tcp_timer.c revision 5d8f37ad78fc66901af50c762029a501561f3b23 
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. Neither the name of the University nor the names of its contributors
14 *    may be used to endorse or promote products derived from this software
15 *    without specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27 * SUCH DAMAGE.
28 *
29 *	@(#)tcp_timer.c	8.1 (Berkeley) 6/10/93
30 * tcp_timer.c,v 1.2 1994/08/02 07:49:10 davidg Exp
31 */
32
33#include <slirp.h>
34
35#ifdef LOG_ENABLED
36struct   tcpstat tcpstat;        /* tcp statistics */
37#endif
38
39u_int32_t        tcp_now;                /* for RFC 1323 timestamps */
40
41static struct tcpcb *tcp_timers(register struct tcpcb *tp, int timer);
42
43/*
44 * Fast timeout routine for processing delayed acks
45 */
46void
47tcp_fasttimo(void)
48{
49	register struct socket *so;
50	register struct tcpcb *tp;
51
52	DEBUG_CALL("tcp_fasttimo");
53
54	so = tcb.so_next;
55	if (so)
56	for (; so != &tcb; so = so->so_next)
57		if ((tp = (struct tcpcb *)so->so_tcpcb) &&
58		    (tp->t_flags & TF_DELACK)) {
59			tp->t_flags &= ~TF_DELACK;
60			tp->t_flags |= TF_ACKNOW;
61			STAT(tcpstat.tcps_delack++);
62			(void) tcp_output(tp);
63		}
64}
65
66/*
67 * Tcp protocol timeout routine called every 500 ms.
68 * Updates the timers in all active tcb's and
69 * causes finite state machine actions if timers expire.
70 */
71void
72tcp_slowtimo(void)
73{
74	register struct socket *ip, *ipnxt;
75	register struct tcpcb *tp;
76	register int i;
77
78	DEBUG_CALL("tcp_slowtimo");
79
80	/*
81	 * Search through tcb's and update active timers.
82	 */
83	ip = tcb.so_next;
84	if (ip == 0)
85	   return;
86	for (; ip != &tcb; ip = ipnxt) {
87		ipnxt = ip->so_next;
88		tp = sototcpcb(ip);
89		if (tp == 0)
90			continue;
91		for (i = 0; i < TCPT_NTIMERS; i++) {
92			if (tp->t_timer[i] && --tp->t_timer[i] == 0) {
93				tcp_timers(tp,i);
94				if (ipnxt->so_prev != ip)
95					goto tpgone;
96			}
97		}
98		tp->t_idle++;
99		if (tp->t_rtt)
100		   tp->t_rtt++;
101tpgone:
102		;
103	}
104	tcp_iss += TCP_ISSINCR/PR_SLOWHZ;		/* increment iss */
105#ifdef TCP_COMPAT_42
106	if ((int)tcp_iss < 0)
107		tcp_iss = 0;				/* XXX */
108#endif
109	tcp_now++;					/* for timestamps */
110}
111
112/*
113 * Cancel all timers for TCP tp.
114 */
115void
116tcp_canceltimers(struct tcpcb *tp)
117{
118	register int i;
119
120	for (i = 0; i < TCPT_NTIMERS; i++)
121		tp->t_timer[i] = 0;
122}
123
124const int tcp_backoff[TCP_MAXRXTSHIFT + 1] =
125   { 1, 2, 4, 8, 16, 32, 64, 64, 64, 64, 64, 64, 64 };
126
127/*
128 * TCP timer processing.
129 */
130static struct tcpcb *
131tcp_timers(register struct tcpcb *tp, int timer)
132{
133	register int rexmt;
134
135	DEBUG_CALL("tcp_timers");
136
137	switch (timer) {
138
139	/*
140	 * 2 MSL timeout in shutdown went off.  If we're closed but
141	 * still waiting for peer to close and connection has been idle
142	 * too long, or if 2MSL time is up from TIME_WAIT, delete connection
143	 * control block.  Otherwise, check again in a bit.
144	 */
145	case TCPT_2MSL:
146		if (tp->t_state != TCPS_TIME_WAIT &&
147		    tp->t_idle <= TCP_MAXIDLE)
148			tp->t_timer[TCPT_2MSL] = TCPTV_KEEPINTVL;
149		else
150			tp = tcp_close(tp);
151		break;
152
153	/*
154	 * Retransmission timer went off.  Message has not
155	 * been acked within retransmit interval.  Back off
156	 * to a longer retransmit interval and retransmit one segment.
157	 */
158	case TCPT_REXMT:
159
160		/*
161		 * XXXXX If a packet has timed out, then remove all the queued
162		 * packets for that session.
163		 */
164
165		if (++tp->t_rxtshift > TCP_MAXRXTSHIFT) {
166			/*
167			 * This is a hack to suit our terminal server here at the uni of canberra
168			 * since they have trouble with zeroes... It usually lets them through
169			 * unharmed, but under some conditions, it'll eat the zeros.  If we
170			 * keep retransmitting it, it'll keep eating the zeroes, so we keep
171			 * retransmitting, and eventually the connection dies...
