Re: [PATCH v2 3/3] dt-bindings: arm: msm: Convert kpss-gcc driver Documentation to yaml

[Bjorn Andersson <bjorn.andersson@linaro.org>]

On Sat 30 Apr 02:42 CDT 2022, Ansuel Smith wrote:

> On Sat, Apr 30, 2022 at 04:40:54PM +0200, Krzysztof Kozlowski wrote:
> > On 30/04/2022 08:01, Ansuel Smith wrote:
> > > Convert kpss-gcc driver Documentation to yaml.
> > > Add #clock-cells additional binding to required bindings and example
> > > as it's a required binding for clock-output-names.
> > > 
> > > Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
> > 
> > 
> > (...)
> > 
> > > +properties:
> > > +  compatible:
> > > +    items:
> > > +      - enum:
> > > +          - qcom,kpss-gcc-ipq8064
> > > +          - qcom,kpss-gcc-apq8064
> > > +          - qcom,kpss-gcc-msm8974
> > > +          - qcom,kpss-gcc-msm8960
> > > +      - const: qcom,kpss-gcc
> > > +
> > > +  reg:
> > > +    maxItems: 1
> > > +
> > > +  clocks:
> > > +    items:
> > > +      - description: phandle to pll8_vote
> > > +      - description: phandle to pxo_board
> > > +
> > > +  clock-names:
> > > +    items:
> > > +      - const: pll8_vote
> > > +      - const: pxo
> > > +
> > > +  clock-output-names:
> > > +    const: acpu_l2_aux
> > 
> > It does not make sense having a constant output name. What is the
> > meaning this property in such case? The original binding did not enforce it.
> > 
> > 
> > 
> > Best regards,
> > Krzysztof
> 
> Mh. Should I just drop the const and put a description referring to an
> advised name? The driver with the kpss-gcc hardcode the name to
> acpu_l2_aux that's why I thought it was a correct conversion using a
> const but I assume this is another problem of not making a correct 1:1
> conversion and adding fixes on pure conversion.
> Think I should drop it and put a description to it. (and then later fix
> it when I will push the other series with all the tweaks)
> 
> What do you think?
> 

The typical reason for using clock-output-names is that we have some
consumer that finds the clock based on global name lookup. Over time
we've been moving these to use .fw_name or .index based lookup, which
removes this problem.

But I don't see that being the case here. So my suggestion is that you
just drop clock-output-names from the binding, which will solve
Krzysztof's objection.


From there we can review what needs to be done in the Linux driver to
work with the improved binding.

Regards,
Bjorn