X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=datapath%2Fflow_table.c;h=c8bd9d1eddc4d8c663f74585b50e75bc22a49b09;hb=003ce655b7116d18c86a74c50391e54990346931;hp=e8e8511be909d6ad90f223497af6aa8dfe1b63c0;hpb=a097c0b230333cbc038be4ec67bf34b5b26c5112;p=sliver-openvswitch.git

diff --git a/datapath/flow_table.c b/datapath/flow_table.c
index e8e8511be..c8bd9d1ed 100644
--- a/datapath/flow_table.c
+++ b/datapath/flow_table.c
@@ -25,7 +25,7 @@
 #include <linux/if_vlan.h>
 #include <net/llc_pdu.h>
 #include <linux/kernel.h>
-#include <linux/jhash.h>
+#include <linux/hash.h>
 #include <linux/jiffies.h>
 #include <linux/llc.h>
 #include <linux/module.h>
@@ -46,7 +46,16 @@
 
 #include "vlan.h"
 
+#define TBL_MIN_BUCKETS		1024
+#define MASK_ARRAY_SIZE_MIN	16
+#define REHASH_INTERVAL		(10 * 60 * HZ)
+
+#define MC_HASH_SHIFT		8
+#define MC_HASH_ENTRIES		(1u << MC_HASH_SHIFT)
+#define MC_HASH_SEGS		((sizeof(uint32_t) * 8) / MC_HASH_SHIFT)
+
 static struct kmem_cache *flow_cache;
+struct kmem_cache *flow_stats_cache __read_mostly;
 
 static u16 range_n_bytes(const struct sw_flow_key_range *range)
 {
@@ -56,8 +65,10 @@ static u16 range_n_bytes(const struct sw_flow_key_range *range)
 void ovs_flow_mask_key(struct sw_flow_key *dst, const struct sw_flow_key *src,
 		       const struct sw_flow_mask *mask)
 {
-	const long *m = (long *)((u8 *)&mask->key + mask->range.start);
-	const long *s = (long *)((u8 *)src + mask->range.start);
+	const long *m = (const long *)((const u8 *)&mask->key +
+				mask->range.start);
+	const long *s = (const long *)((const u8 *)src +
+				mask->range.start);
 	long *d = (long *)((u8 *)dst + mask->range.start);
 	int i;
 
@@ -72,16 +83,40 @@ void ovs_flow_mask_key(struct sw_flow_key *dst, const struct sw_flow_key *src,
 struct sw_flow *ovs_flow_alloc(void)
 {
 	struct sw_flow *flow;
+	struct flow_stats *stats;
+	int node;
 
 	flow = kmem_cache_alloc(flow_cache, GFP_KERNEL);
 	if (!flow)
 		return ERR_PTR(-ENOMEM);
 
-	spin_lock_init(&flow->lock);
 	flow->sf_acts = NULL;
 	flow->mask = NULL;
+	flow->stats_last_writer = NUMA_NO_NODE;
+
+	/* Initialize the default stat node. */
+	stats = kmem_cache_alloc_node(flow_stats_cache,
+				      GFP_KERNEL | __GFP_ZERO, 0);
+	if (!stats)
+		goto err;
+
+	spin_lock_init(&stats->lock);
+
+	RCU_INIT_POINTER(flow->stats[0], stats);
+
+	for_each_node(node)
+		if (node != 0)
+			RCU_INIT_POINTER(flow->stats[node], NULL);
 
 	return flow;
+err:
+	kmem_cache_free(flow_cache, flow);
+	return ERR_PTR(-ENOMEM);
+}
+
+int ovs_flow_tbl_count(struct flow_table *table)
+{
+	return table->count;
 }
 
 static struct flex_array *alloc_buckets(unsigned int n_buckets)
@@ -109,7 +144,13 @@ static struct flex_array *alloc_buckets(unsigned int n_buckets)
 
 static void flow_free(struct sw_flow *flow)
 {
-	kfree((struct sf_flow_acts __force *)flow->sf_acts);
+	int node;
+
+	kfree((struct sw_flow_actions __force *)flow->sf_acts);
+	for_each_node(node)
+		if (flow->stats[node])
+			kmem_cache_free(flow_stats_cache,
+					(struct flow_stats __force *)flow->stats[node]);
 	kmem_cache_free(flow_cache, flow);
 }
 
