[PATCH 00/13] Balancing the scan rate of major caches V2

[PATCH 00/13] Balancing the scan rate of major caches V2

[Wu Fengguang]

Changes since V1:
- better broken up of patches
- replace pages_more_aged with age_ge/age_gt
- expanded shrink_slab interface
- rewrite kswapd rebalance logic to be simple and robust


This patch balances the aging rates of active_list/inactive_list/slab.

It started out as an effort to enable the adaptive read-ahead to handle large
number of concurrent readers. Then I found it envolves much more stuffs, and
deserves a standalone patchset to address the balancing problem as a whole.


The whole picture of balancing:

- In each node, inactive_list scan rates are synced with each other
  It is done in the direct/kswapd reclaim path.

- In each zone, active_list scan rate always follows that of inactive_list

- Slab cache scan rates always follow that of the current node.
  Since shrink_slab() can be called from different CPUs, that effectly sync
  slab cache scan rates with that of the most scanned node.


The patches can be grouped as follows:

- balancing stuffs
mm-revert-vmscan-balancing-fix.patch
mm-simplify-kswapd-reclaim-code.patch
mm-balance-zone-aging-supporting-facilities.patch
mm-balance-zone-aging-in-direct-reclaim.patch
mm-balance-zone-aging-in-kswapd-reclaim.patch
mm-balance-slab-aging.patch
mm-balance-active-inactive-list-aging.patch


- pure code cleanups
mm-remove-unnecessary-variable-and-loop.patch
mm-remove-swap-cluster-max-from-scan-control.patch
mm-accumulate-nr-scanned-reclaimed-in-scan-control.patch
mm-turn-bool-variables-into-flags-in-scan-control.patch

- debug code
mm-page-reclaim-debug-traces.patch

- a minor fix
mm-scan-accounting-fix.patch

Thanks,
Wu Fengguang

--
Dept. Automation                University of Science and Technology of China

[PATCH 01/13] mm: restore sc.nr_to_reclaim

[Wu Fengguang]

[-- Attachment #1: mm-revert-vmscan-balancing-fix.patch --]
[-- Type: text/plain, Size: 1331 bytes --]

Keep it before the real fine grained scan patch is ready :)

The following patches really needs small scan quantities, at least in
normal situation.

Signed-off-by: Wu Fengguang <wfg@mail.ustc.edu.cn>
---

 mm/vmscan.c |    8 ++++++++
 1 files changed, 8 insertions(+)

--- linux-2.6.15-rc5-mm1.orig/mm/vmscan.c
+++ linux-2.6.15-rc5-mm1/mm/vmscan.c
@@ -63,6 +63,9 @@ struct scan_control {
 
 	unsigned long nr_mapped;	/* From page_state */
 
+	/* How many pages shrink_cache() should reclaim */
+	int nr_to_reclaim;
+
 	/* Ask shrink_caches, or shrink_zone to scan at this priority */
 	unsigned int priority;
 
@@ -898,6 +901,7 @@ static void shrink_cache(struct zone *zo
 		if (current_is_kswapd())
 			mod_page_state(kswapd_steal, nr_freed);
 		mod_page_state_zone(zone, pgsteal, nr_freed);
+		sc->nr_to_reclaim -= nr_freed;
 
 		spin_lock_irq(&zone->lru_lock);
 		/*
@@ -1097,6 +1101,8 @@ shrink_zone(struct zone *zone, struct sc
 	else
 		nr_inactive = 0;
 
+	sc->nr_to_reclaim = sc->swap_cluster_max;
+
 	while (nr_active || nr_inactive) {
 		if (nr_active) {
 			sc->nr_to_scan = min(nr_active,
@@ -1110,6 +1116,8 @@ shrink_zone(struct zone *zone, struct sc
 					(unsigned long)sc->swap_cluster_max);
 			nr_inactive -= sc->nr_to_scan;
 			shrink_cache(zone, sc);
+			if (sc->nr_to_reclaim <= 0)
+				break;
 		}
 	}
 

--

[PATCH 02/13] mm: simplify kswapd reclaim code

[Wu Fengguang]

[-- Attachment #1: mm-simplify-kswapd-reclaim-code.patch --]
[-- Type: text/plain, Size: 4423 bytes --]

Simplify the kswapd reclaim code for the new balancing logic.

Signed-off-by: Nick Piggin <npiggin@suse.de>
Signed-off-by: Wu Fengguang <wfg@mail.ustc.edu.cn>
---


 mm/vmscan.c |  100 ++++++++++++++++++++----------------------------------------
 1 files changed, 34 insertions(+), 66 deletions(-)

--- linux-2.6.15-rc5-mm1.orig/mm/vmscan.c
+++ linux-2.6.15-rc5-mm1/mm/vmscan.c
@@ -1309,47 +1309,18 @@ loop_again:
 	}
 
 	for (priority = DEF_PRIORITY; priority >= 0; priority--) {
-		int end_zone = 0;	/* Inclusive.  0 = ZONE_DMA */
 		unsigned long lru_pages = 0;
 
+		all_zones_ok = 1;
+		sc.nr_scanned = 0;
+		sc.nr_reclaimed = 0;
+		sc.priority = priority;
+		sc.swap_cluster_max = nr_pages ? nr_pages : SWAP_CLUSTER_MAX;
+
 		/* The swap token gets in the way of swapout... */
 		if (!priority)
 			disable_swap_token();
 
-		all_zones_ok = 1;
-
-		if (nr_pages == 0) {
-			/*
-			 * Scan in the highmem->dma direction for the highest
-			 * zone which needs scanning
-			 */
-			for (i = pgdat->nr_zones - 1; i >= 0; i--) {
-				struct zone *zone = pgdat->node_zones + i;
-
-				if (!populated_zone(zone))
-					continue;
-
-				if (zone->all_unreclaimable &&
-						priority != DEF_PRIORITY)
-					continue;
-
-				if (!zone_watermark_ok(zone, order,
-						zone->pages_high, 0, 0)) {
-					end_zone = i;
-					goto scan;
-				}
-			}
-			goto out;
-		} else {
-			end_zone = pgdat->nr_zones - 1;
-		}
-scan:
-		for (i = 0; i <= end_zone; i++) {
-			struct zone *zone = pgdat->node_zones + i;
-
-			lru_pages += zone->nr_active + zone->nr_inactive;
-		}
-
 		/*
 		 * Now scan the zone in the dma->highmem direction, stopping
 		 * at the last zone which needs scanning.
@@ -1359,51 +1330,49 @@ scan:
 		 * pages behind kswapd's direction of progress, which would
 		 * cause too much scanning of the lower zones.
 		 */
-		for (i = 0; i <= end_zone; i++) {
+		for (i = 0; i < pgdat->nr_zones; i++) {
 			struct zone *zone = pgdat->node_zones + i;
-			int nr_slab;
 
 			if (!populated_zone(zone))
 				continue;
 
+			if (nr_pages == 0) {	/* Not software suspend */
+				if (zone_watermark_ok(zone, order,
+					zone->pages_high, 0, 0))
+					continue;
+
+				all_zones_ok = 0;
+			}
+
 			if (zone->all_unreclaimable && priority != DEF_PRIORITY)
 				continue;
 
-			if (nr_pages == 0) {	/* Not software suspend */
-				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;
-			sc.nr_reclaimed = 0;
-			sc.priority = priority;
-			sc.swap_cluster_max = nr_pages? nr_pages : SWAP_CLUSTER_MAX;
-			atomic_inc(&zone->reclaim_in_progress);
+			lru_pages += zone->nr_active + zone->nr_inactive;
+
 			shrink_zone(zone, &sc);
-			atomic_dec(&zone->reclaim_in_progress);
-			reclaim_state->reclaimed_slab = 0;
-			nr_slab = shrink_slab(sc.nr_scanned, GFP_KERNEL,
-						lru_pages);
-			sc.nr_reclaimed += reclaim_state->reclaimed_slab;
-			total_reclaimed += sc.nr_reclaimed;
-			total_scanned += sc.nr_scanned;
-			if (zone->all_unreclaimable)
-				continue;
-			if (nr_slab == 0 && zone->pages_scanned >=
+
+			if (zone->pages_scanned >=
 				    (zone->nr_active + zone->nr_inactive) * 4)
 				zone->all_unreclaimable = 1;
-			/*
-			 * If we've done a decent amount of scanning and
-			 * the reclaim ratio is low, start doing writepage
-			 * even in laptop mode
-			 */
-			if (total_scanned > SWAP_CLUSTER_MAX * 2 &&
-			    total_scanned > total_reclaimed+total_reclaimed/2)
-				sc.may_writepage = 1;
 		}
+		reclaim_state->reclaimed_slab = 0;
+		shrink_slab(sc.nr_scanned, GFP_KERNEL, lru_pages);
+		sc.nr_reclaimed += reclaim_state->reclaimed_slab;
+		total_reclaimed += sc.nr_reclaimed;
+		total_scanned += sc.nr_scanned;
+
+		/*
+		 * If we've done a decent amount of scanning and
+		 * the reclaim ratio is low, start doing writepage
+		 * even in laptop mode
+		 */
+		if (total_scanned > SWAP_CLUSTER_MAX * 2 &&
+		    total_scanned > total_reclaimed+total_reclaimed/2)
+			sc.may_writepage = 1;
+
 		if (nr_pages && to_free > total_reclaimed)
 			continue;	/* swsusp: need to do more work */
 		if (all_zones_ok)
@@ -1424,7 +1393,6 @@ scan:
 		if ((total_reclaimed >= SWAP_CLUSTER_MAX) && (!nr_pages))
 			break;
 	}
-out:
 	for (i = 0; i < pgdat->nr_zones; i++) {
 		struct zone *zone = pgdat->node_zones + i;
 

--

[PATCH 03/13] mm: supporting variables and functions for balanced zone aging

[Wu Fengguang]

[-- Attachment #1: mm-balance-zone-aging-supporting-facilities.patch --]
[-- Type: text/plain, Size: 5485 bytes --]

The zone aging rates are currently imbalanced, the gap can be as large as 3
times, which can severely damage read-ahead requests and shorten their
effective life time.

This patch adds three variables in struct zone
	- aging_total
	- aging_milestone
	- page_age
to keep track of page aging rate, and keep it in sync on page reclaim time.

The aging_total is just a per-zone counter-part to the per-cpu
pgscan_{kswapd,direct}_{zone name}. But it is not direct comparable between
zones, so the aging_milestone/page_age are maintained based on aging_total.

The page_age is a normalized value that can be direct compared between zones
with the helper macro age_ge/age_gt. The goal of balancing logics are to keep
this normalized value in sync between zones.

