Created attachment 117925 [details] crash dump saved to /sys/class/drm/card0/error GPU hang in either the steam interface or launching Dota 2. Also occurs in other GL applications (but not stellarium). Mesa version: 10.6.4 Linux version: 4.1.5 (Debian unstable linux-image-4.1.0-2-amd64 version 4.1.5-1) libdrm-intel1: 2.4.64 Additional info: This issue started to appear when upgrading from Mesa 10.3 to Mesa 10.6. May be the same bug as 89279, but I can't tell. Dmesg output: [ 485.530632] [drm] stuck on render ring [ 485.531158] [drm] GPU HANG: ecode 6:0:0x84fefffc, in dota_linux [1933], reason: Ring hung, action: reset [ 485.531160] [drm] GPU hangs can indicate a bug anywhere in the entire gfx stack, including userspace. [ 485.531161] [drm] Please file a _new_ bug report on bugs.freedesktop.org against DRI -> DRM/Intel [ 485.531162] [drm] drm/i915 developers can then reassign to the right component if it's not a kernel issue. [ 485.531163] [drm] The gpu crash dump is required to analyze gpu hangs, so please always attach it. [ 485.531164] [drm] GPU crash dump saved to /sys/class/drm/card0/error [ 485.533388] drm/i915: Resetting chip after gpu hang [ 491.515238] [drm] stuck on render ring [ 491.515754] [drm] GPU HANG: ecode 6:0:0x84fefffc, in dota_linux [1933], reason: Ring hung, action: reset [ 491.517139] drm/i915: Resetting chip after gpu hang [ 498.499306] [drm] stuck on render ring [ 498.499779] [drm] GPU HANG: ecode 6:0:0x84fefffc, in steam [1662], reason: Ring hung, action: reset [ 498.501864] drm/i915: Resetting chip after gpu hang [ 504.507969] [drm] stuck on render ring [ 504.508712] [drm] GPU HANG: ecode 6:0:0x95fefffc, in steam [1662], reason: Ring hung, action: reset [ 504.510830] drm/i915: Resetting chip after gpu hang [ 505.065428] CIPCServer::Thr[1686]: segfault at 0 ip 00000000efcdcfc4 sp 00000000eee38ab0 error 4 in steamclient.so[ef20b000+10e2000]
Hardware: i5-2557M
Bug https://bugs.freedesktop.org/show_bug.cgi?id=91650 seems similar.
Still occurs with: Kernel: 4.1.7 Mesa: 10.6.7
The problem has disappeared when testing configuration: Kernel: 4.1.8 Mesa: 10.6.8 I have to suspect the kernel upgrade made the difference. Clueless guess is that the following patch in 4.1.8 fixed it: drm/i915: Always mark the object as dirty when used by the GPU commit 51bc140431e233284660b1d22c47dec9ecdb521e upstream.
Thanks.
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.