[alsa-devel] AC97 slot information

[Jaroslav Kysela]

On Mon, 30 Nov 2009, Takashi Iwai wrote:

> At Sun, 29 Nov 2009 16:50:58 +0000,
> Russell King - ARM Linux wrote:
>>
>> When opening an AC'97 codec under ALSA using snd_ac97_pcm_open(),
>> specifying the number of channels desired (2, 4 or 6) is there a
>> way to read back which slots will be in use?
>
> Hm, a quick look through the code, there seems no such way right now.
> Actually, slots subsets are determined in snd_ac97_pcm_assign(), and
> the open function just accepts the reduced slots in there.
>
>> Currently, in the AACI driver, I use a lookup table to convert from
>> channels to a slot mask - if this information is already available
>> from ALSA, I'd much rather use it rather than inventing keeping my
>> own solution.
>>
>> Also, should the 'slots' value passed to snd_ac97_pcm_open() be
>> those in pcm->r[].slots or should it be selected from the number
>> of channels?
>
> pcm->r[].slots, I suppose.  Well, but snd_ac97_pcm_open() behaves a
> bit differently from what I thought.  It doesn't check cfg value
> (usually the number of channels (2, 4 or 6) is passed) to reduce the
> assigned slots, so it'll try always the full slots.  Then reducing
> the slots bits in the caller look mandatory right now...
>
> Jaroslav, is it the intended behavior, or just a missing piece?

The slot management is left to the lowlevel driver - see ac97_pcm_defs in 
intel8x0.c driver for example (all slots are configured in the 
snd_ac97_pcm_open() call - even if less channels are requested). The cfg 
value is not used for channel count handling (because we really do not 
know the ac97 slot -> real channel position mapping in the universal ac97 
routines - it depends how the ac97 controller is set). Actually, I don't 
have any idea to simplify this mapping and if it's worth to add handling 
of this mapping to the AC97 driver tree.

 					Jaroslav

-----
Jaroslav Kysela <perex at perex.cz>
Linux Kernel Sound Maintainer
ALSA Project, Red Hat, Inc.