One can check the balanced aging progress by running:
                        tar c / | cat > /dev/null &
                        watch -n1 'grep "age " /proc/zoneinfo'

Signed-off-by: Wu Fengguang <wfg@mail.ustc.edu.cn>
---

 include/linux/mmzone.h |   14 ++++++++++++++
 mm/page_alloc.c        |   11 +++++++++++
 mm/vmscan.c            |   43 +++++++++++++++++++++++++++++++++++++++++++
 3 files changed, 68 insertions(+)

--- linux-2.6.15-rc5-mm1.orig/include/linux/mmzone.h
+++ linux-2.6.15-rc5-mm1/include/linux/mmzone.h
@@ -149,6 +149,20 @@ struct zone {
 	unsigned long		pages_scanned;	   /* since last reclaim */
 	int			all_unreclaimable; /* All pages pinned */
 
+	/* Fields for balanced page aging:
+	 * aging_total     - The accumulated number of activities that may
+	 *                   cause page aging, that is, make some pages closer
+	 *                   to the tail of inactive_list.
+	 * aging_milestone - A snapshot of total_scan every time a full
+	 *                   inactive_list of pages become aged.
+	 * page_age        - A normalized value showing the percent of pages
+	 *                   have been aged.  It is compared between zones to
+	 *                   balance the rate of page aging.
+	 */
+	unsigned long		aging_total;
+	unsigned long		aging_milestone;
+	unsigned long		page_age;
+
 	/*
 	 * Does the allocator try to reclaim pages from the zone as soon
 	 * as it fails a watermark_ok() in __alloc_pages?
--- linux-2.6.15-rc5-mm1.orig/mm/vmscan.c
+++ linux-2.6.15-rc5-mm1/mm/vmscan.c
@@ -123,6 +123,48 @@ static long total_memory;
 static LIST_HEAD(shrinker_list);
 static DECLARE_RWSEM(shrinker_rwsem);
 
+#ifdef CONFIG_HIGHMEM64G
+#define		PAGE_AGE_SHIFT  8
+#elif BITS_PER_LONG == 32
+#define		PAGE_AGE_SHIFT  12
+#elif BITS_PER_LONG == 64
+#define		PAGE_AGE_SHIFT  20
+#else
+#error unknown BITS_PER_LONG
+#endif
+#define		PAGE_AGE_SIZE   (1 << PAGE_AGE_SHIFT)
+#define		PAGE_AGE_MASK   (PAGE_AGE_SIZE - 1)
+
+/*
+ * The simplified code is:
+ * 	age_ge: (a->page_age >= b->page_age)
+ * 	age_gt: (a->page_age > b->page_age)
+ * The complexity deals with the wrap-around problem.
+ * Two page ages not close enough(gap >= 1/8) should also be ignored:
+ * they are out of sync and the comparison may be nonsense.
+ */
+#define age_ge(a, b) \
+	(((a->page_age - b->page_age) & PAGE_AGE_MASK) < PAGE_AGE_SIZE / 8)
+#define age_gt(a, b) \
+	(((b->page_age - a->page_age) & PAGE_AGE_MASK) > PAGE_AGE_SIZE * 7 / 8)
+
+/*
+ * Keep track of the percent of cold pages that have been scanned / aged.
+ * It's not really ##%, but a high resolution normalized value.
+ */
+static inline void update_zone_age(struct zone *z, int nr_scan)
+{
+	unsigned long len = z->nr_inactive | 1;
+
+	z->aging_total += nr_scan;
+
+	if (z->aging_total - z->aging_milestone > len)
+		z->aging_milestone += len;
+
+	z->page_age = ((z->aging_total - z->aging_milestone)
+						<< PAGE_AGE_SHIFT) / len;
+}
+
 /*
  * Add a shrinker callback to be called from the vm
  */
@@ -887,6 +929,7 @@ static void shrink_cache(struct zone *zo
 					     &page_list, &nr_scan);
 		zone->nr_inactive -= nr_taken;
 		zone->pages_scanned += nr_scan;
+		update_zone_age(zone, nr_scan);
 		spin_unlock_irq(&zone->lru_lock);
 
 		if (nr_taken == 0)
--- linux-2.6.15-rc5-mm1.orig/mm/page_alloc.c
+++ linux-2.6.15-rc5-mm1/mm/page_alloc.c
@@ -1522,6 +1522,8 @@ void show_free_areas(void)
 			" active:%lukB"
 			" inactive:%lukB"
 			" present:%lukB"
+			" aging:%lukB"
+			" age:%lu"
 			" pages_scanned:%lu"
 			" all_unreclaimable? %s"
 			"\n",
@@ -1533,6 +1535,8 @@ void show_free_areas(void)
 			K(zone->nr_active),
 			K(zone->nr_inactive),
 			K(zone->present_pages),
+			K(zone->aging_total),
+			zone->page_age,
 			zone->pages_scanned,
 			(zone->all_unreclaimable ? "yes" : "no")
 			);
@@ -2144,6 +2148,9 @@ static void __init free_area_init_core(s
 		zone->nr_scan_inactive = 0;
 		zone->nr_active = 0;
 		zone->nr_inactive = 0;
+		zone->aging_total = 0;
+		zone->aging_milestone = 0;
+		zone->page_age = 0;
 		atomic_set(&zone->reclaim_in_progress, 0);
 		if (!size)
 			continue;
@@ -2292,6 +2299,8 @@ static int zoneinfo_show(struct seq_file
 			   "\n        high     %lu"
 			   "\n        active   %lu"
 			   "\n        inactive %lu"
+			   "\n        aging    %lu"
+			   "\n        age      %lu"
 			   "\n        scanned  %lu (a: %lu i: %lu)"
 			   "\n        spanned  %lu"
 			   "\n        present  %lu",
@@ -2301,6 +2310,8 @@ static int zoneinfo_show(struct seq_file
 			   zone->pages_high,
 			   zone->nr_active,
 			   zone->nr_inactive,
+			   zone->aging_total,
+			   zone->page_age,
 			   zone->pages_scanned,
 			   zone->nr_scan_active, zone->nr_scan_inactive,
 			   zone->spanned_pages,

--

[PATCH 04/13] mm: balance zone aging in direct reclaim path

[Wu Fengguang]

[-- Attachment #1: mm-balance-zone-aging-in-direct-reclaim.patch --]
[-- Type: text/plain, Size: 2390 bytes --]

Add 10 extra priorities to the direct page reclaim path, which makes 10 round of
balancing effort(reclaim only from the least aged local/headless zone) before
falling back to the reclaim-all scheme.

Ten rounds should be enough to get enough free pages in normal cases, which
prevents unnecessarily disturbing remote nodes. If further restrict the first
round of page allocation to local zones, we might get what the early zone
reclaim patch want: memory affinity/locality.

Signed-off-by: Wu Fengguang <wfg@mail.ustc.edu.cn>
---

 mm/vmscan.c |   31 ++++++++++++++++++++++++++++---
 1 files changed, 28 insertions(+), 3 deletions(-)

--- linux-2.6.15-rc5-mm1.orig/mm/vmscan.c
+++ linux-2.6.15-rc5-mm1/mm/vmscan.c
@@ -1189,6 +1189,7 @@ static void
 shrink_caches(struct zone **zones, struct scan_control *sc)
 {
 	int i;
+	struct zone *z = NULL;
 
 	for (i = 0; zones[i] != NULL; i++) {
 		struct zone *zone = zones[i];
@@ -1203,11 +1204,34 @@ shrink_caches(struct zone **zones, struc
 		if (zone->prev_priority > sc->priority)
 			zone->prev_priority = sc->priority;
 
-		if (zone->all_unreclaimable && sc->priority != DEF_PRIORITY)
+		if (zone->all_unreclaimable && sc->priority < DEF_PRIORITY)
 			continue;	/* Let kswapd poll it */
 
+		/*
+		 * Balance page aging in local zones and following headless
+		 * zones.
+		 */
+		if (sc->priority > DEF_PRIORITY) {
+			if (zone->zone_pgdat != zones[0]->zone_pgdat) {
+				cpumask_t cpu = node_to_cpumask(
+						zone->zone_pgdat->node_id);
+				if (!cpus_empty(cpu))
+					break;
+			}
+
+			if (!z)
+				z = zone;
+			else if (age_gt(z, zone))
+				z = zone;
+
+			continue;
+		}
+
 		shrink_zone(zone, sc);
 	}
+
+	if (z)
+		shrink_zone(z, sc);
 }
  
 /*
@@ -1251,7 +1275,8 @@ int try_to_free_pages(struct zone **zone
 		lru_pages += zone->nr_active + zone->nr_inactive;
 	}
 
-	for (priority = DEF_PRIORITY; priority >= 0; priority--) {
+	/* The added 10 priorities are for scan rate balancing */
+	for (priority = DEF_PRIORITY + 10; priority >= 0; priority--) {
 		sc.nr_mapped = read_page_state(nr_mapped);
 		sc.nr_scanned = 0;
 		sc.nr_reclaimed = 0;
@@ -1285,7 +1310,7 @@ int try_to_free_pages(struct zone **zone
 		}
 
 		/* Take a nap, wait for some writeback to complete */
-		if (sc.nr_scanned && priority < DEF_PRIORITY - 2)
+		if (sc.nr_scanned && priority < DEF_PRIORITY)
 			blk_congestion_wait(WRITE, HZ/10);
 	}
 out:

--

[PATCH 05/13] mm: balance zone aging in kswapd reclaim path

[Wu Fengguang]

[-- Attachment #1: mm-balance-zone-aging-in-kswapd-reclaim.patch --]
[-- Type: text/plain, Size: 4182 bytes --]

The vm subsystem is rather complex. System memory is divided into zones,
lower zones act as fallback of higher zones in memory allocation.  The page
reclaim algorithm should generally keep zone aging rates in sync. But if a
zone under watermark has many unreclaimable pages, it has to be scanned much
more to get enough free pages. While doing this,

- lower zones should also be scanned more, since their pages are also usable
  for higher zone allocations.
- higher zones should not be scanned just to keep the aging in sync, which
  can evict large amount of pages without saving the problem(and may well
  worsen it).

With that in mind, the patch does the rebalance in kswapd as follows:
1) reclaim from the lowest zone when
	- under pages_high
	- under pages_high+lowmem_reserve, and less/equal aged than highest
	  zone(or out of sync with it)
2) reclaim from higher zones when
	- under pages_high+lowmem_reserve, and less/equal aged than its
	  immediate lower neighbor(or out of sync with it)

Note that the zone age is a normalized value in range 0-4096 on i386/4G. 4096
corresponds to a full scan of one zone. And the comparison of ages are only
deemed ok if the gap is less than 4096/8, or they will be regarded as out of
sync.

