b=14265
i=deen
i=scjody

Fix build failes for RHEL3.

i=liang
i=isaac
 - Landing b_usocklnd on HEAD
Bugzilla   : 12302
Description: new userspace socklnd
Details    : Old userspace tcpnal that resided in lnet/ulnds/socklnd replaced
             with new one - usocklnd.

*  fixed some initialiser struct formatting in socklnd_lib-linux.c

Bump keepalive interval to send fewer total keepalives.
Trying to determine if this is cause of BG/L EMAC "flood" issue at
cleanup time.

*    minor socknal cleanups

- don't LASSERT in ll_release on NULL lld with NFS export (4655, 5760)
- hold NS lock when calling handle_ast_error->del_waiting_lock (5746)
- fix setattr mtime regression from lovcleanup merge (4829, 5669)
- workaround for 2.6 crash in ll_unhash_aliases (5687, 5210)
- small ext3 extents cleanups and fixes (5733)
- improved mballoc code, several small races and bugs fixed (5733, 5638)
- kernel version 43: fix remove_suid bugs in both 2.4 and 2.6 (5695)
- avoid needless client->OST connect, fix handle mismatch (5317)
- fix lustre/lustrefs init scripts for SuSE (patch from Scali, 5702)
- don't hold the pinger_sem in ptlrpc_pinger_sending_on_import
- change obd_increase_kms to obd_adjust_kms (up or down) (5654)

	  - removed CFS_DECL_SPIN*
	    just use 'spinlock_t' and initialise with spin_lock_init()

	  - removed CFS_DECL_MUTEX*
	    just use 'struct semaphore' and initialise with init_mutex()

	  - removed CFS_DECL_RWSEM*
	    just use 'struct rw_semaphore' and initialise with init_rwsem()

	  - renamed cfs_sleep_chan -> cfs_waitq
	            cfs_sleep_link -> cfs_waitlink

          - fixed race in linux version of arch-independent socknal
  	    (the ENOMEM/EAGAIN decision).

	  - Didn't fix problems in Darwin version of arch-independent socknal
 	    (resetting socket callbacks, eager ack hack, ENOMEM/EAGAIN decision)

	  - removed libcfs types from non-socknal header files (only some types
	    in the header files had been changed; the .c files hadn't been
	    updated at all).

	  - Updated lustre b1_4 to match

Two compile fixes for recent obdfilter changes.
Compile warning fixups for 2.6/x86_64.
b=5545

do the reasonable thing and enable multi-frag only on non-HIGHMEM platforms,
where there's no kmap-deadlock risk.

- keepalives enabled; starting after 30 seconds, and sending up to
  10 keepalives, every 2 seconds
- enable multi-frag I/O to improve IP-over-IB performance, with the
  vanishingly small risk of kmap-induced deadlock

- remove ancient Sandia copyright statements from Portals
  (our tree has been rewritten so many times that it's not even
   recognizably the same software)
- Remove the LLNL copyright from the qswnal (for the same reason)
- Remove our copyright from header files which only implement
  structures and prototypes, which are widely recognized as
  non-copyrightable

*   Removed S_???NAL in favour of a single S_NAL

*   NB liblustre needs attention

b=bug4165(pid)
r=Peter

- Assign the constant number to kernel portals pid,
  this constant is LUSTRE_SRV_PTL_PID(999999) defined
  in lustre_idl.h;
  In the user mode(liblustre), use the getpid()

- Modify the struct ptlrpc_peer's field peer_nid
  from ptl_nid_t to ptl_process_id_t, which include pid.
  Modify the related code with peer_nid.

- Change the snprintf remote_uuid.uuid to
  include pid info in target_handle_connect;

- In ksocknal_api_startup, use LUSTRE_SRV_PTL_PID to PtlNIInit.

- In tcpnal_init, b->lib_nal->libnal_ni.ni_pid.pid=0;
  This line should be deleted.

- In ptlrpc_get_connection, add the pid compare
  when iterate the conn_list & conn_unused_list

- Change request_in_callback while assign req->rq_peer.peer_id

- In ptlrpc_uuid_to_peer, give default value
  (LUSTRE_SRV_PTL_PID) to peer->peer_id.pid

- Change each entry to call PtlNIInit/PtlPut

- Change the entry to call PtlMEAttach,
  use c_peer value replace PTL_PID_ANY , it is the most important!

- b1_4 is basically the parent of HEAD, because that's the direction
  that changes flow, as strange as that sounds.  so there's a
  HEAD_BASE tag which sits on b1_4

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

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.

    Superbug 2871:
    Sub-bugs fixed 2869 2870 3034 3036 3037 3035 3038 3135

r=phil

land b_cray_portals_merge on HEAD.  This consists of four families of changes:

1) Eric's work to change some APIs in Lustre's portals to come into line with
   Cray's portals and the spec.

2) add --with-cray-portals= to the build and use it to cut up the build
   when we're building lustre against external includes for Cray's portals.

3) Move some facilities from portals.o into libcfs.o so that Lustre can
   consume them when it is running against Cray's portals.o.

4) Fix up the liblustre build.  These changes have also made it to
   b_cray_delivery recently.

b=2464

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

   forwarding interface from a single struct iovec for the message header +
   payload, to passing the message header by pointer, and the message body
   by ptl_kiov_t (== page,offset,size).  This restores portals router
   performance by allowing it to use socknal zero copy again.

*  Applied the socknal connection load balancing fix from Bug 2817

b=2578

  routers (i.e. gateways) without forgetting network topology
  (i.e. adding/deleting routes).

* Socknal and qswnal automatically notify their local router and make an
  upcall when they detect peer death.

* portals router load balances over equivalent routes

* ENETUNREACH returned when a NAL thinks the router isn't loaded.

* Improved socknal network failure detection.

* /proc/sys/socknal/* interface

  - timeout is the socknal I/O timeout (50 by default) in seconds .

  - eager_ack is a boolean (set by default) that enables setting
    TCP_QUICKACK after every incoming message to ensure peer zero-copy
    sends complete quickly.

  - zero_copy is the size (2k by default) in bytes, below which message
    fragments are copied into a socket, rather than using zero-copy sends.
    Setting this above PAGE_SIZE will disable zero copy.

* Socknal autoconnect option to create all peer connections eagerly or not.
  If more than one autoconnect reaches the same peer NID, with the eager
  option (this was the previous default), all connections will be made when
  one is required, otherwise only one at a time will be attempted (the new
  default).  NB socknal still load balances over all established
  connections.

* Generalised portals upcall to "cmd <action> [params]".  Current upcalls
  are...

  - upcall LBUG file fn line

  - upcall ROUTER_NOTIFY <nal> <nid> <timestamp> up|down
    where <nal> is the kernel NAL number (defined in <linux/kp30.h>)
          <nid> is the peer's NID in 0xhex
          <timestamp> is seconds since 1/1/1970

* Added 'lctl --net' option, to make 1-line network lctl commands easy.

* Added Light-weight event tracing.

- remove extN/ files