mm, compaction: fast_find_migrateblock() should return pfn in the target zone

At present, pages not in the target zone are added to cc->migratepages
list in isolate_migratepages_block().  As a result, pages may migrate
between nodes unintentionally.

This would be a serious problem for older kernels without commit
a984226f457f849e ("mm: memcontrol: remove the pgdata parameter of
mem_cgroup_page_lruvec"), because it can corrupt the lru list by
handling pages in list without holding proper lru_lock.

Avoid returning a pfn outside the target zone in the case that it is
not aligned with a pageblock boundary.  Otherwise
isolate_migratepages_block() will handle pages not in the target zone.

Link: https://lkml.kernel.org/r/[email protected]
Fixes: 70b44595eafe ("mm, compaction: use free lists to quickly locate a migration source")
Signed-off-by: Rei Yamamoto <[email protected]>
Reviewed-by: Miaohe Lin <[email protected]>
Acked-by: Mel Gorman <[email protected]>
Reviewed-by: Oscar Salvador <[email protected]>
Cc: Don Dutile <[email protected]>
Cc: Wonhyuk Yang <[email protected]>
Cc: Rei Yamamoto <[email protected]>
Cc: <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>

1 files changed, 2 insertions, 0 deletions

diff --git a/mm/compaction.c b/mm/compaction.c
index 0e0ef9e44530..1f89b969c12b 100644
--- a/mm/compaction.c
+++ b/mm/compaction.c
@@ -1848,6 +1848,8 @@ static unsigned long fast_find_migrateblock(struct compact_control *cc)
 
 				update_fast_start_pfn(cc, free_pfn);
 				pfn = pageblock_start_pfn(free_pfn);
+				if (pfn < cc->zone->zone_start_pfn)
+					pfn = cc->zone->zone_start_pfn;
 				cc->fast_search_fail = 0;
 				found_block = true;
 				set_pageblock_skip(freepage);