On Wed, Dec 14, 2022 at 07:22:40PM +0100, Sebastian Reichel wrote:
> Hi,
>
> This adds initial rk3588(s) DT including two different board
> devicetrees. All required driver changes have been merged into
> the respective maintainer trees. There is one warning from the
> DT check:
>
> $ make CHECK_DTBS=y rockchip/rk3588-evb1-v10.dtb rockchip/rk3588s-rock-5a.dtb
> DTC_CHK arch/arm64/boot/dts/rockchip/rk3588-evb1-v10.dtb
> /home/sre/src/collabora/rode/linux-rockchip-upstream/arch/arm64/boot/dts/rockchip/rk3588-evb1-v10.dtb:
> ethernet@fe1c0000: Unevaluated properties are not allowed ('interrupt-names', 'interrupts', 'mdio',
> 'power-domains', 'reg', 'reset-names', 'resets', 'rx-queues-config', 'snps,axi-config', 'snps,mixed-burst',
> 'snps,mtl-rx-config', 'snps,mtl-tx-config', 'snps,tso', 'stmmac-axi-config', 'tx-queues-config' were unexpected)
> From schema: /home/sre/src/collabora/rode/linux-rockchip-upstream/Documentation/devicetree/bindings/net/rockchip-dwmac.yaml
>
> This is for gmac1. gmac0 has the same properties and there is no warning. Also
> rk3588s (and thus the Rock 5A) has only gmac1 and there is no warning for the
> Rock 5A. It looks like for some reason the referenced "snps,dwmac.yaml#"
> is only checked for the first node. I think it's a bug in dt-validate.
> Also the same issue can be seen with rk356x.
I believe the issue is fixed in v2022.12. I didn't reproduce it. The
problem is with disabled nodes. In some cases 'required' failing in turn
triggers unevaluated property failures. So we now filter out both
errors. But that's still not enough in some cases, so the only thing
that works is just removing disabled nodes before validation. That's not
great if it is something where the bootloader enables the node and also
means whomever enables a block for a board gets to fix the issues.
BTW, I do also see this warning:
arch/arm64/boot/dts/rockchip/rk3588-evb1-v10.dtb:0:0: /i2c@feaa0000/rtc@51: failed to match any schema with compatible: ['haoyu,hym8563']
Rob
