Re: [PATCH v5 01/14] dt-bindings: pci: cadence: Extend compatible for new RP configuration

[Krzysztof Kozlowski <krzk@kernel.org>]

On 01/07/2025 13:56, Manikandan Karunakaran Pillai wrote:
> 
>>>> On 2025/6/30 15:30, Krzysztof Kozlowski wrote:
>>>>> EXTERNAL EMAIL
>>>>>
>>>>> On Mon, Jun 30, 2025 at 12:15:48PM +0800, hans.zhang@cixtech.com
>> wrote:
>>>>>> From: Manikandan K Pillai <mpillai@cadence.com>
>>>>>>
>>>>>> Document the compatible property for HPA (High Performance
>>>> Architecture)
>>>>>> PCIe controller RP configuration.
>>>>>
>>>>> I don't see Conor's comment addressed:
>>>>>
>>>>> https://urldefense.com/v3/__https://lore.kernel.org/linux-
>>>> devicetree/20250424-elm-magma-
>>>> b791798477ab@spud/__;!!EHscmS1ygiU1lA!Bo-
>>>>
>> ayMVqCWXSbSgFpsBZzgk1ADft8pqRQbuOeAhIuAjz0zI015s4dmzxgaWKycqKMn
>>>> 1cejS8kKZvjF5xDAse$
>>>>>
>>>>> You cannot just send someone's work and bypassing the review feedback.
>>>
>>> I thought the comment was implicitly addressed when the device drivers
>> were separated out based on other review comments in this patch.
>>> To make it more clear, in the next patch I will add the following description
>> for the dt-binding patch
>>>
>>> "The High performance architecture is different from legacy architecture
>> controller in design of register banks,
>>> register definitions, hardware sequences of initialization and is considered as
>> a different device due to the
>>> large number of changes required in the device driver and hence adding a
>> new compatible."
>> That's still vague. Anyway this does not address other concern that the
>> generic compatible is discouraged and we expect specific compatibles. We
>> already said that and what? You send the same patch.
>>
>> So no, don't send the same patch.
> 
> 
> Hi Kryzsztof,
> 
> Are you suggesting to create new file for both RC and EP for HPA host like:
> cdns,cdns-pcie-hpa-host.yaml
> cdns,cdns-pcie-hpa-ep.yaml
> And during the commit log, explain why you need to create a new file for HPA, and not use the legacy one.

No, there was no such suggestions in any previous or current
discussions. IIRC, this was simply rejected previously. I consider this
rejected still, with the same arguments: you should use specific SoC
compatibles. The generic compatible alone is rather legacy approach and
we have been commenting on this sooooo many times.



Best regards,
Krzysztof