Created attachment 130198 [details] log Kernel drm-tip: 2017y-03m-13d-16h-06m-45s, config 2327. CPU: Celeron(R) CPU G3900 GPU: 00:02.0 0300: 8086:1902 (rev 06) (prog-if 00 [VGA controller]) The test fails 90% of the time.
2x-long-cursor-vs-flip-atomic is also affected with the same failure mode: (kms_cursor_legacy:3135) CRITICAL: Test assertion failure function two_screens_cursor_vs_flip, file kms_cursor_legacy.c:1221: (kms_cursor_legacy:3135) CRITICAL: Failed assertion: vbl.sequence <= vblank_start[child] + 5 * vrefresh[child] / 8 (kms_cursor_legacy:3135) CRITICAL: Last errno: 9, Bad file descriptor (kms_cursor_legacy:3135) CRITICAL: error: 256427 > 256426
This report looks related to this bug: https://intel-gfx-ci.01.org/CI/CI_DRM_2350/fi-skl-6770hq/igt@kms_pipe_crc_basic@nonblocking-crc-pipe-b-frame-sequence.html
Something changed between drm-tip: 2017y-03m-13d and 2017y-03m-16d and 2x tests are SKIP now. The system has only 1 VGA screen connected.
https://bugs.freedesktop.org/show_bug.cgi?id=100195 - this bug's associated series: https://patchwork.freedesktop.org/series/21226/ brings down the failure rate from 90% to <10%
Failure rate 1/36 run(s) (2%) Increasing the severity because it affects our CI.
All 2x-*-atomic fail since commit 8d2b47dde8a097e6fef2ebb5042cbb267cc75adf Author: Maarten Lankhorst <maarten.lankhorst@linux.intel.com> Date: Thu Feb 2 08:41:42 2017 +0100 drm/i915: Enable atomic support by default on supported platforms. Failure rate with that commit: 2x-cursor-vs-flip-atomic: (fail 7 + dmesg-fail 5) / 32 = 12/32 2x-long-cursor-vs-flip-atomic: (fail 22 + dmesg-fail 7) / 32 = 29/32
Ok so they've always failed, before they were skipped. Patch from Ville should definitely help.
Which patch do you mean? I tried this one: https://patchwork.freedesktop.org/series/21226/ (see comment 21226). It helped somewhat, I will check if failure mode was different after it was applied.
Is this still an issue with latest drm-tip?
The bug has been seen once, and not reproduced in the following 74 runs. When was the patch from Ville pushed to drm-tip? If you think this should be fixed, then we can close the bug and whitelist the test on this platform, but if we are wrong, more noise will happen on pre-merge.
With latest drm-tip, kms_cursor_legacy/2x-cursor-vs-flip-atomic is SKIP. I'm still using just one external VGA screen - perhaps the correct outcome is SKIP indeed. Unfortunately, I will not be able to dig deeper into this issue. Using 2 screens (DVI+VGA) only fails 2/64 times, so I assume it's unrelated to this report.
Statistics: Failure rate 1/163 run(s) (0%)
Potentially fixed by: commit 93aa2a1c25e562cc0ca69c3175333fe33fdf055b Author: Ville Syrjälä <ville.syrjala@linux.intel.com> AuthorDate: Tue Mar 14 17:10:50 2017 +0200 Commit: Ville Syrjälä <ville.syrjala@linux.intel.com> CommitDate: Wed Mar 22 22:04:50 2017 +0200 drm/i915: Fix SKL cursor watermarks Use intel_wm_plane_visible() to determine cursor visibility for SKL+ also. Previously SKL+ would check the actual visibility which now conflicts with the assumptions in intel_legacy_cursor_update(). We also change SKL+ to compute the cursor watermarks based on the unclipped cursor size, just as we do on all the other platforms. Using the clipped size could now result in garbage results. Testcase: igt/kms_chv_cursor_fail Fixes: a5509abda48e ("drm/i915: Fix legacy cursor vs. watermarks for ILK-BDW ") Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=100195 Marking resolved and closing.
Whitelisted for CI
Just as final confirmation, issue didn't reproduce in over 50 runs. Closing this.
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.