[Bug 102671] [CI][IGT] igt at kms_atomic_transition@plane-all-transition-nonblocking Failed assertion: ret == 0

[bugzilla-daemon at freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=102671

Maarten Lankhorst <bugs at mblankhorst.nl> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED

--- Comment #9 from Maarten Lankhorst <bugs at mblankhorst.nl> ---
Seems to occur randomly, but I probably know what's going on now.

kms_atomic_transitions.plane_all_transitions calls wm_setup_plane which
indiscriminately calls igt_plane_set_fb(plane, NULL); this is fine for planes
which were bound to the crtc, but for unbound planes we have a separate
completion event to prevent use-after-free. This is tested in the
'plane-use-after-nonblocking-unbind' subtest. When we call
igt_plane_set_fb(plane, NULL) on a disabled plane, by accident we update its
properties nonblockingly.

The unbound plane is not associated with a crtc, and thus has its own
completion event on commit.

The nonblocking update does drm_atomic_helper_wait_for_flip_done(), followed
some time later by drm_atomic_helper_commit_hw_done(). The commit_hw_done
signals the unbound planes, so if the new atomic commit runs between
wait_for_flip_done() and hw_done(), we get the -EBUSY here.

Fix is not blindly calling igt_plane_set_fb(plane, NULL) on disabled planes.

-- 
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/20171016/16b42b8a/attachment.html>