Fix replay-single.sh and recovery-small.sh so they can run with the
default lmv config or "-f cfg/local.sh" config.

use assignment instead of memset to clear the conn_cnt.

  move primary cowed inode to cowed dir.
2)move MSG_CONNECT_INITIAL check before obd_uuid equal check, otherwise,
  it will bring some problems in cmobd_setup.
3)set imp_conn_cnt to 0 in ptlrpc_disconnect_import

2)some minor fix in smfs cache_hook.

robert's fixes:
 - add CONNECT_INITAL flag for the initial connect message from a client.  When
   server sees an old client reconnect with that flag, then it needs to set the
   handles as if it was a new client.

 - set the initial imp_conn_cnt from the server's conn_cnt. When client
   connects with INITIAL set, the server sends back the old conn_cnt+1.

 - timeout for inter-MDS requests is set to obd_timeout / 2

- mds_open() logs parent/name couple. this helps recovery debugging

1) fix some bugs in create_indirect inode.

1) add IOC_OBD_SNAP_ADD
3) mirror s_flags in sb for P_DIROPS

1)some fix for snap
2)add snap_add command

1)some fix in fsfilt_snap_ext3 and fsfilt_snap_smfs.
2)add smfs_cow.c in smfs.
3)remove snapfs in Makefile.in and configure.in

  #8 from CMD acceptance series

VS: ----------------------------------------------------------------------

  lock on part of dir only
- minor API changes to pass the lock pointer

- configure script sets HAVE_LOOKUP_RAW if kernel has the feature

  then it doesn't take exclusive lock on it; in the meantime another client
  can enlarge directory and 1st client will run splitting w/o exclusive lock

- additional debug message

- test 4a of sanity-lmv.sh to check this case

  this should prevent lookup serialization
- vfs-pdirops-2.4.20-rh.patch exports lock_dir() and unlock_dir() primitives

lock handle which was never used