Have TCP Inflight disable itself if the RTT is below a certain

threshold.  Inflight doesn't make sense on a LAN as it has
trouble figuring out the maximal bandwidth because of the coarse
tick granularity.

The sysctl net.inet.tcp.inflight.rttthresh specifies the threshold
in milliseconds below which inflight will disengage.  It defaults
to 10ms.

Tested by:	Joao Barros <joao.barros-at-gmail.com>,
		Rich Murphey <rich-at-whiteoaklabs.com>
Sponsored by:	TCP/IP Optimization Fundraise 2005

6 files changed, 38 insertions, 2 deletions

diff --git a/sys/netinet/tcp_input.c b/sys/netinet/tcp_input.c
index e71b6bd..d50333b 100644
--- a/sys/netinet/tcp_input.c
+++ b/sys/netinet/tcp_input.c
@@ -1193,10 +1193,16 @@ after_listen:
 				 */
 				if ((to.to_flags & TOF_TS) != 0 &&
 				    to.to_tsecr) {
+					if (!tp->t_rttlow ||
+					    tp->t_rttlow > ticks - to.to_tsecr)
+						tp->t_rttlow = ticks - to.to_tsecr;
 					tcp_xmit_timer(tp,
 					    ticks - to.to_tsecr + 1);
 				} else if (tp->t_rtttime &&
 					    SEQ_GT(th->th_ack, tp->t_rtseq)) {
+					if (!tp->t_rttlow ||
+					    tp->t_rttlow > ticks - tp->t_rtttime)
+						tp->t_rttlow = ticks - tp->t_rtttime;
 					tcp_xmit_timer(tp,
 							ticks - tp->t_rtttime);
 				}
@@ -2077,8 +2083,12 @@ process_ACK:
 		 */
 		if ((to.to_flags & TOF_TS) != 0 &&
 		    to.to_tsecr) {
+			if (!tp->t_rttlow || tp->t_rttlow > ticks - to.to_tsecr)
+				tp->t_rttlow = ticks - to.to_tsecr;
 			tcp_xmit_timer(tp, ticks - to.to_tsecr + 1);
 		} else if (tp->t_rtttime && SEQ_GT(th->th_ack, tp->t_rtseq)) {
+			if (!tp->t_rttlow || tp->t_rttlow > ticks - tp->t_rtttime)
+				tp->t_rttlow = ticks - tp->t_rtttime;
 			tcp_xmit_timer(tp, ticks - tp->t_rtttime);
 		}
 		tcp_xmit_bandwidth_limit(tp, th->th_ack);
diff --git a/sys/netinet/tcp_reass.c b/sys/netinet/tcp_reass.c
index e71b6bd..d50333b 100644
--- a/sys/netinet/tcp_reass.c
+++ b/sys/netinet/tcp_reass.c
@@ -1193,10 +1193,16 @@ after_listen:
 				 */
 				if ((to.to_flags & TOF_TS) != 0 &&
 				    to.to_tsecr) {
+					if (!tp->t_rttlow ||
+					    tp->t_rttlow > ticks - to.to_tsecr)
+						tp->t_rttlow = ticks - to.to_tsecr;
 					tcp_xmit_timer(tp,
 					    ticks - to.to_tsecr + 1);
 				} else if (tp->t_rtttime &&
 					    SEQ_GT(th->th_ack, tp->t_rtseq)) {
+					if (!tp->t_rttlow ||
+					    tp->t_rttlow > ticks - tp->t_rtttime)
+						tp->t_rttlow = ticks - tp->t_rtttime;
 					tcp_xmit_timer(tp,
 							ticks - tp->t_rtttime);
 				}
@@ -2077,8 +2083,12 @@ process_ACK:
 		 */
 		if ((to.to_flags & TOF_TS) != 0 &&
 		    to.to_tsecr) {
+			if (!tp->t_rttlow || tp->t_rttlow > ticks - to.to_tsecr)
+				tp->t_rttlow = ticks - to.to_tsecr;
 			tcp_xmit_timer(tp, ticks - to.to_tsecr + 1);
 		} else if (tp->t_rtttime && SEQ_GT(th->th_ack, tp->t_rtseq)) {
+			if (!tp->t_rttlow || tp->t_rttlow > ticks - tp->t_rtttime)
+				tp->t_rttlow = ticks - tp->t_rtttime;
 			tcp_xmit_timer(tp, ticks - tp->t_rtttime);
 		}
 		tcp_xmit_bandwidth_limit(tp, th->th_ack);
diff --git a/sys/netinet/tcp_subr.c b/sys/netinet/tcp_subr.c
index 9267ea2..e61d541 100644
--- a/sys/netinet/tcp_subr.c
+++ b/sys/netinet/tcp_subr.c
@@ -193,6 +193,11 @@ static int	tcp_inflight_debug = 0;
 SYSCTL_INT(_net_inet_tcp_inflight, OID_AUTO, debug, CTLFLAG_RW,
     &tcp_inflight_debug, 0, "Debug TCP inflight calculations");
 
