Bug#536896: linux-image-2.6.26-2-amd64: audio popping with gstreamer using apps when using amd64 kernel and i386 userspace

[Takashi Iwai <tiwai@suse.de>]

At Sun, 11 Apr 2010 21:25:16 +0100,
Ben Hutchings wrote:
> 
> [1  <text/plain; UTF-8 (quoted-printable)>]
> [Earlier messages can be found at http://bugs.debian.org/536896 ]
> 
> On Mon, 2010-04-12 at 01:08 +0800, James Andrewartha wrote:
> > On Thu, 25 Feb 2010, Moritz Muehlenhoff wrote:
> > 
> > > On Tue, Jul 14, 2009 at 09:57:54PM +0800, James Andrewartha wrote:
> > > > Package: linux-image-2.6.26-2-amd64
> > > > Version: 2.6.26-17
> > > > Severity: normal
> > > > 
> > > > All audio from gstreamer-using apps (eg Totem and Banshee) pops when I use a
> > > > 2.6.{25,26,28,30} amd64 kernel and i386 userspace. 2.6.24-etchnhalf.1-amd64
> > > > does not have this problem, nor does 2.6.26-2-686. It doesn't occur when using
> > > > non-gstreamer apps like mplayer or mpd, nor with various gst-launch pipelines
> > > > suggested by #gstreamer like gst-launch playbin uri=file:///home/trs80/a.mp3
> > > 
> > > Hi,
> > > The next release of Debian (6.0, code name Squeeze) will be based
> > > on 2.6.32. Please test the current 2.6.32 from unstable/testing and tell
> > > us whether the problem persists. If so, we should report it upstream
> > > to the kernel.org developers.
> > > 
> > > The 2.6.32 kernel is available from packages.debian.org and can
> > > be installed in both Debian stable, testing and unstable
> > > installations.
> > 
> > I've bisected it, and the bad patch is 130755108ba03461f69da990e54e02a254accd23:
> 
> Thanks for taking the time to do this.
> 
> > Author: Takashi Iwai <tiwai@suse.de>  2008-01-09 02:08:14
> > Committer: Jaroslav Kysela <perex@perex.cz>  2008-02-01 01:29:47
> > Parent: d948035a928400ae127c873fbf771389bee18949 ([ALSA] Remove PCM xfer_align sw params)
> > 
> >     [ALSA] PCM - clean up snd_pcm_lib_read/write
> >     
> >     Introduce a common helper function for snd_pcm_lib_read and snd_pcm_lib_write
> >     for cleaning up the code.
> >     
> >     Signed-off-by: Takashi Iwai <tiwai@suse.de>
> >     Signed-off-by: Jaroslav Kysela <perex@perex.cz>
> > 
> > I don't know anything about this code, but I'm happy to deal with upstream 
> > if you'd prefer me to.
> 
> The above commit is supposed to be cleanup, but it has at least one
> semantic change: snd_pcm_mmap_control::avail_min no longer applies to
> non-blocking file handles.  I don't know whether this is was an
> intentional or unintentional change, but it wasn't commented.  I also
> don't know whether this can explain the popping, but I expect that it
> has changed the timing of audio I/O.

The above change is essentially a fix of the buggy behavior for
non-blocking access.  avail_min is the definition for wake-up
behavior, and it doesn't define the blocking behavior.
But, it's possible that this changes the timing, indeed.  If so, it
implies that the app expects somehow wrongly.


thanks,

Takashi