Re: [PATCH v5 04/15] dt-bindings: spi: dw: Add AMD Pensando Elba SoC SPI Controller bindings

[Serge Semin]

On Tue, Jun 21, 2022 at 09:00:36AM +0200, Krzysztof Kozlowski wrote:
> On 20/06/2022 22:04, Serge Semin wrote:
> > On Mon, Jun 20, 2022 at 09:46:25PM +0200, Krzysztof Kozlowski wrote:
> >> On 20/06/2022 21:30, Serge Semin wrote:
> >>> On Mon, Jun 13, 2022 at 12:56:47PM -0700, Brad Larson wrote:
> >>>> From: Brad Larson <blarson@xxxxxxx>
> >>>>
> >>>> The AMD Pensando Elba SoC has integrated the DW APB SPI Controller
> >>>>
> >>>> Signed-off-by: Brad Larson <blarson@xxxxxxx>
> >>>> ---
> >>>>  Documentation/devicetree/bindings/spi/snps,dw-apb-ssi.yaml | 2 ++
> >>>>  1 file changed, 2 insertions(+)
> >>>>
> >>>> diff --git a/Documentation/devicetree/bindings/spi/snps,dw-apb-ssi.yaml b/Documentation/devicetree/bindings/spi/snps,dw-apb-ssi.yaml
> >>>> index e25d44c218f2..2a55b947cffc 100644
> >>>> --- a/Documentation/devicetree/bindings/spi/snps,dw-apb-ssi.yaml
> >>>> +++ b/Documentation/devicetree/bindings/spi/snps,dw-apb-ssi.yaml
> >>>> @@ -73,6 +73,8 @@ properties:
> >>>>                - renesas,r9a06g032-spi # RZ/N1D
> >>>>                - renesas,r9a06g033-spi # RZ/N1S
> >>>>            - const: renesas,rzn1-spi   # RZ/N1
> >>>
> >>>> +      - description: AMD Pensando Elba SoC SPI Controller
> >>>> +        const: amd,pensando-elba-spi
> >>>
> >>> Not enough. The driver requires to have a phandle reference to the
> >>> Pensando System Controller. So the property like
> >>> "amd,pensando-elba-syscon" is also needed to be added to the DT schema
> >>> otherwise should the dt-schema tool correctly handle the
> >>> "unevaluatedProperties: false" setting (Rob says it isn't fully
> >>> supported at the moment), the dtbs_check procedure will fail on your
> >>> dts evaluation.
> >>
> > 
> >> The property was here before, now removed, so I assume it was also
> >> removed from the driver and DTS. Isn't that the case?
> > 
> > Ah, the property has been indeed removed. The driver now searches for
> > the system controller by the next compatible string:
> > "amd,pensando-elba-syscon" using the
> > syscon_regmap_lookup_by_compatible() method. My mistake. Sorry for the
> > noise.
> > 

> > * Though personally I'd prefer to have a property with the phandle
> > reference in order to signify the connection between the system controller
> > and the SPI-controller. Otherwise the implicit DT bindings like having
> > the "amd,pensando-elba-syscon"-compatible syscon gets to be
> > hidden behind the DT scene. But seeing we have already got the Microsemi
> > platform with such semantic, I can't insist on fixing this.
> 
> I agree entirely, this should be explicit syscon-type property. Looking
> up for compatibles:
>  - creates hidden (not expressed via bindings) dependency between nodes,
>  - is not portable and several people struggled with it later and needed
> backward-compatible code (many examples, let's just give recent one: [1])
> 
> 
> [1]
> https://lore.kernel.org/all/20220619151225.209029-10-tmaimon77@xxxxxxxxx/

Seems even more reasonable now. Thanks for providing a bright example
justifying the property-based approach.

@Brad, could you get back the property with a phandle to the syscon
DT-node? (No need in adding the CS CSR address as the phandle argument,
just a phandle.)

-Sergey

> 
> 
> Best regards,
> Krzysztof