tcp: Make RFC 6191 support configurable

Add a default-on per-VIMAGE sysctl for RFC 6191 connection recycling.
This makes it possible to merge the change to older branches where it
can be switched off by default to minimize risk.

MFC after:	1 week
Sponsored by:	Klara, Inc.
Sponsored by:	Modirum MDPay
Reviewed by:	pouria, marius.h_lden.org, tuexen
Differential Revision:	https://reviews.freebsd.org/D57045

2 files changed, 39 insertions, 21 deletions

diff --git a/sys/netinet/tcp_timewait.c b/sys/netinet/tcp_timewait.c
index 4f4ca445fa46..276733066c30 100644
--- a/sys/netinet/tcp_timewait.c
+++ b/sys/netinet/tcp_timewait.c
@@ -91,6 +91,11 @@
 
 #include <security/mac/mac_framework.h>
 
+VNET_DEFINE(int, tcp_do_rfc6191) = 1;
+SYSCTL_INT(_net_inet_tcp, OID_AUTO, rfc3465, CTLFLAG_VNET | CTLFLAG_RW,
+    &VNET_NAME(tcp_do_rfc6191), 0,
+    "Enable RFC 6191 (Reduced TIME-WAIT State)");
+
 static u_int
 tcp_eff_msl(struct tcpcb *tp)
 {
@@ -223,29 +228,40 @@ tcp_twcheck(struct inpcb *inp, struct tcpopt *to, struct tcphdr *th,
 	}
 
 	/*
-	 * If a new connection request is received
-	 * while in TIME_WAIT, drop the old connection
-	 * and start over if allowed by RFC 6191.
+	 * If a new connection request is received while in TIME_WAIT,
+	 * drop the old connection and start over if appropriate.
+	 *
+	 * The original rule is to start over if and only if the sequence
+	 * number of the new connection is greater than the last sequence
+	 * number seen on the old connection.
+	 *
+	 * Additionally, RFC 6191 allows restarting if the new connection
+	 * has TCP timestamps enabled and either the old one didn't, or it
+	 * did but the timestamp on the incoming SYN is greater than the
+	 * last timestamp seen on the old connection.
+	 *
 	 * Allow UDP port number changes in this case.
 	 */
-	if (((thflags & (TH_SYN | TH_ACK)) == TH_SYN) &&
-	    ((((tp->t_flags & TF_RCVD_TSTMP) != 0) &&
-	      ((to->to_flags & TOF_TS) != 0) &&
-	      TSTMP_LT(tp->ts_recent, to->to_tsval)) ||
-	     (((tp->t_flags & TF_RCVD_TSTMP) == 0) &&
-	      ((to->to_flags & TOF_TS) != 0) &&
-	      (V_tcp_tolerate_missing_ts == 0)) ||
-	     SEQ_GT(th->th_seq, tp->rcv_nxt))) {
-		/*
-		 * In case we can't upgrade our lock just pretend we have
-		 * lost this packet.
-		 */
-		if (INP_TRY_UPGRADE(inp) == 0)
-			goto drop;
-		if ((tp = tcp_close(tp)) != NULL)
-			INP_WUNLOCK(inp);
-		TCPSTAT_INC(tcps_tw_recycles);
-		return (true);
+	if ((thflags & (TH_SYN | TH_ACK)) == TH_SYN) {
+		bool rfc6191 = false;
+
+		if ((to->to_flags & TOF_TS) != 0 && V_tcp_do_rfc6191) {
+			rfc6191 = (tp->t_flags & TF_RCVD_TSTMP) != 0 ?
+			    TSTMP_LT(tp->ts_recent, to->to_tsval) :
+			    V_tcp_tolerate_missing_ts == 0;
+		}
+		if (rfc6191 || SEQ_GT(th->th_seq, tp->rcv_nxt)) {
+			/*
+			 * In case we can't upgrade our lock just pretend
+			 * we have lost this packet.
+			 */
+			if (INP_TRY_UPGRADE(inp) == 0)
+				goto drop;
+			if ((tp = tcp_close(tp)) != NULL)
+				INP_WUNLOCK(inp);
+			TCPSTAT_INC(tcps_tw_recycles);
+			return (true);
+		}
 	}
 
 	/*
diff --git a/sys/netinet/tcp_var.h b/sys/netinet/tcp_var.h
index a1b0519ceac3..5b3733e8e91e 100644
--- a/sys/netinet/tcp_var.h
+++ b/sys/netinet/tcp_var.h
@@ -1308,6 +1308,7 @@ VNET_DECLARE(int, tcp_tolerate_missing_ts);
 VNET_DECLARE(int, tcp_do_rfc3042);
 VNET_DECLARE(int, tcp_do_rfc3390);
 VNET_DECLARE(int, tcp_do_rfc3465);
+VNET_DECLARE(int, tcp_do_rfc6191);
 VNET_DECLARE(int, tcp_do_sack);
 VNET_DECLARE(int, tcp_do_tso);
 VNET_DECLARE(int, tcp_ecn_maxretries);
@@ -1358,6 +1359,7 @@ VNET_DECLARE(struct inpcbinfo, tcbinfo);
 #define	V_tcp_do_rfc3042		VNET(tcp_do_rfc3042)
 #define	V_tcp_do_rfc3390		VNET(tcp_do_rfc3390)
 #define	V_tcp_do_rfc3465		VNET(tcp_do_rfc3465)
+#define	V_tcp_do_rfc6191		VNET(tcp_do_rfc6191)
 #define	V_tcp_do_sack			VNET(tcp_do_sack)
 #define	V_tcp_do_tso			VNET(tcp_do_tso)
 #define	V_tcp_ecn_maxretries		VNET(tcp_ecn_maxretries)