Created attachment 44432 [details] First hang Got this hang twice, seems to be triggered by Firefox going to a certain webpage or some other combination, but otherwise it's not easy to reproduce. Second time was after rebooting the system and doing a session restore in Firefox. Display was frozen, but mouse cursor still worked. After sshing into the machine, I could get a display in console by killing X and resetting the device (I did echo 0>enabled, echo 1, and an echo 1>reset). Starting X didn't work because X detected a hung GPU, but could switch between X and console. This is with 2.6.38-rc8, xf86-video-intel 2.14.0, intel-dri 7.10.1 and libdrm 2.4.23.
Created attachment 44433 [details] Second hang Hang after a reboot and Firefox session restore. I guess it was triggered by a spam message on a forum, but it's deleted by another admin now, so can't use it to trigger the crash anymore, assuming it was caused by it. If it was, then it might be something related to glyphs/fonts.
Buffer: 07c80000 12288 0002 0000 000098ae X dirty render uncached (fence: 4) Batch: 0x05a802a0: 0x54f00006: XY_SRC_COPY_BLT (rgb enabled, alpha enabled, src tile 0, dst tile 0) 0x05a802a4: 0x03cc0200: format 8888, dst pitch 512, clipping disabled 0x05a802a8: 0x00000000: dst (0,0) 0x05a802ac: 0x00120068: dst (104,18) 0x05a802b0: 0x07c80000: dst offset 0x07c80000 0x05a802b4: 0x000005f9: src (1529,0) 0x05a802b8: 0x00002000: src pitch 8192 0x05a802bc: 0x07c00000: src offset 0x07c00000 Interesting. :(
And a very similar tail of blits in the second trace. Indan, can you add Option "DebugFlushCaches" "True" Option "DebugFlushBatches" "True| to the Section "Device" in your xorg.conf and so double check that it does fall over on the XY_SRC_COPY_BLT that looks the most likely suspect?
(In reply to comment #3) > And a very similar tail of blits in the second trace. Indan, can you add > > Option "DebugFlushCaches" "True" > Option "DebugFlushBatches" "True| > > to the Section "Device" in your xorg.conf and so double check that it does > fall over on the XY_SRC_COPY_BLT that looks the most likely suspect? Is there a simple X test program to trigger that thing? Cause I can't easily trigger this hang, that I got two hangs just after each other was plain "luck". As usual I'm willing to test patches. (If you don't hear from me the following days I'm gone camping.)
x11perf -copywin is about as simple as you can get and unlikely to trigger the bug. Within cairo you can find cairo-perf-trace and some firefox traces at http://cgit.freedesktop.org/cairo-traces which hopefully reproduce the bug. Otherwise, hopefully we can capture the hang with cairo-trace firefox.
I tried several traces, but they didn't cause a hang. I also captured Firefox traces when a hang happened, but replaying those didn't hang the system either. On the bright side, I gathered more dumps in the meantime, and even better, with 2.6.39-rc2 the system hangs immediately at X start. With luck it's the same bug. And my brightness fix didn't go upstream, so backlight brightness is still wonky sometimes. I must admit it's getting very tiresome to keep my crappy GPU functioning without regressions. :-( I think I'll try either an old intel X driver, or the git version, when I can muster up some motivation.
This issue is affecting a hardware component which is not being actively worked on anymore. Moving the assignee to the dri-devel list as contact, to give this issue a better coverage.
It seems to be fixed somewhere along the way.
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.