Date: Thu, 28 Apr 2011 10:30:47 -0700 (PDT)
From: Christian Kujau <lists@nerdbynature.de>
To: Dave Chinner <david@fromorbit.com>
cc: LKML <linux-kernel@vger.kernel.org>, xfs@oss.sgi.com
Subject: Re: 2.6.39-rc4+: oom-killer busy killing tasks
In-Reply-To: <20110427102824.GI12436@dastard>
Message-ID: <alpine.DEB.2.01.1104281008320.18728@trent.utfs.org>
References: <alpine.DEB.2.01.1104211841510.18728@trent.utfs.org> <20110424234655.GC12436@dastard> <alpine.DEB.2.01.1104242245090.18728@trent.utfs.org> <alpine.DEB.2.01.1104250015480.18728@trent.utfs.org> <20110427022655.GE12436@dastard>
 <alpine.DEB.2.01.1104270042510.18728@trent.utfs.org> <20110427102824.GI12436@dastard>
User-Agent: Alpine 2.01 (DEB 1266 2009-07-14)
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 1825
Lines: 57

Hi David,

sorry for the late response:

On Wed, 27 Apr 2011 at 20:28, Dave Chinner wrote:
> hmmmm. Speaking of which - have you changed any of the XFS tunables
> in /proc/sys/fs/xfs/ on your machine (specifically
> xfssyncd_centisecs)?

No, I did not change anything underneath /proc/sys/fs/xfs.
 
> Not easy because there are tree-wide changes that need to be
> preserved (e.g. block layer plugging changes) while others around it
> would need to be reverted....

Yeah, I noticed. I wanted to do something like this:

  $ for c in `git log --pretty=oneline --no-merges v2.6.38..v2.6.39-rc2 fs/xfs | awk '{print $1}'`; do
       git revert $c
    done

...but this broke on the 1st or 2nd commit because of unresolved changes.

Now I'm in the middle of another bisect attempt, trying to reproduce but 
Im getting (probably totally unrelated) warnings here:

BUG: sleeping function called from invalid context at /usr/local/src/linux-2.6-git/drivers/ide/ide-io.c:447
in_atomic(): 1, irqs_disabled(): 0, pid: 0, name: swapper
INFO: lockdep is turned off.
[...]
 
> Can you check if there are any blocked tasks nearing OOM (i.e. "echo
> w > /proc/sysrq-trigger") so we can see if XFS inode reclaim is
> stuck somewhere?

In one instance when I mounted with "noatime", no log messages made it to 
the disk, but I could see "hung tasks":

   http://nerdbynature.de/bits/2.6.39-rc4/oom/oom-6.JPG

I will try sysrq-w later on.

FWIW, a regression bug has been opened:

  https://bugzilla.kernel.org/show_bug.cgi?id=34012

Thanks,
Christian.
-- 
BOFH excuse #284:

Electrons on a bender
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/