Bug 93122

Please attach the dmesg etc. so we can at least have a chance at looking at what type of machine it is.

The testcase SKIPs. There isn't anything in dmesg ofc.

For full details please read

http://benchsrv.fi.intel.com/archive/results/CI_IGT_test/CI_DRM_848/snb-dellxps/html/snb-dellxps@CI_DRM_848@1/igt@pm_rpm@basic-pci-d3-state.html

Unfortunately I can't get at a working log any more because CI deletes old logs :(

Correction, link was wrong. Working testrun:

http://benchsrv.fi.intel.com/archive/results/CI_IGT_test/CI_DRM_843/snb-dellxps/html/snb-dellxps@CI_DRM_843@1/igt@pm_rpm@basic-pci-d3-state.html

Unassigning since Jani only assigned to me for additional info. I can't repro this on my local SNB, there pm_rpm@basic-rte works perfectly.

Looking at the CI results page, I don't see any SKIPs happening recently, especially for this machine. What I see is that in some cases the test was not run, and I'm not sure if this is a problem.

Can we close this?

As a note, if we skip at enable_one_screen_with_type(ANY), it means that the program didn't find any connected connector for modeset, which could be either a bug or just a machine without real connectors connected.

(In reply to Paulo Zanoni from comment #5)
> Looking at the CI results page, I don't see any SKIPs happening recently,
> especially for this machine. What I see is that in some cases the test was
> not run, and I'm not sure if this is a problem.
> 
> Can we close this?

It's appeared in CI patch results today on the ml, so it is still definitely a problem.

Running the test ./pm_rpm --run-subtest basic-pci-d3-state is passing on my side on an SNB platform using the following Configuration:


Kernel 4.5.0-nightly

commit 4b39223f6e3bef4dfa678f7239dcd87c38e20e96
Author: Matt Roper <matthew.d.roper@intel.com>
Date:   Mon Mar 21 11:44:01 2016 -0700
    drm-intel-nightly: 2016y-03m-21d-18h-43m-18s UTC integration manifest
--------------------------------------------------------------------------

Hardware Information

Platform: SNB
Motherboard Model: OptiPlex 990
CPU: Core i7-2600 CPU@ 3.40GHz
GPU Card: 2nd Generation Core processor family Integrated graphics controller
Memory ram: 8GB
Hard Drive capacity: 2.00 TB
---------------------------------------------------------------------------

FWIW, it looks like this test has sporadically inconsistent behavior across all platforms, not sure whether the root cause it's the same though. If someone looks into this further, maybe it's worth considering other cases as well (e.g. dmesg fails on SNB-x220t and occasional skips across all).

http://benchsrv.fi.intel.com/archive/results/CI_IGT_test/igt@pm_rpm@basic-pci-d3-state.html for reference

We have seen it only on SNB and APL (but this is expected result at this stage). Can you confirm what are the other impacted platforms as well as attaching more information whenever this bug is also being seen?

What's the next step for this bug? It's been open for 3 months with no comment.  Perhaps it's time to close it?  Is the bug still being seen in CI runs?

The root cause is https://patchwork.freedesktop.org/patch/88669/

Pretending my analysis is perfect,

commit 12c100bfa5d9103b6c4d43636fee09c31e75605a
Author: Ville Syrjälä <ville.syrjala@linux.intel.com>
Date:   Mon May 23 17:42:48 2016 +0300

    drm/i915: Never fully mask the the EI up rps interrupt on SNB/IVB

and note that we worked around this bug in the meantime by avoiding the looping batch on snb inside igt.

Closing resolved+fixed with assumption that SKIP is expected behavior for basic-pci-d3-state on SNB. See https://intel-gfx-ci.01.org/CI/igt@pm_rpm@basic-pci-d3-state.html. If assumption is not correct then this bug shall be reopened against IGT component, not kernel.