Bug 75079 - [PNV/BSW Bisected]igt/kms_pipe_crc_basic/read-crc-pipe-A-frame-sequence is slow
Summary: [PNV/BSW Bisected]igt/kms_pipe_crc_basic/read-crc-pipe-A-frame-sequence is slow
Status: CLOSED FIXED
Alias: None
Product: DRI
Classification: Unclassified
Component: DRM/Intel (show other bugs)
Version: unspecified
Hardware: All Linux (All)
: highest major
Assignee: Antti Koskipaa
QA Contact: Intel GFX Bugs mailing list
URL:
Whiteboard:
Keywords:
: 77475 78282 (view as bug list)
Depends on:
Blocks:
 
Reported: 2014-02-17 03:34 UTC by lu hua
Modified: 2016-10-07 08:50 UTC (History)
5 users (show)

See Also:
i915 platform:
i915 features:


Attachments
dmesg (85.38 KB, text/plain)
2014-02-17 03:34 UTC, lu hua
no flags Details
dmesg(BSW) (122.64 KB, text/plain)
2014-11-06 09:03 UTC, lu hua
no flags Details

Description lu hua 2014-02-17 03:34:15 UTC
Created attachment 94186 [details]
dmesg

system Environment:
--------------------------
Platform: Pineview
Kernel:(drm-intel-fixes)f51a44b9a6c4982cc25bfb3727de9bb893621ebc

Bug detailed description:
-----------------------------
kms_pipe_crc_basic/read-crc-pipe-A-frame-sequence is slow on pineview with -fixes and -nightly kernel.
Run more than 30 minutes, It doesn't exit testing.

The latest known good commit: 3670d6dbff7b829d71ce84d6c9cb88cd5abdbc53
The latest known bad commit: 959783684e2010326d91db2aa55f424f7e041c84

output:
IGT-Version: 1.5-g9597836 (i686) (Linux: 3.13.0-rc8_drm-intel-fixes_f51a44_20140217+ i686)
read-crc-pipe-A-frame-sequence: Testing connector LVDS-1 using pipe A

Reproduce steps:
----------------------------
1. ./kms_pipe_crc_basic --run-subtest read-crc-pipe-A-frame-sequence
Comment 1 Damien Lespiau 2014-02-18 15:28:48 UTC
can you strace the test and confirm that it's blcoking in a read() syscall? If so, this can be alleviated by implementing poll() support for this debugsfs file to be able to timeout.
Comment 2 lu hua 2014-03-13 04:54:40 UTC
Bisect shows: 417987353ec2bf637839990553967474ccf1528e is the first bad commit.
commit 417987353ec2bf637839990553967474ccf1528e
Author: Damien Lespiau <damien.lespiau@intel.com>
Date:   Mon Jan 27 16:32:09 2014 +0000

    kms_pipe_crc_basic: Port to the new modeset API

    Signed-off-by: Damien Lespiau <damien.lespiau@intel.com>
Comment 3 Daniel Vetter 2014-03-27 16:28:12 UTC
I think this is another case of the new framework's failure to handle the "encoder can't be used on all pipes" restriction correctly. I see the hang on all my gen2/3 lvds machines. If we restrict the pipe of any connector on modern platforms it'll likely blow up exactly the same way.

Poking Damien harder.
Comment 4 Daniel Vetter 2014-05-15 14:46:59 UTC
*** Bug 78282 has been marked as a duplicate of this bug. ***
Comment 5 Daniel Vetter 2014-05-15 14:59:52 UTC
This will block igt_kms testing on chv since that platform reintroduced pipe limits on certain outputs ...
Comment 6 Rodrigo Vivi 2014-06-05 17:03:27 UTC
*** Bug 77475 has been marked as a duplicate of this bug. ***
Comment 7 Guo Jinxian 2014-07-30 08:48:12 UTC
This bug still able to reproduce on BSW on latest -nightly(eb2af97968f0321622cf3aeb29991ad78cfdc1ca)

[root@x-bsw01 tests]# time ./kms_plane --run-subtest plane-position-covered-pipe-A-plane-1
IGT-Version: 1.7-ge48c495 (x86_64) (Linux: 3.16.0-rc6_drm-intel-nightly_eb2af9_20140730+ x86_64)
Testing connector HDMI-A-3 using pipe A plane 1
Subtest plane-position-covered-pipe-A-plane-1: TIMEOUT

real    0m5.984s
user    0m0.018s
sys     0m0.151s
Comment 8 Guo Jinxian 2014-08-21 03:24:32 UTC
Test timeout while run cases below on BSW on latest -nightly(49ceddc7b3b0e3f7eb21b9fb4c9b840cff630047)

