Re: btrfs io errors on 3.4rc1

[Chris Mason]

On Mon, Apr 02, 2012 at 06:51:31PM -0400, Dave Jones wrote:
> On Mon, Apr 02, 2012 at 06:39:19PM -0400, Chris Mason wrote:
>  > On Mon, Apr 02, 2012 at 06:33:50PM -0400, Dave Jones wrote:
>  > > On Mon, Apr 02, 2012 at 06:28:02PM -0400, Chris Mason wrote:
>  > >  
>  > >  > > x86-64.
>  > >  > > 
>  > >  > > dmesg below.  (ignore the rpc oops, reported elsewhere, it's unrelated)
>  > >  > 
>  > >  > Well, there really are no btrfs messages in there at all.  Do you have
>  > >  > free space for a clean copy of the btrfs partition?  Trying to figure
>  > >  > out if you have a stale corruption (on two boxes seems really unlikely).
>  > >  > I definitely can't reproduce it here.
>  > > 
>  > > Don't really have any free space for it (It's the / partition on both)
>  > > 
>  > > Is there a wip btrfs.fsck yet ? One of the machines is just a scratch testbox,
>  > > so I'm happy to run it even if it'll potentially make things worse.
>  > 
>  > The current btrfsck repair mode will fix problems in the extent
>  > allocation tree, but that probably isn't what you're seeing.  I'd just
>  > run it read only and see if it finds any problems.
>  > 
>  > (You'll need the FS completely unmounted though).
> 
> I'll start a bisect later to see if I can narrow it down at least.

Ok, a directed bisect of the major suspects.  Josef changed the extent
buffer eio code in this commit (jump to the commit before it):

commit ea466794084f55d8fcc100711cf17923bf57e962
Author: Josef Bacik <josef@xxxxxxxxxx>
Date:   Mon Mar 26 21:57:36 2012 -0400

    Btrfs: deal with read errors on extent buffers differently

You can jump to the commit before my big blocks change:

commit 727011e07cbdf87772fcc1999cccd15cc915eb62
Author: Chris Mason <chris.mason@xxxxxxxxxx>
Date:   Fri Aug 6 13:21:20 2010 -0400

    Btrfs: allow metadata blocks larger than the page size

The SUSE code would be my other suspect.  It came from this commit:

commit 1d4284bd6e8d7dd1d5521a6747bdb6dc1caf0225
Merge: b5d67f6 65139ed
Author: Chris Mason <chris.mason@xxxxxxxxxx>
Date:   Wed Mar 28 20:31:37 2012 -0400

    Merge branch 'error-handling' into for-linus

commit 65139ed99234d8505948cdb7a835452eb5c191f9
Author: David Sterba <dsterba@xxxxxxx>
Date:   Fri Feb 17 12:26:09 2012 +0100

    btrfs: disallow unequal data/metadata blocksize for mixed block
groups

So if you git reset --hard b5d67f6, you'll get rid of the suse code.

Thanks Dave!

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