memory hotplug: make pages from movable zone always isolatable
authorShaohua Li <shaohua.li@intel.com>
Tue, 22 Sep 2009 00:01:18 +0000 (17:01 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Tue, 22 Sep 2009 14:17:26 +0000 (07:17 -0700)
Pages on movable zone have two types, MIGRATE_MOVABLE and MIGRATE_RESERVE,
both them can be movable, because only movable memory allocation can get
pages from movable zone.  This makes pages in movable zone always be able
to migrate.

Signed-off-by: Shaohua Li <shaohua.li@intel.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Christoph Lameter <cl@linux-foundation.org>
Cc: Yakui Zhao <yakui.zhao@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
mm/page_alloc.c

index 8a83027117256daf4070f6445cb66a2fc6c34b7d..7b4d4e4ea64f5e5ea7d193b8a8b4bde8e3642761 100644 (file)
@@ -4906,13 +4906,16 @@ int set_migratetype_isolate(struct page *page)
        struct zone *zone;
        unsigned long flags;
        int ret = -EBUSY;
+       int zone_idx;
 
        zone = page_zone(page);
+       zone_idx = zone_idx(zone);
        spin_lock_irqsave(&zone->lock, flags);
        /*
         * In future, more migrate types will be able to be isolation target.
         */
-       if (get_pageblock_migratetype(page) != MIGRATE_MOVABLE)
+       if (get_pageblock_migratetype(page) != MIGRATE_MOVABLE &&
+           zone_idx != ZONE_MOVABLE)
                goto out;
        set_pageblock_migratetype(page, MIGRATE_ISOLATE);
        move_freepages_block(zone, page, MIGRATE_ISOLATE);