Bug 94599 - Semi-random Broadwell-U GPU hangs
Summary: Semi-random Broadwell-U GPU hangs
Status: CLOSED WORKSFORME
Alias: None
Product: DRI
Classification: Unclassified
Component: DRM/Intel (show other bugs)
Version: unspecified
Hardware: x86-64 (AMD64) Linux (All)
: medium major
Assignee: Intel GFX Bugs mailing list
QA Contact: Intel GFX Bugs mailing list
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2016-03-17 19:07 UTC by Vadim A. Misbakh-Soloviov (mva)
Modified: 2017-07-24 23:16 UTC (History)
2 users (show)

See Also:
i915 platform: BDW
i915 features: display/LVDS


Attachments
GPU crash dump (3.26 MB, text/plain)
2016-03-17 19:09 UTC, Vadim A. Misbakh-Soloviov (mva)
no flags Details
dmesg cutoff (14.24 KB, text/plain)
2016-03-17 19:18 UTC, Vadim A. Misbakh-Soloviov (mva)
no flags Details

Description Vadim A. Misbakh-Soloviov (mva) 2016-03-17 19:07:17 UTC
Hi there! Probably I chose wrong i915 platform when created bug, so, if it is, change it to the right one, please.
I'm running i7-5700HQ APU with Broadwell-U (HD 5600, afair) GPU on board. Unfortunatelly, I was unable to setup userland (Xorg) part by just installing corresponding distro package with driver with SNA enabled, so I: 1) installed xorg-driver from yesterday git master, disabled SNA and enabled UXA (both on "./configure"-time, so I didn't touch xorg.conf at all and using Xorg autodetection), added "video.use_native_backlight=1 i915.semaphores=1 drm_kms_helper.poll=N i915.enable_rc6=1 i915.enable_fbc=0 intel_idle.max_cstate=1 acpi_osi=Linux i915.enable_execlists=0" to kernel commandline, and also "export hiz=false INTEL_HIZ=0 INTEL_SEPARATE_STENCIL=0 vblank_mode=0 LIBGL_ALWAYS_SOFTWARE=1;" to /etc/profile.
Without that changes I experienced system freezes (so even SysRq-magic didn't work) in few moments after enabling compositing effects in any of my DEs (KDE5, Enlightenment). After described changes (not sure, what exactly helped) all kinda started to work fine (including compositing), but from time to time, during work I still experience system freezes (although, thankfully, not all of them are total, so, say, now I was able to switch to tty console and writing that report in lynx).  
So, I opened dmesg | less and inspected what happened. And found there suggestion to report "_new_ bug", so here I am.

P.S. Sorry, I'm not sure how lynx will handle text I just entered if I press "add attachment", so I'll include /sys/class/drm/card0/error and dmesg cutoff on next update.

Additional info: Uname: Linux-4.5.0, xorg-server-1.18.2, mesa-11.1.2
Also, just before freezing I was in X11 terminal emulator (with enabled compositing (semitransparent background)) and edited a file in a vim. And there was only Firefox and Thunderbird processes run under in inactive windows.
P.P.S. I'll be happy to provide any additional info on request (although, I still can't figure out exactly conditions to 100% reproducibility)
Comment 1 Vadim A. Misbakh-Soloviov (mva) 2016-03-17 19:09:36 UTC
Created attachment 122387 [details]
GPU crash dump
Comment 2 Vadim A. Misbakh-Soloviov (mva) 2016-03-17 19:18:31 UTC
Created attachment 122388 [details]
dmesg cutoff
Comment 3 Jari Tahvanainen 2017-03-10 06:58:27 UTC
Vadim - sorry about this loooong delay until getting back to you. Assumable you have updated your kernel to newer. Do you still face this problem? If yes then please provide new dmesg and crash dump, like you did originally.
Comment 4 Jari Tahvanainen 2017-04-19 14:47:19 UTC
Timeout - marking resolved+worksforme


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.