Re: [PATCH V2 4/6] dt-bindings: timestamp: Add Tegra234 support

[Krzysztof Kozlowski <krzk@xxxxxxxxxx>]

On 09/03/2023 19:49, Dipen Patel wrote:
> On 3/8/23 10:16 PM, Krzysztof Kozlowski wrote:
>> On 08/03/2023 21:09, Dipen Patel wrote:
>>> On 3/8/23 11:05 AM, Krzysztof Kozlowski wrote:
>>>> On 08/03/2023 19:45, Dipen Patel wrote:
>>>>> On 2/16/23 6:17 AM, Krzysztof Kozlowski wrote:
>>>>>> On 14/02/2023 12:55, Dipen Patel wrote:
>>>>>>> Added timestamp provider support for the Tegra234 in devicetree
>>>>>>> bindings.
>>>>>>
>>>>>> 1. Your commit does much more. You need to explain it why you drop some
>>>>>> property.
>>>>> ACK, will address it next patch
>>>>>>
>>>>>> 2. Bindings go before its usage (in the patchset).
>>>>> Ack...
>>>>>>
>>>>>> 3. Please use scripts/get_maintainers.pl to get a list of necessary
>>>>>> people and lists to CC.  It might happen, that command when run on an
>>>>>> older kernel, gives you outdated entries.  Therefore please be sure you
>>>>>> base your patches on recent Linux kernel.
>>>>> It is based on recent linux at the time patch series was sent...
>>>>
>>>> That's good but then why you do not use scripts/get_maintainers.pl? The
>>>> hint about recent kernel was just a hint... Just do not invent addresses
>>>> by yourself and use the tool to get them right.
>>>>
>>> I will take a note for the next patch series to add any missing people. The current
>>> list of people/group is what historically helped review this new timestamp/hte subsystem.
>>>
>>>> (...)
>>>>
>>>>>>> +  properties:
>>>>>>> +    compatible:
>>>>>>> +      contains:
>>>>>>> +        enum:
>>>>>>> +          - nvidia,tegra194-gte-aon
>>>>>>
>>>>>> This is an ABI break. Does your driver handle it?
>>>>> yes, handling patch is part of this patch series.
>>>>
>>>> Can you point me to the code which does it? I see "return -ENODEV;", so
>>>> I think you do not handle ABI break. I could miss something but since
>>>> you disagree with me, please at least bring some arguments...
>>> Refer to patch https://patchwork.kernel.org/project/timestamp/patch/20230214115553.10416-3-dipenp@xxxxxxxxxx/
>>> which has compatible properties added and also code changes to reflect addition/deletion of some
>>> properties.
>>
>> I referred to the code which breaks the ABI.
>>
>>>
>>> I am not sure I have understood about ABI break comment. How else one should handle if
>>> there is no related gpio controller property found?
>>
>> In a way it does not break existing users? There are many ways to handle
>> it, but I don't know your code to point you.
> 
> It is new subsystem and has only one driver which uses it so far. 

We do not talk about subsystem, but Tegra SoC, which is not new. Unless
you meant this is new SoC/DTS?

> This was a decision taken
> after review comments (By Thierry, also in the mailing list) to add this property (nvidia,gpio-controller)
> and necessary changes have been made to existing user. From now on, it has to follow this change.

What is "it" which has to follow? There are rules for stable ABI and
commit msg does not explain why they should not be followed.

> 
>>
>>> I am assuming you are referring to the
>>> below code from the patch 2 (link above) when you said "return -ENODEV".
>>
>>
>> Your bindings patch points to ABI break without any
>> explanation/justification. Then your code #2 patch actually breaks it,
>> also without any justification.
> I am going to add explanation/justification in the commit message in the next patch series. But to give
> you context, discussion happened here https://patchwork.ozlabs.org/project/linux-gpio/patch/20221103174523.29592-3-dipenp@xxxxxxxxxx/

Either too many messages (and I missed something) or I could not find
why ABI break is accepted and justified.


Best regards,
Krzysztof