Subject: Re: TCP ECN diff
To: Rui Paulo <rpaulo@fnop.net>
From: Rui Paulo <rpaulo@fnop.net>
List: tech-net
Date: 08/31/2006 20:55:27
--Apple-Mail-3-505784705
Content-Transfer-Encoding: 7bit
Content-Type: text/plain;
charset=US-ASCII;
format=flowed
On Aug 31, 2006, at 8:32 PM, Rui Paulo wrote:
> This is only the kernel patch.
Attached is the whole patch.
--Apple-Mail-3-505784705
Content-Transfer-Encoding: 7bit
Content-Type: application/octet-stream;
x-unix-mode=0755;
name=netbsd-tcp-ecn.diff
Content-Disposition: attachment;
filename=netbsd-tcp-ecn.diff
Index: lib/libc/gen/sysctl.3
===================================================================
RCS file: /cvsroot/src/lib/libc/gen/sysctl.3,v
retrieving revision 1.178
diff -u -p -r1.178 sysctl.3
--- lib/libc/gen/sysctl.3 11 Aug 2006 19:17:47 -0000 1.178
+++ lib/libc/gen/sysctl.3 31 Aug 2006 19:53:47 -0000
@@ -1335,6 +1335,10 @@ Global number of TCP SACK holes.
Global maximum number of TCP SACK holes.
.It Li tcp.sack.maxholes
Maximum number of TCP SACK holes allowed per connection.
+.It Li tcp.ecn.enable
+If set to 1, enables RFC 3168 Explicit Congestion Notification.
+.It Li tcp.ecn.maxretries
+Number of times to retry sending the ECN-setup packet.
.It Li tcp.sendspace
The default TCP send buffer size.
.It Li tcp.slowhz
Index: sbin/sysctl/sysctl.8
===================================================================
RCS file: /cvsroot/src/sbin/sysctl/sysctl.8,v
retrieving revision 1.151
diff -u -p -r1.151 sysctl.8
--- sbin/sysctl/sysctl.8 8 Aug 2006 22:11:42 -0000 1.151
+++ sbin/sysctl/sysctl.8 31 Aug 2006 19:53:47 -0000
@@ -459,6 +459,8 @@ privilege can change the value.
.It net.inet.tcp.sack.globalholes integer no
.It net.inet.tcp.sack.globalmaxholes integer yes
.It net.inet.tcp.sack.maxholes integer yes
+.It net.inet.tcp.ecn.enable integer yes
+.It net.inet.tcp.ecn.maxretries integer yes
.It net.inet.tcp.sendspace integer yes
.It net.inet.tcp.slowhz integer no
.It net.inet.tcp.syn_bucket_limit integer yes
Index: sys/netinet/ip.h
===================================================================
RCS file: /cvsroot/src/sys/netinet/ip.h,v
retrieving revision 1.27
diff -u -p -r1.27 ip.h
--- sys/netinet/ip.h 10 Dec 2005 23:36:23 -0000 1.27
+++ sys/netinet/ip.h 31 Aug 2006 19:53:47 -0000
@@ -76,11 +76,6 @@ struct ip {
#define IPTOS_THROUGHPUT 0x08
#define IPTOS_RELIABILITY 0x04
/* IPTOS_LOWCOST 0x02 XXX */
-#if 1
-/* ECN RFC3168 obsoletes RFC2481, and these will be deprecated soon. */
-#define IPTOS_CE 0x01 /* congestion experienced */
-#define IPTOS_ECT 0x02 /* ECN-capable transport */
-#endif
/*
* Definitions for IP precedence (also in ip_tos) (hopefully unused)
Index: sys/netinet/ip_ecn.c
===================================================================
RCS file: /cvsroot/src/sys/netinet/ip_ecn.c,v
retrieving revision 1.14
diff -u -p -r1.14 ip_ecn.c
--- sys/netinet/ip_ecn.c 11 Dec 2005 12:24:57 -0000 1.14
+++ sys/netinet/ip_ecn.c 31 Aug 2006 19:53:47 -0000
@@ -67,10 +67,10 @@ ip_ecn_ingress(int mode, u_int8_t *outer
*outer = *inner;
switch (mode) {
case ECN_ALLOWED: /* ECN allowed */
