svn commit: r342280 - head/sys/netinet

[Michael Tuexen]

Author: tuexen
Date: Thu Dec 20 16:05:30 2018
New Revision: 342280
URL: https://svnweb.freebsd.org/changeset/base/342280

Log:
  Fix a regression in the TCP handling of received segments.
  When receiving TCP segments the stack protects itself by limiting
  the resources allocated for a TCP connections. This patch adds
  an exception to these limitations for the TCP segement which is the next
  expected in-sequence segment. Without this patch, TCP connections
  may stall and finally fail in some cases of packet loss.
  
  Reported by:		jhb@
  Reviewed by:		jtl@, rrs@
  MFC after:		3 days
  Sponsored by:		Netflix, Inc.
  Differential Revision:	https://reviews.freebsd.org/D18580

Modified:
  head/sys/netinet/tcp_reass.c

Modified: head/sys/netinet/tcp_reass.c
==============================================================================
--- head/sys/netinet/tcp_reass.c	Thu Dec 20 15:53:27 2018	(r342279)
+++ head/sys/netinet/tcp_reass.c	Thu Dec 20 16:05:30 2018	(r342280)
@@ -579,7 +579,8 @@ tcp_reass(struct tcpcb *tp, struct tcphdr *th, tcp_seq
 	 */
 	lenofoh = tcp_reass_overhead_of_chain(m, &mlast);
 	sb = &tp->t_inpcb->inp_socket->so_rcv;
-	if ((sb->sb_mbcnt + tp->t_segqmbuflen + lenofoh) > sb->sb_mbmax) {
+	if ((th->th_seq != tp->rcv_nxt || !TCPS_HAVEESTABLISHED(tp->t_state)) &&
+	    (sb->sb_mbcnt + tp->t_segqmbuflen + lenofoh) > sb->sb_mbmax) {
 		/* No room */
 		TCPSTAT_INC(tcps_rcvreassfull);
 #ifdef TCP_REASS_COUNTERS
@@ -588,6 +589,11 @@ tcp_reass(struct tcpcb *tp, struct tcphdr *th, tcp_seq
 #ifdef TCP_REASS_LOGGING
 		tcp_log_reassm(tp, NULL, NULL, th->th_seq, lenofoh, TCP_R_LOG_LIMIT_REACHED, 0);
 #endif
+		if ((s = tcp_log_addrs(&tp->t_inpcb->inp_inc, th, NULL, NULL))) {
+			log(LOG_DEBUG, "%s; %s: mbuf count limit reached, "
+			    "segment dropped\n", s, __func__);
+			free(s, M_TCPLOG);
+		}
 		m_freem(m);
 		*tlenp = 0;
 #ifdef TCP_REASS_LOGGING
@@ -936,6 +942,20 @@ tcp_reass(struct tcpcb *tp, struct tcphdr *th, tcp_seq
 	 * is understood.
 	 */
 new_entry:
+	if (th->th_seq == tp->rcv_nxt && TCPS_HAVEESTABLISHED(tp->t_state)) {
+		tp->rcv_nxt += *tlenp;
+		flags = th->th_flags & TH_FIN;
+		TCPSTAT_INC(tcps_rcvoopack);
+		TCPSTAT_ADD(tcps_rcvoobyte, *tlenp);
+		SOCKBUF_LOCK(&so->so_rcv);
+		if (so->so_rcv.sb_state & SBS_CANTRCVMORE) {
+			m_freem(m);
+		} else {
+			sbappendstream_locked(&so->so_rcv, m, 0);
+		}
+		sorwakeup_locked(so);
+		return (flags);
+	}
 	if (tcp_new_limits) {
 		if ((tp->t_segqlen > tcp_reass_queue_guard) &&
 		    (*tlenp < MSIZE)) {
@@ -960,9 +980,7 @@ new_entry:
 			return (0);
 		}
 	} else {
-
-		if ((th->th_seq != tp->rcv_nxt || !TCPS_HAVEESTABLISHED(tp->t_state)) &&
-		    tp->t_segqlen >= min((so->so_rcv.sb_hiwat / tp->t_maxseg) + 1,
+		if (tp->t_segqlen >= min((so->so_rcv.sb_hiwat / tp->t_maxseg) + 1,
 					 tcp_reass_maxqueuelen)) {
 			TCPSTAT_INC(tcps_rcvreassfull);
 			*tlenp = 0;