Bug 26808

Summary:

[855] GPU hung (drm i915 intel) on Ubuntu 10.04 with drm-intel-next kernel

Product:

DRI

Reporter:

Geir Ove Myhr <gomyhr>

Component:

DRM/Intel

Assignee:

Chris Wilson <chris>

Status:

CLOSED FIXED

QA Contact:

Severity:

normal

Priority:

medium

CC:

arthapex, axet, geroxp, moikkis, orion

Version:

unspecified

Hardware:

x86 (IA32)

OS:

Linux (All)

Whiteboard:

i915 platform:

i915 features:

Attachments:

Description	Flags
i915_error_state	none
dmesg output	none
Xorg.0.log	none

Description Geir Ove Myhr 2010-02-28 13:47:06 UTC

Originally reported by Gero Mudersbach at:
  https://bugs.launchpad.net/bugs/528467

[Problem]
Another GPU hang bug on 855GM with captured i915_error_state. I am not good enough at reading the output of intel_error_decode to tell if this a cache flushing problem like bug 26746. 

[Original report]

Binary package hint: xserver-xorg-video-intel i915_

With the alpha 3 version of the Lucid Lynx live CD I get a GPU hung shortly after the splash screen of the the desktop (KDE) is shown. 

the following messages are from dmesg:
[drm:i915_hangcheck_elapsed] * ERROR * Hangcheck timer elapsed... GPU hung
render error detected , EIR: 0x00000000
i915: waking up sleeping processes
reboot required

The loginmanager then tries to respawn the xserver several times until there is a non blinking cursor shown at the botton of the screen. 

On some other boot tries with the live cd the xserver freezes completely and it is not possible to switch to another console.

Hardware info:
00:02.0 VGA compatible controller: Intel Corporation 82852/855GM Integrated Graphics Device (rev 02)
        Subsystem: ASUSTeK Computer Inc. Device 1712                                                
        Flags: bus master, fast devsel, latency 0, IRQ 16                                           
        Memory at f0000000 (32-bit, prefetchable) [size=128M]                                       
        Memory at feb80000 (32-bit, non-prefetchable) [size=512K]                                   
        I/O ports at dc00 [size=8]                                                                  
        Capabilities: [d0] Power Management version 1                                               
        Kernel driver in use: i915                                                                  
        Kernel modules: i915                                                                        

00:02.1 Display controller: Intel Corporation 82852/855GM Integrated Graphics Device (rev 02)
        Subsystem: ASUSTeK Computer Inc. Device 1712                                         
        Flags: bus master, fast devsel, latency 0                                            
        Memory at e8000000 (32-bit, prefetchable) [size=128M]                                
        Memory at fea80000 (32-bit, non-prefetchable) [size=512K]                            
        Capabilities: [d0] Power Management version 1

BTW: using "vesa" in xorg.conf and "nomodeset" at the kernel command line seems to be a workaround until the bug is fixed in the intel driver.

Comment 1 Geir Ove Myhr 2010-02-28 13:48:02 UTC

Created attachment 33652 [details]
i915_error_state

Comment 2 Geir Ove Myhr 2010-02-28 13:48:50 UTC

Created attachment 33653 [details]
dmesg output

Comment 3 Geir Ove Myhr 2010-02-28 13:49:39 UTC

Created attachment 33654 [details]
Xorg.0.log

Comment 4 Geir Ove Myhr 2010-02-28 13:51:39 UTC

Assigning to Chris Wilson, since he is into this kind of problems. Let me know if I shouldn't to this.

Comment 5 Chris Wilson 2010-03-02 11:59:13 UTC

It is not an obvious cache-flushing bug like the others, IPEHR matches the previous instruction. Nothing stands out as being the cause, if you can grab more i915_error_states that may help establish a pattern.

Comment 6 Bob McElrath 2010-05-24 15:22:05 UTC

There is also some display corruption in the upper right, and in some instances the display is completely corrupted.  Here is another i915_error_state from Ubuntu 10.04.

Time: 1274739074 s 997040 us
EIR: 0x00000000
  PGTBL_ER: 0x00000000
  INSTPM: 0x00000000
  IPEIR: 0x00000000
  IPEHR: 0xff17111b
  INSTDONE: 0xffe5fafd
  ACTHD: 0x00002e14
  INSTPS: 0x8011e032
  INSTDONE1: 0x000fffff

In one instance where X did not crash immediately, the Xorg.0.log says:

(WW) intel(0): i830_uxa_prepare_access: bo map failed

I'm downgrading to 9.10 now.  This is unusable.

Comment 7 Chris Wilson 2010-07-24 04:30:57 UTC

Gah, mishandling of tiling. We told the GPU that we were rendering to a tiled buffer, but did not allocate sufficient space for one. Old bug.

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.