From mboxrd@z Thu Jan 1 00:00:00 1970
From: bugzilla-daemon@freedesktop.org
Subject: [Bug 91278] Tonga GPU lock/reset fail with Unigine Valley
Date: Sat, 10 Oct 2015 10:33:39 +0000
Message-ID:
References:
Mime-Version: 1.0
Content-Type: multipart/mixed; boundary="===============0861362247=="
Return-path:
Received: from culpepper.freedesktop.org (unknown [131.252.210.165])
by gabe.freedesktop.org (Postfix) with ESMTP id 77C956E334
for ; Sat, 10 Oct 2015 03:33:39 -0700 (PDT)
In-Reply-To:
List-Unsubscribe: ,
List-Archive:
List-Post:
List-Help:
List-Subscribe: ,
Errors-To: dri-devel-bounces@lists.freedesktop.org
Sender: "dri-devel"
To: dri-devel@lists.freedesktop.org
List-Id: dri-devel@lists.freedesktop.org
--===============0861362247==
Content-Type: multipart/alternative; boundary="1444473219.6e511.10084"; charset="UTF-8"
--1444473219.6e511.10084
Date: Sat, 10 Oct 2015 10:33:39 +0000
MIME-Version: 1.0
Content-Type: text/plain; charset="UTF-8"
https://bugs.freedesktop.org/show_bug.cgi?id=91278
--- Comment #33 from Mathias Tillman ---
(In reply to Andy Furniss from comment #32)
> (In reply to Mathias Tillman from comment #31)
> > Small update: I have been able to run it for a total of 7 hours (2 + 2 + 3)
> > without a hang on the latest drm-next-4.4-wip. Tried the latest
> > drm-fixes-4.3 and it hung after about 30 minutes, so I definitely think
> > there's something in 4.4 that fixes it. I will do a bisect to see if I can
> > figure out what, more exactly, is the fix.
>
> I'll try over the weekend - no locks so far.
>
> Bisecting would be a pain for me as I managed to very long lucky runs
> previously, so wouldn't be able to easily call good.
>
> I notice that enable_scheduler is now on by default, maybe flipping that
> would be quicker than bisect. Long ago I did try some older kernel with that
> enabled and though it didn't fix IIRC it took longer to lock than was normal
> at that time.
Just ran it again with the scheduler disabled (from code, not through the
module parameter) on 4.4-wip, and sure enough, after about 30 minutes it hung.
So it looks like there's something in the scheduler that either makes it not
happen as often, or not at all (will need to confirm this).
--
You are receiving this mail because:
You are the assignee for the bug.
--1444473219.6e511.10084
Date: Sat, 10 Oct 2015 10:33:39 +0000
MIME-Version: 1.0
Content-Type: text/html; charset="UTF-8"
Comment # 33
on bug 91278
from Mathias Tillman
(In reply to Andy Furniss from comment #32)
> (In reply to Mathias Tillman from comment #31)
> > Small update: I have been able to run it for a total of 7 hours (2 + 2 + 3)
> > without a hang on the latest drm-next-4.4-wip. Tried the latest
> > drm-fixes-4.3 and it hung after about 30 minutes, so I definitely think
> > there's something in 4.4 that fixes it. I will do a bisect to see if I can
> > figure out what, more exactly, is the fix.
>
> I'll try over the weekend - no locks so far.
>
> Bisecting would be a pain for me as I managed to very long lucky runs
> previously, so wouldn't be able to easily call good.
>
> I notice that enable_scheduler is now on by default, maybe flipping that
> would be quicker than bisect. Long ago I did try some older kernel with that
> enabled and though it didn't fix IIRC it took longer to lock than was normal
> at that time.
Just ran it again with the scheduler disabled (from code, not through the
module parameter) on 4.4-wip, and sure enough, after about 30 minutes it hung.
So it looks like there's something in the scheduler that either makes it not
happen as often, or not at all (will need to confirm this).
You are receiving this mail because:
- You are the assignee for the bug.
--1444473219.6e511.10084--
--===============0861362247==
Content-Type: text/plain; charset="utf-8"
MIME-Version: 1.0
Content-Transfer-Encoding: base64
Content-Disposition: inline
X19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX19fX18KZHJpLWRldmVs
IG1haWxpbmcgbGlzdApkcmktZGV2ZWxAbGlzdHMuZnJlZWRlc2t0b3Aub3JnCmh0dHA6Ly9saXN0
cy5mcmVlZGVza3RvcC5vcmcvbWFpbG1hbi9saXN0aW5mby9kcmktZGV2ZWwK
--===============0861362247==--