[patch] fix CONFIG_DEBUG_HIGHMEM assert in copy_page_range()

[Ingo Molnar <mingo@elte.hu>]

copy_page_range() triggers a CONFIG_DEBUG_HIGHMEM assert, because if a
preemption occurs right at a PMD_SIZE boundary the kmap/kunmap doesnt
match. This isnt a real problem because atomic_kunmap doesnt do anything
if CONFIG_DEBUG_HIGHMEM is not set and the atomic_map's are still
correct, but it's ugly nevertheless.

tested patch is against 2.6.9-rc2-mm4.

	Ingo

Signed-off-by: Ingo Molnar <mingo@elte.hu>

--- linux/mm/memory.c.orig	
+++ linux/mm/memory.c	
@@ -337,14 +337,6 @@ skip_copy_pte_range:
 				set_pte(dst_pte, pte);
 				page_dup_rmap(page);
 cont_copy_pte_range_noset:
-				address += PAGE_SIZE;
-				if (address >= end) {
-					pte_unmap_nested(src_pte);
-					pte_unmap(dst_pte);
-					goto out_unlock;
-				}
-				src_pte++;
-				dst_pte++;
 				/*
 				 * We are holding two locks at this point -
 				 * any of them could generate latencies in
@@ -364,6 +356,14 @@ cont_copy_pte_range_noset:
 					dst_pte = pte_offset_map(dst_pmd, address);
 					src_pte = pte_offset_map_nested(src_pmd, address);
 				}
+				address += PAGE_SIZE;
+				if (address >= end) {
+					pte_unmap_nested(src_pte);
+					pte_unmap(dst_pte);
+					goto out_unlock;
+				}
+				src_pte++;
+				dst_pte++;
 			} while ((unsigned long)src_pte & PTE_TABLE_MASK);
 			pte_unmap_nested(src_pte-1);
 			pte_unmap(dst_pte-1);