a) Flags2 left little-endian. SMBFLG2_... definitions switched to  
cpu_to_le16(...), which kills the need of conversions in ->Flags2.
	b) FILE_SYSTEM_UNIX_INFO (never used anywhere) is left little-endian;
when users appear, they can convert themselves.
	c) bugfix: in CIFSSessSetup() we used SecurityBlobLength of response
without conversion.  That would screw bigendian clients with servers that
give wcnt=4.
	d) bugfix: in CIFSNTLMSSPNegotiateSessSetup() we use 
SecurityBlob2->NegotiateFlags without conversion.  Again, problem on
big-endian.
	e) bugfix: SecurityBlob->MessageType is *not* host-endian.
	f) bugfix: ->sendmsg() expects its last argument to be equal to sum
of iovec lengths, so the value passed to kernel_sendmsg()/sock_sendmsg()
would better be correct...
Signed-off-by: Al Viro <viro@parcelfarce.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>