Re: [PATCH 29/36] dt-bindings: arm: Convert Renesas board/soc bindings to json-schema

[Rob Herring]

On Mon, Oct 8, 2018 at 10:13 AM Geert Uytterhoeven <geert@xxxxxxxxxxxxxx> wrote:
>
> Hi Rob,
>
> On Mon, Oct 8, 2018 at 4:57 PM Rob Herring <robh@xxxxxxxxxx> wrote:
> > On Mon, Oct 8, 2018 at 2:47 AM Geert Uytterhoeven <geert@xxxxxxxxxxxxxx> wrote:
> > > On Fri, Oct 5, 2018 at 6:59 PM Rob Herring <robh@xxxxxxxxxx> wrote:
> > > > Convert Renesas SoC bindings to DT schema format using json-schema.
>
> > > > --- /dev/null
> > > > +++ b/Documentation/devicetree/bindings/arm/shmobile.yaml
> > > > @@ -0,0 +1,205 @@
>
> > > > +      - description: Kingfisher (SBEV-RCAR-KF-M03)
> > > > +        items:
> > > > +          - const: shimafuji,kingfisher
> > > > +          - enum:
> > > > +              - renesas,h3ulcb
> > > > +              - renesas,m3ulcb
> > > > +          - enum:
> > > > +              - renesas,r8a7795
> > > > +              - renesas,r8a7796
> > >
> > > This looks a bit funny: all other entries have the "const" last, and
> > > use it for the
> > > SoC number. May be correct, though.
> > > To clarify, this is an extension board that can fit both the [HM]3ULCB
> > > boards (actually also the new M3NULCB, I think).
> >
> > This being Kingfisher?
>
> Correct.
>
> > I wrote this based on dts files in the tree. There's 2 combinations that I see:
> >
> > "shimafuji,kingfisher", "renesas,h3ulcb", "renesas,r8a7795"
> > "shimafuji,kingfisher", "renesas,m3ulcb", "renesas,r8a7796"
> >
> > The schema allows 4 combinations (1 * 2 * 2). I have no idea if the
> > other combinations are possible. If not, then we could rewrite this as
> > 2 entries with 3 const values each.
>
> I expect there will soon be a third one:
>
>     "shimafuji,kingfisher", "renesas,m3nulcb", "renesas,r8a77965"
>
> Technically, {h3,m3,m3n}ulcb are the same board (although there may be
> minor revision differences), with a different SiP mounted.
> But they are called/marketed depending on which SiP is mounted.
>
> And on top of that, you can plug in a Kingfisher daughterboard. Could be an
> overlay ;-)

We probably shouldn't have put kingfisher as a top-level compatible
then. But we did, so not really much point to discuss that now.

As to whether there's a better way to express it in the schema, I'm
not sure. I don't think there's a way with json-schema to express a
list, but the 1st item is optional.

Rob