On 06/12/2023 13:31, Krzysztof Kozlowski wrote:
> On 06/12/2023 12:38, Varadarajan Narayanan wrote:
>> On Tue, Nov 28, 2023 at 03:01:12PM +0100, Krzysztof Kozlowski wrote:
>>> On 28/11/2023 11:14, Varadarajan Narayanan wrote:
>>>> On Tue, Nov 28, 2023 at 09:51:50AM +0100, Krzysztof Kozlowski wrote:
>>>>> On 28/11/2023 09:46, Varadarajan Narayanan wrote:
>>>>>> From: Varadarajan Narayanan <quic_varada@xxxxxxxxxxx>
>>>>>>
>>>>>> The earlier patch ec4f047679d5, incorrectly used 'l2'
>>>>>> as the vdda-pll-supply. However, 'l5' is the correct
>>>>>> ldo that supplies power to the USB PHY.
>>>>>>
>>>>>> Fixes: ec4f047679d5 ("arm64: dts: qcom: ipq9574: Enable USB")
>>>>>
>>>>> Doesn't this depend on the driver change?
>>>>
>>>> Yes, will mention in the cover letter.
>>>
>>> This commit should have it in its changelog ---
>>>
>>>>
>>>>> It affects both existing
>>>>> kernel and backports which you claim here should happen.
>>>>
>>>> Ok. Will include stable@xxxxxxxxxxxxxxx in the next revision.
>>>
>>> I wasn't speaking about Cc. You indicated this should be backported.
>>> Then please backport it, without previous commit, and check the result.
>>> Is stable tree working correctly or not?
>>
>> Without the previous commit, it would fail in both the latest
>> and stable tree. (Please see below for the error messages and
>> stack dump)
>>
>> The previous commit is necessary for this commit to work.
>
> Yep, exactly. It's visible from the patches. I don't know how to solve
> this exactly. The Fixes tag here is logically correct, but then any
> backporting must include previous commit. Dependency can be provided in
> cc-stable tag, but you did not cc-stable, I suppose on purpose.
>
> If this is chosen by AUTOSEL, are you going to check if backport
> includes previous patch and object/review such backport?
One more point. Except issues with backporting, you did not annotate any
dependency so patches can be applied independently. This will lead to
non-bisectable tree or even broken tree. What's more DTS goes always via
separate tree and branches, so this patch must be delayed.
You always must explicitly mention such dependencies and changes to
default applying process, so maintainers know what to do. Nothing like
this was explained anywhere here.
Best regards,
Krzysztof