- *outer &= ~IPTOS_CE;
+ *outer &= ~IPTOS_ECN_CE;
break;
case ECN_FORBIDDEN: /* ECN forbidden */
- *outer &= ~(IPTOS_ECT | IPTOS_CE);
+ *outer &= ~(IPTOS_ECN_ECT0 | IPTOS_ECN_CE);
break;
case ECN_NOCARE: /* no consideration to ECN */
break;
@@ -88,8 +88,8 @@ ip_ecn_egress(int mode, const u_int8_t *
switch (mode) {
case ECN_ALLOWED:
- if (*outer & IPTOS_CE)
- *inner |= IPTOS_CE;
+ if (*outer & IPTOS_ECN_CE)
+ *inner |= IPTOS_ECN_CE;
break;
case ECN_FORBIDDEN: /* ECN forbidden */
case ECN_NOCARE: /* no consideration to ECN */
Index: sys/netinet/tcp.h
===================================================================
RCS file: /cvsroot/src/sys/netinet/tcp.h,v
retrieving revision 1.21
diff -u -p -r1.21 tcp.h
--- sys/netinet/tcp.h 10 Dec 2005 23:36:23 -0000 1.21
+++ sys/netinet/tcp.h 31 Aug 2006 19:53:47 -0000
@@ -66,8 +66,8 @@ struct tcphdr {
#define TH_PUSH 0x08
#define TH_ACK 0x10
#define TH_URG 0x20
-#define TH_ECE 0x40 /* (unimplemented) */
-#define TH_CWR 0x80 /* (unimplemented) */
+#define TH_ECE 0x40
+#define TH_CWR 0x80
u_int16_t th_win; /* window */
u_int16_t th_sum; /* checksum */
u_int16_t th_urp; /* urgent pointer */
Index: sys/netinet/tcp_input.c
===================================================================
RCS file: /cvsroot/src/sys/netinet/tcp_input.c,v
retrieving revision 1.243
diff -u -p -r1.243 tcp_input.c
--- sys/netinet/tcp_input.c 7 Jun 2006 22:34:01 -0000 1.243
+++ sys/netinet/tcp_input.c 31 Aug 2006 19:53:48 -0000
@@ -70,7 +70,7 @@
*/
/*-
- * Copyright (c) 1997, 1998, 1999, 2001, 2005 The NetBSD Foundation, Inc.
+ * Copyright (c) 1997, 1998, 1999, 2001, 2005, 2006 The NetBSD Foundation, Inc.
* All rights reserved.
*
* This code is derived from software contributed to The NetBSD Foundation
@@ -78,6 +78,8 @@
* Facility, NASA Ames Research Center.
* This code is derived from software contributed to The NetBSD Foundation
* by Charles M. Hannum.
+ * This code is derived from software contributed to The NetBSD Foundation
+ * by Rui Paulo.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
@@ -236,6 +238,8 @@ __KERNEL_RCSID(0, "$NetBSD: tcp_input.c,
#endif
#endif /* FAST_IPSEC*/
+static inline void tcp_congestion_exp(struct tcpcb *);
+
int tcprexmtthresh = 3;
int tcp_log_refused;
@@ -405,6 +409,28 @@ tcpipqent_free(struct ipqent *ipqe)
splx(s);
}
+/*
+ * Halve the congestion window and reduce the
+ * slow start threshold.
+ *
+ * Optionally, mark the packet.
+ */
+static inline void
+tcp_congestion_exp(struct tcpcb *tp)
+{
+ u_int win;
+
+ win = min(tp->snd_wnd, tp->snd_cwnd) / 2 / tp->t_segsz;
+ if (win < 2)
+ win = 2;
+
+ tp->snd_ssthresh = win * tp->t_segsz;
+ tp->snd_recover = tp->snd_max;
+ tp->snd_cwnd = tp->snd_ssthresh;
+ if (TCP_ECN_ALLOWED(tp))
+ tp->t_flags |= TF_ECN_SND_CWR;
+}
+
int
tcp_reass(struct tcpcb *tp, struct tcphdr *th, struct mbuf *m, int *tlen)
{
@@ -977,6 +1003,7 @@ tcp_input(struct mbuf *m, ...)
int af; /* af on the wire */
struct mbuf *tcp_saveti = NULL;
uint32_t ts_rtt;
+ uint8_t iptos;
MCLAIM(m, &tcp_rx_mowner);
va_start(ap, m);
@@ -1033,6 +1060,7 @@ tcp_input(struct mbuf *m, ...)
