xfs_growfs causing "Metadata corruption detected at xfs_agf_read_verify"

xfs_growfs causing "Metadata corruption detected at xfs_agf_read_verify"

[Richard W.M. Jones]

With recent kernels (or maybe xfsprogs?) we've noticed strange
filesystem corruption after using xfs_growfs.

Examples:

https://bugzilla.redhat.com/show_bug.cgi?id=1315895 (on aarch64)
https://www.redhat.com/archives/libguestfs/2016-March/msg00113.html (x86_64)
https://rwmj.wordpress.com/2015/11/04/virt-builder-fedora-23-image/#comment-15668 (x86_64)

Do these errors look familiar?  I don't yet have a simple test case,
still working on that.

Rich.

-- 
Richard Jones, Virtualization Group, Red Hat http://people.redhat.com/~rjones
Read my programming and virtualization blog: http://rwmj.wordpress.com
virt-builder quickly builds VMs from scratch
http://libguestfs.org/virt-builder.1.html

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

Re: xfs_growfs causing "Metadata corruption detected at xfs_agf_read_verify"

[Eric Sandeen]

On 3/17/16 5:48 PM, Richard W.M. Jones wrote:
> 
> With recent kernels (or maybe xfsprogs?) we've noticed strange
> filesystem corruption after using xfs_growfs.
> 
> Examples:
> 
> https://bugzilla.redhat.com/show_bug.cgi?id=1315895 (on aarch64)
> https://www.redhat.com/archives/libguestfs/2016-March/msg00113.html (x86_64)
> https://rwmj.wordpress.com/2015/11/04/virt-builder-fedora-23-image/#comment-15668 (x86_64)
> 
> Do these errors look familiar?  I don't yet have a simple test case,
> still working on that.

try:
http://oss.sgi.com/pipermail/xfs/2016-March/047683.html

-Eric

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

Re: xfs_growfs causing "Metadata corruption detected at xfs_agf_read_verify"

[Richard W.M. Jones]

On Thu, Mar 17, 2016 at 05:55:52PM -0500, Eric Sandeen wrote:
> On 3/17/16 5:48 PM, Richard W.M. Jones wrote:
> > 
> > With recent kernels (or maybe xfsprogs?) we've noticed strange
> > filesystem corruption after using xfs_growfs.
> > 
> > Examples:
> > 
> > https://bugzilla.redhat.com/show_bug.cgi?id=1315895 (on aarch64)
> > https://www.redhat.com/archives/libguestfs/2016-March/msg00113.html (x86_64)
> > https://rwmj.wordpress.com/2015/11/04/virt-builder-fedora-23-image/#comment-15668 (x86_64)
> > 
> > Do these errors look familiar?  I don't yet have a simple test case,
> > still working on that.
> 
> try:
> http://oss.sgi.com/pipermail/xfs/2016-March/047683.html

I've tried this patch now, but I still can't get a solid reproducer
(one that doesn't involve using aarch64).

I just tried to create a filesystem on kernel 4.6.0, grow it on 4.6.0,
then copy it to another machine with 4.4.4.  However that didn't
reproduce the bug, so I couldn't verify if the patch fixes the bug.
(Maybe 4.4.4 is too new?)

Can you help me to understand under what circumstances this bug would
appear, as far as you understand what's happening?

Rich.

-- 
Richard Jones, Virtualization Group, Red Hat http://people.redhat.com/~rjones
Read my programming and virtualization blog: http://rwmj.wordpress.com
virt-p2v converts physical machines to virtual machines.  Boot with a
live CD or over the network (PXE) and turn machines into KVM guests.
http://libguestfs.org/virt-v2v

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

Re: xfs_growfs causing "Metadata corruption detected at xfs_agf_read_verify"

[Dave Chinner]

On Fri, Mar 18, 2016 at 12:31:07PM +0000, Richard W.M. Jones wrote:
> On Thu, Mar 17, 2016 at 05:55:52PM -0500, Eric Sandeen wrote:
> > On 3/17/16 5:48 PM, Richard W.M. Jones wrote:
> > > 
> > > With recent kernels (or maybe xfsprogs?) we've noticed strange
> > > filesystem corruption after using xfs_growfs.
> > > 
> > > Examples:
> > > 
> > > https://bugzilla.redhat.com/show_bug.cgi?id=1315895 (on aarch64)
> > > https://www.redhat.com/archives/libguestfs/2016-March/msg00113.html (x86_64)
> > > https://rwmj.wordpress.com/2015/11/04/virt-builder-fedora-23-image/#comment-15668 (x86_64)
> > > 
> > > Do these errors look familiar?  I don't yet have a simple test case,
> > > still working on that.
> > 
> > try:
> > http://oss.sgi.com/pipermail/xfs/2016-March/047683.html
> 
> I've tried this patch now, but I still can't get a solid reproducer
> (one that doesn't involve using aarch64).
> 
> I just tried to create a filesystem on kernel 4.6.0, grow it on 4.6.0,
> then copy it to another machine with 4.4.4.  However that didn't
> reproduce the bug, so I couldn't verify if the patch fixes the bug.
> (Maybe 4.4.4 is too new?)

4.4.4 has the fix already in it, through the stable kernel backport
process. i.e. the original kernel commit had this in it:

	 cc: <stable@vger.kernel.org> # 3.10 - 4.4

And so the fix is automatically appearing in stable kernels as they
are released. Hence the number of current supported kernels you can
reproduce this on is shrinking....

> Can you help me to understand under what circumstances this bug would
> appear, as far as you understand what's happening?

Take a stock 4.5.0 kernel, mkfs, mount, growfs, unmount go back to a
stock 4.4.0 kernel, mount (fails), xfs_repair v4.3.0 (warns, fixes),
xfs_repair v4.5.0 (clean, fixes).

Cheers,

Dave.

-- 
Dave Chinner
david@fromorbit.com

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs