[PATCH 3/6] dt-bindings: gpu: powervr-rogue: Add MediaTek MT8173 GPU

[Adam Ford]

On Thu, Jun 13, 2024 at 4:10 AM Chen-Yu Tsai <wenst at chromium.org> wrote:
>
> On Tue, Jun 4, 2024 at 12:18 PM Chen-Yu Tsai <wenst at chromium.org> wrote:
> >
> > On Fri, May 31, 2024 at 9:37 PM Frank Binns <Frank.Binns at imgtec.com> wrote:
> > >
> > > Hi ChenYu,
> > >
> > > On Thu, 2024-05-30 at 16:35 +0800, Chen-Yu Tsai wrote:
> > > > The MediaTek MT8173 comes with a PowerVR Rogue GX6250, which is one
> > > > of the Series6XT GPUs, another sub-family of the Rogue family.
> > >
> > > I've added Adam Ford who sent out some DT related patches [1] for the Renesas
> > > variant of GX6250 and the GX6650 (another Series6XT GPU).
> > >
> > > >
> > > > This was part of the very first few versions of the PowerVR submission,
> > > > but was later dropped. The compatible string has been updated to follow
> > > > the new naming scheme adopted for the AXE series.
> > > >
> > > > In a previous iteration of the PowerVR binding submission [1], the
> > > > number of clocks required for the 6XT family was mentioned to be
> > > > always 3. This is also reflected here.
> > > >
> > > > [1] https://lore.kernel.org/dri-devel/6eeccb26e09aad67fb30ffcd523c793a43c79c2a.camel@imgtec.com/
> > > >
> > > > Signed-off-by: Chen-Yu Tsai <wenst at chromium.org>
> > > > ---
> > > >  .../bindings/gpu/img,powervr-rogue.yaml       | 24 +++++++++++++++----
> > > >  1 file changed, 20 insertions(+), 4 deletions(-)
> > > >
> > > > diff --git a/Documentation/devicetree/bindings/gpu/img,powervr-rogue.yaml b/Documentation/devicetree/bindings/gpu/img,powervr-rogue.yaml
> > > > index 256e252f8087..48aa205b66b4 100644
> > > > --- a/Documentation/devicetree/bindings/gpu/img,powervr-rogue.yaml
> > > > +++ b/Documentation/devicetree/bindings/gpu/img,powervr-rogue.yaml
> > > > @@ -12,10 +12,17 @@ maintainers:
> > > >
> > > >  properties:
> > > >    compatible:
> > > > -    items:
> > > > -      - enum:
> > > > -          - ti,am62-gpu
> > > > -      - const: img,img-axe # IMG AXE GPU model/revision is fully discoverable
> > > > +    oneOf:
> > > > +      - items:
> > > > +          - enum:
> > > > +              - mediatek,mt8173-gpu
> > > > +          # PowerVR 6XT GPU model/revision is fully discoverable
> > > > +          - const: img,powervr-6xt
> > > > +      - items:
> > > > +          - enum:
> > > > +              - ti,am62-gpu
> > > > +          # IMG AXE GPU model/revision is fully discoverable
> > > > +          - const: img,img-axe
> > >
> > > The Series6XT GPU models have differing numbers of power domains (either 2, 4 or
> > > 5). Whereas, the AXE GPUs have a single power domain, so I assume there should
> > > be a related change here.
> > >
> > > The GX6250 has two power domains (lets call them A and B). There's a constraint
> > > that if domain B is powered then domain A must also be powered.
> > >
> > > In patch 6 [2] it's setting the power domain to MT8173_POWER_DOMAIN_MFG, which I
> > > believe corresponds to power domain B. I assume this works because the MTK power
> > > controller driver is encoding the constraint above, meaning that when we disable
> > > or enable MT8173_POWER_DOMAIN_MFG it's also disabling/enabling MT8173_POWER_DOMA
> > > IN_MFG_2D (domain A).
> >
> > It could also be that the power domains are split in the glue layer and there
> > is some sequencing handled there. I'll reach out to MediaTek to see if they
> > can dig up some design specifics.
>
> Unfortunately they said they no longer have that information.
>
> > I assume you would like to see the separate power domains properly modeled
> > in the device tree?
>
> So how should we go about this? Adam, do you have this information for
> your platform?

In the Renesas platform, I only had to add one clock [1] and one
power-domain [2] to get the GPU to come up.  In Renesas' downstream
driver, they only use one clock, but Geert, the Renesas tree
maintainer, sent me the three clocks to assign to the 6xt graphics if
it's determined that 3 clocks are required.  In terms of the power
domain, there appear to be 2, but one is dependent on another, so
turning on the 'B' power domain turns on the 'A' power domain
automatically.  I should note that I can't get the graphics to
function, since the 6xt isn't supported yet, so there could be some
elements missing that I am unaware of.

adam

[1] - https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git/commit/drivers/clk/renesas?h=next-20240613&id=f7b0dfffd3e0897ca73916a0c3d3fb61c61df51e
[2] - https://patchwork.kernel.org/project/linux-renesas-soc/patch/20240227034539.193573-3-aford173@gmail.com/



>
> Thanks
> ChenYu
>
> >
> > Thanks
> > ChenYu
> >
> > > Thanks
> > > Frank
> > >
> > > [1] https://lists.freedesktop.org/archives/dri-devel/2024-February/443548.html
> > > [2] https://lists.freedesktop.org/archives/dri-devel/2024-May/455833.html
> > >
> > > >
> > > >    reg:
> > > >      maxItems: 1
> > > > @@ -56,6 +63,15 @@ allOf:
> > > >        properties:
> > > >          clocks:
> > > >            maxItems: 1
> > > > +  - if:
> > > > +      properties:
> > > > +        compatible:
> > > > +          contains:
> > > > +            const: img,powervr-6xt
> > > > +    then:
> > > > +      properties:
> > > > +        clocks:
> > > > +          minItems: 3
> > > >
> > > >  examples:
> > > >    - |