netlink: Convert netlink_lookup() to use RCU protected hash table

Heavy Netlink users such as Open vSwitch spend a considerable amount of
time in netlink_lookup() due to the read-lock on nl_table_lock. Use of
RCU relieves the lock contention.

Makes use of the new resizable hash table to avoid locking on the
lookup.

The hash table will grow if entries exceeds 75% of table size up to a
total table size of 64K. It will automatically shrink if usage falls
below 30%.

Also splits nl_table_lock into a separate mutex to protect hash table
mutations and allow synchronize_rcu() to sleep while waiting for readers
during expansion and shrinking.

Before:
   9.16%  kpktgend_0  [openvswitch]      [k] masked_flow_lookup
   6.42%  kpktgend_0  [pktgen]           [k] mod_cur_headers
   6.26%  kpktgend_0  [pktgen]           [k] pktgen_thread_worker
   6.23%  kpktgend_0  [kernel.kallsyms]  [k] memset
   4.79%  kpktgend_0  [kernel.kallsyms]  [k] netlink_lookup
   4.37%  kpktgend_0  [kernel.kallsyms]  [k] memcpy
   3.60%  kpktgend_0  [openvswitch]      [k] ovs_flow_extract
   2.69%  kpktgend_0  [kernel.kallsyms]  [k] jhash2

After:
  15.26%  kpktgend_0  [openvswitch]      [k] masked_flow_lookup
   8.12%  kpktgend_0  [pktgen]           [k] pktgen_thread_worker
   7.92%  kpktgend_0  [pktgen]           [k] mod_cur_headers
   5.11%  kpktgend_0  [kernel.kallsyms]  [k] memset
   4.11%  kpktgend_0  [openvswitch]      [k] ovs_flow_extract
   4.06%  kpktgend_0  [kernel.kallsyms]  [k] _raw_spin_lock
   3.90%  kpktgend_0  [kernel.kallsyms]  [k] jhash2
   [...]
   0.67%  kpktgend_0  [kernel.kallsyms]  [k] netlink_lookup
Signed-off-by: Thomas Graf <tgraf@suug.ch>
Reviewed-by: Nikolay Aleksandrov <nikolay@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net/netlink/af_netlink.c

@@ -58,7 +58,9 @@
 #include <linux/mutex.h>
 #include <linux/vmalloc.h>
 #include <linux/if_arp.h>
+#include <linux/rhashtable.h>
 #include <asm/cacheflush.h>
+#include <linux/hash.h>
 
 #include <net/net_namespace.h>
 #include <net/sock.h>
@@ -100,6 +102,18 @@ static atomic_t nl_table_users = ATOMIC_INIT(0);
 
 #define nl_deref_protected(X) rcu_dereference_protected(X, lockdep_is_held(&nl_table_lock));
 
+/* Protects netlink socket hash table mutations */
+DEFINE_MUTEX(nl_sk_hash_lock);
+
+static int lockdep_nl_sk_hash_is_held(void)
+{
+#ifdef CONFIG_LOCKDEP
+	return (debug_locks) ? lockdep_is_held(&nl_sk_hash_lock) : 1;
+#else
+	return 1;
+#endif
+}
+
 static ATOMIC_NOTIFIER_HEAD(netlink_chain);
 
 static DEFINE_SPINLOCK(netlink_tap_lock);
@@ -110,11 +124,6 @@ static inline u32 netlink_group_mask(u32 group)
 	return group ? 1 << (group - 1) : 0;
 }
 
-static inline struct hlist_head *nl_portid_hashfn(struct nl_portid_hash *hash, u32 portid)
-{
-	return &hash->table[jhash_1word(portid, hash->rnd) & hash->mask];
-}
-
 int netlink_add_tap(struct netlink_tap *nt)
 {
 	if (unlikely(nt->dev->type != ARPHRD_NETLINK))
@@ -983,105 +992,48 @@ netlink_unlock_table(void)
 		wake_up(&nl_table_wait);
 }
 
