tcp: early retransmit: delayed fast retransmit

Implementing the advanced early retransmit (sysctl_tcp_early_retrans==2).
Delays the fast retransmit by an interval of RTT/4. We borrow the
RTO timer to implement the delay. If we receive another ACK or send
a new packet, the timer is cancelled and restored to original RTO
value offset by time elapsed.  When the delayed-ER timer fires,
we enter fast recovery and perform fast retransmit.

Signed-off-by: Yuchung Cheng <ycheng@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

5 files changed, 74 insertions, 11 deletions

diff --git a/include/linux/tcp.h b/include/linux/tcp.h
index 7859b416d46..d9b42c5be08 100644
--- a/include/linux/tcp.h
+++ b/include/linux/tcp.h
@@ -372,7 +372,8 @@ struct tcp_sock {
 		repair      : 1,
 		unused      : 1;
 	u8	repair_queue;
-	u8	do_early_retrans:1;/* Enable RFC5827 early-retransmit  */
+	u8	do_early_retrans:1,/* Enable RFC5827 early-retransmit  */
+		early_retrans_delayed:1; /* Delayed ER timer installed */
 
 /* RTT measurement */
 	u32	srtt;		/* smoothed round trip time << 3	*/
diff --git a/include/net/tcp.h b/include/net/tcp.h
index 685437a16c9..5283aa4bfa2 100644
--- a/include/net/tcp.h
+++ b/include/net/tcp.h
@@ -500,6 +500,8 @@ extern void tcp_send_delayed_ack(struct sock *sk);
 
 /* tcp_input.c */
 extern void tcp_cwnd_application_limited(struct sock *sk);
+extern void tcp_resume_early_retransmit(struct sock *sk);
+extern void tcp_rearm_rto(struct sock *sk);
 
 /* tcp_timer.c */
 extern void tcp_init_xmit_timers(struct sock *);
@@ -805,6 +807,7 @@ static inline void tcp_enable_early_retrans(struct tcp_sock *tp)
 {
 	tp->do_early_retrans = sysctl_tcp_early_retrans &&
 		!sysctl_tcp_thin_dupack && sysctl_tcp_reordering == 3;
+	tp->early_retrans_delayed = 0;
 }
 
 static inline void tcp_disable_early_retrans(struct tcp_sock *tp)
diff --git a/net/ipv4/tcp_input.c b/net/ipv4/tcp_input.c
index e042cabb695..7096790e06b 100644
--- a/net/ipv4/tcp_input.c
+++ b/net/ipv4/tcp_input.c
@@ -2344,6 +2344,27 @@ static inline int tcp_dupack_heuristics(const struct tcp_sock *tp)
 	return tcp_is_fack(tp) ? tp->fackets_out : tp->sacked_out + 1;
 }
 
+static bool tcp_pause_early_retransmit(struct sock *sk, int flag)
+{
+	struct tcp_sock *tp = tcp_sk(sk);
+	unsigned long delay;
+
+	/* Delay early retransmit and entering fast recovery for
+	 * max(RTT/4, 2msec) unless ack has ECE mark, no RTT samples
+	 * available, or RTO is scheduled to fire first.
+	 */
+	if (sysctl_tcp_early_retrans < 2 || (flag & FLAG_ECE) || !tp->srtt)
+		return false;
+
+	delay = max_t(unsigned long, (tp->srtt >> 5), msecs_to_jiffies(2));
+	if (!time_after(inet_csk(sk)->icsk_timeout, (jiffies + delay)))
+		return false;
+
+	inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS, delay, TCP_RTO_MAX);
+	tp->early_retrans_delayed = 1;
+	return true;
+}
+
 static inline int tcp_skb_timedout(const struct sock *sk,
 				   const struct sk_buff *skb)
 {
@@ -2451,7 +2472,7 @@ static inline int tcp_head_timedout(const struct sock *sk)
  * Main question: may we further continue forward transmission
  * with the same cwnd?
  */
-static int tcp_time_to_recover(struct sock *sk)
+static int tcp_time_to_recover(struct sock *sk, int flag)
 {
 	struct tcp_sock *tp = tcp_sk(sk);
 	__u32 packets_out;
@@ -2505,7 +2526,7 @@ static int tcp_time_to_recover(struct sock *sk)
 	if (tp->do_early_retrans && !tp->retrans_out && tp->sacked_out &&
 	    (tp->packets_out == (tp->sacked_out + 1) && tp->packets_out < 4) &&
 	    !tcp_may_send_now(sk))
-		return 1;
+		return !tcp_pause_early_retransmit(sk, flag);
 
 	return 0;
 }
@@ -3172,7 +3193,7 @@ static void tcp_fastretrans_alert(struct sock *sk, int pkts_acked,
 		if (icsk->icsk_ca_state <= TCP_CA_Disorder)
 			tcp_try_undo_dsack(sk);
 