On exit, the code ensures:
1) the lowest zone will be pages_high ok
2) at least one zone will be pages_high+lowmem_reserve ok
3) a very strong force of rebalancing with the exception of
	- some lower zones are unreclaimable: we must let them go ahead
	  alone, leaving higher zones back
	- shrink_zone() scans too much and creates huge imbalance in one
	  run(Nick is working on this)

The logic can deal with known normal/abnormal situations gracefully:
1) Normal case
	- zone ages are cyclicly tied together: taking over each other, and
	  keeping close enough

2) A Zone is unreclaimable, scanned much more, and become out of sync
	- if ever a troublesome zone is being overscanned, the logic brings
	  its lower neighbors ahead together, leaving higher neighbors back.
	- the aging tie between the two groups is broken, and the relevant
	  zones are reclaimed when pages_high+lowmem_reserve not ok, just as
	  before the patch.
	- at some time the zone ages meet again and back to normal
	- a possiblely better strategy, as soon as the pressure disappeared,
	  might be relunctant to reclaim from the already overscanned lower
	  group, and let the higher group slowly catch up.

3) Zone is truncated
	- will not reclaim from it until under watermark

With this patch, the meaning of zone->pages_high+lowmem_reserve changed from
the _required_ watermark to the _recommended_ watermark. Someone might be
willing to increase them somehow.

Signed-off-by: Wu Fengguang <wfg@mail.ustc.edu.cn>
---

 mm/vmscan.c |   25 ++++++++++++++++++++-----
 1 files changed, 20 insertions(+), 5 deletions(-)

--- linux-2.6.15-rc5-mm1.orig/mm/vmscan.c
+++ linux-2.6.15-rc5-mm1/mm/vmscan.c
@@ -1359,6 +1359,7 @@ static int balance_pgdat(pg_data_t *pgda
 	int total_scanned, total_reclaimed;
 	struct reclaim_state *reclaim_state = current->reclaim_state;
 	struct scan_control sc;
+	struct zone *prev_zone = pgdat->node_zones;
 
 loop_again:
 	total_scanned = 0;
@@ -1374,6 +1375,9 @@ loop_again:
 		struct zone *zone = pgdat->node_zones + i;
 
 		zone->temp_priority = DEF_PRIORITY;
+
+		if (populated_zone(zone))
+			prev_zone = zone;
 	}
 
 	for (priority = DEF_PRIORITY; priority >= 0; priority--) {
@@ -1404,14 +1408,25 @@ loop_again:
 			if (!populated_zone(zone))
 				continue;
 
-			if (nr_pages == 0) {	/* Not software suspend */
-				if (zone_watermark_ok(zone, order,
-					zone->pages_high, 0, 0))
-					continue;
+			if (nr_pages) 	/* software suspend */
+				goto scan_swspd;
 
-				all_zones_ok = 0;
+			if (zone < prev_zone &&
+					!zone_watermark_ok(zone, order,
+						zone->pages_high, 0, 0)) {
+			} else if (!age_gt(zone, prev_zone) &&
+					!zone_watermark_ok(zone, order,
+						zone->pages_high,
+						pgdat->nr_zones - 1, 0)) {
+			} else {
+				prev_zone = zone;
+				continue;
 			}
 
+			prev_zone = zone;
+			all_zones_ok = 0;
+
+scan_swspd:
 			if (zone->all_unreclaimable && priority != DEF_PRIORITY)
 				continue;
 

--

[PATCH 06/13] mm: balance slab aging

[Wu Fengguang]

[-- Attachment #1: mm-balance-slab-aging.patch --]
[-- Type: text/plain, Size: 8679 bytes --]

The current slab shrinking code is way too fragile.
Let it manage aging pace by itself, and provide a simple and robust interface.

The design considerations:
- use the same syncing facilities as that of the zones
- keep the age of slabs in line with that of the largest zone
  this in effect makes aging rate of slabs follow that of the most aged node.

- reserve a minimal number of unused slabs
  the size of reservation depends on vm pressure

- shrink more slab caches only when vm pressure is high
  the old logic, `mmap pages found' - `shrink more caches' - `avoid swapping',
  sounds not quite logical, so the code is removed.

- let sc->nr_scanned record the exact number of cold pages scanned
  it is no longer used by the slab cache shrinking algorithm, but good for other
  algorithms(e.g. the active_list/inactive_list balancing).

Signed-off-by: Wu Fengguang <wfg@mail.ustc.edu.cn>
---

 fs/drop-pagecache.c |    2 
 include/linux/mm.h  |    7 +--
 mm/vmscan.c         |  106 +++++++++++++++++++++-------------------------------
 3 files changed, 48 insertions(+), 67 deletions(-)

--- linux-2.6.15-rc5-mm1.orig/include/linux/mm.h
+++ linux-2.6.15-rc5-mm1/include/linux/mm.h
@@ -798,7 +798,9 @@ struct shrinker {
 	shrinker_t		shrinker;
 	struct list_head	list;
 	int			seeks;	/* seeks to recreate an obj */
-	long			nr;	/* objs pending delete */
+	unsigned long		aging_total;
+	unsigned long		aging_milestone;
+	unsigned long		page_age;
 	struct shrinker_stats	*s_stats;
 };
 
@@ -1080,8 +1082,7 @@ int in_gate_area_no_task(unsigned long a
 
 int drop_pagecache_sysctl_handler(struct ctl_table *, int, struct file *,
 					void __user *, size_t *, loff_t *);
-int shrink_slab(unsigned long scanned, gfp_t gfp_mask,
-			unsigned long lru_pages);
+int shrink_slab(struct zone *zone, int priority, gfp_t gfp_mask);
 
 #endif /* __KERNEL__ */
 #endif /* _LINUX_MM_H */
--- linux-2.6.15-rc5-mm1.orig/mm/vmscan.c
+++ linux-2.6.15-rc5-mm1/mm/vmscan.c
@@ -165,6 +165,18 @@ static inline void update_zone_age(struc
 						<< PAGE_AGE_SHIFT) / len;
 }
 
+static inline void update_slab_age(struct shrinker *s,
+					unsigned long len, int nr_scan)
+{
+	s->aging_total += nr_scan;
+
+	if (s->aging_total - s->aging_milestone > len)
+		s->aging_milestone += len;
+
+	s->page_age = ((s->aging_total - s->aging_milestone)
+						<< PAGE_AGE_SHIFT) / len;
+}
+
 /*
  * Add a shrinker callback to be called from the vm
  */
@@ -176,7 +188,9 @@ struct shrinker *set_shrinker(int seeks,
         if (shrinker) {
 	        shrinker->shrinker = theshrinker;
 	        shrinker->seeks = seeks;
-	        shrinker->nr = 0;
+	        shrinker->aging_total = 0;
+	        shrinker->aging_milestone = 0;
+	        shrinker->page_age = 0;
 		shrinker->s_stats = alloc_percpu(struct shrinker_stats);
 		if (!shrinker->s_stats) {
 			kfree(shrinker);
@@ -204,6 +218,7 @@ void remove_shrinker(struct shrinker *sh
 EXPORT_SYMBOL(remove_shrinker);
 
 #define SHRINK_BATCH 128
+#define SLAB_RESERVE 1000
 /*
  * Call the shrink functions to age shrinkable caches
  *
@@ -212,76 +227,49 @@ EXPORT_SYMBOL(remove_shrinker);
  * percentages of the lru and ageable caches.  This should balance the seeks
  * generated by these structures.
  *
- * If the vm encounted mapped pages on the LRU it increase the pressure on
- * slab to avoid swapping.
+ * @priority reflects the vm pressure, the lower the value, the more to
+ * shrink.
  *
- * We do weird things to avoid (scanned*seeks*entries) overflowing 32 bits.
- *
- * `lru_pages' represents the number of on-LRU pages in all the zones which
- * are eligible for the caller's allocation attempt.  It is used for balancing
- * slab reclaim versus page reclaim.
+ * @zone is better to be the least over-scanned one (normally the highest
+ * zone).
  *
  * Returns the number of slab objects which we shrunk.
  */
-int shrink_slab(unsigned long scanned, gfp_t gfp_mask, unsigned long lru_pages)
+int shrink_slab(struct zone *zone, int priority, gfp_t gfp_mask)
 {
 	struct shrinker *shrinker;
 	int ret = 0;
 
-	if (scanned == 0)
-		scanned = SWAP_CLUSTER_MAX;
-
 	if (!down_read_trylock(&shrinker_rwsem))
 		return 1;	/* Assume we'll be able to shrink next time */
 
 	list_for_each_entry(shrinker, &shrinker_list, list) {
-		unsigned long long delta;
-		unsigned long total_scan;
-		unsigned long max_pass = (*shrinker->shrinker)(0, gfp_mask);
-
-		delta = (4 * scanned) / shrinker->seeks;
-		delta *= max_pass;
-		do_div(delta, lru_pages + 1);
-		shrinker->nr += delta;
-		if (shrinker->nr < 0) {
-			printk(KERN_ERR "%s: nr=%ld\n",
-					__FUNCTION__, shrinker->nr);
-			shrinker->nr = max_pass;
-		}
-
-		/*
-		 * Avoid risking looping forever due to too large nr value:
-		 * never try to free more than twice the estimate number of
-		 * freeable entries.
-		 */
-		if (shrinker->nr > max_pass * 2)
-			shrinker->nr = max_pass * 2;
-
-		total_scan = shrinker->nr;
-		shrinker->nr = 0;
-
-		while (total_scan >= SHRINK_BATCH) {
-			long this_scan = SHRINK_BATCH;
-			int shrink_ret;
+		while (!zone || age_gt(zone, shrinker)) {
 			int nr_before;
+			int nr_after;
 
 			nr_before = (*shrinker->shrinker)(0, gfp_mask);
-			shrink_ret = (*shrinker->shrinker)(this_scan, gfp_mask);
-			if (shrink_ret == -1)
+			if (nr_before < SLAB_RESERVE * priority / DEF_PRIORITY)
+				break;
+
+			nr_after = (*shrinker->shrinker)(SHRINK_BATCH, gfp_mask);
+			if (nr_after == -1)
 				break;
-			if (shrink_ret < nr_before) {
-				ret += nr_before - shrink_ret;
-				shrinker_stat_add(shrinker, nr_freed,
-					(nr_before - shrink_ret));
+
+			if (nr_after < nr_before) {
+				int nr_freed = nr_before - nr_after;
+
+				ret += nr_freed;
+				shrinker_stat_add(shrinker, nr_freed, nr_freed);
 			}
-			shrinker_stat_add(shrinker, nr_req, this_scan);
-			mod_page_state(slabs_scanned, this_scan);
-			total_scan -= this_scan;
+			shrinker_stat_add(shrinker, nr_req, SHRINK_BATCH);
+			mod_page_state(slabs_scanned, SHRINK_BATCH);
+			update_slab_age(shrinker, nr_before * DEF_PRIORITY * 2,
+					SHRINK_BATCH * shrinker->seeks *
+					(DEF_PRIORITY + priority));
 
 			cond_resched();
 		}
-
-		shrinker->nr += total_scan;
 	}
 	up_read(&shrinker_rwsem);
 	return ret;
@@ -487,11 +475,6 @@ static int shrink_list(struct list_head 
 
