Rather than forcing us to take the inode->i_lock just in order to bump
the number.
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>