Created attachment 79658 [details]
dmesg after X freeze

When watching a flash video (opera + flash-11.2.202.262 ) the kernel log starts
filling up with 
[ 7009.603310] radeon 0000:01:00.0: GPU fault detected: 146 0x0e677004
[ 7009.603313] radeon 0000:01:00.0:   VM_CONTEXT1_PROTECTION_FAULT_ADDR  
0x00000000
[ 7009.603316] radeon 0000:01:00.0:   VM_CONTEXT1_PROTECTION_FAULT_STATUS
0x00000000

to sometimes eventually hang the GPU. When this starts appearing there is
graphic corruption everywhere. I also managed to reproduce this when playing
back a video using mplayer with the vdpau backend. ( EnableLinuxHWVideoDecode=1
is commented out in /etc/adobe/mms.cfg)

I managed to get a dmesg output after the GPU hang (attached) and then tried to
save what is in /sys/kernel/debug/dri/0 but not knowing which was important,
tried to do it in a bash for cycle, and got a complete hang.

linux-3.10-rc1
mesa, libdrm, radeon ddx from git.

might be related to https://bugs.freedesktop.org/show_bug.cgi?id=62959 , but
piglit just finished fine. (fine = didn't hang in this case, still a bunch of
"radeon_gem_object_create:69 alloc size 1365Mb bigger than 256Mb limit" in the
logs).
Additional info : I have a dual screen setup.

Anything more you need, I'll be happy to provide.