Re: [PATCH 12/36] dt-bindings: arm: Convert cpu binding to json-schema

[Rob Herring]

On Mon, Dec 3, 2018 at 6:40 AM Will Deacon <will.deacon@xxxxxxx> wrote:
>
> On Fri, Nov 30, 2018 at 12:00:05PM -0600, Rob Herring wrote:
> > On Thu, Nov 8, 2018 at 2:49 AM Michal Simek <michal.simek@xxxxxxxxxx> wrote:
> > >
> > > Hi Rob,
> > >
> > > On 05. 10. 18 18:58, Rob Herring wrote:
> > > > Convert ARM CPU binding to DT schema format using json-schema.
> > > >
> > > > Cc: Mark Rutland <mark.rutland@xxxxxxx>
> > > > Cc: Matthias Brugger <matthias.bgg@xxxxxxxxx>
> > > > Cc: devicetree@xxxxxxxxxxxxxxx
> > > > Cc: linux-arm-kernel@xxxxxxxxxxxxxxxxxxx
> > > > Cc: linux-mediatek@xxxxxxxxxxxxxxxxxxx
> > > > Signed-off-by: Rob Herring <robh@xxxxxxxxxx>
> > > > ---
> > > >  .../devicetree/bindings/arm/cpus.txt          | 490 -----------------
> > > >  .../devicetree/bindings/arm/cpus.yaml         | 503 ++++++++++++++++++
> > > >  2 files changed, 503 insertions(+), 490 deletions(-)
> > > >  delete mode 100644 Documentation/devicetree/bindings/arm/cpus.txt
> > > >  create mode 100644 Documentation/devicetree/bindings/arm/cpus.yaml
> >
> > [...]
> >
> > > I have take a look at xilinx part of this and try to build it for arm64
> > > platforms and I see errors coming from this cpu description.
> > > /root/linux/arch/arm64/boot/dts/xilinx/zynqmp-zcu100-revC.dt.yaml:
> > > cpu@0:compatible: ['arm,cortex-a53', 'arm,armv8'] is too long
> > > /root/linux/arch/arm64/boot/dts/xilinx/zynqmp-zcu100-revC.dt.yaml:
> > > cpu@0:compatible: Additional items are not allowed ('arm,armv8' was
> > > unexpected)
> >
> > Thanks for actually giving this a spin!
> >
> > > Based on grep this is used in a lot of places
> > > compatible = "arm,cortex-a53", "arm,armv8";
> > >
> > > Should this be moved to just simple?
> > > compatible = "arm,cortex-a53";
> >
> > I'd normally go with the majority which would be to keep it. However,
> > 'arm,armv8' is of questionable value, isn't actually documented, and
> > doesn't exist for any other version of the architecture. So we should
> > kill it IMO.
>
> I'd prefer to keep it around, since that's what's used to describe the CPUs
> on the fastmodel iirc.

We can and should keep it for that purpose, but do we need it as a
fallback? For real cores though, we have mixture of with and without
and one of those need to be fixed.

Rob