/* We do the checksum after PCB lookup... */
len = ntohs(ip->ip_len);
tlen = len - toff;
+ iptos = ip->ip_tos;
break;
#endif
#ifdef INET6
@@ -1080,6 +1108,7 @@ tcp_input(struct mbuf *m, ...)
/* We do the checksum after PCB lookup... */
len = m->m_pkthdr.len;
tlen = len - toff;
+ iptos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
break;
#endif
default:
@@ -1587,6 +1616,31 @@ after_listen:
tcp_del_sackholes(tp, th);
}
+ if (TCP_ECN_ALLOWED(tp)) {
+ switch (iptos & IPTOS_ECN_MASK) {
+ case IPTOS_ECN_CE:
+ tp->t_flags |= TF_ECN_SND_ECE;
+ tcpstat.tcps_ecn_ce++;
+ break;
+ case IPTOS_ECN_ECT0:
+ tcpstat.tcps_ecn_ect++;
+ break;
+ case IPTOS_ECN_ECT1:
+ /* XXX: ignore for now -- rpaulo */
+ break;
+ }
+
+ if (tiflags & TH_CWR)
+ tp->t_flags &= ~TF_ECN_SND_ECE;
+
+ /*
+ * Congestion experienced.
+ * Ignore if we are already trying to recover.
+ */
+ if ((tiflags & TH_ECE) && SEQ_GEQ(tp->snd_una, tp->snd_recover))
+ tcp_congestion_exp(tp);
+ }
+
if (opti.ts_present && opti.ts_ecr) {
/*
* Calculate the RTT from the returned time stamp and the
@@ -1617,7 +1671,8 @@ after_listen:
* the socket buffer and note that we need a delayed ack.
*/
if (tp->t_state == TCPS_ESTABLISHED &&
- (tiflags & (TH_SYN|TH_FIN|TH_RST|TH_URG|TH_ACK)) == TH_ACK &&
+ (tiflags & (TH_SYN|TH_FIN|TH_RST|TH_URG|TH_ECE|TH_CWR|TH_ACK))
+ == TH_ACK &&
(!opti.ts_present || TSTMP_GEQ(opti.ts_val, tp->ts_recent)) &&
th->th_seq == tp->rcv_nxt &&
tiwin && tiwin == tp->snd_wnd &&
@@ -1788,6 +1843,8 @@ after_listen:
* Otherwise this is an acceptable SYN segment
* initialize tp->rcv_nxt and tp->irs
* if seg contains ack then advance tp->snd_una
+ * if seg contains a ECE and ECN support is enabled, the stream
+ * is ECN capable.
* if SYN has been acked change to ESTABLISHED else SYN_RCVD state
* arrange for segment to be acked (eventually)
* continue processing rest of data/controls, beginning with URG
@@ -1811,6 +1868,12 @@ after_listen:
if (SEQ_LT(tp->snd_high, tp->snd_una))
tp->snd_high = tp->snd_una;
TCP_TIMER_DISARM(tp, TCPT_REXMT);
+
+ if ((tiflags & TH_ECE) && tcp_do_ecn) {
+ tp->t_flags |= TF_ECN_PERMIT;
+ tcpstat.tcps_ecn_shs++;
+ }
+
}
tp->irs = th->th_seq;
tcp_rcvseqinit(tp);
@@ -2200,6 +2263,9 @@ after_listen:
* to keep a constant cwnd packets in the
* network.
*
+ * When using TCP ECN, notify the peer that
+ * we reduced the cwnd.
