Andrew Davis [Mon, 21 Aug 2023 13:59:10 +0000 (08:59 -0500)]
configs: Enable CONFIG_DM_SCSI in am57xx_hs_evm_usb
This should have already been enabled but was missed when converting the
base platform defconfig, fix this here.
Fixes: 3c5aa6caccab ("configs: Enable CONFIG_BLK in am57xx_evm and am57xx_hs_evm") Signed-off-by: Andrew Davis <afd@ti.com> Reviewed-by: Tom Rini <trini@konsulko.com> Reviewed-by: Simon Glass <sjg@chromium.org>
Ilias Apalodimas [Thu, 24 Aug 2023 14:21:09 +0000 (17:21 +0300)]
efi_loader: delete handle from events when a protocol is uninstalled
When a notification event is registered for a protocol the handle of the
protocol is added in our event notification list. When all the protocols
of the handle are uninstalled we delete the handle but we do not remove
it from the event notification list.
Clean up the protocol removal functions and add a wrapper which
- Removes the to-be deleted handle from any lists it participates
- Remove the handle if no more protocols are present
Signed-off-by: Ilias Apalodimas <ilias.apalodimas@linaro.org> Reviewed-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
Nishanth Menon [Thu, 24 Aug 2023 15:40:36 +0000 (10:40 -0500)]
doc: board: ti: k3: Convert to sphinx-prompt
Sphinx-prompt provides a handy scheme to provide documentation that
renders nicely and yet provides a scheme to copy paste for users without
having to hand-edit the copied text as is the result of code-block
[1] https://lore.kernel.org/all/87fs48rgto.fsf@baylibre.com/ Reported-by: Simon Glass <sjg@chromium.org> Suggested-by: Mattijs Korpershoek <mkorpershoek@baylibre.com> Signed-off-by: Nishanth Menon <nm@ti.com> Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
Nishanth Menon [Thu, 24 Aug 2023 15:40:35 +0000 (10:40 -0500)]
doc: sphinx: Add sphinx-prompt
Sphinx-prompt[1] helps bring-in '.. prompt::' option that allows a
better rendered documentation, yet be able to copy paste without
picking up the prompt from rendered documentation.
Signed-off-by: Sumit Garg <sumit.garg@linaro.org>
Use code-block. Fix length of two heading underlines. Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
Do not assume that partitions are numbered continuously starting at 1.
Only a single partition table type can exist on a block device. If we found
a GPT partition table, we must not re-enumerate with the MBR partition
driver which would find the protective partition.
Fixes: 12fc1f3bb223 ("cmd: gpt: add eMMC and GPT support") Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com> Reviewed-by: Simon Glass <sjg@chromium.org>
If hash_calculate is invoked with region_count = 0, it will try to hash
INT_MAX regions. We should check this parameter.
* Avoid a comparison with different signedness.
* Check that region_count is at least 1.
* Avoid a superfluous assignment.
Fixes: b37b46f042cc ("rsa: Use checksum algorithms from struct hash_algo") Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com> Reviewed-by: Simon Glass <sjg@chromium.org>
Patrick Delaunay [Mon, 10 Jul 2023 08:44:10 +0000 (10:44 +0200)]
ARM: dts: stm32mp13: remove shmem for scmi-optee
CFG_STM32MP1_SCMI_SHM_SYSRAM will be disabled by default for STM32MP13x
SoCs in next OP-TEE version and the OP-TEE SMCI server uses the OP-TEE
native shared memory registered by clients.
To be compatible by default with this configuration this patch removes
the shared memory in the SCMI configuration and the associated reserved
memory in SRAM.
Signed-off-by: Patrick Delaunay <patrick.delaunay@foss.st.com> Reviewed-by: Patrice Chotard <patrice.chotard@foss.st.com>
Patrick Delaunay [Mon, 10 Jul 2023 08:44:09 +0000 (10:44 +0200)]
ARM: dts: stm32mp15: remove shmem for scmi-optee
Since OP-TEE commit 89ba3422ee80 ("plat-stm32mp1: scmi_server: default
use OP-TEE shared memory"), integrated in OP-TEE 3.22.0-rc1
the default configuration for STM32MP15x SoCs changes,
CFG_STM32MP1_SCMI_SHM_SYSRAM is disabled by default and the OP-TEE SMCI
server uses ithe OP-TEE native shared memory registered by clients.
