[alsa-devel] Regression: snd_hda_intel driver causes quake3 to hang

[Takashi Iwai]

At Thu, 19 Feb 2009 16:55:39 +0000,
Charlie Baylis wrote:
> 
> On Tue, Feb 17, 2009 at 01:05:46PM +0100, Takashi Iwai wrote:
> > At Tue, 17 Feb 2009 11:49:17 +0000, Charlie Baylis wrote:
> > > 
> > > 
> > > Hi
> > > 
> > > Some time ago, quake3 stopped running correctly on my PC. This occured when
> > > changing kernel from 2.6.25 to 2.6.27.  The symptom is that within a
> > > fraction of a second of gameplay, the game freezes. Switching to a console
> > > allows the game to be killed. I am using standard kernel.org kernels with
> > > no additional patches.
> > 
> > Could you check whether it happens with 2.6.29-rc5?  On 2.6.29, the delayed
> > handling is done via its own workqueue, so this might be no longer a problem.
> > 
> > Also you could try position_fix=1 instead of bdl_pos_adj. 
> > 
> > The problem is likely your hardware does unreliable DMA position reporting.
> > Detecting such a broken hardware would be possible, but it'll be a bit
> > tricky...
> 
> I tried the latest git (shortly after 2.6.29-rc5) and I found that the same
> symptoms occur. This happens with the driver defaults, or with bdl_pos_adj=0,
> or with position_fix=1. So 2.6.29 is worse, because there seems to be no
> workaround.
> 
> Is there a way to find out whether my hardware really is broken? I had no
> problems until the bdl_pos_adj workaround/feature was added.

On 2.6.29, bdl_pos_adj=0 disables that.  So it'll be a workaround for
you, at least.


Takashi