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 | 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 | * |
616d42db | 17 | * @(#)tcp_timer.c 7.14 (Berkeley) %G% |
8ae0e4b4 | 18 | */ |
f03530e6 | 19 | |
20666ad3 JB |
20 | #include "param.h" |
21 | #include "systm.h" | |
22 | #include "mbuf.h" | |
23 | #include "socket.h" | |
24 | #include "socketvar.h" | |
25 | #include "protosw.h" | |
26 | #include "errno.h" | |
6e7edb25 BJ |
27 | |
28 | #include "../net/if.h" | |
c124e997 | 29 | #include "../net/route.h" |
f4d55810 | 30 | |
20666ad3 JB |
31 | #include "in.h" |
32 | #include "in_pcb.h" | |
33 | #include "in_systm.h" | |
34 | #include "ip.h" | |
35 | #include "ip_var.h" | |
36 | #include "tcp.h" | |
37 | #include "tcp_fsm.h" | |
38 | #include "tcp_seq.h" | |
39 | #include "tcp_timer.h" | |
40 | #include "tcp_var.h" | |
41 | #include "tcpip.h" | |
f03530e6 | 42 | |
8a36cf82 MK |
43 | int tcp_keepidle = TCPTV_KEEP_IDLE; |
44 | int tcp_keepintvl = TCPTV_KEEPINTVL; | |
45 | int tcp_maxidle; | |
f03530e6 BJ |
46 | /* |
47 | * Fast timeout routine for processing delayed acks | |
48 | */ | |
49 | tcp_fasttimo() | |
50 | { | |
b8b9174f BJ |
51 | register struct inpcb *inp; |
52 | register struct tcpcb *tp; | |
53 | int s = splnet(); | |
b8b9174f | 54 | |
fd5dc5f0 BJ |
55 | inp = tcb.inp_next; |
56 | if (inp) | |
57 | for (; inp != &tcb; inp = inp->inp_next) | |
b8b9174f BJ |
58 | if ((tp = (struct tcpcb *)inp->inp_ppcb) && |
59 | (tp->t_flags & TF_DELACK)) { | |
60 | tp->t_flags &= ~TF_DELACK; | |
61 | tp->t_flags |= TF_ACKNOW; | |
35f3fc10 | 62 | tcpstat.tcps_delack++; |
b8b9174f BJ |
63 | (void) tcp_output(tp); |
64 | } | |
65 | splx(s); | |
f03530e6 BJ |
66 | } |
67 | ||
68 | /* | |
69 | * Tcp protocol timeout routine called every 500 ms. | |
70 | * Updates the timers in all active tcb's and | |
71 | * causes finite state machine actions if timers expire. | |
72 | */ | |
73 | tcp_slowtimo() | |
74 | { | |
1e977657 | 75 | register struct inpcb *ip, *ipnxt; |
f03530e6 BJ |
76 | register struct tcpcb *tp; |
77 | int s = splnet(); | |
f03530e6 | 78 | register int i; |
f03530e6 | 79 | |
8a36cf82 | 80 | tcp_maxidle = TCPTV_KEEPCNT * tcp_keepintvl; |
f03530e6 BJ |
81 | /* |
82 | * Search through tcb's and update active timers. | |
83 | */ | |
4aed14e3 BJ |
84 | ip = tcb.inp_next; |
85 | if (ip == 0) { | |
86 | splx(s); | |
87 | return; | |
88 | } | |
039b88d5 MK |
89 | for (; ip != &tcb; ip = ipnxt) { |
90 | ipnxt = ip->inp_next; | |
f03530e6 | 91 | tp = intotcpcb(ip); |
37de812c BJ |
92 | if (tp == 0) |
93 | continue; | |
a6503abf | 94 | for (i = 0; i < TCPT_NTIMERS; i++) { |
1e977657 | 95 | if (tp->t_timer[i] && --tp->t_timer[i] == 0) { |
f03530e6 BJ |
96 | (void) tcp_usrreq(tp->t_inpcb->inp_socket, |
97 | PRU_SLOWTIMO, (struct mbuf *)0, | |
755d8841 | 98 | (struct mbuf *)i, (struct mbuf *)0); |
1e977657 BJ |
99 | if (ipnxt->inp_prev != ip) |
100 | goto tpgone; | |
101 | } | |
f03530e6 | 102 | } |
405c9168 BJ |
103 | tp->t_idle++; |
104 | if (tp->t_rtt) | |
105 | tp->t_rtt++; | |
1e977657 | 106 | tpgone: |
039b88d5 | 107 | ; |
f03530e6 | 108 | } |
a6503abf | 109 | tcp_iss += TCP_ISSINCR/PR_SLOWHZ; /* increment iss */ |
367bed15 | 110 | #ifdef TCP_COMPAT_42 |
c50542f3 MK |
111 | if ((int)tcp_iss < 0) |
112 | tcp_iss = 0; /* XXX */ | |
113 | #endif | |
f03530e6 BJ |
114 | splx(s); |
115 | } | |
116 | ||
117 | /* | |
a6503abf | 118 | * Cancel all timers for TCP tp. |
f03530e6 | 119 | */ |
0974b45c | 120 | tcp_canceltimers(tp) |