To be compatible by default with this configuration and the next OP-TEE
version, this patch removes the SHMEM in the SCMI configuration and the
associated reserved memory in the last 4KByte page of SRAM,
in the STM32MP15 device tree.
Signed-off-by: Patrick Delaunay <patrick.delaunay@foss.st.com> Reviewed-by: Patrice Chotard <patrice.chotard@foss.st.com>
Tim Harvey [Thu, 13 Jul 2023 18:56:07 +0000 (11:56 -0700)]
phy: phy-imx8mq-usb: add vbus regulator support
Add support for enabling and disabling vbus-supply regulator found
on several imx8mp boards in the usb3_phy0 and usb3_phy1 nodes.
Signed-off-by: Tim Harvey <tharvey@gateworks.com> Reviewed-by: Adam Ford <aford173@gmail.com> Reviewed-by: Marek Vasut <marex@denx.de> Tested-by: Marcel Ziswiler <marcel.ziswiler@toradex.com>
usb: dwc3: Fix remove function if there is no ulpi_reset gpio
As ulpi_reset gpio is now optional, we need to check it when doing
the 'dwc3_generic_remove' function. Check if it is declared before
accessing the ulpi_reset.
Fixes: 237d1f60b1d ("usb: dwc3: Use the devm_gpiod_get_optional()
API for reset gpio")
Reported-by: Thomas Nizan <tnizan@witekio.com> Signed-off-by: Venkatesh Yadav Abbarapu <venkatesh.abbarapu@amd.com>
Tom Rini [Wed, 23 Aug 2023 15:03:46 +0000 (11:03 -0400)]
Merge tag 'doc-2023-10-rc4' of https://source.denx.de/u-boot/custodians/u-boot-efi
Pull request for doc-2023-10-rc4
Documentation:
* Fix documentation for TI boards
* Describe running on VirtualBox
* List build dependencies for building documenation
* Add references to U-Boot talks
Other:
* Fix error handling in the setexpr command (printf_str)
Documentation extended with specific information for VirtualBox
The configuration for EFI is enhanced to start U-Boot at VirtualBox (x86_64)
with all features like booting with help of boot scripts and extended linux.
This documentation describes how to use U-Boot at VirtualBox to boot
Linux systems with a some simple example.
Signed-off-by: Thomas Mittelstaedt <thomas.mittelstaedt@de.bosch.com>
doc: board: ti: k3: Fix up OpenOCD references and debug info
Fix minor path and config macro name updates to sync with latest
OpenOCD and U-Boot configurations.
Fixes: effe50854a69 ("doc: board: ti: k3: Add a guide to debugging with OpenOCD") Signed-off-by: Jonathan Humphreys <j-humphreys@ti.com> Signed-off-by: Nishanth Menon <nm@ti.com> Reviewed-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
Nishanth Menon [Tue, 22 Aug 2023 16:41:01 +0000 (11:41 -0500)]
doc: board: ti: j721e: provide image alt text
Provide alternative text for image.
Fixes: 3b83dff183b5 ("doc: board: ti: j721e: Convert the image format to svg") Reported-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com> Signed-off-by: Nishanth Menon <nm@ti.com>
Nishanth Menon [Tue, 22 Aug 2023 16:41:00 +0000 (11:41 -0500)]
doc: board: ti: j7200: provide image alt text
Provide alternative text for image.
Fixes: f4ade09a1e76 ("doc: board: ti: j7200: Convert the image format to svg") Reported-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com> Signed-off-by: Nishanth Menon <nm@ti.com>
Fixes: 34f76921d801 ("doc: board: ti: am62x: Convert the image format to svg") Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com> Signed-off-by: Nishanth Menon <nm@ti.com>
Chanho Park [Fri, 18 Aug 2023 05:11:03 +0000 (14:11 +0900)]
riscv: cpu: make riscv_cpu_probe to EVT_DM_POST_INIT_R callback
Since the Patch 55171aedda88, VisionFive2 booting has been broken [1].