Case list:
igt/kms_pipe_crc_basic/hang-read-crc-pipe-B
igt/kms_pipe_crc_basic/hang-read-crc-pipe-C
igt/kms_pipe_crc_basic/suspend-read-crc-pipe-B
igt/kms_pipe_crc_basic/suspend-read-crc-pipe-C

Output:
[root@x-hsw24 tests]# time ./kms_pipe_crc_basic --run-subtest hang-read-crc-pipe-B
IGT-Version: 1.7-g4d2f511 (x86_64) (Linux: 3.17.0-rc1_drm-intel-nightly_49cedd_20140821+ x86_64)
hang-read-crc-pipe-B: Testing connector VGA-1 using pipe B
Subtest hang-read-crc-pipe-B: TIMEOUT

real    0m11.384s
user    0m0.002s
sys     0m0.008s
Comment 9 Paulo Zanoni 2014-10-09 12:49:06 UTC
Cc'ing Damien on the bug due to comment #3...
Comment 10 Antti Koskipaa 2014-11-05 10:53:07 UTC
(In reply to Guo Jinxian from comment #8)
> Test timeout while run cases below on BSW on latest
> -nightly(49ceddc7b3b0e3f7eb21b9fb4c9b840cff630047)
> 
> Case list:
> igt/kms_pipe_crc_basic/hang-read-crc-pipe-B
> igt/kms_pipe_crc_basic/hang-read-crc-pipe-C
> igt/kms_pipe_crc_basic/suspend-read-crc-pipe-B
> igt/kms_pipe_crc_basic/suspend-read-crc-pipe-C
> 
> Output:
> [root@x-hsw24 tests]# time ./kms_pipe_crc_basic --run-subtest
> hang-read-crc-pipe-B
> IGT-Version: 1.7-g4d2f511 (x86_64) (Linux:
> 3.17.0-rc1_drm-intel-nightly_49cedd_20140821+ x86_64)
> hang-read-crc-pipe-B: Testing connector VGA-1 using pipe B
> Subtest hang-read-crc-pipe-B: TIMEOUT
> 
> real    0m11.384s
> user    0m0.002s
> sys     0m0.008s

What is this? You say BSW, hostname says hsw. BSW does not support VGA. So please retest on BSW and attach the correct log.
Comment 11 lu hua 2014-11-06 09:03:56 UTC
Created attachment 109014 [details]
dmesg(BSW)


> What is this? You say BSW, hostname says hsw. BSW does not support VGA. So
> please retest on BSW and attach the correct log.

Both BDW and BSW have this issue. comment 6's bug 77475 mentioned HSW. Comment 8 says BSW also has this issue, in order to illustrate BSW's test status. 

Test on BSW with latest -nightly kernel, the timeout is unstable.
output:
[root@x-bsw01 tests]# time ./kms_pipe_crc_basic --run-subtest read-crc-pipe-A-frame-sequence
IGT-Version: 1.8-ga973aab (x86_64) (Linux: 3.18.0-rc3_drm-intel-nightly_0642a5_20141106+ x86_64)
read-crc-pipe-A-frame-sequence: Testing connector eDP-1 using pipe A
Test assertion failure function test_read_crc_for_output, file kms_pipe_crc_basic.c:188:
Failed assertion: igt_crc_equal(&crcs[j], &crcs[j + 1])
Subtest read-crc-pipe-A-frame-sequence: FAIL (2.051s)

real    0m4.587s
user    0m0.010s
sys     0m0.074s
[root@x-bsw01 tests]# time ./kms_pipe_crc_basic --run-subtest read-crc-pipe-A-frame-sequence
IGT-Version: 1.8-ga973aab (x86_64) (Linux: 3.18.0-rc3_drm-intel-nightly_0642a5_20141106+ x86_64)
read-crc-pipe-A-frame-sequence: Testing connector eDP-1 using pipe A
Subtest read-crc-pipe-A-frame-sequence: TIMEOUT (6.993s)

real    0m9.537s
user    0m0.015s
sys     0m0.062s
Comment 12 Daniel Vetter 2014-11-14 09:15:51 UTC
Hm, I guess I didn't manage to guilt-trip Damien into fixing this.

[RFC] More structure for igt_kms tests
http://lists.freedesktop.org/archives/intel-gfx/2014-July/048891.html

is what we need. I guess we need to track this in a Jira.
Comment 13 Daniel Vetter 2014-11-14 09:17:58 UTC
Actually the original regression is fixed iirc from Ville, this is just plain and simple a bsw/bdw crc bug. Please file new bug reports for this.
Comment 14 lu hua 2014-11-17 03:18:38 UTC
Test on PNV the timeout goes away, but some subcases fail, file bug 86365 track it.
Test on BSW, some subcases still timeout, they will be tracked in bug 83519. Close this bug.
Comment 15 Jari Tahvanainen 2016-10-07 08:50:20 UTC
Closing verified+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.