tcp_timer.c revision 6283 
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 1.2 1994/08/02 07:49:10 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
47#include <net/if.h>
48#include <net/route.h>
49
50#include <netinet/in.h>
51#include <netinet/in_systm.h>
52#include <netinet/ip.h>
53#include <netinet/in_pcb.h>
54#include <netinet/ip_var.h>
55#include <netinet/tcp.h>
56#include <netinet/tcp_fsm.h>
57#include <netinet/tcp_seq.h>
58#include <netinet/tcp_timer.h>
59#include <netinet/tcp_var.h>
60#include <netinet/tcpip.h>
61
62int	tcp_keepidle = TCPTV_KEEP_IDLE;
63int	tcp_keepintvl = TCPTV_KEEPINTVL;
64int	tcp_maxidle;
65#endif /* TUBA_INCLUDE */
66/*
67 * Fast timeout routine for processing delayed acks
68 */
69void
70tcp_fasttimo()
71{
72	register struct inpcb *inp;
73	register struct tcpcb *tp;
74	int s = splnet();
75
76	inp = tcb.inp_next;
77	if (inp)
78	for (; inp != &tcb; inp = inp->inp_next)
79		if ((tp = (struct tcpcb *)inp->inp_ppcb) &&
80		    (tp->t_flags & TF_DELACK)) {
81			tp->t_flags &= ~TF_DELACK;
82			tp->t_flags |= TF_ACKNOW;
83			tcpstat.tcps_delack++;
84			(void) tcp_output(tp);
85		}
86	splx(s);
87}
88
89/*
90 * Tcp protocol timeout routine called every 500 ms.
91 * Updates the timers in all active tcb's and
92 * causes finite state machine actions if timers expire.
93 */
94void
95tcp_slowtimo()
96{
97	register struct inpcb *ip, *ipnxt;
98	register struct tcpcb *tp;
99	int s = splnet();
100	register int i;
101
102	tcp_maxidle = TCPTV_KEEPCNT * tcp_keepintvl;
103	/*
104	 * Search through tcb's and update active timers.
105	 */
106	ip = tcb.inp_next;
107	if (ip == 0) {
108		splx(s);
109		return;
110	}
111	for (; ip != &tcb; ip = ipnxt) {
112		ipnxt = ip->inp_next;
113		tp = intotcpcb(ip);
114		if (tp == 0)
115			continue;
116		for (i = 0; i < TCPT_NTIMERS; i++) {
117			if (tp->t_timer[i] && --tp->t_timer[i] == 0) {
118				(void) tcp_usrreq(tp->t_inpcb->inp_socket,
119				    PRU_SLOWTIMO, (struct mbuf *)0,
120				    (struct mbuf *)i, (struct mbuf *)0);
121				if (ipnxt->inp_prev != ip)
122					goto tpgone;
123			}
124		}
125		tp->t_idle++;
126#ifdef TTCP
127		tp->t_duration++;
128#endif
129		if (tp->t_rtt)
130			tp->t_rtt++;
131tpgone:
132		;
133	}
134	tcp_iss += TCP_ISSINCR/PR_SLOWHZ;		/* increment iss */
135#ifdef TCP_COMPAT_42
136	if ((int)tcp_iss < 0)
137		tcp_iss = 0;				/* XXX */
138#endif
139	tcp_now++;					/* for timestamps */
140	splx(s);
141}
142#ifndef TUBA_INCLUDE
143
144/*
145 * Cancel all timers for TCP tp.
146 */
147void
148tcp_canceltimers(tp)
149	struct tcpcb *tp;
150{
151	register int i;
152
153	for (i = 0; i < TCPT_NTIMERS; i++)
154		tp->t_timer[i] = 0;
155}
156
157int	tcp_backoff[TCP_MAXRXTSHIFT + 1] =
158    { 1, 2, 4, 8, 16, 32, 64, 64, 64, 64, 64, 64, 64 };
159
160/*
161 * TCP timer processing.
162 */
163struct tcpcb *
164tcp_timers(tp, timer)
165	register struct tcpcb *tp;
166	int timer;
167{
168	register int rexmt;
169
170	switch (timer) {
171
172	/*
173	 * 2 MSL timeout in shutdown went off.  If we're closed but
174	 * still waiting for peer to close and connection has been idle
175	 * too long, or if 2MSL time is up from TIME_WAIT, delete connection
176	 * control block.  Otherwise, check again in a bit.
177	 */
178	case TCPT_2MSL:
179		if (tp->t_state != TCPS_TIME_WAIT &&
180		    tp->t_idle <= tcp_maxidle)
181			tp->t_timer[TCPT_2MSL] = tcp_keepintvl;
182		else
183			tp = tcp_close(tp);
184		break;
185
186	/*
187	 * Retransmission timer went off.  Message has not
188	 * been acked within retransmit interval.  Back off
189	 * to a longer retransmit interval and retransmit one segment.
190	 */
191	case TCPT_REXMT:
192		if (++tp->t_rxtshift > TCP_MAXRXTSHIFT) {
193			tp->t_rxtshift = TCP_MAXRXTSHIFT;
194			tcpstat.tcps_timeoutdrop++;
195			tp = tcp_drop(tp, tp->t_softerror ?
196			    tp->t_softerror : ETIMEDOUT);
197			break;
198		}
199		tcpstat.tcps_rexmttimeo++;
200		rexmt = TCP_REXMTVAL(tp) * tcp_backoff[tp->t_rxtshift];
201		TCPT_RANGESET(tp->t_rxtcur, rexmt,
202		    tp->t_rttmin, TCPTV_REXMTMAX);
203		tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
204		/*
205		 * If losing, let the lower level know and try for
206		 * a better route.  Also, if we backed off this far,
207		 * our srtt estimate is probably bogus.  Clobber it
208		 * so we'll take the next rtt measurement as our srtt;
209		 * move the current srtt into rttvar to keep the current
210		 * retransmit times until then.
211		 */
212		if (tp->t_rxtshift > TCP_MAXRXTSHIFT / 4) {
213			in_losing(tp->t_inpcb);
214			tp->t_rttvar += (tp->t_srtt >> TCP_RTT_SHIFT);
215			tp->t_srtt = 0;
216		}
217		tp->snd_nxt = tp->snd_una;
218#ifdef TTCP
219		/*
220		 * Force a segment to be sent.
221		 */
222		tp->t_flags |= TF_ACKNOW;
223#endif
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