@@ -120,13 +161,18 @@ static void rcu_free_flow_callback(struct rcu_head *rcu)
 	flow_free(flow);
 }
 
+static void rcu_free_sw_flow_mask_cb(struct rcu_head *rcu)
+{
+	struct sw_flow_mask *mask = container_of(rcu, struct sw_flow_mask, rcu);
+
+	kfree(mask);
+}
+
 void ovs_flow_free(struct sw_flow *flow, bool deferred)
 {
 	if (!flow)
 		return;
 
-	ovs_sw_flow_mask_del_ref(flow->mask, deferred);
-
 	if (deferred)
 		call_rcu(&flow->rcu, rcu_free_flow_callback);
 	else
@@ -138,93 +184,161 @@ static void free_buckets(struct flex_array *buckets)
 	flex_array_free(buckets);
 }
 
-static void __flow_tbl_destroy(struct flow_table *table)
+
+static void __table_instance_destroy(struct table_instance *ti)
 {
-	int i;
+	free_buckets(ti->buckets);
+	kfree(ti);
+}
 
-	if (table->keep_flows)
-		goto skip_flows;
+static struct table_instance *table_instance_alloc(int new_size)
+{
+	struct table_instance *ti = kmalloc(sizeof(*ti), GFP_KERNEL);
 
-	for (i = 0; i < table->n_buckets; i++) {
-		struct sw_flow *flow;
-		struct hlist_head *head = flex_array_get(table->buckets, i);
-		struct hlist_node *n;
-		int ver = table->node_ver;
+	if (!ti)
+		return NULL;
 
-		hlist_for_each_entry_safe(flow, n, head, hash_node[ver]) {
-			hlist_del(&flow->hash_node[ver]);
-			ovs_flow_free(flow, false);
-		}
+	ti->buckets = alloc_buckets(new_size);
+
+	if (!ti->buckets) {
+		kfree(ti);
+		return NULL;
 	}
+	ti->n_buckets = new_size;
+	ti->node_ver = 0;
+	ti->keep_flows = false;
+	get_random_bytes(&ti->hash_seed, sizeof(u32));
 
-	BUG_ON(!list_empty(table->mask_list));
-	kfree(table->mask_list);
+	return ti;
+}
 
-skip_flows:
-	free_buckets(table->buckets);
-	kfree(table);
+static void mask_array_rcu_cb(struct rcu_head *rcu)
+{
+	struct mask_array *ma = container_of(rcu, struct mask_array, rcu);
+
+	kfree(ma);
 }
 
-static struct flow_table *__flow_tbl_alloc(int new_size)
+static struct mask_array *tbl_mask_array_alloc(int size)
 {
-	struct flow_table *table = kmalloc(sizeof(*table), GFP_KERNEL);
+	struct mask_array *new;
 
-	if (!table)
+	new = kzalloc(sizeof(struct mask_array) +
+		      sizeof(struct sw_flow_mask *) * size, GFP_KERNEL);
+	if (!new)
 		return NULL;
 
-	table->buckets = alloc_buckets(new_size);
+	new->count = 0;
+	new->max = size;
 
-	if (!table->buckets) {
-		kfree(table);
-		return NULL;
+	return new;
+}
+
+static int tbl_mask_array_realloc(struct flow_table *tbl, int size)
+{
+	struct mask_array *old;
+	struct mask_array *new;
+
+	new = tbl_mask_array_alloc(size);
+	if (!new)
+		return -ENOMEM;
+
+	old = ovsl_dereference(tbl->mask_array);
+	if (old) {
+		int i;
+
+		for (i = 0; i < old->max; i++) {
+			if (old->masks[i])
+				new->masks[new->count++] = old->masks[i];
+		}
 	}
-	table->n_buckets = new_size;
-	table->count = 0;
-	table->node_ver = 0;
-	table->keep_flows = false;
-	get_random_bytes(&table->hash_seed, sizeof(u32));
-	table->mask_list = NULL;
+	rcu_assign_pointer(tbl->mask_array, new);
+
+	if (old)
+		call_rcu(&old->rcu, mask_array_rcu_cb);
 