+ *
* If we are using TCP/SACK, then enter
* Fast Recovery if the receiver SACKs
* data that is tcprexmtthresh * MSS
@@ -2213,9 +2279,8 @@ after_listen:
(++tp->t_dupacks == tcprexmtthresh ||
TCP_FACK_FASTRECOV(tp))) {
tcp_seq onxt;
- u_int win;
- if (tcp_do_newreno &&
+ if ((tcp_do_newreno || tcp_do_ecn) &&
SEQ_LT(th->th_ack, tp->snd_high)) {
/*
* False fast retransmit after
@@ -2227,12 +2292,7 @@ after_listen:
}
onxt = tp->snd_nxt;
- win = min(tp->snd_wnd, tp->snd_cwnd) /
- 2 / tp->t_segsz;
- if (win < 2)
- win = 2;
- tp->snd_ssthresh = win * tp->t_segsz;
- tp->snd_recover = tp->snd_max;
+ tcp_congestion_exp(tp);
tp->t_partialacks = 0;
TCP_TIMER_DISARM(tp, TCPT_REXMT);
tp->t_rtttime = 0;
@@ -3809,6 +3869,9 @@ syn_cache_get(struct sockaddr *src, stru
if ((sc->sc_flags & SCF_SACK_PERMIT) && tcp_do_sack)
tp->t_flags |= TF_WILL_SACK;
+ if ((sc->sc_flags & SCF_ECN_PERMIT) && tcp_do_ecn)
+ tp->t_flags |= TF_ECN_PERMIT;
+
#ifdef TCP_SIGNATURE
if (sc->sc_flags & SCF_SIGNATURE)
tp->t_flags |= TF_SIGNATURE;
@@ -4101,6 +4164,13 @@ syn_cache_add(struct sockaddr *src, stru
}
if ((tb.t_flags & TF_SACK_PERMIT) && tcp_do_sack)
sc->sc_flags |= SCF_SACK_PERMIT;
+
+ /*
+ * ECN setup packet recieved.
+ */
+ if ((th->th_flags & (TH_ECE|TH_CWR)) && tcp_do_ecn)
+ sc->sc_flags |= SCF_ECN_PERMIT;
+
#ifdef TCP_SIGNATURE
if (tb.t_flags & TF_SIGNATURE)
sc->sc_flags |= SCF_SIGNATURE;
@@ -4128,7 +4198,7 @@ syn_cache_respond(struct syn_cache *sc,
#ifdef INET6
struct ip6_hdr *ip6 = NULL;
#endif
- struct tcpcb *tp;
+ struct tcpcb *tp = NULL;
struct tcphdr *th;
u_int hlen;
struct socket *so;
@@ -4269,6 +4339,55 @@ syn_cache_respond(struct syn_cache *sc,
optp += 4;
}
+ /*
+ * Send ECN SYN-ACK setup packet.
+ * Routes can be asymetric, so, even if we receive a packet
+ * with ECE and CWR set, we must not assume no one will block
+ * the ECE packet we are about to send.
+ */
+ if ((sc->sc_flags & SCF_ECN_PERMIT) && tp &&
+ SEQ_GEQ(tp->snd_nxt, tp->snd_max)) {
+ th->th_flags |= TH_ECE;
+ tcpstat.tcps_ecn_shs++;
+
+ /*
+ * draft-ietf-tcpm-ecnsyn-00.txt
+ *
+ * "[...] a TCP node MAY respond to an ECN-setup
+ * SYN packet by setting ECT in the responding
+ * ECN-setup SYN/ACK packet, indicating to routers
+ * that the SYN/ACK packet is ECN-Capable.
+ * This allows a congested router along the path
+ * to mark the packet instead of dropping the
+ * packet as an indication of congestion."
+ *
+ * "[...] There can be a great benefit in setting
+ * an ECN-capable codepoint in SYN/ACK packets [...]
+ * Congestion is most likely to occur in
+ * the server-to-client direction. As a result,
+ * setting an ECN-capable codepoint in SYN/ACK
+ * packets can reduce the occurence of three-second
+ * retransmit timeouts resulting from the drop
+ * of SYN/ACK packets."
+ *
+ * Page 4 and 6, January 2006.
+ */
+
+ switch (sc->sc_src.sa.sa_family) {
+#ifdef INET
+ case AF_INET:
+ ip->ip_tos |= IPTOS_ECN_ECT0;
+ break;
+#endif
+#ifdef INET6
+ case AF_INET6:
+ ip6->ip6_flow |= htonl(IPTOS_ECN_ECT0 << 20);
+ break;
+#endif
+ }
+ tcpstat.tcps_ecn_ect++;
+ }
+
#ifdef TCP_SIGNATURE
if (sc->sc_flags & SCF_SIGNATURE) {
struct secasvar *sav;
Index: sys/netinet/tcp_output.c
===================================================================
RCS file: /cvsroot/src/sys/netinet/tcp_output.c,v
retrieving revision 1.142
diff -u -p -r1.142 tcp_output.c
--- sys/netinet/tcp_output.c 25 Mar 2006 13:34:35 -0000 1.142
+++ sys/netinet/tcp_output.c 31 Aug 2006 19:53:49 -0000
@@ -70,7 +70,7 @@
*/
/*-
- * Copyright (c) 1997, 1998, 2001, 2005 The NetBSD Foundation, Inc.
