b=2818

b=2167

a failed object creation instead of in advance.  Otherwise, if there is a
down OST the following one will have twice as many objects allocated on it
(observed with watch-lustre on ALC while running IOR with one OST inactive
on the MDS).

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

 - don't build mpi programs by default, as Bob's suggestion

 - another inode drop fix
 - fix a stupid error of last checkin

 - supress a warning of glimpse callback
 - remove the debug code accidently checked in sometime before

 - don't blindly drop inode
 - fix iget
 - root inode checking

 - b#1179: fix byteorder definition
 - minor code cleanup in liblustre.h

 - fix broken readdir test

 - must update kms after write

 - follow the file size glimpse/probe change

to incorporate change of lib_finalize function in portals/portals/lib-msg.c

merge gmnal code in b_llp2 to HEAD

*  Fixed obdecho persistent object size at 1 MByte.

*  Changed test_brw data verification tests to work on a fixed 4k block
   rather than on page size to allow different page sized servers and
   clients.

   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