-	return table;
+	return 0;
 }
 
-struct flow_table *ovs_flow_tbl_alloc(int new_size)
+int ovs_flow_tbl_init(struct flow_table *table)
 {
-	struct flow_table *table = __flow_tbl_alloc(new_size);
+	struct table_instance *ti;
+	struct mask_array *ma;
 
-	if (!table)
-		return NULL;
+	table->mask_cache = __alloc_percpu(sizeof(struct mask_cache_entry) *
+					  MC_HASH_ENTRIES, __alignof__(struct mask_cache_entry));
+	if (!table->mask_cache)
+		return -ENOMEM;
 
-	table->mask_list = kmalloc(sizeof(struct list_head), GFP_KERNEL);
-	if (!table->mask_list) {
-		table->keep_flows = true;
-		__flow_tbl_destroy(table);
-		return NULL;
-	}
-	INIT_LIST_HEAD(table->mask_list);
+	ma = tbl_mask_array_alloc(MASK_ARRAY_SIZE_MIN);
+	if (!ma)
+		goto free_mask_cache;
+
+	ti = table_instance_alloc(TBL_MIN_BUCKETS);
+	if (!ti)
+		goto free_mask_array;
+
+	rcu_assign_pointer(table->ti, ti);
+	rcu_assign_pointer(table->mask_array, ma);
+	table->last_rehash = jiffies;
+	table->count = 0;
+	return 0;
 
-	return table;
+free_mask_array:
+	kfree((struct mask_array __force *)table->mask_array);
+free_mask_cache:
+	free_percpu(table->mask_cache);
+	return -ENOMEM;
 }
 
 static void flow_tbl_destroy_rcu_cb(struct rcu_head *rcu)
 {
-	struct flow_table *table = container_of(rcu, struct flow_table, rcu);
+	struct table_instance *ti = container_of(rcu, struct table_instance, rcu);
 
-	__flow_tbl_destroy(table);
+	__table_instance_destroy(ti);
 }
 
-void ovs_flow_tbl_destroy(struct flow_table *table, bool deferred)
+static void table_instance_destroy(struct table_instance *ti, bool deferred)
 {
-	if (!table)
+	int i;
+
+	if (!ti)
 		return;
 
+	if (ti->keep_flows)
+		goto skip_flows;
+
+	for (i = 0; i < ti->n_buckets; i++) {
+		struct sw_flow *flow;
+		struct hlist_head *head = flex_array_get(ti->buckets, i);
+		struct hlist_node *n;
+		int ver = ti->node_ver;
+
+		hlist_for_each_entry_safe(flow, n, head, hash_node[ver]) {
+			hlist_del_rcu(&flow->hash_node[ver]);
+			ovs_flow_free(flow, deferred);
+		}
+	}
+
+skip_flows:
 	if (deferred)
-		call_rcu(&table->rcu, flow_tbl_destroy_rcu_cb);
+		call_rcu(&ti->rcu, flow_tbl_destroy_rcu_cb);
 	else
-		__flow_tbl_destroy(table);
+		__table_instance_destroy(ti);
+}
+
+/* No need for locking this function is called from RCU callback or
+ * error path. */
+void ovs_flow_tbl_destroy(struct flow_table *table)
+{
+	struct table_instance *ti = (struct table_instance __force *)table->ti;
+
+	free_percpu(table->mask_cache);
+	kfree((struct mask_array __force *)table->mask_array);
+	table_instance_destroy(ti, false);
 }
 
-struct sw_flow *ovs_flow_tbl_dump_next(struct flow_table *table,
+struct sw_flow *ovs_flow_tbl_dump_next(struct table_instance *ti,
 				       u32 *bucket, u32 *last)
 {
 	struct sw_flow *flow;
@@ -232,10 +346,10 @@ struct sw_flow *ovs_flow_tbl_dump_next(struct flow_table *table,
 	int ver;
 	int i;
 
