tcp_timer.c revision 7770 
1/*
2 * Copyright (c) 1982, 1986, 1988, 1990, 1993
3 *	The Regents of the University of California.  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 *	@(#)tcp_timer.c	8.1 (Berkeley) 6/10/93
34 * $Id: tcp_timer.c,v 1.5 1995/04/09 01:29:27 davidg Exp $
35 */
36
37#ifndef TUBA_INCLUDE
38#include <sys/param.h>
39#include <sys/systm.h>
40#include <sys/malloc.h>
41#include <sys/mbuf.h>
42#include <sys/socket.h>
43#include <sys/socketvar.h>
44#include <sys/protosw.h>
45#include <sys/errno.h>
46#include <sys/queue.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;
76
77	s = splnet();
78
79	for (inp = tcb.lh_first; inp != NULL; inp = inp->inp_list.le_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	}
88	splx(s);
89}
90
91/*
92 * Tcp protocol timeout routine called every 500 ms.
93 * Updates the timers in all active tcb's and
94 * causes finite state machine actions if timers expire.
95 */
96void
97tcp_slowtimo()
98{
99	register struct inpcb *ip, *ipnxt;
100	register struct tcpcb *tp;
101	register int i;
102	int s;
103
104	s = splnet();
105
106	tcp_maxidle = TCPTV_KEEPCNT * tcp_keepintvl;
107
108	ip = tcb.lh_first;
109	if (ip == NULL) {
110		splx(s);
111		return;
112	}
113	/*
114	 * Search through tcb's and update active timers.
115	 */
116	for (; ip != NULL; ip = ipnxt) {
117		ipnxt = ip->inp_list.le_next;
118		tp = intotcpcb(ip);
119		if (tp == 0)
120			continue;
121		for (i = 0; i < TCPT_NTIMERS; i++) {
122			if (tp->t_timer[i] && --tp->t_timer[i] == 0) {
123				if (tcp_usrreq(tp->t_inpcb->inp_socket,
124				    PRU_SLOWTIMO, (struct mbuf *)0,
125				    (struct mbuf *)i, (struct mbuf *)0) == NULL)
126					goto tpgone;
127			}
128		}
129		tp->t_idle++;
130		tp->t_duration++;
131		if (tp->t_rtt)
132			tp->t_rtt++;
133tpgone:
134		;
135	}
136	tcp_iss += TCP_ISSINCR/PR_SLOWHZ;		/* increment iss */
137#ifdef TCP_COMPAT_42
138	if ((int)tcp_iss < 0)
139		tcp_iss = 0;				/* XXX */
140#endif
141	tcp_now++;					/* for timestamps */
142	splx(s);
143}
144#ifndef TUBA_INCLUDE
145
146/*
147 * Cancel all timers for TCP tp.
148 */
149void
150tcp_canceltimers(tp)
151	struct tcpcb *tp;
152{
153	register int i;
154
155	for (i = 0; i < TCPT_NTIMERS; i++)
156		tp->t_timer[i] = 0;
157}
158
159int	tcp_backoff[TCP_MAXRXTSHIFT + 1] =
160    { 1, 2, 4, 8, 16, 32, 64, 64, 64, 64, 64, 64, 64 };
161
162/*
163 * TCP timer processing.
164 */
165struct tcpcb *
166tcp_timers(tp, timer)
167	register struct tcpcb *tp;
168	int timer;
169{
170	register int rexmt;
171
172	switch (timer) {
173
174	/*
175	 * 2 MSL timeout in shutdown went off.  If we're closed but
176	 * still waiting for peer to close and connection has been idle
177	 * too long, or if 2MSL time is up from TIME_WAIT, delete connection
178	 * control block.  Otherwise, check again in a bit.
179	 */
180	case TCPT_2MSL:
181		if (tp->t_state != TCPS_TIME_WAIT &&
182		    tp->t_idle <= tcp_maxidle)
183			tp->t_timer[TCPT_2MSL] = tcp_keepintvl;
184		else
185			tp = tcp_close(tp);
186		break;
187
188	/*
189	 * Retransmission timer went off.  Message has not
190	 * been acked within retransmit interval.  Back off
191	 * to a longer retransmit interval and retransmit one segment.
192	 */
193	case TCPT_REXMT:
194		if (++tp->t_rxtshift > TCP_MAXRXTSHIFT) {
195			tp->t_rxtshift = TCP_MAXRXTSHIFT;
196			tcpstat.tcps_timeoutdrop++;
197			tp = tcp_drop(tp, tp->t_softerror ?
