Re: [PATCH v1 1/2] arm64: dts: add missing sram-supply to mali gpu nodes

[AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>]

Il 15/02/26 11:02, Onur Özkan ha scritto:
> Some arm64 Mali GPU DT nodes define mali-supply but
> are missing sram-supply.
> 
> As discussed in [1], sram-supply should be explicitly
> defined for all relevant compatibles. If there is no
> separate SRAM rail, it can reference the same regulator
> as mali-supply.
> 
> This change adds the missing sram-supply properties to
> affected DT files. This aligns DTs with expected regulator
> modeling and avoids relying on implicit fallback behavior.
> 
> [1]: https://lore.kernel.org/all/20260213155937.6af75786@nimda/
> 

Sorry, at least for MediaTek, this makes no sense. The SRAM supply on MTK platforms
is managed by the mediatek-regulator-coupler driver, as it scales in a peculiar way
in relation to the GPU core supply.

Besides, you're cloning the GPU core supply to the sram supply, which is, well,
simply wrong... and you're changing devicetrees of SoCs that are running on
panfrost (not on panthor), which will anyway ignore the existance of the SRAM
supply.

That's also going to give all sorts of binding warnings, I guess, but I didn't
check anyway.

While the change makes sense, as in, I agree with Mark in that the SRAM must have
a power supply in order to actually work... I'm afraid that what you're doing here
needs a bit more work before it becomes acceptable.

Especially - faking the vgpu-core regulator (not just on MTK, but on all) to also
be the sram regulator leads to confusion and is, well, plain wrong.

Cheers,
Angelo