Created attachment 21181 [details]
"bt full" stacks showing which functions is being hit in the infiniate loop plus their params
When I got back from work my 965 laptop was running the "busy spheres" screensaver. When I tried to wake it up Xorg went into an infinite loop hogging 100% CPU. I sampled some "bt full" stacks and saved xorg_log/dmesg (attached).
Created attachment 21182 [details]
xorg_log saved while CPU spin was in effect
Created attachment 21183 [details]
dmesg saved when CPU spin was still running
I was using compiz when this happened. These are my package versions as seen in "dpkg -l":
intel driver 2:2.4.1-1ubuntu10
Is the size parameter in a "reasonable" range? I have no idea what that function does but the value seems pretty big?
#2 0xb79dc29b in drmCommandWrite (fd=12, drmCommandIndex=5, data=0xbf825998, size=3212990872) at xf86drm.c:2305
Created attachment 21195 [details]
xorg_log, dmesg and gdb backtraces show inf loop (second repro)
I left this laptop running the same screensaver again overnight. When I woke up it was still animating the screensaver but when I tried to wake it up it froze. Infinite loop, exact same stacks etc.
So basically this bug is reproducible (at least on this machine).
Created attachment 21196 [details]
output of /proc/dri/0/*
Does this only happen with the busy spheres screen saver? If so it might actually be a DRI driver bug...
reassign to Haihao.
Can you reproduce this with DRI2? If so, could you attach the output of intel_gpu_dump so we can try to figure out what's going on?
Feedback timeout. Just left this running for a day on a GM965 here, and it appears to be fine.