> some weeks ago when we had a locking problem in the audio
> filter code I found that the device softc is passed around
> for no good reason. The locking stuff would indeed be much easier
> if a filter pipeline was just a thing of its own, not referring
> to pointers which might be volatile.
> I've done the change in my tree (patch appended), and havn't
> seen any problems for more than 2 months.
> Is this OK to commit?

sys/dev/audiomix.c in kent-audio2 branch refers softc.

-- 
TAMURA Kent <kent_2006 at hauN.org> <kent at NetBSD.org>