Bug 103067

Summary: GPU hang and (sometimes) reset on 4.13
Product: DRI Reporter: Nikola <nikolam3244>
Component: DRM/IntelAssignee: Intel GFX Bugs mailing list <intel-gfx-bugs>
Status: CLOSED DUPLICATE QA Contact: Intel GFX Bugs mailing list <intel-gfx-bugs>
Severity: normal    
Priority: medium CC: intel-gfx-bugs
Version: unspecified   
Hardware: x86-64 (AMD64)   
OS: Linux (All)   
Whiteboard:
i915 platform: i915 features:
Attachments:
Description Flags
Output of /sys/class/drm/card0/error
none
Dmesg output none

Description Nikola 2017-10-02 22:57:01 UTC
Created attachment 134628 [details]
Output of /sys/class/drm/card0/error

I updated Arch Linux' kernel to 4.13.3-1-ARCH and GUI was completely unusable.
I would boot up the system, log in, launch i3wm, open Chromium and system would just hang. Nothing would work, not even magic SysRq keys.

I am not sure that the issue is with Chromium, since I would open it up straight when I log in. Also, it doesn't matter what is open on it. Last time crashed just with one single "New Tab" - but it recovered (the only time it did recover, out of 4 boots).

Crash occurs just couple minutes after launching i3wm. 
Crash does not occur in CLI.

I have reverted to 4.12.13-1-ARCH, and crashing does not occur.

Error log is attached.
Comment 1 Nikola 2017-10-02 22:57:49 UTC
Created attachment 134629 [details]
Dmesg output
Comment 2 Nikola 2017-10-02 23:56:45 UTC
I just took a look at other recent issues, and #102912 sounds very similar to my 
issue. (https://bugs.freedesktop.org/show_bug.cgi?id=102912)

>If a second tty is opened before login, its possible to switch to tty2, than
>the GPU resets until the next freeze.
This would explain why GPU reset only on 4th boot - I had tty open

>If no other tty is started, only power-off works, no reboot, no magic keys.
Exactly.

I also saw Elizabeth's comment, about #103034#c2, and I will try that out soon.
(https://bugs.freedesktop.org/show_bug.cgi?id=103034#c2)
Comment 3 Nikola 2017-10-03 00:21:18 UTC
Yup, booting with intel_iommu=igfx_off seems to fix the issue.

*** This bug has been marked as a duplicate of bug 103034 ***

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.