Re: [PATCH 2/9] dt-bindings: phy: qcom,m31: Document qcom,m31 USB phy

[Varadarajan Narayanan]

On Sat, Jun 17, 2023 at 10:48:41AM +0200, Krzysztof Kozlowski wrote:
> On 15/06/2023 07:27, Varadarajan Narayanan wrote:
> >>> +          - enum:
> >>> +              - qcom,m31-usb-hsphy
> >>
> >> I am confused what's this. If m31 is coming from some IP block provider,
> >> then you are using wrong vendor prefix.
> >> https://www.m31tech.com/download_file/M31_USB.pdf
> >>
> >>
> >>> +              - qcom,ipq5332-m31-usb-hsphy
> >>
> >> This confuses me even more. IPQ m31?
> >
> > Will change this to m31,usb-hsphy and m31,ipq5332-usb-hsphy respectively.
> > Will that be acceptable?
>
> m31,ipq5332 seems wrong, as m31 did not create ipq5332. Does the m31
> device have some name/version/model? If it is not really known, then I
> would just propose to go with qcom,ipq5332-usb-hsphy.
>
> Skip generic compatible ("usb-hsphy") entirely.

Ok.

> And then we have... existing bindings qcom,usb-hs-phy.yaml. Don't create
> something similar with difference in the hyphen. Just use device
> specific compatible thus device specific filename.

qcom,usb-hs-phy.yaml seems to be for ULPI mode phy and the
driver we are introducing is for UTMI. We would have to
modify phy-qcom-usb-hs.c to accomodate M31. Will that be
acceptable to phy-qcom-usb-hs.c owners/maintainers?

> >>> +
> >>> +  reg:
> >>> +    description:
> >>> +      Offset and length of the M31 PHY register set
> >>
> >> Drop description, obvious.
> >
> > Ok.
> >
> >>> +    maxItems: 2
> >>> +
> >>> +  reg-names:
> >>> +    items:
> >>> +      - const: m31usb_phy_base
> >>> +      - const: qscratch_base
> >>
> >> Drop "_base" from both.
> >
> > Ok. Will drop qscratch_base. This is in the controller space.
> > Should not come here.
>
> Then drop reg-names entirely.

Ok.

> >>> +
> >>> +  phy_type:
> >>> +    oneOf:
> >>> +      - items:
> >>> +          - enum:
> >>> +              - utmi
> >>> +              - ulpi
> >>
> >> This does not belong to phy, but to USB node.
> >
> > This is used by the driver to set a bit during phy init. Hence
> > have it as a replication of the USB node's entry. If this is not
> > permissible, is there some way to get this from the USB node,
> > or any other alternative mechanism?
>
> Shouldn't USB controller choose what type of PHY type it wants?

Will remove this. IPQ5332 uses it in UTMI mode only.

> >>> +
> >>> +additionalProperties: false
> >>> +
> >>> +examples:
> >>> +  - |
> >>> +    #include <dt-bindings/clock/qcom,ipq5332-gcc.h>
> >>> +    hs_m31phy_0: hs_m31phy@5b00 {
> >>
> >> Node names should be generic. See also explanation and list of examples
> >> in DT specification:
> >> https://devicetree-specification.readthedocs.io/en/latest/chapter2-devicetree-basics.html#generic-names-recommendation
> >>
> >> Also, no underscores in node names.
> >
> > Will change this as usbphy0:hs_m31phy@7b000
>
> This does not solve my comments. I did not write "label" but "node name".

Sorry. will fix it.

Thanks
Varada