Hello all,
  This patch is an attempt to add a numa ware add_memory functionality
to x86_64 using CONFIG_SPARSEMEM.  The add memory function today just
grabs the pgdat from node 0 and adds the memory there.  On a numa system
this is functional but not optimal/correct. 

  The SRAT can expose future memory locality.  This information is
already tracked by the nodes_add data structure (it keeps the
memory/node locality information) from the SRAT code.  The code in
srat.c is built around RESERVE_HOTADD.  This patch is a little subtle in
the way it uses the existing code for use with sparsemem.  Perhaps
acpi_numa_memory_affinity_init needs a larger refactor to fit both
RESERVE_HOTADD and sparsemem.  

  This patch still hotadd_percent as a flag to the whole srat parsing
code to disable and contain broken bios.  It's functionality is retained
and an on off switch to sparsemem hot-add.  Without changing the safety
mechanisms build into the current SRAT code I have provided a path for
the sparsemem hot-add path to get to the nodes_add data for use at
runtime. 

  This is a 1st run at the patch, it works with 2.6.17

Signed-off-by:  Keith Mannthey <kmannth@us.ibm.com>