VisionFive2 board requires to enable CONFIG_TIMER_EARLY but booting went
to panic from initr_dm_devices due to lack of a timer device.
Thus, we need to move riscv_cpu_probe function in order to register
the timer earlier than initr_dm_devices.
Fixes: 7fe32b3442f0 ("event: Convert arch_cpu_init_dm() to use events") Cc: Simon Glass <sjg@chromium.org> Cc: Bin Meng <bmeng.cn@gmail.com> Signed-off-by: Chanho Park <chanho61.park@samsung.com> Tested-by: Milan P. Stanić <mps@arvanta.net> Tested-by: Roland Ruckerbauer <mail@ruabmbua.dev> Tested-by: Roland Ruckerbauer <mail@ruabmbua.dev>
Chanho Park [Fri, 18 Aug 2023 05:11:02 +0000 (14:11 +0900)]
dm: event: add EVT_DM_POST_INIT_R event type
This patch introduces EVT_DM_POST_INIT_R event type for handling hooks
after relocation.
Fixes: 55171aedda88 ("dm: Emit the arch_cpu_init_dm() even only before relocation") Suggested-by: Simon Glass <sjg@chromium.org> Cc: Bin Meng <bmeng.cn@gmail.com> Signed-off-by: Chanho Park <chanho61.park@samsung.com> Tested-by: Milan P. Stanić <mps@arvanta.net> Reviewed-by: Simon Glass <sjg@chromium.org> Tested-by: Roland Ruckerbauer <mail@ruabmbua.dev> Tested-by: Roland Ruckerbauer <mail@ruabmbua.dev>
Fixed missing event name in event.c: Signed-off-by: Simon Glass <sjg@chromium.org>
Marek Vasut [Sun, 13 Aug 2023 19:52:19 +0000 (21:52 +0200)]
test: cpu: Handle both 32bit and 64bit CPUs
Handle both 32bit and 64bit systems, i.e. sandbox and sandbox64
the same way drivers/cpu/cpu_sandbox.c does, that is in case
CONFIG_PHYS_64BIT is enabled, assume 64bit address width, else
assume 32bit address width. This fixes ut_dm_dm_test_cpu test
failure on sandbox64.
Signed-off-by: Marek Vasut <marek.vasut+renesas@mailbox.org> Reviewed-by: Simon Glass <sjg@chromium.org>
Marek Vasut [Sun, 13 Aug 2023 05:15:05 +0000 (07:15 +0200)]
configs: sandbox64: Enable video 16bpp and 24bpp support
Align the sandbox64 defconfig with sandbox defconfig. Enable missing
16bpp and 24bpp video support. This fixes ut_dm_dm_test_video_bmp16
and ut_dm_dm_test_video_bmp24 tests .
Signed-off-by: Marek Vasut <marek.vasut+renesas@mailbox.org> Reviewed-by: Simon Glass <sjg@chromium.org>
Marek Vasut [Sun, 13 Aug 2023 03:32:08 +0000 (05:32 +0200)]
test: dm: pinmux: Handle %pa in pinctrl-single mux output
The pinctrl-single driver uses %pa to print register value
in its single_get_pin_muxing() output. Handle this properly
in the test based on CONFIG_PHYS_64BIT .
Signed-off-by: Marek Vasut <marek.vasut+renesas@mailbox.org> Reviewed-by: Simon Glass <sjg@chromium.org>
Paul Barker [Fri, 18 Aug 2023 13:17:21 +0000 (14:17 +0100)]
arm: rmobile: Fix off-by-one error in cpuinfo
In rmobile_cpuinfo_idx() there is an off-by-one error in accessing the
rmobile_cpuinfo array.
At the end of the loop, i is equal to the array size, i.e.
rmobile_cpuinfo[i] accesses one entry past the end of the array. The
last entry in the array is a fallback value so the loop should count to
ARRAY_SIZE(rmobile_cpuinfo) - 1 instead, this will leave i equal to the
index of the fallback value if no match is found.
