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.
The current slab shrinking code is way too fragile. It should manage aging
pace by itself, and provide a simple and robust interface.
THE METHOD
==========
This patch adds three variables in struct zone and shrinker
- 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 main design aspects of page balancing:
- In the kswapd reclaim path,
- reclaim also the least aged zone to help it catch up with the most
aged zone;
- the reclaim for watermark is avoided whenever possible by calling
watermark_ok() with classzone_idx = 0;
- in this way, the reclaims for aging are garuanteed to be more than
reclaims for watermark if there's a big imbalance, so that there
will not be any chance for a growing gap.
- In the direct reclaim path,
- reclaim only the least aged local/headless zone in the first two
rounds;
- two rounds are enough in normal to get enough free pages;
- and prevents unnecessarily disturbing remote nodes.
- In shrink_cache()/shrink_zone()/shrink_list(),
- zone->nr_scan_inactive is purged, the requested zone will be
scanned at least SWAP_CLUSTER_MAX pages;
- sc->nr_scanned reflects exactly the number of cold pages scanned.
- In shrink_slab(),
- keep the age of slabs in line with that of the largest zone, using
the same syncing logic as that of the zones;
- a minimal number of unused slabs are reserved in normal;
- the slabs are shrinked more if vm pressure is high;
- `mmap pages found' - `shrink more slabs' - `avoid swapping',
I don't think they are strongly related, so the code is removed.
THE RESULT
==========
- It does not touch the time critical page alloc path, so will not hurt
performance at all.
- The shrink of slabs become simple, consistent and robust. The callers no
longer have to worry about the tricky aging thing.
- Experiments with 64M/512M/2G memory show that it keeps the zone ages in
sync perfectly. One can check the effects by running:
tar c / | cat > /dev/null &
watch -n1 'grep "age " /proc/zoneinfo'
Signed-off-by: Wu Fengguang <[email protected]>
---
include/linux/mm.h | 4
include/linux/mmzone.h | 14 +++
mm/page_alloc.c | 16 ++-
mm/vmscan.c | 222 +++++++++++++++++++++++++++++++------------------
4 files changed, 173 insertions(+), 83 deletions(-)
--- linux-2.6.15-rc2-mm1.orig/include/linux/mmzone.h
+++ linux-2.6.15-rc2-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-rc2-mm1.orig/include/linux/mm.h
+++ linux-2.6.15-rc2-mm1/include/linux/mm.h
@@ -775,7 +775,9 @@ 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;
};
--- linux-2.6.15-rc2-mm1.orig/mm/page_alloc.c
+++ linux-2.6.15-rc2-mm1/mm/page_alloc.c
@@ -1468,6 +1468,8 @@ void show_free_areas(void)
" active:%lukB"
" inactive:%lukB"
" present:%lukB"
+ " aging:%lukB"
+ " age:%lu"
" pages_scanned:%lu"
" all_unreclaimable? %s"
"\n",
@@ -1479,6 +1481,8 @@ 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")
);
@@ -2089,9 +2093,11 @@ 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;
+ zone->aging_milestone = 0;
+ zone->page_age = 0;
atomic_set(&zone->reclaim_in_progress, 0);
if (!size)
continue;
@@ -2240,7 +2246,9 @@ static int zoneinfo_show(struct seq_file
"\n high %lu"
"\n active %lu"
"\n inactive %lu"
- "\n scanned %lu (a: %lu i: %lu)"
+ "\n aging %lu"
+ "\n age %lu"
+ "\n scanned %lu (a: %lu)"
"\n spanned %lu"
"\n present %lu",
zone->free_pages,
@@ -2249,8 +2257,10 @@ 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->nr_scan_active,
zone->spanned_pages,
zone->present_pages);
seq_printf(m,
--- linux-2.6.15-rc2-mm1.orig/mm/vmscan.c
+++ linux-2.6.15-rc2-mm1/mm/vmscan.c
@@ -123,6 +123,56 @@ 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_MASK ((1 << PAGE_AGE_SHIFT) - 1)
+
+/*
+ * The simplified code is: (a->page_age > b->page_age)
+ * The complexity deals with the wrap-around problem.
+ * Two page ages not close enough should also be ignored:
+ * they are out of sync and the comparison may be nonsense.
+ */
+#define pages_more_aged(a, b) \
+ ((b->page_age - a->page_age) & PAGE_AGE_MASK) > \
+ PAGE_AGE_MASK - (1 << (PAGE_AGE_SHIFT - 3)) \
+
+/*
+ * 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 + z->free_pages;
+
+ 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;
+}
+
+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
*/
@@ -134,7 +184,9 @@ 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);
@@ -170,80 +222,61 @@ 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.
- *
- * 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.
+ * If the vm pressure is high, shrink the slabs more.
*
* Returns the number of slab objects which we shrunk.