f03530e6 BJ |
121 | struct tcpcb *tp; |
122 | { | |
f03530e6 BJ |
123 | register int i; |
124 | ||
a6503abf BJ |
125 | for (i = 0; i < TCPT_NTIMERS; i++) |
126 | tp->t_timer[i] = 0; | |
f03530e6 BJ |
127 | } |
128 | ||
a6bbda13 MK |
129 | int tcp_backoff[TCP_MAXRXTSHIFT + 1] = |
130 | { 1, 2, 4, 8, 16, 32, 64, 64, 64, 64, 64, 64, 64 }; | |
8b6ad229 | 131 | |
f03530e6 | 132 | /* |
405c9168 | 133 | * TCP timer processing. |
f03530e6 | 134 | */ |
0e3936fa | 135 | struct tcpcb * |
a6503abf | 136 | tcp_timers(tp, timer) |
f03530e6 | 137 | register struct tcpcb *tp; |
a6503abf | 138 | int timer; |
f03530e6 | 139 | { |
2a89e5a6 | 140 | register int rexmt; |
f03530e6 | 141 | |
0974b45c | 142 | switch (timer) { |
f03530e6 | 143 | |
405c9168 | 144 | /* |
6209c5c4 MK |
145 | * 2 MSL timeout in shutdown went off. If we're closed but |
146 | * still waiting for peer to close and connection has been idle | |
147 | * too long, or if 2MSL time is up from TIME_WAIT, delete connection | |
148 | * control block. Otherwise, check again in a bit. | |
405c9168 | 149 | */ |
a6503abf | 150 | case TCPT_2MSL: |
6209c5c4 | 151 | if (tp->t_state != TCPS_TIME_WAIT && |
8a36cf82 MK |
152 | tp->t_idle <= tcp_maxidle) |
153 | tp->t_timer[TCPT_2MSL] = tcp_keepintvl; | |
6209c5c4 MK |
154 | else |
155 | tp = tcp_close(tp); | |
0e3936fa | 156 | break; |
f03530e6 | 157 | |
405c9168 BJ |
158 | /* |
159 | * Retransmission timer went off. Message has not | |
160 | * been acked within retransmit interval. Back off | |
eeaf00e3 | 161 | * to a longer retransmit interval and retransmit one segment. |
405c9168 | 162 | */ |
a6503abf | 163 | case TCPT_REXMT: |
a6bbda13 MK |
164 | if (++tp->t_rxtshift > TCP_MAXRXTSHIFT) { |
165 | tp->t_rxtshift = TCP_MAXRXTSHIFT; | |
35f3fc10 | 166 | tcpstat.tcps_timeoutdrop++; |
0e3936fa SL |
167 | tp = tcp_drop(tp, ETIMEDOUT); |
168 | break; | |
fd5dc5f0 | 169 | } |
35f3fc10 | 170 | tcpstat.tcps_rexmttimeo++; |
7cc62c26 MK |
171 | rexmt = ((tp->t_srtt >> 2) + tp->t_rttvar) >> 1; |
172 | rexmt *= tcp_backoff[tp->t_rxtshift]; | |
a6bbda13 MK |
173 | TCPT_RANGESET(tp->t_rxtcur, rexmt, TCPTV_MIN, TCPTV_REXMTMAX); |
174 | tp->t_timer[TCPT_REXMT] = tp->t_rxtcur; | |
6a4fd140 | 175 | /* |
a6bbda13 MK |
176 | * If losing, let the lower level know and try for |
177 | * a better route. Also, if we backed off this far, | |
178 | * our srtt estimate is probably bogus. Clobber it | |
179 | * so we'll take the next rtt measurement as our srtt; | |
180 | * move the current srtt into rttvar to keep the current | |
181 | * retransmit times until then. | |
6a4fd140 | 182 | */ |
a6bbda13 | 183 | if (tp->t_rxtshift > TCP_MAXRXTSHIFT / 4) { |
9d866d2f | 184 | #if BSD>=43 |
6a4fd140 | 185 | in_losing(tp->t_inpcb); |
9d866d2f | 186 | #endif |
a6bbda13 MK |
187 | tp->t_rttvar += (tp->t_srtt >> 2); |
188 | tp->t_srtt = 0; | |
189 | } | |
405c9168 | 190 | tp->snd_nxt = tp->snd_una; |
7872cc0a | 191 | /* |
7cc62c26 | 192 | * If timing a segment in this window, stop the timer. |
7872cc0a | 193 | */ |
7cc62c26 | 194 | tp->t_rtt = 0; |
2e5a76f2 MK |
195 | /* |
196 | * Close the congestion window down to one segment | |
197 | * (we'll open it by one segment for each ack we get). | |
198 | * Since we probably have a window's worth of unacked | |
199 | * data accumulated, this "slow start" keeps us from | |
200 | * dumping all that data as back-to-back packets (which | |
201 | * might overwhelm an intermediate gateway). | |
6f854ef4 MK |
202 | * |