172			 * (this only happens on incoming data)
173			 *
174			 * So, if we were gonna drop the connection from too many retransmits,
175			 * don't... instead halve the t_maxseg, which might break up the NULLs and
176			 * let them through
177			 *
178			 * *sigh*
179			 */
180
181			tp->t_maxseg >>= 1;
182			if (tp->t_maxseg < 32) {
183				/*
184				 * We tried our best, now the connection must die!
185				 */
186				tp->t_rxtshift = TCP_MAXRXTSHIFT;
187				STAT(tcpstat.tcps_timeoutdrop++);
188				tp = tcp_drop(tp, tp->t_softerror);
189				/* tp->t_softerror : ETIMEDOUT); */ /* XXX */
190				return (tp); /* XXX */
191			}
192
193			/*
194			 * Set rxtshift to 6, which is still at the maximum
195			 * backoff time
196			 */
197			tp->t_rxtshift = 6;
198		}
199		STAT(tcpstat.tcps_rexmttimeo++);
200		rexmt = TCP_REXMTVAL(tp) * tcp_backoff[tp->t_rxtshift];
201		TCPT_RANGESET(tp->t_rxtcur, rexmt,
202		    (short)tp->t_rttmin, TCPTV_REXMTMAX); /* XXX */
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		/*
219		 * If timing a segment in this window, stop the timer.
220		 */
221		tp->t_rtt = 0;
222		/*
223		 * Close the congestion window down to one segment
224		 * (we'll open it by one segment for each ack we get).
225		 * Since we probably have a window's worth of unacked
226		 * data accumulated, this "slow start" keeps us from
227		 * dumping all that data as back-to-back packets (which
228		 * might overwhelm an intermediate gateway).
229		 *
230		 * There are two phases to the opening: Initially we
231		 * open by one mss on each ack.  This makes the window
232		 * size increase exponentially with time.  If the
233		 * window is larger than the path can handle, this
234		 * exponential growth results in dropped packet(s)
235		 * almost immediately.  To get more time between
236		 * drops but still "push" the network to take advantage
237		 * of improving conditions, we switch from exponential
238		 * to linear window opening at some threshold size.
239		 * For a threshold, we use half the current window
240		 * size, truncated to a multiple of the mss.
241		 *
242		 * (the minimum cwnd that will give us exponential
243		 * growth is 2 mss.  We don't allow the threshold
244		 * to go below this.)
245		 */
246		{
247		u_int win = min(tp->snd_wnd, tp->snd_cwnd) / 2 / tp->t_maxseg;
248		if (win < 2)
249			win = 2;
250		tp->snd_cwnd = tp->t_maxseg;
251		tp->snd_ssthresh = win * tp->t_maxseg;
252		tp->t_dupacks = 0;
253		}
254		(void) tcp_output(tp);
255		break;
256
257	/*
258	 * Persistence timer into zero window.
259	 * Force a byte to be output, if possible.
260	 */
261	case TCPT_PERSIST:
262		STAT(tcpstat.tcps_persisttimeo++);
263		tcp_setpersist(tp);
264		tp->t_force = 1;
265		(void) tcp_output(tp);
266		tp->t_force = 0;
267		break;
268
269	/*
270	 * Keep-alive timer went off; send something
271	 * or drop connection if idle for too long.
272	 */
273	case TCPT_KEEP:
274		STAT(tcpstat.tcps_keeptimeo++);
275		if (tp->t_state < TCPS_ESTABLISHED)
276			goto dropit;
277
278/*		if (tp->t_socket->so_options & SO_KEEPALIVE && */
279		if ((SO_OPTIONS) && tp->t_state <= TCPS_CLOSE_WAIT) {
280		    	if (tp->t_idle >= TCPTV_KEEP_IDLE + TCP_MAXIDLE)
281				goto dropit;
282			/*
283			 * Send a packet designed to force a response
284			 * if the peer is up and reachable:
285			 * either an ACK if the connection is still alive,
286			 * or an RST if the peer has closed the connection
287			 * due to timeout or reboot.
288			 * Using sequence number tp->snd_una-1
289			 * causes the transmitted zero-length segment
290			 * to lie outside the receive window;
291			 * by the protocol spec, this requires the
292			 * correspondent TCP to respond.
293			 */
294			STAT(tcpstat.tcps_keepprobe++);
295#ifdef TCP_COMPAT_42
296			/*
297			 * The keepalive packet must have nonzero length
298			 * to get a 4.2 host to respond.
299			 */
300			tcp_respond(tp, &tp->t_template, (struct mbuf *)NULL,
301			    tp->rcv_nxt - 1, tp->snd_una - 1, 0);
302#else
303			tcp_respond(tp, &tp->t_template, (struct mbuf *)NULL,
304			    tp->rcv_nxt, tp->snd_una - 1, 0);
305#endif
306			tp->t_timer[TCPT_KEEP] = TCPTV_KEEPINTVL;
307		} else
308			tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_IDLE;
309		break;
310
311	dropit:
312		STAT(tcpstat.tcps_keepdrops++);
313		tp = tcp_drop(tp, 0); /* ETIMEDOUT); */
314		break;
315	}
316
317	return (tp);
318}
319