 		BUG_ON(PageActive(page));
 
-		sc->nr_scanned++;
-		/* Double the slab pressure for mapped and swapcache pages */
-		if (page_mapped(page) || PageSwapCache(page))
-			sc->nr_scanned++;
-
 		if (PageWriteback(page))
 			goto keep_locked;
 
@@ -936,6 +919,7 @@ static void shrink_cache(struct zone *zo
 			goto done;
 
 		max_scan -= nr_scan;
+		sc->nr_scanned += nr_scan;
 		if (current_is_kswapd())
 			mod_page_state_zone(zone, pgscan_kswapd, nr_scan);
 		else
@@ -1254,7 +1238,6 @@ int try_to_free_pages(struct zone **zone
 	int total_scanned = 0, total_reclaimed = 0;
 	struct reclaim_state *reclaim_state = current->reclaim_state;
 	struct scan_control sc;
-	unsigned long lru_pages = 0;
 	int i;
 
 	delay_prefetch();
@@ -1272,7 +1255,6 @@ int try_to_free_pages(struct zone **zone
 			continue;
 
 		zone->temp_priority = DEF_PRIORITY;
-		lru_pages += zone->nr_active + zone->nr_inactive;
 	}
 
 	/* The added 10 priorities are for scan rate balancing */
@@ -1285,7 +1267,8 @@ int try_to_free_pages(struct zone **zone
 		if (!priority)
 			disable_swap_token();
 		shrink_caches(zones, &sc);
-		shrink_slab(sc.nr_scanned, gfp_mask, lru_pages);
+		if (zone_idx(zones[0]))
+			shrink_slab(zones[0], priority, gfp_mask);
 		if (reclaim_state) {
 			sc.nr_reclaimed += reclaim_state->reclaimed_slab;
 			reclaim_state->reclaimed_slab = 0;
@@ -1381,8 +1364,6 @@ loop_again:
 	}
 
 	for (priority = DEF_PRIORITY; priority >= 0; priority--) {
-		unsigned long lru_pages = 0;
-
 		all_zones_ok = 1;
 		sc.nr_scanned = 0;
 		sc.nr_reclaimed = 0;
@@ -1433,7 +1414,6 @@ scan_swspd:
 			zone->temp_priority = priority;
 			if (zone->prev_priority > priority)
 				zone->prev_priority = priority;
-			lru_pages += zone->nr_active + zone->nr_inactive;
 
 			shrink_zone(zone, &sc);
 
@@ -1442,7 +1422,7 @@ scan_swspd:
 				zone->all_unreclaimable = 1;
 		}
 		reclaim_state->reclaimed_slab = 0;
-		shrink_slab(sc.nr_scanned, GFP_KERNEL, lru_pages);
+		shrink_slab(prev_zone, priority, GFP_KERNEL);
 		sc.nr_reclaimed += reclaim_state->reclaimed_slab;
 		total_reclaimed += sc.nr_reclaimed;
 		total_scanned += sc.nr_scanned;
--- linux-2.6.15-rc5-mm1.orig/fs/drop-pagecache.c
+++ linux-2.6.15-rc5-mm1/fs/drop-pagecache.c
@@ -47,7 +47,7 @@ static void drop_slab(void)
 	int nr_objects;
 
 	do {
-		nr_objects = shrink_slab(1000, GFP_KERNEL, 1000);
+		nr_objects = shrink_slab(NULL, 0, GFP_KERNEL);
 	} while (nr_objects > 10);
 }
 

--

[PATCH 07/13] mm: balance active/inactive list scan rates

[Wu Fengguang]

[-- Attachment #1: mm-balance-active-inactive-list-aging.patch --]
[-- Type: text/plain, Size: 9517 bytes --]

shrink_zone() has two major design goals:
1) let active/inactive lists have equal scan rates
2) do the scans in small chunks

But the implementation has some problems:
- reluctant to scan small zones
  the callers often have to dip into low priority to free memory.

- the balance is quite rough
  the break statement in the loop breaks it.

- may scan few pages in one batch
  refill_inactive_zone can be called twice to scan 32 and 1 pages.

The new design:
1) keep perfect balance
   let active_list follow inactive_list in scan rate

2) always scan in SWAP_CLUSTER_MAX sized chunks
   simple and efficient

3) will scan at least one chunk
   the expected behavior from the callers

The perfect balance may or may not yield better performance, though it
a) is a more understandable and dependable behavior
b) together with inter-zone balancing, makes the zoned memories consistent

The atomic reclaim_in_progress is there to prevent most concurrent reclaims.
If concurrent reclaims did happen, there will be no fatal errors.


I tested the patch with the following commands:
	dd if=/dev/zero of=hot bs=1M seek=800 count=1
	dd if=/dev/zero of=cold bs=1M seek=50000 count=1
	./test-aging.sh; ./active-inactive-aging-rate.sh

Before the patch:
-----------------------------------------------------------------------------
active/inactive sizes on 2.6.14-2-686-smp:
0/1000          = 0 / 1241
563/1000        = 73343 / 130108
887/1000        = 137348 / 154816

active/inactive scan rates:
dma      38/1000        = 7731 / (198924 + 0)
normal   465/1000       = 2979780 / (6394740 + 0)
high     680/1000       = 4354230 / (6396786 + 0)

             total       used       free     shared    buffers     cached
Mem:          2027       1978         49          0          4       1923
-/+ buffers/cache:         49       1977
Swap:            0          0          0
-----------------------------------------------------------------------------

After the patch, the scan rates and the size ratios are kept roughly the same
for all zones:
-----------------------------------------------------------------------------
active/inactive sizes on 2.6.15-rc3-mm1:
0/1000          = 0 / 961
236/1000        = 38385 / 162429
319/1000        = 70607 / 221101

active/inactive scan rates:
dma      0/1000         = 0 / (42176 + 0)
normal   234/1000       = 1714688 / (7303456 + 1088)
high     317/1000       = 3151936 / (9933792 + 96)
             
             total       used       free     shared    buffers     cached
Mem:          2020       1969         50          0          5       1908
-/+ buffers/cache:         54       1965
Swap:            0          0          0
-----------------------------------------------------------------------------

script test-aging.sh:
------------------------------
#!/bin/zsh
cp cold /dev/null&

while {pidof cp > /dev/null};
do
        cp hot /dev/null
done
------------------------------

script active-inactive-aging-rate.sh:
-----------------------------------------------------------------------------
#!/bin/sh

echo active/inactive sizes on `uname -r`:
egrep '(active|inactive)' /proc/zoneinfo |
while true
do
	read name value
	[[ -z $name ]] && break
	eval $name=$value
	[[ $name = "inactive" ]] && echo -e "$((active * 1000 / (1 + inactive)))/1000  \t= $active / $inactive"
done

while true
do
	read name value
	[[ -z $name ]] && break
	eval $name=$value
done < /proc/vmstat

echo
echo active/inactive scan rates:
echo -e "dma \t $((pgrefill_dma * 1000 / (1 + pgscan_kswapd_dma + pgscan_direct_dma)))/1000 \t= $pgrefill_dma / ($pgscan_kswapd_dma + $pgscan_direct_dma)"
echo -e "normal \t $((pgrefill_normal * 1000 / (1 + pgscan_kswapd_normal + pgscan_direct_normal)))/1000 \t= $pgrefill_normal / ($pgscan_kswapd_normal + $pgscan_direct_normal)"
echo -e "high \t $((pgrefill_high * 1000 / (1 + pgscan_kswapd_high + pgscan_direct_high)))/1000 \t= $pgrefill_high / ($pgscan_kswapd_high + $pgscan_direct_high)"

echo
free -m
-----------------------------------------------------------------------------

Signed-off-by: Wu Fengguang <wfg@mail.ustc.edu.cn>
---

 include/linux/mmzone.h |    3 --
 include/linux/swap.h   |    2 -
 mm/page_alloc.c        |    5 +---
 mm/vmscan.c            |   52 +++++++++++++++++++++++++++----------------------
 4 files changed, 33 insertions(+), 29 deletions(-)

--- linux-2.6.15-rc5-mm1.orig/mm/vmscan.c
+++ linux-2.6.15-rc5-mm1/mm/vmscan.c
@@ -907,7 +907,7 @@ static void shrink_cache(struct zone *zo
 		int nr_scan;
 		int nr_freed;
 
-		nr_taken = isolate_lru_pages(sc->swap_cluster_max,
+		nr_taken = isolate_lru_pages(sc->nr_to_scan,
 					     &zone->inactive_list,
 					     &page_list, &nr_scan);
 		zone->nr_inactive -= nr_taken;
@@ -1101,56 +1101,56 @@ refill_inactive_zone(struct zone *zone, 
 
 /*
  * This is a basic per-zone page freer.  Used by both kswapd and direct reclaim.
+ * The reclaim process:
+ * a) scan always in batch of SWAP_CLUSTER_MAX pages
+ * b) scan inactive list at least one batch
+ * c) balance the scan rate of active/inactive list
+ * d) finish on either scanned or reclaimed enough pages
  */
 static void
 shrink_zone(struct zone *zone, struct scan_control *sc)
 {
+	unsigned long long next_scan_active;
 	unsigned long nr_active;
 	unsigned long nr_inactive;
 
 	atomic_inc(&zone->reclaim_in_progress);
 
