[Bug 69340] Recent mesa git revisions cause frequent gpu hangs on radeonsi

bugzilla-daemon at freedesktop.org bugzilla-daemon at freedesktop.org
Sat Sep 14 07:35:44 PDT 2013


https://bugs.freedesktop.org/show_bug.cgi?id=69340

--- Comment #5 from Dave Witbrodt <dawitbro at sbcglobal.net> ---
Created attachment 85830
  --> https://bugs.freedesktop.org/attachment.cgi?id=85830&action=edit
Sample of dmesg output

I began seeing these symptoms on my HD 7850 PITCAIRN when I attempted to
upgrade Mesa from commit 6b5c802c (Sep. 2) to commit 2937d704 (Sep. 6).  I had
upgraded libdrm from 2.4.46 to commit 58d00888 at the same time.

The attached dmesg output looks a lot like the bug reported here by José, so I
hope I'm not interfering with an unrelated problem.

I have been having trouble finding time to investigate, which is why I did not
report this myself sooner.  I am using a stable 3.10 kernel with DRM cherry
picks from 3.11 and upcoming 3.12 -- which is not appropriate for use when
reporting bugs.  I also did not rebuild the entire X stack when I upgraded, but
just libdrm and Mesa.  There was a lot of due diligence I needed to perform
before filing a bug report here...

Anyway, I'm glad to see others reporting this -- now I feel less alone.  Not a
lot happened in Mesa between 6b5c802c and 2937d704, so if it turns out that my
Frankenstein kernel is not to blame, and rebuilding the X stack doesn't help,
then I'm going to bisect Mesa.  I have "good" and "bad" commits to use, and I'm
really interested in seeing whether the one big Radeon change in that interval
is the culprit:


commit a81beee37e0dd7b75422448420e8e8b0b4b76c1e
Author: Alex Deucher <alexander.deucher at amd.com>
Date:   Fri Sep 6 16:43:34 2013 -0400

    radeon/winsys: pad IBs to a multiple of 8 DWs

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.freedesktop.org/archives/dri-devel/attachments/20130914/164c9533/attachment.html>


More information about the dri-devel mailing list