tcp_timer.c revision 1.26
1 /* $NetBSD: tcp_timer.c,v 1.26 1997/12/17 06:04:17 thorpej Exp $ */ 2 3 /* 4 * Copyright (c) 1982, 1986, 1988, 1990, 1993 5 * The Regents of the University of California. All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. All advertising materials mentioning features or use of this software 16 * must display the following acknowledgement: 17 * This product includes software developed by the University of 18 * California, Berkeley and its contributors. 19 * 4. Neither the name of the University nor the names of its contributors 20 * may be used to endorse or promote products derived from this software 21 * without specific prior written permission. 22 * 23 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 24 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 25 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 26 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 28 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 29 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 30 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 31 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 32 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 33 * SUCH DAMAGE. 34 * 35 * @(#)tcp_timer.c 8.1 (Berkeley) 6/10/93 36 */ 37 38 #ifndef TUBA_INCLUDE 39 #include <sys/param.h> 40 #include <sys/systm.h> 41 #include <sys/malloc.h> 42 #include <sys/mbuf.h> 43 #include <sys/socket.h> 44 #include <sys/socketvar.h> 45 #include <sys/protosw.h> 46 #include <sys/errno.h> 47 48 #include <net/if.h> 49 #include <net/route.h> 50 51 #include <netinet/in.h> 52 #include <netinet/in_systm.h> 53 #include <netinet/ip.h> 54 #include <netinet/in_pcb.h> 55 #include <netinet/ip_var.h> 56 #include <netinet/tcp.h> 57 #include <netinet/tcp_fsm.h> 58 #include <netinet/tcp_seq.h> 59 #include <netinet/tcp_timer.h> 60 #include <netinet/tcp_var.h> 61 #include <netinet/tcpip.h> 62 63 int tcp_keepidle = TCPTV_KEEP_IDLE; 64 int tcp_keepintvl = TCPTV_KEEPINTVL; 65 int tcp_keepcnt = TCPTV_KEEPCNT; /* max idle probes */ 66 int tcp_maxpersistidle = TCPTV_KEEP_IDLE; /* max idle time in persist */ 67 int tcp_maxidle; 68 #else /* TUBA_INCLUDE */ 69 70 extern int tcp_keepcnt; 71 extern int tcp_maxpersistidle; 72 #endif /* TUBA_INCLUDE */ 73 74 /* 75 * Fast timeout routine for processing delayed acks 76 */ 77 void 78 tcp_fasttimo() 79 { 80 register struct inpcb *inp; 81 register struct tcpcb *tp; 82 int s; 83 84 s = splsoftnet(); 85 inp = tcbtable.inpt_queue.cqh_first; 86 if (inp) /* XXX */ 87 for (; inp != (struct inpcb *)&tcbtable.inpt_queue; 88 inp = inp->inp_queue.cqe_next) { 89 if ((tp = intotcpcb(inp)) != NULL && 90 (tp->t_flags & TF_DELACK)) { 91 tp->t_flags |= TF_ACKNOW; 92 (void) tcp_output(tp); 93 } 94 } 95 splx(s); 96 } 97 98 /* 99 * Tcp protocol timeout routine called every 500 ms. 100 * Updates the timers in all active tcb's and 101 * causes finite state machine actions if timers expire. 102 */ 103 void 104 tcp_slowtimo() 105 { 106 register struct inpcb *inp, *ninp; 107 register struct tcpcb *tp; 108 int s; 109 register long i; 110 static int syn_cache_last = 0; 111 112 s = splsoftnet(); 113 tcp_maxidle = tcp_keepcnt * tcp_keepintvl; 114 /* 115 * Search through tcb's and update active timers. 