Reap FIN_WAIT_2 connections marked SOCANTRCVMORE faster. This mitigate

potential issues where the peer does not close, potentially leaving
thousands of connections in FIN_WAIT_2. This is controlled by a new sysctl
fast_finwait2_recycle, which is disabled by default.

Reviewed by: gnn, silby.

8 files changed, 53 insertions, 10 deletions

diff --git a/sys/netinet/tcp_input.c b/sys/netinet/tcp_input.c
index 298dec1..01f2316 100644
--- a/sys/netinet/tcp_input.c
+++ b/sys/netinet/tcp_input.c
@@ -2301,8 +2301,12 @@ process_ACK:
 		 * compressed state.
 		 */
 				if (so->so_rcv.sb_state & SBS_CANTRCVMORE) {
+					int timeout;
+
 					soisdisconnected(so);
-					callout_reset(tp->tt_2msl, tcp_maxidle,
+					timeout = (tcp_fast_finwait2_recycle) ? 
+						tcp_finwait2_timeout : tcp_maxidle;
+					callout_reset(tp->tt_2msl, timeout,
 						      tcp_timer_2msl, tp);
 				}
 				tp->t_state = TCPS_FIN_WAIT_2;
diff --git a/sys/netinet/tcp_reass.c b/sys/netinet/tcp_reass.c
index 298dec1..01f2316 100644
--- a/sys/netinet/tcp_reass.c
+++ b/sys/netinet/tcp_reass.c
@@ -2301,8 +2301,12 @@ process_ACK:
 		 * compressed state.
 		 */
 				if (so->so_rcv.sb_state & SBS_CANTRCVMORE) {
+					int timeout;
+
 					soisdisconnected(so);
-					callout_reset(tp->tt_2msl, tcp_maxidle,
+					timeout = (tcp_fast_finwait2_recycle) ? 
+						tcp_finwait2_timeout : tcp_maxidle;
+					callout_reset(tp->tt_2msl, timeout,
 						      tcp_timer_2msl, tp);
 				}
 				tp->t_state = TCPS_FIN_WAIT_2;
diff --git a/sys/netinet/tcp_subr.c b/sys/netinet/tcp_subr.c
index 4257603..14287b2 100644
--- a/sys/netinet/tcp_subr.c
+++ b/sys/netinet/tcp_subr.c
@@ -329,6 +329,7 @@ tcp_init(void)
 	tcp_rexmit_min = TCPTV_MIN;
 	tcp_rexmit_slop = TCPTV_CPU_VAR;
 	tcp_inflight_rttthresh = TCPTV_INFLIGHT_RTTTHRESH;
+	tcp_finwait2_timeout = TCPTV_FINWAIT2_TIMEOUT;
 
 	INP_INFO_LOCK_INIT(&tcbinfo, "tcp");
 	LIST_INIT(&tcb);
diff --git a/sys/netinet/tcp_timer.c b/sys/netinet/tcp_timer.c
index 86d9c6a..33e4806 100644
--- a/sys/netinet/tcp_timer.c
+++ b/sys/netinet/tcp_timer.c
@@ -96,6 +96,15 @@ static int	always_keepalive = 1;
 SYSCTL_INT(_net_inet_tcp, OID_AUTO, always_keepalive, CTLFLAG_RW,
     &always_keepalive , 0, "Assume SO_KEEPALIVE on all TCP connections");
 
+int    tcp_fast_finwait2_recycle = 0;
+SYSCTL_INT(_net_inet_tcp, OID_AUTO, fast_finwait2_recycle, CTLFLAG_RW, 
+	   &tcp_fast_finwait2_recycle, 0, "Recycle closed FIN_WAIT_2 connections faster");
+
+int    tcp_finwait2_timeout;
+SYSCTL_PROC(_net_inet_tcp, OID_AUTO, finwait2_timeout, CTLTYPE_INT|CTLFLAG_RW,
+	    &tcp_finwait2_timeout, 0, sysctl_msec_to_ticks, "I", "");
+
+
 static int	tcp_keepcnt = TCPTV_KEEPCNT;
 	/* max idle probes */
 int	tcp_maxpersistidle;
@@ -211,13 +220,24 @@ tcp_timer_2msl(xtp)
 	 * still waiting for peer to close and connection has been idle
 	 * too long, or if 2MSL time is up from TIME_WAIT, delete connection
 	 * control block.  Otherwise, check again in a bit.
+	 *
+	 * If fastrecycle of FIN_WAIT_2, in FIN_WAIT_2 and receiver has closed, 
+	 * there's no point in hanging onto FIN_WAIT_2 socket. Just close it. 
+	 * Ignore fact that there were recent incoming segments.
 	 */
-	if (tp->t_state != TCPS_TIME_WAIT &&
-	    (ticks - tp->t_rcvtime) <= tcp_maxidle)
-		callout_reset(tp->tt_2msl, tcp_keepintvl,
-			      tcp_timer_2msl, tp);
-	else
-		tp = tcp_close(tp);
+	if (tcp_fast_finwait2_recycle && tp->t_state == TCPS_FIN_WAIT_2 &&
+	    tp->t_inpcb && tp->t_inpcb->inp_socket && 
+	    (tp->t_inpcb->inp_socket->so_rcv.sb_state & SBS_CANTRCVMORE)) {
+		tcpstat.tcps_finwait2_drops++;
+		tp = tcp_close(tp);             
+	} else {
+		if (tp->t_state != TCPS_TIME_WAIT &&
+		   (ticks - tp->t_rcvtime) <= tcp_maxidle)
+		       callout_reset(tp->tt_2msl, tcp_keepintvl,
+				     tcp_timer_2msl, tp);
+	       else
+		       tp = tcp_close(tp);
+       }
 
