tcp_timer.c revision 1.16 
1/*	$NetBSD: tcp_timer.c,v 1.16 1996/09/09 14:51:22 mycroft 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;
76
77	s = splsoftnet();
78	inp = tcbtable.inpt_queue.cqh_first;
79	if (inp)						/* XXX */
80	for (; inp != (struct inpcb *)&tcbtable.inpt_queue;
81	    inp = inp->inp_queue.cqe_next) {
82		if ((tp = (struct tcpcb *)inp->inp_ppcb) &&
83		    (tp->t_flags & TF_DELACK)) {
84			tp->t_flags &= ~TF_DELACK;
85			tp->t_flags |= TF_ACKNOW;
86			tcpstat.tcps_delack++;
87			(void) tcp_output(tp);
88		}
89	}
90	splx(s);
91}
92
93/*
94 * Tcp protocol timeout routine called every 500 ms.
95 * Updates the timers in all active tcb's and
96 * causes finite state machine actions if timers expire.
97 */
98void
99tcp_slowtimo()
100{
101	register struct inpcb *inp, *ninp;
102	register struct tcpcb *tp;
103	int s;
104	register long i;
105
106	s = splsoftnet();
107	tcp_maxidle = TCPTV_KEEPCNT * tcp_keepintvl;
108	/*
109	 * Search through tcb's and update active timers.
110	 */
111	inp = tcbtable.inpt_queue.cqh_first;
112	if (inp == (struct inpcb *)0) {				/* XXX */
113		splx(s);
114		return;
115	}
116	for (; inp != (struct inpcb *)&tcbtable.inpt_queue; inp = ninp) {
117		ninp = inp->inp_queue.cqe_next;
118		tp = intotcpcb(inp);
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				(void) tcp_usrreq(tp->t_inpcb->inp_socket,
124				    PRU_SLOWTIMO, (struct mbuf *)0,
125				    (struct mbuf *)i, (struct mbuf *)0,
126				    (struct proc *)0);
127				/* XXX NOT MP SAFE */
128				if ((ninp == (void *)&tcbtable.inpt_queue &&
129				    tcbtable.inpt_queue.cqh_last != inp) ||
130				    ninp->inp_queue.cqe_prev != inp)
131					goto tpgone;
132			}
133		}
134		tp->t_idle++;
135		if (tp->t_rtt)
136			tp->t_rtt++;
137tpgone:
138		;
139	}
140	tcp_iss += TCP_ISSINCR/PR_SLOWHZ;		/* increment iss */
141#ifdef TCP_COMPAT_42
142	if ((int)tcp_iss < 0)
143		tcp_iss = 0;				/* XXX */
144#endif
145	tcp_now++;					/* for timestamps */
146	splx(s);
147}
148#ifndef TUBA_INCLUDE
149
150/*
151 * Cancel all timers for TCP tp.
152 */
153void
154tcp_canceltimers(tp)
155	struct tcpcb *tp;
156{
157	register int i;
158
159	for (i = 0; i < TCPT_NTIMERS; i++)
160		tp->t_timer[i] = 0;
161}
162
163int	tcp_backoff[TCP_MAXRXTSHIFT + 1] =
164    { 1, 2, 4, 8, 16, 32, 64, 64, 64, 64, 64, 64, 64 };
165
166/*
167 * TCP timer processing.
168 */
169struct tcpcb *
170tcp_timers(tp, timer)
171	register struct tcpcb *tp;
172	int timer;
173{
174	register int rexmt;
175
176	switch (timer) {
177
178	/*
179	 * 2 MSL timeout in shutdown went off.  If we're closed but
180	 * still waiting for peer to close and connection has been idle
181	 * too long, or if 2MSL time is up from TIME_WAIT, delete connection
182	 * control block.  Otherwise, check again in a bit.
183	 */
184	case TCPT_2MSL:
185		if (tp->t_state != TCPS_TIME_WAIT &&
186		    tp->t_idle <= tcp_maxidle)
187			tp->t_timer[TCPT_2MSL] = tcp_keepintvl;
188		else
189			tp = tcp_close(tp);
190		break;
191
192	/*
193	 * Retransmission timer went off.  Message has not
194	 * been acked within retransmit interval.  Back off
195	 * to a longer retransmit interval and retransmit one segment.
196	 */
197	case TCPT_REXMT:
198		if (++tp->t_rxtshift > TCP_MAXRXTSHIFT) {
199			tp->t_rxtshift = TCP_MAXRXTSHIFT;
200			tcpstat.tcps_timeoutdrop++;
201			tp = tcp_drop(tp, tp->t_softerror ?
202			    tp->t_softerror : ETIMEDOUT);
203			break;
204		}
205		tcpstat.tcps_rexmttimeo++;
206		rexmt = TCP_REXMTVAL(tp) * tcp_backoff[tp->t_rxtshift];
207		TCPT_RANGESET(tp->t_rxtcur, rexmt, tp->t_rttmin,
208		    TCPTV_REXMTMAX);
209		tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
210		/*
211		 * If losing, let the lower level know and try for
212		 * a better route.  Also, if we backed off this far,
213		 * our srtt estimate is probably bogus.  Clobber it
214		 * so we'll take the next rtt measurement as our srtt;
215		 * move the current srtt into rttvar to keep the current
216		 * retransmit times until then.
217		 */
218		if (tp->t_rxtshift > TCP_MAXRXTSHIFT / 4) {
219			in_losing(tp->t_inpcb);
220			tp->t_rttvar += (tp->t_srtt >> TCP_RTT_SHIFT);
221			tp->t_srtt = 0;
222		}
223		tp->snd_nxt = tp->snd_una;
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 (TCPS_HAVEESTABLISHED(tp->t_state) == 0)
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