Commit | Line | Data |
---|---|---|
8ae0e4b4 | 1 | /* |
0880b18e | 2 | * Copyright (c) 1982, 1986 Regents of the University of California. |
2b6b6284 | 3 | * All rights reserved. |
8ae0e4b4 | 4 | * |
2b6b6284 KB |
5 | * Redistribution and use in source and binary forms are permitted |
6 | * provided that this notice is preserved and that due credit is given | |
7 | * to the University of California at Berkeley. The name of the University | |
8 | * may not be used to endorse or promote products derived from this | |
9 | * software without specific prior written permission. This software | |
10 | * is provided ``as is'' without express or implied warranty. | |
11 | * | |
22340ea2 | 12 | * @(#)tcp_timer.c 7.11.1.2 (Berkeley) %G% |
8ae0e4b4 | 13 | */ |
f03530e6 | 14 | |
20666ad3 JB |
15 | #include "param.h" |
16 | #include "systm.h" | |
17 | #include "mbuf.h" | |
18 | #include "socket.h" | |
19 | #include "socketvar.h" | |
20 | #include "protosw.h" | |
21 | #include "errno.h" | |
6e7edb25 BJ |
22 | |
23 | #include "../net/if.h" | |
c124e997 | 24 | #include "../net/route.h" |
f4d55810 | 25 | |
20666ad3 JB |
26 | #include "in.h" |
27 | #include "in_pcb.h" | |
28 | #include "in_systm.h" | |
29 | #include "ip.h" | |
30 | #include "ip_var.h" | |
31 | #include "tcp.h" | |
32 | #include "tcp_fsm.h" | |
33 | #include "tcp_seq.h" | |
34 | #include "tcp_timer.h" | |
35 | #include "tcp_var.h" | |
36 | #include "tcpip.h" | |
f03530e6 | 37 | |
8b5a83bb | 38 | int tcpnodelack = 0; |
8a36cf82 MK |
39 | int tcp_keepidle = TCPTV_KEEP_IDLE; |
40 | int tcp_keepintvl = TCPTV_KEEPINTVL; | |
41 | int tcp_maxidle; | |
f03530e6 BJ |
42 | /* |
43 | * Fast timeout routine for processing delayed acks | |
44 | */ | |
45 | tcp_fasttimo() | |
46 | { | |
b8b9174f BJ |
47 | register struct inpcb *inp; |
48 | register struct tcpcb *tp; | |
49 | int s = splnet(); | |
b8b9174f | 50 | |
fd5dc5f0 BJ |
51 | inp = tcb.inp_next; |
52 | if (inp) | |
53 | for (; inp != &tcb; inp = inp->inp_next) | |
b8b9174f BJ |
54 | if ((tp = (struct tcpcb *)inp->inp_ppcb) && |
55 | (tp->t_flags & TF_DELACK)) { | |
56 | tp->t_flags &= ~TF_DELACK; | |
57 | tp->t_flags |= TF_ACKNOW; | |
35f3fc10 | 58 | tcpstat.tcps_delack++; |
b8b9174f BJ |
59 | (void) tcp_output(tp); |
60 | } | |
61 | splx(s); | |
f03530e6 BJ |
62 | } |
63 | ||
64 | /* | |
65 | * Tcp protocol timeout routine called every 500 ms. | |
66 | * Updates the timers in all active tcb's and | |
67 | * causes finite state machine actions if timers expire. | |
68 | */ | |
69 | tcp_slowtimo() | |
70 | { | |
1e977657 | 71 | register struct inpcb *ip, *ipnxt; |
f03530e6 BJ |
72 | register struct tcpcb *tp; |
73 | int s = splnet(); | |
f03530e6 | 74 | register int i; |
f03530e6 | 75 | |
8a36cf82 | 76 | tcp_maxidle = TCPTV_KEEPCNT * tcp_keepintvl; |
