The code incorrectly failed memory registration if the buffer was not
page aligned.  Also, the length field is mangled causing the hardware
to think the registration is much larger than it really is.

The fix is to remove the page alignment restriction as well the
incorrect length adjustment.  Also make sure that all buffers after
the first start at a page boundary, and all buffers except the last
end on a page boundary.
Signed-off-by: Don Wood <donald.e.wood@intel.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>