mm/compaction: change the timing to check to drop the spinlock
authorJoonsoo Kim <iamjoonsoo.kim@lge.com>
Mon, 7 Apr 2014 22:37:05 +0000 (15:37 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Mon, 7 Apr 2014 23:35:51 +0000 (16:35 -0700)
It is odd to drop the spinlock when we scan (SWAP_CLUSTER_MAX - 1) th
pfn page.  This may results in below situation while isolating
migratepage.

1. try isolate 0x0 ~ 0x200 pfn pages.
2. When low_pfn is 0x1ff, ((low_pfn+1) % SWAP_CLUSTER_MAX) == 0, so drop
   the spinlock.
3. Then, to complete isolating, retry to aquire the lock.

I think that it is better to use SWAP_CLUSTER_MAX th pfn for checking the
criteria about dropping the lock.  This has no harm 0x0 pfn, because, at
this time, locked variable would be false.

Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Mel Gorman <mgorman@suse.de>
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>
mm/compaction.c

index 3a1828541dc0ba4861f0827e1479d4fef5cb31cf..0eb9f99196ce7f03ca368b099cf79ba5f6b10515 100644 (file)
@@ -487,7 +487,7 @@ isolate_migratepages_range(struct zone *zone, struct compact_control *cc,
        cond_resched();
        for (; low_pfn < end_pfn; low_pfn++) {
                /* give a chance to irqs before checking need_resched() */
-               if (locked && !((low_pfn+1) % SWAP_CLUSTER_MAX)) {
+               if (locked && !(low_pfn % SWAP_CLUSTER_MAX)) {
                        if (should_release_lock(&zone->lru_lock)) {
                                spin_unlock_irqrestore(&zone->lru_lock, flags);
                                locked = false;