1 /* 2 * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995 3 * The Regents of the University of California. 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 8.2 (Berkeley) 5/24/95 34 * $Id: tcp_timer.c,v 1.23 1997/08/02 14:32:57 bde Exp $ 35 */ 36 37 #include "opt_tcpdebug.h" 38 39 #ifndef TUBA_INCLUDE 40 #include <sys/param.h> 41 #include <sys/systm.h> 42 #include <sys/kernel.h> 43 #include <sys/sysctl.h> 44 #include <sys/socket.h> 45 #include <sys/socketvar.h> 46 #include <sys/protosw.h> 47 48 #include <machine/cpu.h> /* before tcp_seq.h, for tcp_random18() */ 49 50 #include <net/route.h> 51 52 #include <netinet/in.h> 53 #include <netinet/in_systm.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 #ifdef TCPDEBUG 63 #include <netinet/tcp_debug.h> 64 #endif 65 66 int tcp_keepinit = TCPTV_KEEP_INIT; 67 SYSCTL_INT(_net_inet_tcp, TCPCTL_KEEPINIT, keepinit, 68 CTLFLAG_RW, &tcp_keepinit , 0, ""); 69 70 int tcp_keepidle = TCPTV_KEEP_IDLE; 71 SYSCTL_INT(_net_inet_tcp, TCPCTL_KEEPIDLE, keepidle, 72 CTLFLAG_RW, &tcp_keepidle , 0, ""); 73 74 static int tcp_keepintvl = TCPTV_KEEPINTVL; 75 SYSCTL_INT(_net_inet_tcp, TCPCTL_KEEPINTVL, keepintvl, 76 CTLFLAG_RW, &tcp_keepintvl , 0, ""); 77 78 static int always_keepalive = 0; 79 SYSCTL_INT(_net_inet_tcp, OID_AUTO, always_keepalive, 80 CTLFLAG_RW, &always_keepalive , 0, ""); 81 82 static int tcp_keepcnt = TCPTV_KEEPCNT; 83 /* max idle probes */ 84 static int tcp_maxpersistidle = TCPTV_KEEP_IDLE; 85 /* max idle time in persist */ 86 int tcp_maxidle; 87 #else /* TUBA_INCLUDE */ 88 89 static int tcp_maxpersistidle; 90 #endif /* TUBA_INCLUDE */ 91 92 /* 93 * Fast timeout routine for processing delayed acks 94 */ 95 void 96 tcp_fasttimo() 97 { 98 register struct inpcb *inp; 99 register struct tcpcb *tp; 100 int s; 101 102 s = splnet(); 103 104 for (inp = tcb.lh_first; inp != NULL; inp = inp->inp_list.le_next) { 105 if ((tp = (struct tcpcb *)inp->inp_ppcb) && 106 (tp->t_flags & TF_DELACK)) { 107 tp->t_flags &= ~TF_DELACK; 108 tp->t_flags |= TF_ACKNOW; 109 tcpstat.tcps_delack++; 110 (void) tcp_output(tp); 111 } 112 } 113 splx(s); 114 } 115 116 /* 117 * Tcp protocol timeout routine called every 500 ms. 118 * Updates the timers in all active tcb's and 119 * causes finite state machine actions if timers expire. 120 */ 121 void 122 tcp_slowtimo() 123 { 124 register struct inpcb *ip, *ipnxt; 125 register struct tcpcb *tp; 126 register int i; 127 int s; 128 #ifdef TCPDEBUG 129 int ostate; 130 #endif 131 132 s = splnet(); 133 134 tcp_maxidle = tcp_keepcnt * tcp_keepintvl; 135 136 ip = tcb.lh_first; 137 if (ip == NULL) { 138 splx(s); 139 return; 140 } 141 /* 142 * Search through tcb's and update active timers. 143 */ 144 for (; ip != NULL; ip = ipnxt) { 145 ipnxt = ip->inp_list.le_next; 146 tp = intotcpcb(ip); 147 if (tp == 0 || tp->t_state == TCPS_LISTEN) 148 continue; 149 for (i = 0; i < TCPT_NTIMERS; i++) { 150 if (tp->t_timer[i] && --tp->t_timer[i] == 0) { 151 #ifdef TCPDEBUG 152 ostate = tp->t_state; 153 #endif 154 tp = tcp_timers(tp, i); 155 if (tp == NULL) 156 goto tpgone; 157 #ifdef TCPDEBUG 158 if (tp->t_inpcb->inp_socket->so_options 159 & SO_DEBUG) 160 tcp_trace(TA_USER, ostate, tp, 161 (struct tcpiphdr *)0, 162 