+	next_scan_active = sc->nr_scanned;
+
 	/*
 	 * Add one to `nr_to_scan' just to make sure that the kernel will
 	 * slowly sift through the active list.
 	 */
-	zone->nr_scan_active += (zone->nr_active >> sc->priority) + 1;
-	nr_active = zone->nr_scan_active;
-	if (nr_active >= sc->swap_cluster_max)
-		zone->nr_scan_active = 0;
-	else
-		nr_active = 0;
-
-	zone->nr_scan_inactive += (zone->nr_inactive >> sc->priority) + 1;
-	nr_inactive = zone->nr_scan_inactive;
-	if (nr_inactive >= sc->swap_cluster_max)
-		zone->nr_scan_inactive = 0;
-	else
-		nr_inactive = 0;
+	nr_active = zone->nr_scan_active + 1;
+	nr_inactive = (zone->nr_inactive >> sc->priority) + SWAP_CLUSTER_MAX;
+	nr_inactive &= ~(SWAP_CLUSTER_MAX - 1);
 
+	sc->nr_to_scan = SWAP_CLUSTER_MAX;
 	sc->nr_to_reclaim = sc->swap_cluster_max;
 
-	while (nr_active || nr_inactive) {
-		if (nr_active) {
-			sc->nr_to_scan = min(nr_active,
-					(unsigned long)sc->swap_cluster_max);
-			nr_active -= sc->nr_to_scan;
+	while (nr_active >= SWAP_CLUSTER_MAX * 1024 || nr_inactive) {
+		if (nr_active >= SWAP_CLUSTER_MAX * 1024) {
+			nr_active -= SWAP_CLUSTER_MAX * 1024;
 			refill_inactive_zone(zone, sc);
 		}
 
 		if (nr_inactive) {
-			sc->nr_to_scan = min(nr_inactive,
-					(unsigned long)sc->swap_cluster_max);
-			nr_inactive -= sc->nr_to_scan;
+			nr_inactive -= SWAP_CLUSTER_MAX;
 			shrink_cache(zone, sc);
 			if (sc->nr_to_reclaim <= 0)
 				break;
 		}
 	}
 
-	throttle_vm_writeout();
+	next_scan_active = (sc->nr_scanned - next_scan_active) * 1024ULL *
+					(unsigned long long)zone->nr_active;
+	do_div(next_scan_active, zone->nr_inactive | 1);
+	zone->nr_scan_active = nr_active + (unsigned long)next_scan_active;
 
 	atomic_dec(&zone->reclaim_in_progress);
+
+	throttle_vm_writeout();
 }
 
 /*
@@ -1191,6 +1191,9 @@ shrink_caches(struct zone **zones, struc
 		if (zone->all_unreclaimable && sc->priority < DEF_PRIORITY)
 			continue;	/* Let kswapd poll it */
 
+		if (atomic_read(&zone->reclaim_in_progress))
+			continue;
+
 		/*
 		 * Balance page aging in local zones and following headless
 		 * zones.
@@ -1411,6 +1414,9 @@ scan_swspd:
 			if (zone->all_unreclaimable && priority != DEF_PRIORITY)
 				continue;
 
+			if (atomic_read(&zone->reclaim_in_progress))
+				continue;
+
 			zone->temp_priority = priority;
 			if (zone->prev_priority > priority)
 				zone->prev_priority = priority;
--- linux-2.6.15-rc5-mm1.orig/mm/page_alloc.c
+++ linux-2.6.15-rc5-mm1/mm/page_alloc.c
@@ -2145,7 +2145,6 @@ static void __init free_area_init_core(s
 		INIT_LIST_HEAD(&zone->active_list);
 		INIT_LIST_HEAD(&zone->inactive_list);
 		zone->nr_scan_active = 0;
-		zone->nr_scan_inactive = 0;
 		zone->nr_active = 0;
 		zone->nr_inactive = 0;
 		zone->aging_total = 0;
@@ -2301,7 +2300,7 @@ static int zoneinfo_show(struct seq_file
 			   "\n        inactive %lu"
 			   "\n        aging    %lu"
 			   "\n        age      %lu"
-			   "\n        scanned  %lu (a: %lu i: %lu)"
+			   "\n        scanned  %lu (a: %lu)"
 			   "\n        spanned  %lu"
 			   "\n        present  %lu",
 			   zone->free_pages,
@@ -2313,7 +2312,7 @@ static int zoneinfo_show(struct seq_file
 			   zone->aging_total,
 			   zone->page_age,
 			   zone->pages_scanned,
-			   zone->nr_scan_active, zone->nr_scan_inactive,
+			   zone->nr_scan_active / 1024,
 			   zone->spanned_pages,
 			   zone->present_pages);
 		seq_printf(m,
--- linux-2.6.15-rc5-mm1.orig/include/linux/swap.h
+++ linux-2.6.15-rc5-mm1/include/linux/swap.h
@@ -111,7 +111,7 @@ enum {
 	SWP_SCANNING	= (1 << 8),	/* refcount in scan_swap_map */
 };
 
-#define SWAP_CLUSTER_MAX 32
+#define SWAP_CLUSTER_MAX 32		/* must be power of 2 */
 
 #define SWAP_MAP_MAX	0x7fff
 #define SWAP_MAP_BAD	0x8000
--- linux-2.6.15-rc5-mm1.orig/include/linux/mmzone.h
+++ linux-2.6.15-rc5-mm1/include/linux/mmzone.h
@@ -142,8 +142,7 @@ struct zone {
 	spinlock_t		lru_lock;	
 	struct list_head	active_list;
 	struct list_head	inactive_list;
-	unsigned long		nr_scan_active;
-	unsigned long		nr_scan_inactive;
+	unsigned long		nr_scan_active;	/* x1024 to be more precise */
 	unsigned long		nr_active;
 	unsigned long		nr_inactive;
 	unsigned long		pages_scanned;	   /* since last reclaim */

--

[PATCH 08/13] mm: remove unnecessary variable and loop

[Wu Fengguang]

[-- Attachment #1: mm-remove-unnecessary-variable-and-loop.patch --]
[-- Type: text/plain, Size: 3856 bytes --]

shrink_cache() and refill_inactive_zone() do not need loops.

Simplify them to scan one chunk at a time.

Signed-off-by: Wu Fengguang <wfg@mail.ustc.edu.cn>
---

 mm/vmscan.c |   92 ++++++++++++++++++++++++++++--------------------------------
 1 files changed, 43 insertions(+), 49 deletions(-)

--- linux-2.6.15-rc5-mm1.orig/mm/vmscan.c
+++ linux-2.6.15-rc5-mm1/mm/vmscan.c
@@ -895,63 +895,58 @@ static void shrink_cache(struct zone *zo
 {
 	LIST_HEAD(page_list);
 	struct pagevec pvec;
-	int max_scan = sc->nr_to_scan;
+	struct page *page;
+	int nr_taken;
+	int nr_scan;
+	int nr_freed;
 
 	pagevec_init(&pvec, 1);
 
 	lru_add_drain();
 	spin_lock_irq(&zone->lru_lock);
-	while (max_scan > 0) {
-		struct page *page;
-		int nr_taken;
-		int nr_scan;
-		int nr_freed;
-
-		nr_taken = isolate_lru_pages(sc->nr_to_scan,
-					     &zone->inactive_list,
-					     &page_list, &nr_scan);
-		zone->nr_inactive -= nr_taken;
-		zone->pages_scanned += nr_scan;
-		update_zone_age(zone, nr_scan);
-		spin_unlock_irq(&zone->lru_lock);
+	nr_taken = isolate_lru_pages(sc->nr_to_scan,
+				     &zone->inactive_list,
+				     &page_list, &nr_scan);
+	zone->nr_inactive -= nr_taken;
+	zone->pages_scanned += nr_scan;
+	update_zone_age(zone, nr_scan);
+	spin_unlock_irq(&zone->lru_lock);
 
-		if (nr_taken == 0)
-			goto done;
+	if (nr_taken == 0)
+		return;
 
-		max_scan -= nr_scan;
-		sc->nr_scanned += nr_scan;
-		if (current_is_kswapd())
-			mod_page_state_zone(zone, pgscan_kswapd, nr_scan);
-		else
-			mod_page_state_zone(zone, pgscan_direct, nr_scan);
-		nr_freed = shrink_list(&page_list, sc);
-		if (current_is_kswapd())
-			mod_page_state(kswapd_steal, nr_freed);
-		mod_page_state_zone(zone, pgsteal, nr_freed);
-		sc->nr_to_reclaim -= nr_freed;
+	sc->nr_scanned += nr_scan;
+	if (current_is_kswapd())
+		mod_page_state_zone(zone, pgscan_kswapd, nr_scan);
+	else
+		mod_page_state_zone(zone, pgscan_direct, nr_scan);
+	nr_freed = shrink_list(&page_list, sc);
+	if (current_is_kswapd())
+		mod_page_state(kswapd_steal, nr_freed);
+	mod_page_state_zone(zone, pgsteal, nr_freed);
+	sc->nr_to_reclaim -= nr_freed;
 
-		spin_lock_irq(&zone->lru_lock);
-		/*
-		 * Put back any unfreeable pages.
-		 */
-		while (!list_empty(&page_list)) {
-			page = lru_to_page(&page_list);
-			if (TestSetPageLRU(page))
-				BUG();
-			list_del(&page->lru);
-			if (PageActive(page))
-				add_page_to_active_list(zone, page);
-			else
-				add_page_to_inactive_list(zone, page);
-			if (!pagevec_add(&pvec, page)) {
-				spin_unlock_irq(&zone->lru_lock);
-				__pagevec_release(&pvec);
-				spin_lock_irq(&zone->lru_lock);
-			}
+	spin_lock_irq(&zone->lru_lock);
+	/*
+	 * Put back any unfreeable pages.
+	 */
+	while (!list_empty(&page_list)) {
+		page = lru_to_page(&page_list);
+		if (TestSetPageLRU(page))
+			BUG();
+		list_del(&page->lru);
+		if (PageActive(page))
+			add_page_to_active_list(zone, page);
+		else
+			add_page_to_inactive_list(zone, page);
+		if (!pagevec_add(&pvec, page)) {
+			spin_unlock_irq(&zone->lru_lock);
+			__pagevec_release(&pvec);
+			spin_lock_irq(&zone->lru_lock);
 		}
-  	}
+	}
 	spin_unlock_irq(&zone->lru_lock);
-done:
+
 	pagevec_release(&pvec);
 }
 
@@ -978,7 +973,6 @@ refill_inactive_zone(struct zone *zone, 
 	int pgmoved;
 	int pgdeactivate = 0;
 	int pgscanned;
-	int nr_pages = sc->nr_to_scan;
 	LIST_HEAD(l_hold);	/* The pages which were snipped off */
 	LIST_HEAD(l_inactive);	/* Pages to go onto the inactive_list */
 	LIST_HEAD(l_active);	/* Pages to go onto the active_list */
@@ -991,7 +985,7 @@ refill_inactive_zone(struct zone *zone, 
 
 	lru_add_drain();
 	spin_lock_irq(&zone->lru_lock);
-	pgmoved = isolate_lru_pages(nr_pages, &zone->active_list,
+	pgmoved = isolate_lru_pages(sc->nr_to_scan, &zone->active_list,
 				    &l_hold, &pgscanned);
 	zone->pages_scanned += pgscanned;
 	zone->nr_active -= pgmoved;

--

[PATCH 09/13] mm: remove swap_cluster_max from scan_control

[Wu Fengguang]

[-- Attachment #1: mm-remove-swap-cluster-max-from-scan-control.patch --]
[-- Type: text/plain, Size: 2384 bytes --]

The use of sc.swap_cluster_max is weird and redundant.