f03530e6 BJ |
77 | /* |
78 | * Search through tcb's and update active timers. | |
79 | */ | |
4aed14e3 BJ |
80 | ip = tcb.inp_next; |
81 | if (ip == 0) { | |
82 | splx(s); | |
83 | return; | |
84 | } | |
039b88d5 MK |
85 | for (; ip != &tcb; ip = ipnxt) { |
86 | ipnxt = ip->inp_next; | |
f03530e6 | 87 | tp = intotcpcb(ip); |
37de812c BJ |
88 | if (tp == 0) |
89 | continue; | |
a6503abf | 90 | for (i = 0; i < TCPT_NTIMERS; i++) { |
1e977657 | 91 | if (tp->t_timer[i] && --tp->t_timer[i] == 0) { |
f03530e6 BJ |
92 | (void) tcp_usrreq(tp->t_inpcb->inp_socket, |
93 | PRU_SLOWTIMO, (struct mbuf *)0, | |
755d8841 | 94 | (struct mbuf *)i, (struct mbuf *)0); |
1e977657 BJ |
95 | if (ipnxt->inp_prev != ip) |
96 | goto tpgone; | |
97 | } | |
f03530e6 | 98 | } |
405c9168 BJ |
99 | tp->t_idle++; |
100 | if (tp->t_rtt) | |
101 | tp->t_rtt++; | |
1e977657 | 102 | tpgone: |
039b88d5 | 103 | ; |
f03530e6 | 104 | } |
a6503abf | 105 | tcp_iss += TCP_ISSINCR/PR_SLOWHZ; /* increment iss */ |
367bed15 | 106 | #ifdef TCP_COMPAT_42 |
c50542f3 MK |
107 | if ((int)tcp_iss < 0) |
108 | tcp_iss = 0; /* XXX */ | |
109 | #endif | |
f03530e6 BJ |
110 | splx(s); |
111 | } | |
112 | ||
113 | /* | |
a6503abf | 114 | * Cancel all timers for TCP tp. |
f03530e6 | 115 | */ |
0974b45c | 116 | tcp_canceltimers(tp) |
f03530e6 BJ |
117 | struct tcpcb *tp; |
118 | { | |
f03530e6 BJ |
119 | register int i; |
120 | ||
a6503abf BJ |
121 | for (i = 0; i < TCPT_NTIMERS; i++) |
122 | tp->t_timer[i] = 0; | |
f03530e6 BJ |
123 | } |
124 | ||
a6bbda13 MK |
125 | int tcp_backoff[TCP_MAXRXTSHIFT + 1] = |
126 | { 1, 2, 4, 8, 16, 32, 64, 64, 64, 64, 64, 64, 64 }; | |
8b6ad229 | 127 | |
f03530e6 | 128 | /* |
405c9168 | 129 | * TCP timer processing. |
f03530e6 | 130 | */ |
0e3936fa | 131 | struct tcpcb * |
a6503abf | 132 | tcp_timers(tp, timer) |
f03530e6 | 133 | register struct tcpcb *tp; |
a6503abf | 134 | int timer; |
f03530e6 | 135 | { |
2a89e5a6 | 136 | register int rexmt; |
f03530e6 | 137 | |
0974b45c | 138 | switch (timer) { |
f03530e6 | 139 | |
405c9168 | 140 | /* |
6209c5c4 MK |
141 | * 2 MSL timeout in shutdown went off. If we're closed but |
142 | * still waiting for peer to close and connection has been idle | |
143 | * too long, or if 2MSL time is up from TIME_WAIT, delete connection | |
144 | * control block. Otherwise, check again in a bit. | |
405c9168 | 145 | */ |
a6503abf | 146 | case TCPT_2MSL: |
6209c5c4 | 147 | if (tp->t_state != TCPS_TIME_WAIT && |
8a36cf82 MK |
148 | tp->t_idle <= tcp_maxidle) |
149 | tp->t_timer[TCPT_2MSL] = tcp_keepintvl; | |
6209c5c4 MK |
150 | else |
151 | tp = tcp_close(tp); | |
0e3936fa | 152 | break; |
f03530e6 | 153 | |
405c9168 BJ |
154 | /* |
155 | * Retransmission timer went off. Message has not | |
