32 bit compatibility code sometimes needs to copy unaligned data across
kernel/user boundary and currently there is no architecture independent API
to do it.

(1) Introduce new APIs __{get,put}_user_unaligned. These APIs are
    necessary because the optimal way to copy unaligned data across
    kernel/user boundary is different on different architectures.
    Some architectures don't even care about alignment.
    On some __put_user is faster than __copy_to_user for small sizes.
(2) Optimize __{get,put}_user_unaligned for ia64, x86-64, s390, ppc64.
(3) Fix compat_filldir64() which is broken on big-endian machines

Thanks to Arnd Bergmann <arnd@arndb.de> for his help.
Signed-off-by: Gordon Jin <gordon.jin@intel.com>
Signed-off-by: Arun Sharma <arun.sharma@intel.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>