Re: [PATCH 1/4] dt-bindings: clock: mediatek: Add clocks for MT8196 mfgpll

[Conor Dooley <conor@kernel.org>]

[-- Attachment #1: Type: text/plain, Size: 3329 bytes --]

On Tue, Sep 30, 2025 at 05:57:00PM +0200, Nicolas Frattaroli wrote:
> On Monday, 29 September 2025 19:31:36 Central European Summer Time Conor Dooley wrote:
> > On Mon, Sep 29, 2025 at 02:13:20PM +0200, Nicolas Frattaroli wrote:
> > > The clock controllers for mfgpll, mfgpll-sc0, and mfgpll-sc1 all need
> > > CLK_TOP_MFG_EB to be on if their clock control registers are touched in
> > > any way.
> > > 
> > > This was not known at the time this binding was written, as this
> > > dependency only came to light when I started poking at the MFlexGraphics
> > > hardware, where this undocumented peculiarity made itself known through
> > > SErrors being thrown during register reads.
> > > 
> > > Add a clocks property to the binding to describe this relationship, and
> > > mark it as required for the affected clocks.
> > > 
> > > Fixes: dd240e95f1be ("dt-bindings: clock: mediatek: Describe MT8196 clock controllers")
> > > Signed-off-by: Nicolas Frattaroli <nicolas.frattaroli@collabora.com>
> > > ---
> > >  .../bindings/clock/mediatek,mt8196-sys-clock.yaml  | 28 ++++++++++++++++++++++
> > >  1 file changed, 28 insertions(+)
> > > 
> > > diff --git a/Documentation/devicetree/bindings/clock/mediatek,mt8196-sys-clock.yaml b/Documentation/devicetree/bindings/clock/mediatek,mt8196-sys-clock.yaml
> > > index 660ab64f390d2e722b7d3e25cf057926da318bc0..41aacd8d5f69050eebdf8392f7b652427632f491 100644
> > > --- a/Documentation/devicetree/bindings/clock/mediatek,mt8196-sys-clock.yaml
> > > +++ b/Documentation/devicetree/bindings/clock/mediatek,mt8196-sys-clock.yaml
> > > @@ -45,6 +45,9 @@ properties:
> > >    reg:
> > >      maxItems: 1
> > >  
> > > +  clocks:
> > > +    maxItems: 1
> > > +
> > >    '#clock-cells':
> > >      const: 1
> > >  
> > > @@ -90,6 +93,23 @@ required:
> > >  
> > >  additionalProperties: false
> > >  
> > > +allOf:
> > > +  - if:
> > > +      properties:
> > > +        compatible:
> > > +          contains:
> > > +            enum:
> > > +              - mediatek,mt8196-mfgpll-pll-ctrl
> > > +              - mediatek,mt8196-mfgpll-sc0-pll-ctrl
> > > +              - mediatek,mt8196-mfgpll-sc1-pll-ctrl
> > > +    then:
> > > +      properties:
> > > +        clocks:
> > > +          items:
> > > +            - description: mfg_eb clock
> > > +      required:
> > > +        - clocks
> > 
> > Don't you want an else: properties: clocks: false here?
> 
> Possibly. I'm never quite sure how strict bindings should be when
> it comes to stuff like this. On the one hand, none of the other
> compatibles described in it use any clocks that we know of
> right now.

It's a bit case-by-case I suppose, but anything that is invalid and can
be easily prevented should be. Particularly in cases where information
about what's correct is harder to find.

> On the other, if we have a second set of compatibles that also
> needs clocks, but in a different way, would we repeat that for
> each such if/then condition? Or would be reformulate this as
> some oneOf/anyOf construct specifically for the clock property?

Depends on what's simpler to do. Sometimes it's better to have if/then
on a per property basis and sometimes separate entries under the allOf
per platform is. Usually depends on how unique each platform is.

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 228 bytes --]