tcp: disable TFO blackhole logic by default

Multiple complaints have been raised from the TFO users on the internet
stating that the TFO blackhole logic is too aggressive and gets falsely
triggered too often.
(e.g. https://blog.apnic.net/2021/07/05/tcp-fast-open-not-so-fast/)
Considering that most middleboxes no longer drop TFO packets, we decide
to disable the blackhole logic by setting
/proc/sys/net/ipv4/tcp_fastopen_blackhole_timeout_set to 0 by default.

Fixes: cf1ef3f ("net/tcp_fastopen: Disable active side TFO in certain scenarios")
Signed-off-by: Wei Wang <weiwan@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Acked-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Documentation/networking/ip-sysctl.rst

@@ -826,7 +826,7 @@ tcp_fastopen_blackhole_timeout_sec - INTEGER
 	initial value when the blackhole issue goes away.
 	0 to disable the blackhole detection.
 
-	By default, it is set to 1hr.
+	By default, it is set to 0 (feature is disabled).
 
 tcp_fastopen_key - list of comma separated 32-digit hexadecimal INTEGERs
 	The list consists of a primary key and an optional backup key. The

net/ipv4/tcp_fastopen.c

@@ -507,6 +507,9 @@ void tcp_fastopen_active_disable(struct sock *sk)
 {
 	struct net *net = sock_net(sk);
 
+	if (!sock_net(sk)->ipv4.sysctl_tcp_fastopen_blackhole_timeout)
+		return;
+
 	/* Paired with READ_ONCE() in tcp_fastopen_active_should_disable() */
 	WRITE_ONCE(net->ipv4.tfo_active_disable_stamp, jiffies);
 
@@ -526,10 +529,14 @@ void tcp_fastopen_active_disable(struct sock *sk)
 bool tcp_fastopen_active_should_disable(struct sock *sk)
 {
 	unsigned int tfo_bh_timeout = sock_net(sk)->ipv4.sysctl_tcp_fastopen_blackhole_timeout;
-	int tfo_da_times = atomic_read(&sock_net(sk)->ipv4.tfo_active_disable_times);
 	unsigned long timeout;
+	int tfo_da_times;
 	int multiplier;
 
+	if (!tfo_bh_timeout)
+		return false;
+
+	tfo_da_times = atomic_read(&sock_net(sk)->ipv4.tfo_active_disable_times);
 	if (!tfo_da_times)
 		return false;
 

net/ipv4/tcp_ipv4.c

@@ -2965,7 +2965,7 @@ static int __net_init tcp_sk_init(struct net *net)
 	net->ipv4.sysctl_tcp_comp_sack_nr = 44;
 	net->ipv4.sysctl_tcp_fastopen = TFO_CLIENT_ENABLE;
 	spin_lock_init(&net->ipv4.tcp_fastopen_ctx_lock);
-	net->ipv4.sysctl_tcp_fastopen_blackhole_timeout = 60 * 60;
+	net->ipv4.sysctl_tcp_fastopen_blackhole_timeout = 0;
 	atomic_set(&net->ipv4.tfo_active_disable_times, 0);
 
 	/* Reno is always built in */