ARM: mm: don't free memmap entries in the middle of a MAX_ORDER_NR_PAGES block

On some machines, nodes do not always start on MAX_ORDER_NR_PAGES boundaries. The rest of the VM assumes that pages within a MAX_ORDER_NR_PAGES range always have valid memmap entries. So, we round down the bank_start in free_unused_memmap to ensure that there are valid page structs for all pages within a MAX_ORDER_NR_PAGES block of pages.

---
 arch/arm/mm/init.c |    3 ++-
 1 files changed, 2 insertions(+), 1 deletions(-)

index cddd684..1290fbe 100644
@@ -424,7 +424,8 @@ static void __init free_unused_memmap(struct meminfo *mi)
 	for_each_bank(i, mi) {
 		struct membank *bank = &mi->bank[i];
 
-		bank_start = bank_pfn_start(bank);
+		bank_start = round_down(bank_pfn_start(bank),
+					MAX_ORDER_NR_PAGES);
 
 		/*
 		 * If we had a previous bank, and there is a space

1.7.3.1