svn commit: r268506 - stable/9/sys/netinet

[Hiren Panchasara]

Author: hiren
Date: Thu Jul 10 21:02:58 2014
New Revision: 268506
URL: http://svnweb.freebsd.org/changeset/base/268506

Log:
  MFC r256920
  
  The TCP delayed ACK logic isn't aware of LRO passing up large aggregated
  segments thinking it received only one segment. This causes it to enable
  the delay the ACK for 100ms to wait for another segment which may never
  come because all the data was received already.
  
  Doing delayed ACK for LRO segments is bogus for two reasons: a) it pushes
  us further away from acking every other packet; b) it introduces additional
  delay in responding to the sender.  The latter is especially bad because it
  is in the nature of LRO to aggregated all segments of a burst with no more
  coming until an ACK is sent back.
  
  Change the delayed ACK logic to detect LRO segments by being larger than
  the MSS for this connection and issuing an immediate ACK for them to keep
  the ACK clock ticking without interruption.

Modified:
  stable/9/sys/netinet/tcp_input.c

Modified: stable/9/sys/netinet/tcp_input.c
==============================================================================
--- stable/9/sys/netinet/tcp_input.c	Thu Jul 10 20:59:54 2014	(r268505)
+++ stable/9/sys/netinet/tcp_input.c	Thu Jul 10 21:02:58 2014	(r268506)
@@ -502,10 +502,13 @@ do { \
  *	  the ack that opens up a 0-sized window and
  *		- delayed acks are enabled or
  *		- this is a half-synchronized T/TCP connection.
+ *	- the segment size is not larger than the MSS and LRO wasn't used
+ *	  for this segment.
  */
-#define DELAY_ACK(tp)							\
+#define DELAY_ACK(tp, tlen)						\
 	((!tcp_timer_active(tp, TT_DELACK) &&				\
 	    (tp->t_flags & TF_RXWIN0SENT) == 0) &&			\
+	    (tlen <= tp->t_maxopd) &&					\
 	    (V_tcp_delack_enabled || (tp->t_flags & TF_NEEDSYN)))
 
 /*
@@ -1850,7 +1853,7 @@ tcp_do_segment(struct mbuf *m, struct tc
 			}
 			/* NB: sorwakeup_locked() does an implicit unlock. */
 			sorwakeup_locked(so);
-			if (DELAY_ACK(tp)) {
+			if (DELAY_ACK(tp, tlen)) {
 				tp->t_flags |= TF_DELACK;
 			} else {
 				tp->t_flags |= TF_ACKNOW;
@@ -1938,7 +1941,7 @@ tcp_do_segment(struct mbuf *m, struct tc
 			 * If there's data, delay ACK; if there's also a FIN
 			 * ACKNOW will be turned on later.
 			 */
-			if (DELAY_ACK(tp) && tlen != 0)
+			if (DELAY_ACK(tp, tlen) && tlen != 0)
 				tcp_timer_activate(tp, TT_DELACK,
 				    tcp_delacktime);
 			else
@@ -2905,7 +2908,7 @@ dodata:							/* XXX */
 		if (th->th_seq == tp->rcv_nxt &&
 		    LIST_EMPTY(&tp->t_segq) &&
 		    TCPS_HAVEESTABLISHED(tp->t_state)) {
-			if (DELAY_ACK(tp))
+			if (DELAY_ACK(tp, tlen))
 				tp->t_flags |= TF_DELACK;
 			else
 				tp->t_flags |= TF_ACKNOW;