[Bug 109495] [CI][DRMTIP] igt at kms_* - fail - Failed assertion: igt_ioctl((fd), ((((2U|1U) << (((0+8)+8)+14)) | ((('d')) << (0+8)) | (((0xB2)) << 0) | ((((sizeof(struct drm_mode_create_dumb)))) << ((0+8)+8)))), (&create)) == 0

[bugzilla-daemon at freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=109495

--- Comment #10 from Stanislav Lisovskiy <stanislav.lisovskiy at intel.com> ---
Update: 

With the help of Chamelium, I can now easily reproduce this anytime and
currently the issue consists of two separate issues:

1) There seems to be some hardware bouncing involved for hpd pins, like very
display goind to unplug and plug in a very short time. The kernel seems to not
able filter those as those events are not exceeding the required threshold for
irq storm detection in a fixed period of time, i.e those anyway get processed
like connector going to disconnected and/or connected state quickly.

2) Sometimes connector goes to disconnected state in "unlucky" time, when igt
does refresh the output, because currently it refreshes the outputs
periodically in igt_output_pipe_select, during that time kernel manages to
return some kind of garbage mode, containing all 0, which is then saved by igt
and used during framebuffer creation => causing the asserting as
drm_mode_create_dumb rejects creation of 0 size, 0 height and width buffer.

IMHO this should have fixes both in kernel(against small very short hpd pin
bounces, not detectable by storm detection) and also in igt, as it should not
refresh the outputs on the fly during test execution, but use cached version
for
outputs, if output happens to be disconnected in the middle of a test, this
will at least fail on commit, i.e in a controlled way.

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/intel-gfx-bugs/attachments/20190920/46a63f5f/attachment.html>