Re: [PATCH 1/2] dt-bindings: arm64: dts: mediatek: Add mt8395-evk board

[Macpaul Lin <macpaul.lin@mediatek.com>]

On 9/5/23 18:58, AngeloGioacchino Del Regno and Krzsztof Kozlowski wrote:
> Il 05/09/23 12:36, Krzysztof Kozlowski ha scritto:
>> On 05/09/2023 11:36, Macpaul Lin wrote:
>>>
>>>
>>> On 9/4/23 20:11, Krzysztof Kozlowski wrote:
>>>>
>>>>
>>>> External email : Please do not click links or open attachments until 
>>>> you
>>>> have verified the sender or the content.
>>>>
>>>> On 04/09/2023 11:50, Macpaul Lin wrote:
>>>>>
>>>>>
>>>>> On 9/4/23 17:33, Krzysztof Kozlowski wrote:
>>>>>>
>>>>>>
>>>>>> External email : Please do not click links or open attachments 
>>>>>> until you
>>>>>> have verified the sender or the content.
>>>>>>
>>>>>> On 04/09/2023 11:20, Macpaul Lin wrote:
>>>>>>> Add bindings for the MediaTek mt8395-evk board.
>>>>>>> The mt8359-evk board is also named as "Genio 1200-EVK".
>>>>>>> MT8195 and MT8395 are the same family series SoC could share
>>>>>>
>>>>>> How can be the same and have different numbers? You sill need 
>>>>>> dedicated
>>>>>> compatible.
>>>>>>
>>>>>
>>>>> The SoCs mt8195 and mt8395 are designed for different market 
>>>>> application
>>>>> and physical characteristics, using different efuse values for
>>>>> distinction. The booting flow and configurations are controllered 
>>>>> by the
>>>>> boot loaders, firmware, and TF-A. Therefore, the part numbers and
>>>>> procurement channels are different. The detail information of these
>>>>> efuse values is proprietary, so I cant disclose it futher. Hence the
>>>>> most of peripheral drivers and base address are almost the same.
>>>>
>>>> 1. Drivers? So we talk about compatibility, not the same.
>>>> 2. "almost the same" is not the same. Follow the guidelines for writing
>>>> bindings.
>>>>
>>>
>>> Thanks for the review.
>>>
>>> After internal confirmation and discussion, it can be confirmed that the
>>> MT8195 and MT8395 are identical SoCs from to binding's perspective.
>>
>> I am sorry, but I really do not care what you internally discussed about
>> bindings. I do not think your internal review respect existing
>> guidelines. You talked about drivers, not "bindings perspective", so
>> your internal discussion is clearly discussing something else.
>>
>>> MediaTek hope the mt8395 boards could directly use mt8195.dtsi, without
>>> the need to create a separate mt8395.dtsi to include mt8195.dtsi.
>>> Therefore, we hope to fully adopt the bindings of mt8195. However, I
>>> will submit a revised patch for compatible since they are different 
>>> boards.
>>
>> You can disagree but then I expect arguments from your side.
>>
> 
> In short - they're the same chip, as in, they behave the same on a 
> *hardware*
> perspective; what changes is the bootchain (plus stricter security from 
> TF-A)
> and allowable temperature ranges for operation, that's practically it...
> 
> ...so yes the compatible for the "new soc" must be documented, but that's
> practically just a revision, *not a new soc* at all.
> 
> (though, I agree that seeing a different name as in 1 -> 3 can be totally
> confusing)
> 
> The drivers difference that Macpaul hinted to are about drivers needing 
> some
> SMC calls instead of direct MMIO manipulation, so, something like two 
> bindings
> for something like two drivers will need to add a 8395 compatible; 
> speaking of
> what we would have in a devicetree for this SoC, that'd be exactly 99% 
> identical
> to mt8195.dtsi.
> 
> Anyway, drivers are drivers, bindings describe hardware - and the hw is, 
> again,
> the same...
> 
> Hope that this makes things clearer! :-)
> 
> Cheers,
> Angelo

Thanks for your patience and clarification.
I'll submit new SOC binding and revise the patches for mt8395 board.

Thanks
Macpaul Lin