tcp_timer.c revision 1.4 
1/*
2 * Copyright (c) 1982, 1986, 1988, 1990 Regents of the University of California.
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 *    notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 *    notice, this list of conditions and the following disclaimer in the
12 *    documentation and/or other materials provided with the distribution.
13 * 3. All advertising materials mentioning features or use of this software
14 *    must display the following acknowledgement:
15 *	This product includes software developed by the University of
16 *	California, Berkeley and its contributors.
17 * 4. Neither the name of the University nor the names of its contributors
18 *    may be used to endorse or promote products derived from this software
19 *    without specific prior written permission.
20 *
21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * SUCH DAMAGE.
32 *
33 *	from: @(#)tcp_timer.c	7.18 (Berkeley) 6/28/90
34 *	$Id: tcp_timer.c,v 1.4 1993/12/18 00:42:09 mycroft Exp $
35 */
36
37#include <sys/param.h>
38#include <sys/systm.h>
39#include <sys/malloc.h>
40#include <sys/mbuf.h>
41#include <sys/socket.h>
42#include <sys/socketvar.h>
43#include <sys/protosw.h>
44#include <sys/errno.h>
45
46#include <net/if.h>
47#include <net/route.h>
48
49#include <netinet/in.h>
50#include <netinet/in_systm.h>
51#include <netinet/ip.h>
52#include <netinet/in_pcb.h>
53#include <netinet/ip_var.h>
54#include <netinet/tcp.h>
55#include <netinet/tcp_fsm.h>
56#include <netinet/tcp_seq.h>
57#include <netinet/tcp_timer.h>
58#include <netinet/tcp_var.h>
59#include <netinet/tcpip.h>
60
61int	tcp_keepidle = TCPTV_KEEP_IDLE;
62int	tcp_keepintvl = TCPTV_KEEPINTVL;
63int	tcp_maxidle;
64/*
65 * Fast timeout routine for processing delayed acks
66 */
67tcp_fasttimo()
68{
69	register struct inpcb *inp;
70	register struct tcpcb *tp;
71	int s = splnet();
72
73	inp = tcb.inp_next;
74	if (inp)
75	for (; inp != &tcb; inp = inp->inp_next)
76		if ((tp = (struct tcpcb *)inp->inp_ppcb) &&
77		    (tp->t_flags & TF_DELACK)) {
78			tp->t_flags &= ~TF_DELACK;
79			tp->t_flags |= TF_ACKNOW;
80			tcpstat.tcps_delack++;
81			(void) tcp_output(tp);
82		}
83	splx(s);
84}
85
86/*
87 * Tcp protocol timeout routine called every 500 ms.
88 * Updates the timers in all active tcb's and
89 * causes finite state machine actions if timers expire.
90 */
91tcp_slowtimo()
92{
93	register struct inpcb *ip, *ipnxt;
94	register struct tcpcb *tp;
95	int s = splnet();
96	register int i;
97
98	tcp_maxidle = TCPTV_KEEPCNT * tcp_keepintvl;
99	/*
100	 * Search through tcb's and update active timers.
101	 */
102	ip = tcb.inp_next;
103	if (ip == 0) {
104		splx(s);
105		return;
106	}
107	for (; ip != &tcb; ip = ipnxt) {
108		ipnxt = ip->inp_next;
109		tp = intotcpcb(ip);
110		if (tp == 0)
111			continue;
112		for (i = 0; i < TCPT_NTIMERS; i++) {
113			if (tp->t_timer[i] && --tp->t_timer[i] == 0) {
114				(void) tcp_usrreq(tp->t_inpcb->inp_socket,
115				    PRU_SLOWTIMO, (struct mbuf *)0,
116				    (struct mbuf *)i, (struct mbuf *)0);
117				if (ipnxt->inp_prev != ip)
118					goto tpgone;
119			}
120		}
121		tp->t_idle++;
122		if (tp->t_rtt)
123			tp->t_rtt++;
124tpgone:
125		;
126	}
127	tcp_iss += TCP_ISSINCR/PR_SLOWHZ;		/* increment iss */
128#ifdef TCP_COMPAT_42
129	if ((int)tcp_iss < 0)
130		tcp_iss = 0;				/* XXX */
131#endif
132	splx(s);
133}
134
135/*
136 * Cancel all timers for TCP tp.
137 */
138tcp_canceltimers(tp)
139	struct tcpcb *tp;
140{
141	register int i;
142
143	for (i = 0; i < TCPT_NTIMERS; i++)
144		tp->t_timer[i] = 0;
145}
146
147int	tcp_backoff[TCP_MAXRXTSHIFT + 1] =
148    { 1, 2, 4, 8, 16, 32, 64, 64, 64, 64, 64, 64, 64 };
149
150/*
151 * TCP timer processing.
152 */
153struct tcpcb *
154tcp_timers(tp, timer)
155	register struct tcpcb *tp;
156	int timer;
157{
158	register int rexmt;
159
160	switch (timer) {
161
162	/*
163	 * 2 MSL timeout in shutdown went off.  If we're closed but
164	 * still waiting for peer to close and connection has been idle
165	 * too long, or if 2MSL time is up from TIME_WAIT, delete connection
166	 * control block.  Otherwise, check again in a bit.
