Re: [PATCH 2/3] dt-bindings: usb: atmel: Add Microchip LAN966x compatible string

[Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>]

On 20/05/2022 15:52, Alexandre Belloni wrote:
> Hello,
> 
> On 20/05/2022 15:38:36+0200, Krzysztof Kozlowski wrote:
>> On 20/05/2022 15:02, Herve Codina wrote:
>>> On Fri, 20 May 2022 14:50:24 +0200
>>> Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org> wrote:
>>>
>>>> On 20/05/2022 14:21, Herve Codina wrote:
>>>>>>> I think it makes sense to keep 'microchip,lan966x-udc' for the USB
>>>>>>> device controller (same controller on LAN9662 and LAN9668) and so
>>>>>>> keeping the same rules as for other common parts.    
>>>>>>
>>>>>> Having wildcard was rather a mistake and we already started correcting
>>>>>> it, so keeping the "mistake" neither gives you consistency, nor
>>>>>> correctness...
>>>>>>  
>>>>>
>>>>> I think that the "family" compatible should be present.
>>>>> This one allows to define the common parts in the common
>>>>> .dtsi file (lan966x.dtsi in our case).
>>>>>
>>>>> What do you think about:
>>>>> - microchip,lan9662-udc
>>>>> - microchip,lan9668-udc
>>>>> - microchip,lan966-udc  <-- Family
>>>>>
>>>>> lan966 is defined as the family compatible string since (1) in
>>>>> bindings/arm/atmel-at91.yaml and in Documentation/arm/microchip.rst
>>>>>   
>>>>
>>>> You can add some family compatible, if it makes sense. I don't get why
>>>> do you mention it - we did not discuss family names, but using
>>>> wildcards... Just please do not add wildcards.
>>>
>>> Well, I mentioned it as I will only use the family compatible string
>>> and not the SOC (lan9662 or lan9668) compatible string in lan966x.dtsi.
>>> In this case, the family compatible string can be seen as a kind of
>>> "wildcard".
>>
>> I understood as "the "family" compatible should be present" as you want
>> to add it as a fallback. It would be okay (assuming devices indeed share
>> family design). If you want to use it as the only one, then it is again
>> not a recommended approach. Please use specific compatibles.
>>
>> I mean, why do we have this discussion? What is the benefit for you to
>> implement something not-recommended by Devicetree spec and style?
>>
> 
> Honestly, I would just go for microchip,lan9662-udc. There is no
> difference between lan9662 and lan9668 apart from the number of switch
> ports.

Thank you, and maybe that was misunderstanding - I do not propose to add
additional lan9668 compatible, if it is not actually needed.


Best regards,
Krzysztof