tcp_timer.c revision 1.7 
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 *	from: @(#)tcp_timer.c	8.1 (Berkeley) 6/10/93
34 *	$Id: tcp_timer.c,v 1.7 1994/05/13 06:06:46 mycroft 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		if (tp->t_rtt)
127			tp->t_rtt++;
128tpgone:
129		;
130	}
131	tcp_iss += TCP_ISSINCR/PR_SLOWHZ;		/* increment iss */
132#ifdef TCP_COMPAT_42
133	if ((int)tcp_iss < 0)
134		tcp_iss = 0;				/* XXX */
135#endif
136	tcp_now++;					/* for timestamps */
137	splx(s);
138}
139#ifndef TUBA_INCLUDE
140
141/*
142 * Cancel all timers for TCP tp.
143 */
144void
145tcp_canceltimers(tp)
146	struct tcpcb *tp;
147{
148	register int i;
149
150	for (i = 0; i < TCPT_NTIMERS; i++)
151		tp->t_timer[i] = 0;
152}
153
154int	tcp_backoff[TCP_MAXRXTSHIFT + 1] =
155    { 1, 2, 4, 8, 16, 32, 64, 64, 64, 64, 64, 64, 64 };
156
157/*
158 * TCP timer processing.
159 */
160struct tcpcb *
161tcp_timers(tp, timer)
162	register struct tcpcb *tp;
163	int timer;
164{
165	register int rexmt;
166
167	switch (timer) {
168
169	/*
170	 * 2 MSL timeout in shutdown went off.  If we're closed but
171	 * still waiting for peer to close and connection has been idle
172	 * too long, or if 2MSL time is up from TIME_WAIT, delete connection
173	 * control block.  Otherwise, check again in a bit.
174	 */
175	case TCPT_2MSL:
176		if (tp->t_state != TCPS_TIME_WAIT &&
177		    tp->t_idle <= tcp_maxidle)
178			tp->t_timer[TCPT_2MSL] = tcp_keepintvl;
179		else
180			tp = tcp_close(tp);
181		break;
182
183	/*
184	 * Retransmission timer went off.  Message has not
185	 * been acked within retransmit interval.  Back off
186	 * to a longer retransmit interval and retransmit one segment.
187	 */
188	case TCPT_REXMT:
189		if (++tp->t_rxtshift > TCP_MAXRXTSHIFT) {
190			tp->t_rxtshift = TCP_MAXRXTSHIFT;
191			tcpstat.tcps_timeoutdrop++;
192			tp = tcp_drop(tp, tp->t_softerror ?
193			    tp->t_softerror : ETIMEDOUT);
194			break;
195		}
196		tcpstat.tcps_rexmttimeo++;
197		rexmt = TCP_REXMTVAL(tp) * tcp_backoff[tp->t_rxtshift];
198		TCPT_RANGESET(tp->t_rxtcur, rexmt,
199		    tp->t_rttmin, TCPTV_REXMTMAX);
200		tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
201		/*
202		 * If losing, let the lower level know and try for
203		 * a better route.  Also, if we backed off this far,
204		 * our srtt estimate is probably bogus.  Clobber it
205		 * so we'll take the next rtt measurement as our srtt;
206		 * move the current srtt into rttvar to keep the current
207		 * retransmit times until then.
208		 */
209		if (tp->t_rxtshift > TCP_MAXRXTSHIFT / 4) {
210			in_losing(tp->t_inpcb);
211			tp->t_rttvar += (tp->t_srtt >> TCP_RTT_SHIFT);
212			tp->t_srtt = 0;
213		}
214		tp->snd_nxt = tp->snd_una;
215		/*
216		 * If timing a segment in this window, stop the timer.
217		 */
218		tp->t_rtt = 0;
219		/*
220		 * Close the congestion window down to one segment
221		 * (we'll open it by one segment for each ack we get).
222		 * Since we probably have a window's worth of unacked
223		 * data accumulated, this "slow start" keeps us from
224		 * dumping all that data as back-to-back packets (which
225		 * might overwhelm an intermediate gateway).
226		 *
227		 * There are two phases to the opening: Initially we
228		 * open by one mss on each ack.  This makes the window
229		 * size increase exponentially with time.  If the
230		 * window is larger than the path can handle, this
231		 * exponential growth results in dropped packet(s)
232		 * almost immediately.  To get more time between
233		 * drops but still "push" the network to take advantage
234		 * of improving conditions, we switch from exponential
235		 * to linear window opening at some threshhold size.
236		 * For a threshhold, we use half the current window
237		 * size, truncated to a multiple of the mss.
238		 *
239		 * (the minimum cwnd that will give us exponential
240		 * growth is 2 mss.  We don't allow the threshhold
241		 * to go below this.)
242		 */
243		{
244		u_int win = min(tp->snd_wnd, tp->snd_cwnd) / 2 / tp->t_maxseg;
245		if (win < 2)
246			win = 2;
247		tp->snd_cwnd = tp->t_maxseg;
248		tp->snd_ssthresh = win * tp->t_maxseg;
249		tp->t_dupacks = 0;
250		}
251		(void) tcp_output(tp);
252		break;
253
254	/*
255	 * Persistance timer into zero window.
256	 * Force a byte to be output, if possible.
257	 */
258	case TCPT_PERSIST:
259		tcpstat.tcps_persisttimeo++;
260		tcp_setpersist(tp);
261		tp->t_force = 1;
262		(void) tcp_output(tp);
263		tp->t_force = 0;
264		break;
265
266	/*
267	 * Keep-alive timer went off; send something
268	 * or drop connection if idle for too long.
269	 */
270	case TCPT_KEEP:
271		tcpstat.tcps_keeptimeo++;
272		if (tp->t_state < TCPS_ESTABLISHED)
273			goto dropit;
274		if (tp->t_inpcb->inp_socket->so_options & SO_KEEPALIVE &&
275		    tp->t_state <= TCPS_CLOSE_WAIT) {
276		    	if (tp->t_idle >= tcp_keepidle + tcp_maxidle)
277				goto dropit;
278			/*
279			 * Send a packet designed to force a response
280			 * if the peer is up and reachable:
281			 * either an ACK if the connection is still alive,
282			 * or an RST if the peer has closed the connection
283			 * due to timeout or reboot.
284			 * Using sequence number tp->snd_una-1
285			 * causes the transmitted zero-length segment
286			 * to lie outside the receive window;
287			 * by the protocol spec, this requires the
288			 * correspondent TCP to respond.
289			 */
290			tcpstat.tcps_keepprobe++;
291#ifdef TCP_COMPAT_42
292			/*
293			 * The keepalive packet must have nonzero length
294			 * to get a 4.2 host to respond.
295			 */
296			tcp_respond(tp, tp->t_template, (struct mbuf *)NULL,
297			    tp->rcv_nxt - 1, tp->snd_una - 1, 0);
298#else
299			tcp_respond(tp, tp->t_template, (struct mbuf *)NULL,
300			    tp->rcv_nxt, tp->snd_una - 1, 0);
301#endif
302			tp->t_timer[TCPT_KEEP] = tcp_keepintvl;
303		} else
304			tp->t_timer[TCPT_KEEP] = tcp_keepidle;
305		break;
306	dropit:
307		tcpstat.tcps_keepdrops++;
308		tp = tcp_drop(tp, ETIMEDOUT);
309		break;
310	}
311	return (tp);
312}
313#endif /* TUBA_INCLUDE */
314