System Environment: -------------------------- Platform: SKL ChromiumOS Libdrm: libdrm-2.4.61-13-g97be70b45eccc37e98a1cecf360593f36956ea42 Mesa: (master)4fd42a7c2798d03476c84b79cb855984a15c222c Kernel: 4.1.0-rc5_drm-intel-nightly_5267b9_20150527+ Bug detailed description: ----------------------------- Run WebGL Demo Dynamic Cubemap by SKL chromium for several minutes, then system hang is occured. This bug is occured by SKL SPT-H C0 stepping, with Bios B084.R00. But can not reproduce by SKL PCH-LP B0 stepping, with Bios V85.R00. Reproduce steps: ---------------------------- Open Chromium browser and run WebGL demo Dynamic Cubemap . chromium-browser --enable-webgl --ignore-gpu-blacklist --user-data-dir "http://webglsamples.googlecode.com/hg/dynamic-cubemap/dynamic-cubemap.html"
For PCIID: Can be reproduce by VGA compatible controller: Intel Corporation Device 191b (rev 04) can not reproduced by VGA compatible controller: Intel Corporation Device 191e (rev 02)
Do you get messages on console just before the hang? Could you setup netconsole and get a full log for the issue?
There is no special message just by hang. Below is the detail: Command: root@x-sklh01:/# chromium-browser --enable-webgl --ignore-gpu-blacklist --user-data-dir "http://webglsamples.googlecode.com/hg/dynamic-cubemap/dynamic-cubemap.html" ** (update-notifier:6660): WARNING **: not starting for system user ATTENTION: default value of option force_s3tc_enable overridden by environment. SSH is also hang, and can not connect to the test machine after system hang.
(In reply to shuo.wang from comment #1) > For PCIID: > > Can be reproduce by VGA compatible controller: Intel Corporation Device 191b > (rev 04) > > can not reproduced by VGA compatible controller: Intel Corporation Device > 191e (rev 02) Could you please test if ./gem_concurrent_blit --run-subtest gtt-blt-overwrite-source-read-rcs-forked ..fails with 191b and works with 191e? Thanks
I just tried it by 191b, it is passed. root@x-sklh01:/GFX/Test/Intel_gpu_tools/intel-gpu-tools/tests# ./gem_concurrent_blit --run-subtest gtt-blt-overwrite-source-read-rcs-forked IGT-Version: 1.10-g0d16473 (x86_64) (Linux: 4.1.0-rc5_drm-intel-nightly_5267b9_20150527+ x86_64) using 2x512 buffers, each 1MiB Subtest gtt-blt-overwrite-source-read-rcs-forked: SUCCESS (1.576s)
Can you please disable GPU reset (i915.reset=0). If it doesn't hard hang with that, please attach the error state. If it still hard hangs, please let us know.
OK, I will try it tomorrow since the machine is being used for testing now.
Created attachment 116389 [details] i915_error_state of WebGL after disable GPU reset After disable GPU reset with set i915.reset=0, during running WebGL demo, hard hang is not occured, ssh still work, but keyboard does not work any more.
Thank you Ben, hard hang is not occured any more. (In reply to Ben Widawsky from comment #6) > Can you please disable GPU reset (i915.reset=0). If it doesn't hard hang > with that, please attach the error state. > > If it still hard hangs, please let us know.
Created attachment 116390 [details] dmesg after GPU reset is disable
Moving to Mesa. Bug 90854 tracks the system hang in the kernel.
Agrees to track GPU hang as Mesa bug here, and track system hang as Kernel bug at bug#90854.
*** This bug has been marked as a duplicate of bug 90685 ***
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.