Bug 100256 - [HSW][BAT] gvt_basic@invalid-placeholder-test incomplete in CI
Summary: [HSW][BAT] gvt_basic@invalid-placeholder-test incomplete in CI
Status: CLOSED WORKSFORME
Alias: None
Product: DRI
Classification: Unclassified
Component: DRM/iGVT-g (show other bugs)
Version: DRI git
Hardware: Other All
: highest critical
Assignee: Terrence Xu
QA Contact: Terrence Xu
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2017-03-17 15:37 UTC by Jani Saarinen
Modified: 2017-06-09 06:20 UTC (History)
3 users (show)

See Also:
i915 platform: HSW
i915 features:


Attachments

Comment 1 Terrence Xu 2017-03-20 11:50:03 UTC
Hi Chris, can you provide more information for gvt_basic? it is gvt-d or gvt-g?
Comment 2 Chris Wilson 2017-03-20 13:07:59 UTC
It is neither.
Comment 3 Martin Peres 2017-03-23 12:36:41 UTC
Failure rate 1/59 run(s) (1%)

No occurences for the past 10 days. Raising the priority to get attention!
Comment 4 Dorota Czaplejewicz 2017-03-26 10:03:12 UTC
260 runs passed (skipped) without hangups on HSW-4770R.

Kernel: drm-tip: 2017y-03m-23d-22h-16m-00s
Config: CI 2393
Comment 5 Terrence Xu 2017-03-27 11:54:11 UTC
(In reply to Martin Peres from comment #3)
> Failure rate 1/59 run(s) (1%)
> 
> No occurences for the past 10 days. Raising the priority to get attention!

Hi Martin, I still confuse for meaning of failure for this case, can you explain it? It return unexpected skipped? or any failure information?
Comment 6 Terrence Xu 2017-03-27 11:54:49 UTC
(In reply to Dorota Czaplejewicz from comment #4)
> 260 runs passed (skipped) without hangups on HSW-4770R.
> 
> Kernel: drm-tip: 2017y-03m-23d-22h-16m-00s
> Config: CI 2393

Dorota, can you attach the detail config here? :)
Comment 7 Dorota Czaplejewicz 2017-03-27 16:30:46 UTC
Sure, the config I used is here:
https://intel-gfx-ci.01.org/CI/CI_DRM_2393/kernel.config.bz2

Sorry for not stating what CI this is.
Comment 8 Terrence Xu 2017-03-28 07:23:43 UTC
Hi Dorota, I use your config to build kernel and run the gvt_basic case on it, both Haswell platform and Skylake platform are run fail with skipped as below:


IGT-Version: 1.17-ge631bb5 (x86_64) (Linux: 4.11.0-rc3-CI-CI_DRM_2393+ x86_64)
Reloading i915 with enable_gvt=1

Test requirement not met in function __real_main29, file gvt_basic.c:34:
Test requirement: igt_gvt_load_module()
Last errno: 9, Bad file descriptor
Subtest invalid-placeholder-test: SKIP


should we need any specific configurations for host grub?
Comment 9 Dorota Czaplejewicz 2017-03-28 10:54:25 UTC
Terrence, judging by the CI: https://intel-gfx-ci.01.org/CI/igt@gvt_basic@invalid-placeholder-test.html
meaning of failure is "fail", the meaning of no failure is "skip". In this case, your execution would not be a failure but an expected skip.
Comment 10 Terrence Xu 2017-03-30 06:02:40 UTC
(In reply to Dorota Czaplejewicz from comment #9)
> Terrence, judging by the CI:
> https://intel-gfx-ci.01.org/CI/igt@gvt_basic@invalid-placeholder-test.html
> meaning of failure is "fail", the meaning of no failure is "skip". In this
> case, your execution would not be a failure but an expected skip.

So I cannot reproduce the issue in my side, it always returns skip in my haswell desktop.
Comment 11 Martin Peres 2017-04-12 16:23:20 UTC
(In reply to Martin Peres from comment #3)
> Failure rate 1/59 run(s) (1%)
> 
> No occurences for the past 10 days. Raising the priority to get attention!

