i915_wait_irq error: system hangs while running heavy graphics applications

Bug #48932 reported by Stephen Irons
8
Affects Status Importance Assigned to Milestone
linux-source-2.6.15 (Ubuntu)
Fix Released
Undecided
Unassigned

Bug Description

While running Celestia, Flightgear and TORCS, my system occasionally hangs. The sound stop, mouse and keyboard freeze and the bottom section of the display goes stripey.

After a few seconds, the screen goes blank and the X server cross-shaped cursor appears, shortly before the Ubuntu rotating disk 'wait' cursor. It looks as though GDM is trying to restart. However, the keyboard and mouse are still frozen. It is not possible to activate a virtual terminal using the keyboard.

The system does still respond to the power key and turns off after a few seconds.

After restarting, syslog reports the following:

Jun 2 20:33:52 localhost -- MARK --
Jun 2 20:53:52 localhost -- MARK --
Jun 2 21:01:48 localhost kernel: [4297572.942000] [drm:i915_wait_irq] *ERROR* i915_wait_irq: EBUSY -- rec: 413596 emitted: 413599
Jun 2 21:01:52 localhost gdm[4894]: Error reinitilizing server
Jun 2 21:01:53 localhost shutdown[8424]: shutting down for system halt
Jun 2 21:02:37 localhost syslogd 1.4.1#17ubuntu7: restart.

Looking at a number of these reports, there is no consistency between what is logged just before the *ERROR* i915_wait_irq message. The time of the previous message is also variable: between 15 seconds and 3.5 minutes.

Hardware:
Dell Inspiron 6400
Core Duo 1.6 GHz
1G RAM
Intel 945GM integrated graphics

Software:
Ubuntu Dapper installed from Beta 2 install (now called alternate) CD and updated to final alternate CD and archives on 2006-06-01.
Kernel: 2.6.15-23-696-smp
xserver-xorg-driver-i810: 1:1.4.1.3-0ubuntu6
915resolution: 0.5-1ubuntu6

While running, RAM is sitting at about 400M occupied. According to the CPU meters, one CPU is ticking over at 1GHz, the other at full speed (1.6GHz). According to system monitor application, this CPU is about 80% loaded, but the update rate is slow so I have no idea what the loading is like in the last few seconds before the crash.

The crash occurs even when there are no other application open; however, I have not yet tried to reduce the number of daemons.

Revision history for this message
Hannes Hauswedell (hannes-hauswedell) wrote :

I can confirm this bug.
I have two up-to-date dapper laptops, a Dell Inspiron 640m and a Dell Inspiron 1300. My experiences:
When i start a full-screen OpenGl application , like Quake3 the screen first remains black, but pressing Ctrl+Esc redraws the screen, showing the menu. everything works fine then.
When i start Warcraft3 in Wine, i sometimes have the same situation (menu is black, but if i manage to start a game via keyboard-shortcuts everything works ok). most of the times though, the screen becomes black, the sound plays, then stops, i cannot use keyboard shortcuts , the application, the x-server and everathing with it hangs. i somtimes see the mouse in a little area in the center of the screen. the rest stays black, but redraws once in while (the resolution seems to change). i have no way of getting out of this, neither Ctrl+Alt+Backspace nor Ctrl+Alt+F* work.
BUT i am able to login in via ssh and to perform normal actions like reboot. (note that killing or attempting to restart X does not work)
via ssh i found out that one message is logged:

[17183271.528000] [drm:i915_wait_irq] *ERROR* i915_wait_irq: EBUSY -- rec: 736021 emitted: 736024

so i am pretty sure its the same issue.

Revision history for this message
Hannes Hauswedell (hannes-hauswedell) wrote :

i wanted to add that trying to run an opengl application in wine always causes this error on the inspiron1300, whereas on the 640m they only happen 'most' of the time.

this really is an annoying issue, i would be glad if someone looked into it.
if more information is required please let me know.

Revision history for this message
Hannes Hauswedell (hannes-hauswedell) wrote :

after having updated
kernel-image to 2.6.15-26,
xorg-driver-i810 to 1.6.0-0ubuntu3 and
xorg-core to 1.0.2-0ubuntu10.1

the described problems are gone now :)

the original poster should check this, then the bug can be closed.

Changed in linux-source-2.6.15:
status: Unconfirmed → Fix Released
Revision history for this message
Stephen Irons (stephen-irons) wrote :

I have updated my whole system to the latest on the official repositories:

* linux-image-2.6.15-26-686 version 2.6.15-26.46
* xserver-xorg-core version 1:1.0.2-0ubuntu10.1

and downloaded from ubuntu.compiz.net
* xserver-xorg-driver-i810 version 1:1.6.0-0ubuntu3

as suggested.

