Bug 23692 - [945GM MacMini] Reproducible freeze with intel 2.8.1
Summary: [945GM MacMini] Reproducible freeze with intel 2.8.1
Alias: None
Product: xorg
Classification: Unclassified
Component: Driver/intel (show other bugs)
Version: unspecified
Hardware: x86 (IA32) Linux (All)
: medium major
Assignee: Wang Zhenyu
QA Contact: Xorg Project Team
Depends on:
Reported: 2009-09-03 11:21 UTC by Chris King
Modified: 2010-07-17 07:04 UTC (History)
0 users

See Also:
i915 platform:
i915 features:

all attachments (is there any other way to do more than one?) (328.42 KB, application/x-tar)
2009-09-03 11:21 UTC, Chris King
no flags Details

Description Chris King 2009-09-03 11:21:03 UTC
Created attachment 29187 [details]
all attachments (is there any other way to do more than one?)

My symptoms are similar to many others but I have not yet found one other report with this reproducibility, or my exact hardware, so I'm opening a new one in hopes of providing some very specific info.


X.org is "pristine", except that tiling must be disabled or the screen is garbled (Option "Tiling" "false").  (However the bug is reproducible in the same way even if the screen is garbled.)

X starts fine.  I use no display manager, and my window manager (wmii) uses no "special effects" (font is "fixed", drawing is done with Xlib).  I can also start a terminal (uxrvt, again, no "special effects", same font).

As soon as I start any program (Pidgin and Firefox have been my test cases) which uses "special effects", e.g. antialiased fonts, the X server freezes: graphics come to a halt (but remain on the screen), keyboard and mouse clicks are ignored, however the mouse cursor still moves (although nothing responds to it).  I cannot switch to another VT.

The rest of the OS still seems to run: I was able to grab debug output from after the crash using a script I had started in the background (I confirmed this by timestamping the script's output and comparing it afterward to the time which had frozen on the screen).  VT switching (again, with a script) does not seem to work.  I do not have an easy way to log in remotely to my machine or I would be able to test this more directly.


machine: Mac Mini Core Solo
graphics card: Intel Corporation Mobile 945GM/GMS, 943/940GML Express Integrated Graphics Controller (rev 03) (8086:27a2)
OS: mix of Debian testing (squeeze) and unstable (sid)
kernel: Linux 2.6.30 with Debian patches
X.org: 7.3
libdrm/libdrm-intel: 2.4.12
xserver-xorg-video-intel: 2.8.1 (though also froze with 2.8.0)
Mesa: 7.5
acceleration method: UXA (before 2.8, I used XAA with rare intermittent freezes)

Attached are three Intel GPU dumps (right after starting X, right before the freeze, and right after the freeze) -- I have also a fourth dump from several seconds after the freeze but it is identical to the dump from right after the freeze.

Also attached is my xorg.conf and my Xorg.0.log from the freeze.  /var/log/debug, kern.log, messages, and syslog have the following combined wisdom from before the freeze (nothing after):

Sep  3 13:31:30 mac kernel: [drm] Initialized drm 1.1.0 20060810
Sep  3 13:31:30 mac kernel: pci 0000:00:02.0: setting latency timer to 64
Sep  3 13:31:30 mac kernel: [drm] Initialized i915 1.6.0 20080730 for 0000:00:02.0 on minor 0

When I have time, I will try various combinations of disabling DRI, DRI2, Render, and acceleration.  From past trials I recall NoAccel having no effect, and some combination of the other options fixing the problem but resulting in unusably slow rendering (worse than VESA, which I am using now).

Is there any other information I can give to help?

Thanks for your help; I hope we can find a fix or workaround soon.
Comment 1 Wang Zhenyu 2009-09-08 23:47:15 UTC
Does this issue also exist when KMS is on? i915.modeset=1
Comment 2 Chris King 2009-09-09 16:28:41 UTC
Enabling KMS causes X to blank the display upon start.  Neither switching VTs nor Ctrl-Alt-Bksp restored the screen, so presumably either X, the kernel, or the video card locked up.
Comment 3 Chris King 2009-09-09 18:57:38 UTC
After testing various combinations of DRI, DRI2, Render, and NoAccel, I found:

* NoAccel is ignored by the intel driver.
* Disabling DRI (regardless of the other settings) fixes the problem, but results in extremely slow rendering (slower than with VESA).
* So long as DRI was enabled, the same freeze occured, with the following exceptions:
 - The first time I tested with both DRI2 and Render disabled, I could not trigger the freeze.  The second time I tested however the freeze triggered as normal.
 - Disabling Render seemed to cause strange artifacts to appear, most notably on the mouse cursor (which at times was replaced with a 64x64 block of pseudo-garbage).

Are there any other (perhaps undocumented) options which may make a difference which I should try?
Comment 4 Wang Zhenyu 2009-09-28 00:40:03 UTC
We have other reporter on that upstream components have no more freeze issue on 945G Mac, could you test them too? kernel, libdrm, mesa, xf86-video-intel.
Comment 5 Wang Zhenyu 2009-12-21 21:55:50 UTC
lower severity. Any testing news?
Comment 6 Chris Wilson 2010-07-17 07:04:24 UTC
Timeout, presuming fixed.

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.