Bug 75141

Summary: [SNB/IVB/HSW/BYT/BDW Bisected]igt/gem_flink_race/flink_close fails
Product: DRI Reporter: lu hua <huax.lu>
Component: DRM/IntelAssignee: Paulo Zanoni <przanoni>
Status: CLOSED WORKSFORME QA Contact: Intel GFX Bugs mailing list <intel-gfx-bugs>
Severity: major    
Priority: high CC: ben, intel-gfx-bugs
Version: unspecified   
Hardware: All   
OS: Linux (All)   
Whiteboard:
i915 platform: i915 features:
Attachments:
Description Flags
dmesg none

Description lu hua 2014-02-18 08:06:25 UTC
Created attachment 94264 [details]
dmesg

System Environment:
--------------------------
Platform: Sandybridge
kernel:   (drm-intel-nightly)1be8f2b4dd6d3db00af24d4891c82d2650bd282d

Bug detailed description:
---------------------------
Clean boot system, run ./gem_flink_race --run-subtest flink_close, It fails. Run the 2nd cycle, It passes.
It fails on Sandybridge with -nightly and -queued kernel.

The latest known good commit: c461562e84d180fb691af57f93a42bd9cc7eb69c
The latest known bad commit:  4c0e552882114d1edb588242d45035246ab078a0

IGT-Version: 1.5-g9597836 (i686) (Linux: 3.13.0_drm-intel-nightly_1be8f2_20140218+ i686)
leaked 1 objects
Test assertion failure function test_flink_close, file gem_flink_race.c:195:
Last errno: 2, No such file or directory
Failed assertion: obj_count == 0
Subtest flink_close: FAIL

Reproduce steps:
----------------------------
1. clean boot system
2. ./gem_flink_race --run-subtest flink_close
Comment 1 lu hua 2014-02-20 01:31:51 UTC
It also happens on Ivybridge, Haswell, Baytrail and Broadwell.
Comment 2 Shuang He 2014-02-28 14:34:46 UTC
Hello, I'm debuging the autobisect functionality in PRTS system, following are the bisect result I get from the system and I have manually verified that:
commit 8ebc02a54c22b7a83a34c923153861848183cd96
Author: Ben Widawsky 
Date: Fri Feb 21 13:15:35 2014 -0800

quick_dump: Add missed broadwell autodetect

Signed-off-by: Ben Widawsky
Comment 3 Daniel Vetter 2014-03-03 07:23:03 UTC
This is strange indeed. Does the igt commit you (and ptrs) bisected to _really_ make the difference on all platforms?

From a quick look it shouldn't affect the testcase at all ...

Also please check that you don't have anything else running (like X or other testcases) on the relevant testbox. The testcase shouldn't spuriously fail like that with just 1 leaked object (if it fails it tends to leak more than that and also fail sometimes later on).
Comment 4 lu hua 2014-03-04 03:16:04 UTC
(In reply to comment #3)
> This is strange indeed. Does the igt commit you (and ptrs) bisected to
> _really_ make the difference on all platforms?
> 
> From a quick look it shouldn't affect the testcase at all ...
> 
> Also please check that you don't have anything else running (like X or other
> testcases) on the relevant testbox. The testcase shouldn't spuriously fail
> like that with just 1 leaked object (if it fails it tends to leak more than
> that and also fail sometimes later on).

Double check the bisect commit is right.
Comment 5 Daniel Vetter 2014-03-04 18:48:19 UTC
To triple check. Reverting the offending igt commit with

$ git revert 8ebc02a54c22b7a83a34c923153861848183cd96

does indeed fix the testcase, i.e.

# ./gem_flink_race --run-subtest flink_close

on all platforms you've reported this bug against?

If this is true the universe seriously hates us ...
Comment 6 lu hua 2014-03-05 06:57:59 UTC
revert 8ebc02a54c22b7a83a34c923153861848183cd96
#./gem_flink_race --run-subtest flink_close
output
IGT-Version: 1.5-g9397936 (i686) (Linux: 3.14.0-rc5_drm-intel-nightly_3d805d_20140305+ i686)
leaked 0 objects
Subtest flink_close: SUCCESS
Comment 7 Daniel Vetter 2014-03-14 16:02:01 UTC
(In reply to comment #6)
> revert 8ebc02a54c22b7a83a34c923153861848183cd96
> #./gem_flink_race --run-subtest flink_close
> output
> IGT-Version: 1.5-g9397936 (i686) (Linux:
> 3.14.0-rc5_drm-intel-nightly_3d805d_20140305+ i686)
> leaked 0 objects
> Subtest flink_close: SUCCESS

Can you please retest whether latest i-g-t works?
Comment 8 lu hua 2014-03-17 05:22:44 UTC
It still fails.
output:
IGT-Version: 1.6-g8fe0c50 (x86_64) (Linux: 3.14.0-rc6_drm-intel-nightly_b18596_20140314+ x86_64)
leaked 1 objects
Test assertion failure function test_flink_close, file gem_flink_race.c:195:
Last errno: 2, No such file or directory
Failed assertion: obj_count == 0
Subtest flink_close: FAIL
Comment 9 Ben Widawsky 2014-03-25 04:07:03 UTC
Is this failing on the latest -nightly? PPGTT has been disabled.
Comment 10 lu hua 2014-03-25 07:43:03 UTC
It still fails.
output:
IGT-Version: 1.6-g7a81094 (x86_64) (Linux: 3.14.0-rc7_drm-intel-nightly_842955_20140325+ x86_64)
leaked 1 objects
Test assertion failure function test_flink_close, file gem_flink_race.c:196:
Last errno: 2, No such file or directory
Failed assertion: obj_count == 0
Subtest flink_close: FAIL
Comment 11 Daniel Vetter 2014-03-26 19:03:49 UTC
Tested on both snb and ivb with latest igt, works well for me. Please retest and make sure you don't have anything funny going on. Also _really_ strange that this fails across the board.
Comment 12 lu hua 2014-03-28 07:42:57 UTC
Test on IVB with latest igt and nightly kernel:
# ./gem_flink_race --run-subtest flink_close
IGT-Version: 1.6-g9048aa1 (x86_64) (Linux: 3.14.0-rc7_drm-intel-nightly_10c69a_20140328+ x86_64)
leaked 1 objects
Test assertion failure function test_flink_close, file gem_flink_race.c:196:
Last errno: 2, No such file or directory
Failed assertion: obj_count == 0
Subtest flink_close: FAIL
# ./gem_flink_race --run-subtest flink_close
IGT-Version: 1.6-g9048aa1 (x86_64) (Linux: 3.14.0-rc7_drm-intel-nightly_10c69a_20140328+ x86_64)
start new ssh tunnel and quit
leaked 0 objects
Subtest flink_close: SUCCESS

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.