tcp_timer.c revision 1.10 
1/*	$NetBSD: tcp_timer.c,v 1.10 1995/04/13 06:36:49 cgd Exp $	*/
2
3/*
4 * Copyright (c) 1982, 1986, 1988, 1990, 1993
5 *	The Regents of the University of California.  All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 *    notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 *    notice, this list of conditions and the following disclaimer in the
14 *    documentation and/or other materials provided with the distribution.
15 * 3. All advertising materials mentioning features or use of this software
16 *    must display the following acknowledgement:
17 *	This product includes software developed by the University of
18 *	California, Berkeley and its contributors.
19 * 4. Neither the name of the University nor the names of its contributors
20 *    may be used to endorse or promote products derived from this software
21 *    without specific prior written permission.
22 *
23 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
24 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33 * SUCH DAMAGE.
34 *
35 *	@(#)tcp_timer.c	8.1 (Berkeley) 6/10/93
36 */
37
38#ifndef TUBA_INCLUDE
39#include <sys/param.h>
40#include <sys/systm.h>
41#include <sys/malloc.h>
42#include <sys/mbuf.h>
43#include <sys/socket.h>
44#include <sys/socketvar.h>
45#include <sys/protosw.h>
46#include <sys/errno.h>
47
48#include <net/if.h>
49#include <net/route.h>
50
51#include <netinet/in.h>
52#include <netinet/in_systm.h>
53#include <netinet/ip.h>
54#include <netinet/in_pcb.h>
55#include <netinet/ip_var.h>
56#include <netinet/tcp.h>
57#include <netinet/tcp_fsm.h>
58#include <netinet/tcp_seq.h>
59#include <netinet/tcp_timer.h>
60#include <netinet/tcp_var.h>
61#include <netinet/tcpip.h>
62
63int	tcp_keepidle = TCPTV_KEEP_IDLE;
64int	tcp_keepintvl = TCPTV_KEEPINTVL;
65int	tcp_maxidle;
66#endif /* TUBA_INCLUDE */
67/*
68 * Fast timeout routine for processing delayed acks
69 */
70void
71tcp_fasttimo()
72{
73	register struct inpcb *inp;
74	register struct tcpcb *tp;
75	int s = splnet();
76
77	inp = tcb.inp_next;
78	if (inp)
79	for (; inp != &tcb; inp = inp->inp_next)
80		if ((tp = (struct tcpcb *)inp->inp_ppcb) &&
81		    (tp->t_flags & TF_DELACK)) {
82			tp->t_flags &= ~TF_DELACK;
83			tp->t_flags |= TF_ACKNOW;
84			tcpstat.tcps_delack++;
85			(void) tcp_output(tp);
86		}
87	splx(s);
88}
89
90/*
91 * Tcp protocol timeout routine called every 500 ms.
92 * Updates the timers in all active tcb's and
93 * causes finite state machine actions if timers expire.
94 */
95void
96tcp_slowtimo()
97{
98	register struct inpcb *ip, *ipnxt;
99	register struct tcpcb *tp;
100	int s = splnet();
101	register long i;
102
103	tcp_maxidle = TCPTV_KEEPCNT * tcp_keepintvl;
104	/*
105	 * Search through tcb's and update active timers.
106	 */
107	ip = tcb.inp_next;
108	if (ip == 0) {
109		splx(s);
110		return;
111	}
112	for (; ip != &tcb; ip = ipnxt) {
113		ipnxt = ip->inp_next;
114		tp = intotcpcb(ip);
115		if (tp == 0)
116			continue;
117		for (i = 0; i < TCPT_NTIMERS; i++) {
118			if (tp->t_timer[i] && --tp->t_timer[i] == 0) {
119				(void) tcp_usrreq(tp->t_inpcb->inp_socket,
120				    PRU_SLOWTIMO, (struct mbuf *)0,
121				    (struct mbuf *)i, (struct mbuf *)0);
122				if (ipnxt->inp_prev != ip)
123					goto tpgone;
124			}
125		}
126		tp->t_idle++;
127		if (tp->t_rtt)
128			tp->t_rtt++;
129tpgone:
130		;
131	}
132	tcp_iss += TCP_ISSINCR/PR_SLOWHZ;		/* increment iss */
133#ifdef TCP_COMPAT_42
134	if ((int)tcp_iss < 0)
135		tcp_iss = 0;				/* XXX */
136#endif
137	tcp_now++;					/* for timestamps */
138	splx(s);
139}
140#ifndef TUBA_INCLUDE
141
142/*
143 * Cancel all timers for TCP tp.
144 */
145void
146tcp_canceltimers(tp)
147	struct tcpcb *tp;
148{
149	register int i;
150
151	for (i = 0; i < TCPT_NTIMERS; i++)
152		tp->t_timer[i] = 0;
153}
154
155int	tcp_backoff[TCP_MAXRXTSHIFT + 1] =
156    { 1, 2, 4, 8, 16, 32, 64, 64, 64, 64, 64, 64, 64 };
157
158/*
159 * TCP timer processing.
160 */
161struct tcpcb *
162tcp_timers(tp, timer)
163	register struct tcpcb *tp;
164	int timer;
165{
166	register int rexmt;
167
168	switch (timer) {
169
170	/*
171	 * 2 MSL timeout in shutdown went off.  If we're closed but
172	 * still waiting for peer to close and connection has been idle
173	 * too long, or if 2MSL time is up from TIME_WAIT, delete connection
174	 * control block.  Otherwise, check again in a bit.
