On Sat, Nov 25, 2023 at 01:25:22PM +0100, Luca Weiss wrote:
> On Montag, 16. Oktober 2023 16:44:28 CET Stephan Gerhold wrote:
> > On Mon, Oct 16, 2023 at 03:16:14PM +0200, Loic Poulain wrote:
> > > On Mon, 16 Oct 2023 at 07:35, Krzysztof Kozlowski
> > >
> > > <[email protected]> wrote:
> > > > On 15/10/2023 22:03, Luca Weiss wrote:
> > > > > Add a compatible for the iris subnode in the WCNSS PIL.
> > > > >
> > > > > Signed-off-by: Luca Weiss <[email protected]>
> > > > > ---
> > > > >
> > > > > Documentation/devicetree/bindings/remoteproc/qcom,wcnss-pil.yaml | 1
> > > > > +
> > > > > 1 file changed, 1 insertion(+)
> > > > >
> > > > > diff --git
> > > > > a/Documentation/devicetree/bindings/remoteproc/qcom,wcnss-pil.yaml
> > > > > b/Documentation/devicetree/bindings/remoteproc/qcom,wcnss-pil.yaml
> > > > > index 45eb42bd3c2c..0e5e0b7a0610 100644
> > > > > --- a/Documentation/devicetree/bindings/remoteproc/qcom,wcnss-pil.yaml
> > > > > +++ b/Documentation/devicetree/bindings/remoteproc/qcom,wcnss-pil.yaml
> > > > >
> > > > > @@ -111,6 +111,7 @@ properties:
> > > > > - qcom,wcn3660
> > > > > - qcom,wcn3660b
> > > > > - qcom,wcn3680
> > > > >
> > > > > + - qcom,wcn3680b
> > > >
> > > > Looks like this should be made as compatible with qcom,wcn3680 (so with
> > > > fallback).
> > >
> > > Yes, agree, let's do a regular fallback as there is nothing 'b'
> > > specific in the driver:
> > > `compatible = "qcom,wcn3680b", "qcom,wcn3680";`
> > >
> > > And yes, we should also have done that for qcom,wcn3660b...
> >
> > I don't think this would have worked properly for qcom,wcn3660b:
> >
> > - It's not compatible with "qcom,wcn3660", because they have different
> > regulator voltage requirements. wcn3660(a?) needs vddpa with
> > 2.9-3.0V, but wcn3660b needs 3.3V. That's why wcn3660b uses the
> > wcn3680_data in qcom_wcnss.iris.c. Otherwise if you would run an
> > older kernel that knows "qcom,wcn3660" but not "qcom,wcn3660b" it
> > would apply the wrong voltage.
> >
> > - It's not compatible with "qcom,wcn3680" either because that is used
> > as indication if 802.11ac is supported (wcn3660b doesn't).
> >
> > The main question here is: What does the current "qcom,wcn3680"
> > compatible actually represent? It's defined with vddpa = 3.3V in the
> > driver, which would suggest that:
> >
> > 1. It's actually meant to represent WCN3680B, which needs 3.3V vddpa
> > like WCN3660B, or
> >
> > 2. WCN3680(A?) has different requirements than WCN3660(A?) and also
> > needs 3.3V vddpa. But then what is the difference between
> > WCN3680(A?) and WCN3680B? Is there even a variant without ...B?
> >
> > There is public documentation for WCN3660B and WCN3680B but the non-B
> > variants are shrouded in mystery.
>
> Hi Stephan (and everyone),
>
> Do you have a suggestion how to move this patchset forward? Is the fallback
> compatible that was suggested okay for the wcn3680b situation?
>
> compatible = "qcom,wcn3680b", "qcom,wcn3680";
>
> If so, I'll make v2 with that implemented.
>
Personally, I would just go with what exists already and use the
existing "qcom,wcn3680" compatible as-is. No one has provided evidence
that there is actually something like a WCN3680(A), without a 'B' at the
end. Also, all existing users of the "qcom,wcn3680" compatible upstream
actually seem to refer to WCN3680B (I'm quite sure apq8039-t2 has
WCN3680B, MSM8974 isn't entirely clear but iirc you mentioned there was
some schematic of a MSM8974 device which says WCN3680B as well).
It would be nice to have the compatible perfectly correct and complete,
but I see no point to extend it with additional information that we
can't properly verify.
Or do you actually need a different compatible to customize something in
the driver?
Thanks,
Stephan