+static int	tcp_inflight_rttthresh;
+SYSCTL_PROC(_net_inet_tcp_inflight, OID_AUTO, rttthresh, CTLTYPE_INT|CTLFLAG_RW,
+    &tcp_inflight_rttthresh, 0, sysctl_msec_to_ticks, "I",
+    "RTT threshold below which inflight will deactivate itself");
+
 static int	tcp_inflight_min = 6144;
 SYSCTL_INT(_net_inet_tcp_inflight, OID_AUTO, min, CTLFLAG_RW,
     &tcp_inflight_min, 0, "Lower-bound for TCP inflight window");
@@ -253,6 +258,7 @@ tcp_init()
 	tcp_msl = TCPTV_MSL;
 	tcp_rexmit_min = TCPTV_MIN;
 	tcp_rexmit_slop = TCPTV_CPU_VAR;
+	tcp_inflight_rttthresh = TCPTV_INFLIGHT_RTTTHRESH;
 
 	INP_INFO_LOCK_INIT(&tcbinfo, "tcp");
 	LIST_INIT(&tcb);
@@ -1937,7 +1943,7 @@ tcp_xmit_bandwidth_limit(struct tcpcb *tp, tcp_seq ack_seq)
 	 * If inflight_enable is disabled in the middle of a tcp connection,
 	 * make sure snd_bwnd is effectively disabled.
 	 */
-	if (tcp_inflight_enable == 0) {
+	if (tcp_inflight_enable == 0 || tp->t_rttlow < tcp_inflight_rttthresh) {
 		tp->snd_bwnd = TCP_MAXWIN << TCP_MAX_WINSHIFT;
 		tp->snd_bandwidth = 0;
 		return;
diff --git a/sys/netinet/tcp_timer.h b/sys/netinet/tcp_timer.h
index c62060f..d3d7c94 100644
--- a/sys/netinet/tcp_timer.h
+++ b/sys/netinet/tcp_timer.h
@@ -86,6 +86,9 @@
 #define	TCPTV_KEEPINTVL	( 75*hz)		/* default probe interval */
 #define	TCPTV_KEEPCNT	8			/* max probes before drop */
 
+#define	TCPTV_INFLIGHT_RTTTHRESH (10*hz/1000)	/* below which inflight
+						   disengages, in msec */
+
 /*
  * Minimum retransmit timer is 3 ticks, for algorithmic stability.
  * TCPT_RANGESET() will add another TCPTV_CPU_VAR to deal with
diff --git a/sys/netinet/tcp_timewait.c b/sys/netinet/tcp_timewait.c
index 9267ea2..e61d541 100644
--- a/sys/netinet/tcp_timewait.c
+++ b/sys/netinet/tcp_timewait.c
@@ -193,6 +193,11 @@ static int	tcp_inflight_debug = 0;
 SYSCTL_INT(_net_inet_tcp_inflight, OID_AUTO, debug, CTLFLAG_RW,
     &tcp_inflight_debug, 0, "Debug TCP inflight calculations");
 
+static int	tcp_inflight_rttthresh;
+SYSCTL_PROC(_net_inet_tcp_inflight, OID_AUTO, rttthresh, CTLTYPE_INT|CTLFLAG_RW,
+    &tcp_inflight_rttthresh, 0, sysctl_msec_to_ticks, "I",
+    "RTT threshold below which inflight will deactivate itself");
+
 static int	tcp_inflight_min = 6144;
 SYSCTL_INT(_net_inet_tcp_inflight, OID_AUTO, min, CTLFLAG_RW,
     &tcp_inflight_min, 0, "Lower-bound for TCP inflight window");
@@ -253,6 +258,7 @@ tcp_init()
 	tcp_msl = TCPTV_MSL;
 	tcp_rexmit_min = TCPTV_MIN;
 	tcp_rexmit_slop = TCPTV_CPU_VAR;
+	tcp_inflight_rttthresh = TCPTV_INFLIGHT_RTTTHRESH;
 
 	INP_INFO_LOCK_INIT(&tcbinfo, "tcp");
 	LIST_INIT(&tcb);
@@ -1937,7 +1943,7 @@ tcp_xmit_bandwidth_limit(struct tcpcb *tp, tcp_seq ack_seq)
 	 * If inflight_enable is disabled in the middle of a tcp connection,
 	 * make sure snd_bwnd is effectively disabled.
 	 */
-	if (tcp_inflight_enable == 0) {
+	if (tcp_inflight_enable == 0 || tp->t_rttlow < tcp_inflight_rttthresh) {
 		tp->snd_bwnd = TCP_MAXWIN << TCP_MAX_WINSHIFT;
 		tp->snd_bandwidth = 0;
 		return;
diff --git a/sys/netinet/tcp_var.h b/sys/netinet/tcp_var.h
index b1515dd..f9050df 100644
--- a/sys/netinet/tcp_var.h
+++ b/sys/netinet/tcp_var.h
@@ -199,6 +199,7 @@ struct tcpcb {
 	tcp_seq sack_newdata;		/* New data xmitted in this recovery
 					   episode starts at this seq number */
 	struct sackhint	sackhint;	/* SACK scoreboard hint */
+	int	t_rttlow;		/* smallest observerved RTT */
 };
 
 #define IN_FASTRECOVERY(tp)	(tp->t_flags & TF_FASTRECOVERY)