+ * Copyright (c) 1997, 1998, 2001, 2005, 2006 The NetBSD Foundation, Inc.
* All rights reserved.
*
* This code is derived from software contributed to The NetBSD Foundation
@@ -78,6 +78,8 @@
* Facility, NASA Ames Research Center.
* This code is derived from software contributed to The NetBSD Foundation
* by Charles M. Hannum.
+ * This code is derived from software contributed to The NetBSD Foundation
+ * by Rui Paulo.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
@@ -1235,6 +1237,58 @@ send:
bcopy(mtod(tp->t_template, caddr_t), mtod(m, caddr_t), iphdrlen);
/*
+ * If we are starting a connection, send ECN setup
+ * SYN packet. If we are on a retransmit, we may
+ * resend those bits a number of times as per
+ * RFC 3168.
+ */
+ if (tp->t_state == TCPS_SYN_SENT && tcp_do_ecn) {
+ if (tp->t_flags & TF_SYN_REXMT) {
+ if (tp->t_ecn_retries--)
+ flags |= TH_ECE|TH_CWR;
+ } else {
+ flags |= TH_ECE|TH_CWR;
+ tp->t_ecn_retries = tcp_ecn_maxretries;
+ }
+ }
+
+ if (TCP_ECN_ALLOWED(tp)) {
+ /*
+ * If the peer has ECN, mark data packets
+ * ECN capable. Ignore pure ack packets, retransmissions
+ * and window probes.
+ */
+ if (len > 0 && SEQ_GEQ(tp->snd_nxt, tp->snd_max) &&
+ !(tp->t_force && len == 1)) {
+ switch (af) {
+#ifdef INET
+ case AF_INET:
+ tp->t_inpcb->inp_ip.ip_tos |= IPTOS_ECN_ECT0;
+ break;
+#endif
+#ifdef INET6
+ case AF_INET6:
+ ip6->ip6_flow |= htonl(IPTOS_ECN_ECT0 << 20);
+ break;
+#endif
+ }
+ tcpstat.tcps_ecn_ect++;
+ }
+
+ /*
+ * Reply with proper ECN notifications.
+ */
+ if (tp->t_flags & TF_ECN_SND_CWR) {
+ flags |= TH_CWR;
+ tp->t_flags &= ~TF_ECN_SND_CWR;
+ }
+ if (tp->t_flags & TF_ECN_SND_ECE) {
+ flags |= TH_ECE;
+ }
+ }
+
+
+ /*
* If we are doing retransmissions, then snd_nxt will
* not reflect the first unsent octet. For ACK only
* packets, we do not want the sequence number of the
Index: sys/netinet/tcp_subr.c
===================================================================
RCS file: /cvsroot/src/sys/netinet/tcp_subr.c,v
retrieving revision 1.198
diff -u -p -r1.198 tcp_subr.c
--- sys/netinet/tcp_subr.c 15 Apr 2006 02:30:39 -0000 1.198
+++ sys/netinet/tcp_subr.c 31 Aug 2006 19:53:49 -0000
@@ -184,6 +184,7 @@ int tcp_do_win_scale = 1; /* RFC1323 win
int tcp_do_timestamps = 1; /* RFC1323 timestamps */
int tcp_do_newreno = 1; /* Use the New Reno algorithms */
int tcp_ack_on_push = 0; /* set to enable immediate ACK-on-PUSH */
+int tcp_do_ecn = 0; /* Explicit Congestion Notification */
#ifndef TCP_INIT_WIN
#define TCP_INIT_WIN 0 /* initial slow start window */
#endif
@@ -204,6 +205,7 @@ int tcp_do_loopback_cksum = 0;
int tcp_sack_tp_maxholes = 32;
int tcp_sack_globalmaxholes = 1024;
int tcp_sack_globalholes = 0;
+int tcp_ecn_maxretries = 1;
/* tcb hash */
Index: sys/netinet/tcp_usrreq.c
===================================================================
RCS file: /cvsroot/src/sys/netinet/tcp_usrreq.c,v
retrieving revision 1.119
diff -u -p -r1.119 tcp_usrreq.c
--- sys/netinet/tcp_usrreq.c 23 Jul 2006 22:06:13 -0000 1.119
+++ sys/netinet/tcp_usrreq.c 31 Aug 2006 19:53:50 -0000
@@ -30,7 +30,7 @@
*/
/*-
- * Copyright (c) 1997, 1998, 2005 The NetBSD Foundation, Inc.
