xfs
[Top] [All Lists]

Re: 2.6.27.7 vanilla, project quota enabled and process stuck in D state

To: Christoph Hellwig <hch@xxxxxxxxxxxxx>
Subject: Re: 2.6.27.7 vanilla, project quota enabled and process stuck in D state (repeatable every time)
From: Dave Chinner <david@xxxxxxxxxxxxx>
Date: Fri, 5 Dec 2008 08:34:01 +1100
Cc: Arkadiusz Miskiewicz <arekm@xxxxxxxx>, xfs@xxxxxxxxxxx
In-reply-to: <20081204123206.GA6935@xxxxxxxxxxxxx>
Mail-followup-to: Christoph Hellwig <hch@xxxxxxxxxxxxx>, Arkadiusz Miskiewicz <arekm@xxxxxxxx>, xfs@xxxxxxxxxxx
References: <200812021949.55463.arekm@xxxxxxxx> <200812031406.41882.arekm@xxxxxxxx> <20081203213028.GW18236@disturbed> <200812032242.29326.arekm@xxxxxxxx> <20081203220934.GA32301@disturbed> <20081204123206.GA6935@xxxxxxxxxxxxx>
User-agent: Mutt/1.5.18 (2008-05-17)
On Thu, Dec 04, 2008 at 07:32:06AM -0500, Christoph Hellwig wrote:
> On Thu, Dec 04, 2008 at 09:09:34AM +1100, Dave Chinner wrote:
> > Is that the unlock of the inodes is using the incorrect lock
> > type for the unlock, (inodes lock XFS_ILOCK_EXCL, unlocked XFS_ILOCK_SHARED)
> > which means they don't get unlocked and the next attempt to do anything
> > with those inodes will hang.
> > 
> > Compile-tested-only patch below that should fix the problem.
> 
> Yeah, that also explains why my patch fixes it :)  I'd say let's put
> yours into 2.6.28 and -stable, and I'll rediff mine ontop for the 2.6.29
> queue.  I'll also write a testcase for xfsqa based on Arkadiusz's
> report.

I agree that this is probably the best approach - your fix is the
better long term solution, I think.

SGI folk, can we get my patch pushed to linus and stable ASAP?
Probably be an idea to add a:

Tested-by: Arkadiusz Miskiewicz <arekm@xxxxxxxx>

tag to it as well to make it easy for the stable review process....

Cheers,

Dave.
-- 
Dave Chinner
david@xxxxxxxxxxxxx

<Prev in Thread] Current Thread [Next in Thread>