From owner-freebsd-net  Fri Jul 28 15:44:12 2000
Delivered-To: freebsd-net@freebsd.org
Received: from bubba.whistle.com (bubba.whistle.com [207.76.205.7])
	by hub.freebsd.org (Postfix) with ESMTP
	id 95D3237B63A; Fri, 28 Jul 2000 15:44:05 -0700 (PDT)
	(envelope-from archie@whistle.com)
Received: (from archie@localhost)
	by bubba.whistle.com (8.9.3/8.9.3) id PAA28204;
	Fri, 28 Jul 2000 15:44:03 -0700 (PDT)
	(envelope-from archie)
From: Archie Cobbs <archie@whistle.com>
Message-Id: <200007282244.PAA28204@bubba.whistle.com>
Subject: Re: sub-optimal tcp_ouput() performance in the face of ENOBUFS
In-Reply-To: <Pine.BSF.4.21.0007272130290.16889-100000@achilles.silby.com> from
 Mike Silbersack at "Jul 27, 2000 09:35:57 pm"
To: freebsd-net@freebsd.org
Date: Fri, 28 Jul 2000 15:44:03 -0700 (PDT)
Cc: silby@silby.com, jlemon@freebsd.org,
	wollman@khavrinen.lcs.mit.edu
X-Mailer: ELM [version 2.4ME+ PL68 (25)]
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Sender: owner-freebsd-net@FreeBSD.ORG
Precedence: bulk
X-Loop: FreeBSD.org

Mike Silbersack writes:
> > It seems like there is sub-optimal behaviour in tcp_output(),
> > and I'm wondering what other people think.
> > 
> > Consider an output interface whose transmit queue is full.
> > tcp_output() calls ip_output(), and it will return ENOBUFS.
> > Here's where this is handled (tcp_output, line 863):

OK, here's a first stab at a patch.. reviews requested..

This dramatically improves performance for me when I lower the
max queue length. Still get good throughput even with 25% of the
transmitted packets returning ENOBUFS.

However, I haven't tested the edge conditions: SYN and FIN,
persist state, etc.

Thanks,
-Archie

___________________________________________________________________________
Archie Cobbs   *   Whistle Communications, Inc.  *   http://www.whistle.com

Index: tcp_output.c
===================================================================
RCS file: /home/ncvs/src/sys/netinet/tcp_output.c,v
retrieving revision 1.39.2.4
diff -u -r1.39.2.4 tcp_output.c
--- tcp_output.c	2000/07/15 07:14:31	1.39.2.4
+++ tcp_output.c	2000/07/28 22:25:21
@@ -860,6 +860,21 @@
 	    (so->so_options & SO_DONTROUTE), 0);
     }
 	if (error) {
+
+		/*
+		 * We know that the packet was lost, so back out the
+		 * sequence number advance, if any.
+		 */
+		if (tp->t_force == 0 || !callout_active(tp->tt_persist)) {
+			/*
+			 * No need to check for TH_FIN here because
+			 * the TF_SENTFIN flag handles that case.
+			 */
+			if (flags & TH_SYN)
+				tp->snd_nxt--;
+			tp->snd_nxt -= len;
+		}
+
 out:
 		if (error == ENOBUFS) {
 	                if (!callout_active(tp->tt_rexmt) &&


To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-net" in the body of the message