116 */ 117 inp = tcbtable.inpt_queue.cqh_first; 118 if (inp == (struct inpcb *)0) { /* XXX */ 119 splx(s); 120 return; 121 } 122 for (; inp != (struct inpcb *)&tcbtable.inpt_queue; inp = ninp) { 123 ninp = inp->inp_queue.cqe_next; 124 tp = intotcpcb(inp); 125 if (tp == 0 || tp->t_state == TCPS_LISTEN) 126 continue; 127 for (i = 0; i < TCPT_NTIMERS; i++) { 128 if (tp->t_timer[i] && --tp->t_timer[i] == 0) { 129 (void) tcp_usrreq(tp->t_inpcb->inp_socket, 130 PRU_SLOWTIMO, (struct mbuf *)0, 131 (struct mbuf *)i, (struct mbuf *)0, 132 (struct proc *)0); 133 /* XXX NOT MP SAFE */ 134 if ((ninp == (void *)&tcbtable.inpt_queue && 135 tcbtable.inpt_queue.cqh_last != inp) || 136 ninp->inp_queue.cqe_prev != inp) 137 goto tpgone; 138 } 139 } 140 tp->t_idle++; 141 if (tp->t_rtt) 142 tp->t_rtt++; 143 tpgone: 144 ; 145 } 146 #if NRND == 0 /* Do we need to do this when using random() ? */ 147 tcp_iss_seq += TCP_ISSINCR; /* increment iss */ 148 #ifdef TCP_COMPAT_42 149 if ((int)tcp_iss_seq < 0) 150 tcp_iss_seq = 0; /* XXX */ 151 #endif 152 #endif 153 tcp_now++; /* for timestamps */ 154 if (++syn_cache_last >= tcp_syn_cache_interval) { 155 syn_cache_timer(syn_cache_last); 156 syn_cache_last = 0; 157 } 158 splx(s); 159 } 160 #ifndef TUBA_INCLUDE 161 162 /* 163 * Cancel all timers for TCP tp. 164 */ 165 void 166 tcp_canceltimers(tp) 167 struct tcpcb *tp; 168 { 169 register int i; 170 171 for (i = 0; i < TCPT_NTIMERS; i++) 172 tp->t_timer[i] = 0; 173 } 174 175 int tcp_backoff[TCP_MAXRXTSHIFT + 1] = 176 { 1, 2, 4, 8, 16, 32, 64, 64, 64, 64, 64, 64, 64 }; 177 178 int tcp_totbackoff = 511; /* sum of tcp_backoff[] */ 179 180 /* 181 * TCP timer processing. 182 */ 183 struct tcpcb * 184 tcp_timers(tp, timer) 185 register struct tcpcb *tp; 186 int timer; 187 { 188 189 switch (timer) { 190 191 /* 192 * 2 MSL timeout in shutdown went off. If we're closed but 193 * still waiting for peer to close and connection has been idle 194 * too long, or if 2MSL time is up from TIME_WAIT, delete connection 195 * control block. Otherwise, check again in a bit. 196 */ 197 case TCPT_2MSL: 198 if (tp->t_state != TCPS_TIME_WAIT && 199 tp->t_idle <= tcp_maxidle) 200 tp->t_timer[TCPT_2MSL] = tcp_keepintvl; 201 else 202 tp = tcp_close(tp); 203 break; 204 205 /* 206 * Retransmission timer went off. Message has not 207 * been acked within retransmit interval. Back off 208 * to a longer retransmit interval and retransmit one segment. 209 */ 210 case TCPT_REXMT: 211 if (++tp->t_rxtshift > TCP_MAXRXTSHIFT) { 212 tp->t_rxtshift = TCP_MAXRXTSHIFT; 213 tcpstat.tcps_timeoutdrop++; 214 tp = tcp_drop(tp, tp->t_softerror ? 215 tp->t_softerror : ETIMEDOUT); 216 break; 217 } 218 tcpstat.tcps_rexmttimeo++; 219 TCPT_RANGESET(tp->t_rxtcur, 220 TCP_REXMTVAL(tp) * tcp_backoff[tp->t_rxtshift], 221 tp->t_rttmin, TCPTV_REXMTMAX); 222 tp->t_timer[TCPT_REXMT] = tp->t_rxtcur; 223 /* 224 * If losing, let the lower level know and try for 225 * a better route. Also, if we backed off this far, 226 * our srtt estimate is probably bogus. Clobber it 227 * so we'll take the next rtt measurement as our srtt; 228 * move the current srtt into rttvar to keep the current 229 * retransmit times until then. 230 */ 231 if (tp->t_rxtshift > TCP_MAXRXTSHIFT / 4) { 232 in_losing(tp->t_inpcb); 233 tp->t_rttvar += (tp->t_srtt >> TCP_RTT_SHIFT); 234 tp->t_srtt = 0; 235 } 236 tp->snd_nxt = tp->snd_una; 237 /* 238 * If timing a segment in this window, stop the timer. 239 */ 240 tp->t_rtt = 0; 241 /* 242 * Close the congestion window down to the initial window 243 * (we'll open it by one segment for each ack we get). 