Originally reported by Irihapeti at: https://bugs.launchpad.net/bugs/477256 [Problem] This GPU hang only happens every few hours during normal use with xfce with desktop effects enabled. It doesn't seem to happen without desktop effects or with openbox or lxde, but since it doesn't happen very often, it is hard to say for sure. Irihapeti can trigger it more often by using a script to stress the video system. One of the two attached batchbuffer dumps is with default Lucid kernel (2.6.32) and the other with 2.6.33-rc5. Both are with recent xorg-edgers which give recent git versions of the xorg, driver, mesa, etc. Irihapeti found that freezes started after the kernel commit [e517a5e97080bbe52857bd0d7df9b66602d53c4d] agp/intel: Fix the pre-9xx chipset flush. (in the 2.6.31 stable tree with intel driver 2.9.0) but since the kernel and driver has progressed quite a bit since then, this freeze may be of a different nature. [Original bug report from Ubuntu Karmic (2.6.31 kernel and driver 2.9.0)] Binary package hint: xserver-xorg-video-intel After installing Xubuntu Karmic on this laptop, I'm having regular Xserver lockups, as in once per session. The mouse pointer still moves, but clicking does nothing and the keyboard is disabled. I can use the Alt-SysRq-REISUB commands, but the reboot that follows doesn't complete. The only way to clear this is by a complete power-down. The time before the freeze occurs varies but would usually be less than 15 minutes. I have tried using the standard Xserver-xorg-video-intel with DRI disabled, and a retro version of the same package, but the problem persists. Also tried adding "nomodeset" to the kernel command line when booting, which likewise made no difference. Hardy has run without issue on this machine, and as far as I can tell, Jaunty also. (I ran it for an hour or more from the live CD without problems.) The batchbuffer dump archive is attached. I reinstalled the standard Xserver-xorg-video-intel package for this test. PS: This was spun off from Launchpad bug 447892, since it doesn't seem to be the same freeze.
Created attachment 32911 [details] batchbuffer dump with 2.6.33-rc5
Created attachment 32912 [details] batchbuffer dump with kernel 2.6.32
Created attachment 32913 [details] dmesg output
Created attachment 32914 [details] Xorg.0.log
This looks like the CPU/GPU coherency issue with a twist. Both gpu dumps have recorded PGTBL_ERR=0x49, which is unusual. Those seem to be the unaligned fb issue, but there are several other preceding oopses in the dmesg as all. I'd like to see this bug confirmed on a recent kernel with i915_error_state capturing the batch buffer before declaring it to indeed be the same bug.
Created attachment 33988 [details] i915_error_state from intel-drm-next kernel This one has PGTBL_ER: 0x00000000, so it may be different from the rest. Also, IPEHR doesn't match the instruction before the ACTHD address, so I guess this is another GPU/CPU incoherency hang.
Created attachment 33989 [details] dmesg output from when i915_error_state was captured
Created attachment 33990 [details] Xorg.0.log from when i915_error_state was captured
(In reply to comment #5) > This looks like the CPU/GPU coherency issue with a twist. Both gpu dumps have > recorded PGTBL_ERR=0x49, which is unusual. We have several automatic bug reports in ubuntu with PGTBL_ER=0x49, so it shouldn't be too unusual. https://bugs.launchpad.net/ubuntu/+source/xserver-xorg-video-intel/+bug/534108 https://bugs.launchpad.net/ubuntu/+source/xserver-xorg-video-intel/+bug/534051 https://bugs.launchpad.net/ubuntu/+source/xserver-xorg-video-intel/+bug/534002 https://bugs.launchpad.net/ubuntu/+source/xserver-xorg-video-intel/+bug/533976 The first three are on i845 with 2.6.32 drm (which is what we have most bug reports from) and the last one is i855 on 2.6.31. I haven't seen it in on 2.6.33 drm or drm-intel-next yet.
Another look: 0x077ac270: 0x7d8e0001: 3DSTATE_BUFFER_INFO 0x077ac274: 0x03800c40: color, tiling = fenced, pitch=3136 0x077ac278: 0x07af6000: address Old tiling bug.
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.