Bug 14364 - [G33 915GM TTM] glean abort with intel_bufmgr_ttm.c:626: Error -16 waiting for fence fence buffers.
Summary: [G33 915GM TTM] glean abort with intel_bufmgr_ttm.c:626: Error -16 waiting fo...
Status: CLOSED FIXED
Alias: None
Product: Mesa
Classification: Unclassified
Component: Drivers/DRI/i915 (show other bugs)
Version: unspecified
Hardware: Other Linux (All)
: medium normal
Assignee: Thomas Hellström
QA Contact:
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2008-02-03 19:08 UTC by Shuang He
Modified: 2008-02-06 07:21 UTC (History)
2 users (show)

See Also:
i915 platform:
i915 features:


Attachments
xorg conf (3.92 KB, text/plain)
2008-02-03 19:08 UTC, Shuang He
Details
xorg log (20.82 KB, text/plain)
2008-02-03 19:09 UTC, Shuang He
Details
Patch to re-report last breadcrumb on a poll. (2.46 KB, patch)
2008-02-04 03:25 UTC, Thomas Hellström
Details | Splinter Review

Description Shuang He 2008-02-03 19:08:25 UTC
Created attachment 14130 [details]
xorg conf

System Environment:
--------------------------
--Platform: FC6
--Xf86_video_intel:be1b568bc2379856c3eaaed365002512bebc218c
--Mesa:30c9d96ebf038864485139064c47aeb0d0319598 
--DRM:733e07663e50087ca1e9af8e9b5def556521e3b5
--Kernel:2.6.22

Bug detailed description:
--------------------------
running glean abort with "intel_bufmgr_ttm.c:626: Error -16 waiting for fence fence buffers".and it seems it's not abort at the same case every time.


Reproduce steps:
----------------
1. start X
2. ./glean -r log 


Current result:
----------------
glean abort with "intel_bufmgr_ttm.c:626: Error -16 waiting for fence fence buffers".


Expected result:
----------------
glean should run without problem
Comment 1 Shuang He 2008-02-03 19:09:55 UTC
Created attachment 14131 [details]
xorg log
Comment 2 Shuang He 2008-02-03 23:04:41 UTC
after some bisect work, following drm commit seems to bring this issue in:

commit f1edb7ad91d8b92057ffa02eb162e3740d05a147
Author: Thomas Hellstrom <thomas-at-tungstengraphics-dot-com>
Date:   Wed Jan 30 22:06:02 2008 +0100

    Simplify the fencing code and differentiate between flushes and
    waiting types.
    Add a "command_stream_barrier" method to the bo driver.
Comment 3 Thomas Hellström 2008-02-04 03:25:29 UTC
Created attachment 14135 [details] [review]
Patch to re-report last breadcrumb on a poll.

Does this patch fix the problem?
If not, what is the related "dmesg" output?


/Thomas
Comment 4 Shuang He 2008-02-04 17:51:06 UTC
(In reply to comment #3)
> Created an attachment (id=14135) [details]
> Patch to re-report last breadcrumb on a poll.
> 
> Does this patch fix the problem?
> If not, what is the related "dmesg" output?
> 
> 
> /Thomas
> 

great, this patch seems working for me.
Comment 5 Thomas Hellström 2008-02-05 01:46:31 UTC
I've commited a variant of that patch and I'm marking the bug as resolved.
Please reopen if the problem persists.

The problem appears to be that the breadcrumb sequence turns up in the status page before the corresponding fence is actually queued on the fence ring.

/Thomas
Comment 6 Shuang He 2008-02-05 19:08:38 UTC
verified, thanks
Comment 7 Thomas Hellström 2008-02-06 07:21:20 UTC
Thanks. 
Closing.


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.