Index: linux-2.6.16-rc1-mm1/include/linux/swap.h
===================================================================
--- linux-2.6.16-rc1-mm1.orig/include/linux/swap.h	2006-01-18 13:41:51.000000000 -0800
+++ linux-2.6.16-rc1-mm1/include/linux/swap.h	2006-01-19 17:52:27.000000000 -0800
@@ -178,6 +178,7 @@ extern int vm_swappiness;
 
 #ifdef CONFIG_NUMA
 extern int zone_reclaim_mode;
+extern int zone_reclaim_interval;
 extern int zone_reclaim(struct zone *, gfp_t, unsigned int);
 #else
 #define zone_reclaim_mode 0
Index: linux-2.6.16-rc1-mm1/mm/vmscan.c
===================================================================
--- linux-2.6.16-rc1-mm1.orig/mm/vmscan.c	2006-01-19 17:52:20.000000000 -0800
+++ linux-2.6.16-rc1-mm1/mm/vmscan.c	2006-01-19 18:05:14.000000000 -0800
@@ -1816,17 +1816,20 @@ module_init(kswapd_init)
  *
  * If non-zero call zone_reclaim when the number of free pages falls below
  * the watermarks.
- *
- * In the future we may add flags to the mode. However, the page allocator
- * should only have to check that zone_reclaim_mode != 0 before calling
- * zone_reclaim().
  */
 int zone_reclaim_mode __read_mostly;
 
+#define RECLAIM_OFF 0
+#define RECLAIM_ZONE (1<<0)	/* Run shrink_cache on the zone */
+#define RECLAIM_WRITE (1<<1)	/* Writeout pages during reclaim */
+#define RECLAIM_SWAP (1<<2)	/* Swap pages out during reclaim */
+#define RECLAIM_SLAB (1<<3)   /* Do a global slab shrink if the zone is out of memory */
+
 /*
  * Mininum time between zone reclaim scans
  */
-#define ZONE_RECLAIM_INTERVAL 30*HZ
+int zone_reclaim_interval __read_mostly = 30*HZ;
+
 /*
  * Try to free up some pages from this zone through reclaim.
  */
@@ -1838,7 +1841,7 @@ int zone_reclaim(struct zone *zone, gfp_
 	struct scan_control sc;
 
 	if (time_before(jiffies,
-		zone->last_unsuccessful_zone_reclaim + ZONE_RECLAIM_INTERVAL))
+		zone->last_unsuccessful_zone_reclaim + zone_reclaim_interval))
 			return 0;
 
 	if (!(gfp_mask & __GFP_WAIT) ||
@@ -1848,8 +1851,8 @@ int zone_reclaim(struct zone *zone, gfp_
 		atomic_read(&zone->reclaim_in_progress) > 0)
 			return 0;
 
-	sc.may_writepage = 0;
-	sc.may_swap = 0;
+	sc.may_writepage = !!(zone_reclaim_mode & RECLAIM_WRITE);
+	sc.may_swap = !!(zone_reclaim_mode & RECLAIM_SWAP);
 	sc.nr_scanned = 0;
 	sc.nr_reclaimed = 0;
 	sc.priority = 0;
@@ -1868,7 +1871,15 @@ int zone_reclaim(struct zone *zone, gfp_
 	p->flags |= PF_MEMALLOC;
 	reclaim_state.reclaimed_slab = 0;
 	p->reclaim_state = &reclaim_state;
-	shrink_zone(zone, &sc);
+
+	if (zone_reclaim_mode & (RECLAIM_ZONE|RECLAIM_SWAP|RECLAIM_WRITE))
+		shrink_zone(zone, &sc);
+
+	if (sc.nr_reclaimed < nr_pages && (zone_reclaim_mode & RECLAIM_SLAB)) {
+		if (shrink_slab(sc.nr_scanned, gfp_mask, order) &&
+			zone_watermark_ok(z, order, mark,idx, flags))
+				sc.nr_reclaimed = nr_pages + 1;
+	}
 	p->reclaim_state = NULL;
 	current->flags &= ~PF_MEMALLOC;
 
Index: linux-2.6.16-rc1-mm1/Documentation/sysctl/vm.txt
===================================================================
--- linux-2.6.16-rc1-mm1.orig/Documentation/sysctl/vm.txt	2006-01-18 12:05:38.000000000 -0800
+++ linux-2.6.16-rc1-mm1/Documentation/sysctl/vm.txt	2006-01-19 17:54:33.000000000 -0800
@@ -126,6 +126,16 @@ the high water marks for each per cpu pa
 
 zone_reclaim_mode:
 
+This allows to set more or less agressive forms of reclaiming memory
+when a zone runs out of memory.
+
+This is a ORed together value of
+
+1	= Zone reclaim without swapout
+2	= Zone reclaim writes dirty pages out
+4	= Zone reclaim swaps pages
+8	= Slab reclaim when zone is out of memory
+
 This is set during bootup to 1 if it is determined that pages from
 remote zones will cause a significant performance reduction. The
 page allocator will then reclaim easily reusable pages (those page
@@ -135,6 +145,15 @@ The user can override this setting. It m
 off zone reclaim if the system is used for a file server and all
 of memory should be used for caching files from disk.
 
-It may be beneficial to switch this on if one wants to do zone
-reclaim regardless of the numa distances in the system.
+It may be advisable to set Slab reclaim if the system makes heavy
+use of files and builds up large slab caches and no longer has
+sufficient local memory available. Note that the slab shrink is global
+and may free slab entries on other nodes.
+
+================================================================
+
+zone_reclaim_interval:
 
+The time allowed for off node allocations after zone reclaim
+has failed. This is 30 seconds by default. Time is set in ticks.
+30 seconds are 30*1024 = 30720.
Index: linux-2.6.16-rc1-mm1/kernel/sysctl.c
===================================================================
--- linux-2.6.16-rc1-mm1.orig/kernel/sysctl.c	2006-01-18 13:42:00.000000000 -0800
+++ linux-2.6.16-rc1-mm1/kernel/sysctl.c	2006-01-19 17:52:27.000000000 -0800
@@ -888,6 +888,15 @@ static ctl_table vm_table[] = {
 		.proc_handler	= &proc_dointvec,
 		.strategy	= &zero,
 	},
+	{
+		.ctl_name	= VM_ZONE_RECLAIM_INTERVAL,
+		.procname	= "zone_reclaim_interval",
+		.data		= &zone_reclaim_interval,
+		.maxlen		= sizeof(zone_reclaim_interval),
+		.mode		= 0644,
+		.proc_handler	= &proc_dointvec,
+		.strategy	= &zero,
+	},
 #endif
 	{ .ctl_name = 0 }
 };
Index: linux-2.6.16-rc1-mm1/include/linux/sysctl.h
===================================================================
--- linux-2.6.16-rc1-mm1.orig/include/linux/sysctl.h	2006-01-18 13:42:00.000000000 -0800
+++ linux-2.6.16-rc1-mm1/include/linux/sysctl.h	2006-01-19 17:52:27.000000000 -0800
@@ -184,6 +184,7 @@ enum
 	VM_DROP_PAGECACHE=29,	/* int: nuke lots of pagecache */
 	VM_PERCPU_PAGELIST_FRACTION=30,/* int: fraction of pages in each percpu_pagelist */
 	VM_ZONE_RECLAIM_MODE=31,/* reclaim local zone memory before going off node */
+	VM_ZONE_RECLAIM_INTERVAL=32,/* time period to wait after reclaim failure */
 };