-	ver = table->node_ver;
-	while (*bucket < table->n_buckets) {
+	ver = ti->node_ver;
+	while (*bucket < ti->n_buckets) {
 		i = 0;
-		head = flex_array_get(table->buckets, *bucket);
+		head = flex_array_get(ti->buckets, *bucket);
 		hlist_for_each_entry_rcu(flow, head, hash_node[ver]) {
 			if (i < *last) {
 				i++;
@@ -251,25 +365,23 @@ struct sw_flow *ovs_flow_tbl_dump_next(struct flow_table *table,
 	return NULL;
 }
 
-static struct hlist_head *find_bucket(struct flow_table *table, u32 hash)
+static struct hlist_head *find_bucket(struct table_instance *ti, u32 hash)
 {
-	hash = jhash_1word(hash, table->hash_seed);
-	return flex_array_get(table->buckets,
-				(hash & (table->n_buckets - 1)));
+	hash = jhash_1word(hash, ti->hash_seed);
+	return flex_array_get(ti->buckets,
+				(hash & (ti->n_buckets - 1)));
 }
 
-static void __tbl_insert(struct flow_table *table, struct sw_flow *flow)
+static void table_instance_insert(struct table_instance *ti, struct sw_flow *flow)
 {
 	struct hlist_head *head;
 
-	head = find_bucket(table, flow->hash);
-	hlist_add_head_rcu(&flow->hash_node[table->node_ver], head);
-
-	table->count++;
+	head = find_bucket(ti, flow->hash);
+	hlist_add_head_rcu(&flow->hash_node[ti->node_ver], head);
 }
 
-static void flow_table_copy_flows(struct flow_table *old,
-				  struct flow_table *new)
+static void flow_table_copy_flows(struct table_instance *old,
+				  struct table_instance *new)
 {
 	int old_ver;
 	int i;
@@ -285,47 +397,54 @@ static void flow_table_copy_flows(struct flow_table *old,
 		head = flex_array_get(old->buckets, i);
 
 		hlist_for_each_entry(flow, head, hash_node[old_ver])
-			__tbl_insert(new, flow);
+			table_instance_insert(new, flow);
 	}
 
-	new->mask_list = old->mask_list;
 	old->keep_flows = true;
 }
 
-static struct flow_table *__flow_tbl_rehash(struct flow_table *table,
+static struct table_instance *table_instance_rehash(struct table_instance *ti,
 					    int n_buckets)
 {
-	struct flow_table *new_table;
+	struct table_instance *new_ti;
 
-	new_table = __flow_tbl_alloc(n_buckets);
-	if (!new_table)
-		return ERR_PTR(-ENOMEM);
+	new_ti = table_instance_alloc(n_buckets);
+	if (!new_ti)
+		return NULL;
 
-	flow_table_copy_flows(table, new_table);
+	flow_table_copy_flows(ti, new_ti);
 
-	return new_table;
+	return new_ti;
 }
 
-struct flow_table *ovs_flow_tbl_rehash(struct flow_table *table)
+int ovs_flow_tbl_flush(struct flow_table *flow_table)
 {
-	return __flow_tbl_rehash(table, table->n_buckets);
-}
+	struct table_instance *old_ti;
+	struct table_instance *new_ti;
 
-struct flow_table *ovs_flow_tbl_expand(struct flow_table *table)
-{
-	return __flow_tbl_rehash(table, table->n_buckets * 2);
+	old_ti = ovsl_dereference(flow_table->ti);
+	new_ti = table_instance_alloc(TBL_MIN_BUCKETS);
+	if (!new_ti)
+		return -ENOMEM;
+
+	rcu_assign_pointer(flow_table->ti, new_ti);
+	flow_table->last_rehash = jiffies;
+	flow_table->count = 0;
+
+	table_instance_destroy(old_ti, true);
+	return 0;
 }
 
 static u32 flow_hash(const struct sw_flow_key *key, int key_start,
 		     int key_end)
 {
-	u32 *hash_key = (u32 *)((u8 *)key + key_start);
+	const u32 *hash_key = (const u32 *)((const u8 *)key + key_start);
 	int hash_u32s = (key_end - key_start) >> 2;
 
