Commit | Line | Data |
---|---|---|
05df3b37 WJ |
1 | /* |
2 | * Copyright (c) 1982, 1986, 1988, 1990 Regents of the University of California. | |
3 | * 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 7.18 (Berkeley) 6/28/90 | |
34 | */ | |
35 | ||
36 | #include "param.h" | |
37 | #include "systm.h" | |
38 | #include "malloc.h" | |
39 | #include "mbuf.h" | |
40 | #include "socket.h" | |
41 | #include "socketvar.h" | |
42 | #include "protosw.h" | |
43 | #include "errno.h" | |
44 | ||
45 | #include "../net/if.h" | |
46 | #include "../net/route.h" | |
47 | ||
48 | #include "in.h" | |
49 | #include "in_systm.h" | |
50 | #include "ip.h" | |
51 | #include "in_pcb.h" | |
52 | #include "ip_var.h" | |
53 | #include "tcp.h" | |
54 | #include "tcp_fsm.h" | |
55 | #include "tcp_seq.h" | |
56 | #include "tcp_timer.h" | |
57 | #include "tcp_var.h" | |
58 | #include "tcpip.h" | |
59 | ||
60 | int tcp_keepidle = TCPTV_KEEP_IDLE; | |
61 | int tcp_keepintvl = TCPTV_KEEPINTVL; | |
62 | int tcp_maxidle; | |
63 | /* | |
64 | * Fast timeout routine for processing delayed acks | |
65 | */ | |
66 | tcp_fasttimo() | |
67 | { | |
68 | register struct inpcb *inp; | |
69 | register struct tcpcb *tp; | |
70 | int s = splnet(); | |
71 | ||
72 | inp = tcb.inp_next; | |
73 | if (inp) | |
74 | for (; inp != &tcb; inp = inp->inp_next) | |
75 | if ((tp = (struct tcpcb *)inp->inp_ppcb) && | |
76 | (tp->t_flags & TF_DELACK)) { | |
77 | tp->t_flags &= ~TF_DELACK; | |
78 | tp->t_flags |= TF_ACKNOW; | |
79 | tcpstat.tcps_delack++; | |
80 | (void) tcp_output(tp); | |
81 | } | |
82 | splx(s); | |
83 | } | |
84 | ||
85 | /* | |
86 | * Tcp protocol timeout routine called every 500 ms. | |
87 | * Updates the timers in all active tcb's and | |
88 | * causes finite state machine actions if timers expire. | |
89 | */ | |
90 | tcp_slowtimo() | |
91 | { | |
92 | register struct inpcb *ip, *ipnxt; | |
93 | register struct tcpcb *tp; | |
94 | int s = splnet(); | |
95 | register int i; | |
96 | ||
97 | tcp_maxidle = TCPTV_KEEPCNT * tcp_keepintvl; | |
98 | /* | |
99 | * Search through tcb's and update active timers. | |
100 | */ | |
101 | ip = tcb.inp_next; | |
102 | if (ip == 0) { | |
103 | splx(s); | |
104 | return; | |
105 | } | |
106 | for (; ip != &tcb; ip = ipnxt) { | |
107 | ipnxt = ip->inp_next; | |
108 | tp = intotcpcb(ip); | |
109 | if (tp == 0) | |
110 | continue; | |
111 | for (i = 0; i < TCPT_NTIMERS; i++) { | |
112 | if (tp->t_timer[i] && --tp->t_timer[i] == 0) { | |
113 | (void) tcp_usrreq(tp->t_inpcb->inp_socket, | |
114 | PRU_SLOWTIMO, (struct mbuf *)0, | |
115 | (struct mbuf *)i, (struct mbuf *)0); | |
116 | if (ipnxt->inp_prev != ip) | |
117 | goto tpgone; | |
118 | } | |
119 | } | |
120 | tp->t_idle++; | |
121 | if (tp->t_rtt) | |
122 | tp->t_rtt++; | |
123 | tpgone: | |
124 | ; | |
125 | } | |
126 | tcp_iss += TCP_ISSINCR/PR_SLOWHZ; /* increment iss */ | |
127 | #ifdef TCP_COMPAT_42 | |
128 | if ((int)tcp_iss < 0) | |
129 | tcp_iss = 0; /* XXX */ | |
130 | #endif | |
131 | splx(s); | |
132 | } | |
133 | ||
134 | /* | |
135 | * Cancel all timers for TCP tp. | |
