Re: [PATCH 4/8] dt-bindings: Add Tegra264 clock and reset definitions

[Krzysztof Kozlowski <krzk@kernel.org>]

On 08/05/2025 09:59, Thierry Reding wrote:
> On Thu, May 08, 2025 at 09:49:12AM +0200, Krzysztof Kozlowski wrote:
>> On 08/05/2025 09:46, Thierry Reding wrote:
>>> On Thu, May 08, 2025 at 09:40:38AM +0200, Krzysztof Kozlowski wrote:
>>>> On 08/05/2025 09:39, Krzysztof Kozlowski wrote:
>>>>> On 07/05/2025 16:37, Thierry Reding wrote:
>>>>>> From: Thierry Reding <treding@nvidia.com>
>>>>>>
>>>>>> Signed-off-by: Thierry Reding <treding@nvidia.com>
>>>>>
>>>>> Missing commit msg
>>>>>
>>>>>> ---
>>>>>>  include/dt-bindings/clock/tegra264-clock.h | 9 +++++++++
>>>>>>  include/dt-bindings/reset/tegra264-reset.h | 7 +++++++
>>>>>>  2 files changed, 16 insertions(+)
>>>>>>  create mode 100644 include/dt-bindings/clock/tegra264-clock.h
>>>>>>  create mode 100644 include/dt-bindings/reset/tegra264-reset.h
>>>>>
>>>>>
>>>>> Filename equal to the compatible. That's the standard convention for all
>>>>> the headers since some years.
>>>>
>>>> Huh, I cannot find the binding in this patchset. Where is the actual
>>>> binding added?
>>>
>>> The bindings for this are in
>>>
>>>   Documentation/devicetree/bindings/firmware/nvidia,tegra186-bpmp.yaml
>>
>> There is no tegra264 in that binding.
> 
> That's part of an earlier series I sent out (and linked to from the
> cover letter). It's here:
> 
> 	https://lore.kernel.org/linux-tegra/20250506133118.1011777-1-thierry.reding@gmail.com/T/#m96bb396b352659581a9e71a4610c51e6ab4d5b6a


Then this patch belongs there. Standard rules apply: binding headers go
with the binding itself and the binding itself go with driver patch via
driver subsystem tree. At least usually. Nothing here is different than
all other vendors who follow such convention.

> 
>> The header always goes with the binding and the drivers.
>>
>>>
>>> There's no 1:1 mapping to a compatible for this because BPMP is many
>>> things. It's a clock provider, a reset provider, a power domain
>>
>> Sure, that's fine.
>>
>>> provider. These definitions reflect the IDs assigned by the BPMP ABI
>>> and we've used this structure ever since this was introduced back in
>>> 2016.
>>>
>>> I don't think changing the convention for this is a net advantage.
>>
>> Headers still should match the compatible one way or another. Can be
>> nvidia,tegra264.h
>> (because -clock is redundant and you do not want to use the actual
>> compatible)
> 
> I get it. You want consistency. But what about consistency with earlier
> chip generations?


I will fix them after finishing my time machine. :)

> Do you want me to go and rename all of these files?


No, I don't want to change them, but I would be fine if someone does the
change (although someone else might claim this is a churn). That ship
has sailed, but at least we can start with new bindings.

Best regards,
Krzysztof