 	/* Make sure number of hash bytes are multiple of u32. */
 	BUILD_BUG_ON(sizeof(long) % sizeof(u32));
 
-	return jhash2(hash_key, hash_u32s, 0);
+	return arch_fast_hash2(hash_key, hash_u32s, 0);
 }
 
 static int flow_key_start(const struct sw_flow_key *key)
@@ -341,8 +460,8 @@ static bool cmp_key(const struct sw_flow_key *key1,
 		    const struct sw_flow_key *key2,
 		    int key_start, int key_end)
 {
-	const long *cp1 = (long *)((u8 *)key1 + key_start);
-	const long *cp2 = (long *)((u8 *)key2 + key_start);
+	const long *cp1 = (const long *)((const u8 *)key1 + key_start);
+	const long *cp2 = (const long *)((const u8 *)key2 + key_start);
 	long diffs = 0;
 	int i;
 
@@ -369,9 +488,10 @@ bool ovs_flow_cmp_unmasked_key(const struct sw_flow *flow,
 	return cmp_key(&flow->unmasked_key, key, key_start, key_end);
 }
 
-static struct sw_flow *masked_flow_lookup(struct flow_table *table,
+static struct sw_flow *masked_flow_lookup(struct table_instance *ti,
 					  const struct sw_flow_key *unmasked,
-					  struct sw_flow_mask *mask)
+					  struct sw_flow_mask *mask,
+					  u32 *n_mask_hit)
 {
 	struct sw_flow *flow;
 	struct hlist_head *head;
@@ -382,9 +502,10 @@ static struct sw_flow *masked_flow_lookup(struct flow_table *table,
 
 	ovs_flow_mask_key(&masked_key, unmasked, mask);
 	hash = flow_hash(&masked_key, key_start, key_end);
-	head = find_bucket(table, hash);
-	hlist_for_each_entry_rcu(flow, head, hash_node[table->node_ver]) {
-		if (flow->mask == mask &&
+	head = find_bucket(ti, hash);
+	(*n_mask_hit)++;
+	hlist_for_each_entry_rcu(flow, head, hash_node[ti->node_ver]) {
+		if (flow->mask == mask && flow->hash == hash &&
 		    flow_cmp_masked_key(flow, &masked_key,
 					  key_start, key_end))
 			return flow;
@@ -392,109 +513,290 @@ static struct sw_flow *masked_flow_lookup(struct flow_table *table,
 	return NULL;
 }
 
-struct sw_flow *ovs_flow_tbl_lookup(struct flow_table *tbl,
-				    const struct sw_flow_key *key)
+
+static struct sw_flow *flow_lookup(struct flow_table *tbl,
+				   struct table_instance *ti,
+				   struct mask_array *ma,
+				   const struct sw_flow_key *key,
+				   u32 *n_mask_hit,
+				   u32 *index)
 {
-	struct sw_flow *flow = NULL;
-	struct sw_flow_mask *mask;
+	struct sw_flow *flow;
+	int i;
 