PRU_SLOWTIMO); 163 #endif 164 } 165 } 166 tp->t_idle++; 167 tp->t_duration++; 168 if (tp->t_rtt) 169 tp->t_rtt++; 170 tpgone: 171 ; 172 } 173 tcp_iss += TCP_ISSINCR/PR_SLOWHZ; /* increment iss */ 174 #ifdef TCP_COMPAT_42 175 if ((int)tcp_iss < 0) 176 tcp_iss = TCP_ISSINCR; /* XXX */ 177 #endif 178 tcp_now++; /* for timestamps */ 179 splx(s); 180 } 181 #ifndef TUBA_INCLUDE 182 183 /* 184 * Cancel all timers for TCP tp. 185 */ 186 void 187 tcp_canceltimers(tp) 188 struct tcpcb *tp; 189 { 190 register int i; 191 192 for (i = 0; i < TCPT_NTIMERS; i++) 193 tp->t_timer[i] = 0; 194 } 195 196 int tcp_backoff[TCP_MAXRXTSHIFT + 1] = 197 { 1, 2, 4, 8, 16, 32, 64, 64, 64, 64, 64, 64, 64 }; 198 199 static int tcp_totbackoff = 511; /* sum of tcp_backoff[] */ 200 201 /* 202 * TCP timer processing. 203 */ 204 struct tcpcb * 205 tcp_timers(tp, timer) 206 register struct tcpcb *tp; 207 int timer; 208 { 209 register int rexmt; 210 211 switch (timer) { 212 213 /* 214 * 2 MSL timeout in shutdown went off. If we're closed but 215 * still waiting for peer to close and connection has been idle 216 * too long, or if 2MSL time is up from TIME_WAIT, delete connection 217 * control block. Otherwise, check again in a bit. 218 */ 219 case TCPT_2MSL: 220 if (tp->t_state != TCPS_TIME_WAIT && 221 tp->t_idle <= tcp_maxidle) 222 tp->t_timer[TCPT_2MSL] = tcp_keepintvl; 223 else 224 tp = tcp_close(tp); 225 break; 226 227 /* 228 * Retransmission timer went off. Message has not 229 * been acked within retransmit interval. Back off 230 * to a longer retransmit interval and retransmit one segment. 231 */ 232 case TCPT_REXMT: 233 if (++tp->t_rxtshift > TCP_MAXRXTSHIFT) { 234 tp->t_rxtshift = TCP_MAXRXTSHIFT; 235 tcpstat.tcps_timeoutdrop++; 236 tp = tcp_drop(tp, tp->t_softerror ? 237 tp->t_softerror : ETIMEDOUT); 238 break; 239 } 240 tcpstat.tcps_rexmttimeo++; 241 rexmt = TCP_REXMTVAL(tp) * tcp_backoff[tp->t_rxtshift]; 242 TCPT_RANGESET(tp->t_rxtcur, rexmt, 243 tp->t_rttmin, TCPTV_REXMTMAX); 244 tp->t_timer[TCPT_REXMT] = tp->t_rxtcur; 245 /* 246 * If losing, let the lower level know and try for 247 * a better route. Also, if we backed off this far, 248 * our srtt estimate is probably bogus. Clobber it 249 * so we'll take the next rtt measurement as our srtt; 250 * move the current srtt into rttvar to keep the current 251 * retransmit times until then. 252 */ 253 if (tp->t_rxtshift > TCP_MAXRXTSHIFT / 4) { 254 in_losing(tp->t_inpcb); 255 tp->t_rttvar += (tp->t_srtt >> TCP_RTT_SHIFT); 256 tp->t_srtt = 0; 257 } 258 tp->snd_nxt = tp->snd_una; 259 /* 260 * Force a segment to be sent. 261 */ 262 tp->t_flags |= TF_ACKNOW; 263 /* 264 * If timing a segment in this window, stop the timer. 265 */ 266 tp->t_rtt = 0; 267 /* 268 * Close the congestion window down to one segment 269 * (we'll open it by one segment for each ack we get). 270 * Since we probably have a window's worth of unacked 271 * data accumulated, this "slow start" keeps us from 272 * dumping all that data as back-to-back packets (which 273 * might overwhelm an intermediate gateway). 