Nick Piggin wrote: [Sun Nov 06 2005, 07:58:26PM EST]
Hmm, right - in __free_pages_bulk.
Could we make a different call here, or is the full array of bad_range
checks required?
Not the full array. Just the pfn_valid call. Seems CONFIG_HOLES_IN_ZONE is
already in page_alloc.c, perhaps just in __free_pages_bulk as a replacement
for the bad_range call which isn't within a BUG_ON check. It's somewhat of a wart but already there. Otherwise we might want arch_holes_in_zone inline which is only required by ia64 and noop for other arches.
The only place I didn't look closely is the BUG_ON in expand. I'll do that
tomorrow.