snd_hda: loud distorted sound with NVidia MCP51

[Andrew Pantyukhin]

On 12/22/06, Andrew Pantyukhin <infofarmer at freebsd.org> wrote:
> On 12/19/06, Ariff Abdullah <ariff at freebsd.org> wrote:
> > On Mon, 18 Dec 2006 23:57:00 +0300
> > "Andrew Pantyukhin" <infofarmer at FreeBSD.org> wrote:
> > > On 12/18/06, Hiroharu Tamaru <tamaru at myn.rcast.u-tokyo.ac.jp> wrote:
> > > > This is already commited. So look for this line
> > > >               if (subvendor == 0x81cb1043) {
> > > > and replace it with
> > > >               if (subvendor == 0x81cb1043 || subvendor ==
> > > >               0x10b81734) {
> > > > and have a try.
> > >
> > > This actually works very well! Thank you very much!
> >
> > Fix committed
>
> I confirm that -current works great. Thank you!
>
> > > Ariff, I'll provide additional testing as needed.
> > >
> > > Also there's another problem, independent of the
> > > patch. Sometimes the card starts playing, but
> > > just as about half a second passes the output
> > > stops (mpg123 exits) and kernel says this:
> > >
> > > pcm0:virtual:0:dsp0.v0: play interrupt timeout, channel dead
> >
> > Could you try this patch (attached). snd_hda supports polling mode
> > instead of interrupt, so you might give it a try as well, _after_
> > the given patch failed to address this issue. sysctl
> > dev.pcm.0.polling.
>
> I'm still trying to find a way to reproduce it. I
> could hardly do it a couple of times since the first
> one. When I'm sure I can cause it, I'll try the fix
> and tell you.

So. I applied the patch and the error disappeared. But,
it feels like some glitch is still there. In the rare
situations when I would expect the error - when the
laptop has just booted and I have just kldloaded
snd_hda - instead of an error I get a 2-10 sec delay
when I start mpg123 before I hear any output. The
output starts from the beginning, i.e. nothing is lost,
there's just a delay.

A delay is much better than an error, as mpg123 does
not exit. Thanks! I'm ready to do further testing.