fcmh generation problems

The method I’m exploring now is one where the clients will flush their bmap caches and release bdbufs for bmaps belonging to the truncated FID. We must ensure that no async I/O’s issued prior to truncate are sent to sliod before acquiring new bdbufs. Therefore we must lock the fcmh and iteratively free its bmaps prior to sending any truncate RPC to the MDS.