136 | */ | |
137 | tcp_canceltimers(tp) | |
138 | struct tcpcb *tp; | |
139 | { | |
140 | register int i; | |
141 | ||
142 | for (i = 0; i < TCPT_NTIMERS; i++) | |
143 | tp->t_timer[i] = 0; | |
144 | } | |
145 | ||
146 | int tcp_backoff[TCP_MAXRXTSHIFT + 1] = | |
147 | { 1, 2, 4, 8, 16, 32, 64, 64, 64, 64, 64, 64, 64 }; | |
148 | ||
149 | /* | |
150 | * TCP timer processing. | |
151 | */ | |
152 | struct tcpcb * | |
153 | tcp_timers(tp, timer) | |
154 | register struct tcpcb *tp; | |
155 | int timer; | |
156 | { | |
157 | register int rexmt; | |
158 | ||
159 | switch (timer) { | |
160 | ||
161 | /* | |
162 | * 2 MSL timeout in shutdown went off. If we're closed but | |
163 | * still waiting for peer to close and connection has been idle | |
164 | * too long, or if 2MSL time is up from TIME_WAIT, delete connection | |
165 | * control block. Otherwise, check again in a bit. | |
166 | */ | |
167 | case TCPT_2MSL: | |
168 | if (tp->t_state != TCPS_TIME_WAIT && | |
169 | tp->t_idle <= tcp_maxidle) | |
170 | tp->t_timer[TCPT_2MSL] = tcp_keepintvl; | |
171 | else | |
172 | tp = tcp_close(tp); | |
173 | break; | |
174 | ||
175 | /* | |
176 | * Retransmission timer went off. Message has not | |
177 | * been acked within retransmit interval. Back off | |
178 | * to a longer retransmit interval and retransmit one segment. | |
179 | */ | |
180 | case TCPT_REXMT: | |
181 | if (++tp->t_rxtshift > TCP_MAXRXTSHIFT) { | |
182 | tp->t_rxtshift = TCP_MAXRXTSHIFT; | |
183 | tcpstat.tcps_timeoutdrop++; | |
184 | tp = tcp_drop(tp, tp->t_softerror ? | |
185 | tp->t_softerror : ETIMEDOUT); | |
186 | break; | |
187 | } | |
188 | tcpstat.tcps_rexmttimeo++; | |
189 | rexmt = TCP_REXMTVAL(tp) * tcp_backoff[tp->t_rxtshift]; | |
190 | TCPT_RANGESET(tp->t_rxtcur, rexmt, | |
191 | tp->t_rttmin, TCPTV_REXMTMAX); | |
192 | tp->t_timer[TCPT_REXMT] = tp->t_rxtcur; | |
193 | /* | |
194 | * If losing, let the lower level know and try for | |
195 | * a better route. Also, if we backed off this far, | |
196 | * our srtt estimate is probably bogus. Clobber it | |
197 | * so we'll take the next rtt measurement as our srtt; | |
198 | * move the current srtt into rttvar to keep the current | |
199 | * retransmit times until then. | |
200 | */ | |
201 | if (tp->t_rxtshift > TCP_MAXRXTSHIFT / 4) { | |
202 | in_losing(tp->t_inpcb); | |
203 | tp->t_rttvar += (tp->t_srtt >> TCP_RTT_SHIFT); | |
204 | tp->t_srtt = 0; | |
205 | } | |
206 | tp->snd_nxt = tp->snd_una; | |
207 | /* | |
208 | * If timing a segment in this window, stop the timer. | |
209 | */ | |
210 | tp->t_rtt = 0; | |
211 | /* | |
212 | * Close the congestion window down to one segment | |
213 | * (we'll open it by one segment for each ack we get). | |
214 | * Since we probably have a window's worth of unacked | |
215 | * data accumulated, this "slow start" keeps us from | |
216 | * dumping all that data as back-to-back packets (which | |
217 | * might overwhelm an intermediate gateway). | |
218 | * | |
219 | * There are two phases to the opening: Initially we | |
220 | * open by one mss on each ack. This makes the window | |
