svn commit: r308943 - head/sys/netinet

[Hiren Panchasara]

Author: hiren
Date: Mon Nov 21 20:53:11 2016
New Revision: 308943
URL: https://svnweb.freebsd.org/changeset/base/308943

Log:
  For RTT calculations mid-session, we explicitly ignore ACKs with tsecr of 0 as
  many borken middle-boxes tend to do that. But during 3whs, in syncache_expand(),
  we don't do that which causes us to send a RST to such a client. Relax this
  constraint by only using tsecr to compare against timestamp that we sent when it
  is not 0. As a result, we'd now accept the final ACK of 3whs with tsecr of 0.
  
  Reviewed by:	    jtl, gnn
  Sponsored by:	    Limelight Networks
  Differential Revision:	https://reviews.freebsd.org/D8552

Modified:
  head/sys/netinet/tcp_syncache.c

Modified: head/sys/netinet/tcp_syncache.c
==============================================================================
--- head/sys/netinet/tcp_syncache.c	Mon Nov 21 19:47:37 2016	(r308942)
+++ head/sys/netinet/tcp_syncache.c	Mon Nov 21 20:53:11 2016	(r308943)
@@ -1069,10 +1069,17 @@ syncache_expand(struct in_conninfo *inc,
 	}
 
 	/*
-	 * If timestamps were negotiated the reflected timestamp
-	 * must be equal to what we actually sent in the SYN|ACK.
+	 * If timestamps were negotiated, the reflected timestamp
+	 * must be equal to what we actually sent in the SYN|ACK
+	 * except in the case of 0. Some boxes are known for sending
+	 * broken timestamp replies during the 3whs (and potentially
+	 * during the connection also).
+	 *
+	 * Accept the final ACK of 3whs with reflected timestamp of 0
+	 * instead of sending a RST and deleting the syncache entry.
 	 */
-	if ((to->to_flags & TOF_TS) && to->to_tsecr != sc->sc_ts) {
+	if ((to->to_flags & TOF_TS) && to->to_tsecr &&
+	    to->to_tsecr != sc->sc_ts) {
 		if ((s = tcp_log_addrs(inc, th, NULL, NULL)))
 			log(LOG_DEBUG, "%s; %s: TSECR %u != TS %u, "
 			    "segment rejected\n",