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 1.10 1995/11/03 22:19:50 olah Exp $ 35 */ 36 37 #ifndef TUBA_INCLUDE 38 #include <sys/param.h> 39 #include <sys/systm.h> 40 #include <sys/kernel.h> 41 #include <sys/sysctl.h> 42 #include <sys/malloc.h> 43 #include <sys/mbuf.h> 44 #include <sys/socket.h> 45 #include <sys/socketvar.h> 46 #include <sys/protosw.h> 47 #include <sys/errno.h> 48 #include <sys/queue.h> 49 50 #include <machine/cpu.h> /* before tcp_seq.h, for tcp_random18() */ 51 52 #include <net/if.h> 53 #include <net/route.h> 54 55 #include <netinet/in.h> 56 #include <netinet/in_systm.h> 57 #include <netinet/ip.h> 58 #include <netinet/in_pcb.h> 59 #include <netinet/ip_var.h> 60 #include <netinet/tcp.h> 61 #include <netinet/tcp_fsm.h> 62 #include <netinet/tcp_seq.h> 63 #include <netinet/tcp_timer.h> 64 #include <netinet/tcp_var.h> 65 #include <netinet/tcpip.h> 66 67 int tcp_keepidle = TCPTV_KEEP_IDLE; 68 SYSCTL_INT(_net_inet_tcp, TCPCTL_KEEPIDLE, keepidle, 69 CTLFLAG_RW, &tcp_keepidle , 0, ""); 70 71 int tcp_keepintvl = TCPTV_KEEPINTVL; 72 SYSCTL_INT(_net_inet_tcp, TCPCTL_KEEPINTVL, keepintvl, 73 CTLFLAG_RW, &tcp_keepintvl , 0, ""); 74 75 int tcp_keepcnt = TCPTV_KEEPCNT; /* max idle probes */ 76 int tcp_maxpersistidle = TCPTV_KEEP_IDLE; /* max idle time in persist */ 77 int tcp_maxidle; 78 #else /* TUBA_INCLUDE */ 79 80 extern int tcp_maxpersistidle; 81 #endif /* TUBA_INCLUDE */ 82 83 /* 84 * Fast timeout routine for processing delayed acks 85 */ 86 void 87 tcp_fasttimo() 88 { 89 register struct inpcb *inp; 90 register struct tcpcb *tp; 91 int s; 92 93 s = splnet(); 94 95 for (inp = tcb.lh_first; inp != NULL; inp = inp->inp_list.le_next) { 96 if ((tp = (struct tcpcb *)inp->inp_ppcb) && 97 (tp->t_flags & TF_DELACK)) { 98 tp->t_flags &= ~TF_DELACK; 99 tp->t_flags |= TF_ACKNOW; 100 tcpstat.tcps_delack++; 101 (void) tcp_output(tp); 102 } 103 } 104 splx(s); 105 } 106 107 /* 108 * Tcp protocol timeout routine called every 500 ms. 109 * Updates the timers in all active tcb's and 110 * causes finite state machine actions if timers expire. 111 */ 112 void 113 tcp_slowtimo() 114 { 115 register struct inpcb *ip, *ipnxt; 116 register struct tcpcb *tp; 117 register int i; 118 int s; 119 120 s = splnet(); 121 122 tcp_maxidle = tcp_keepcnt * tcp_keepintvl; 123 124 ip = tcb.lh_first; 125 if (ip == NULL) { 126 splx(s); 127 return; 128 } 129 /* 130 * Search through tcb's and update active timers. 131 */ 132 for (; ip != NULL; ip = ipnxt) { 133 ipnxt = ip->inp_list.le_next; 134 tp = intotcpcb(ip); 135 if (tp == 0 || tp->t_state == TCPS_LISTEN) 136 continue; 137 for (i = 0; i < TCPT_NTIMERS; i++) { 138 if (tp->t_timer[i] && --tp->t_timer[i] == 0) { 139 if (tcp_usrreq(tp->t_inpcb->inp_socket, 140 PRU_SLOWTIMO, (struct mbuf *)0, 141 (struct mbuf *)i, (struct mbuf *)0) == NULL) 142 goto tpgone; 143 } 144 } 145 tp->t_idle++; 146 tp->t_duration++; 147 if (tp->t_rtt) 148 tp->t_rtt++; 149 tpgone: 150 ; 151 } 152 tcp_iss += TCP_ISSINCR/PR_SLOWHZ; /* increment iss */ 153 #ifdef TCP_COMPAT_42 154 if ((int)tcp_iss < 0) 155 tcp_iss = TCP_ISSINCR; /* XXX */ 156 #endif 157 tcp_now++; /* for timestamps */ 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 register int rexmt; 189 190 switch (timer) { 191 192 /* 193 * 2 MSL timeout in shutdown went off. If we're closed but 194 * still waiting for peer to close and connection has been idle 195 * too long, or if 2MSL time is up from TIME_WAIT, delete connection 196 * control block. Otherwise, check again in a bit. 197 */ 198 case TCPT_2MSL: 199 if (tp->t_state != TCPS_TIME_WAIT && 200 tp->t_idle <= tcp_maxidle) 201 tp->t_timer[TCPT_2MSL] = tcp_keepintvl; 202 else 203 tp = tcp_close(tp); 204 break; 205 206 /* 207 * Retransmission timer went off. Message has not 208 * been acked within retransmit interval. Back off 209 * to a longer retransmit interval and retransmit one segment. 210 */ 211 case TCPT_REXMT: 212 if (++tp->t_rxtshift > TCP_MAXRXTSHIFT) { 213 tp->t_rxtshift = TCP_MAXRXTSHIFT; 214 tcpstat.tcps_timeoutdrop++; 215 tp = tcp_drop(tp, tp->t_softerror ? 