vmscan: only defer compaction for failed order and higher

Currently a failed order-9 (transparent hugepage) compaction can lead to
memory compaction being temporarily disabled for a memory zone.  Even if
we only need compaction for an order 2 allocation, eg.  for jumbo frames
networking.

The fix is relatively straightforward: keep track of the highest order at
which compaction is succeeding, and only defer compaction for orders at
which compaction is failing.

Signed-off-by: Rik van Riel <[email protected]>
Cc: Andrea Arcangeli <[email protected]>
Acked-by: Mel Gorman <[email protected]>
Cc: Johannes Weiner <[email protected]>
Cc: Minchan Kim <[email protected]>
Cc: KOSAKI Motohiro <[email protected]>
Cc: Hillf Danton <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Author: Rik van Riel

include/linux/compaction.h

@@ -34,20 +34,26 @@ extern unsigned long compaction_suitable(struct zone *zone, int order);
  * allocation success. 1 << compact_defer_limit compactions are skipped up
  * to a limit of 1 << COMPACT_MAX_DEFER_SHIFT
  */
-static inline void defer_compaction(struct zone *zone)
+static inline void defer_compaction(struct zone *zone, int order)
 {
 	zone->compact_considered = 0;
 	zone->compact_defer_shift++;
 
+	if (order < zone->compact_order_failed)
+		zone->compact_order_failed = order;
+
 	if (zone->compact_defer_shift > COMPACT_MAX_DEFER_SHIFT)
 		zone->compact_defer_shift = COMPACT_MAX_DEFER_SHIFT;
 }
 
 /* Returns true if compaction should be skipped this time */
-static inline bool compaction_deferred(struct zone *zone)
+static inline bool compaction_deferred(struct zone *zone, int order)
 {
 	unsigned long defer_limit = 1UL << zone->compact_defer_shift;
 
+	if (order < zone->compact_order_failed)
+		return false;
+
 	/* Avoid possible overflow */
 	if (++zone->compact_considered > defer_limit)
 		zone->compact_considered = defer_limit;
@@ -73,11 +79,11 @@ static inline unsigned long compaction_suitable(struct zone *zone, int order)
 	return COMPACT_SKIPPED;
 }
 
-static inline void defer_compaction(struct zone *zone)
+static inline void defer_compaction(struct zone *zone, int order)
 {
 }
 
-static inline bool compaction_deferred(struct zone *zone)
+static inline bool compaction_deferred(struct zone *zone, int order)
 {
 	return 1;
 }

include/linux/mmzone.h

@@ -365,6 +365,7 @@ struct zone {
 	 */
 	unsigned int		compact_considered;
 	unsigned int		compact_defer_shift;
+	int			compact_order_failed;
 #endif
 
 	ZONE_PADDING(_pad1_)

mm/compaction.c

@@ -695,9 +695,19 @@ static int __compact_pgdat(pg_data_t *pgdat, struct compact_control *cc)
 		INIT_LIST_HEAD(&cc->freepages);
 		INIT_LIST_HEAD(&cc->migratepages);
 
-		if (cc->order < 0 || !compaction_deferred(zone))
+		if (cc->order < 0 || !compaction_deferred(zone, cc->order))
 			compact_zone(zone, cc);
 
+		if (cc->order > 0) {
+			int ok = zone_watermark_ok(zone, cc->order,
+						low_wmark_pages(zone), 0, 0);
+			if (ok && cc->order > zone->compact_order_failed)
+				zone->compact_order_failed = cc->order + 1;
+			/* Currently async compaction is never deferred. */
+			else if (!ok && cc->sync)
+				defer_compaction(zone, cc->order);
+		}
+
 		VM_BUG_ON(!list_empty(&cc->freepages));
 		VM_BUG_ON(!list_empty(&cc->migratepages));
 	}

mm/page_alloc.c

@@ -1990,7 +1990,7 @@ __alloc_pages_direct_compact(gfp_t gfp_mask, unsigned int order,
 	if (!order)
 		return NULL;
 
-	if (compaction_deferred(preferred_zone)) {
+	if (compaction_deferred(preferred_zone, order)) {
 		*deferred_compaction = true;
 		return NULL;
 	}
@@ -2012,6 +2012,8 @@ __alloc_pages_direct_compact(gfp_t gfp_mask, unsigned int order,
 		if (page) {
 			preferred_zone->compact_considered = 0;
 			preferred_zone->compact_defer_shift = 0;
+			if (order >= preferred_zone->compact_order_failed)
+				preferred_zone->compact_order_failed = order + 1;
 			count_vm_event(COMPACTSUCCESS);
 			return page;
 		}
@@ -2028,7 +2030,7 @@ __alloc_pages_direct_compact(gfp_t gfp_mask, unsigned int order,
 		 * defer if the failure was a sync compaction failure.
 		 */
 		if (sync_migration)
-			defer_compaction(preferred_zone);
+			defer_compaction(preferred_zone, order);
 
 		cond_resched();
 	}

mm/vmscan.c

@@ -2198,7 +2198,7 @@ static inline bool compaction_ready(struct zone *zone, struct scan_control *sc)
 	 * If compaction is deferred, reclaim up to a point where
 	 * compaction will have a chance of success when re-enabled
 	 */
-	if (compaction_deferred(zone))
+	if (compaction_deferred(zone, sc->order))
 		return watermark_ok;
 
 	/* If compaction is not ready to start, keep reclaiming */