a reflink fix

[Darrick J. Wong]

On Wed, Jan 13, 2016 at 05:13:48PM +0100, Christoph Hellwig wrote:
> This fixes the issues I saw with generic/168 over NFS.  Turns out it
> wasn't anything specific to that particular test cases, we just needed
> a sufficiently aged refcount btree to trigger it.

Thanks for the patch; I've worked the fixes into the overall patchset.

I also wrote a quick reproducer which will appear as xfs/807 in my next
xfstests drop.

--D