198			    tp->t_softerror : ETIMEDOUT);
199			break;
200		}
201		tcpstat.tcps_rexmttimeo++;
202		rexmt = TCP_REXMTVAL(tp) * tcp_backoff[tp->t_rxtshift];
203		TCPT_RANGESET(tp->t_rxtcur, rexmt,
204		    tp->t_rttmin, TCPTV_REXMTMAX);
205		tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
206		/*
207		 * If losing, let the lower level know and try for
208		 * a better route.  Also, if we backed off this far,
209		 * our srtt estimate is probably bogus.  Clobber it
210		 * so we'll take the next rtt measurement as our srtt;
211		 * move the current srtt into rttvar to keep the current
212		 * retransmit times until then.
213		 */
214		if (tp->t_rxtshift > TCP_MAXRXTSHIFT / 4) {
215			in_losing(tp->t_inpcb);
216			tp->t_rttvar += (tp->t_srtt >> TCP_RTT_SHIFT);
217			tp->t_srtt = 0;
218		}
219		tp->snd_nxt = tp->snd_una;
220		/*
221		 * Force a segment to be sent.
222		 */
223		tp->t_flags |= TF_ACKNOW;
224		/*
225		 * If timing a segment in this window, stop the timer.
226		 */
227		tp->t_rtt = 0;
228		/*
229		 * Close the congestion window down to one segment
230		 * (we'll open it by one segment for each ack we get).
231		 * Since we probably have a window's worth of unacked
232		 * data accumulated, this "slow start" keeps us from
233		 * dumping all that data as back-to-back packets (which
234		 * might overwhelm an intermediate gateway).
235		 *
236		 * There are two phases to the opening: Initially we
237		 * open by one mss on each ack.  This makes the window
238		 * size increase exponentially with time.  If the
239		 * window is larger than the path can handle, this
240		 * exponential growth results in dropped packet(s)
241		 * almost immediately.  To get more time between
242		 * drops but still "push" the network to take advantage
243		 * of improving conditions, we switch from exponential
244		 * to linear window opening at some threshhold size.
245		 * For a threshhold, we use half the current window
246		 * size, truncated to a multiple of the mss.
247		 *
248		 * (the minimum cwnd that will give us exponential
249		 * growth is 2 mss.  We don't allow the threshhold
250		 * to go below this.)
251		 */
252		{
253		u_int win = min(tp->snd_wnd, tp->snd_cwnd) / 2 / tp->t_maxseg;
254		if (win < 2)
255			win = 2;
256		tp->snd_cwnd = tp->t_maxseg;
257		tp->snd_ssthresh = win * tp->t_maxseg;
258		tp->t_dupacks = 0;
259		}
260		(void) tcp_output(tp);
261		break;
262
263	/*
264	 * Persistance timer into zero window.
265	 * Force a byte to be output, if possible.
266	 */
267	case TCPT_PERSIST:
268		tcpstat.tcps_persisttimeo++;
269		tcp_setpersist(tp);
270		tp->t_force = 1;
271		(void) tcp_output(tp);
272		tp->t_force = 0;
273		break;
274
275	/*
276	 * Keep-alive timer went off; send something
277	 * or drop connection if idle for too long.
278	 */
279	case TCPT_KEEP:
280		tcpstat.tcps_keeptimeo++;
281		if (tp->t_state < TCPS_ESTABLISHED)
282			goto dropit;
283		if (tp->t_inpcb->inp_socket->so_options & SO_KEEPALIVE &&
284		    tp->t_state <= TCPS_CLOSE_WAIT) {
285		    	if (tp->t_idle >= tcp_keepidle + tcp_maxidle)
286				goto dropit;
287			/*
288			 * Send a packet designed to force a response
289			 * if the peer is up and reachable:
290			 * either an ACK if the connection is still alive,
291			 * or an RST if the peer has closed the connection
292			 * due to timeout or reboot.
293			 * Using sequence number tp->snd_una-1
294			 * causes the transmitted zero-length segment
295			 * to lie outside the receive window;
296			 * by the protocol spec, this requires the
297			 * correspondent TCP to respond.
298			 */
299			tcpstat.tcps_keepprobe++;
300#ifdef TCP_COMPAT_42
301			/*
302			 * The keepalive packet must have nonzero length
303			 * to get a 4.2 host to respond.
304			 */
305			tcp_respond(tp, tp->t_template, (struct mbuf *)NULL,
306			    tp->rcv_nxt - 1, tp->snd_una - 1, 0);
307#else
308			tcp_respond(tp, tp->t_template, (struct mbuf *)NULL,
309			    tp->rcv_nxt, tp->snd_una - 1, 0);
310#endif
311			tp->t_timer[TCPT_KEEP] = tcp_keepintvl;
312		} else
313			tp->t_timer[TCPT_KEEP] = tcp_keepidle;
314		break;
315	dropit:
316		tcpstat.tcps_keepdrops++;
317		tp = tcp_drop(tp, ETIMEDOUT);
318		break;
319	}
320	return (tp);
321}
322#endif /* TUBA_INCLUDE */
323