On 01/15/2017 09:08 PM, Jarkko Nikula wrote:
On 01/15/2017 12:05 PM, Pavel Machek wrote:
So regression seems to be between v4.9 and v4.10. Any ideas?
Interesting... seems there are no sound relevant changes after v4.9.
Looks like there are only three commits after v4.9 for sound/soc which are built for Nokia N900:
e411b0b5eb9b65257a050eac333d181d6e00e2c6 e7aa450fe17890e59db7d3c2d8eff5b6b41fc531 63c3194b82530bd71fd49db84eb7ab656b8d404a
Maybe something not related to sound/soc could broke it?
Lets see.
a9042defa29a01cc538b742eab047848e9b5ae14 -- works ok. ce38207f161513ee3d2bd3860489f07ebe65bc78 --
alsactl: set_control:1328: failed to obtain info for control #229 (No such file or directory)
I'm unable to test older kernels at the moment since modules don't load by some reason (bluetooth: Unknown symbol _GLOBAL_OFFSET_TABLE_ (err 0)). Probably some recent regression in my debian/testing installation since the same configs were working before.
However v4.9.4 works and with it I can see there is an issue with these aic3x controls:
amixer: Unable to find simple control 'Left DAC_L1 Mixer Line',0
amixer: Unable to find simple control 'Right DAC_R1 Mixer Line',0
These are certainly not the regression between 4.6 and 4.9, more like more like regression since 2.6: 'Line Switch' -> 'LineL Switch': 54f01916297bafc18bd7df4e2300a0544a84fce3
'LineL Switch' -> 'DACL1 Switch': c3b79e05b4d9ab2e7c3ba281261ea87ab5b71a92
At quick look I don't see why above commits could cause it. Maybe regression happened earlier between v4.6 and v4.9?
It might be that there were some work to fix control names for different aic versions to match with the HW or support for new codec is added and some control gone missing for 3x?
I would compare the 'amixer controls' for the working and regressed kernel to see what differs. From there git blame or smthing will for sure point to the correct commit.