The callers should just set sc.priority/sc.nr_to_reclaim, and let
shrink_zone() decide the proper loop parameters.

Signed-off-by: Wu Fengguang <wfg@mail.ustc.edu.cn>
---

 mm/vmscan.c |   15 ++++-----------
 1 files changed, 4 insertions(+), 11 deletions(-)

--- linux-2.6.15-rc5-mm1.orig/mm/vmscan.c
+++ linux-2.6.15-rc5-mm1/mm/vmscan.c
@@ -76,12 +76,6 @@ struct scan_control {
 
 	/* Can pages be swapped as part of reclaim? */
 	int may_swap;
-
-	/* This context's SWAP_CLUSTER_MAX. If freeing memory for
-	 * suspend, we effectively ignore SWAP_CLUSTER_MAX.
-	 * In this context, it doesn't matter that we scan the
-	 * whole list at once. */
-	int swap_cluster_max;
 };
 
 #define lru_to_page(_head) (list_entry((_head)->prev, struct page, lru))
@@ -1121,7 +1115,6 @@ shrink_zone(struct zone *zone, struct sc
 	nr_inactive &= ~(SWAP_CLUSTER_MAX - 1);
 
 	sc->nr_to_scan = SWAP_CLUSTER_MAX;
-	sc->nr_to_reclaim = sc->swap_cluster_max;
 
 	while (nr_active >= SWAP_CLUSTER_MAX * 1024 || nr_inactive) {
 		if (nr_active >= SWAP_CLUSTER_MAX * 1024) {
@@ -1260,7 +1253,7 @@ int try_to_free_pages(struct zone **zone
 		sc.nr_scanned = 0;
 		sc.nr_reclaimed = 0;
 		sc.priority = priority;
-		sc.swap_cluster_max = SWAP_CLUSTER_MAX;
+		sc.nr_to_reclaim = SWAP_CLUSTER_MAX;
 		if (!priority)
 			disable_swap_token();
 		shrink_caches(zones, &sc);
@@ -1272,7 +1265,7 @@ int try_to_free_pages(struct zone **zone
 		}
 		total_scanned += sc.nr_scanned;
 		total_reclaimed += sc.nr_reclaimed;
-		if (total_reclaimed >= sc.swap_cluster_max) {
+		if (total_reclaimed >= SWAP_CLUSTER_MAX) {
 			ret = 1;
 			goto out;
 		}
@@ -1284,7 +1277,7 @@ int try_to_free_pages(struct zone **zone
 		 * that's undesirable in laptop mode, where we *want* lumpy
 		 * writeout.  So in laptop mode, write out the whole world.
 		 */
-		if (total_scanned > sc.swap_cluster_max + sc.swap_cluster_max/2) {
+		if (total_scanned > SWAP_CLUSTER_MAX * 3 / 2) {
 			wakeup_pdflush(laptop_mode ? 0 : total_scanned);
 			sc.may_writepage = 1;
 		}
@@ -1365,7 +1358,7 @@ loop_again:
 		sc.nr_scanned = 0;
 		sc.nr_reclaimed = 0;
 		sc.priority = priority;
-		sc.swap_cluster_max = nr_pages ? nr_pages : SWAP_CLUSTER_MAX;
+		sc.nr_to_reclaim = nr_pages ? nr_pages : SWAP_CLUSTER_MAX;
 
 		/* The swap token gets in the way of swapout... */
 		if (!priority)

--

[PATCH 10/13] mm: let sc.nr_scanned/sc.nr_reclaimed accumulate

[Wu Fengguang]

[-- Attachment #1: mm-accumulate-nr-scanned-reclaimed-in-scan-control.patch --]
[-- Type: text/plain, Size: 4611 bytes --]

Now that there's no need to keep track of nr_scanned/nr_reclaimed for every
single round of shrink_zone(), remove the total_scanned/total_reclaimed and
let nr_scanned/nr_reclaimed accumulate between shrink_zone() calls.

Signed-off-by: Wu Fengguang <wfg@mail.ustc.edu.cn>
---

 mm/vmscan.c |   36 ++++++++++++++----------------------
 1 files changed, 14 insertions(+), 22 deletions(-)

--- linux-2.6.15-rc5-mm1.orig/mm/vmscan.c
+++ linux-2.6.15-rc5-mm1/mm/vmscan.c
@@ -1225,7 +1225,6 @@ int try_to_free_pages(struct zone **zone
 {
 	int priority;
 	int ret = 0;
-	int total_scanned = 0, total_reclaimed = 0;
 	struct reclaim_state *reclaim_state = current->reclaim_state;
 	struct scan_control sc;
 	int i;
@@ -1235,6 +1234,8 @@ int try_to_free_pages(struct zone **zone
 	sc.gfp_mask = gfp_mask;
 	sc.may_writepage = 0;
 	sc.may_swap = 1;
+	sc.nr_scanned = 0;
+	sc.nr_reclaimed = 0;
 
 	inc_page_state(allocstall);
 
@@ -1250,8 +1251,6 @@ int try_to_free_pages(struct zone **zone
 	/* The added 10 priorities are for scan rate balancing */
 	for (priority = DEF_PRIORITY + 10; priority >= 0; priority--) {
 		sc.nr_mapped = read_page_state(nr_mapped);
-		sc.nr_scanned = 0;
-		sc.nr_reclaimed = 0;
 		sc.priority = priority;
 		sc.nr_to_reclaim = SWAP_CLUSTER_MAX;
 		if (!priority)
@@ -1263,9 +1262,7 @@ int try_to_free_pages(struct zone **zone
 			sc.nr_reclaimed += reclaim_state->reclaimed_slab;
 			reclaim_state->reclaimed_slab = 0;
 		}
-		total_scanned += sc.nr_scanned;
-		total_reclaimed += sc.nr_reclaimed;
-		if (total_reclaimed >= SWAP_CLUSTER_MAX) {
+		if (sc.nr_reclaimed >= SWAP_CLUSTER_MAX) {
 			ret = 1;
 			goto out;
 		}
@@ -1277,13 +1274,13 @@ int try_to_free_pages(struct zone **zone
 		 * that's undesirable in laptop mode, where we *want* lumpy
 		 * writeout.  So in laptop mode, write out the whole world.
 		 */
-		if (total_scanned > SWAP_CLUSTER_MAX * 3 / 2) {
-			wakeup_pdflush(laptop_mode ? 0 : total_scanned);
+		if (sc.nr_scanned > SWAP_CLUSTER_MAX * 3 / 2) {
+			wakeup_pdflush(laptop_mode ? 0 : sc.nr_scanned);
 			sc.may_writepage = 1;
 		}
 
 		/* Take a nap, wait for some writeback to complete */
-		if (sc.nr_scanned && priority < DEF_PRIORITY)
+		if (priority < DEF_PRIORITY)
 			blk_congestion_wait(WRITE, HZ/10);
 	}
 out:
@@ -1329,18 +1326,17 @@ static int balance_pgdat(pg_data_t *pgda
 	int all_zones_ok;
 	int priority;
 	int i;
-	int total_scanned, total_reclaimed;
 	struct reclaim_state *reclaim_state = current->reclaim_state;
 	struct scan_control sc;
 	struct zone *prev_zone = pgdat->node_zones;
 
 loop_again:
-	total_scanned = 0;
-	total_reclaimed = 0;
 	sc.gfp_mask = GFP_KERNEL;
 	sc.may_writepage = 0;
 	sc.may_swap = 1;
 	sc.nr_mapped = read_page_state(nr_mapped);
+	sc.nr_scanned = 0;
+	sc.nr_reclaimed = 0;
 
 	inc_page_state(pageoutrun);
 
@@ -1355,8 +1351,6 @@ loop_again:
 
 	for (priority = DEF_PRIORITY; priority >= 0; priority--) {
 		all_zones_ok = 1;
-		sc.nr_scanned = 0;
-		sc.nr_reclaimed = 0;
 		sc.priority = priority;
 		sc.nr_to_reclaim = nr_pages ? nr_pages : SWAP_CLUSTER_MAX;
 
@@ -1417,19 +1411,17 @@ scan_swspd:
 		reclaim_state->reclaimed_slab = 0;
 		shrink_slab(prev_zone, priority, GFP_KERNEL);
 		sc.nr_reclaimed += reclaim_state->reclaimed_slab;
-		total_reclaimed += sc.nr_reclaimed;
-		total_scanned += sc.nr_scanned;
 
 		/*
 		 * If we've done a decent amount of scanning and
 		 * the reclaim ratio is low, start doing writepage
 		 * even in laptop mode
 		 */
-		if (total_scanned > SWAP_CLUSTER_MAX * 2 &&
-		    total_scanned > total_reclaimed+total_reclaimed/2)
+		if (sc.nr_scanned > SWAP_CLUSTER_MAX * 2 &&
+		    sc.nr_scanned > sc.nr_reclaimed + sc.nr_reclaimed / 2)
 			sc.may_writepage = 1;
 
-		if (nr_pages && to_free > total_reclaimed)
+		if (nr_pages && to_free > sc.nr_reclaimed)
 			continue;	/* swsusp: need to do more work */
 		if (all_zones_ok)
 			break;		/* kswapd: all done */
@@ -1437,7 +1429,7 @@ scan_swspd:
 		 * OK, kswapd is getting into trouble.  Take a nap, then take
 		 * another pass across the zones.
 		 */
-		if (total_scanned && priority < DEF_PRIORITY - 2)
+		if (priority < DEF_PRIORITY - 2)
 			blk_congestion_wait(WRITE, HZ/10);
 
 		/*
@@ -1446,7 +1438,7 @@ scan_swspd:
 		 * matches the direct reclaim path behaviour in terms of impact
 		 * on zone->*_priority.
 		 */
-		if ((total_reclaimed >= SWAP_CLUSTER_MAX) && (!nr_pages))
+		if (sc.nr_reclaimed >= SWAP_CLUSTER_MAX && !nr_pages)
 			break;
 	}
 	for (i = 0; i < pgdat->nr_zones; i++) {
@@ -1459,7 +1451,7 @@ scan_swspd:
 		goto loop_again;
 	}
 
-	return total_reclaimed;
+	return sc.nr_reclaimed;
 }
 
 /*

--

[PATCH 11/13] mm: fold sc.may_writepage and sc.may_swap into sc.flags

[Wu Fengguang]

[-- Attachment #1: mm-fold-bool-variables-into-flags-in-scan-control.patch --]
[-- Type: text/plain, Size: 2436 bytes --]

Fold bool values into flags to make struct scan_control more compact.