-static bool netlink_compare(struct net *net, struct sock *sk)
-{
-	return net_eq(sock_net(sk), net);
-}
-
-static struct sock *netlink_lookup(struct net *net, int protocol, u32 portid)
+struct netlink_compare_arg
 {
-	struct netlink_table *table = &nl_table[protocol];
-	struct nl_portid_hash *hash = &table->hash;
-	struct hlist_head *head;
-	struct sock *sk;
-
-	read_lock(&nl_table_lock);
-	head = nl_portid_hashfn(hash, portid);
-	sk_for_each(sk, head) {
-		if (table->compare(net, sk) &&
-		    (nlk_sk(sk)->portid == portid)) {
-			sock_hold(sk);
-			goto found;
-		}
-	}
-	sk = NULL;
-found:
-	read_unlock(&nl_table_lock);
-	return sk;
-}
+	struct net *net;
+	u32 portid;
+};
 
-static struct hlist_head *nl_portid_hash_zalloc(size_t size)
+static bool netlink_compare(void *ptr, void *arg)
 {
-	if (size <= PAGE_SIZE)
-		return kzalloc(size, GFP_ATOMIC);
-	else
-		return (struct hlist_head *)
-			__get_free_pages(GFP_ATOMIC | __GFP_ZERO,
-					 get_order(size));
-}
+	struct netlink_compare_arg *x = arg;
+	struct sock *sk = ptr;
 
-static void nl_portid_hash_free(struct hlist_head *table, size_t size)
-{
-	if (size <= PAGE_SIZE)
-		kfree(table);
-	else
-		free_pages((unsigned long)table, get_order(size));
+	return nlk_sk(sk)->portid == x->portid &&
+	       net_eq(sock_net(sk), x->net);
 }
 
-static int nl_portid_hash_rehash(struct nl_portid_hash *hash, int grow)
+static struct sock *__netlink_lookup(struct netlink_table *table, u32 portid,
+				     struct net *net)
 {
-	unsigned int omask, mask, shift;
-	size_t osize, size;
-	struct hlist_head *otable, *table;
-	int i;
-
-	omask = mask = hash->mask;
-	osize = size = (mask + 1) * sizeof(*table);
-	shift = hash->shift;
-
-	if (grow) {
-		if (++shift > hash->max_shift)
-			return 0;
-		mask = mask * 2 + 1;
-		size *= 2;
-	}
-
-	table = nl_portid_hash_zalloc(size);
-	if (!table)
-		return 0;
-
-	otable = hash->table;
-	hash->table = table;
-	hash->mask = mask;
-	hash->shift = shift;
-	get_random_bytes(&hash->rnd, sizeof(hash->rnd));
-
-	for (i = 0; i <= omask; i++) {
-		struct sock *sk;
-		struct hlist_node *tmp;
+	struct netlink_compare_arg arg = {
+		.net = net,
+		.portid = portid,
+	};
+	u32 hash;
 
-		sk_for_each_safe(sk, tmp, &otable[i])
-			__sk_add_node(sk, nl_portid_hashfn(hash, nlk_sk(sk)->portid));
-	}
+	hash = rhashtable_hashfn(&table->hash, &portid, sizeof(portid));
 
-	nl_portid_hash_free(otable, osize);
-	hash->rehash_time = jiffies + 10 * 60 * HZ;
-	return 1;
+	return rhashtable_lookup_compare(&table->hash, hash,
+					 &netlink_compare, &arg);
 }
 
-static inline int nl_portid_hash_dilute(struct nl_portid_hash *hash, int len)
+static struct sock *netlink_lookup(struct net *net, int protocol, u32 portid)
 {
-	int avg = hash->entries >> hash->shift;
-
-	if (unlikely(avg > 1) && nl_portid_hash_rehash(hash, 1))
-		return 1;
+	struct netlink_table *table = &nl_table[protocol];
+	struct sock *sk;
 
-	if (unlikely(len > avg) && time_after(jiffies, hash->rehash_time)) {
-		nl_portid_hash_rehash(hash, 0);
-		return 1;
-	}
+	rcu_read_lock();
+	sk = __netlink_lookup(table, portid, net);
+	if (sk)
+		sock_hold(sk);
+	rcu_read_unlock();
 
