[alsa-devel] SB X-Fi driver fixes (Re: Testers wanted: New SB X-Fi driver)

[Takashi Iwai]

At Tue, 09 Jun 2009 08:24:41 +0200,
I wrote:
> 
> At Tue, 09 Jun 2009 07:44:29 +0200,
> Gernot Kohlhaas wrote:
> > 
> > Takashi Iwai schrieb:
> > > At Tue, 09 Jun 2009 07:23:56 +0200,
> > > Gernot Kohlhaas wrote:
> > >> Takashi Iwai schrieb:
> > >>> Also, do you mean that the default output works without 1/3 second
> > >>> distortion?  Or, the distortion happens regardless of your PCM device
> > >>> setup?
> > >>>
> > >>> I'd like to concentrate on debugging the default setup at first,
> > >>> so better to reduce the risk of brokenness of ladspa plugin.
> > >>> If you get any severe problems with the default setup, please let me
> > >>> know more details.
> > >>>
> > >>>
> > >>> thanks,
> > >>>
> > >>> Takashi
> > >>>
> > >> No, this distortion happens "everywhere" when I compile the module with
> > >> --debug=full.
> > > 
> > > That's weird.  Do you get any kernel messages?
> > > 
> > Unfortunately not. Just the "standard" ones (when loading the module).
> 
> Hm, please double-check later.  I really don't see any possible cause
> of this.
> 
> > >> For the crashes: They only happen when I use the LADSPA low pass filter
> > >> AND when I am fast forwarding in xmms. When I don't use fast forwarding,
> > >> everything works fine. But I will try use_system_timer=1 when I get back
> > >> from work today.
> > >> It really seems like the plug-in is doing some weird stuff.
> > > 
> > > But the plugin itself cannot lead to a system crash.  It's the driver.
> > > 
> > > Do I understand correctly that your system locks up when this happens,
> > > right?  If yes, the cause might be the mmap mode access.
> > > 
> > Yes, it locks up. No kernel panic or something like that. A "hard"
> > lock-up/freeze.
> 
> OK, this implies either a spin deadlock or a dead-end in the irq
> handler.
> 
> Anyway, please try the very latest alsa-driver-snapshot.  If it still
> doesn't work, let me know whether use_system_timer=1 works around the
> crash.  If yes, just try to comment out all spin_*() calls in
> cttimer.c.

Also, show your ~/.asoundrc setup so that I can try to reproduce
the same bug with my test machine.


thanks,

Takashi