274 * 275 * There are two phases to the opening: Initially we 276 * open by one mss on each ack. This makes the window 277 * size increase exponentially with time. If the 278 * window is larger than the path can handle, this 279 * exponential growth results in dropped packet(s) 280 * almost immediately. To get more time between 281 * drops but still "push" the network to take advantage 282 * of improving conditions, we switch from exponential 283 * to linear window opening at some threshhold size. 284 * For a threshhold, we use half the current window 285 * size, truncated to a multiple of the mss. 286 * 287 * (the minimum cwnd that will give us exponential 288 * growth is 2 mss. We don't allow the threshhold 289 * to go below this.) 290 */ 291 { 292 u_int win = min(tp->snd_wnd, tp->snd_cwnd) / 2 / tp->t_maxseg; 293 if (win < 2) 294 win = 2; 295 tp->snd_cwnd = tp->t_maxseg; 296 tp->snd_ssthresh = win * tp->t_maxseg; 297 tp->t_dupacks = 0; 298 } 299 (void) tcp_output(tp); 300 break; 301 302 /* 303 * Persistance timer into zero window. 304 * Force a byte to be output, if possible. 305 */ 306 case TCPT_PERSIST: 307 tcpstat.tcps_persisttimeo++; 308 /* 309 * Hack: if the peer is dead/unreachable, we do not 310 * time out if the window is closed. After a full 311 * backoff, drop the connection if the idle time 312 * (no responses to probes) reaches the maximum 313 * backoff that we would use if retransmitting. 314 */ 315 if (tp->t_rxtshift == TCP_MAXRXTSHIFT) { 316 u_long maxidle = TCP_REXMTVAL(tp); 317 if (maxidle < tp->t_rttmin) 318 maxidle = tp->t_rttmin; 319 maxidle *= tcp_totbackoff; 320 if (tp->t_idle >= tcp_maxpersistidle || 321 tp->t_idle >= maxidle) { 322 tcpstat.tcps_persistdrop++; 323 tp = tcp_drop(tp, ETIMEDOUT); 324 break; 325 } 326 } 327 tcp_setpersist(tp); 328 tp->t_force = 1; 329 (void) tcp_output(tp); 330 tp->t_force = 0; 331 break; 332 333 /* 334 * Keep-alive timer went off; send something 335 * or drop connection if idle for too long. 336 */ 337 case TCPT_KEEP: 338 tcpstat.tcps_keeptimeo++; 339 if (tp->t_state < TCPS_ESTABLISHED) 340 goto dropit; 341 if ((always_keepalive || 342 tp->t_inpcb->inp_socket->so_options & SO_KEEPALIVE) && 343 tp->t_state <= TCPS_CLOSING) { 344 if (tp->t_idle >= tcp_keepidle + tcp_maxidle) 345 goto dropit; 346 /* 347 * Send a packet designed to force a response 348 * if the peer is up and reachable: 349 * either an ACK if the connection is still alive, 350 * or an RST if the peer has closed the connection 351 * due to timeout or reboot. 352 * Using sequence number tp->snd_una-1 353 * causes the transmitted zero-length segment 354 * to lie outside the receive window; 355 * by the protocol spec, this requires the 356 * correspondent TCP to respond. 357 */ 358 tcpstat.tcps_keepprobe++; 359 #ifdef TCP_COMPAT_42 360 /* 361 * The keepalive packet must have nonzero length 362 * to get a 4.2 host to respond. 363 */ 364 tcp_respond(tp, tp->t_template, (struct mbuf *)NULL, 365 tp->rcv_nxt - 1, tp->snd_una - 1, 0); 366 #else 367 tcp_respond(tp, tp->t_template, (struct mbuf *)NULL, 368 tp->rcv_nxt, tp->snd_una - 1, 0); 369 #endif 370 tp->t_timer[TCPT_KEEP] = tcp_keepintvl; 371 } else 372 tp->t_timer[TCPT_KEEP] = tcp_keepidle; 373 break; 374 dropit: 375 tcpstat.tcps_keepdrops++; 376 tp = tcp_drop(tp, ETIMEDOUT); 377 break; 378 } 379 return (tp); 380 } 381 #endif /* TUBA_INCLUDE */ 382