tcp: add sysctl to tolerate TCP segments missing timestamps

When timestamp support has been negotiated, TCP segements received
without a timestamp should be discarded. However, there are broken
TCP implementations (for example, stacks used by Omniswitch 63xx and
64xx models), which send TCP segments without timestamps although
they negotiated timestamp support.
This patch adds a sysctl variable which tolerates such TCP segments
and allows to interoperate with broken stacks.

Reviewed by:		jtl@, rscheff@
Differential Revision:	https://reviews.freebsd.org/D28142
Sponsored by:		Netflix, Inc.
PR:			252449
MFC after:		1 week

8 files changed, 58 insertions, 18 deletions

diff --git a/share/man/man4/tcp.4 b/share/man/man4/tcp.4
index 24b2e2d24d19..dee4040c0a98 100644
--- a/share/man/man4/tcp.4
+++ b/share/man/man4/tcp.4
@@ -34,7 +34,7 @@
 .\"     From: @(#)tcp.4	8.1 (Berkeley) 6/5/93
 .\" $FreeBSD$
 .\"
-.Dd December 19, 2020
+.Dd January 14, 2021
 .Dt TCP 4
 .Os
 .Sh NAME
@@ -369,8 +369,17 @@ MIB.
 .Bl -tag -width ".Va TCPCTL_DO_RFC1323"
 .It Dv TCPCTL_DO_RFC1323
 .Pq Va rfc1323
-Implement the window scaling and timestamp options of RFC 1323
+Implement the window scaling and timestamp options of RFC 1323/RFC 7323
 (default is true).
+.It Va tolerate_missing_ts
+Tolerate the missing of timestamps (RFC 1323/RFC 7323) for
+.Tn TCP
+segments belonging to
+.Tn TCP
+connections for which support of
+.Tn TCP
+timestamps has been negotiated.
+(default is 0, i.e., the missing of timestamps is not tolerated).
 .It Dv TCPCTL_MSSDFLT
 .Pq Va mssdflt
 The default value used for the maximum segment size
@@ -728,12 +737,20 @@ when trying to use a TCP function block that is not available;
 .Xr tcp_functions 9
 .Rs
 .%A "V. Jacobson"
-.%A "R. Braden"
+.%A "B. Braden"
 .%A "D. Borman"
 .%T "TCP Extensions for High Performance"
 .%O "RFC 1323"
 .Re
 .Rs
+.%A "D. Borman"
+.%A "B. Braden"
+.%A "V. Jacobson"
+.%A "R. Scheffenegger"
+.%T "TCP Extensions for High Performance"
+.%O "RFC 7323"
+.Re
+.Rs
 .%A "A. Heffernan"
 .%T "Protection of BGP Sessions via the TCP MD5 Signature Option"
 .%O "RFC 2385"
diff --git a/sys/netinet/tcp_input.c b/sys/netinet/tcp_input.c
index dda37d4795c9..283d42b594bd 100644
--- a/sys/netinet/tcp_input.c
+++ b/sys/netinet/tcp_input.c
@@ -1692,11 +1692,12 @@ tcp_do_segment(struct mbuf *m, struct tcphdr *th, struct socket *so,
 	/*
 	 * If timestamps were negotiated during SYN/ACK and a
 	 * segment without a timestamp is received, silently drop
-	 * the segment, unless it is a RST segment.
+	 * the segment, unless it is a RST segment or missing timestamps are
+	 * tolerated.
 	 * See section 3.2 of RFC 7323.
 	 */
 	if ((tp->t_flags & TF_RCVD_TSTMP) && !(to.to_flags & TOF_TS)) {
-		if ((thflags & TH_RST) != 0) {
+		if (((thflags & TH_RST) != 0) || V_tcp_tolerate_missing_ts) {
 			if ((s = tcp_log_addrs(inc, th, NULL, NULL))) {
 				log(LOG_DEBUG, "%s; %s: Timestamp missing, "
 				    "segment processed normally\n",
diff --git a/sys/netinet/tcp_stacks/bbr.c b/sys/netinet/tcp_stacks/bbr.c
index e59d9b9ff168..49ed7e992fba 100644
--- a/sys/netinet/tcp_stacks/bbr.c
+++ b/sys/netinet/tcp_stacks/bbr.c
@@ -11463,11 +11463,12 @@ bbr_do_segment_nounlock(struct mbuf *m, struct tcphdr *th, struct socket *so,
 	/*
 	 * If timestamps were negotiated during SYN/ACK and a
 	 * segment without a timestamp is received, silently drop
-	 * the segment, unless it is a RST segment.
+	 * the segment, unless it is a RST segment or missing timestamps are
+	 * tolerated.
 	 * See section 3.2 of RFC 7323.
 	 */
 	if ((tp->t_flags & TF_RCVD_TSTMP) && !(to.to_flags & TOF_TS) &&
-	    ((thflags & TH_RST) == 0)) {
+	    ((thflags & TH_RST) == 0) && (V_tcp_tolerate_missing_ts == 0)) {
 		retval = 0;
 		goto done_with_input;
 	}
diff --git a/sys/netinet/tcp_stacks/rack.c b/sys/netinet/tcp_stacks/rack.c
index 7c81e8b3a2bc..8e00cae85496 100644
--- a/sys/netinet/tcp_stacks/rack.c
+++ b/sys/netinet/tcp_stacks/rack.c
@@ -10879,11 +10879,12 @@ rack_do_segment_nounlock(struct mbuf *m, struct tcphdr *th, struct socket *so,
 	/*
 	 * If timestamps were negotiated during SYN/ACK and a
 	 * segment without a timestamp is received, silently drop
-	 * the segment, unless it is a RST segment.
+	 * the segment, unless it is a RST segment or missing timestamps are
+	 * tolerated.
 	 * See section 3.2 of RFC 7323.
 	 */
 	if ((tp->t_flags & TF_RCVD_TSTMP) && !(to.to_flags & TOF_TS) &&
-	    ((thflags & TH_RST) == 0)) {
+	    ((thflags & TH_RST) == 0) && (V_tcp_tolerate_missing_ts == 0)) {
 		way_out = 5;
 		retval = 0;
 		goto done_with_input;
diff --git a/sys/netinet/tcp_subr.c b/sys/netinet/tcp_subr.c
index b1ccb9a1acbe..b39d2b43d3a2 100644
--- a/sys/netinet/tcp_subr.c
+++ b/sys/netinet/tcp_subr.c
@@ -256,6 +256,11 @@ SYSCTL_INT(_net_inet_tcp, TCPCTL_DO_RFC1323, rfc1323, CTLFLAG_VNET | CTLFLAG_RW,
     &VNET_NAME(tcp_do_rfc1323), 0,
     "Enable rfc1323 (high performance TCP) extensions");
 
