block allocations for the refcount btree

[Christoph Hellwig]

On Wed, Mar 02, 2016 at 01:21:01PM -0800, Darrick J. Wong wrote:
> Ok.  I think the problem is that making changes to the refcount btree eats
> up our entire reservation in certain cases.  Can you try the following bandaid?
> This should give us enough room to handle splitting the btree at both ends
> of a range that we're refcount-changing.

This seems to work in general.  I ran into one log related hang when
running -g auto on nfs, but it's not been reproducible so far.