-	list_for_each_entry_rcu(mask, tbl->mask_list, list) {
-		flow = masked_flow_lookup(tbl, key, mask);
-		if (flow)  /* Found */
+	for (i = 0; i < ma->max; i++) {
+		struct sw_flow_mask *mask;
+
+		mask = rcu_dereference_ovsl(ma->masks[i]);
+		if (mask) {
+			flow = masked_flow_lookup(ti, key, mask, n_mask_hit);
+			if (flow) { /* Found */
+				*index = i;
+				return flow;
+			}
+		}
+	}
+
+	return NULL;
+}
+
+/*
+ * mask_cache maps flow to probable mask. This cache is not tightly
+ * coupled cache, It means updates to  mask list can result in inconsistent
+ * cache entry in mask cache.
+ * This is per cpu cache and is divided in MC_HASH_SEGS segments.
+ * In case of a hash collision the entry is hashed in next segment.
+ * */
+struct sw_flow *ovs_flow_tbl_lookup_stats(struct flow_table *tbl,
+					  const struct sw_flow_key *key,
+					  u32 skb_hash,
+					  u32 *n_mask_hit)
+{
+	struct mask_array *ma = rcu_dereference_ovsl(tbl->mask_array);
+	struct table_instance *ti = rcu_dereference_ovsl(tbl->ti);
+	struct mask_cache_entry  *entries, *ce, *del;
+	struct sw_flow *flow;
+	u32 hash = skb_hash;
+	int seg;
+
+	*n_mask_hit = 0;
+	if (unlikely(!skb_hash)) {
+		u32 __always_unused mask_index;
+
+		return flow_lookup(tbl, ti, ma, key, n_mask_hit, &mask_index);
+	}
+
+	del = NULL;
+	entries = this_cpu_ptr(tbl->mask_cache);
+
+	for (seg = 0; seg < MC_HASH_SEGS; seg++) {
+		int index;
+
+		index = hash & (MC_HASH_ENTRIES - 1);
+		ce = &entries[index];
+
+		if (ce->skb_hash == skb_hash) {
+			struct sw_flow_mask *mask;
+			struct sw_flow *flow;
+
+			mask = rcu_dereference_ovsl(ma->masks[ce->mask_index]);
+			if (mask) {
+				flow = masked_flow_lookup(ti, key, mask,
+							  n_mask_hit);
+				if (flow)  /* Found */
+					return flow;
+
+			}
+			del = ce;
 			break;
+		}
+
+		if (!del || (del->skb_hash && !ce->skb_hash) ||
+		    (rcu_dereference_ovsl(ma->masks[del->mask_index]) &&
+		    !rcu_dereference_ovsl(ma->masks[ce->mask_index]))) {
+			del = ce;
+		}
+
+		hash >>= MC_HASH_SHIFT;
 	}
 
+	flow = flow_lookup(tbl, ti, ma, key, n_mask_hit, &del->mask_index);
+	if (flow)
+		del->skb_hash = skb_hash;
+
 	return flow;
 }
 
-void ovs_flow_tbl_insert(struct flow_table *table, struct sw_flow *flow)
+struct sw_flow *ovs_flow_tbl_lookup(struct flow_table *tbl,
+				    const struct sw_flow_key *key)
 {
-	flow->hash = flow_hash(&flow->key, flow->mask->range.start,
-			flow->mask->range.end);
-	__tbl_insert(table, flow);
-}
+	struct table_instance *ti = rcu_dereference_ovsl(tbl->ti);
+	struct mask_array *ma = rcu_dereference_ovsl(tbl->mask_array);
+	u32 __always_unused n_mask_hit;
+	u32 __always_unused index;
 
-void ovs_flow_tbl_remove(struct flow_table *table, struct sw_flow *flow)
-{
-	BUG_ON(table->count == 0);
-	hlist_del_rcu(&flow->hash_node[table->node_ver]);
-	table->count--;
+	n_mask_hit = 0;
+	return flow_lookup(tbl, ti, ma, key, &n_mask_hit, &index);
 }
 
-struct sw_flow_mask *ovs_sw_flow_mask_alloc(void)
+int ovs_flow_tbl_num_masks(const struct flow_table *table)
 {
-	struct sw_flow_mask *mask;
+	struct mask_array *ma;
 
-	mask = kmalloc(sizeof(*mask), GFP_KERNEL);
-	if (mask)
-		mask->ref_count = 0;
+	ma = rcu_dereference_ovsl(table->mask_array);
+	return ma->count;
+}
 
-	return mask;
+static struct table_instance *table_instance_expand(struct table_instance *ti)
+{
+	return table_instance_rehash(ti, ti->n_buckets * 2);
 }
 