+VNET_DEFINE(int, tcp_tolerate_missing_ts) = 0;
+SYSCTL_INT(_net_inet_tcp, OID_AUTO, tolerate_missing_ts, CTLFLAG_VNET | CTLFLAG_RW,
+    &VNET_NAME(tcp_tolerate_missing_ts), 0,
+    "Tolerate missing TCP timestamps");
+
 VNET_DEFINE(int, tcp_ts_offset_per_conn) = 1;
 SYSCTL_INT(_net_inet_tcp, OID_AUTO, ts_offset_per_conn, CTLFLAG_VNET | CTLFLAG_RW,
     &VNET_NAME(tcp_ts_offset_per_conn), 0,
diff --git a/sys/netinet/tcp_syncache.c b/sys/netinet/tcp_syncache.c
index 4697374084ab..e59a41fc1102 100644
--- a/sys/netinet/tcp_syncache.c
+++ b/sys/netinet/tcp_syncache.c
@@ -1231,18 +1231,30 @@ syncache_expand(struct in_conninfo *inc, struct tcpopt *to, struct tcphdr *th,
 		/*
 		 * If timestamps were negotiated during SYN/ACK and a
 		 * segment without a timestamp is received, silently drop
-		 * the segment.
+		 * the segment, unless the missing timestamps are tolerated.
 		 * See section 3.2 of RFC 7323.
 		 */
 		if ((sc->sc_flags & SCF_TIMESTAMP) &&
 		    !(to->to_flags & TOF_TS)) {
-			SCH_UNLOCK(sch);
-			if ((s = tcp_log_addrs(inc, th, NULL, NULL))) {
-				log(LOG_DEBUG, "%s; %s: Timestamp missing, "
-				    "segment silently dropped\n", s, __func__);
-				free(s, M_TCPLOG);
+			if (V_tcp_tolerate_missing_ts) {
+				if ((s = tcp_log_addrs(inc, th, NULL, NULL))) {
+					log(LOG_DEBUG,
+					    "%s; %s: Timestamp missing, "
+					    "segment processed normally\n",
+					    s, __func__);
+					free(s, M_TCPLOG);
+				}
+			} else {
+				SCH_UNLOCK(sch);
+				if ((s = tcp_log_addrs(inc, th, NULL, NULL))) {
+					log(LOG_DEBUG,
+					    "%s; %s: Timestamp missing, "
+					    "segment silently dropped\n",
+					    s, __func__);
+					free(s, M_TCPLOG);
+				}
+				return (-1);  /* Do not send RST */
 			}
-			return (-1);  /* Do not send RST */
 		}
 
