[PATCH net-next 2/8] tcp: undo initial congestion window on false SYN timeout

[Yuchung Cheng <ycheng@google.com>]

Linux implements RFC6298 and use an initial congestion window of 1
upon establishing the connection if the SYN packet is retransmitted 2
or more times. In cellular networks SYN timeouts are often spurious
if the wireless radio was dormant or idle. Also some network path
is longer than the default SYN timeout. Having a minimal cwnd on
both cases are detrimental to TCP startup performance.

This patch extends TCP undo feature (RFC3522 aka TCP Eifel) to detect
spurious SYN timeout via TCP timestamps. Since tp->retrans_stamp
records the initial SYN timestamp instead of first retransmission, we
have to implement a different undo code additionally. The detection
also must happen before tcp_ack() as retrans_stamp is reset when
SYN is acknowledged.

Note this patch covers both active regular and fast open.

Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
---
 net/ipv4/tcp_input.c   | 16 ++++++++++++++++
 net/ipv4/tcp_metrics.c |  2 +-
 2 files changed, 17 insertions(+), 1 deletion(-)

diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c
index e2cbfc3ffa3f..695f840acc14 100644
--- a/net/ipv4/tcp_input.c
+++ b/net/ipv4/tcp_input.c
@@ -5748,6 +5748,21 @@ static void smc_check_reset_syn(struct tcp_sock *tp)
 #endif
 }
 
+static void tcp_try_undo_spurious_syn(struct sock *sk)
+{
+	struct tcp_sock *tp = tcp_sk(sk);
+	u32 syn_stamp;
+
+	/* undo_marker is set when SYN or SYNACK times out. The timeout is
+	 * spurious if the ACK's timestamp option echo value matches the
+	 * original SYN timestamp.
+	 */
+	syn_stamp = tp->retrans_stamp;
+	if (tp->undo_marker && syn_stamp && tp->rx_opt.saw_tstamp &&
+	    syn_stamp == tp->rx_opt.rcv_tsecr)
+		tp->undo_marker = 0;
+}
+
 static int tcp_rcv_synsent_state_process(struct sock *sk, struct sk_buff *skb,
 					 const struct tcphdr *th)
 {
@@ -5815,6 +5830,7 @@ static int tcp_rcv_synsent_state_process(struct sock *sk, struct sk_buff *skb,
 		tcp_ecn_rcv_synack(tp, th);
 
 		tcp_init_wl(tp, TCP_SKB_CB(skb)->seq);
+		tcp_try_undo_spurious_syn(sk);
 		tcp_ack(sk, skb, FLAG_SLOWPATH);
 
 		/* Ok.. it's good. Set up sequence numbers and
diff --git a/net/ipv4/tcp_metrics.c b/net/ipv4/tcp_metrics.c
index f262f2cace29..d4d687330e2b 100644
--- a/net/ipv4/tcp_metrics.c
+++ b/net/ipv4/tcp_metrics.c
@@ -517,7 +517,7 @@ void tcp_init_metrics(struct sock *sk)
 	 * initRTO, we only reset cwnd when more than 1 SYN/SYN-ACK
 	 * retransmission has occurred.
 	 */
-	if (tp->total_retrans > 1)
+	if (tp->total_retrans > 1 && tp->undo_marker)
 		tp->snd_cwnd = 1;
 	else
 		tp->snd_cwnd = tcp_init_cwnd(tp, dst);
-- 
2.21.0.593.g511ec345e18-goog