net-tcp: extend tcp_tw_reuse sysctl to enable loopback only optimization

This changes the /proc/sys/net/ipv4/tcp_tw_reuse from a boolean
to an integer.

It now takes the values 0, 1 and 2, where 0 and 1 behave as before,
while 2 enables timewait socket reuse only for sockets that we can
prove are loopback connections:
  ie. bound to 'lo' interface or where one of source or destination
  IPs is 127.0.0.0/8, ::ffff:127.0.0.0/104 or ::1.

This enables quicker reuse of ephemeral ports for loopback connections
- where tcp_tw_reuse is 100% safe from a protocol perspective
(this assumes no artificially induced packet loss on 'lo').

This also makes estblishing many loopback connections *much* faster
(allocating ports out of the first half of the ephemeral port range
is significantly faster, then allocating from the second half)

Without this change in a 32K ephemeral port space my sample program
(it just establishes and closes [::1]:ephemeral -> [::1]:server_port
connections in a tight loop) fails after 32765 connections in 24 seconds.
With it enabled 50000 connections only take 4.7 seconds.

This is particularly problematic for IPv6 where we only have one local
address and cannot play tricks with varying source IP from 127.0.0.0/8
pool.

Signed-off-by: Maciej Żenczykowski <maze@google.com>
Cc: Neal Cardwell <ncardwell@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Cc: Wei Wang <weiwan@google.com>
Change-Id: I0377961749979d0301b7b62871a32a4b34b654e1
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

2 files changed, 36 insertions, 4 deletions

diff --git a/net/ipv4/sysctl_net_ipv4.c b/net/ipv4/sysctl_net_ipv4.c
index d2eed3ddcb0a..d06247ba08b2 100644
--- a/net/ipv4/sysctl_net_ipv4.c
+++ b/net/ipv4/sysctl_net_ipv4.c
@@ -30,6 +30,7 @@
 
 static int zero;
 static int one = 1;
+static int two = 2;
 static int four = 4;
 static int thousand = 1000;
 static int gso_max_segs = GSO_MAX_SEGS;
@@ -845,7 +846,9 @@ static struct ctl_table ipv4_net_table[] = {
 		.data		= &init_net.ipv4.sysctl_tcp_tw_reuse,
 		.maxlen		= sizeof(int),
 		.mode		= 0644,
-		.proc_handler	= proc_dointvec
+		.proc_handler	= proc_dointvec_minmax,
+		.extra1		= &zero,
+		.extra2		= &two,
 	},
 	{
 		.procname	= "tcp_max_tw_buckets",
diff --git a/net/ipv4/tcp_ipv4.c b/net/ipv4/tcp_ipv4.c
index 749b0ef9f405..633963e228bc 100644
--- a/net/ipv4/tcp_ipv4.c
+++ b/net/ipv4/tcp_ipv4.c
@@ -110,8 +110,38 @@ static u32 tcp_v4_init_ts_off(const struct net *net, const struct sk_buff *skb)
 
 int tcp_twsk_unique(struct sock *sk, struct sock *sktw, void *twp)
 {
+	const struct inet_timewait_sock *tw = inet_twsk(sktw);
 	const struct tcp_timewait_sock *tcptw = tcp_twsk(sktw);
 	struct tcp_sock *tp = tcp_sk(sk);
+	int reuse = sock_net(sk)->ipv4.sysctl_tcp_tw_reuse;
+
+	if (reuse == 2) {
+		/* Still does not detect *everything* that goes through
+		 * lo, since we require a loopback src or dst address
+		 * or direct binding to 'lo' interface.
+		 */
+		bool loopback = false;
+		if (tw->tw_bound_dev_if == LOOPBACK_IFINDEX)
+			loopback = true;
+#if IS_ENABLED(CONFIG_IPV6)
+		if (tw->tw_family == AF_INET6) {
+			if (ipv6_addr_loopback(&tw->tw_v6_daddr) ||
+			    (ipv6_addr_v4mapped(&tw->tw_v6_daddr) &&
+			     (tw->tw_v6_daddr.s6_addr[12] == 127)) ||
+			    ipv6_addr_loopback(&tw->tw_v6_rcv_saddr) ||
+			    (ipv6_addr_v4mapped(&tw->tw_v6_rcv_saddr) &&
+			     (tw->tw_v6_rcv_saddr.s6_addr[12] == 127)))
+				loopback = true;
+		} else
+#endif
+		{
+			if (ipv4_is_loopback(tw->tw_daddr) ||
+			    ipv4_is_loopback(tw->tw_rcv_saddr))
+				loopback = true;
+		}
+		if (!loopback)
+			reuse = 0;
+	}
 
 	/* With PAWS, it is safe from the viewpoint
 	   of data integrity. Even without PAWS it is safe provided sequence
@@ -125,8 +155,7 @@ int tcp_twsk_unique(struct sock *sk, struct sock *sktw, void *twp)
 	   and use initial timestamp retrieved from peer table.
 	 */
 	if (tcptw->tw_ts_recent_stamp &&
-	    (!twp || (sock_net(sk)->ipv4.sysctl_tcp_tw_reuse &&
-			     get_seconds() - tcptw->tw_ts_recent_stamp > 1))) {
+	    (!twp || (reuse && get_seconds() - tcptw->tw_ts_recent_stamp > 1))) {
 		tp->write_seq = tcptw->tw_snd_nxt + 65535 + 2;
 		if (tp->write_seq == 0)
 			tp->write_seq = 1;
@@ -2529,7 +2558,7 @@ static int __net_init tcp_sk_init(struct net *net)
 	net->ipv4.sysctl_tcp_orphan_retries = 0;
 	net->ipv4.sysctl_tcp_fin_timeout = TCP_FIN_TIMEOUT;
 	net->ipv4.sysctl_tcp_notsent_lowat = UINT_MAX;
-	net->ipv4.sysctl_tcp_tw_reuse = 0;
+	net->ipv4.sysctl_tcp_tw_reuse = 2;
 
 	cnt = tcp_hashinfo.ehash_mask + 1;
 	net->ipv4.tcp_death_row.sysctl_max_tw_buckets = (cnt + 1) / 2;