Re: [PATCH v3 4/5] dt-bindings: cpufreq: cpufreq-qcom-hw: Add SDX75 compatible

[Pavan Kondeti]

On Fri, Jun 09, 2023 at 11:17:08AM +0200, Konrad Dybcio wrote:
> 
> 
> On 9.06.2023 07:00, Pavan Kondeti wrote:
> > On Wed, Jun 07, 2023 at 05:17:48PM +0530, Rohit Agarwal wrote:
> >> Add compatible for EPSS CPUFREQ-HW on SDX75.
> >>
> >> Signed-off-by: Rohit Agarwal <quic_rohiagar@xxxxxxxxxxx>
> >> Acked-by: Viresh Kumar <viresh.kumar@xxxxxxxxxx>
> >> ---
> >>  Documentation/devicetree/bindings/cpufreq/cpufreq-qcom-hw.yaml | 1 +
> >>  1 file changed, 1 insertion(+)
> >>
> >> diff --git a/Documentation/devicetree/bindings/cpufreq/cpufreq-qcom-hw.yaml b/Documentation/devicetree/bindings/cpufreq/cpufreq-qcom-hw.yaml
> >> index a6b3bb8..866ed2d 100644
> >> --- a/Documentation/devicetree/bindings/cpufreq/cpufreq-qcom-hw.yaml
> >> +++ b/Documentation/devicetree/bindings/cpufreq/cpufreq-qcom-hw.yaml
> >> @@ -36,6 +36,7 @@ properties:
> >>                - qcom,sa8775p-cpufreq-epss
> >>                - qcom,sc7280-cpufreq-epss
> >>                - qcom,sc8280xp-cpufreq-epss
> >> +              - qcom,sdx75-cpufreq-epss
> >>                - qcom,sm6375-cpufreq-epss
> >>                - qcom,sm8250-cpufreq-epss
> >>                - qcom,sm8350-cpufreq-epss
> > 
> > This is a very basic question, not completely related to this patch.
> > Apologies in advance.
> > 
> > What is the rationale for adding a new soc string under compatible and
> > using it in the new soc device tree? Is it meant for documentation purpose?
> > i.e one know what all SoCs / boards supported by this device node.
> It's two-fold:
> 
> 1. The device tree describes the hardware, and for lack of better terms (e.g.
>    an SoC-specific version number of the block that is identical to all other
>    implementations of that revision on all SoCs that use it), we tend to
>    associate it with the SoC it's been (first) found on.
> 
> 2. In case we ever needed to introduce a SoC-specific quirk, we can just add
>    an of_is_compatible-sorta check to the driver and not have to update the
>    device trees. This is very important for keeping backwards compatibility,
>    as it's assumed that not everybody may be running the latest one. This
>    means we have to avoid ABI breaks (unless we have *very* good reasons, like
>    "this would have never worked anyway" or "it was not described properly
>    and worked on this occasion by pure luck")
> 

Thanks Konrad for the explanation. The #2 is a clear winner here. It
makes complete sense. In devices like USB, we have PID/VID through which
quirks can be implemented later. So I guess the same analogy applies here.
Like you said in (1), the devices are identified with SoC compatible
string.

Thanks,
Pavan