[RFC] Remove temp_priority

[Martin Bligh <mbligh@google.com>]

[-- Attachment #1: Type: text/plain, Size: 218 bytes --]

This is not tested yet. What do you think?

This patch removes temp_priority, as it is racy. We're setting
prev_priority from it, and yet temp_priority could have been
set back to DEF_PRIORITY by another reclaimer.




[-- Attachment #2: 2.6.18-prio_fix --]
[-- Type: text/plain, Size: 3484 bytes --]

diff -aurpN -X /home/mbligh/.diff.exclude linux-2.6.18/include/linux/mmzone.h 2.6.18-prio_fix/include/linux/mmzone.h
--- linux-2.6.18/include/linux/mmzone.h	2006-09-20 12:24:41.000000000 -0700
+++ 2.6.18-prio_fix/include/linux/mmzone.h	2006-10-17 10:27:48.000000000 -0700
@@ -199,13 +199,9 @@ struct zone {
 	 * under - it drives the swappiness decision: whether to unmap mapped
 	 * pages.
 	 *
-	 * temp_priority is used to remember the scanning priority at which
-	 * this zone was successfully refilled to free_pages == pages_high.
-	 *
-	 * Access to both these fields is quite racy even on uniprocessor.  But
-	 * it is expected to average out OK.
+	 * Access to this field is quite racy even on uniprocessor. It needs
+	 * to be fixed.
 	 */
-	int temp_priority;
 	int prev_priority;
 
 
diff -aurpN -X /home/mbligh/.diff.exclude linux-2.6.18/mm/page_alloc.c 2.6.18-prio_fix/mm/page_alloc.c
--- linux-2.6.18/mm/page_alloc.c	2006-09-20 12:24:42.000000000 -0700
+++ 2.6.18-prio_fix/mm/page_alloc.c	2006-10-17 10:25:55.000000000 -0700
@@ -2016,7 +2016,7 @@ static void __meminit free_area_init_cor
 		zone->zone_pgdat = pgdat;
 		zone->free_pages = 0;
 
-		zone->temp_priority = zone->prev_priority = DEF_PRIORITY;
+		zone->prev_priority = DEF_PRIORITY;
 
 		zone_pcp_init(zone);
 		INIT_LIST_HEAD(&zone->active_list);
diff -aurpN -X /home/mbligh/.diff.exclude linux-2.6.18/mm/vmscan.c 2.6.18-prio_fix/mm/vmscan.c
--- linux-2.6.18/mm/vmscan.c	2006-09-20 12:24:42.000000000 -0700
+++ 2.6.18-prio_fix/mm/vmscan.c	2006-10-17 10:25:24.000000000 -0700
@@ -934,7 +934,6 @@ static unsigned long shrink_zones(int pr
 		if (!cpuset_zone_allowed(zone, __GFP_HARDWALL))
 			continue;
 
-		zone->temp_priority = priority;
 		if (zone->prev_priority > priority)
 			zone->prev_priority = priority;
 
@@ -984,7 +983,6 @@ unsigned long try_to_free_pages(struct z
 		if (!cpuset_zone_allowed(zone, __GFP_HARDWALL))
 			continue;
 
-		zone->temp_priority = DEF_PRIORITY;
 		lru_pages += zone->nr_active + zone->nr_inactive;
 	}
 
@@ -1021,6 +1019,8 @@ unsigned long try_to_free_pages(struct z
 		if (sc.nr_scanned && priority < DEF_PRIORITY - 2)
 			blk_congestion_wait(WRITE, HZ/10);
 	}
+	if (priority < 0)
+		priority = 0;
 out:
 	for (i = 0; zones[i] != 0; i++) {
 		struct zone *zone = zones[i];
@@ -1028,7 +1028,7 @@ out:
 		if (!cpuset_zone_allowed(zone, __GFP_HARDWALL))
 			continue;
 
-		zone->prev_priority = zone->temp_priority;
+		zone->prev_priority = priority;
 	}
 	return ret;
 }
@@ -1075,12 +1075,6 @@ loop_again:
 	sc.may_writepage = !laptop_mode;
 	count_vm_event(PAGEOUTRUN);
 
-	for (i = 0; i < pgdat->nr_zones; i++) {
-		struct zone *zone = pgdat->node_zones + i;
-
-		zone->temp_priority = DEF_PRIORITY;
-	}
-
 	for (priority = DEF_PRIORITY; priority >= 0; priority--) {
 		int end_zone = 0;	/* Inclusive.  0 = ZONE_DMA */
 		unsigned long lru_pages = 0;
@@ -1140,7 +1134,6 @@ scan:
 			if (!zone_watermark_ok(zone, order, zone->pages_high,
 					       end_zone, 0))
 				all_zones_ok = 0;
-			zone->temp_priority = priority;
 			if (zone->prev_priority > priority)
 				zone->prev_priority = priority;
 			sc.nr_scanned = 0;
@@ -1182,11 +1175,13 @@ scan:
 		if (nr_reclaimed >= SWAP_CLUSTER_MAX)
 			break;
 	}
+	if (priority < 0)
+		priority = 0;
 out:
 	for (i = 0; i < pgdat->nr_zones; i++) {
 		struct zone *zone = pgdat->node_zones + i;
 
-		zone->prev_priority = zone->temp_priority;
+		zone->prev_priority = priority;
 	}
 	if (!all_zones_ok) {
 		cond_resched();