Re: [PATCH 1/2] dt-bindings: clock: Add reg-clock-controller

[Yangfl]

Rob Herring <robh@xxxxxxxxxx> 于2023年4月18日周二 04:49写道：
>
> On Sat, Apr 15, 2023 at 02:12:59AM +0800, David Yang wrote:
> > Add DT bindings documentation for reg-clock-controller, collection of
> > basic clocks common to many platforms.
> >
> > Signed-off-by: David Yang <mmyangfl@xxxxxxxxx>
> > ---
> >  .../bindings/clock/reg-clock-controller.yaml  | 245 ++++++++++++++++++
> >  1 file changed, 245 insertions(+)
> >  create mode 100644 Documentation/devicetree/bindings/clock/reg-clock-controller.yaml
> >
> > diff --git a/Documentation/devicetree/bindings/clock/reg-clock-controller.yaml b/Documentation/devicetree/bindings/clock/reg-clock-controller.yaml
> > new file mode 100644
> > index 000000000000..a6a7e0b05821
> > --- /dev/null
> > +++ b/Documentation/devicetree/bindings/clock/reg-clock-controller.yaml
> > @@ -0,0 +1,245 @@
> > +# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
> > +%YAML 1.2
> > +---
> > +$id: http://devicetree.org/schemas/clock/reg-clock-controller.yaml#
> > +$schema: http://devicetree.org/meta-schemas/core.yaml#
> > +
> > +title: Simple straight-forward register-based clocks
>
> 'simple' means extended one property at a time with little foresight. In
> the end, 'simple' bindings are never simple. s/simple/generic/ as well.
>
> When we first started the clock binding, we had exactly this. There's
> still bindings for at least some of it. It turned out to be a bad
> idea because it was difficult to get correct and complete. So this
> binding is exactly what we don't want.
>
> Rob

Thanks. So in the next version v2, clock-controller are made merely
container of clocks. The real jobs are done by sub clock nodes, so
that clocks, with or without hardware-specific clock types
"compatible", can be declared independently and individually as dt
nodes, instead of hardware clock "controllers" with quite arbitrary
clock identifiers listed under include/dt-bindings/clock/.