Re: [PATCH 3/3] ASoC: amd: fix for variable set but not used warning

[Mukunda,Vijendar]

On 7/9/22 1:03 PM, Christophe JAILLET wrote:
> Le 07/07/2022 à 15:26, Vijendar Mukunda a écrit :
>> Fix below kernel warning.
>>>>> sound/soc/amd/acp-es8336.c:200:13: warning: variable 'ret' set but
>>>>> not used [-Wunused-but-set-variable]
>>
> 
> Hi,
> this patch, looks odd to me.
> 
> 
>> Signed-off-by: Vijendar Mukunda <Vijendar.Mukunda@xxxxxxx>
>> Reported-by: kernel test robot <lkp@xxxxxxxxx>
>> ---
>>   sound/soc/amd/acp-es8336.c | 3 +++
>>   1 file changed, 3 insertions(+)
>>
>> diff --git a/sound/soc/amd/acp-es8336.c b/sound/soc/amd/acp-es8336.c
>> index 90f4e5809c72..e1479ae684e9 100644
>> --- a/sound/soc/amd/acp-es8336.c
>> +++ b/sound/soc/amd/acp-es8336.c
>> @@ -206,6 +206,8 @@ static int st_es8336_late_probe(struct
>> snd_soc_card *card)
>>           dev_err(card->dev, "can not find codec dev\n");
> 
> The next devm_acpi_dev_add_driver_gpios() will fail, should we return
> immediately?
> 
>>         ret = devm_acpi_dev_add_driver_gpios(codec_dev,
>> acpi_es8336_gpios);
>> +    if (ret)
>> +        dev_warn(card->dev, "Failed to add driver gpios\n");
> 
> Should we return immediately?
As it required to support Machine driver differed probe , we shouldn't
return immediately.
We are checking gpiod_get_optional() return status. If still error is
reported, then return is invoked after checking whether return error
code is -EPROBE_DEFER.

We found similar implementation in other platforms machine driver code
as well.

>>         gpio_pa = gpiod_get_optional(codec_dev, "pa-enable",
>> GPIOD_OUT_LOW);
>>       if (IS_ERR(gpio_pa)) {
>> @@ -213,6 +215,7 @@ static int st_es8336_late_probe(struct
>> snd_soc_card *card)
>>                       "could not get pa-enable GPIO\n");
>>           gpiod_put(gpio_pa);
>>           put_device(codec_dev);
>> +        return ret;
> 
> Here, we return 'ret' which is what is returned by
> devm_acpi_dev_add_driver_gpios(). So there doesn't seem to be a link
> with this block which checks for gpiod_get_optional() errors.
> 
> Moreover, returning an error for something that is optional
> (gpiod_get_optional()) looks strange.
> 
> Finally, should an error be returned, maybe PTR_ERR(gpio_pa) would be
> better here.

Machine driver deferred probing should be supported.
err code PTR_ERR(gpio_pa) is compared with -EPROBE_DIFFER and same err
returned from dev_err_probe() API.

same code can also be modified as below

if (IS_ERR(gpio_pa)) {
	gpiod_put(gpio_pa);
	put_device(codec_dev);
	return dev_err_probe(card->dev, PTR_ERR(gpio_pa),
			    "could not get pa-enable GPIO\n");		

}





> 
> 
> Just my 2c.
> 
> CJ
> 
>>       }
>>       return 0;
>>   }
>