Bug 89536

Summary: GPU lockup and kernel OOPS when using PRIME
Product: DRI Reporter: Rafael Ristovski <rafael.ristovski>
Component: DRM/RadeonAssignee: Default DRI bug account <dri-devel>
Status: RESOLVED MOVED QA Contact:
Severity: critical    
Priority: high CC: rafael.ristovski
Version: XOrg git   
Hardware: x86-64 (AMD64)   
OS: Linux (All)   
Whiteboard:
i915 platform: i915 features:
Attachments:
Description Flags
Two of the kernel hangs, note one is an OOPS, while the other is not, I was able to shut down the machine normally after the first OOPS as the kernel was still responsive to power button events.
none
lspci output
none
Xorg log
none
dmesg, ends when login prompt appears (non-graphical)
none
Another crash, this time with running 4.0.0 kernel, which also gave me a trace.
none
Same crash, this time with 3.19.2-1 kernel. Notice the mention of i915 none

Description Rafael Ristovski 2015-03-11 16:44:26 UTC
Created attachment 114225 [details]
Two of the kernel hangs, note one is an OOPS, while the other is not, I was able to shut down the machine normally after the first OOPS as the kernel was still responsive to power button events.

When running OpenGL applications with PRIME, the screen freezes and a kernel OOPS is generated. This seems to happen either when the application starts, or the window of the application is re-drawn (moving workspaces, minimizing/maximizing) Crashes seem to be inconsistent as sometimes the kernel just freezes and no OOPS is generated.

OS Details:
Archlinux with kernel 3.19.1-1
xorg-server 1.17.1-3
xf86-video-intel 2.99.917-4
xf86-video-ati 1:7.5.0-2
libdrm 2.4.59-1

Below is an attached kernel log with _TWO_ separate kernel hang, both caused by running an OpenGL application with PRIME.
Comment 1 Alex Deucher 2015-03-11 16:54:20 UTC
Please attach your full dmesg from boot, xorg log, and lspci -nn output.
Comment 2 Rafael Ristovski 2015-03-11 17:02:30 UTC
Created attachment 114226 [details]
lspci output
Comment 3 Rafael Ristovski 2015-03-11 17:03:22 UTC
Created attachment 114227 [details]
Xorg log
Comment 4 Rafael Ristovski 2015-03-11 17:04:45 UTC
Created attachment 114228 [details]
dmesg, ends when login prompt appears (non-graphical)
Comment 5 Rafael Ristovski 2015-03-12 17:09:59 UTC
Created attachment 114265 [details]
Another crash, this time with running 4.0.0 kernel, which also gave me a trace.
Comment 6 Rafael Ristovski 2015-03-21 12:35:44 UTC
Created attachment 114503 [details]
Same crash, this time with 3.19.2-1 kernel. Notice the mention of i915
Comment 7 Rafael Ristovski 2015-03-21 13:24:04 UTC
I just downgraded, PRIME worked flawlessly up until 3.19, I was able to test up until 3.17.6-1 since the next one after that was the latest kernel in the [testing] repo aka 3.19.2-1
Comment 8 Michel Dänzer 2015-03-25 07:37:33 UTC
(In reply to Rafael Ristovski from comment #7)
> I just downgraded, PRIME worked flawlessly up until 3.19, I was able to test
> up until 3.17.6-1 since the next one after that was the latest kernel in the
> [testing] repo aka 3.19.2-1

Can you narrow it down further, ideally bisect?
Comment 9 Rafael Ristovski 2015-03-25 13:54:34 UTC
(In reply to Michel Dänzer from comment #8)
> (In reply to Rafael Ristovski from comment #7)
> > I just downgraded, PRIME worked flawlessly up until 3.19, I was able to test
> > up until 3.17.6-1 since the next one after that was the latest kernel in the
> > [testing] repo aka 3.19.2-1
> 
> Can you narrow it down further, ideally bisect?

I guess I could try, but not in the near future as I lack free time. Will report back once done or when any progress is made.
Comment 10 Martin Peres 2019-11-19 09:02:29 UTC
-- GitLab Migration Automatic Message --

This bug has been migrated to freedesktop.org's GitLab instance and has been closed from further activity.

You can subscribe and participate further through the new bug through this link to our GitLab instance: https://gitlab.freedesktop.org/drm/amd/issues/593.

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.