ok - applied to tip/x86/core - lets see whether Suresh's system now
works fine with this change.
i'd suggest e820_end_of_ram_pfn().
we now have symmetric behavior on 32-bit and 64-bit, e820_end() returns
the true end of RAM. (64-bit used to return the last PFN mentioned in
the e820 map - i.e. ACPI tables and other end of RAM items could have
been included)
Ingo
--