Sorry, I did not get any email about this issue...

We got an incomplete in CI, instead of the usual skip. This bug report is actually quite stupid, now that I think of it. It is more likely than the previous test hanged the machine with a slight delay and/or the test never managed to get written to disk.

The current failure rate is 1 in 190 runs, so basically, we have no useful information, so I will close this bug.
Comment 12 Jani Saarinen 2017-04-28 13:56:06 UTC
This caused again incomplete in: https://patchwork.freedesktop.org/series/23702/

Re-opened to reconsider.
Comment 13 Chuanxiao Dong 2017-05-03 06:34:46 UTC
the test case is called gvt_basic, but cannot see the connection between GVT driver and the test case gvt_basic.

It is better if someone can help us to understand how the gvt_basic test case is related to the GVT driver
Comment 14 Chuanxiao Dong 2017-05-04 09:41:39 UTC
(In reply to Jani Saarinen from comment #12)
> This caused again incomplete in:
> https://patchwork.freedesktop.org/series/23702/
> 
> Re-opened to reconsider.

from the attached link, seems this series passed the test
Comment 15 Jani Saarinen 2017-05-04 10:05:08 UTC
It was re-run. But problem still there. 
See mail on intel-gfx:
> Test gem_exec_flush:
>         Subgroup basic-batch-kernel-default-uc:
>                 pass       -> FAIL       (fi-snb-2600) fdo#100007
> Test gvt_basic:
>         Subgroup invalid-placeholder-test:
>                 skip       -> INCOMPLETE (fi-hsw-4770r)
I guess just need to re-open:
https://bugs.freedesktop.org/show_bug.cgi?id=100256
blah...done, marked for the future.
Comment 16 Chuanxiao Dong 2017-05-10 01:12:17 UTC
(In reply to Jani Saarinen from comment #15)
> It was re-run. But problem still there. 
> See mail on intel-gfx:
> > Test gem_exec_flush:
> >         Subgroup basic-batch-kernel-default-uc:
> >                 pass       -> FAIL       (fi-snb-2600) fdo#100007
> > Test gvt_basic:
> >         Subgroup invalid-placeholder-test:
> >                 skip       -> INCOMPLETE (fi-hsw-4770r)
> I guess just need to re-open:
> https://bugs.freedesktop.org/show_bug.cgi?id=100256
> blah...done, marked for the future.

Thanks Jani.

I checked a kernel dmesg from: https://intel-gfx-ci.01.org/CI/Patchwork_4588/fi-hsw-4770r/igt@gvt_basic@invalid-placeholder-test.html, which should be in incomplete case for gvt_basic test.

From the kernel dmesg, I can see even gvt-g is not enabled:
[    3.388092] [drm:intel_gvt_init [i915]] GVT-g is disabled by kernel params

so would you please let me know what does this gvt_basic do with GVT-g?
Comment 17 Chris Wilson 2017-05-11 07:30:50 UTC
The test does nothing more than a module load with i915.enable_gvt=1. There is not a single error to suggest that the gvt initialisation code is at fault, indeed on hsw the initialisation code for gvt is not even run. This is just your irregular module load fail without information to discern the cause.
Comment 18 Chuanxiao Dong 2017-06-01 02:03:00 UTC
(In reply to Chris Wilson from comment #17)
> The test does nothing more than a module load with i915.enable_gvt=1. There
> is not a single error to suggest that the gvt initialisation code is at
> fault, indeed on hsw the initialisation code for gvt is not even run. This
> is just your irregular module load fail without information to discern the
> cause.

Hi Chris,

As we reviewed from the patch "drm/i915/gvt: Add gvt options sanitize function", for the platforms which are not supported, enable_gvt will be cleared, and keep the i915 module loading successfully.

With this design, do you think what should be the next step for this issue?
Comment 19 Jani Saarinen 2017-06-09 06:20:17 UTC
Lets close this as not reproducible.
As written. This is just your irregular module load fail without information to discern the cause.

Last seen: 2017-03-13
Statistics: Failure rate 1/372 run(s) (0%)


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.