167	 */
168	case TCPT_2MSL:
169		if (tp->t_state != TCPS_TIME_WAIT &&
170		    tp->t_idle <= tcp_maxidle)
171			tp->t_timer[TCPT_2MSL] = tcp_keepintvl;
172		else
173			tp = tcp_close(tp);
174		break;
175
176	/*
177	 * Retransmission timer went off.  Message has not
178	 * been acked within retransmit interval.  Back off
179	 * to a longer retransmit interval and retransmit one segment.
180	 */
181	case TCPT_REXMT:
182		if (++tp->t_rxtshift > TCP_MAXRXTSHIFT) {
183			tp->t_rxtshift = TCP_MAXRXTSHIFT;
184			tcpstat.tcps_timeoutdrop++;
185			tp = tcp_drop(tp, tp->t_softerror ?
186			    tp->t_softerror : ETIMEDOUT);
187			break;
188		}
189		tcpstat.tcps_rexmttimeo++;
190		rexmt = TCP_REXMTVAL(tp) * tcp_backoff[tp->t_rxtshift];
191		TCPT_RANGESET(tp->t_rxtcur, rexmt,
192		    tp->t_rttmin, TCPTV_REXMTMAX);
193		tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
194		/*
195		 * If losing, let the lower level know and try for
196		 * a better route.  Also, if we backed off this far,
197		 * our srtt estimate is probably bogus.  Clobber it
198		 * so we'll take the next rtt measurement as our srtt;
199		 * move the current srtt into rttvar to keep the current
200		 * retransmit times until then.
201		 */
202		if (tp->t_rxtshift > TCP_MAXRXTSHIFT / 4) {
203			in_losing(tp->t_inpcb);
204			tp->t_rttvar += (tp->t_srtt >> TCP_RTT_SHIFT);
205			tp->t_srtt = 0;
206		}
207		tp->snd_nxt = tp->snd_una;
208		/*
209		 * If timing a segment in this window, stop the timer.
210		 */
211		tp->t_rtt = 0;
212		/*
213		 * Close the congestion window down to one segment
214		 * (we'll open it by one segment for each ack we get).
215		 * Since we probably have a window's worth of unacked
216		 * data accumulated, this "slow start" keeps us from
217		 * dumping all that data as back-to-back packets (which
218		 * might overwhelm an intermediate gateway).
219		 *
220		 * There are two phases to the opening: Initially we
221		 * open by one mss on each ack.  This makes the window
222		 * size increase exponentially with time.  If the
223		 * window is larger than the path can handle, this
224		 * exponential growth results in dropped packet(s)
225		 * almost immediately.  To get more time between
226		 * drops but still "push" the network to take advantage
227		 * of improving conditions, we switch from exponential
228		 * to linear window opening at some threshhold size.
229		 * For a threshhold, we use half the current window
230		 * size, truncated to a multiple of the mss.
231		 *
232		 * (the minimum cwnd that will give us exponential
233		 * growth is 2 mss.  We don't allow the threshhold
234		 * to go below this.)
235		 */
236		{
237		u_int win = min(tp->snd_wnd, tp->snd_cwnd) / 2 / tp->t_maxseg;
238		if (win < 2)
239			win = 2;
240		tp->snd_cwnd = tp->t_maxseg;
241		tp->snd_ssthresh = win * tp->t_maxseg;
242		tp->t_dupacks = 0;
243		}
244		(void) tcp_output(tp);
245		break;
246
247	/*
248	 * Persistance timer into zero window.
249	 * Force a byte to be output, if possible.
250	 */
251	case TCPT_PERSIST:
252		tcpstat.tcps_persisttimeo++;
253		tcp_setpersist(tp);
254		tp->t_force = 1;
255		(void) tcp_output(tp);
256		tp->t_force = 0;
257		break;
258
259	/*
260	 * Keep-alive timer went off; send something
261	 * or drop connection if idle for too long.
262	 */
263	case TCPT_KEEP:
264		tcpstat.tcps_keeptimeo++;
265		if (tp->t_state < TCPS_ESTABLISHED)
266			goto dropit;
267		if (tp->t_inpcb->inp_socket->so_options & SO_KEEPALIVE &&
268		    tp->t_state <= TCPS_CLOSE_WAIT) {
269		    	if (tp->t_idle >= tcp_keepidle + tcp_maxidle)
270				goto dropit;
271			/*
272			 * Send a packet designed to force a response
273			 * if the peer is up and reachable:
274			 * either an ACK if the connection is still alive,
275			 * or an RST if the peer has closed the connection
276			 * due to timeout or reboot.
277			 * Using sequence number tp->snd_una-1
278			 * causes the transmitted zero-length segment
279			 * to lie outside the receive window;
280			 * by the protocol spec, this requires the
281			 * correspondent TCP to respond.
282			 */
283			tcpstat.tcps_keepprobe++;
284#ifdef TCP_COMPAT_42
285			/*
286			 * The keepalive packet must have nonzero length
287			 * to get a 4.2 host to respond.
288			 */
289			tcp_respond(tp, tp->t_template, (struct mbuf *)NULL,
290			    tp->rcv_nxt - 1, tp->snd_una - 1, 0);
291#else
292			tcp_respond(tp, tp->t_template, (struct mbuf *)NULL,
293			    tp->rcv_nxt, tp->snd_una - 1, 0);
294#endif
295			tp->t_timer[TCPT_KEEP] = tcp_keepintvl;
296		} else
297			tp->t_timer[TCPT_KEEP] = tcp_keepidle;
298		break;
299	dropit:
300		tcpstat.tcps_keepdrops++;
301		tp = tcp_drop(tp, ETIMEDOUT);
302		break;
303	}
304	return (tp);
305}
306