From time to time, I get a *HUGE* *LOT* of the following messages:

May 25 08:29:48 blatt [drm:via_cmdbuf_wait] *ERROR* via_cmdbuf_wait timed out
hw 1fb300 cur_addr 17b300 next_addr 1fb402

When I get those messages, the system becomes unworkably slow.

I'm not yet quite sure what's causing these messages, but when I get those
messages, I usually had the notebook running unattendedly for some time.

On the unichrome bugtracker, I found
https://sourceforge.net/tracker/index.php?func=detail&aid=1157219&group_id=102048&atid=630967
which implied that the via_agp kernel module isn't loaded. That's not true for
my system - or maybe it is, but then it would have been unloaded automatically
while X was running.

When I get these messages, the "hw", "cur_addr" and "next_addr" addresses do
NOT change over a run. Ie.,at May 25, I had:

May 25 08:29:48 blatt [drm:via_cmdbuf_wait] *ERROR* via_cmdbuf_wait timed out
hw 1fb300 cur_addr 17b300 next_addr 1fb402

So, at May 25, I always had "1fb300", "17b300" and "1fb402". A few days ago, I
had:

May 14 09:26:30 blatt [drm:via_cmdbuf_wait] *ERROR* via_cmdbuf_wait timed out
hw 6f000 cur_addr 400 next_addr 80502

etc.pp.

I'm using 

alexander@blatt /var/log $ uname -a
Linux blatt 2.6.16-suspend2-r6.043.security-models #1 PREEMPT Sat May 20
19:48:42 CEST 2006 i686 Intel(R) Celeron(R) M processor         1.50GHz
GNU/Linux

on a Gentoo Linux system with xorg-server 1.0.2 from xorg.freedesktop.org.