203 | * There are two phases to the opening: Initially we | |
204 | * open by one mss on each ack. This makes the window | |
205 | * size increase exponentially with time. If the | |
206 | * window is larger than the path can handle, this | |
207 | * exponential growth results in dropped packet(s) | |
208 | * almost immediately. To get more time between | |
209 | * drops but still "push" the network to take advantage | |
210 | * of improving conditions, we switch from exponential | |
211 | * to linear window opening at some threshhold size. | |
212 | * For a threshhold, we use half the current window | |
213 | * size, truncated to a multiple of the mss. | |
214 | * | |
215 | * (the minimum cwnd that will give us exponential | |
216 | * growth is 2 mss. We don't allow the threshhold | |
217 | * to go below this.) | |
2e5a76f2 | 218 | */ |
6f854ef4 MK |
219 | { |
220 | u_int win = MIN(tp->snd_wnd, tp->snd_cwnd) / 2 / tp->t_maxseg; | |
221 | if (win < 2) | |
222 | win = 2; | |
2e5a76f2 | 223 | tp->snd_cwnd = tp->t_maxseg; |
6f854ef4 MK |
224 | tp->snd_ssthresh = win * tp->t_maxseg; |
225 | } | |
405c9168 | 226 | (void) tcp_output(tp); |
0e3936fa | 227 | break; |
f03530e6 | 228 | |
405c9168 BJ |
229 | /* |
230 | * Persistance timer into zero window. | |
231 | * Force a byte to be output, if possible. | |
232 | */ | |
a6503abf | 233 | case TCPT_PERSIST: |
35f3fc10 | 234 | tcpstat.tcps_persisttimeo++; |
a13c006d | 235 | tcp_setpersist(tp); |
405c9168 BJ |
236 | tp->t_force = 1; |
237 | (void) tcp_output(tp); | |
238 | tp->t_force = 0; | |
0e3936fa | 239 | break; |
f03530e6 | 240 | |
405c9168 BJ |
241 | /* |
242 | * Keep-alive timer went off; send something | |
243 | * or drop connection if idle for too long. | |
244 | */ | |
a6503abf | 245 | case TCPT_KEEP: |
35f3fc10 | 246 | tcpstat.tcps_keeptimeo++; |
f3cdd721 BJ |
247 | if (tp->t_state < TCPS_ESTABLISHED) |
248 | goto dropit; | |
c0200c65 MK |
249 | if (tp->t_inpcb->inp_socket->so_options & SO_KEEPALIVE && |
250 | tp->t_state <= TCPS_CLOSE_WAIT) { | |
8a36cf82 | 251 | if (tp->t_idle >= tcp_keepidle + tcp_maxidle) |
f3cdd721 | 252 | goto dropit; |
3bd14c98 | 253 | /* |
35f3fc10 MK |
254 | * Send a packet designed to force a response |
255 | * if the peer is up and reachable: | |
256 | * either an ACK if the connection is still alive, | |
257 | * or an RST if the peer has closed the connection | |
258 | * due to timeout or reboot. | |
259 | * Using sequence number tp->snd_una-1 | |
260 | * causes the transmitted zero-length segment | |
261 | * to lie outside the receive window; | |
262 | * by the protocol spec, this requires the | |
263 | * correspondent TCP to respond. | |
3bd14c98 | 264 | */ |
35f3fc10 | 265 | tcpstat.tcps_keepprobe++; |
eeef4ac3 MK |
266 | #ifdef TCP_COMPAT_42 |
267 | /* | |
268 | * The keepalive packet must have nonzero length | |
269 | * to get a 4.2 host to respond. | |
270 | */ | |
8b6ad229 | 271 | tcp_respond(tp, tp->t_template, |
eeef4ac3 MK |
272 | tp->rcv_nxt - 1, tp->snd_una - 1, 0); |
273 | #else | |
274 | tcp_respond(tp, tp->t_template, | |
275 | tp->rcv_nxt, tp->snd_una - 1, 0); | |
276 | #endif | |
8a36cf82 MK |
277 | tp->t_timer[TCPT_KEEP] = tcp_keepintvl; |
278 | } else | |
279 | tp->t_timer[TCPT_KEEP] = tcp_keepidle; | |
0e3936fa | 280 | break; |
f3cdd721 | 281 | dropit: |
35f3fc10 | 282 | tcpstat.tcps_keepdrops++; |
0e3936fa SL |
283 | tp = tcp_drop(tp, ETIMEDOUT); |
284 | break; | |
f03530e6 | 285 | } |
0e3936fa | 286 | return (tp); |
f03530e6 | 287 | } |