-void ovs_sw_flow_mask_add_ref(struct sw_flow_mask *mask)
+/* Remove 'mask' from the mask list, if it is not needed any more. */
+static void flow_mask_remove(struct flow_table *tbl, struct sw_flow_mask *mask)
 {
-	mask->ref_count++;
+	if (mask) {
+		/* ovs-lock is required to protect mask-refcount and
+		 * mask list.
+		 */
+		ASSERT_OVSL();
+		BUG_ON(!mask->ref_count);
+		mask->ref_count--;
+
+		if (!mask->ref_count) {
+			struct mask_array *ma;
+			int i;
+
+			ma = ovsl_dereference(tbl->mask_array);
+			for (i = 0; i < ma->max; i++) {
+				if (mask == ovsl_dereference(ma->masks[i])) {
+					RCU_INIT_POINTER(ma->masks[i], NULL);
+					ma->count--;
+					goto free;
+				}
+			}
+			BUG();
+free:
+			call_rcu(&mask->rcu, rcu_free_sw_flow_mask_cb);
+		}
+	}
 }
 
-static void rcu_free_sw_flow_mask_cb(struct rcu_head *rcu)
+/* Must be called with OVS mutex held. */
+void ovs_flow_tbl_remove(struct flow_table *table, struct sw_flow *flow)
 {
-	struct sw_flow_mask *mask = container_of(rcu, struct sw_flow_mask, rcu);
+	struct table_instance *ti = ovsl_dereference(table->ti);
 
-	kfree(mask);
+	BUG_ON(table->count == 0);
+	hlist_del_rcu(&flow->hash_node[ti->node_ver]);
+	table->count--;
+
+	/* RCU delete the mask. 'flow->mask' is not NULLed, as it should be
+	 * accessible as long as the RCU read lock is held. */
+	flow_mask_remove(table, flow->mask);
 }
 
-void ovs_sw_flow_mask_del_ref(struct sw_flow_mask *mask, bool deferred)
+static struct sw_flow_mask *mask_alloc(void)
 {
-	if (!mask)
-		return;
+	struct sw_flow_mask *mask;
 
-	BUG_ON(!mask->ref_count);
-	mask->ref_count--;
+	mask = kmalloc(sizeof(*mask), GFP_KERNEL);
+	if (mask)
+		mask->ref_count = 1;
 
-	if (!mask->ref_count) {
-		list_del_rcu(&mask->list);
-		if (deferred)
-			call_rcu(&mask->rcu, rcu_free_sw_flow_mask_cb);
-		else
-			kfree(mask);
-	}
+	return mask;
 }
 
 static bool mask_equal(const struct sw_flow_mask *a,
 		       const struct sw_flow_mask *b)
 {
-	u8 *a_ = (u8 *)&a->key + a->range.start;
-	u8 *b_ = (u8 *)&b->key + b->range.start;
+	const u8 *a_ = (const u8 *)&a->key + a->range.start;
+	const u8 *b_ = (const u8 *)&b->key + b->range.start;
 
 	return  (a->range.end == b->range.end)
 		&& (a->range.start == b->range.start)
 		&& (memcmp(a_, b_, range_n_bytes(&a->range)) == 0);
 }
 
-struct sw_flow_mask *ovs_sw_flow_mask_find(const struct flow_table *tbl,
+static struct sw_flow_mask *flow_mask_find(const struct flow_table *tbl,
 					   const struct sw_flow_mask *mask)
 {
-	struct list_head *ml;
+	struct mask_array *ma;
+	int i;
 
-	list_for_each(ml, tbl->mask_list) {
-		struct sw_flow_mask *m;
-		m = container_of(ml, struct sw_flow_mask, list);
-		if (mask_equal(mask, m))
-			return m;
+	ma = ovsl_dereference(tbl->mask_array);
+	for (i = 0; i < ma->max; i++) {
+		struct sw_flow_mask *t;
+
+		t = ovsl_dereference(ma->masks[i]);
+		if (t && mask_equal(mask, t))
+			return t;
 	}
 