216 tp->t_softerror : ETIMEDOUT); 217 break; 218 } 219 tcpstat.tcps_rexmttimeo++; 220 rexmt = TCP_REXMTVAL(tp) * tcp_backoff[tp->t_rxtshift]; 221 TCPT_RANGESET(tp->t_rxtcur, rexmt, 222 tp->t_rttmin, TCPTV_REXMTMAX); 223 tp->t_timer[TCPT_REXMT] = tp->t_rxtcur; 224 /* 225 * If losing, let the lower level know and try for 226 * a better route. Also, if we backed off this far, 227 * our srtt estimate is probably bogus. Clobber it 228 * so we'll take the next rtt measurement as our srtt; 229 * move the current srtt into rttvar to keep the current 230 * retransmit times until then. 231 */ 232 if (tp->t_rxtshift > TCP_MAXRXTSHIFT / 4) { 233 in_losing(tp->t_inpcb); 234 tp->t_rttvar += (tp->t_srtt >> TCP_RTT_SHIFT); 235 tp->t_srtt = 0; 236 } 237 tp->snd_nxt = tp->snd_una; 238 /* 239 * If timing a segment in this window, stop the timer. 240 */ 241 tp->t_rtt = 0; 242 /* 243 * Close the congestion window down to one segment 244 * (we'll open it by one segment for each ack we get). 245 * Since we probably have a window's worth of unacked 246 * data accumulated, this "slow start" keeps us from 247 * dumping all that data as back-to-back packets (which 248 * might overwhelm an intermediate gateway). 249 * 250 * There are two phases to the opening: Initially we 251 * open by one mss on each ack. This makes the window 252 * size increase exponentially with time. If the 253 * window is larger than the path can handle, this 254 * exponential growth results in dropped packet(s) 255 * almost immediately. To get more time between 256 * drops but still "push" the network to take advantage 257 * of improving conditions, we switch from exponential 258 * to linear window opening at some threshhold size. 259 * For a threshhold, we use half the current window 260 * size, truncated to a multiple of the mss. 261 * 262 * (the minimum cwnd that will give us exponential 263 * growth is 2 mss. We don't allow the threshhold 264 * to go below this.) 265 */ 266 { 267 u_int win = min(tp->snd_wnd, tp->snd_cwnd) / 2 / tp->t_maxseg; 268 if (win < 2) 269 win = 2; 270 tp->snd_cwnd = tp->t_maxseg; 271 tp->snd_ssthresh = win * tp->t_maxseg; 272 tp->t_dupacks = 0; 273 } 274 (void) tcp_output(tp); 275 break; 276 277 /* 278 * Persistance timer into zero window. 279 * Force a byte to be output, if possible. 280 */ 281 case TCPT_PERSIST: 282 tcpstat.tcps_persisttimeo++; 283 /* 284 * Hack: if the peer is dead/unreachable, we do not 285 * time out if the window is closed. After a full 286 * backoff, drop the connection if the idle time 287 * (no responses to probes) reaches the maximum 288 * backoff that we would use if retransmitting. 289 */ 290 if (tp->t_rxtshift == TCP_MAXRXTSHIFT && 291 (tp->t_idle >= tcp_maxpersistidle || 292 tp->t_idle >= TCP_REXMTVAL(tp) * tcp_totbackoff)) { 293 tcpstat.tcps_persistdrop++; 294 tp = tcp_drop(tp, ETIMEDOUT); 295 break; 296 } 297 tcp_setpersist(tp); 298 tp->t_force = 1; 299 (void) tcp_output(tp); 300 tp->t_force = 0; 301 break; 302 303 /* 304 * Keep-alive timer went off; send something 305 * or drop connection if idle for too long. 306 */ 307 case TCPT_KEEP: 308 tcpstat.tcps_keeptimeo++; 309 if (tp->t_state < TCPS_ESTABLISHED) 310 goto dropit; 311 if (tp->t_inpcb->inp_socket->so_options & SO_KEEPALIVE && 312 tp->t_state <= TCPS_CLOSE_WAIT) { 313 if (tp->t_idle >= tcp_keepidle + tcp_maxidle) 314 goto dropit; 315 /* 316 * Send a packet designed to force a response 317 * if the peer is up and reachable: 318 * either an ACK if the connection is still alive, 319 * or an RST if the peer has closed the connection 320 * due to timeout or reboot. 321 * Using sequence number tp->snd_una-1 322 * causes the transmitted zero-length segment 323 * to lie outside the receive window; 324 * by the protocol spec, this requires the 325 * correspondent TCP to respond. 326 */ 327 tcpstat.tcps_keepprobe++; 328 #ifdef TCP_COMPAT_42 329 /* 330 * The keepalive packet must have nonzero length 331 * to get a 4.2 host to respond. 332 */ 333 tcp_respond(tp, tp->t_template, (struct mbuf *)NULL, 334 tp->rcv_nxt - 1, tp->snd_una - 1, 0); 335 #else 336 tcp_respond(tp, tp->t_template, (struct mbuf *)NULL, 337 tp->rcv_nxt, tp->snd_una - 1, 0); 338 #endif 339 tp->t_timer[TCPT_KEEP] = tcp_keepintvl; 340 } else 341 tp->t_timer[TCPT_KEEP] = tcp_keepidle; 342 break; 343 dropit: 344 tcpstat.tcps_keepdrops++; 345 tp = tcp_drop(tp, ETIMEDOUT); 346 break; 347 } 348 return (tp); 349 } 350 #endif /* TUBA_INCLUDE */ 351