+ * Copyright (c) 1997, 1998, 2005, 2006 The NetBSD Foundation, Inc.
* All rights reserved.
*
* This code is derived from software contributed to The NetBSD Foundation
@@ -38,6 +38,8 @@
* Facility, NASA Ames Research Center.
* This code is derived from software contributed to The NetBSD Foundation
* by Charles M. Hannum.
+ * This code is derived from software contributed to The NetBSD Foundation
+ * by Rui Paulo.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
@@ -1384,7 +1386,7 @@ static void
sysctl_net_inet_tcp_setup2(struct sysctllog **clog, int pf, const char *pfname,
const char *tcpname)
{
- const struct sysctlnode *sack_node;
+ const struct sysctlnode *sack_node, *ecn_node;
#ifdef TCP_DEBUG
extern struct tcp_debug tcp_debug[TCP_NDEBUG];
extern int tcp_debx;
@@ -1475,6 +1477,12 @@ sysctl_net_inet_tcp_setup2(struct sysctl
SYSCTL_DESCR("RFC2018 Selective ACKnowledgement tunables"),
NULL, 0, NULL, 0,
CTL_NET, pf, IPPROTO_TCP, TCPCTL_SACK, CTL_EOL);
+ sysctl_createv(clog, 0, NULL, &ecn_node,
+ CTLFLAG_PERMANENT,
+ CTLTYPE_NODE, "ecn",
+ SYSCTL_DESCR("RFC3168 Explicit Congestion Notification"),
+ NULL, 0, NULL, 0,
+ CTL_NET, pf, IPPROTO_TCP, TCPCTL_ECN, CTL_EOL);
sysctl_createv(clog, 0, NULL, NULL,
CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
CTLTYPE_INT, "win_scale",
@@ -1600,6 +1608,23 @@ sysctl_net_inet_tcp_setup2(struct sysctl
CTL_NET, pf, IPPROTO_TCP, CTL_CREATE,
CTL_EOL);
+ sysctl_createv(clog, 0, NULL, &ecn_node,
+ CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
+ CTLTYPE_INT, "enable",
+ SYSCTL_DESCR("Enable TCP Explicit Congestion "
+ "Notification"),
+ NULL, 0, &tcp_do_ecn, 0,
+ CTL_NET, pf, IPPROTO_TCP, TCPCTL_ECN,
+ CTL_CREATE, CTL_EOL);
+ sysctl_createv(clog, 0, NULL, &ecn_node,
+ CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
+ CTLTYPE_INT, "maxretries",
+ SYSCTL_DESCR("Number of times to retry ECN setup "
+ "before disabling ECN on the connection"),
+ NULL, 0, &tcp_ecn_maxretries, 0,
+ CTL_NET, pf, IPPROTO_TCP, TCPCTL_ECN,
+ CTL_CREATE, CTL_EOL);
+
/* SACK gets it's own little subtree. */
sysctl_createv(clog, 0, NULL, &sack_node,
CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
Index: sys/netinet/tcp_var.h
===================================================================
RCS file: /cvsroot/src/sys/netinet/tcp_var.h,v
retrieving revision 1.136
diff -u -p -r1.136 tcp_var.h
--- sys/netinet/tcp_var.h 22 Jul 2006 17:45:03 -0000 1.136
+++ sys/netinet/tcp_var.h 31 Aug 2006 19:53:50 -0000
@@ -217,6 +217,9 @@ struct tcpcb {
#define TF_REASSEMBLING 0x1000 /* we're busy reassembling */
#define TF_DEAD 0x2000 /* dead and to-be-released */
#define TF_PMTUD_PEND 0x4000 /* Path MTU Discovery pending */
+#define TF_ECN_PERMIT 0x10000 /* other side said is ECN-ready */
+#define TF_ECN_SND_CWR 0x20000 /* ECN CWR in queue */
+#define TF_ECN_SND_ECE 0x40000 /* ECN ECE in queue */
#define TF_SIGNATURE 0x400000 /* require MD5 digests (RFC2385) */
@@ -320,9 +323,16 @@ struct tcpcb {
u_int t_pmtud_nextmtu; /* Advertised Next-Hop MTU from ICMP */
u_short t_pmtud_ip_len; /* IP length from ICMP payload */
u_short t_pmtud_ip_hl; /* IP header length from ICMP payload */
+
+ uint8_t t_ecn_retries; /* # of ECN setup retries */
};
/*
+ * Macros to aid ECN TCP.
