[Bug 65387] [HSW] igt/gem_seqno_wrap randomly aborted on -dinq

[bugzilla-daemon at freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=65387

--- Comment #9 from Mika Kuoppala <mika.kuoppala at intel.com> ---
(In reply to comment #8)
> So, I think inter-ring serialisation is broken. The batch is passed from
> render -> bsd -> blt -> vecs. In order for that object to be removed from
> the active list, vecs must have completed its request early. Indeed we can
> tell that much by the RING registers )which is all we have :(

gem_seqno_wrap doesn't submit anything to bsd nor vecs rings.

Apparently gem_seqno_wrap doesn't even success to submit any batches.
The first time it tries to alter the driver seqno, it get -EAGAIN
back, most likely from __wait_seqno due to differing reset counts.

I suspect that that the _previous_ test that was run has not completed
cleanly. It has still work in rings when the test ends.

So either previous test has injected something which hangs the bsd ring
and gem_seqno_wrap is a victim here as it will find hung gpu when it starts. Or
calling gem_seqno_init while there are legimate work in the rings causes gpu
hang.

Lu, could you find out what test is run before gem_seqno_wrap?
And perhaps alter the test order to see the effects.

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.freedesktop.org/archives/intel-gfx-bugs/attachments/20130614/6c54148a/attachment.html>