*/
-static int shrink_slab(unsigned long scanned, gfp_t gfp_mask,
- unsigned long lru_pages)
+static int shrink_slab(gfp_t gfp_mask)
{
struct shrinker *shrinker;
- int ret = 0;
-
- if (scanned == 0)
- scanned = SWAP_CLUSTER_MAX;
+ struct pglist_data *pgdat;
+ struct zone *zone;
+ int n;
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;
- }
+ /* find the major zone for the slabs to catch up age with */
+ pgdat = NODE_DATA(numa_node_id());
+ zone = pgdat->node_zones;
+ for (n = 1; n < pgdat->nr_zones; n++) {
+ struct zone *z = pgdat->node_zones + n;
- /*
- * 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;
+ if (zone->present_pages < z->present_pages)
+ zone = z;
+ }
- while (total_scan >= SHRINK_BATCH) {
- long this_scan = SHRINK_BATCH;
- int shrink_ret;
+ n = 0;
+ list_for_each_entry(shrinker, &shrinker_list, list) {
+ while (pages_more_aged(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 <= SHRINK_BATCH * zone->prev_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;
+
+ n += 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,
+ SHRINK_BATCH * shrinker->seeks *
+ zone->prev_priority);
cond_resched();
}
-
- shrinker->nr += total_scan;
}
up_read(&shrinker_rwsem);
- return ret;
+ return n;
}
/* Called without lock on whether page is mapped, so answer is unstable */
@@ -446,11 +479,6 @@ 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;
@@ -894,11 +922,13 @@ 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)
goto done;
+ sc->nr_scanned += nr_scan;
max_scan -= nr_scan;
if (current_is_kswapd())
mod_page_state_zone(zone, pgscan_kswapd, nr_scan);
@@ -1101,12 +1131,7 @@ shrink_zone(struct zone *zone, struct sc
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_inactive = (zone->nr_inactive >> sc->priority) + 1;
sc->nr_to_reclaim = sc->swap_cluster_max;
@@ -1153,6 +1178,7 @@ 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];
@@ -1170,8 +1196,31 @@ shrink_caches(struct zone **zones, struc
if (zone->all_unreclaimable && sc->priority != DEF_PRIORITY)
continue; /* Let kswapd poll it */
+ /*
+ * Balance page aging in local zones and following headless
+ * zones in the first two rounds of direct reclaim.
+ */
+ if (sc->priority > DEF_PRIORITY - 2) {
+ 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 (pages_more_aged(z, zone))
+ z = zone;
+
+ continue;
+ }
+
shrink_zone(zone, sc);
}
+
+ if (z)
+ shrink_zone(z, sc);
}
/*
@@ -1194,7 +1243,6 @@ 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();
@@ -1212,7 +1260,6 @@ int try_to_free_pages(struct zone **zone
continue;
zone->temp_priority = DEF_PRIORITY;
- lru_pages += zone->nr_active + zone->nr_inactive;
}
for (priority = DEF_PRIORITY; priority >= 0; priority--) {
@@ -1222,7 +1269,7 @@ int try_to_free_pages(struct zone **zone
sc.priority = priority;
sc.swap_cluster_max = SWAP_CLUSTER_MAX;
shrink_caches(zones, &sc);
- shrink_slab(sc.nr_scanned, gfp_mask, lru_pages);
+ shrink_slab(gfp_mask);
if (reclaim_state) {
sc.nr_reclaimed += reclaim_state->reclaimed_slab;
reclaim_state->reclaimed_slab = 0;
@@ -1296,6 +1343,23 @@ 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 *youngest_zone;
+ struct zone *oldest_zone;
+
+ youngest_zone = oldest_zone = NULL;
+ for (i = 0; i < pgdat->nr_zones; i++) {
+ struct zone *zone = pgdat->node_zones + i;
+
+ if (zone->present_pages == 0)
+ continue;
+
+ if (!oldest_zone)
+ youngest_zone = oldest_zone = zone;
+ else if (pages_more_aged(zone, oldest_zone))
+ oldest_zone = zone;
+ else if (pages_more_aged(youngest_zone, zone))
+ youngest_zone = zone;
+ }
loop_again:
total_scanned = 0;
@@ -1314,9 +1378,6 @@ loop_again:
}
for (priority = DEF_PRIORITY; priority >= 0; priority--) {
- unsigned long lru_pages = 0;
- int first_low_zone = 0;
-
all_zones_ok = 1;
sc.nr_scanned = 0;
sc.nr_reclaimed = 0;
@@ -1331,13 +1392,17 @@ loop_again:
continue;
if (nr_pages == 0) { /* Not software suspend */
- if (zone_watermark_ok(zone, order,
- zone->pages_high, first_low_zone, 0))
+ if (!zone_watermark_ok(zone, order,
+ zone->pages_high,
+ 0, 0)) {
+ all_zones_ok = 0;
+ } else if (zone == youngest_zone &&
+ pages_more_aged(oldest_zone,
+ youngest_zone)) {
+ /* if (priority > DEF_PRIORITY - 2) */
+ /* all_zones_ok = 0; */
+ } else
continue;
-
- all_zones_ok = 0;
- if (first_low_zone < i)
- first_low_zone = i;
}
if (zone->all_unreclaimable && priority != DEF_PRIORITY)
@@ -1346,7 +1411,6 @@ loop_again:
zone->temp_priority = priority;
if (zone->prev_priority > priority)
zone->prev_priority = priority;
- lru_pages += zone->nr_active + zone->nr_inactive;
atomic_inc(&zone->reclaim_in_progress);
shrink_zone(zone, &sc);
@@ -1357,7 +1421,7 @@ loop_again:
zone->all_unreclaimable = 1;
}
reclaim_state->reclaimed_slab = 0;
- shrink_slab(sc.nr_scanned, GFP_KERNEL, lru_pages);
+ shrink_slab(GFP_KERNEL);
sc.nr_reclaimed += reclaim_state->reclaimed_slab;
total_reclaimed += sc.nr_reclaimed;
total_scanned += sc.nr_scanned;
--