[Bug 94829] X crash when starting minetest with compositing and DRI3 PRIME

[bugzilla-daemon at freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=94829

--- Comment #11 from Chris Wilson <chris at chris-wilson.co.uk> ---
Step 1, commit 65dc4176d84e1d1764a6fa1edc972aa7f1dcd2ba
Author: Chris Wilson <chris at chris-wilson.co.uk>
Date:   Tue Apr 5 17:53:55 2016 +0100

    sna/present: Prevent reporting an incomplete event


It looks very suspicious that we get a sequence of happy flips (both submit,
completion events) then a failure with -ENOMEM after ~800 flips without much of
a hint of a leak.

Hmm, ah, what appears to be happening is that the event queue is filled up with
future vblanks. The first hint of trouble is:

[ 67019.025] sna_present_queue: target msc=4008810, seq=4008810
(last_msc=4008683)
[ 67019.025] sna_wait_vblank(pipe=0, waiting until seq=4008810)
[ 67019.025] sna_present_queue: vblank enqueue failed, faking
[ 67019.025] msc_to_delay: sleep 127 frames, 2113 ms
[ 67019.025] sna_fake_vblank(event=954, target_msc=4008810, msc=4008683,
delay=2113ms)

And they continue to submit more and more vblanks into the future, just prior
to the crash (when we try to flip onto a full queue) we are up to:

[ 67023.142] sna_present_queue_vblank(pipe=0, event=1563, msc=4009173, last
swap=4008929)
[ 67023.142] sna_present_queue: target msc=4009173, seq=4009173
(last_msc=4008929)
[ 67023.142] sna_wait_vblank(pipe=0, waiting until seq=4009173)
[ 67023.142] sna_present_queue: vblank enqueue failed, faking
[ 67023.142] msc_to_delay: sleep 244 frames, 4057 ms
[ 67023.142] sna_fake_vblank(event=1563, target_msc=4009173, msc=4008929,
delay=4057ms)

It feels like it queuing vblanks twice as fast as vrefresh.

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/intel-gfx-bugs/attachments/20160405/676ea2e7/attachment.html>