Signed-off-by: Paul Barker <paul.barker.ct@bp.renesas.com> Reviewed-by: Biju Das <biju.das.jz@bp.renesas.com> Reviewed-by: Marek Vasut <marek.vasut+renesas@mailbox.org>
Tom Rini [Sat, 19 Aug 2023 14:13:28 +0000 (10:13 -0400)]
Merge tag 'doc-2023-10-rc3-2' of https://source.denx.de/u-boot/custodians/u-boot-efi
Pull request doc-2023-10-rc3-2
Documentation:
* csf_examples: csf.sh: Remove unneeded export ATF_LOAD_ADDR line
* printf() codes: correct format specifier for unsigned int
* Fix typos in clk.h, irq.h.
* Correct description of proftool
Other:
* Quieten test for erofs filesystem presence
* spl: don't assume NVMe partition 1 exists
Fabio Estevam [Tue, 15 Aug 2023 13:48:01 +0000 (10:48 -0300)]
doc: csf_examples: csf.sh: Remove unneeded export ATF_LOAD_ADDR line
Originally, exporting the ATF_LOAD_ADDR was required, but since binman has
been used to generate the flash.bin, it is no longer needed to do
such manual export.
There is no requirement that a partition 1 exists in a partition table.
We should not try to retrieve information about it.
We should not even try reading with partition number
CONFIG_SYS_NVME_BOOT_PARTITION here as this is done in the fs_set_blk_dev()
call anyway.
Fixes: 8ce6a2e17577 ("spl: blk: Support loading images from fs") Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com> Reviewed-by: Tom Rini <trini@konsulko.com>
Jonas Karlman [Thu, 17 Aug 2023 21:52:48 +0000 (21:52 +0000)]
rockchip: rk3566-anbernic-rgxx3: Rename defconfig to include SoC name
Rename defconfig to include SoC name, use similar pattern as other
RK356x boards: <soc>-<name>.dts -> <name>-<soc>_defconfig
Suggested-by: Kever Yang <kever.yang@rock-chips.com> Signed-off-by: Jonas Karlman <jonas@kwiboo.se> Reviewed-by: Kever Yang <kever.yang@rock-chips.com>
Tom Rini [Fri, 18 Aug 2023 14:05:04 +0000 (10:05 -0400)]
Merge tag 'tegra-for-2023.10-rc1' of https://source.denx.de/u-boot/custodians/u-boot-tegra
ARM: tegra: Changes for v2023.10-rc1
This adds support for various new Tegra30 boards (ASUS, LG and HTC) and
has some other minor enhancements, such as enabling the poweroff command
on several Tegra210 and Tegra186 boards.
Jonas Karlman [Thu, 17 Aug 2023 06:04:34 +0000 (06:04 +0000)]
doc: rockchip: Add supported RK3566/RK3568 boards
Update Rockchip documentation to include RK3566/RK3568 boards already
supported. Also list Pine64 boards under RK3566 and drop defconfig to
match other listed boards.
Signed-off-by: Abdellatif El Khlifi <abdellatif.elkhlifi@arm.com> Cc: Tom Rini <trini@konsulko.com> Cc: Simon Glass <sjg@chromium.org> Reviewed-by: Simon Glass <sjg@chromium.org>
Tom Rini [Mon, 7 Aug 2023 21:08:11 +0000 (17:08 -0400)]
arm: Add arch/arm/dts/Makefile specifically to MAINTAINERS
In order to reduce the number of people that are cc'd on a patch for
simply touching arch/arm/dts/Makefile (which is a big common file) add
an entry specifically to MAINTAINERS under the ARM entry.
