[solved] Random hang, forcing reboot - suspect graphics driver or card

Question

Stats:
[nathanb@nathanb-box ~] kate --version
Qt: 4.8.4
KDE Development Platform: 4.9.5
Kate: 3.9.5
[nathanb@nathanb-box ~] uname -a
Linux nathanb-box 3.6.11-1-ARCH #1 SMP PREEMPT Tue Dec 18 08:57:15 CET 2012 x86_64 GNU/Linux
Computer was running fine until today, when upon trying to unlock it there was a hang. I could move the mouse cursor, and the keyboard LEDs would toggle, but there was nothing on the screen (except the cursor).
Throughout the day, I've seen random graphical corruption + hang.
Looking through the system logs, before each hang I see something like the following:
Jan 7 16:02:17 localhost kernel: [ 3398.654742] [drm] nouveau 0000:03:00.0: EvoCh 1 Mthd 0x0094 Data 0xcafe0000 (0x0000 0x07)
Jan 7 16:02:17 localhost kernel: [ 3398.671174] [drm] nouveau 0000:03:00.0: PFIFO0: (unknown bits 0x80000000)
Jan 7 16:02:17 localhost kernel: [ 3398.671174] [drm] nouveau 0000:03:00.0: PFIFO0: ch 0 subc 0 mthd 0x0000 data 0x00000000
Jan 7 16:02:17 localhost kernel: [ 3398.671188] [drm] nouveau 0000:03:00.0: PFIFO0: (unknown bits 0x80044000)
Jan 7 16:02:17 localhost kernel: [ 3398.671188] [drm] nouveau 0000:03:00.0: PFIFO0: ch 0 subc 0 mthd 0x0000 data 0x00000000
Jan 7 16:02:17 localhost kernel: [ 3398.671205] [drm] nouveau 0000:03:00.0: PFIFO0: (unknown bits 0x80004000)
Jan 7 16:02:17 localhost kernel: [ 3398.671205] [drm] nouveau 0000:03:00.0: PFIFO0: ch 0 subc 0 mthd 0x0000 data 0x00000000
Jan 7 16:02:17 localhost kernel: [ 3398.671225] [drm] nouveau 0000:03:00.0: PFIFO: read fault at 0x0000000000 [PT_NOT_PRESENT] from PFIFO/PFIFO on channel 0x00000c8000
Jan 7 16:02:17 localhost kernel: [ 3398.671312] [drm] nouveau 0000:03:00.0: EvoCh 2 Mthd 0x0094 Data 0xcafe0000 (0x0000 0x07)
Jan 7 16:02:17 localhost kernel: [ 3398.903679] [drm] nouveau 0000:03:00.0: GPU lockup - switching to software fbcon
Jan 7 16:02:17 localhost kernel: [ 3398.903768] [drm] nouveau 0000:03:00.0: PFIFO: write fault at 0x00000a0000 [PAGE_NOT_PRESENT] from BAR1/BAR_WRITE on channel 0x0000072000
Jan 7 16:02:17 localhost kernel: [ 3398.919619] [drm] nouveau 0000:03:00.0: PFIFO: write fault at 0x00002f2000 [PAGE_NOT_PRESENT] from BAR3/BAR_WRITE on channel 0x0000068000
That line reading "GPU lockup" seems especially suspicious.
This is a fairly new graphics card; purchased in the last six months. I'm using nouveau, obviously, with stock kernel. Have never seen anything like this with this graphics card before today. Don't think I've done a system update yet this year.
Some Googling turned up a possible match (I had the URL, but then somehow it got dropped, oh well), but that's not identical since the problem described in the bug is the X server not starting, while mine will start and run for a few hours before barfing. It does involve the "GPU lockup" message, and is a software bug, but that bug is theoretically fixed as of the 3.6 kernel (I'm using 3.6.11).
I'm going to try installing the nvidia proprietary drivers, but that's a pain in my rear.
Does this problem sound familiar to anyone? Any upstream bug I could track?
Last edited by nathanb (2013-05-17 19:05:03)

Answer 1

Follow-up: the nvidia proprietary drivers fixed the problem.