soreuseport: TCP/IPv4 implementation

Allow multiple listener sockets to bind to the same port.

Motivation for soresuseport would be something like a web server
binding to port 80 running with multiple threads, where each thread
might have it's own listener socket.  This could be done as an
alternative to other models: 1) have one listener thread which
dispatches completed connections to workers. 2) accept on a single
listener socket from multiple threads.  In case #1 the listener thread
can easily become the bottleneck with high connection turn-over rate.
In case #2, the proportion of connections accepted per thread tends
to be uneven under high connection load (assuming simple event loop:
while (1) { accept(); process() }, wakeup does not promote fairness
among the sockets.  We have seen the  disproportion to be as high
as 3:1 ratio between thread accepting most connections and the one
accepting the fewest.  With so_reusport the distribution is
uniform.

Signed-off-by: Tom Herbert <therbert@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

1 files changed, 22 insertions, 6 deletions

diff --git a/net/ipv4/inet_hashtables.c b/net/ipv4/inet_hashtables.c
index fa3ae814871..0ce0595d986 100644
--- a/net/ipv4/inet_hashtables.c
+++ b/net/ipv4/inet_hashtables.c
@@ -39,6 +39,7 @@ struct inet_bind_bucket *inet_bind_bucket_create(struct kmem_cache *cachep,
 		write_pnet(&tb->ib_net, hold_net(net));
 		tb->port      = snum;
 		tb->fastreuse = 0;
+		tb->fastreuseport = 0;
 		tb->num_owners = 0;
 		INIT_HLIST_HEAD(&tb->owners);
 		hlist_add_head(&tb->node, &head->chain);
@@ -151,16 +152,16 @@ static inline int compute_score(struct sock *sk, struct net *net,
 	if (net_eq(sock_net(sk), net) && inet->inet_num == hnum &&
 			!ipv6_only_sock(sk)) {
 		__be32 rcv_saddr = inet->inet_rcv_saddr;
-		score = sk->sk_family == PF_INET ? 1 : 0;
+		score = sk->sk_family == PF_INET ? 2 : 1;
 		if (rcv_saddr) {
 			if (rcv_saddr != daddr)
 				return -1;
-			score += 2;
+			score += 4;
 		}
 		if (sk->sk_bound_dev_if) {
 			if (sk->sk_bound_dev_if != dif)
 				return -1;
-			score += 2;
+			score += 4;
 		}
 	}
 	return score;
@@ -176,6 +177,7 @@ static inline int compute_score(struct sock *sk, struct net *net,
 
 struct sock *__inet_lookup_listener(struct net *net,
 				    struct inet_hashinfo *hashinfo,
+				    const __be32 saddr, __be16 sport,
 				    const __be32 daddr, const unsigned short hnum,
 				    const int dif)
 {
@@ -183,17 +185,29 @@ struct sock *__inet_lookup_listener(struct net *net,
 	struct hlist_nulls_node *node;
 	unsigned int hash = inet_lhashfn(net, hnum);
 	struct inet_listen_hashbucket *ilb = &hashinfo->listening_hash[hash];
-	int score, hiscore;
+	int score, hiscore, matches = 0, reuseport = 0;
+	u32 phash = 0;
 
 	rcu_read_lock();
 begin:
 	result = NULL;
-	hiscore = -1;
+	hiscore = 0;
 	sk_nulls_for_each_rcu(sk, node, &ilb->head) {
 		score = compute_score(sk, net, hnum, daddr, dif);
 		if (score > hiscore) {
 			result = sk;
 			hiscore = score;
+			reuseport = sk->sk_reuseport;
+			if (reuseport) {
+				phash = inet_ehashfn(net, daddr, hnum,
+						     saddr, sport);
+				matches = 1;
+			}
+		} else if (score == hiscore && reuseport) {
+			matches++;
+			if (((u64)phash * matches) >> 32 == 0)
+				result = sk;
+			phash = next_pseudo_random32(phash);
 		}
 	}
 	/*
@@ -501,7 +515,8 @@ int __inet_hash_connect(struct inet_timewait_death_row *death_row,
 			inet_bind_bucket_for_each(tb, node, &head->chain) {
 				if (net_eq(ib_net(tb), net) &&
 				    tb->port == port) {
-					if (tb->fastreuse >= 0)
+					if (tb->fastreuse >= 0 ||
+					    tb->fastreuseport >= 0)
 						goto next_port;
 					WARN_ON(hlist_empty(&tb->owners));
 					if (!check_established(death_row, sk,
@@ -518,6 +533,7 @@ int __inet_hash_connect(struct inet_timewait_death_row *death_row,
 				break;
 			}
 			tb->fastreuse = -1;
+			tb->fastreuseport = -1;
 			goto ok;
 
 		next_port: