Commit | Line | Data |
---|---|---|
8ae0e4b4 | 1 | /* |
cc072043 | 2 | * Copyright (c) 1982, 1986, 1988 Regents of the University of California. |
2b6b6284 | 3 | * All rights reserved. |
8ae0e4b4 | 4 | * |
2b6b6284 | 5 | * Redistribution and use in source and binary forms are permitted |
616d42db KB |
6 | * provided that the above copyright notice and this paragraph are |
7 | * duplicated in all such forms and that any documentation, | |
8 | * advertising materials, and other materials related to such | |
9 | * distribution and use acknowledge that the software was developed | |
10 | * by the University of California, Berkeley. The name of the | |
11 | * University may not be used to endorse or promote products derived | |
12 | * from this software without specific prior written permission. | |
13 | * THIS SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY EXPRESS OR | |
14 | * IMPLIED WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED | |
15 | * WARRANTIES OF MERCHANTIBILITY AND FITNESS FOR A PARTICULAR PURPOSE. | |
2b6b6284 | 16 | * |
2b25f79c | 17 | * @(#)tcp_timer.c 7.16 (Berkeley) %G% |
8ae0e4b4 | 18 | */ |
f03530e6 | 19 | |
20666ad3 JB |
20 | #include "param.h" |
21 | #include "systm.h" | |
cc072043 | 22 | #include "malloc.h" |
20666ad3 JB |
23 | #include "mbuf.h" |
24 | #include "socket.h" | |
25 | #include "socketvar.h" | |
26 | #include "protosw.h" | |
27 | #include "errno.h" | |
6e7edb25 BJ |
28 | |
29 | #include "../net/if.h" | |
c124e997 | 30 | #include "../net/route.h" |
f4d55810 | 31 | |
20666ad3 | 32 | #include "in.h" |
20666ad3 JB |
33 | #include "in_systm.h" |
34 | #include "ip.h" | |
2b25f79c | 35 | #include "in_pcb.h" |
20666ad3 JB |
36 | #include "ip_var.h" |
37 | #include "tcp.h" | |
38 | #include "tcp_fsm.h" | |
39 | #include "tcp_seq.h" | |
40 | #include "tcp_timer.h" | |
41 | #include "tcp_var.h" | |
42 | #include "tcpip.h" | |
f03530e6 | 43 | |
8a36cf82 MK |
44 | int tcp_keepidle = TCPTV_KEEP_IDLE; |
45 | int tcp_keepintvl = TCPTV_KEEPINTVL; | |
46 | int tcp_maxidle; | |
f03530e6 BJ |
47 | /* |
48 | * Fast timeout routine for processing delayed acks | |
49 | */ | |
50 | tcp_fasttimo() | |
51 | { | |
b8b9174f BJ |
52 | register struct inpcb *inp; |
53 | register struct tcpcb *tp; | |
54 | int s = splnet(); | |
b8b9174f | 55 | |
fd5dc5f0 BJ |
56 | inp = tcb.inp_next; |
57 | if (inp) | |
58 | for (; inp != &tcb; inp = inp->inp_next) | |
b8b9174f BJ |
59 | if ((tp = (struct tcpcb *)inp->inp_ppcb) && |
60 | (tp->t_flags & TF_DELACK)) { | |
61 | tp->t_flags &= ~TF_DELACK; | |
62 | tp->t_flags |= TF_ACKNOW; | |
35f3fc10 | 63 | tcpstat.tcps_delack++; |
b8b9174f BJ |
64 | (void) tcp_output(tp); |
65 | } | |
66 | splx(s); | |
f03530e6 BJ |
67 | } |
68 | ||
69 | /* | |
70 | * Tcp protocol timeout routine called every 500 ms. | |
71 | * Updates the timers in all active tcb's and | |
72 | * causes finite state machine actions if timers expire. | |
73 | */ | |
74 | tcp_slowtimo() | |
75 | { | |
1e977657 | 76 | register struct inpcb *ip, *ipnxt; |
f03530e6 BJ |
77 | register struct tcpcb *tp; |
78 | int s = splnet(); | |
f03530e6 | 79 | register int i; |
f03530e6 | 80 | |
8a36cf82 | 81 | tcp_maxidle = TCPTV_KEEPCNT * tcp_keepintvl; |
f03530e6 BJ |
82 | /* |
83 | * Search through tcb's and update active timers. | |
84 | */ | |
4aed14e3 BJ |
85 | ip = tcb.inp_next; |
86 | if (ip == 0) { | |
87 | splx(s); | |
88 | return; | |
89 | } | |
039b88d5 MK |
90 | for (; ip != &tcb; ip = ipnxt) { |
91 | ipnxt = ip->inp_next; | |
f03530e6 | 92 | tp = intotcpcb(ip); |
37de812c BJ |
93 | if (tp == 0) |
94 | continue; | |
a6503abf | 95 | for (i = 0; i < TCPT_NTIMERS; i++) { |
1e977657 | 96 | if (tp->t_timer[i] && --tp->t_timer[i] == 0) { |
f03530e6 BJ |
97 | (void) tcp_usrreq(tp->t_inpcb->inp_socket, |
98 | PRU_SLOWTIMO, (struct mbuf *)0, | |
755d8841 | 99 | (struct mbuf *)i, (struct mbuf *)0); |
1e977657 BJ |
100 | if (ipnxt->inp_prev != ip) |
101 | goto tpgone; | |
102 | } | |
f03530e6 | 103 | } |
405c9168 BJ |
104 | tp->t_idle++; |
105 | if (tp->t_rtt) | |
106 | tp->t_rtt++; | |
1e977657 | 107 | tpgone: |
039b88d5 | 108 | ; |
f03530e6 | 109 | } |
a6503abf | 110 | tcp_iss += TCP_ISSINCR/PR_SLOWHZ; /* increment iss */ |
367bed15 | 111 | #ifdef TCP_COMPAT_42 |
c50542f3 MK |
112 | if ((int)tcp_iss < 0) |
113 | tcp_iss = 0; /* XXX */ | |
114 | #endif | |
f03530e6 BJ |
115 | splx(s); |
116 | } | |
117 | ||
118 | /* | |
a6503abf | 119 | * Cancel all timers for TCP tp. |
f03530e6 | 120 | */ |
0974b45c | 121 | tcp_canceltimers(tp) |
f03530e6 BJ |
122 | struct tcpcb *tp; |
123 | { | |
f03530e6 BJ |
124 | register int i; |
125 | ||
a6503abf BJ |
126 | for (i = 0; i < TCPT_NTIMERS; i++) |
127 | tp->t_timer[i] = 0; | |
f03530e6 BJ |
128 | } |
129 | ||
a6bbda13 MK |
130 | int tcp_backoff[TCP_MAXRXTSHIFT + 1] = |
131 | { 1, 2, 4, 8, 16, 32, 64, 64, 64, 64, 64, 64, 64 }; | |
8b6ad229 | 132 | |
f03530e6 | 133 | /* |
405c9168 | 134 | * TCP timer processing. |
f03530e6 | 135 | */ |
0e3936fa | 136 | struct tcpcb * |
a6503abf | 137 | tcp_timers(tp, timer) |
f03530e6 | 138 | register struct tcpcb *tp; |
a6503abf | 139 | int timer; |
f03530e6 | 140 | { |
2a89e5a6 | 141 | register int rexmt; |
f03530e6 | 142 | |
0974b45c | 143 | switch (timer) { |
f03530e6 | 144 | |
405c9168 | 145 | /* |
6209c5c4 MK |
146 | * 2 MSL timeout in shutdown went off. If we're closed but |
147 | * still waiting for peer to close and connection has been idle | |
148 | * too long, or if 2MSL time is up from TIME_WAIT, delete connection | |
149 | * control block. Otherwise, check again in a bit. | |
405c9168 | 150 | */ |
a6503abf | 151 | case TCPT_2MSL: |
6209c5c4 | 152 | if (tp->t_state != TCPS_TIME_WAIT && |
8a36cf82 MK |
153 | tp->t_idle <= tcp_maxidle) |
154 | tp->t_timer[TCPT_2MSL] = tcp_keepintvl; | |
