View Single Post
Old 03-31-11, 12:26 PM   #1
phz_swe
Registered User
 
Join Date: Jan 2008
Posts: 2
Default VDPAU -> pixel snow on screen until reboot

Geforce 9400GT, driver versions when problem has occured: 260.19.44-1, 270.30-1 (from Debian Sid/Experimental respectively). Before 260.19.44-1, I used 195.36.31-7, which was the previous version in Debian Sid.

Quite randomly (after a week's uptime or so) when viewing video with VDPAU, a lot of pixel artefacts begin to appear, not only on the video surface. Thereafter there are a lot of pixel artefacts all over the screen in X until reboot (just restarting X doesn't help).

See pictures:
http://tinypic.com/r/2s79dzk/7 - glxgears
http://tinypic.com/r/s16cdu/7 - video and glxgears (which looks better this way, for some reason)
Also notice that the pixel corruption isn't localized to glxgears/the video window. The Pidgin conversation looks obfuscated, but that's not deliberate. All areas of the screen are affected. (An interesting personal note is that Google Chrome has a lot more obfuscation problems than e.g. Opera, which I guess has something to do with GPU rendering.)

"Quite randomly" I said: it might be connected to incomplete video files and such, but I can't tell for sure. The last time it happened it was with a regular xvid file I've encoded myself, which has played without any problems before.

After the snow has appeared once during a session, subsequent video file playbacks are much more likely (certain?) to fail. Since I can't predict when it will happen the first time, I haven't gotten a bug trace of the initial occurence, but several of the subsequent playback crashes.

Sometimes it hangs the entire X session. ssh shows that X takes ~100% CPU during these hangs. Sometimes MPlayer exits by itself. Sometimes it is possible to press "q" to quit MPlayer with a couple of seconds delay.

I attach MPlayer bug traces with VDPAU_TRACE=1 and VDPAU_NVIDIA_DEBUG=3. I took three bug traces since the crash appeared in some different ways. I ran "startx -- -logverbose 6".
-nvidia-bug_trace-mplayer_catch.txt - I could switch to the terminal window and kill MPlayer with Ctrl+c.
-nvidia-bug_trace-mplayer_all_by_itself_1.txt - MPlayer exited by itself.
-nvidia-bug_trace-mplayer_all_by_itself_2.txt - MPlayer exited by itself, but in a slightly different way message-wise.

It has occured with very different video files. Once it has started to snow, any file seems to be able to reproduce the freeze behaviour.

I experience a), d) and e) of http://www.nvnews.net/vbulletin/showthread.php?t=148113 , but the problem didn't exist for me with the mentioned driver version, so I guess only the symptoms and not the causes are related.
Attached Files
File Type: gz nvidia-bug-report.log.gz (109.8 KB, 39 views)
File Type: txt nvidia-bug_trace-mplayer_catch.txt (79.5 KB, 60 views)
File Type: txt nvidia-bug_trace-mplayer_all_by_itself_1.txt (53.9 KB, 59 views)
File Type: txt nvidia-bug_trace-mplayer_all_by_itself_2.txt (59.1 KB, 55 views)
phz_swe is offline   Reply With Quote