-	return 0;
+	return sk;
 }
 
 static const struct proto_ops netlink_ops;
@@ -1113,22 +1065,10 @@ netlink_update_listeners(struct sock *sk)
 static int netlink_insert(struct sock *sk, struct net *net, u32 portid)
 {
 	struct netlink_table *table = &nl_table[sk->sk_protocol];
-	struct nl_portid_hash *hash = &table->hash;
-	struct hlist_head *head;
 	int err = -EADDRINUSE;
-	struct sock *osk;
-	int len;
 
-	netlink_table_grab();
-	head = nl_portid_hashfn(hash, portid);
-	len = 0;
-	sk_for_each(osk, head) {
-		if (table->compare(net, osk) &&
-		    (nlk_sk(osk)->portid == portid))
-			break;
-		len++;
-	}
-	if (osk)
+	mutex_lock(&nl_sk_hash_lock);
+	if (__netlink_lookup(table, portid, net))
 		goto err;
 
 	err = -EBUSY;
@@ -1136,26 +1076,31 @@ static int netlink_insert(struct sock *sk, struct net *net, u32 portid)
 		goto err;
 
 	err = -ENOMEM;
-	if (BITS_PER_LONG > 32 && unlikely(hash->entries >= UINT_MAX))
+	if (BITS_PER_LONG > 32 && unlikely(table->hash.nelems >= UINT_MAX))
 		goto err;
 
-	if (len && nl_portid_hash_dilute(hash, len))
-		head = nl_portid_hashfn(hash, portid);
-	hash->entries++;
 	nlk_sk(sk)->portid = portid;
-	sk_add_node(sk, head);
+	sock_hold(sk);
+	rhashtable_insert(&table->hash, &nlk_sk(sk)->node, GFP_KERNEL);
 	err = 0;
-
 err:
-	netlink_table_ungrab();
+	mutex_unlock(&nl_sk_hash_lock);
 	return err;
 }
 
 static void netlink_remove(struct sock *sk)
 {
+	struct netlink_table *table;
+
+	mutex_lock(&nl_sk_hash_lock);
+	table = &nl_table[sk->sk_protocol];
+	if (rhashtable_remove(&table->hash, &nlk_sk(sk)->node, GFP_KERNEL)) {
+		WARN_ON(atomic_read(&sk->sk_refcnt) == 1);
+		__sock_put(sk);
+	}
+	mutex_unlock(&nl_sk_hash_lock);
+
 	netlink_table_grab();
-	if (sk_del_node_init(sk))
-		nl_table[sk->sk_protocol].hash.entries--;
 	if (nlk_sk(sk)->subscriptions)
 		__sk_del_bind_node(sk);
 	netlink_table_ungrab();
@@ -1311,6 +1256,9 @@ static int netlink_release(struct socket *sock)
 	}
 	netlink_table_ungrab();
 
+	/* Wait for readers to complete */
+	synchronize_net();
+
 	kfree(nlk->groups);
 	nlk->groups = NULL;
 
@@ -1326,30 +1274,22 @@ static int netlink_autobind(struct socket *sock)
 	struct sock *sk = sock->sk;
 	struct net *net = sock_net(sk);
 	struct netlink_table *table = &nl_table[sk->sk_protocol];
-	struct nl_portid_hash *hash = &table->hash;
-	struct hlist_head *head;
-	struct sock *osk;
 	s32 portid = task_tgid_vnr(current);
 	int err;
 	static s32 rover = -4097;
 
 retry:
 	cond_resched();
-	netlink_table_grab();
-	head = nl_portid_hashfn(hash, portid);
-	sk_for_each(osk, head) {
-		if (!table->compare(net, osk))
-			continue;
-		if (nlk_sk(osk)->portid == portid) {
+	rcu_read_lock();
+	if (__netlink_lookup(table, portid, net)) {
 		/* Bind collision, search negative portid values. */
 		portid = rover--;
 		if (rover > -4097)
 			rover = -4097;
-			netlink_table_ungrab();
+		rcu_read_unlock();
 		goto retry;
 	}
-	}
-	netlink_table_ungrab();
+	rcu_read_unlock();
 
 	err = netlink_insert(sk, net, portid);
 	if (err == -EADDRINUSE)
@@ -2953,14 +2893,18 @@ static struct sock *netlink_seq_socket_idx(struct seq_file *seq, loff_t pos)
 {
 	struct nl_seq_iter *iter = seq->private;
 	int i, j;
+	struct netlink_sock *nlk;
 	struct sock *s;
 	loff_t off = 0;
 
 	for (i = 0; i < MAX_LINKS; i++) {
-		struct nl_portid_hash *hash = &nl_table[i].hash;
+		struct rhashtable *ht = &nl_table[i].hash;
+		const struct bucket_table *tbl = rht_dereference(ht->tbl, ht);
+
+		for (j = 0; j < tbl->size; j++) {
+			rht_for_each_entry_rcu(nlk, tbl->buckets[j], node) {
+				s = (struct sock *)nlk;
 
