Freeze/X lockup during VDPAU playback

[Licaon]

Originally Posted by mlauss

I have this hang with *every* clip; the time it takes to trigger is random but it
happens usually within an hour after booting up; mplayer then is dead.
After killing it, it works for days without a hitch (although black areas show color artifacts).
Btw, the 195.36.X drivers are fine.

can you read http://www.nvnews.net/vbulletin/showthread.php?t=46678 and generate a good report with the debug ( -logverbose 6 ) info as instructed after the issue appears ?

[Stephen Warren]

Everyone experiencing a problem should generate and post an nvidia-bug-report.

Note that just because multiple people are seeing a hang doesn't mean that any of these issues are the same or related. Equally, bad hardware can affect multiple people. (I myself used to blame my home server's instability on bad reiserfs kernel code, but eventually my power supply went completely bad, I replaced it, and then never had any problems again. Just an example)

Unfortunately, I'm unable to debug any problem I can't reproduce.

[mlauss]

This issue is gone for me with the newest beta driver (270.26).
I've been pressing the "f" button on mplayer for 6hrs now (so that
it rapidly switches between windowed and fullscreen) and not once
did it freeze.

EDIT: spoke too soon.

[Civilus]

Stephen Warren
I've got 2 kinds of issues:
1) X sometimes (really sometimes - I don't figured out how to reproduce it with 100% chance) hangs at start, screen corruption before or just after modeswithch, after a while monitor turns off, like this: http://imageshack.us/photo/my-images/16/mg5197.jpg/ (pattern always same if it matters). I can ssh to machine but I don't think I can get any useful bugreport.

The last only error in Xorg.0.conf is:
[ 48.885] (EE) NVIDIA(0): WAIT: (E, 0, 0x857d, 0)

It boots only if I add NVreg_RegistryDwords="PerfLevelSrc=0x2222" to modules.d/nvidia.conf

I've attached nvidia-bugreport for this. Again.

p.s. backtrace when X hangs up at start:

#0  0x00007f71bf53ebdf in ?? () from /usr/lib64/xorg/modules/drivers/nvidia_drv.so
#1  0x00007f71bf4dcc00 in ?? () from /usr/lib64/xorg/modules/drivers/nvidia_drv.so
#2  0x00007f71bf4dcee9 in ?? () from /usr/lib64/xorg/modules/drivers/nvidia_drv.so
#3  0x00007f71bf4cf430 in ?? () from /usr/lib64/xorg/modules/drivers/nvidia_drv.so
#4  0x00007f71bf4d28d8 in ?? () from /usr/lib64/xorg/modules/drivers/nvidia_drv.so
#5  0x00007f71bf844f06 in ?? () from /usr/lib64/xorg/modules/drivers/nvidia_drv.so
#6  0x00007f71bf8451ae in ?? () from /usr/lib64/xorg/modules/drivers/nvidia_drv.so
#7  0x00007f71bf8384bf in ?? () from /usr/lib64/xorg/modules/drivers/nvidia_drv.so
#8  0x0000000000434efb in AddScreen ()
#9  0x0000000000495536 in InitOutput ()
#10 0x000000000042469e in _start ()

So nothing useful. Where can I get debug symbols for nvidia_drv?
2) Blackscreen and possibly OOPS (almost always if PowerMizer is on). It reboots after a while, but I still can't get any backtraces. I don't have COM port on my mobo, and I don't have IEEE1394 on other machines, I've tried to set up kdump, but for some reason it doesn't work yet.

Any suggestions how to get some useful info?

Next thing I'll try is to recompile whole system with -O0 -g -ggdb and attach to hang X process with gdb, maybe I would manage to get something interesting with stacktrace/backtrace.

[mlauss]

Here's a bug report and vdpau trace when the X hang occured.
It's pretty easy to reproduce: KDE-4.6, firefox and a few xterms on one virtual desktop,
place a few other xterms on another virtual desktop, start mplayer fullscreen on a second head, direct kwin to pin it to all desktops and start switching between desktops with the slide effect.

It's much harder to reproduce when KDE desktop effects are disabled.