block allocations for the refcount btree

[Christoph Hellwig]

On Sat, Feb 13, 2016 at 01:33:10PM +1100, Dave Chinner wrote:
> > allocate an extent (or rather just a single block) in xfs_alloc_ag_vextent
> > as called from xfs_refcountbt_alloc_block.  The reservation helps us to
> > ensure this block is always available, but we still need to account for
> > that in xfs_trans_reserve(), which we currently don't do for itruncate
> > transactions.  
> 
> Ok, so we may have two different issues with a similar failure
> symptom. As it is, I don't think this is a show stopper - we're
> expecting to find these sorts of issues as we go along (hence the
> experimental tag on the feature) and I think, at this point, getting
> review and an initial merge done is more important...

This triggers 100% reproducible over NFS, and as outlined I'm
also pretty sure about the root cause.  I don't think this is something
to be ignored.