Signed-off-by: Tom Rini <trini@konsulko.com> Reviewed-by: Simon Glass <sjg@chromium.org>
Tom Rini [Wed, 16 Aug 2023 15:23:58 +0000 (11:23 -0400)]
Merge tag 'u-boot-stm32-20230816' of https://source.denx.de/u-boot/custodians/u-boot-stm
DHSOM: Power cycle Buck3 in reset
DHCOM: Switch DWMAC RMII clock to MCO2
stm32f746: fix display pinmux
stm32mp: psci: Inhibit PDDS because CSTBYDIS is set
stm32mp1: DT alignment with v6.4
stm32mp1: add splashscreen with STMicroelectronics logo
stm32mp1: clk: remove error for disabled clock in stm32mp1_clk_get_parent
serial: stm32: Extend TC timeout
Marek Vasut [Wed, 26 Jul 2023 23:58:07 +0000 (01:58 +0200)]
ARM: dts: stm32: Switch DWMAC RMII clock to MCO2 on DHCOM
The DHCOM SoM has two options for supplying ETHRX clock to the DWMAC
block and PHY. Either (1) ETHCK_K generates 50 MHz clock on ETH_CLK
pad for the PHY and the same 50 MHz clock are fed back to ETHRX via
internal eth_clk_fb clock connection OR (2) ETH_CLK is not used at
all, MCO2 generates 50 MHz clock on MCO2 output pad for the PHY and
the same MCO2 clock are fed back into ETHRX via ETH_RX_CLK input pad
using external pad-to-pad connection.
Option (1) has two downsides. ETHCK_K is supplied directly from either
PLL3_Q or PLL4_P, hence the PLL output is limited to exactly 50 MHz and
since the same PLL output is also used to supply SDMMC blocks, the
performance of SD and eMMC access is affected. The second downside is
that using this option, the EMI of the SoM is higher.
Option (2) solves both of those problems, so implement it here. In this
case, the PLL4_P is no longer limited and can be operated faster, at
100 MHz, which improves SDMMC performance (read performance is improved
from ~41 MiB/s to ~57 MiB/s with dd if=/dev/mmcblk1 of=/dev/null bs=64M
count=1). The EMI interference also decreases.
Ported from Linux kernel commit 73ab99aad50cd ("ARM: dts: stm32: Switch DWMAC RMII clock to MCO2 on DHCOM")
Signed-off-by: Marek Vasut <marex@denx.de> Reviewed-by: Patrice Chotard <patrice.chotard@foss.st.com>
Marek Vasut [Thu, 6 Jul 2023 21:32:27 +0000 (23:32 +0200)]
ARM: stm32: Inhibit PDDS because CSTBYDIS is set
The PWR_MPUCR CSTBYDIS bit is set, therefore the CA cores can never
enter CStandby state and would always end up in CStop state. Clear
the PDDS bit, which indicates the CA cores can enter CStandby state
as it makes little sense to keep it set with CSTBYDIS also set.
This does however fix a problem too. When both PWR_MPUCR and PWR_MCUCR
PDDS bits are set, then the chip enters CStandby state even though the
PWR_MCUCR CSTBYDIS is set. Clearing the PWR_MPUCR PDDS prevents that
from happening.
Signed-off-by: Marek Vasut <marex@denx.de> Reviewed-by: Patrick Delaunay <patrick.delaunay@foss.st.com>
Patrick Delaunay [Fri, 23 Jun 2023 13:05:16 +0000 (15:05 +0200)]
clk: stm32mp1: remove error for disabled clock in stm32mp1_clk_get_parent
To disabled a clock in clock tree initialization for a mux of STM32MP15,
the selected clock source index is set with the latest possible index for
the number of bit used. Today this valid configuration cause a error
in U-Boot messages, for example with CLK_ETH_DISABLED, when this clock
is not needed for the used ETH PHY without crystal:
no parents defined for clk id 123
This patch change the level of this message to avoid this trace for
valid clock tree.
Signed-off-by: Patrick Delaunay <patrick.delaunay@foss.st.com> Reviewed-by: Patrice Chotard <patrice.chotard@foss.st.com>
Marek Vasut [Wed, 17 May 2023 22:02:39 +0000 (00:02 +0200)]
ARM: stm32: Power cycle Buck3 in reset on DHSOM
In case the DHSOM is in suspend state and either reset button is pushed
or IWDG2 triggers a watchdog reset, then DRAM initialization could fail
as follows:
Avoid this failure by not keeping any Buck regulators enabled during reset,
let the SoC and DRAMs power cycle fully. Since the change which keeps Buck3
VDD enabled during reset is ST specific, move this addition to ST specific
SPL board initialization so that it wouldn't affect the DHSOM .