Signed-off-by: Wu Fengguang <wfg@mail.ustc.edu.cn>
---

 mm/vmscan.c |   22 ++++++++++------------
 1 files changed, 10 insertions(+), 12 deletions(-)

--- linux-2.6.15-rc5-mm1.orig/mm/vmscan.c
+++ linux-2.6.15-rc5-mm1/mm/vmscan.c
@@ -72,12 +72,12 @@ struct scan_control {
 	/* This context's GFP mask */
 	gfp_t gfp_mask;
 
-	int may_writepage;
-
-	/* Can pages be swapped as part of reclaim? */
-	int may_swap;
+	unsigned long flags;
 };
 
+#define SC_MAY_WRITEPAGE	0x1
+#define SC_MAY_SWAP		0x2	/* Can pages be swapped as part of reclaim? */
+
 #define lru_to_page(_head) (list_entry((_head)->prev, struct page, lru))
 
 #ifdef ARCH_HAS_PREFETCH
@@ -483,7 +483,7 @@ static int shrink_list(struct list_head 
 		 * Try to allocate it some swap space here.
 		 */
 		if (PageAnon(page) && !PageSwapCache(page)) {
-			if (!sc->may_swap)
+			if (!(sc->flags & SC_MAY_SWAP))
 				goto keep_locked;
 			if (!add_to_swap(page, GFP_ATOMIC))
 				goto activate_locked;
@@ -514,7 +514,7 @@ static int shrink_list(struct list_head 
 				goto keep_locked;
 			if (!may_enter_fs)
 				goto keep_locked;
-			if (laptop_mode && !sc->may_writepage)
+			if (laptop_mode && !(sc->flags & SC_MAY_WRITEPAGE))
 				goto keep_locked;
 
 			/* Page is dirty, try to write it out here */
@@ -1232,8 +1232,7 @@ int try_to_free_pages(struct zone **zone
 	delay_prefetch();
 
 	sc.gfp_mask = gfp_mask;
-	sc.may_writepage = 0;
-	sc.may_swap = 1;
+	sc.flags = SC_MAY_SWAP;
 	sc.nr_scanned = 0;
 	sc.nr_reclaimed = 0;
 
@@ -1276,7 +1275,7 @@ int try_to_free_pages(struct zone **zone
 		 */
 		if (sc.nr_scanned > SWAP_CLUSTER_MAX * 3 / 2) {
 			wakeup_pdflush(laptop_mode ? 0 : sc.nr_scanned);
-			sc.may_writepage = 1;
+			sc.flags |= SC_MAY_WRITEPAGE;
 		}
 
 		/* Take a nap, wait for some writeback to complete */
@@ -1332,8 +1331,7 @@ static int balance_pgdat(pg_data_t *pgda
 
 loop_again:
 	sc.gfp_mask = GFP_KERNEL;
-	sc.may_writepage = 0;
-	sc.may_swap = 1;
+	sc.flags = SC_MAY_SWAP;
 	sc.nr_mapped = read_page_state(nr_mapped);
 	sc.nr_scanned = 0;
 	sc.nr_reclaimed = 0;
@@ -1419,7 +1417,7 @@ scan_swspd:
 		 */
 		if (sc.nr_scanned > SWAP_CLUSTER_MAX * 2 &&
 		    sc.nr_scanned > sc.nr_reclaimed + sc.nr_reclaimed / 2)
-			sc.may_writepage = 1;
+			sc.flags |= SC_MAY_WRITEPAGE;
 
 		if (nr_pages && to_free > sc.nr_reclaimed)
 			continue;	/* swsusp: need to do more work */

--

[PATCH 12/13] mm: add page reclaim debug traces

[Wu Fengguang]

[-- Attachment #1: mm-page-reclaim-debug-traces.patch --]
[-- Type: text/plain, Size: 5231 bytes --]

Show the detailed steps of direct/kswapd page reclaim.

To enable the printk traces:
# echo y > /debug/debug_page_reclaim

Sample lines:

reclaim zone3 from kswapd for watermark, prio 12, scan-reclaimed 32-32, age 2626, active to scan 6542, hot+cold+free pages 8842+283558+352
reclaim zone2 from kswapd for aging, prio 12, scan-reclaimed 32-32, age 2626, active to scan 8018, hot+cold+free pages 1693+200036+10360
reclaim zone3 from kswapd for watermark, prio 12, scan-reclaimed 64-64, age 2627, active to scan 7564, hot+cold+free pages 8842+283526+384
reclaim zone2 from kswapd for aging, prio 12, scan-reclaimed 32-32, age 2627, active to scan 8296, hot+cold+free pages 1693+200018+10360
reclaim zone3 from kswapd for watermark, prio 12, scan-reclaimed 64-63, age 2628, active to scan 8587, hot+cold+free pages 8843+283495+416
reclaim zone2 from kswapd for aging, prio 12, scan-reclaimed 32-32, age 2628, active to scan 8574, hot+cold+free pages 1693+200014+10392
reclaim zone3 from kswapd for watermark, prio 12, scan-reclaimed 64-63, age 2628, active to scan 9610, hot+cold+free pages 8844+283465+448
reclaim zone2 from kswapd for aging, prio 12, scan-reclaimed 32-32, age 2628, active to scan 8852, hot+cold+free pages 1693+199996+10424
reclaim zone3 from kswapd for watermark, prio 12, scan-reclaimed 64-64, age 2629, active to scan 10633, hot+cold+free pages 8844+283433+480
reclaim zone2 from kswapd for aging, prio 12, scan-reclaimed 32-32, age 2629, active to scan 9130, hot+cold+free pages 1693+199992+10456
reclaim zone3 from kswapd for watermark, prio 12, scan-reclaimed 64-64, age 2630, active to scan 11656, hot+cold+free pages 8844+283401+512
reclaim zone2 from kswapd for aging, prio 12, scan-reclaimed 32-32, age 2630, active to scan 9408, hot+cold+free pages 1693+199974+10488

Signed-off-by: Wu Fengguang <wfg@mail.ustc.edu.cn>
---


 mm/vmscan.c |   68 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++-
 1 files changed, 67 insertions(+), 1 deletion(-)

--- linux-2.6.15-rc5-mm1.orig/mm/vmscan.c
+++ linux-2.6.15-rc5-mm1/mm/vmscan.c
@@ -38,6 +38,7 @@
 #include <asm/div64.h>
 
 #include <linux/swapops.h>
+#include <linux/debugfs.h>
 
 /* possible outcome of pageout() */
 typedef enum {
@@ -78,6 +79,62 @@ struct scan_control {
 #define SC_MAY_WRITEPAGE	0x1
 #define SC_MAY_SWAP		0x2	/* Can pages be swapped as part of reclaim? */
 
+#define SC_RECLAIM_FROM_KSWAPD		0x10
+#define SC_RECLAIM_FROM_DIRECT		0x20
+#define SC_RECLAIM_FOR_WATERMARK	0x40
+#define SC_RECLAIM_FOR_AGING		0x80
+#define SC_RECLAIM_MASK			0xF0
+
+#ifdef CONFIG_DEBUG_FS
+static u32 debug_page_reclaim;
+
+static inline void debug_reclaim(struct scan_control *sc, unsigned long flags)
+{
+	sc->flags = (sc->flags & ~SC_RECLAIM_MASK) | flags;
+}
+
+static inline void debug_reclaim_report(struct scan_control *sc, struct zone *z)
+{
+	if (!debug_page_reclaim)
+		return;
+
+	printk(KERN_DEBUG "reclaim zone%d from %s for %s, "
+			"prio %d, scan-reclaimed %lu-%lu, age %lu, "
+			"active to scan %lu, "
+			"hot+cold+free pages %lu+%lu+%lu\n",
+			zone_idx(z),
+			(sc->flags & SC_RECLAIM_FROM_KSWAPD) ? "kswapd" :
+			((sc->flags & SC_RECLAIM_FROM_DIRECT) ? "direct" :
+								"early"),
+			(sc->flags & SC_RECLAIM_FOR_AGING) ?
+							"aging" : "watermark",
+			sc->priority, sc->nr_scanned, sc->nr_reclaimed,
+			z->page_age,
+			z->nr_scan_active,
+			z->nr_active, z->nr_inactive, z->free_pages);
+
+	if (atomic_read(&z->reclaim_in_progress))
+		printk(KERN_WARNING "reclaim_in_progress=%d\n",
+					atomic_read(&z->reclaim_in_progress));
+}
+
+static inline void debug_reclaim_init(void)
+{
+	debugfs_create_bool("debug_page_reclaim", 0644, NULL,
+							&debug_page_reclaim);
+}
+#else
+static inline void debug_reclaim(struct scan_control *sc, int flags)
+{
+}
+static inline void debug_reclaim_report(struct scan_control *sc, struct zone *z)
+{
+}
+static inline void debug_reclaim_init(void)
+{
+}
+#endif
+
 #define lru_to_page(_head) (list_entry((_head)->prev, struct page, lru))
 
 #ifdef ARCH_HAS_PREFETCH
@@ -1137,6 +1194,7 @@ shrink_zone(struct zone *zone, struct sc
 
 	atomic_dec(&zone->reclaim_in_progress);
 
+	debug_reclaim_report(sc, zone);
 	throttle_vm_writeout();
 }
 
@@ -1201,11 +1259,14 @@ shrink_caches(struct zone **zones, struc
 			continue;
 		}
 
+		debug_reclaim(sc, SC_RECLAIM_FROM_DIRECT);
 		shrink_zone(zone, sc);
 	}
 
-	if (z)
+	if (z) {
+		debug_reclaim(sc, SC_RECLAIM_FROM_DIRECT|SC_RECLAIM_FOR_AGING);
 		shrink_zone(z, sc);
+	}
 }
  
 /*
@@ -1377,10 +1438,14 @@ loop_again:
 			if (zone < prev_zone &&
 					!zone_watermark_ok(zone, order,
 						zone->pages_high, 0, 0)) {
+				debug_reclaim(&sc, SC_RECLAIM_FROM_KSWAPD |
+						   SC_RECLAIM_FOR_WATERMARK);
 			} else if (!age_gt(zone, prev_zone) &&
 					!zone_watermark_ok(zone, order,
 						zone->pages_high,
 						pgdat->nr_zones - 1, 0)) {
+				debug_reclaim(&sc, SC_RECLAIM_FROM_KSWAPD |
+						   SC_RECLAIM_FOR_AGING);
 			} else {
 				prev_zone = zone;
 				continue;
@@ -1607,6 +1672,7 @@ static int __init kswapd_init(void)
 		= find_task_by_pid(kernel_thread(kswapd, pgdat, CLONE_KERNEL));
 	total_memory = nr_free_pagecache_pages();
 	hotcpu_notifier(cpu_callback, 0);
+	debug_reclaim_init();
 	return 0;
 }
 

--

[PATCH 13/13] mm: fix minor scan count bugs

[Wu Fengguang]

[-- Attachment #1: mm-scan-accounting-fix.patch --]
[-- Type: text/plain, Size: 1130 bytes --]

- in isolate_lru_pages(): reports one more scan. Fix it.
- in shrink_cache(): 0 pages taken does not mean 0 pages scanned. Fix it.