* Celestia runs extremely slowly: Jupiter rotates very jerkily, and moving with mouse or cursor takes about 1s before it updates. When I run celestia-gnome from the command line, it reports

ERROR! sizeof(I830DRIRec) does not match passed size from device driver
libGL warning: 3D driver returned no fbconfigs.
libGL error: InitDriver failed
libGL error: reverting to (slow) indirect rendering
render path: 1

* Torcs starts OK, but crashes before you see any of the race track. It reports much the same:

Visual Properties Report
------------------------
Compatibility mode, properties unknown.

ERROR! sizeof(I830DRIRec) does not match passed size from device driver
libGL warning: 3D driver returned no fbconfigs.
libGL error: InitDriver failed
libGL error: reverting to (slow) indirect rendering
freeglut (/usr/lib/torcs/torcs-bin): Unable to create direct context rendering for window '/usr/lib/torcs/torcs-bin'
This may hurt performance.

and, after the crash

WARNING: ssgLoadTexture: Cannot determine file type for './(null)'
OpenAL backend info:
  Vendor: OpenAL Community
  Renderer: Software
  Version: 1.1
  Available sources: 1024 or more
  Available buffers: 1024 or more
  Dynamic Sources: requested: 1003, created: 1003
  #static sources: 21
  #dyn sources : 1003
torcs-bin: indirect_vertex_array.c:1359: __indirect_glTexCoordPointer: Assertion `a != ((void *)0)' failed.
/usr/games/torcs: line 52: 6236 Aborted $LIBDIR/torcs-bin -l $LOCAL_CONF -L $LIBDIR -D $DATADIR $*

The terrifying
ERROR! sizeof(I830DRIRec) does not match passed size from device driver

has been reported
http://lists.debian.org/debian-x/2006/05/msg01108.html

So, the problem as originally described no longer happens :) but Celestia still does not work :(

I do not know enough about the bug management policies in this forum to decide whether this issue should be closed or if I should raise a new issue, or if I should just wait for the debian fix to filter down into ubuntu. Any suggestions?

Revision history for this message
Stephen Irons (stephen-irons) wrote :

I see that the official ubuntu repositories have libgl1-mesa version 6.4.1-0ubuntu8 and libglu1-mesa version 6.4.1-0ubuntu8, whereas

http://lists.debian.org/debian-x/2006/05/msg01108.html

talks about mesa 6.5.something. I will wait for this to filter down into ubuntu.

Revision history for this message
dpchem (peryshkov) wrote :

This bug is still here in Ubuntu 7.10 (on Thinkpad T61) !

Intel X3100 graphic card
Core Duo 2 GHz
Ubuntu 7.10
xserver-xorg-video-intel 2:2.1.1-0ubuntu9

It is even triggered by gnome-screensaver.

I have

[drm:i915_wait_irq] *ERROR* i915_wait_irq: EBUSY -- rec: 1810686 emitted: 1810707

in syslog

After that gdm tries to restart but hangs with a mouse cursor in the middle of the blank screen :-)

Revision history for this message
Rubin (rubin) wrote :

I can confirm this bug still on Thinkpad T61 with current 7.10 (12/4/2007)

To reproduce: open Scorched 3d
click "change game settings"
turn resolution to 1280x768
click troubleshooting/detail
set texture size and tank detail and effects all to large, max, high
click ok
click start a single or multi-player game
click 'apoc'
click "start a target practice apocalypse mod game'
chose tank etc
in game, adjust power to 500 or so, and select doomsday device as a weapon
fire
wait for x to crash

In 7.04, simple things like glxgears would crash it.. now it takes more, so it seems things are improving but the problem is not solved slomehow?

I have had it crash often in secondlife, also. A large window seems to increase the chances, but it also depends on the content.

Revision history for this message
Michael Gratton (mjog) wrote :

Yes, this bug is still happening, albeit occasionally - it just happened on my X61s, Ubuntu 7.10 with Compiz as the WM and while the Lament OpenGL screensaver was running.

Revision history for this message
Michael Gratton (mjog) wrote :

Possibly related to i915 xorg driver, see: http://<email address hidden>/msg29741.html

Revision history for this message
João Pinto (joaopinto) wrote :

Could someone check bug #248505 happening on Hardy is related to this bug ?
Thanks

Revision history for this message
rickyrockrat (rickyrockrat) wrote :

I can confirm this bug on both the intel 82G35 running the latest Hardy (8.04.01) and running kernel 2.6.27-rc7 and GM965 running Gutsy and kernel (2.3.22,14-generic).

This happens consistently with Blender and OpenGL, and yes, this is the same bug as 248505 on Hardy

Revision history for this message
rickyrockrat (rickyrockrat) wrote :

Please see 248505 for fix for this bug for Hardy.

To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.