mm: workingset: per-cgroup cache thrash detection

Cache thrash detection (see a528910e12ec "mm: thrash detection-based
file cache sizing" for details) currently only works on the system
level, not inside cgroups.  Worse, as the refaults are compared to the
global number of active cache, cgroups might wrongfully get all their
refaults activated when their pages are hotter than those of others.

Move the refault machinery from the zone to the lruvec, and then tag
eviction entries with the memcg ID.  This makes the thrash detection
work correctly inside cgroups.

[sergey.senozhatsky@gmail.com: do not return from workingset_activation() with locked rcu and page]
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Reviewed-by: Vladimir Davydov <vdavydov@virtuozzo.com>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

3 files changed, 78 insertions, 44 deletions

diff --git a/mm/memcontrol.c b/mm/memcontrol.c
index 953f0f984392..864e237f32d9 100644
--- a/mm/memcontrol.c
+++ b/mm/memcontrol.c
@@ -268,31 +268,6 @@ static inline bool mem_cgroup_is_root(struct mem_cgroup *memcg)
 	return (memcg == root_mem_cgroup);
 }
 
-/*
- * We restrict the id in the range of [1, 65535], so it can fit into
- * an unsigned short.
- */
-#define MEM_CGROUP_ID_MAX	USHRT_MAX
-
-static inline unsigned short mem_cgroup_id(struct mem_cgroup *memcg)
-{
-	return memcg->css.id;
-}
-
-/*
- * A helper function to get mem_cgroup from ID. must be called under
- * rcu_read_lock().  The caller is responsible for calling
- * css_tryget_online() if the mem_cgroup is used for charging. (dropping
- * refcnt from swap can be called against removed memcg.)
- */
-static inline struct mem_cgroup *mem_cgroup_from_id(unsigned short id)
-{
-	struct cgroup_subsys_state *css;
-
-	css = css_from_id(id, &memory_cgrp_subsys);
-	return mem_cgroup_from_css(css);
-}
-
 #ifndef CONFIG_SLOB
 /*
  * This will be the memcg's index in each cache's ->memcg_params.memcg_caches.
diff --git a/mm/vmscan.c b/mm/vmscan.c
index 08547a7136d3..fd434cc89bea 100644
--- a/mm/vmscan.c
+++ b/mm/vmscan.c
@@ -213,7 +213,7 @@ bool zone_reclaimable(struct zone *zone)
 		zone_reclaimable_pages(zone) * 6;
 }
 
-static unsigned long get_lru_size(struct lruvec *lruvec, enum lru_list lru)
+unsigned long lruvec_lru_size(struct lruvec *lruvec, enum lru_list lru)
 {
 	if (!mem_cgroup_disabled())
 		return mem_cgroup_get_lru_size(lruvec, lru);
@@ -1923,8 +1923,8 @@ static bool inactive_file_is_low(struct lruvec *lruvec)
 	unsigned long inactive;
 	unsigned long active;
 
-	inactive = get_lru_size(lruvec, LRU_INACTIVE_FILE);
-	active = get_lru_size(lruvec, LRU_ACTIVE_FILE);
+	inactive = lruvec_lru_size(lruvec, LRU_INACTIVE_FILE);
+	active = lruvec_lru_size(lruvec, LRU_ACTIVE_FILE);
 
 	return active > inactive;
 }
@@ -2063,7 +2063,7 @@ static void get_scan_count(struct lruvec *lruvec, struct mem_cgroup *memcg,
 	 * system is under heavy pressure.
 	 */
 	if (!inactive_file_is_low(lruvec) &&
-	    get_lru_size(lruvec, LRU_INACTIVE_FILE) >> sc->priority) {
+	    lruvec_lru_size(lruvec, LRU_INACTIVE_FILE) >> sc->priority) {
 		scan_balance = SCAN_FILE;
 		goto out;
 	}
@@ -2089,10 +2089,10 @@ static void get_scan_count(struct lruvec *lruvec, struct mem_cgroup *memcg,
 	 * anon in [0], file in [1]
 	 */
 