+ */
+#define TCP_ECN_ALLOWED(tp) (tp->t_flags & TF_ECN_PERMIT)
+
+/*
* Macros to aid SACK/FACK TCP.
*/
#define TCP_SACK_ENABLED(tp) (tp->t_flags & TF_WILL_SACK)
@@ -483,6 +493,7 @@ struct syn_cache {
#define SCF_TIMESTAMP 0x0002 /* peer will do timestamps */
#define SCF_DEAD 0x0004 /* this entry to be released */
#define SCF_SACK_PERMIT 0x0008 /* peer will do SACK */
+#define SCF_ECN_PERMIT 0x0010 /* peer will do ECN */
#define SCF_SIGNATURE 0x40 /* send MD5 digests */
struct mbuf *sc_ipopts; /* IP options */
@@ -634,6 +645,10 @@ struct tcpstat {
u_quad_t tcps_selfquench; /* # of ENOBUFS we get on output */
u_quad_t tcps_badsig; /* # of drops due to bad signature */
u_quad_t tcps_goodsig; /* # of packets with good signature */
+
+ u_quad_t tcps_ecn_shs; /* # of sucessful ECN handshakes */
+ u_quad_t tcps_ecn_ce; /* # of packets with CE bit */
+ u_quad_t tcps_ecn_ect; /* # of packets with ECT(0) bit */
};
/*
@@ -675,7 +690,8 @@ struct tcpstat {
#define TCPCTL_STATS 30 /* TCP statistics */
#define TCPCTL_DEBUG 31 /* TCP debug sockets */
#define TCPCTL_DEBX 32 /* # of tcp debug sockets */
-#define TCPCTL_MAXID 33
+#define TCPCTL_ECN 33 /* RFC3168 ECN */
+#define TCPCTL_MAXID 34
#define TCPCTL_NAMES { \
{ 0, 0 }, \
@@ -711,6 +727,7 @@ struct tcpstat {
{ "stats", CTLTYPE_STRUCT }, \
{ "debug", CTLTYPE_STRUCT }, \
{ "debx", CTLTYPE_INT }, \
+ { "ecn", CTLTYPE_NODE }, \
}
#ifdef _KERNEL
@@ -733,6 +750,8 @@ extern int tcp_ack_on_push; /* ACK immed
extern int tcp_syn_cache_limit; /* max entries for compressed state engine */
extern int tcp_syn_bucket_limit;/* max entries per hash bucket */
extern int tcp_log_refused; /* log refused connections */
+extern int tcp_do_ecn; /* TCP ECN enabled/disabled? */
+extern int tcp_ecn_maxretries; /* Max ECN setup retries */
extern int tcp_sack_tp_maxholes; /* Max holes per connection. */
extern int tcp_sack_globalmaxholes; /* Max holes per system. */
extern int tcp_sack_globalholes; /* Number of holes present. */
Index: usr.bin/netstat/inet.c
===================================================================
RCS file: /cvsroot/src/usr.bin/netstat/inet.c,v
retrieving revision 1.75
diff -u -p -r1.75 inet.c
--- usr.bin/netstat/inet.c 17 Aug 2006 01:42:57 -0000 1.75
+++ usr.bin/netstat/inet.c 31 Aug 2006 19:53:51 -0000
@@ -391,6 +391,9 @@ tcp_stats(off, name)
p(tcps_badsig, "\t%llu packet%s with bad signature\n");
p(tcps_goodsig, "\t%llu packet%s with good signature\n");
+ p(tcps_ecn_shs, "\t%llu sucessful ECN handshake%s\n");
+ p(tcps_ecn_ce, "\t%llu packet%s with ECN CE bit\n");
+ p(tcps_ecn_ect, "\t%llu packet%s ECN ECT(0) bit\n");
#undef p
#undef ps
#undef p2
--Apple-Mail-3-505784705
Content-Transfer-Encoding: 7bit
Content-Type: text/plain;
charset=US-ASCII;
format=flowed
-- Rui Paulo
--Apple-Mail-3-505784705--