[REGRESSION] rk3588s-evb1-v10 board unable to boot based on linux-next

[Heiko Stübner]

Hi Shawn,

Am Donnerstag, 12. Februar 2026, 14:09:13 Mitteleuropäische Normalzeit schrieb Shawn Lin:
> 在 2026/02/12 星期四 18:48, Heiko Stübner 写道:
> > Am Donnerstag, 12. Februar 2026, 11:35:14 Mitteleuropäische Normalzeit schrieb Shawn Lin:
> >> + Sebastian Reichel
> >>
> >> 在 2026/02/12 星期四 15:59, Shawn Lin 写道:
> >>> Hi Detlev,
> >>>
> >>> commit f61731bd6062 ("arm64: dts: rockchip: Add the vdpu381 Video
> >>> Decoders on RK3588") landed on linux-next breaks rk3588s-evb1-v10.
> >>> Revert it fixes the problem. The full log could be found here[1].
> >>> Also confirm that it doesn't work based on Linus' latest mainline
> >>> as this commit merged into it this cycle.
> >>>
> >>> If anything I could do to help debug it, please let me know.
> >>>
> >>
> >> I did some quick debug and found it's related to a pmdomain patch.
> >> Commit f61731bd6062 happened to enable rkvdec which relys on the
> >> PD_VDEC*, so I have sent a fix patch for you to help review:
> >>
> >> https://lore.kernel.org/linux-rockchip/1770891364-52147-1-git-send-email-shawn.lin@rock-chips.com/T/#u
> > 
> > do you want to provide a patch for the evb1 board to that adds the
> > domain-supplies, so that people can see what is needed to make
> > the vdec work?
> > 
> 
> I do have a plan to add it for evb1 board. Btw，IIRC, peripheral IP
> should be kept as disalbed in soc level dtsi, and then set enabled in
> individual board dts. But rkvdec stuffs in rk3588-base.dtsi seems to
> be an exception?

There is a difference between peripherals _on_ the board, vs. peripherals
that exists _only_ inside the SoC.

Like we keep the sdmmc disabled if there is no card-slot/card on the board.
The video decoders/encoders/etc are only ever inside the SoC with no
relation to the board design itself, so they are _always_ available.
Boards just need to configure the supply relation properly.

With your patch for the pm-domain, we'll make sure, the vdec will not probe
if the board-devicetree is set up wrongly and missing the supply declaration.


Heiko