6209c5c4 MK |
155 | else |
156 | tp = tcp_close(tp); | |
0e3936fa | 157 | break; |
f03530e6 | 158 | |
405c9168 BJ |
159 | /* |
160 | * Retransmission timer went off. Message has not | |
161 | * been acked within retransmit interval. Back off | |
eeaf00e3 | 162 | * to a longer retransmit interval and retransmit one segment. |
405c9168 | 163 | */ |
a6503abf | 164 | case TCPT_REXMT: |
a6bbda13 MK |
165 | if (++tp->t_rxtshift > TCP_MAXRXTSHIFT) { |
166 | tp->t_rxtshift = TCP_MAXRXTSHIFT; | |
35f3fc10 | 167 | tcpstat.tcps_timeoutdrop++; |
0e3936fa SL |
168 | tp = tcp_drop(tp, ETIMEDOUT); |
169 | break; | |
fd5dc5f0 | 170 | } |
35f3fc10 | 171 | tcpstat.tcps_rexmttimeo++; |
7cc62c26 MK |
172 | rexmt = ((tp->t_srtt >> 2) + tp->t_rttvar) >> 1; |
173 | rexmt *= tcp_backoff[tp->t_rxtshift]; | |
a6bbda13 MK |
174 | TCPT_RANGESET(tp->t_rxtcur, rexmt, TCPTV_MIN, TCPTV_REXMTMAX); |
175 | tp->t_timer[TCPT_REXMT] = tp->t_rxtcur; | |
6a4fd140 | 176 | /* |
a6bbda13 MK |
177 | * If losing, let the lower level know and try for |
178 | * a better route. Also, if we backed off this far, | |
179 | * our srtt estimate is probably bogus. Clobber it | |
180 | * so we'll take the next rtt measurement as our srtt; | |
181 | * move the current srtt into rttvar to keep the current | |
182 | * retransmit times until then. | |
6a4fd140 | 183 | */ |
a6bbda13 | 184 | if (tp->t_rxtshift > TCP_MAXRXTSHIFT / 4) { |
9d866d2f | 185 | #if BSD>=43 |
6a4fd140 | 186 | in_losing(tp->t_inpcb); |
9d866d2f | 187 | #endif |
a6bbda13 MK |
188 | tp->t_rttvar += (tp->t_srtt >> 2); |
189 | tp->t_srtt = 0; | |
190 | } | |
405c9168 | 191 | tp->snd_nxt = tp->snd_una; |
7872cc0a | 192 | /* |
7cc62c26 | 193 | * If timing a segment in this window, stop the timer. |
7872cc0a | 194 | */ |
7cc62c26 | 195 | tp->t_rtt = 0; |
2e5a76f2 MK |
196 | /* |
197 | * Close the congestion window down to one segment | |
198 | * (we'll open it by one segment for each ack we get). | |
199 | * Since we probably have a window's worth of unacked | |
200 | * data accumulated, this "slow start" keeps us from | |
201 | * dumping all that data as back-to-back packets (which | |
202 | * might overwhelm an intermediate gateway). | |
6f854ef4 MK |
203 | * |
204 | * There are two phases to the opening: Initially we | |
205 | * open by one mss on each ack. This makes the window | |
206 | * size increase exponentially with time. If the | |
207 | * window is larger than the path can handle, this | |
208 | * exponential growth results in dropped packet(s) | |
209 | * almost immediately. To get more time between | |
210 | * drops but still "push" the network to take advantage | |
211 | * of improving conditions, we switch from exponential | |
212 | * to linear window opening at some threshhold size. | |
213 | * For a threshhold, we use half the current window | |
214 | * size, truncated to a multiple of the mss. | |
215 | * | |