156 | * been acked within retransmit interval. Back off | |
eeaf00e3 | 157 | * to a longer retransmit interval and retransmit one segment. |
405c9168 | 158 | */ |
a6503abf | 159 | case TCPT_REXMT: |
a6bbda13 MK |
160 | if (++tp->t_rxtshift > TCP_MAXRXTSHIFT) { |
161 | tp->t_rxtshift = TCP_MAXRXTSHIFT; | |
35f3fc10 | 162 | tcpstat.tcps_timeoutdrop++; |
0e3936fa SL |
163 | tp = tcp_drop(tp, ETIMEDOUT); |
164 | break; | |
fd5dc5f0 | 165 | } |
35f3fc10 | 166 | tcpstat.tcps_rexmttimeo++; |
7cc62c26 MK |
167 | rexmt = ((tp->t_srtt >> 2) + tp->t_rttvar) >> 1; |
168 | rexmt *= tcp_backoff[tp->t_rxtshift]; | |
a6bbda13 MK |
169 | TCPT_RANGESET(tp->t_rxtcur, rexmt, TCPTV_MIN, TCPTV_REXMTMAX); |
170 | tp->t_timer[TCPT_REXMT] = tp->t_rxtcur; | |
6a4fd140 | 171 | /* |
a6bbda13 MK |
172 | * If losing, let the lower level know and try for |
173 | * a better route. Also, if we backed off this far, | |
174 | * our srtt estimate is probably bogus. Clobber it | |
175 | * so we'll take the next rtt measurement as our srtt; | |
176 | * move the current srtt into rttvar to keep the current | |
177 | * retransmit times until then. | |
6a4fd140 | 178 | */ |
a6bbda13 | 179 | if (tp->t_rxtshift > TCP_MAXRXTSHIFT / 4) { |
9d866d2f | 180 | #if BSD>=43 |
6a4fd140 | 181 | in_losing(tp->t_inpcb); |
9d866d2f | 182 | #endif |
a6bbda13 MK |
183 | tp->t_rttvar += (tp->t_srtt >> 2); |
184 | tp->t_srtt = 0; | |
185 | } | |
405c9168 | 186 | tp->snd_nxt = tp->snd_una; |
7872cc0a | 187 | /* |
7cc62c26 | 188 | * If timing a segment in this window, stop the timer. |
7872cc0a | 189 | */ |
7cc62c26 | 190 | tp->t_rtt = 0; |
2e5a76f2 MK |
191 | /* |
192 | * Close the congestion window down to one segment | |
193 | * (we'll open it by one segment for each ack we get). | |
194 | * Since we probably have a window's worth of unacked | |
195 | * data accumulated, this "slow start" keeps us from | |
196 | * dumping all that data as back-to-back packets (which | |
197 | * might overwhelm an intermediate gateway). | |
6f854ef4 MK |
198 | * |
199 | * There are two phases to the opening: Initially we | |
200 | * open by one mss on each ack. This makes the window | |
201 | * size increase exponentially with time. If the | |
202 | * window is larger than the path can handle, this | |
203 | * exponential growth results in dropped packet(s) | |
204 | * almost immediately. To get more time between | |
205 | * drops but still "push" the network to take advantage | |
206 | * of improving conditions, we switch from exponential | |
207 | * to linear window opening at some threshhold size. | |
208 | * For a threshhold, we use half the current window | |
209 | * size, truncated to a multiple of the mss. | |
210 | * | |
211 | * (the minimum cwnd that will give us exponential | |