Signed-off-by: Wu Fengguang <wfg@mail.ustc.edu.cn>
---

 mm/vmscan.c |   10 ++++++----
 1 files changed, 6 insertions(+), 4 deletions(-)

--- linux-2.6.15-rc5-mm1.orig/mm/vmscan.c
+++ linux-2.6.15-rc5-mm1/mm/vmscan.c
@@ -916,7 +916,8 @@ static int isolate_lru_pages(int nr_to_s
 	struct page *page;
 	int scan = 0;
 
-	while (scan++ < nr_to_scan && !list_empty(src)) {
+	while (scan < nr_to_scan && !list_empty(src)) {
+		scan++;
 		page = lru_to_page(src);
 		prefetchw_prev_lru_page(page, src, flags);
 
@@ -963,14 +964,15 @@ static void shrink_cache(struct zone *zo
 	update_zone_age(zone, nr_scan);
 	spin_unlock_irq(&zone->lru_lock);
 
-	if (nr_taken == 0)
-		return;
-
 	sc->nr_scanned += nr_scan;
 	if (current_is_kswapd())
 		mod_page_state_zone(zone, pgscan_kswapd, nr_scan);
 	else
 		mod_page_state_zone(zone, pgscan_direct, nr_scan);
+
+	if (nr_taken == 0)
+		return;
+
 	nr_freed = shrink_list(&page_list, sc);
 	if (current_is_kswapd())
 		mod_page_state(kswapd_steal, nr_freed);

--

Re: [PATCH 05/13] mm: balance zone aging in kswapd reclaim path

[Wu Fengguang]

Here is a simple test.
It concurrently copies two sparse files of 601M/1.6G in qemu.

The balance is not good enough for now, but is expected to improve much when
shrink_zone() scans small enough in one run(say, < nr_inactive/100).

RESULTS
=======
root ~# ./show-aging-rate.sh
Linux (none) 2.6.15-rc5-mm1 #4 SMP Tue Dec 6 21:27:36 CST 2005 i686 GNU/Linux
             total       used       free     shared    buffers     cached
Mem:          1138       1119         18          0          0       1104
-/+ buffers/cache:         14       1123
Swap:            0          0          0

---------------------------------------------------------------
active/inactive size ratios:
    DMA0:   78 / 1000 =       141 /      1803
 Normal0:  372 / 1000 =     58453 /    156728
HighMem0:  415 / 1000 =     19471 /     46875

active/inactive scan rates:
     DMA:   45 / 1000 =        3238 / (      61920 +        9280)
  Normal:  279 / 1000 =     1509888 / (    5272608 +      133024)
 HighMem:  437 / 1000 =      645536 / (    1430112 +       44032)

---------------------------------------------------------------
inactive size ratios:
    DMA0 /  Normal0:  115 / 10000 =      1803 /    156733
 Normal0 / HighMem0: 33420 / 10000 =    156733 /     46896

inactive scan rates:
     DMA /   Normal:  131 / 10000 = (      61920 +        9280) / (    5272608 +      133024)
  Normal /  HighMem: 36669 / 10000 = (    5272608 +      133024) / (    1430112 +       44032)
root ~# grep "age " /proc/zoneinfo
        age      3085
        age      3072
        age      3072
root ~# grep -E '(low|high|free|protection:) ' /proc/zoneinfo
  pages free     1161
        low      21
        high     25
        protection: (0, 0, 880, 1140)
  pages free     3420
        low      1173
        high     1408
        protection: (0, 0, 0, 2080)
  pages free     132
        low      134
        high     203
        protection: (0, 0, 0, 0)
root ~# vmstat 5 10
procs -----------memory---------- ---swap-- -----io---- --system-- ----cpu----
 r  b   swpd   free   buff  cache   si   so    bi    bo   in    cs us sy id wa
 1  0      0  19544    200 1130844    0    0    40     9 1102    61  6 86  7  1
 1  4      0  16936    208 1133012    0    0     0    22 1043    30  7 93  0  0
 1  0      0  19404    208 1130904    0    0     0     8  994    83  5 95  0  0
 1  0      0  19152    160 1130612    0    0     0     8 1018    85  3 93  0  3
 2  0      0  19152    160 1130952    0    0     0     8  997    56  5 95  0  0
 2  0      0  18372    168 1131896    0    0     0     3 1000    50  5 95  0  0
 1  0      0  18672    112 1131544    0    0     0     7 1014    70  5 95  0  0
 1  0      0  19320    112 1131204    0    0     0     2  989    81  4 96  0  0
 2  0      0  19152    108 1131276    0    0     1     0  996    79  5 95  0  0
 2  0      0  18216     96 1132444    0    0     0     8 1015    84  5 95  0  0

pages_high+lowmem_reserve
203 + 1408 + 2080 + 25 + 1140 = 4856 (x4 = 19424)

SCRIPTS
=======
# cat test-aging2
#!/bin/zsh

while true
do
        cp cold /dev/null
        sleep $((RANDOM%20))
done &

while true
do
        cp hot /dev/null
        sleep $((RANDOM%10))
done

# cat show-aging-rate.sh
#!/bin/sh

uname -a
free -m

echo
echo ---------------------------------------------------------------
echo active/inactive size ratios:

egrep '(zone|active|inactive)' /proc/zoneinfo |
while true
do
        read a b c d
        [[ -z $a ]] && break
        if [[ $c = "zone" ]]; then
                prev_node=$node
                prev_zone=$zone
                node=${b%,}
                zone=$d$node
        else
                eval $a=$b
                if [[ $a = "inactive" ]]; then
                        printf "%8s: %4d / 1000 = %9d / %9d\n" $zone \
                                $((active * 1000 / (1 + inactive))) \
                                $active $inactive
                fi
        fi
done

while true
do
        read name value
        [[ -z $name ]] && break
        eval $name=$value
done < /proc/vmstat

echo
echo active/inactive scan rates:

printf "     DMA: %4d / 1000 = %11d / (%11d + %11d)\n" \
        $((pgrefill_dma * 1000 / (1 + pgscan_kswapd_dma + pgscan_direct_dma))) \
        $pgrefill_dma $pgscan_kswapd_dma $pgscan_direct_dma

[[ $pgscan_kswapd_dma32 != 0 ]] && \
printf "   DMA32: %4d / 1000 = %11d / (%11d + %11d)\n" \
        $((pgrefill_dma32 * 1000 / (1 + pgscan_kswapd_dma32 + pgscan_direct_dma32))) \
        $pgrefill_dma32 $pgscan_kswapd_dma32 $pgscan_direct_dma32

printf "  Normal: %4d / 1000 = %11d / (%11d + %11d)\n" \
        $((pgrefill_normal * 1000 / (1 + pgscan_kswapd_normal + pgscan_direct_normal))) \
        $pgrefill_normal $pgscan_kswapd_normal $pgscan_direct_normal

[[ $pgscan_kswapd_high != 0 ]] && \
printf " HighMem: %4d / 1000 = %11d / (%11d + %11d)\n" \
        $((pgrefill_high * 1000 / (1 + pgscan_kswapd_high + pgscan_direct_high))) \
        $pgrefill_high $pgscan_kswapd_high $pgscan_direct_high

echo
echo ---------------------------------------------------------------
echo inactive size ratios:

egrep '(zone|inactive)' /proc/zoneinfo |
while true
do
        read a b c d
        [[ -z $a ]] && break
        if [[ $c = "zone" ]]; then
                prev_node=$node
                prev_zone=$zone
                node=${b%,}
                zone=$d$node
        else
                prev_inactive=$inactive
                eval $a=$b
                if [[ $prev_node = $node ]]; then
                        printf "%8s / %8s: %4d / 10000 = %9d / %9d\n" \
                                $prev_zone $zone \
                                $((prev_inactive * 10000 / (1 + inactive))) \
                                $prev_inactive $inactive
                fi
        fi
done

echo
echo inactive scan rates:

[[ $pgscan_kswapd_dma != 0 ]] && \
printf "%8s / %8s: %4d / 10000 = (%11d + %11d) / (%11d + %11d)\n" \
        "DMA" "Normal" \
        $(((1 + pgscan_kswapd_dma + pgscan_direct_dma)* 10000 /\
        (1 + pgscan_kswapd_normal + pgscan_direct_normal))) \
        $pgscan_kswapd_dma $pgscan_direct_dma \
        $pgscan_kswapd_normal $pgscan_direct_normal

[[ $pgscan_kswapd_dma32 != 0 ]] && \
printf "%8s / %8s: %4d / 10000 = (%11d + %11d) / (%11d + %11d)\n" \
        "DMA32" "Normal" \
        $(((1 + pgscan_kswapd_dma32 + pgscan_direct_dma32)* 10000 /\
        (1 + pgscan_kswapd_normal + pgscan_direct_normal))) \
        $pgscan_kswapd_dma32 $pgscan_direct_dma32 \
        $pgscan_kswapd_normal $pgscan_direct_normal

[[ $pgscan_kswapd_high != 0 ]] && \
printf "%8s / %8s: %4d / 10000 = (%11d + %11d) / (%11d + %11d)\n" \
        "Normal" "HighMem" \
        $(((1 + pgscan_kswapd_normal + pgscan_direct_normal)* 10000 /\
        (1 + pgscan_kswapd_high + pgscan_direct_high))) \
        $pgscan_kswapd_normal $pgscan_direct_normal \
        $pgscan_kswapd_high $pgscan_direct_high