MFC: r290122

Calculate the correct amount of bytes that are in-flight for a connection as
suggested by RFC 6675.

MFC: r292046
r290122 added 4 bytes and removed 8 in struct sackhint. Add a pad entry of 4
bytes to restore the size.

3 files changed, 40 insertions, 4 deletions

diff --git a/sys/netinet/tcp_input.c b/sys/netinet/tcp_input.c
index b824ba2..b6fddc7 100644
--- a/sys/netinet/tcp_input.c
+++ b/sys/netinet/tcp_input.c
@@ -148,6 +148,11 @@ SYSCTL_VNET_INT(_net_inet_tcp, OID_AUTO, drop_synfin, CTLFLAG_RW,
     &VNET_NAME(drop_synfin), 0,
     "Drop TCP packets with SYN+FIN set");
 
+VNET_DEFINE(int, tcp_do_rfc6675_pipe) = 0;
+SYSCTL_INT(_net_inet_tcp, OID_AUTO, do_pipe, CTLFLAG_VNET | CTLFLAG_RW,
+    &VNET_NAME(tcp_do_rfc6675_pipe), 0,
+    "Use calculated pipe/in-flight bytes per RFC 6675");
+
 VNET_DEFINE(int, tcp_do_rfc3042) = 1;
 #define	V_tcp_do_rfc3042	VNET(tcp_do_rfc3042)
 SYSCTL_VNET_INT(_net_inet_tcp, OID_AUTO, rfc3042, CTLFLAG_RW,
@@ -2455,6 +2460,12 @@ tcp_do_segment(struct mbuf *m, struct tcphdr *th, struct socket *so,
 		    ((to.to_flags & TOF_SACK) ||
 		     !TAILQ_EMPTY(&tp->snd_holes)))
 			tcp_sack_doack(tp, &to, th->th_ack);
+		else
+			/*
+			 * Reset the value so that previous (valid) value
+			 * from the last ack with SACK doesn't get used.
+			 */
+			tp->sackhint.sacked_bytes = 0;
 
 		/* Run HHOOK_TCP_ESTABLISHED_IN helper hooks. */
 		hhook_run_tcp_est_in(tp, th, &to);
@@ -2505,8 +2516,12 @@ tcp_do_segment(struct mbuf *m, struct tcphdr *th, struct socket *so,
 						 * we have less than 1/2 the original window's
 						 * worth of data in flight.
 						 */
-						awnd = (tp->snd_nxt - tp->snd_fack) +
-							tp->sackhint.sack_bytes_rexmit;
+						if (V_tcp_do_rfc6675_pipe)
+							awnd = tcp_compute_pipe(tp);
+						else
+							awnd = (tp->snd_nxt - tp->snd_fack) +
+								tp->sackhint.sack_bytes_rexmit;
+
 						if (awnd < tp->snd_ssthresh) {
 							tp->snd_cwnd += tp->t_maxseg;
 							if (tp->snd_cwnd > tp->snd_ssthresh)
@@ -3798,3 +3813,11 @@ tcp_newreno_partial_ack(struct tcpcb *tp, struct tcphdr *th)
 		tp->snd_cwnd = 0;
 	tp->snd_cwnd += tp->t_maxseg;
 }
+
+int
+tcp_compute_pipe(struct tcpcb *tp)
+{
+	return (tp->snd_max - tp->snd_una +
+		tp->sackhint.sack_bytes_rexmit -
+		tp->sackhint.sacked_bytes);
+}
diff --git a/sys/netinet/tcp_sack.c b/sys/netinet/tcp_sack.c
index 440bd64..28ee45c 100644
--- a/sys/netinet/tcp_sack.c
+++ b/sys/netinet/tcp_sack.c
@@ -368,6 +368,7 @@ tcp_sack_doack(struct tcpcb *tp, struct tcpopt *to, tcp_seq th_ack)
 	 * received new blocks from the other side.
 	 */
 	if (to->to_flags & TOF_SACK) {
+		tp->sackhint.sacked_bytes = 0;	/* reset */
 		for (i = 0; i < to->to_nsacks; i++) {
 			bcopy((to->to_sacks + i * TCPOLEN_SACK),
 			    &sack, sizeof(sack));
@@ -378,8 +379,11 @@ tcp_sack_doack(struct tcpcb *tp, struct tcpopt *to, tcp_seq th_ack)
 			    SEQ_GT(sack.start, th_ack) &&
 			    SEQ_LT(sack.start, tp->snd_max) &&
 			    SEQ_GT(sack.end, tp->snd_una) &&
-			    SEQ_LEQ(sack.end, tp->snd_max))
+			    SEQ_LEQ(sack.end, tp->snd_max)) {
 				sack_blocks[num_sack_blks++] = sack;
+				tp->sackhint.sacked_bytes +=
+				    (sack.end-sack.start);
+			}
 		}
 	}
 	/*
diff --git a/sys/netinet/tcp_var.h b/sys/netinet/tcp_var.h
index cff81c5..2d008c8 100644
--- a/sys/netinet/tcp_var.h
+++ b/sys/netinet/tcp_var.h
@@ -73,7 +73,12 @@ struct sackhint {
 	tcp_seq		last_sack_ack;	/* Most recent/largest sacked ack */
 
 	int		ispare;		/* explicit pad for 64bit alignment */
-	uint64_t	_pad[2];	/* 1 sacked_bytes, 1 TBD */
+	int             sacked_bytes;	/*
+					 * Total sacked bytes reported by the
+					 * receiver via sack option
+					 */
+	uint32_t	_pad1[1];	/* TBD */
+	uint64_t	_pad[1];	/* TBD */
 };
 
 struct tcptemp {
@@ -668,6 +673,9 @@ VNET_DECLARE(int, tcp_ecn_maxretries);
 VNET_DECLARE(struct hhook_head *, tcp_hhh[HHOOK_TCP_LAST + 1]);
 #define	V_tcp_hhh		VNET(tcp_hhh)
 
+VNET_DECLARE(int, tcp_do_rfc6675_pipe);
+#define V_tcp_do_rfc6675_pipe	VNET(tcp_do_rfc6675_pipe)
+
 int	 tcp_addoptions(struct tcpopt *, u_char *);
 int	 tcp_ccalgounload(struct cc_algo *unload_algo);
 struct tcpcb *
@@ -757,6 +765,7 @@ void	 tcp_sack_partialack(struct tcpcb *, struct tcphdr *);
 void	 tcp_free_sackholes(struct tcpcb *tp);
 int	 tcp_newreno(struct tcpcb *, struct tcphdr *);
 u_long	 tcp_seq_subtract(u_long, u_long );
+int	 tcp_compute_pipe(struct tcpcb *);
 
 void	cc_cong_signal(struct tcpcb *tp, struct tcphdr *th, uint32_t type);