-	anon  = get_lru_size(lruvec, LRU_ACTIVE_ANON) +
-		get_lru_size(lruvec, LRU_INACTIVE_ANON);
-	file  = get_lru_size(lruvec, LRU_ACTIVE_FILE) +
-		get_lru_size(lruvec, LRU_INACTIVE_FILE);
+	anon  = lruvec_lru_size(lruvec, LRU_ACTIVE_ANON) +
+		lruvec_lru_size(lruvec, LRU_INACTIVE_ANON);
+	file  = lruvec_lru_size(lruvec, LRU_ACTIVE_FILE) +
+		lruvec_lru_size(lruvec, LRU_INACTIVE_FILE);
 
 	spin_lock_irq(&zone->lru_lock);
 	if (unlikely(reclaim_stat->recent_scanned[0] > anon / 4)) {
@@ -2130,7 +2130,7 @@ out:
 			unsigned long size;
 			unsigned long scan;
 
-			size = get_lru_size(lruvec, lru);
+			size = lruvec_lru_size(lruvec, lru);
 			scan = size >> sc->priority;
 
 			if (!scan && pass && force_scan)
diff --git a/mm/workingset.c b/mm/workingset.c
index 9a26a60368d2..14bc23a7779b 100644
--- a/mm/workingset.c
+++ b/mm/workingset.c
@@ -153,7 +153,8 @@
  */
 
 #define EVICTION_SHIFT	(RADIX_TREE_EXCEPTIONAL_ENTRY + \
-			 ZONES_SHIFT + NODES_SHIFT)
+			 ZONES_SHIFT + NODES_SHIFT +	\
+			 MEM_CGROUP_ID_SHIFT)
 #define EVICTION_MASK	(~0UL >> EVICTION_SHIFT)
 
 /*
@@ -166,9 +167,10 @@
  */
 static unsigned int bucket_order __read_mostly;
 
-static void *pack_shadow(unsigned long eviction, struct zone *zone)
+static void *pack_shadow(int memcgid, struct zone *zone, unsigned long eviction)
 {
 	eviction >>= bucket_order;
+	eviction = (eviction << MEM_CGROUP_ID_SHIFT) | memcgid;
 	eviction = (eviction << NODES_SHIFT) | zone_to_nid(zone);
 	eviction = (eviction << ZONES_SHIFT) | zone_idx(zone);
 	eviction = (eviction << RADIX_TREE_EXCEPTIONAL_SHIFT);
@@ -176,18 +178,21 @@ static void *pack_shadow(unsigned long eviction, struct zone *zone)
 	return (void *)(eviction | RADIX_TREE_EXCEPTIONAL_ENTRY);
 }
 
-static void unpack_shadow(void *shadow, struct zone **zonep,
+static void unpack_shadow(void *shadow, int *memcgidp, struct zone **zonep,
 			  unsigned long *evictionp)
 {
 	unsigned long entry = (unsigned long)shadow;
-	int zid, nid;
+	int memcgid, nid, zid;
 
 	entry >>= RADIX_TREE_EXCEPTIONAL_SHIFT;
 	zid = entry & ((1UL << ZONES_SHIFT) - 1);
 	entry >>= ZONES_SHIFT;
 	nid = entry & ((1UL << NODES_SHIFT) - 1);
 	entry >>= NODES_SHIFT;
+	memcgid = entry & ((1UL << MEM_CGROUP_ID_SHIFT) - 1);
+	entry >>= MEM_CGROUP_ID_SHIFT;
 
+	*memcgidp = memcgid;
 	*zonep = NODE_DATA(nid)->node_zones + zid;
 	*evictionp = entry << bucket_order;
 }
@@ -202,11 +207,20 @@ static void unpack_shadow(void *shadow, struct zone **zonep,
  */
 void *workingset_eviction(struct address_space *mapping, struct page *page)
 {
+	struct mem_cgroup *memcg = page_memcg(page);
 	struct zone *zone = page_zone(page);
+	int memcgid = mem_cgroup_id(memcg);
 	unsigned long eviction;
+	struct lruvec *lruvec;
 
