Re: [PATCH v1 5/7] arm: dts: qcom: mdm9615: remove invalid pmic subnodes compatibles

[Neil Armstrong <neil.armstrong@linaro.org>]

On 29/09/2022 11:12, Krzysztof Kozlowski wrote:
> On 29/09/2022 10:29, Neil Armstrong wrote:
>> Hi,
>>
>> On 28/09/2022 20:03, Krzysztof Kozlowski wrote:
>>> On 28/09/2022 11:14, Neil Armstrong wrote:
>>>> The PMIC is an PM8018, but was compatible with the PM8921. Both compatibles
>>>> was left but it makes no sense anymore the leave both.
>>>
>>> Why? It makes sense for backwards compatibility. If you think it does
>>> not make sense, please say why.
>>
>> We had the same debate at submission 7y ago, some of the pm8018 new compatible
>> were rejected in bindings & drivers so I left both...
>>
>> As of today only the pwrkey bindings is missing, so should I resubmit the pm8018-pwrkey bidings and
>> drop the pm8921-pwrkey compatible ?
> 
> ~7 years ago here:
> https://lore.kernel.org/all/20160624220748.GB11719@dtor-ws/
> you proposed to add something entirely different than we have here now
> and than we talk about.
> 
> In that thread you correctly wrote:
> "My point of view is that the devicetree describes the hardware and need
> to have SoC specific compatible string since it describes the actual
> silicon, and drivers must make sure to handle all the SoC or family
> variants using the compatible string and the match data."

And I'm happy this is still the policy! And I'm tried my best to follow this
in all my DT & bindings submissions, while DT-Schema helped a lot here.

> 
> but implemented it entirely different. Maybe you refer to different mail
> thread, I don't know, but that one is indeed wrong.

In the meantime things got much better, but at that time pushing a SoC bringup
was a pain (I did 2 at the time, the other one is the OX810SE) and I even
mentioned it in a talk ([1] slides 27 to 30).

So I added both to be sure that at some point a driver would probe against
one of the compatible entries...

> 
> The DTS looks correct unless you have some real argument that it is not.
> 
> How this should be fixed? First, drop bogus entries from drivers, then
> document proper compatibles.

What do you mean ? There's no point to keep the PM8921 compatibles, the gpio
and PMIC bindings already enforces to only have the PM8018 compatible.

The only issue is about the PM8018 pwrkey, where the solution would be
to actually re-submit [1] by documenting qcom,pm8018-pwrkey and adding the entry
in the drivers/input/misc/pmic8xxx-pwrkey.c driver.

Or maybe I missed something.

[1] https://www.slideshare.net/superna/elce-2016-neil-armstrong-no-its-never-too-late-to-upstream-your-legacy-linux-based-platform
[2] https://lore.kernel.org/all/1466759887-25394-3-git-send-email-narmstrong@baylibre.com/

> 
> Best regards,
> Krzysztof
> 

Thanks,
Neil