-		if (!tcp_time_to_recover(sk)) {
+		if (!tcp_time_to_recover(sk, flag)) {
 			tcp_try_to_open(sk, flag);
 			return;
 		}
@@ -3271,16 +3292,47 @@ static void tcp_cong_avoid(struct sock *sk, u32 ack, u32 in_flight)
 /* Restart timer after forward progress on connection.
  * RFC2988 recommends to restart timer to now+rto.
  */
-static void tcp_rearm_rto(struct sock *sk)
+void tcp_rearm_rto(struct sock *sk)
 {
-	const struct tcp_sock *tp = tcp_sk(sk);
+	struct tcp_sock *tp = tcp_sk(sk);
 
 	if (!tp->packets_out) {
 		inet_csk_clear_xmit_timer(sk, ICSK_TIME_RETRANS);
 	} else {
-		inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS,
-					  inet_csk(sk)->icsk_rto, TCP_RTO_MAX);
+		u32 rto = inet_csk(sk)->icsk_rto;
+		/* Offset the time elapsed after installing regular RTO */
+		if (tp->early_retrans_delayed) {
+			struct sk_buff *skb = tcp_write_queue_head(sk);
+			const u32 rto_time_stamp = TCP_SKB_CB(skb)->when + rto;
+			s32 delta = (s32)(rto_time_stamp - tcp_time_stamp);
+			/* delta may not be positive if the socket is locked
+			 * when the delayed ER timer fires and is rescheduled.
+			 */
+			if (delta > 0)
+				rto = delta;
+		}
+		inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS, rto,
+					  TCP_RTO_MAX);
 	}
+	tp->early_retrans_delayed = 0;
+}
+
+/* This function is called when the delayed ER timer fires. TCP enters
+ * fast recovery and performs fast-retransmit.
+ */
+void tcp_resume_early_retransmit(struct sock *sk)
+{
+	struct tcp_sock *tp = tcp_sk(sk);
+
+	tcp_rearm_rto(sk);
+
+	/* Stop if ER is disabled after the delayed ER timer is scheduled */
+	if (!tp->do_early_retrans)
+		return;
+
+	tcp_enter_recovery(sk, false);
+	tcp_update_scoreboard(sk, 1);
+	tcp_xmit_retransmit_queue(sk);
 }
 
 /* If we get here, the whole TSO packet has not been acked. */
@@ -3729,6 +3781,9 @@ static int tcp_ack(struct sock *sk, const struct sk_buff *skb, int flag)
 	if (after(ack, tp->snd_nxt))
 		goto invalid_ack;
 
+	if (tp->early_retrans_delayed)
+		tcp_rearm_rto(sk);
+
 	if (after(ack, prior_snd_una))
 		flag |= FLAG_SND_UNA_ADVANCED;
 
diff --git a/net/ipv4/tcp_output.c b/net/ipv4/tcp_output.c
index 834e89fc541..d9473300992 100644
--- a/net/ipv4/tcp_output.c
+++ b/net/ipv4/tcp_output.c
@@ -78,9 +78,8 @@ static void tcp_event_new_data_sent(struct sock *sk, const struct sk_buff *skb)
 		tp->frto_counter = 3;
 
 	tp->packets_out += tcp_skb_pcount(skb);
-	if (!prior_packets)
-		inet_csk_reset_xmit_timer(sk, ICSK_TIME_RETRANS,
-					  inet_csk(sk)->icsk_rto, TCP_RTO_MAX);
+	if (!prior_packets || tp->early_retrans_delayed)
+		tcp_rearm_rto(sk);
 }
 
 /* SND.NXT, if window was not shrunk.
diff --git a/net/ipv4/tcp_timer.c b/net/ipv4/tcp_timer.c
index 34d4a02c2f1..e911e6c523e 100644
--- a/net/ipv4/tcp_timer.c
+++ b/net/ipv4/tcp_timer.c
@@ -319,6 +319,11 @@ void tcp_retransmit_timer(struct sock *sk)
 	struct tcp_sock *tp = tcp_sk(sk);
 	struct inet_connection_sock *icsk = inet_csk(sk);
 
+	if (tp->early_retrans_delayed) {
+		tcp_resume_early_retransmit(sk);
+		return;
+	}
+
 	if (!tp->packets_out)
 		goto out;