-	eviction = atomic_long_inc_return(&zone->inactive_age);
-	return pack_shadow(eviction, zone);
+	/* Page is fully exclusive and pins page->mem_cgroup */
+	VM_BUG_ON_PAGE(PageLRU(page), page);
+	VM_BUG_ON_PAGE(page_count(page), page);
+	VM_BUG_ON_PAGE(!PageLocked(page), page);
+
+	lruvec = mem_cgroup_zone_lruvec(zone, memcg);
+	eviction = atomic_long_inc_return(&lruvec->inactive_age);
+	return pack_shadow(memcgid, zone, eviction);
 }
 
 /**
@@ -221,13 +235,42 @@ void *workingset_eviction(struct address_space *mapping, struct page *page)
 bool workingset_refault(void *shadow)
 {
 	unsigned long refault_distance;
+	unsigned long active_file;
+	struct mem_cgroup *memcg;
 	unsigned long eviction;
+	struct lruvec *lruvec;
 	unsigned long refault;
 	struct zone *zone;
+	int memcgid;
 
-	unpack_shadow(shadow, &zone, &eviction);
+	unpack_shadow(shadow, &memcgid, &zone, &eviction);
 
-	refault = atomic_long_read(&zone->inactive_age);
+	rcu_read_lock();
+	/*
+	 * Look up the memcg associated with the stored ID. It might
+	 * have been deleted since the page's eviction.
+	 *
+	 * Note that in rare events the ID could have been recycled
+	 * for a new cgroup that refaults a shared page. This is
+	 * impossible to tell from the available data. However, this
+	 * should be a rare and limited disturbance, and activations
+	 * are always speculative anyway. Ultimately, it's the aging
+	 * algorithm's job to shake out the minimum access frequency
+	 * for the active cache.
+	 *
+	 * XXX: On !CONFIG_MEMCG, this will always return NULL; it
+	 * would be better if the root_mem_cgroup existed in all
+	 * configurations instead.
+	 */
+	memcg = mem_cgroup_from_id(memcgid);
+	if (!mem_cgroup_disabled() && !memcg) {
+		rcu_read_unlock();
+		return false;
+	}
+	lruvec = mem_cgroup_zone_lruvec(zone, memcg);
+	refault = atomic_long_read(&lruvec->inactive_age);
+	active_file = lruvec_lru_size(lruvec, LRU_ACTIVE_FILE);
+	rcu_read_unlock();
 
 	/*
 	 * The unsigned subtraction here gives an accurate distance
@@ -249,7 +292,7 @@ bool workingset_refault(void *shadow)
 
 	inc_zone_state(zone, WORKINGSET_REFAULT);
 
-	if (refault_distance <= zone_page_state(zone, NR_ACTIVE_FILE)) {
+	if (refault_distance <= active_file) {
 		inc_zone_state(zone, WORKINGSET_ACTIVATE);
 		return true;
 	}
@@ -262,7 +305,23 @@ bool workingset_refault(void *shadow)
  */
 void workingset_activation(struct page *page)
 {
-	atomic_long_inc(&page_zone(page)->inactive_age);
+	struct mem_cgroup *memcg;
+	struct lruvec *lruvec;
+
+	memcg = lock_page_memcg(page);
+	/*
+	 * Filter non-memcg pages here, e.g. unmap can call
+	 * mark_page_accessed() on VDSO pages.
+	 *
+	 * XXX: See workingset_refault() - this should return
+	 * root_mem_cgroup even for !CONFIG_MEMCG.
+	 */
+	if (!mem_cgroup_disabled() && !memcg)
+		goto out;
+	lruvec = mem_cgroup_zone_lruvec(page_zone(page), memcg);
+	atomic_long_inc(&lruvec->inactive_age);
+out:
+	unlock_page_memcg(memcg);
 }
 
 /*