221 | * size increase exponentially with time. If the | |
222 | * window is larger than the path can handle, this | |
223 | * exponential growth results in dropped packet(s) | |
224 | * almost immediately. To get more time between | |
225 | * drops but still "push" the network to take advantage | |
226 | * of improving conditions, we switch from exponential | |
227 | * to linear window opening at some threshhold size. | |
228 | * For a threshhold, we use half the current window | |
229 | * size, truncated to a multiple of the mss. | |
230 | * | |
231 | * (the minimum cwnd that will give us exponential | |
232 | * growth is 2 mss. We don't allow the threshhold | |
233 | * to go below this.) | |
234 | */ | |
235 | { | |
236 | u_int win = min(tp->snd_wnd, tp->snd_cwnd) / 2 / tp->t_maxseg; | |
237 | if (win < 2) | |
238 | win = 2; | |
239 | tp->snd_cwnd = tp->t_maxseg; | |
240 | tp->snd_ssthresh = win * tp->t_maxseg; | |
241 | tp->t_dupacks = 0; | |
242 | } | |
243 | (void) tcp_output(tp); | |
244 | break; | |
245 | ||
246 | /* | |
247 | * Persistance timer into zero window. | |
248 | * Force a byte to be output, if possible. | |
249 | */ | |
250 | case TCPT_PERSIST: | |
251 | tcpstat.tcps_persisttimeo++; | |
252 | tcp_setpersist(tp); | |
253 | tp->t_force = 1; | |
254 | (void) tcp_output(tp); | |
255 | tp->t_force = 0; | |
256 | break; | |
257 | ||
258 | /* | |
259 | * Keep-alive timer went off; send something | |
260 | * or drop connection if idle for too long. | |
261 | */ | |
262 | case TCPT_KEEP: | |
263 | tcpstat.tcps_keeptimeo++; | |
264 | if (tp->t_state < TCPS_ESTABLISHED) | |
265 | goto dropit; | |
266 | if (tp->t_inpcb->inp_socket->so_options & SO_KEEPALIVE && | |
267 | tp->t_state <= TCPS_CLOSE_WAIT) { | |
268 | if (tp->t_idle >= tcp_keepidle + tcp_maxidle) | |
269 | goto dropit; | |
270 | /* | |
271 | * Send a packet designed to force a response | |
272 | * if the peer is up and reachable: | |
273 | * either an ACK if the connection is still alive, | |
274 | * or an RST if the peer has closed the connection | |
275 | * due to timeout or reboot. | |
276 | * Using sequence number tp->snd_una-1 | |
277 | * causes the transmitted zero-length segment | |
278 | * to lie outside the receive window; | |
279 | * by the protocol spec, this requires the | |
280 | * correspondent TCP to respond. | |
281 | */ | |
282 | tcpstat.tcps_keepprobe++; | |
283 | #ifdef TCP_COMPAT_42 | |
284 | /* | |
285 | * The keepalive packet must have nonzero length | |
286 | * to get a 4.2 host to respond. | |
287 | */ | |
288 | tcp_respond(tp, tp->t_template, (struct mbuf *)NULL, | |
289 | tp->rcv_nxt - 1, tp->snd_una - 1, 0); | |
290 | #else | |
291 | tcp_respond(tp, tp->t_template, (struct mbuf *)NULL, | |
292 | tp->rcv_nxt, tp->snd_una - 1, 0); | |
293 | #endif | |
294 | tp->t_timer[TCPT_KEEP] = tcp_keepintvl; | |
295 | } else | |
296 | tp->t_timer[TCPT_KEEP] = tcp_keepidle; | |
297 | break; | |
298 | dropit: | |
299 | tcpstat.tcps_keepdrops++; | |
300 | tp = tcp_drop(tp, ETIMEDOUT); | |
301 | break; | |
302 | } | |
303 | return (tp); | |
304 | } |