Re: [PATCH] Per-superblock unused dentry LRU lists V2

[David Chinner <dgc@sgi.com>]

On Thu, May 25, 2006 at 06:13:12PM +1000, David Chinner wrote:
> On Thu, May 25, 2006 at 12:22:20PM +0530, Balbir Singh wrote:
> > On Thu, May 25, 2006 at 04:33:50PM +1000, David Chinner wrote:
> > > On Thu, May 25, 2006 at 04:15:53PM +1000, David Chinner wrote:
> > > > 
> > > > FWIW, this create/unlink load has been triggering reliable "Busy
> > > > inodes after unmount" errors that I've slowly been tracking down.
> > > > After I fixed the last problem in XFS late last week, I've
> > > > been getting a failure that i think is the unmount/prune_dcache
> > > > races that you and Neil have recently fixed.
> > > 
> > > I just had all 8 filesystems come up with:
> > > 
> > > May 25 15:55:18 budgie kernel: XFS unmount got error 16
> > > May 25 15:55:18 budgie kernel: xfs_fs_put_super: vfsp/0xe00000b006339280 left dangling!
> > > May 25 15:55:18 budgie kernel: VFS: Busy inodes after unmount of dm-9. Self-destruct in 5 seconds.  Have a nice day...
> 
> .....
> 
> > > On the second test iteration. On 2.6.16, it takes about 10 iterations to get one
> > > filesystem to do this. I'll need to look into this one further. I'm going to
> > > reboot the machine and run some dbench tests (which typically don't trigger
> > > this problem) and then come back to this one with added debugging....
> > 
> > Is this with version 2 of your patch?
> 
> That's with version 2 on -rc4-mm3.
> 
> > Could you also try the -mm tree
> > and see if the problem goes away. We have a set of patches to address
> > exactly this problem. 
> 
> Well, the patches overlap and I thought that I had taken into account
> the fixes that were applied to the -mm tree.
> 
> I'm rerunning on 2.6.16 with the s_umount locking fix so I know that it's
> not something else in -mm that is being tripped over....

2.6.16 ran all night on the above test with the s_umount locking fix in it.
The prune_dcache/unmount race was indeed the cause of the errors I have
been seeing over the last week.

I just found a bug in the -mm patch which broke shrink_dcache_parent().
This is the likely cause of the above problems, and I'm about to retest
the -mm kernel with the fixed patch.....

Cheers,

Dave.
-- 
Dave Chinner
R&D Software Enginner
SGI Australian Software Group
-
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/