244 * Since we probably have a window's worth of unacked 245 * data accumulated, this "slow start" keeps us from 246 * dumping all that data as back-to-back packets (which 247 * might overwhelm an intermediate gateway). 248 * 249 * There are two phases to the opening: Initially we 250 * open by one mss on each ack. This makes the window 251 * size increase exponentially with time. If the 252 * window is larger than the path can handle, this 253 * exponential growth results in dropped packet(s) 254 * almost immediately. To get more time between 255 * drops but still "push" the network to take advantage 256 * of improving conditions, we switch from exponential 257 * to linear window opening at some threshhold size. 258 * For a threshhold, we use half the current window 259 * size, truncated to a multiple of the mss. 260 * 261 * (the minimum cwnd that will give us exponential 262 * growth is 2 mss. We don't allow the threshhold 263 * to go below this.) 264 */ 265 { 266 u_int win = min(tp->snd_wnd, tp->snd_cwnd) / 2 / tp->t_segsz; 267 if (win < 2) 268 win = 2; 269 tp->snd_cwnd = TCP_INITIAL_WINDOW(tp->t_segsz); 270 tp->snd_ssthresh = win * tp->t_segsz; 271 tp->t_dupacks = 0; 272 } 273 (void) tcp_output(tp); 274 break; 275 276 /* 277 * Persistance timer into zero window. 278 * Force a byte to be output, if possible. 279 */ 280 case TCPT_PERSIST: 281 /* 282 * Hack: if the peer is dead/unreachable, we do not 283 * time out if the window is closed. After a full 284 * backoff, drop the connection if the idle time 285 * (no responses to probes) reaches the maximum 286 * backoff that we would use if retransmitting. 287 */ 288 if (tp->t_rxtshift == TCP_MAXRXTSHIFT && 289 (tp->t_idle >= tcp_maxpersistidle || 290 tp->t_idle >= TCP_REXMTVAL(tp) * tcp_totbackoff)) { 291 tcpstat.tcps_persistdrops++; 292 tp = tcp_drop(tp, ETIMEDOUT); 293 break; 294 } 295 tcpstat.tcps_persisttimeo++; 296 tcp_setpersist(tp); 297 tp->t_force = 1; 298 (void) tcp_output(tp); 299 tp->t_force = 0; 300 break; 301 302 /* 303 * Keep-alive timer went off; send something 304 * or drop connection if idle for too long. 305 */ 306 case TCPT_KEEP: 307 tcpstat.tcps_keeptimeo++; 308 if (TCPS_HAVEESTABLISHED(tp->t_state) == 0) 309 goto dropit; 310 if (tp->t_inpcb->inp_socket->so_options & SO_KEEPALIVE && 311 tp->t_state <= TCPS_CLOSE_WAIT) { 312 if (tp->t_idle >= tcp_keepidle + tcp_maxidle) 313 goto dropit; 314 /* 315 * Send a packet designed to force a response 316 * if the peer is up and reachable: 317 * either an ACK if the connection is still alive, 318 * or an RST if the peer has closed the connection 319 * due to timeout or reboot. 320 * Using sequence number tp->snd_una-1 321 * causes the transmitted zero-length segment 322 * to lie outside the receive window; 323 * by the protocol spec, this requires the 324 * correspondent TCP to respond. 325 */ 326 tcpstat.tcps_keepprobe++; 327 #ifdef TCP_COMPAT_42 328 /* 329 * The keepalive packet must have nonzero length 330 * to get a 4.2 host to respond. 331 */ 332 (void)tcp_respond(tp, tp->t_template, 333 (struct mbuf *)NULL, tp->rcv_nxt - 1, 334 tp->snd_una - 1, 0); 335 #else 336 (void)tcp_respond(tp, tp->t_template, 337 (struct mbuf *)NULL, tp->rcv_nxt, 338 tp->snd_una - 1, 0); 339 #endif 340 tp->t_timer[TCPT_KEEP] = tcp_keepintvl; 341 } else 342 tp->t_timer[TCPT_KEEP] = tcp_keepidle; 343 break; 344 dropit: 345 tcpstat.tcps_keepdrops++; 346 tp = tcp_drop(tp, ETIMEDOUT); 347 break; 348 } 349 return (tp); 350 } 351 #endif /* TUBA_INCLUDE */ 352
Indexes created Tue Sep 23 14:10:03 GMT 2025