 		/*
diff --git a/sys/netinet/tcp_timewait.c b/sys/netinet/tcp_timewait.c
index d7e3cce44b89..f98927b196fc 100644
--- a/sys/netinet/tcp_timewait.c
+++ b/sys/netinet/tcp_timewait.c
@@ -451,10 +451,11 @@ tcp_twcheck(struct inpcb *inp, struct tcpopt *to, struct tcphdr *th,
 	/*
 	 * If timestamps were negotiated during SYN/ACK and a
 	 * segment without a timestamp is received, silently drop
-	 * the segment.
+	 * the segment, unless the missing timestamps are tolerated.
 	 * See section 3.2 of RFC 7323.
 	 */
-	if (((to->to_flags & TOF_TS) == 0) && (tw->t_recent != 0)) {
+	if (((to->to_flags & TOF_TS) == 0) && (tw->t_recent != 0) &&
+	    (V_tcp_tolerate_missing_ts == 0)) {
 		goto drop;
 	}
 
diff --git a/sys/netinet/tcp_var.h b/sys/netinet/tcp_var.h
index d9235f23706c..6e22d75ac441 100644
--- a/sys/netinet/tcp_var.h
+++ b/sys/netinet/tcp_var.h
@@ -840,6 +840,7 @@ VNET_DECLARE(int, tcp_do_autosndbuf);
 VNET_DECLARE(int, tcp_do_ecn);
 VNET_DECLARE(int, tcp_do_newcwv);
 VNET_DECLARE(int, tcp_do_rfc1323);
+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);
@@ -883,6 +884,7 @@ VNET_DECLARE(struct inpcbinfo, tcbinfo);
 #define	V_tcp_do_autosndbuf		VNET(tcp_do_autosndbuf)
 #define	V_tcp_do_ecn			VNET(tcp_do_ecn)
 #define	V_tcp_do_rfc1323		VNET(tcp_do_rfc1323)
+#define	V_tcp_tolerate_missing_ts	VNET(tcp_tolerate_missing_ts)
 #define V_tcp_ts_offset_per_conn	VNET(tcp_ts_offset_per_conn)
 #define	V_tcp_do_rfc3042		VNET(tcp_do_rfc3042)
 #define	V_tcp_do_rfc3390		VNET(tcp_do_rfc3390)