On 26/07/2024 10:13, Charles Keepax wrote:
> On Thu, Jul 25, 2024 at 12:31:40PM +0200, Javier Carrasco wrote:
>> The instances of the `snd_soc_component_driver` struct are not modified
>> after their declaration, and they are only passed to
>> `devm_snd_soc_register_component()`, which expects a constant
>> `snd_soc_component_driver`.
>>
>> Move all instances of `snd_soc_component_driver` to read-only sections
>> by declaring them const.
>>
>> Signed-off-by: Javier Carrasco <javier.carrasco.cruz@xxxxxxxxx>
>> ---
>> sound/soc/codecs/cs43130.c | 2 +-
>
>> -static struct snd_soc_component_driver soc_component_dev_cs43130 = {
>> +static const struct snd_soc_component_driver soc_component_dev_cs43130 = {
>> .probe = cs43130_probe,
>> .controls = cs43130_snd_controls,
>> .num_controls = ARRAY_SIZE(cs43130_snd_controls),
>
> This won't work for cs43130, whilst what the driver does is
> clearly slightly sketch it directly modifies this struct before
> registering it with ASoC. That would need fixed first before this
> change can be made.
>
> Thanks,
> Charles
Hi Charles,
thanks a lot for pointing this out, somehow I failed to compile cs43130
and sti-sas, which are the only two cases in the kernel that do modify
snd_soc_component_driver after the declaration. The rest don't do that
and I just double checked that they compile cleanly.
Those two cases where modifications are required rely on values that are
not known until they are probed. I think it makes then sense that they
are left as they are, so I will drop both drivers for v2.
Best regards,
Javier Carrasco