 #ifdef TCPDEBUG
 	if (tp != NULL && (tp->t_inpcb->inp_socket->so_options & SO_DEBUG))
diff --git a/sys/netinet/tcp_timer.h b/sys/netinet/tcp_timer.h
index 1e1c239..8d8742de 100644
--- a/sys/netinet/tcp_timer.h
+++ b/sys/netinet/tcp_timer.h
@@ -89,6 +89,8 @@
 #define	TCPTV_INFLIGHT_RTTTHRESH (10*hz/1000)	/* below which inflight
 						   disengages, in msec */
 
+#define TCPTV_FINWAIT2_TIMEOUT (60*hz)         /* FIN_WAIT_2 timeout if no receiver */
+
 /*
  * Minimum retransmit timer is 3 ticks, for algorithmic stability.
  * TCPT_RANGESET() will add another TCPTV_CPU_VAR to deal with
@@ -152,6 +154,9 @@ extern int tcp_backoff[];
 
 struct tcptw;
 
+extern int tcp_finwait2_timeout;
+extern int tcp_fast_finwait2_recycle;
+
 void	tcp_timer_init(void);
 void	tcp_timer_2msl(void *xtp);
 struct tcptw *
diff --git a/sys/netinet/tcp_timewait.c b/sys/netinet/tcp_timewait.c
index 4257603..14287b2 100644
--- a/sys/netinet/tcp_timewait.c
+++ b/sys/netinet/tcp_timewait.c
@@ -329,6 +329,7 @@ tcp_init(void)
 	tcp_rexmit_min = TCPTV_MIN;
 	tcp_rexmit_slop = TCPTV_CPU_VAR;
 	tcp_inflight_rttthresh = TCPTV_INFLIGHT_RTTTHRESH;
+	tcp_finwait2_timeout = TCPTV_FINWAIT2_TIMEOUT;
 
 	INP_INFO_LOCK_INIT(&tcbinfo, "tcp");
 	LIST_INIT(&tcb);
diff --git a/sys/netinet/tcp_usrreq.c b/sys/netinet/tcp_usrreq.c
index c4362cf..57195a4 100644
--- a/sys/netinet/tcp_usrreq.c
+++ b/sys/netinet/tcp_usrreq.c
@@ -1581,9 +1581,14 @@ tcp_usrclosed(tp)
 	if (tp && tp->t_state >= TCPS_FIN_WAIT_2) {
 		soisdisconnected(tp->t_inpcb->inp_socket);
 		/* To prevent the connection hanging in FIN_WAIT_2 forever. */
-		if (tp->t_state == TCPS_FIN_WAIT_2)
-			callout_reset(tp->tt_2msl, tcp_maxidle,
+		if (tp->t_state == TCPS_FIN_WAIT_2) {
+			int timeout;
+
+			timeout = (tcp_fast_finwait2_recycle) ? 
+				tcp_finwait2_timeout : tcp_maxidle;
+			callout_reset(tp->tt_2msl, timeout,
 				      tcp_timer_2msl, tp);
+		}
 	}
 }
 
diff --git a/sys/netinet/tcp_var.h b/sys/netinet/tcp_var.h
index 4c50a82..46d396d 100644
--- a/sys/netinet/tcp_var.h
+++ b/sys/netinet/tcp_var.h
@@ -413,6 +413,8 @@ struct	tcpstat {
 	u_long	tcps_hc_added;		/* entry added to hostcache */
 	u_long	tcps_hc_bucketoverflow;	/* hostcache per bucket limit hit */
 
+	u_long  tcps_finwait2_drops;    /* Drop FIN_WAIT_2 connection after time limit */
+
 	/* SACK related stats */
 	u_long	tcps_sack_recovery_episode; /* SACK recovery episodes */
 	u_long  tcps_sack_rexmits;	    /* SACK rexmit segments   */
@@ -455,6 +457,7 @@ struct	xtcpcb {
 #define	TCPCTL_SACK		14	/* Selective Acknowledgement,rfc 2018 */
 #define	TCPCTL_DROP		15	/* drop tcp connection */
 #define	TCPCTL_MAXID		16
+#define TCPCTL_FINWAIT2_TIMEOUT        17
 
 #define TCPCTL_NAMES { \
 	{ 0, 0 }, \