mm/compaction: fix wrong order check in compact_finished()
authorJoonsoo Kim <iamjoonsoo.kim@lge.com>
Thu, 12 Feb 2015 22:59:50 +0000 (14:59 -0800)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Wed, 18 Mar 2015 12:22:28 +0000 (13:22 +0100)
commit 372549c2a3778fd3df445819811c944ad54609ca upstream.

What we want to check here is whether there is highorder freepage in buddy
list of other migratetype in order to steal it without fragmentation.
But, current code just checks cc->order which means allocation request
order.  So, this is wrong.

Without this fix, non-movable synchronous compaction below pageblock order
would not stopped until compaction is complete, because migratetype of
most pageblocks are movable and high order freepage made by compaction is
usually on movable type buddy list.

There is some report related to this bug. See below link.

  http://www.spinics.net/lists/linux-mm/msg81666.html

Although the issued system still has load spike comes from compaction,
this makes that system completely stable and responsive according to his
report.

stress-highalloc test in mmtests with non movable order 7 allocation
doesn't show any notable difference in allocation success rate, but, it
shows more compaction success rate.

Compaction success rate (Compaction success * 100 / Compaction stalls, %)
18.47 : 28.94

Fixes: 1fb3f8ca0e92 ("mm: compaction: capture a suitable high-order page immediately when it is made available")
Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: David Rientjes <rientjes@google.com>
Cc: Rik van Riel <riel@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
mm/compaction.c

index fb797a32362fecc011c2f2fcf57cb47b875c854f..eeaaa929d1debf611c1cd87adb3e949c701580bb 100644 (file)
@@ -881,7 +881,7 @@ static int compact_finished(struct zone *zone,
                        return COMPACT_PARTIAL;
 
                /* Job done if allocation would set block type */
-               if (cc->order >= pageblock_order && area->nr_free)
+               if (order >= pageblock_order && area->nr_free)
                        return COMPACT_PARTIAL;
        }