216 | * (the minimum cwnd that will give us exponential | |
217 | * growth is 2 mss. We don't allow the threshhold | |
218 | * to go below this.) | |
2e5a76f2 | 219 | */ |
6f854ef4 MK |
220 | { |
221 | u_int win = MIN(tp->snd_wnd, tp->snd_cwnd) / 2 / tp->t_maxseg; | |
222 | if (win < 2) | |
223 | win = 2; | |
2e5a76f2 | 224 | tp->snd_cwnd = tp->t_maxseg; |
6f854ef4 MK |
225 | tp->snd_ssthresh = win * tp->t_maxseg; |
226 | } | |
405c9168 | 227 | (void) tcp_output(tp); |
0e3936fa | 228 | break; |
f03530e6 | 229 | |
405c9168 BJ |
230 | /* |
231 | * Persistance timer into zero window. | |
232 | * Force a byte to be output, if possible. | |
233 | */ | |
a6503abf | 234 | case TCPT_PERSIST: |
35f3fc10 | 235 | tcpstat.tcps_persisttimeo++; |
a13c006d | 236 | tcp_setpersist(tp); |
405c9168 BJ |
237 | tp->t_force = 1; |
238 | (void) tcp_output(tp); | |
239 | tp->t_force = 0; | |
0e3936fa | 240 | break; |
f03530e6 | 241 | |
405c9168 BJ |
242 | /* |
243 | * Keep-alive timer went off; send something | |
244 | * or drop connection if idle for too long. | |
245 | */ | |
a6503abf | 246 | case TCPT_KEEP: |
35f3fc10 | 247 | tcpstat.tcps_keeptimeo++; |
f3cdd721 BJ |
248 | if (tp->t_state < TCPS_ESTABLISHED) |
249 | goto dropit; | |
c0200c65 MK |
250 | if (tp->t_inpcb->inp_socket->so_options & SO_KEEPALIVE && |
251 | tp->t_state <= TCPS_CLOSE_WAIT) { | |
8a36cf82 | 252 | if (tp->t_idle >= tcp_keepidle + tcp_maxidle) |
f3cdd721 | 253 | goto dropit; |
3bd14c98 | 254 | /* |
35f3fc10 MK |
255 | * Send a packet designed to force a response |
256 | * if the peer is up and reachable: | |
257 | * either an ACK if the connection is still alive, | |
258 | * or an RST if the peer has closed the connection | |
259 | * due to timeout or reboot. | |
260 | * Using sequence number tp->snd_una-1 | |
261 | * causes the transmitted zero-length segment | |
262 | * to lie outside the receive window; | |
263 | * by the protocol spec, this requires the | |
264 | * correspondent TCP to respond. | |
3bd14c98 | 265 | */ |
35f3fc10 | 266 | tcpstat.tcps_keepprobe++; |
eeef4ac3 MK |
267 | #ifdef TCP_COMPAT_42 |
268 | /* | |
269 | * The keepalive packet must have nonzero length | |
270 | * to get a 4.2 host to respond. | |
271 | */ | |
cc072043 | 272 | tcp_respond(tp, tp->t_template, (struct mbuf *)NULL, |
eeef4ac3 MK |
273 | tp->rcv_nxt - 1, tp->snd_una - 1, 0); |
274 | #else | |
cc072043 | 275 | tcp_respond(tp, tp->t_template, (struct mbuf *)NULL, |
eeef4ac3 MK |
276 | tp->rcv_nxt, tp->snd_una - 1, 0); |
277 | #endif | |
8a36cf82 MK |
278 | tp->t_timer[TCPT_KEEP] = tcp_keepintvl; |
279 | } else | |
280 | tp->t_timer[TCPT_KEEP] = tcp_keepidle; | |
0e3936fa | 281 | break; |
f3cdd721 | 282 | dropit: |
35f3fc10 | 283 | tcpstat.tcps_keepdrops++; |
0e3936fa SL |
284 | tp = tcp_drop(tp, ETIMEDOUT); |
285 | break; | |
f03530e6 | 286 | } |
0e3936fa | 287 | return (tp); |
f03530e6 | 288 | } |