Re: [PATCH 2/3] dt-bindings: arm: mediatek: convert PCIESYS to the json-schema

[AngeloGioacchino Del Regno]

Il 30/01/24 21:34, Rob Herring ha scritto:
> On Tue, Jan 23, 2024 at 12:20:29PM +0100, AngeloGioacchino Del Regno wrote:
> > Il 23/01/24 09:20, Rafał Miłecki ha scritto:
> > > From: Rafał Miłecki <rafal@xxxxxxxxxx>
> > >
> > > This helps validating DTS files. Introduced changes:
> > > 1. Documented "reg" property
> > > 2. Adjusted "reg" in example
> > >
> > > Signed-off-by: Rafał Miłecki <rafal@xxxxxxxxxx>
> > > ---
> > >    .../arm/mediatek/mediatek,mt7622-pciesys.yaml | 47 +++++++++++++++++++
> > >    .../arm/mediatek/mediatek,pciesys.txt         | 25 ----------
> > >    2 files changed, 47 insertions(+), 25 deletions(-)
> > >    create mode 100644 Documentation/devicetree/bindings/arm/mediatek/mediatek,mt7622-pciesys.yaml
> > >    delete mode 100644 Documentation/devicetree/bindings/arm/mediatek/mediatek,pciesys.txt
> > >
> > > diff --git a/Documentation/devicetree/bindings/arm/mediatek/mediatek,mt7622-pciesys.yaml b/Documentation/devicetree/bindings/arm/mediatek/mediatek,mt7622-pciesys.yaml
> > > new file mode 100644
> > > index 000000000000..7340a2512402
> > > --- /dev/null
> > > +++ b/Documentation/devicetree/bindings/arm/mediatek/mediatek,mt7622-pciesys.yaml
> >
> > I think that we should really move all those clock controller yaml files to their
> > proper directory, which would be
> >
> > Documentation/devicetree/bindings/clock/
> >
> > ...because those are clock controllers anyway and the fact that they do also
> > provide a reset controller doesn't really justify having them in arm/mediatek.
> >
> > Besides, I would appreciate if you could also move mt8186/92/95 and eventual
> > others that are there to clock/.
>
> Yes, please move it.
>
> > > @@ -0,0 +1,47 @@
> > > +# SPDX-License-Identifier: GPL-2.0-only OR BSD-2-Clause
> > > +%YAML 1.2
> > > +---
> > > +$id: http://devicetree.org/schemas/arm/mediatek/mediatek,mt7622-pciesys.yaml#
> > > +$schema: http://devicetree.org/meta-schemas/core.yaml#
> > > +
> > > +title: MediaTek PCIESYS controller
> > > +
> > > +description:
> > > +  The MediaTek PCIESYS controller provides various clocks to the system.
> > > +
> > > +maintainers:
> > > +  - Matthias Brugger <matthias.bgg@xxxxxxxxx>
> > > +
> > > +properties:
> > > +  compatible:
> > > +    items:
> > > +      - enum:
> > > +          - mediatek,mt7622-pciesys
> > > +          - mediatek,mt7629-pciesys
> > > +      - const: syscon
> >
> > I know that there's syscon all over the place and, even if I admit I didn't check,
> > I am fairly sure that there's absolutely no reason to have syscon there, and that
> > the syscon compatible never did anything for (most of, or all of) those clock
> > controllers, at all.
> >
> > I'm not sure - though - if removing syscon during the txt->yaml conversion is
> > acceptable (yeah we'd be cheating a bit), but something makes me say it is, because
> > the bindings couldn't validate before that one as well.
>
> As long as you state why you are removing it in the commit msg.
>
> > Of course you'd have to remove the syscon compatible from the affected device trees
> > as well as omitting it here.
>
> You could also do 'minItems: 1' and 'deprecated' in the 2nd item.

That too. Yes. But I still really want to see the syscon compatible removed from
the device trees on those nodes because it doesn't make any sense to have it :-)

> > However, to be sure that we're doing the right thing here, I have to summon someone
> > that can actually give a definitive answer to what I just said.....
> >
> > Krzysztof, please? :-)