175	 */
176	case TCPT_2MSL:
177		if (tp->t_state != TCPS_TIME_WAIT &&
178		    tp->t_idle <= tcp_maxidle)
179			tp->t_timer[TCPT_2MSL] = tcp_keepintvl;
180		else
181			tp = tcp_close(tp);
182		break;
183
184	/*
185	 * Retransmission timer went off.  Message has not
186	 * been acked within retransmit interval.  Back off
187	 * to a longer retransmit interval and retransmit one segment.
188	 */
189	case TCPT_REXMT:
190		if (++tp->t_rxtshift > TCP_MAXRXTSHIFT) {
191			tp->t_rxtshift = TCP_MAXRXTSHIFT;
192			tcpstat.tcps_timeoutdrop++;
193			tp = tcp_drop(tp, tp->t_softerror ?
194			    tp->t_softerror : ETIMEDOUT);
195			break;
196		}
197		tcpstat.tcps_rexmttimeo++;
198		rexmt = TCP_REXMTVAL(tp) * tcp_backoff[tp->t_rxtshift];
199		TCPT_RANGESET(tp->t_rxtcur, rexmt,
200		    tp->t_rttmin, TCPTV_REXMTMAX);
201		tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
202		/*
203		 * If losing, let the lower level know and try for
204		 * a better route.  Also, if we backed off this far,
205		 * our srtt estimate is probably bogus.  Clobber it
206		 * so we'll take the next rtt measurement as our srtt;
207		 * move the current srtt into rttvar to keep the current
208		 * retransmit times until then.
209		 */
210		if (tp->t_rxtshift > TCP_MAXRXTSHIFT / 4) {
211			in_losing(tp->t_inpcb);
212			tp->t_rttvar += (tp->t_srtt >> TCP_RTT_SHIFT);
213			tp->t_srtt = 0;
214		}
215		tp->snd_nxt = tp->snd_una;
216		/*
217		 * If timing a segment in this window, stop the timer.
218		 */
219		tp->t_rtt = 0;
220		/*
221		 * Close the congestion window down to one segment
222		 * (we'll open it by one segment for each ack we get).
223		 * Since we probably have a window's worth of unacked
224		 * data accumulated, this "slow start" keeps us from
225		 * dumping all that data as back-to-back packets (which
226		 * might overwhelm an intermediate gateway).
227		 *
228		 * There are two phases to the opening: Initially we
229		 * open by one mss on each ack.  This makes the window
230		 * size increase exponentially with time.  If the
231		 * window is larger than the path can handle, this
232		 * exponential growth results in dropped packet(s)
233		 * almost immediately.  To get more time between
234		 * drops but still "push" the network to take advantage
235		 * of improving conditions, we switch from exponential
236		 * to linear window opening at some threshhold size.
237		 * For a threshhold, we use half the current window
238		 * size, truncated to a multiple of the mss.
239		 *
240		 * (the minimum cwnd that will give us exponential
241		 * growth is 2 mss.  We don't allow the threshhold
242		 * to go below this.)
243		 */
244		{
245		u_int win = min(tp->snd_wnd, tp->snd_cwnd) / 2 / tp->t_maxseg;
246		if (win < 2)
247			win = 2;
248		tp->snd_cwnd = tp->t_maxseg;
249		tp->snd_ssthresh = win * tp->t_maxseg;
250		tp->t_dupacks = 0;
251		}
252		(void) tcp_output(tp);
253		break;
254
255	/*
256	 * Persistance timer into zero window.
257	 * Force a byte to be output, if possible.
258	 */
259	case TCPT_PERSIST:
260		tcpstat.tcps_persisttimeo++;
261		tcp_setpersist(tp);
262		tp->t_force = 1;
263		(void) tcp_output(tp);
264		tp->t_force = 0;
265		break;
266
267	/*
268	 * Keep-alive timer went off; send something
269	 * or drop connection if idle for too long.
270	 */
271	case TCPT_KEEP:
272		tcpstat.tcps_keeptimeo++;
273		if (TCPS_HAVEESTABLISHED(tp->t_state) == 0)
274			goto dropit;
275		if (tp->t_inpcb->inp_socket->so_options & SO_KEEPALIVE &&
276		    tp->t_state <= TCPS_CLOSE_WAIT) {
277		    	if (tp->t_idle >= tcp_keepidle + tcp_maxidle)
278				goto dropit;
279			/*
280			 * Send a packet designed to force a response
281			 * if the peer is up and reachable:
282			 * either an ACK if the connection is still alive,
283			 * or an RST if the peer has closed the connection
284			 * due to timeout or reboot.
285			 * Using sequence number tp->snd_una-1
286			 * causes the transmitted zero-length segment
287			 * to lie outside the receive window;
288			 * by the protocol spec, this requires the
289			 * correspondent TCP to respond.
290			 */
291			tcpstat.tcps_keepprobe++;
292#ifdef TCP_COMPAT_42
293			/*
294			 * The keepalive packet must have nonzero length
295			 * to get a 4.2 host to respond.
296			 */
297			tcp_respond(tp, tp->t_template, (struct mbuf *)NULL,
298			    tp->rcv_nxt - 1, tp->snd_una - 1, 0);
299#else
300			tcp_respond(tp, tp->t_template, (struct mbuf *)NULL,
301			    tp->rcv_nxt, tp->snd_una - 1, 0);
302#endif
303			tp->t_timer[TCPT_KEEP] = tcp_keepintvl;
304		} else
305			tp->t_timer[TCPT_KEEP] = tcp_keepidle;
306		break;
307	dropit:
308		tcpstat.tcps_keepdrops++;
309		tp = tcp_drop(tp, ETIMEDOUT);
310		break;
311	}
312	return (tp);
313}
314#endif /* TUBA_INCLUDE */
315