Re: sound/pcm/* bugs (was: Re: page fault panic tracked down (selwakeuppri()) - really sound/pcm/*)

On Wed, 2004-01-14 at 22:28, Don Lewis wrote:
> On 14 Jan, Stefan Ehmann wrote:
> > That seems to fix it - boots fine now.
> > 
> > Looks like we need some further locks. When I try to set vchans to 4
> > with sysctl I get a similar panic.
> > 
> > panic: mutex pcm0:play:0 not owned at
> > /usr/src/sys/dev/sound/pcm/channel.c:706
> > 
> > #15 0xc04e5b57 in panic () at /usr/src/sys/kern/kern_shutdown.c:550
> > #16 0xc04dc4fc in _mtx_assert (m=0xc37ba280, what=0, 
> >     file=0xc07ee3e9 "/usr/src/sys/dev/sound/pcm/channel.c", line=706)
> >     at /usr/src/sys/kern/kern_mutex.c:673
> > #17 0xc07e436f in chn_reset (c=0x2c2, fmt=268435472)
> >     at /usr/src/sys/dev/sound/pcm/channel.c:706
> > #18 0xc07ed1fa in vchan_create (parent=0xc37a8880)
> >     at /usr/src/sys/dev/sound/pcm/vchan.c:273
> 
> Sigh ... I had hoped that your earlier testing had meant that locking
> code was clean.  Here's another patch, but I won't guarantee that you
> won't run into more lock assertion problems.  I guess that I should set
> up some sound hardware here so that I can actually test some of this
> stuff.

This time the system survived sysctl but as soon as sound playback
starts the assertion in channel.c:978 gets triggered.

Looks like setting up sound would be a good idea since there are
probably more traps ahead.