Hi Svyatoslav,
On 3/2/26 3:24 PM, Svyatoslav Ryhel wrote:
пн, 2 бер. 2026 р. о 16:13 Quentin Schulz <[email protected]> пише:
Hi Svyatoslav,
On 2/9/26 7:39 PM, Svyatoslav Ryhel wrote:
Add a condition to enable the PMIC sysreset function via the
system-power-controller device tree property in addition to the existing
Kconfig dependency, provided the PMIC supports it.
Reviewed-by: Quentin Schulz <[email protected]> # RK8xx
I've added in Cc people who have contributed to devices that will lose
support for sysreset once this gets merged.
Because they are missing a system-power-controller property in the RK8xx
PMIC Device Tree node, the following device trees will be impacted:
dts/upstream/src/arm64/amlogic/meson-g12b-odroid-go-ultra.dts
dts/upstream/src/arm64/rockchip/rk3326-anbernic-rg351m.dtsi
dts/upstream/src/arm64/rockchip/rk3326-odroid-go.dtsi
dts/upstream/src/arm64/rockchip/rk3399-evb-ind.dts
dts/upstream/src/arm64/rockchip/rk3566-anbernic-rgxx3.dtsi
dts/upstream/src/arm64/rockchip/rk3566-powkiddy-rk2023.dtsi
dts/upstream/src/arm64/rockchip/rk3566-powkiddy-x55.dts
dts/upstream/src/arm64/rockchip/rk3588-coolpi-cm5.dtsi
dts/upstream/src/arm64/rockchip/rk3588s-coolpi-4b.dts
dts/upstream/src/arm64/rockchip/rk3588s-indiedroid-nova.dts
If it's an oversight, you may want to add the property in a -u-boot.dtsi
and upstream the property to the kernel DTS to avoid carrying it
forever. From my side, this patch does what's expected based on the
(lack of) presence of the property so this is a green light for me
regardless if the above Device Trees are actually wrong.
As far as I could tell, the sysreset driver only implements the poweroff
functionality which can be called by either the poweroff CLI command or
the efi_reset_system_boottime function (no clue when/where this is
called, some EFI thing I guess).
I actually expected such an outcome. If you are fine with it, I will
drop the Rockchip PMIC change for now. It would be quite unpleasant if
a bunch of devices suddenly lose their reset function, especially
since updating upstream trees can be a long process. Once the trees
are updated, anyone working with Rockchip can submit a similar patch.
Well, how does one know whether the property is not there because it
shouldn't be there (the PMIC shouldn't/isn't capable of controlling the
system's power state) or because it's missing?
As far as I'm concerned, we can keep the RK8xx part in this patch. We
shouldn't need to account for broken Device Trees so we can only assume
the aforementioned DTSes are correct. The U-Boot-specific Device Trees
all contain the system-power-controller property so those are fine.
As for the DTSes coming from the kernel, they probably would have
noticed by now since the kernel also adds a poweroff (sys_off) handler
(and for a few PMICs a restart handler too) when this property is
present. Thus, this patch syncs behavior between Linux's and U-Boot's
implementation.
I'm fine keeping this, I just notified other users in case they didn't
know about the property and want to add it to their device. It shouldn't
block this patch though in my opinion.
Note I am only speaking about devices with RK8xx PMICs, I haven't
checked for the other PMICs touched by this patch.
Cheers,
Quentin
