On 09/04/2022 15:18, H. Nikolaus Schaller wrote:
hould have a specific compatible.
>>> Please mention why it does not.
>>
>> Agreed. The "snps,dwc2" should be a fallback string, otherwise there is no way to uniquely identify the JZ4780 implementation of the IP.
>
> Well, there is no specifc implementation and driver for it. So no need to uniquely identify it.
Specific implementation and driver are not arguments here. This does not
matter. It's really unrelated argument.
Bindings are not about implementation in Linux. Implementation can
change, so bindings should also?
Best regards,
Krzysztof
> Am 09.04.2022 um 15:23 schrieb Krzysztof Kozlowski <[email protected]>:
>
> On 09/04/2022 15:18, H. Nikolaus Schaller wrote:
> hould have a specific compatible.
>>>> Please mention why it does not.
>>>
>>> Agreed. The "snps,dwc2" should be a fallback string, otherwise there is no way to uniquely identify the JZ4780 implementation of the IP.
>>
>> Well, there is no specifc implementation and driver for it. So no need to uniquely identify it.
>
> Specific implementation and driver are not arguments here. This does not
> matter. It's really unrelated argument.
The argumentation is in reverse: if there is no need for a specialized driver or implementation,
why is there is a need to define a specialization.
Your argument was:
"there is no way to uniquely identify the JZ4780 implementation of the IP"
My question is:
"what do we need that for?"
> Bindings are not about implementation in Linux. Implementation can
> change, so bindings should also?
No. Implementations should be agnostic.
Le sam., avril 9 2022 at 15:37:51 +0200, H. Nikolaus Schaller
<[email protected]> a ?crit :
>
>
>> Am 09.04.2022 um 15:23 schrieb Krzysztof Kozlowski
>> <[email protected]>:
>>
>> On 09/04/2022 15:18, H. Nikolaus Schaller wrote:
>> hould have a specific compatible.
>>>>> Please mention why it does not.
>>>>
>>>> Agreed. The "snps,dwc2" should be a fallback string, otherwise
>>>> there is no way to uniquely identify the JZ4780 implementation of
>>>> the IP.
>>>
>>> Well, there is no specifc implementation and driver for it. So no
>>> need to uniquely identify it.
>>
>> Specific implementation and driver are not arguments here. This
>> does not
>> matter. It's really unrelated argument.
>
> The argumentation is in reverse: if there is no need for a
> specialized driver or implementation,
> why is there is a need to define a specialization.
>
> Your argument was:
> "there is no way to uniquely identify the JZ4780 implementation of
> the IP"
>
> My question is:
> "what do we need that for?"
You may not need the differenciation now, but if you need it in the
future and you had only the "snps,dwc2" compatible previously, then
you're screwed, since your driver must support older device tree blobs.
-Paul
>> Bindings are not about implementation in Linux. Implementation can
>> change, so bindings should also?
>
> No. Implementations should be agnostic.
>