Re: [PATCH net-next] net: page_pool: fix warning code

[Przemek Kitszel <przemyslaw.kitszel@intel.com>]

On 7/5/24 15:28, Alexander Lobakin wrote:
> From: Przemek Kitszel <przemyslaw.kitszel@intel.com>
> Date: Fri, 5 Jul 2024 15:14:53 +0200
> 
>> On 7/5/24 14:39, Johannes Berg wrote:
>>> On Fri, 2024-07-05 at 14:37 +0200, Alexander Lobakin wrote:
>>>> From: Johannes Berg <johannes@sipsolutions.net>
>>>> Date: Fri, 05 Jul 2024 14:33:31 +0200
>>>>
>>>>> On Fri, 2024-07-05 at 14:32 +0200, Alexander Lobakin wrote:
>>>>>> From: Johannes Berg <johannes@sipsolutions.net>
>>>>>> Date: Fri,  5 Jul 2024 13:42:06 +0200
>>>>>>
>>>>>>> From: Johannes Berg <johannes.berg@intel.com>
>>>>>>>
>>>>>>> WARN_ON_ONCE("string") doesn't really do what appears to
>>>>>>> be intended, so fix that.
>>>>>>>
>>>>>>> Signed-off-by: Johannes Berg <johannes.berg@intel.com>
>>>>>>
>>>>>> "Fixes:" tag?
>>>>>
>>>>> There keep being discussions around this so I have no idea what's the
>>>>> guideline-du-jour ... It changes the code but it's not really an issue?
>>>>
>>>> Hmm, it's an incorrect usage of WARN_ON() (a string is passed instead of
>>>> a warning condition),
>>>
>>> Well, yes, but the intent was clearly to unconditionally trigger a
>>> warning with a message, and the only thing getting lost is the message;
>>> if you look up the warning in the code you still see it. But anyway, I
>>> don't care.
>>>
>>
>> for the record, [1] tells: to the -next
>>
>> [1]
>> https://lore.kernel.org/netdev/20240704072155.2ea340a9@kernel.org/T/#m919e75afc977fd250ec8c4fa37a2fb1e5baadd3f
> 
> But there you have "the exact same binary getting generated", while here
> you clearly have functional and binary changes :>

one could say that in this case the change to the binary is meaningless
--
I think that the most important rationale is that this code change would
require a -net level of testing, and in theory changed binary means
"some adjacent bug my now reveal", and for such changes we have -next.

The above even answers why "no binary changes" [1] above should be still
considered -next material: there could be some compiler version/flags
combo that would generate different binary, and we get back to the prev 
paragraph.

> 
> I'd even say it may confuse readers and make someone think that you
> should pass a string there, not a condition (weak argument, I know).
> 
>>
>>> The tag would be
>>>
>>> Fixes: 90de47f020db ("page_pool: fragment API support for 32-bit arch
>>> with 64-bit DMA")
>>>
>>> if anyone wants it :)
>>>
>>> johannes
> 
> Thanks,
> Olek