-		for (j = 0; j <= hash->mask; j++) {
-			sk_for_each(s, &hash->table[j]) {
 				if (sock_net(s) != seq_file_net(seq))
 					continue;
 				if (off == pos) {
@@ -2976,15 +2920,14 @@ static struct sock *netlink_seq_socket_idx(struct seq_file *seq, loff_t pos)
 }
 
 static void *netlink_seq_start(struct seq_file *seq, loff_t *pos)
-	__acquires(nl_table_lock)
 {
-	read_lock(&nl_table_lock);
+	rcu_read_lock();
 	return *pos ? netlink_seq_socket_idx(seq, *pos - 1) : SEQ_START_TOKEN;
 }
 
 static void *netlink_seq_next(struct seq_file *seq, void *v, loff_t *pos)
 {
-	struct sock *s;
+	struct netlink_sock *nlk;
 	struct nl_seq_iter *iter;
 	struct net *net;
 	int i, j;
@@ -2996,28 +2939,26 @@ static void *netlink_seq_next(struct seq_file *seq, void *v, loff_t *pos)
 
 	net = seq_file_net(seq);
 	iter = seq->private;
-	s = v;
-	do {
-		s = sk_next(s);
-	} while (s && !nl_table[s->sk_protocol].compare(net, s));
-	if (s)
-		return s;
+	nlk = v;
+
+	rht_for_each_entry_rcu(nlk, nlk->node.next, node)
+		if (net_eq(sock_net((struct sock *)nlk), net))
+			return nlk;
 
 	i = iter->link;
 	j = iter->hash_idx + 1;
 
 	do {
-		struct nl_portid_hash *hash = &nl_table[i].hash;
+		struct rhashtable *ht = &nl_table[i].hash;
+		const struct bucket_table *tbl = rht_dereference(ht->tbl, ht);
 
-		for (; j <= hash->mask; j++) {
-			s = sk_head(&hash->table[j]);
-
-			while (s && !nl_table[s->sk_protocol].compare(net, s))
-				s = sk_next(s);
-			if (s) {
+		for (; j < tbl->size; j++) {
+			rht_for_each_entry_rcu(nlk, tbl->buckets[j], node) {
+				if (net_eq(sock_net((struct sock *)nlk), net)) {
 					iter->link = i;
 					iter->hash_idx = j;
-				return s;
+					return nlk;
+				}
 			}
 		}
 
@@ -3028,9 +2969,8 @@ static void *netlink_seq_next(struct seq_file *seq, void *v, loff_t *pos)
 }
 
 static void netlink_seq_stop(struct seq_file *seq, void *v)
-	__releases(nl_table_lock)
 {
-	read_unlock(&nl_table_lock);
+	rcu_read_unlock();
 }
 
 
@@ -3168,9 +3108,17 @@ static struct pernet_operations __net_initdata netlink_net_ops = {
 static int __init netlink_proto_init(void)
 {
 	int i;
-	unsigned long limit;
-	unsigned int order;
 	int err = proto_register(&netlink_proto, 0);
+	struct rhashtable_params ht_params = {
+		.head_offset = offsetof(struct netlink_sock, node),
+		.key_offset = offsetof(struct netlink_sock, portid),
+		.key_len = sizeof(u32), /* portid */
+		.hashfn = arch_fast_hash,
+		.max_shift = 16, /* 64K */
+		.grow_decision = rht_grow_above_75,
+		.shrink_decision = rht_shrink_below_30,
+		.mutex_is_held = lockdep_nl_sk_hash_is_held,
+	};
 
 	if (err != 0)
 		goto out;
@@ -3181,32 +3129,13 @@ static int __init netlink_proto_init(void)
 	if (!nl_table)
 		goto panic;
 
