones <lee@kernel.org>
Tested-by: Xavier Drudis Ferran <xdrudis@tinet.cat>
Reviewed-by: Kever Yang <kever.yang@rock-chips.com>
2 years agoram: rk3399: Fix .set_rate_index() error handling
Lee Jones [Thu, 11 Aug 2022 07:58:46 +0000 (08:58 +0100)]
ram: rk3399: Fix .set_rate_index() error handling

Functions pointed to by this op pointer can return non-zero values
indicating an error.  Ensure any error value is propagated back up the
call-chain.

Signed-off-by: Lee Jones <lee@kernel.org>
Tested-by: Xavier Drudis Ferran <xdrudis@tinet.cat>
Reviewed-by: Kever Yang <kever.yang@rock-chips.com>
2 years agodrivers: ram: rockchip: Fix dram channels calculation for rk3399
Han Pengfei [Sun, 15 May 2022 06:11:59 +0000 (14:11 +0800)]
drivers: ram: rockchip: Fix dram channels calculation for rk3399

Only add the dram channel when we finally setup it successfully at the
last step.

Signed-off-by: Han Pengfei <pengphei@foxmail.com>
Reviewed-by: Kever Yang <kever.yang@rock-chips.com>