Created attachment 40948 [details] dmesg file System Environment: -------------------------- Arch: x86_64 Platform: piketon Libdrm: (master)2.4.22-21-g537703fd4805e9cd352965fce642670986822d22 Mesa: (master)c64447f47de168e82086bc2fb483817b82e59cab Xserver: (server-1.9-branch)xorg-server-1.9.2.902 Xf86_video_intel: (master)2.13.901-25-g9b967807c2d240488a715509649663aac3583532 Kernel: (drm-intel-fixes) 1b39d6f37622f1da70aa2cfd38bfff9a52c13e05 Bug detailed description: ------------------------- Run "x11perf -rgb10text" 2 or 3 times, GPU hangs at first, and then system hangs. This issues happens on piketon, works fine on pineview and huronriver. It's Kernel regression. Reproduce steps: ---------------- 1. xinit& 2. x11perf -rgb10text
Sure it's a kernel regression? The ddx is completely different for pnv vs ilk vs snb and has seen a large flux of patches in the last week. Whereas the only real difference for ilk kernel side is the use of PIPE_CONTROL... So it would be interesting to check behaviour on drm-intel-next.
Sure it's kernel regression. The last known good commit is Kernel: (drm-intel-next)5aa7d52aebfc11760bbc5b081ed621227bb77981. BTW, It works fine on drm-intel-fixes.
Sorry, I am just a little confused. Can you confirm you did mean: Broken: (drm-intel-fixes)1b39d6f37622f1da70aa2cfd38bfff9a52c13e05 Working: (drm-intel-next)5aa7d52aebfc11760bbc5b081ed621227bb77981 Thanks.
Found one bug that is definitely related. Pushed to drm-intel-staging: commit 8c0a6bfef165ccdbf5d73afb9dd660107b0c98d5 Author: Chris Wilson <chris@chris-wilson.co.uk> Date: Thu Dec 9 12:56:37 2010 +0000 drm/i915/ringbuffer: Handle wrapping of the autoreported HEAD If the tail advances beyond the autoreport HEAD value, then we need to fallback to an uncached read of the HEAD register in order to ascertain the correct amount of remaining space in the ringbuffer. Reported-by: Fang, Xun <xunx.fang@intel.com> Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=32259 Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> The reason why that would only affect ilk vs pnv/snb is the disparity between CPU and GPU speeds (plus the extra synchronization overheads on SNB -fixes) and that with the PIPE_CONTROL workaround, ilk can only fit a (relatively) small number of requests.
Sorry, the broken should be Kernel:(drm-intel-next) c57802706aad9f179f2219415717896b3c177845. Thanks for you quickly fix it. Verified with Kernel:(drm-intel-next)8c0a6bfef165ccdbf5d73afb9dd660107b0c98d5.
Closing old verified.
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.