With this change, there's some extra fuzz introduced that a max_addr
specification will get rounded down to a granule boundary and memory
quantity, when using mem=, will be within a granule size of the
requested amount.  Let me know if anyone finds more problems with it.
Signed-off-by: Alex Williamson <alex.williamson@hp.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>