r=adilger

  choose proper obd to match lock that protects directory content

set o_fid in mdc_obj_create for LBUG in mdt_obj_create
b:5234

- removed not needed lookup_by_path() from lustre_smfs.h
- small cleanups about names of variables to make them coherent with another parts.

This decreased stack consumption substantialy and 16K is enough.

that is unsigned long too.
- expanding all ->valid fields to 64bit, as it is exhaused and there is no free bits.

- more cleanups about EXIT, RETURN, etc.

- removed redundant assignment.

 - add a flag in lustre_msg which indicate message type
 - add a security descriptor into most of messages send to MDSs, which
   include the uid, supplementary groups array etc.
 - add group hash
 - MDS retrive local groups database through upcall, and set/merge with
   which sent from client based on policies (here not totally finished)
 - move root_squash test into separate sanity-sec.sh

2)some cleanup in smfs

r=adilger

 - obd_connect() gets one more arg: connect flags
 - client_connect_import() puts this flag in imp_connect_flags
 - ptlrpc_connect_import() sends this flag to peer
 - mds_lmv_connect() passes OBD_OPT_MDS_CONNECTION when connects to MDS
 - mds_connect() looks for OBD_OPT_MDS_CONNECTION and treats
   connection with this flag as MDS-to-MDS and doesn't try
   to connect to own LMV

 this way connection to LMV gets delayed and MDS reconnects to LMV
 after recovery abort

r=adilger

- replay of unlink. landed for testing on china cluster

more diagnostics.

I haven't re-tested the AMD functionality since landing, but I did
test iozone and dbench to ensure sanity.

b_cmobd_merge should not be used for any further development or testing!

  we need this becase client's timeout should be longer then inter-mds recovery may take
- few ugly hacks that allow client to reconnect with old (before failure took place) UUID
- mds_preprw() and mds_preprw() should use l_dput() instead of f_dput()
- SOCKNAL_IO_TIMEOUT has been set to 20 seconds to make discovery of
  stale connection faster
- lconf generates persisten UUIDs for LMV clients
  it each new LMV uses fresh UUID, then each recovered MDS looks like
  new client (target MDS doesn't recognize it's old because of new UUID).
  if target MDS gets restarted, then it'll find more clients in LAST_RCVD
  then it actually had

  MDS node with gotten ino/generation couple
- warning for cross-node rename in lmv_rename(). it isn't implemented yet.
  hence, we'd like to know when we hit this path
- mdc_getattr_name() accepts empty name in order to support getattr by fid request
- mds_get_parent_child_locked() forces MDS_INODELOCK_LOOKUP to be used
  if found dentry points to another MDS node
- mea_name2idx() and raw_name2idx() checks for last not to be zero and warns if it's

NOTE: this should fix the cause of the test #5 failure

  given ino/generation. we need this for cross-node mkdir() which creates
  inode on remote MDS and adds dir entry then. after that remote MDS may
  fail and primary MDS will have to recreate inode with already gotten
  ino and generation numbers
- export refcounting bug in lmv fixed
- mdc_obj_create() stores just received ino/generation back to request.
  if request replay will happen mdt_obj_create() will use stored numbers
  to recreate lost inode
- mds_filter_recovery_request() accepts OST_CREATE request. need to be
  discussed though
- mdt_obj_create() rewritten to support recovery phase
- needless mds_lmv_disconnect() in mds_fs_cleanup() removed
- possible bug in retrieve_generation_numbers() fixed: it should understand
  cross-node dir entries and put proper mdsnum's

NOTE: please, re-apply extN-wantedi patch

cMD_PRE_CMD_SMASH are the relevant tags.

The pre-smashing contents of HEAD will be committed by yury into b_cmd as a
merge of the two branches.  These will be stabilized on b_cmd and eventually
landed onto HEAD.

b=2094, b=3138, b=3335, b=3274, b=3293, b=2862, b=3147, b=2350, b=3313
b=3265, b=3315, b=3301, b=3325, b=3329, b=3125, b=3098

b=3284, b=3285, b=3043, b=3236, revert liblustre group changes

b=3043, b=3258, b=3219, b=2410

b=2912/2922, b=3145 too

b=3116, b=3064, b=3107, b=3122, b=1519, 2.6 kernel fixes, group lock fixes

r=zab (partial)

Better handling for OST eviction.  Fixes for clearing the page cache,
imp_invalid handling, and improved import_event interface. Also
includes the tests and fixes added from 2732; one of those tests does
not yet passed so it is excepted for now.

b=2464

Land initial b_bgl fixes.  Includes compile and a couple of swabbing
fixes.

b=2733, b=2773, b=2529, b=2730, b=2819, b=2814, b=2822, b=1450, b=2676, b=2681
b=2817, b=2706, b=2816, b=1987, b=2884, b=1191, b=2809, b=2765, b=2805, b=1972

land b_size on HEAD for 1.2

The principal change here is ptlrpc_connect_import and recovery are
merged and are now asynchronous and are run on ptlrpcd.

This branch also includes fixes for

  2477 (this was the initial bug)

  2355 recover for initial connections

  1934 (b_1934 was merged directly to b_recovery)

  1901 block all file creations until orphan recovery completes

  2423 client remove rq_connection from request struct

  2640 conf-sanity test_5, and proper cleanup in umount when
       cleanup log is not availale.

  2670 recovery timer race

  2532 mdc_close recovey bug

And few others that werent' filed.

        reconnect after user interruption, and avoid leaking an open-count.

        Also, allocate repmsg _before_ reconstructing a close into it.
r=phik

b=1597
r=shaver