-	if (totalram_pages >= (128 * 1024))
-		limit = totalram_pages >> (21 - PAGE_SHIFT);
-	else
-		limit = totalram_pages >> (23 - PAGE_SHIFT);
-
-	order = get_bitmask_order(limit) - 1 + PAGE_SHIFT;
-	limit = (1UL << order) / sizeof(struct hlist_head);
-	order = get_bitmask_order(min(limit, (unsigned long)UINT_MAX)) - 1;
-
 	for (i = 0; i < MAX_LINKS; i++) {
-		struct nl_portid_hash *hash = &nl_table[i].hash;
-
-		hash->table = nl_portid_hash_zalloc(1 * sizeof(*hash->table));
-		if (!hash->table) {
-			while (i-- > 0)
-				nl_portid_hash_free(nl_table[i].hash.table,
-						 1 * sizeof(*hash->table));
+		if (rhashtable_init(&nl_table[i].hash, &ht_params) < 0) {
+			while (--i > 0)
+				rhashtable_destroy(&nl_table[i].hash);
 			kfree(nl_table);
 			goto panic;
 		}
-		hash->max_shift = order;
-		hash->shift = 0;
-		hash->mask = 0;
-		hash->rehash_time = jiffies;
-
-		nl_table[i].compare = netlink_compare;
 	}
 
 	INIT_LIST_HEAD(&netlink_tap_all);

net/netlink/af_netlink.h

 #ifndef _AF_NETLINK_H
 #define _AF_NETLINK_H
 
+#include <linux/rhashtable.h>
 #include <net/sock.h>
 
 #define NLGRPSZ(x)	(ALIGN(x, sizeof(unsigned long) * 8) / 8)
@@ -47,6 +48,8 @@ struct netlink_sock {
 	struct netlink_ring	tx_ring;
 	atomic_t		mapped;
 #endif /* CONFIG_NETLINK_MMAP */
+
+	struct rhash_head	node;
 };
 
 static inline struct netlink_sock *nlk_sk(struct sock *sk)
@@ -54,21 +57,8 @@ static inline struct netlink_sock *nlk_sk(struct sock *sk)
 	return container_of(sk, struct netlink_sock, sk);
 }
 
-struct nl_portid_hash {
-	struct hlist_head	*table;
-	unsigned long		rehash_time;
-
-	unsigned int		mask;
-	unsigned int		shift;
-
-	unsigned int		entries;
-	unsigned int		max_shift;
-
-	u32			rnd;
-};
-
 struct netlink_table {
-	struct nl_portid_hash	hash;
+	struct rhashtable	hash;
 	struct hlist_head	mc_list;
 	struct listeners __rcu	*listeners;
 	unsigned int		flags;

net/netlink/diag.c

@@ -4,6 +4,7 @@
 #include <linux/netlink.h>
 #include <linux/sock_diag.h>
 #include <linux/netlink_diag.h>
+#include <linux/rhashtable.h>
 
 #include "af_netlink.h"
 
@@ -101,16 +102,20 @@ static int __netlink_diag_dump(struct sk_buff *skb, struct netlink_callback *cb,
 				int protocol, int s_num)
 {
 	struct netlink_table *tbl = &nl_table[protocol];
-	struct nl_portid_hash *hash = &tbl->hash;
+	struct rhashtable *ht = &tbl->hash;
+	const struct bucket_table *htbl = rht_dereference(ht->tbl, ht);
 	struct net *net = sock_net(skb->sk);
 	struct netlink_diag_req *req;
+	struct netlink_sock *nlsk;
 	struct sock *sk;
 	int ret = 0, num = 0, i;
 
 	req = nlmsg_data(cb->nlh);
 
-	for (i = 0; i <= hash->mask; i++) {
-		sk_for_each(sk, &hash->table[i]) {
+	for (i = 0; i < htbl->size; i++) {
+		rht_for_each_entry(nlsk, htbl->buckets[i], ht, node) {
+			sk = (struct sock *)nlsk;
+
 			if (!net_eq(sock_net(sk), net))
 				continue;
 			if (num < s_num) {