Skip site navigation (1)Skip section navigation (2)
Date:      Sat, 25 Aug 2012 09:26:37 +0000 (UTC)
From:      Randall Stewart <rrs@FreeBSD.org>
To:        src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-head@freebsd.org
Subject:   svn commit: r239672 - head/sys/netinet
Message-ID:  <201208250926.q7P9QbcQ014751@svn.freebsd.org>

next in thread | raw e-mail | index | archive | help
Author: rrs
Date: Sat Aug 25 09:26:37 2012
New Revision: 239672
URL: http://svn.freebsd.org/changeset/base/239672

Log:
  This small change takes care of a race condition
  that can occur when both sides close at the same time.
  If that occurs, without this fix the connection enters
  FIN1 on both sides and they will forever send FIN|ACK at
  each other until the connection times out. This is because
  we stopped processing the FIN|ACK and thus did not advance
  the sequence and so never ACK'd each others FIN. This
  fix adjusts it so we *do* process the FIN properly and
  the race goes away ;-)
  
  MFC after:	1 month

Modified:
  head/sys/netinet/tcp_input.c

Modified: head/sys/netinet/tcp_input.c
==============================================================================
--- head/sys/netinet/tcp_input.c	Sat Aug 25 08:31:21 2012	(r239671)
+++ head/sys/netinet/tcp_input.c	Sat Aug 25 09:26:37 2012	(r239672)
@@ -2414,6 +2414,16 @@ tcp_do_segment(struct mbuf *m, struct tc
 						}
 					} else
 						tp->snd_cwnd += tp->t_maxseg;
+					if ((thflags & TH_FIN) &&
+					    (TCPS_HAVERCVDFIN(tp->t_state) == 0)) {
+						/* 
+						 * If its a fin we need to process
+						 * it to avoid a race where both
+						 * sides enter FIN-WAIT and send FIN|ACK
+						 * at the same time.
+						 */
+						break;
+					}
 					(void) tcp_output(tp);
 					goto drop;
 				} else if (tp->t_dupacks == tcprexmtthresh) {
@@ -2453,6 +2463,16 @@ tcp_do_segment(struct mbuf *m, struct tc
 					}
 					tp->snd_nxt = th->th_ack;
 					tp->snd_cwnd = tp->t_maxseg;
+					if ((thflags & TH_FIN) &&
+					    (TCPS_HAVERCVDFIN(tp->t_state) == 0)) {
+						/* 
+						 * If its a fin we need to process
+						 * it to avoid a race where both
+						 * sides enter FIN-WAIT and send FIN|ACK
+						 * at the same time.
+						 */
+						break;
+					}
 					(void) tcp_output(tp);
 					KASSERT(tp->snd_limited <= 2,
 					    ("%s: tp->snd_limited too big",
@@ -2479,6 +2499,16 @@ tcp_do_segment(struct mbuf *m, struct tc
 					    (tp->snd_nxt - tp->snd_una) +
 					    (tp->t_dupacks - tp->snd_limited) *
 					    tp->t_maxseg;
+					if ((thflags & TH_FIN) &&
+					    (TCPS_HAVERCVDFIN(tp->t_state) == 0)) {
+						/* 
+						 * If its a fin we need to process
+						 * it to avoid a race where both
+						 * sides enter FIN-WAIT and send FIN|ACK
+						 * at the same time.
+						 */
+						break;
+					}
 					(void) tcp_output(tp);
 					sent = tp->snd_max - oldsndmax;
 					if (sent > tp->t_maxseg) {



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?201208250926.q7P9QbcQ014751>