Hi Cezary,
On 12/13/22 13:55, Cezary Rojewski wrote:
> On 2022-12-13 1:33 PM, Hans de Goede wrote:
>> For some reason rt5670_i2c_probe() does a pm_runtime_put() at the end
>> of a successful probe. But it has never done a pm_runtime_get() leading
>> to the following error being logged into dmesg:
>>
>> rt5670 i2c-10EC5640:00: Runtime PM usage count underflow!
>>
>> Fix this by removing the unnecessary pm_runtime_put().
>>
>> Fixes: 64e89e5f5548 ("ASoC: rt5670: Add runtime PM support")
>> Signed-off-by: Hans de Goede <hdegoede@xxxxxxxxxx>
>> ---
>> sound/soc/codecs/rt5670.c | 2 --
>> 1 file changed, 2 deletions(-)
>>
>> diff --git a/sound/soc/codecs/rt5670.c b/sound/soc/codecs/rt5670.c
>> index ebac6caeb40a..a230f441559a 100644
>> --- a/sound/soc/codecs/rt5670.c
>> +++ b/sound/soc/codecs/rt5670.c
>> @@ -3311,8 +3311,6 @@ static int rt5670_i2c_probe(struct i2c_client *i2c)
>> if (ret < 0)
>> goto err;
>> - pm_runtime_put(&i2c->dev);
>> -
>> return 0;
>> err:
>> pm_runtime_disable(&i2c->dev);
>
>
> Hello Hans,
>
> Good finding, weird one though. I just analyzed commit pointed by 64e89e5f5548 and it seems that entire change could be reverted. The rt5670 i2c_driver does not assign any PM ops, only ASoC component ones. The later is tied to machine board driver though (and it assigning snd_soc_pm_ops in its descriptor).
>
> From i2s (non-sdw) Realtek codec drivers found in sound/soc/codecs it seems that only rt9120.c defines PM ops and configures the PM for the device. OTOH, there are several which define suspend/resume on ASoC component level and do not touch pm_runtime_* at all.
>
> Thus, voting for a complete revert.
A complete revert seems fine to me too.
Regards,
Hans
