[alsa-devel] Problems with pulseaudio / asla on PS3

[Takashi Iwai]

At Tue, 9 Sep 2008 13:17:06 +0200 (CEST),
Jaroslav Kysela wrote:
> 
> On Tue, 9 Sep 2008, Jaroslav Kysela wrote:
> 
> > On Tue, 9 Sep 2008, Takashi Iwai wrote:
> > 
> > > At Mon, 8 Sep 2008 22:21:36 +0200,
> > > Lennart Poettering wrote:
> > > > 
> > > > Takashi, as Kevin wrote originally, the problem is not that softvol
> > > > doesn't do any format/layout conversions, but that it claims it
> > > > does. PA asks softvol for INTERLEAVED samples, softvol agrees to
> > > > that -- and actually hands us only NONINTERLEAVED samples because it
> > > > doesn't reformat the from the underlying layer.
> > > > 
> > > > So, it *is* a bug in softvol. However, you are right that it doesn't
> > > > do any conversions.
> > > 
> > > Does it happen only with softvol?  For example, try to put any
> > > other conversion, such as, linear plugin instead of softvol.
> > > Could you check whether it works?
> > > 
> > > Basically softvol does nothing special, just like other normal plugins.
> > > So I don't think this is specific to softvol.
> > 
> > Also, output from snd_pcm_dump() after initialization (before samples r/w) 
> > might help us to determine the problem. I'll try to reproduce this 
> > behavior with the snd-dummy module.
> 
> I found the culprit. Lennart was right. The softvol plugin does in-place 
> sample recalculation, thus it cannot change the access type as a side 
> effect (which is true for linear plugins).

Ah, right, that's around mmap shadowing.

> The fix is available here:
> 
> http://git.alsa-project.org/?p=alsa-lib.git;a=commit;h=a13707da6bb0161db855a146c3e4d1d849e4108b
> 
> Also, for debugging purposes, I added snd_pcm_channel_area_t dump to 
> aplay:
> 
> http://git.alsa-project.org/?p=alsa-utils.git;a=commit;h=6bd621177dea5afd74261621c72ff98ca4fc118a

Thanks!

Takashi