Bug 14510 - DRI2 fails to unmap buffer
Summary: DRI2 fails to unmap buffer
Status: NEW
Alias: None
Product: DRI
Classification: Unclassified
Component: DRM/other (show other bugs)
Version: DRI git
Hardware: x86 (IA32) Linux (All)
: medium critical
Assignee: Default DRI bug account
QA Contact:
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2008-02-15 03:15 UTC by Johannes Engel
Modified: 2008-02-21 02:42 UTC (History)
0 users

See Also:
i915 platform:
i915 features:


Attachments
Xorg.0.log with debug enabled (36.80 KB, application/x-bzip)
2008-02-21 01:03 UTC, Johannes Engel
no flags Details
X.Org.log of the crash after starting compiz (77.02 KB, text/x-log)
2008-02-21 01:37 UTC, Johannes Engel
no flags Details

Note You need to log in before you can comment on or make changes to this bug.
Description Johannes Engel 2008-02-15 03:15:40 UTC
Starting glxgears or any application using DRI(2) it runs for about one second and after that fails:

DRM_I830_EXECBUFFER: -16
glxgears: intel_context.c:856: UNLOCK_HARDWARE: Assertion `intel->batch->cliprect_mode != REFERENCES_CLIPRECTS' failed.

dmesg says:
[drm:drm_buffer_object_validate] *ERROR* Timed out waiting for buffer unmap.
[drm:i915_validate_buffer_list] *ERROR* error on handle validate -16

I'm using Mesa, Xorg, drm (lib + kernel modules) from git of today.
intel i945GM with xf86-video-intel branch origin/intel-batchbuffer.
Kernel 2.6.24.2
Comment 1 Kristian Høgsberg 2008-02-15 06:00:27 UTC
Could you:

 - attach the log
 - try running the X server under gdb, set a break point on abort and the attach the backtrace when the assertion fails

Are you trying to run this in direct  rendering mode?  DRI2 is limited to indirect rendering for now (I'm going to post a heads up).
Comment 2 Johannes Engel 2008-02-18 05:27:22 UTC
I recompiled the X server following the howto from your blog. So at the moment the X server does not start anymore. I will investigate further and report my findings. :) Can we leave this one open for some days?
Comment 3 Kristian Høgsberg 2008-02-18 08:34:24 UTC
(In reply to comment #2)
> I recompiled the X server following the howto from your blog. So at the moment
> the X server does not start anymore. I will investigate further and report my
> findings. :) Can we leave this one open for some days?

No problem, let me know what you find.
Comment 4 Johannes Engel 2008-02-21 01:03:56 UTC
Created attachment 14477 [details]
Xorg.0.log with debug enabled

OK, now I got it compiled again. :) Here are my personal test results:

Using DRI2 everything is very slow, even scrolling without compiz.
The Xorg.0.log gets stiffed with messages "taking lock" and "releasing lock". Just after system start it already has 8 MB of this kind. As announced there is no direct rendering possible.
glxgears tells me something about 900 fps, but I do not at all see the gears rotating...
Comment 5 Johannes Engel 2008-02-21 01:37:11 UTC
Created attachment 14479 [details]
X.Org.log of the crash after starting compiz

Starting compiz crashes the X.Org server. I stripped the incredible amount of taking lock and releasing lock. :)
Comment 6 Johannes Engel 2008-02-21 02:42:38 UTC
News: The initially descsribed behaviour occurs only without using DRI2.
Running glxgears I get:

> glxgears
7215 frames in 5.0 seconds = 1442.847 FPS
7235 frames in 5.0 seconds = 1446.931 FPS
DRM_I830_EXECBUFFER: -16
glxgears: intel_context.c:860: UNLOCK_HARDWARE: Assertion `intel->batch->cliprect_mode != REFERENCES_CLIPRECTS' failed.
Abgebrochen

dmesg says:
[drm:drm_buffer_object_validate] *ERROR* Timed out waiting for buffer unmap.
[drm:i915_validate_buffer_list] *ERROR* error on handle validate -16


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.