mm: kswapd: use the order that kswapd was reclaiming at for sleeping_prematurely()
authorMel Gorman <mel@csn.ul.ie>
Thu, 13 Jan 2011 23:46:22 +0000 (15:46 -0800)
committerLinus Torvalds <torvalds@linux-foundation.org>
Fri, 14 Jan 2011 01:32:37 +0000 (17:32 -0800)
Before kswapd goes to sleep, it uses sleeping_prematurely() to check if
there was a race pushing a zone below its watermark.  If the race
happened, it stays awake.  However, balance_pgdat() can decide to reclaim
at order-0 if it decides that high-order reclaim is not working as
expected.  This information is not passed back to sleeping_prematurely().
The impact is that kswapd remains awake reclaiming pages long after it
should have gone to sleep.  This patch passes the adjusted order to
sleeping_prematurely and uses the same logic as balance_pgdat to decide if
it's ok to go to sleep.

Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Reviewed-by: Minchan Kim <minchan.kim@gmail.com>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Reviewed-by: Eric B Munson <emunson@mgebm.net>
Cc: Simon Kirby <sim@hostway.ca>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Shaohua Li <shaohua.li@intel.com>
Cc: Dave Hansen <dave@linux.vnet.ibm.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
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/vmscan.c

index d3488828331ac7411382b56c00180b17907e9f51..46711f080f3830fa0bb5a53bb33884f47066aa28 100644 (file)
@@ -2227,7 +2227,7 @@ static bool pgdat_balanced(pg_data_t *pgdat, unsigned long balanced_pages,
 }
 
 /* is kswapd sleeping prematurely? */
-static int sleeping_prematurely(pg_data_t *pgdat, int order, long remaining)
+static bool sleeping_prematurely(pg_data_t *pgdat, int order, long remaining)
 {
        int i;
        unsigned long balanced = 0;
@@ -2237,7 +2237,7 @@ static int sleeping_prematurely(pg_data_t *pgdat, int order, long remaining)
        if (remaining)
                return 1;
 
-       /* If after HZ/10, a zone is below the high mark, it's premature */
+       /* Check the watermark levels */
        for (i = 0; i < pgdat->nr_zones; i++) {
                struct zone *zone = pgdat->node_zones + i;
 
@@ -2269,7 +2269,7 @@ static int sleeping_prematurely(pg_data_t *pgdat, int order, long remaining)
  * For kswapd, balance_pgdat() will work across all this node's zones until
  * they are all at high_wmark_pages(zone).
  *
- * Returns the number of pages which were actually freed.
+ * Returns the final order kswapd was reclaiming at
  *
  * There is special handling here for zones which are full of pinned pages.
  * This can happen if the pages are all mlocked, or if they are all used by
@@ -2532,7 +2532,13 @@ out:
                }
        }
 
-       return sc.nr_reclaimed;
+       /*
+        * Return the order we were reclaiming at so sleeping_prematurely()
+        * makes a decision on the order we were last reclaiming at. However,
+        * if another caller entered the allocator slow path while kswapd
+        * was awake, order will remain at the higher level
+        */
+       return order;
 }
 
 static void kswapd_try_to_sleep(pg_data_t *pgdat, int order)
@@ -2659,7 +2665,7 @@ static int kswapd(void *p)
                 */
                if (!ret) {
                        trace_mm_vmscan_kswapd_wake(pgdat->node_id, order);
-                       balance_pgdat(pgdat, order, classzone_idx);
+                       order = balance_pgdat(pgdat, order, classzone_idx);
                }
        }
        return 0;