 	return NULL;
 }
 
-/**
- * add a new mask into the mask list.
- * The caller needs to make sure that 'mask' is not the same
- * as any masks that are already on the list.
- */
-void ovs_sw_flow_mask_insert(struct flow_table *tbl, struct sw_flow_mask *mask)
+/* Add 'mask' into the mask list, if it is not already there. */
+static int flow_mask_insert(struct flow_table *tbl, struct sw_flow *flow,
+			    struct sw_flow_mask *new)
 {
-	list_add_rcu(&mask->list, tbl->mask_list);
+	struct sw_flow_mask *mask;
+
+	mask = flow_mask_find(tbl, new);
+	if (!mask) {
+		struct mask_array *ma;
+		int i;
+
+		/* Allocate a new mask if none exsits. */
+		mask = mask_alloc();
+		if (!mask)
+			return -ENOMEM;
+
+		mask->key = new->key;
+		mask->range = new->range;
+
+		/* Add mask to mask-list. */
+		ma = ovsl_dereference(tbl->mask_array);
+		if (ma->count >= ma->max) {
+			int err;
+
+			err = tbl_mask_array_realloc(tbl, ma->max +
+							MASK_ARRAY_SIZE_MIN);
+			if (err) {
+				kfree(mask);
+				return err;
+			}
+			ma = ovsl_dereference(tbl->mask_array);
+		}
+		for (i = 0; i < ma->max; i++) {
+			const struct sw_flow_mask *t;
+
+			t = ovsl_dereference(ma->masks[i]);
+			if (!t) {
+				rcu_assign_pointer(ma->masks[i], mask);
+				ma->count++;
+				break;
+			}
+		}
+	} else {
+		BUG_ON(!mask->ref_count);
+		mask->ref_count++;
+	}
+
+	flow->mask = mask;
+	return 0;
+}
+
+/* Must be called with OVS mutex held. */
+int ovs_flow_tbl_insert(struct flow_table *table, struct sw_flow *flow,
+			struct sw_flow_mask *mask)
+{
+	struct table_instance *new_ti = NULL;
+	struct table_instance *ti;
+	int err;
+
+	err = flow_mask_insert(table, flow, mask);
+	if (err)
+		return err;
+
+	flow->hash = flow_hash(&flow->key, flow->mask->range.start,
+			flow->mask->range.end);
+	ti = ovsl_dereference(table->ti);
+	table_instance_insert(ti, flow);
+	table->count++;
+
+	/* Expand table, if necessary, to make room. */
+	if (table->count > ti->n_buckets)
+		new_ti = table_instance_expand(ti);
+	else if (time_after(jiffies, table->last_rehash + REHASH_INTERVAL))
+		new_ti = table_instance_rehash(ti, ti->n_buckets);
+
+	if (new_ti) {
+		rcu_assign_pointer(table->ti, new_ti);
+		table_instance_destroy(ti, true);
+		table->last_rehash = jiffies;
+	}
+	return 0;
 }
 
 /* Initializes the flow module.
@@ -504,16 +806,28 @@ int ovs_flow_init(void)
 	BUILD_BUG_ON(__alignof__(struct sw_flow_key) % __alignof__(long));
 	BUILD_BUG_ON(sizeof(struct sw_flow_key) % sizeof(long));
 
-	flow_cache = kmem_cache_create("sw_flow", sizeof(struct sw_flow), 0,
-					0, NULL);
+	flow_cache = kmem_cache_create("sw_flow", sizeof(struct sw_flow)
+				       + (num_possible_nodes()
+					  * sizeof(struct flow_stats *)),
+				       0, 0, NULL);
 	if (flow_cache == NULL)
 		return -ENOMEM;
 
+	flow_stats_cache
+		= kmem_cache_create("sw_flow_stats", sizeof(struct flow_stats),
+				    0, SLAB_HWCACHE_ALIGN, NULL);
+	if (flow_stats_cache == NULL) {
+		kmem_cache_destroy(flow_cache);
+		flow_cache = NULL;
+		return -ENOMEM;
+	}
+
 	return 0;
 }
 
 /* Uninitializes the flow module. */
 void ovs_flow_exit(void)
 {
+	kmem_cache_destroy(flow_stats_cache);
 	kmem_cache_destroy(flow_cache);
 }