Hi,

I've got a laptop that has skipping problems, mainly full-duplex
programs. This happens for both native ALSA applications and those using
OSS. Digging a bit into that, I realised that it mainly happens when
using two periods. I wrote a test program (attached) to dump the time I
get the "interrupts" (from user space). The output is attached in the
"timings" file (gettimeofday() microseconds). Basically, the time
increases with near perfect regularity, except for a few cases, where
there's a skip of exactly one period. Looks a lot like lost interrupts
or something like that (at least that was James Courtier-Dutton's
interpretation when he looked at it during LCA2007).

This data was obtained on a Dell D820 laptop with a STAC9200-based
soundcard running kernel 2.6.20 and ALSA 1.0.14rc2, but I've observed
the same problem with all other versions of kernel/ALSA I've tested.
I've also seen that problem on all other HDA-based desktops I have, but
not on any of the i810-based machines I have.

The problem usually does not show up with playback-only applications
(probably because they use >2 periods or something), but it makes VoIP
apps like Ekiga (native ALSA) and OpenWengo (OSS) totally unusable as I
get several xruns per second.

Does anyone know of a fix for that?

	Jean-Marc