Bug 31024 - [SNB] glean/readpixPerf and some olgc cases cause system hang
Summary: [SNB] glean/readpixPerf and some olgc cases cause system hang
Status: VERIFIED FIXED
Alias: None
Product: Mesa
Classification: Unclassified
Component: Drivers/DRI/i965 (show other bugs)
Version: unspecified
Hardware: Other Linux (All)
: highest blocker
Assignee: Wang Zhenyu
QA Contact:
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2010-10-21 01:38 UTC by Yi Sun
Modified: 2010-12-29 18:01 UTC (History)
1 user (show)

See Also:
i915 platform:
i915 features:


Attachments

Description Yi Sun 2010-10-21 01:38:02 UTC
Environment:
------------
platform: HuronRiver
kernel:    2.6.36-rc7
Libdrm:         (master)2.4.22-3-g09b1062628f2cbddb3ebae20e7b3b8a0a93acebf
Mesa:           (master)289900439f0f327910496f6bc362b95930eebb53
Xserver:                (master)xorg-server-1.9.0-177-gd738175eaf1098e29b8afb6de8e99b5098e366a7
Xf86_video_intel:               (master)2.12.902-23-ga1c54f69643671ce296c57d132852e9846cc41d3
Cairo:          (master)0d93468efc7f7337b63c0cd746d5185e14d345f1
Libva:          (master)6c372035373e531d2bb5608861adab3f6b2ff471


Bug detailed description:
--------------------------
The case readpixPerf is timeout. Before killing the case,the machine is able to be login using ssh. But so long as we killed the case, the system hang.
Comment 1 Yi Sun 2010-10-21 01:50:40 UTC
This issue doesn't reproduce on Piketon platform.
Comment 2 Yi Sun 2010-10-21 02:00:50 UTC
The case intel_gpu_tool/gem_pread_after_blit has the same issue.
Comment 3 Gordon Jin 2010-10-28 18:08:59 UTC
This impacts many cases and blocks QA's nightly testing. Eric/Zhenyu, who wanna take this?
Comment 4 Yi Sun 2010-11-09 23:28:53 UTC
The issue in intel_gpu_tool has fix. Now, so long  as run the case glean/readpixPerf on HuronRiver the system hang, but it works well on Calpella platform.
Comment 5 Yi Sun 2010-11-10 00:31:35 UTC
Some cases such as divzero.c zbfunc.c bexact.c logicop.c and drawpix.c have the similar issue in oglc test suite.
Comment 6 Eric Anholt 2010-12-01 18:24:31 UTC
On my system, "glean -r asdf -t readpixPerf -o" triggers a GPU reset and after a couple of flickers I ^C it and graphics is recovered.
Comment 7 Wang Zhenyu 2010-12-01 18:43:35 UTC
yeah, sandybridge reset function works fine. Please test with drm-intel-next. We could close this hang bug if no more hang with that.
Comment 8 Gordon Jin 2010-12-06 00:31:18 UTC
It still hangs on our Huron River with the latest drm-intel-next. 
We'll test the latest mesa.
Comment 9 Yi Sun 2010-12-07 00:53:18 UTC
The case readpixPerf has passed.
And two cases in Oglc test suite still cause system hang: case divzero.c and fpstate.c.
Comment 10 Gordon Jin 2010-12-07 00:57:54 UTC
should be related to this fix:

commit 2b1469340bbf910469449354eeb5c02a9acfedba
Author: Zhenyu Wang <zhenyuw@linux.intel.com>
Date:   Mon Dec 6 15:15:55 2010 +0800

i965: Fix GS state uploading on Sandybridge
        
Need to check the required primitive type for GS on Sandybridge,
and when GS is disabled, the new state has to be issued too, instead
of only updating URB state with no GS entry, that caused hang on
Sandybridge.
		    
This fixes hang issue during conformance suite testing.


Closing this bug, and track the rest 2 oglconform hangs in separate bugs.
Comment 11 Yi Sun 2010-12-29 18:01:02 UTC
The hang issue disappeared.


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.