Bug 109948 - [945G] GPU hang
Summary: [945G] GPU hang
Alias: None
Product: DRI
Classification: Unclassified
Component: DRM/Intel (show other bugs)
Version: unspecified
Hardware: x86-64 (AMD64) Linux (All)
: medium minor
Assignee: Intel GFX Bugs mailing list
QA Contact: Intel GFX Bugs mailing list
Whiteboard: Triaged, ReadyForDev
Depends on:
Reported: 2019-03-09 19:14 UTC by astrosnail
Modified: 2019-07-23 08:54 UTC (History)
1 user (show)

See Also:
i915 platform: I945G
i915 features: GPU hang

Output of dmesg describing the hang (18.41 KB, application/gzip)
2019-03-09 19:14 UTC, astrosnail
no flags Details
The crash dump from /sys/class/drm/card0/error (6.16 KB, application/gzip)
2019-03-09 19:16 UTC, astrosnail
no flags Details
Screenshot at the time of the hang (308.53 KB, image/png)
2019-03-09 19:17 UTC, astrosnail
no flags Details

Note You need to log in before you can comment on or make changes to this bug.
Description astrosnail 2019-03-09 19:14:45 UTC
Created attachment 143599 [details]
Output of dmesg describing the hang

I was watching some video, when suddenly the display froze, went black, and returned with artifacts.
Examining the output of `dmesg` (attached) revealed that my GPU apparently hanged and was reset.
I have *not* been able to reproduce the hang.

The distribution is Arch Linux.
The motherboard is an Asus P5GC-MX/1333, and the output is VGA.
The chipset is 945G or similar (the information I'm getting is inconsistent, between 945G, GC, or GZ)

`uname -a`:
Linux box 5.0.0-arch1-1-ARCH #1 SMP PREEMPT Mon Mar 4 14:11:43 UTC 2019 x86_64 GNU/Linux

`pkg-config --modversion libdrm`:

`glxinfo | grep '^OpenGL version'`:
OpenGL version string: 2.1 Mesa 18.3.4

`pacman -Q xf86-video-intel`:
xf86-video-intel 1:2.99.917+860+g3a2dec17-1

The crash dump and the screenshot of resulting artifacts will be attached soon.
Comment 1 astrosnail 2019-03-09 19:16:26 UTC
Created attachment 143600 [details]
The crash dump from /sys/class/drm/card0/error
Comment 2 astrosnail 2019-03-09 19:17:25 UTC
Created attachment 143601 [details]
Screenshot at the time of the hang
Comment 3 Chris Wilson 2019-03-09 21:13:29 UTC
0x0080c06c:      0x7d060014: 3DSTATE_PIXEL_SHADER_CONSTANTS
0x0080c074:      0xbd800000:    C0.X = -0.062500
0x0080c078:      0xbf000000:    C0.Y = -0.500000
0x0080c07c:      0xbf000000:    C0.Z = -0.500000
0x0080c080:      0x00000000:    C0.W = 0.000000
0x0080c084:      0x3f9507c8:    C1.X = 1.164300
0x0080c088:      0x00000000:    C1.Y = 0.000000
0x0080c08c:      0x3fcc432d:    C1.Z = 1.595800
0x0080c090:      0x00000000:    C1.W = 0.000000
0x0080c094:      0x3f9507c8:    C2.X = 1.164300
0x0080c098:      0xbec890d6:    C2.Y = -0.391730
0x0080c09c:      0xbf501a37:    C2.Z = -0.812900
0x0080c0a0:      0x00000000:    C2.W = 0.000000
0x0080c0a4:      0x3f9507c8:    C3.X = 1.164300
0x0080c0a8:      0x40011687:    C3.Y = 2.017000
0x0080c0ac:      0x00000000:    C3.Z = 0.000000
0x0080c0b0:      0x00000000:    C3.W = 0.000000
0x0080c0b4:      0x00000000:    C4.X = 0.000000
0x0080c0b8:      0x00000000:    C4.Y = 0.000000
0x0080c0bc:      0x00000000:    C4.Z = 0.000000
0x0080c0c0:      0x00000000:    C4.W = 0.000000

  ACTHD: 0x00000000 0080c0a0
  IPEHR: 0xbf501a37

So it tried to execute data, which implies it misparsed the 3DSTATE_PIXEL_SHADER_CONSTANTS command. I wonder if there's a cacheline restriction there or something? The batch buffer doesn't look incoherent or damage, so looks more likely it died in the CS.
Comment 4 Lakshmi 2019-07-19 10:13:07 UTC
Reporter, are you able to reproduce the issue anymore with drmtip? What is the current situation, is this issue still happening?
Comment 5 astrosnail 2019-07-19 11:43:08 UTC
I have not been able to reproduce the hang at all. No further issues seemed to have occurred since.

My guess is that this issue is isolated, and caused by an overclocked fsb.
Comment 6 Francesco Balestrieri 2019-07-23 08:54:27 UTC
Thanks for the update, closing based on your feedback.

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.