212 | * growth is 2 mss. We don't allow the threshhold | |
213 | * to go below this.) | |
2e5a76f2 | 214 | */ |
6f854ef4 MK |
215 | { |
216 | u_int win = MIN(tp->snd_wnd, tp->snd_cwnd) / 2 / tp->t_maxseg; | |
217 | if (win < 2) | |
218 | win = 2; | |
2e5a76f2 | 219 | tp->snd_cwnd = tp->t_maxseg; |
6f854ef4 MK |
220 | tp->snd_ssthresh = win * tp->t_maxseg; |
221 | } | |
405c9168 | 222 | (void) tcp_output(tp); |
0e3936fa | 223 | break; |
f03530e6 | 224 | |
405c9168 BJ |
225 | /* |
226 | * Persistance timer into zero window. | |
227 | * Force a byte to be output, if possible. | |
228 | */ | |
a6503abf | 229 | case TCPT_PERSIST: |
35f3fc10 | 230 | tcpstat.tcps_persisttimeo++; |
a13c006d | 231 | tcp_setpersist(tp); |
405c9168 BJ |
232 | tp->t_force = 1; |
233 | (void) tcp_output(tp); | |
234 | tp->t_force = 0; | |
0e3936fa | 235 | break; |
f03530e6 | 236 | |
405c9168 BJ |
237 | /* |
238 | * Keep-alive timer went off; send something | |
239 | * or drop connection if idle for too long. | |
240 | */ | |
a6503abf | 241 | case TCPT_KEEP: |
35f3fc10 | 242 | tcpstat.tcps_keeptimeo++; |
f3cdd721 BJ |
243 | if (tp->t_state < TCPS_ESTABLISHED) |
244 | goto dropit; | |
c0200c65 MK |
245 | if (tp->t_inpcb->inp_socket->so_options & SO_KEEPALIVE && |
246 | tp->t_state <= TCPS_CLOSE_WAIT) { | |
8a36cf82 | 247 | if (tp->t_idle >= tcp_keepidle + tcp_maxidle) |
f3cdd721 | 248 | goto dropit; |
3bd14c98 | 249 | /* |
35f3fc10 MK |
250 | * Send a packet designed to force a response |
251 | * if the peer is up and reachable: | |
252 | * either an ACK if the connection is still alive, | |
253 | * or an RST if the peer has closed the connection | |
254 | * due to timeout or reboot. | |
255 | * Using sequence number tp->snd_una-1 | |
256 | * causes the transmitted zero-length segment | |
257 | * to lie outside the receive window; | |
258 | * by the protocol spec, this requires the | |
259 | * correspondent TCP to respond. | |
3bd14c98 | 260 | */ |
35f3fc10 | 261 | tcpstat.tcps_keepprobe++; |
eeef4ac3 MK |
262 | #ifdef TCP_COMPAT_42 |
263 | /* | |
264 | * The keepalive packet must have nonzero length | |
265 | * to get a 4.2 host to respond. | |
266 | */ | |
8b6ad229 | 267 | tcp_respond(tp, tp->t_template, |
eeef4ac3 MK |
268 | tp->rcv_nxt - 1, tp->snd_una - 1, 0); |
269 | #else | |
270 | tcp_respond(tp, tp->t_template, | |
271 | tp->rcv_nxt, tp->snd_una - 1, 0); | |
272 | #endif | |
8a36cf82 MK |
273 | tp->t_timer[TCPT_KEEP] = tcp_keepintvl; |
274 | } else | |
275 | tp->t_timer[TCPT_KEEP] = tcp_keepidle; | |
0e3936fa | 276 | break; |
f3cdd721 | 277 | dropit: |
35f3fc10 | 278 | tcpstat.tcps_keepdrops++; |
0e3936fa SL |
279 | tp = tcp_drop(tp, ETIMEDOUT); |
280 | break; | |
f03530e6 | 281 | } |
0e3936fa | 282 | return (tp); |
f03530e6 | 283 | } |