System Environment: ---------------------------------------------------------------------- Platform: SNB (Fedora19) Kernel: 3.12.2 Libdrm: (master)libdrm-2.4.49 Mesa: (10.0)724c07ff1288cebda9a009ca5a72bae5b51e1062 Xserver:(server-1.14-branch)xorg-server-1.14.4 Xf86_video_intel: (master)2.99.906 Cairo:(master)8e11a42e3e9b679dce97ac45cd8b47322536a253 Libva:(master)88ed1ebe960b1c4a7970e12f8df1ed7d7031352a Libva_intel_driver:(master)4a0f76c5b706fccbc85fadaeee9d785cd7b57d5a Bug detailed description: ------------------------------------------------------------------------ Glxgears frozen with gnome-session when running by 50~90 circle on fedora19,It Works ok on Raw X, The problem doesn’t exist on upstream. Please see Xorg.0.log and dmesg.log. This problem reproduced probability is low in Q3, often reproduced in Q4. Xf86 performance -------------------------------------------------------------------- Test 100glxgears.sh on SNB git- 9a8478d :good git- 7468a6b :bad Reproduce steps: -------------------------------------------------------------------- 1. xinit& 2. gnome-session 3. declare -i K=1 while [ $K -le 100 ] do echo $K glxgears & K=$K+1 sleep 1 done killall glxgears
Created attachment 90492 [details] dmesg.log
Created attachment 90493 [details] Xorg.0.log
The problem exists only in f19, not upstream, why is it filed here?
Also can you please specify exactly which of the components need to be from downstream (fedora) and which from upstream to reproduce this? In your system environment only the kernel isn't latest git (and doesn't look like a distro kernel either).
(In reply to comment #4) > Also can you please specify exactly which of the components need to be from > downstream (fedora) and which from upstream to reproduce this? In your > system environment only the kernel isn't latest git (and doesn't look like a > distro kernel either). This problems is caused by Xf86_video_intel. As I know, the good commit is (master)git-9a8478d.
(In reply to comment #3) > The problem exists only in f19, not upstream, why is it filed here? Sorry,I confused you,the problem exists on all Linux.
Since this is a regression in userspace can you pls do the bisect?
(In reply to comment #7) > Since this is a regression in userspace can you pls do the bisect? Okay,by bisected, show that the first bad commit is: f3225fcb38686f3b9701725bf3a11ecf1c100c3f Author: Chris Wilson <chris@chris-wilson.co.uk> Date: Tue Nov 5 08:38:22 2013 +0000 sna: Be move conservative with tiling sizes for older fenced gen The older generations have stricter requirements for alignment of fenced GPU surfaces, so accommodate this by reducing our estimate available space for the temporary tile. Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Bisect fail, you fell down the rabbit hole of chasing a different bug. To confirm yet again, you do not see the bug with the current ddx?
(In reply to comment #9) > Bisect fail, you fell down the rabbit hole of chasing a different bug. To > confirm yet again,you do not see the bug with the current ddx? In this stress test,glxgears loop running 100 times as a round. On 2013Q4,the frame of a glxgears would frozed when running to 30~50 time(only one round),but after a few seconds, then glxgears would work well again. Eerros in demsg:"[65598.315187] [drm:i915_set_reset_status] *ERROR* render ring hung inside bo(0x55e6000 ctx 1) at 0x55e6220" On 2013Q2,it's OK (I have tried 4 rounds,all passed). I'm sorry,the problem doesn't seem caused by xf86. I find that the probability of reproduing the bug is quit lower with old driver. It's difficult to find out which commpent.
Next time you run the stress tests, if it does trigger a hang again, please grab the error state and lets see what's responsible.
I have test it 5 times and 100 rounds per times,The issue doesn't exists on Mesa-10.2,so close it.
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.