Signed-off-by: Marek Vasut <marex@denx.de> Reviewed-by: Patrick Delaunay <patrick.delaunay@foss.st.com>
Tom Rini [Tue, 15 Aug 2023 17:08:17 +0000 (13:08 -0400)]
Merge tag 'efi-2023-10-rc3' of https://source.denx.de/u-boot/custodians/u-boot-efi
Pull request for efi-2023-10-rc3
Documentation:
* Correct description of board_get_usable_ram_top
* Add partition API to HTML documentation
* Describe lmb_is_reserved
* doc/sphinx/requirements.txt: Bump certifi up
UEFI:
* Fix efi_add_known_memory
* Make distro_efi_boot() static
board_get_usable_ram_top() returns a physical address that is stored in
gd->ram_top. The return type of the function should be phys_addr_t like the
current type of gd->ram_top.
Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
In efi_add_known_memory() we currently call board_get_usable_ram_top() with
an incorrect value 0 of parameter total_size. This leads to an incorrect
value for ram_top depending on the code in board_get_usable_ram_top().
Use the value of gd->ram_top instead which is set before relocation by
calling board_get_usable_ram_top().
Fixes: 7b78d6438a2b ("efi_loader: Reserve unaccessible memory") Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
Bin Meng [Thu, 3 Aug 2023 09:30:05 +0000 (17:30 +0800)]
bootmeth: efi: Make distro_efi_boot() static
As it is only called in bootmeth_efi.c
Signed-off-by: Bin Meng <bmeng@tinylab.org> Reviewed-by: Simon Glass <sjg@chromium.org> Reviewed-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
Sam Edwards [Tue, 25 Jul 2023 22:13:05 +0000 (16:13 -0600)]
i2c: mvtwsi: reset controller if stuck in "bus error" state
The MVTWSI controller can act either as a master or slave device. When
acting as a master, the FSM is driven by the CPU. As a slave, the FSM is
driven by the bus directly. In what is (apparently) a safety mechanism,
if the bus transitions our FSM in any improper way, the FSM goes to a
"bus error" state (0x00). I could find no documented or experimental way
to get the FSM out of this state, except for a controller reset.
Since U-Boot only uses the MVTWSI controller as a bus master, this
feature only gets in the way: we do not care what happened on the bus
previously as long as the bus is ready for a new transaction. So, when
trying to start a new transaction, check for this state and reset the
controller if necessary.
Note that this should not be confused with the "deblocking" technique
(used by the `i2c reset` command), which involves pulsing SCL repeatedly
if SDA is found to be held low, in an attempt to force the bus back to
an idle state. This patch only resets the controller in case something
else had previously upset it, and (in principle) results in no
externally-observable change in behavior.
Signed-off-by: Sam Edwards <CFSworks@gmail.com> Reviewed-by: Stefan Roese <sr@denx.de> Reviewed-by: Heiko Schocher <hs@denx.de>
Jonas Karlman [Mon, 14 Aug 2023 00:28:26 +0000 (00:28 +0000)]
pinctrl: rockchip: Fix drive and input schmitt on RK3568
On RK3568 most pins have a configurable drive strength of level 0-5 and
some pins level 0-11. When rk3568_set_drive is called with a strength
value above 7 the drv value written to reg may overflow into the write
enable bits, resulting in a bad configuration.
This cause e.g. ethernet PHY on Radxa CM3-IO board not to work after
drive is configured according to the device tree.
Level 6-11 can be configured using a second reg for some pins, however
the drv value is reused resulting in lower 6 bits being written to reg.
Input schmitt is configured in 2-bit fields on RK3568 compared to
earlier generation and 2'b10 should be used to enable input schmitt.
Change to use regmap_update_bits with a rmask to fix the overflow issue
and closer match the linux driver. Bit shift the drv value used for the
second reg to configure drive strength level 6-11. Also write correct
values for input schmitt setting.
Fixes: 1977d746aa54 ("rockchip: rk3568: add rk3568 pinctrl driver") Signed-off-by: Jonas Karlman <jonas@kwiboo.se> Reviewed-by: Simon Glass <sjg@chromium.org> Reviewed-by: Kever Yang <kever.yang@rock-chips.com>