1 /* 2 * Copyright (c) 1982, 1986, 1988, 1990, 1993 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 8.1 (Berkeley) 6/10/93 34 * $Id: tcp_timer.c,v 1.4 1995/02/16 00:55:42 wollman Exp $ 35 */ 36 37 #ifndef TUBA_INCLUDE 38 #include <sys/param.h> 39 #include <sys/systm.h> 40 #include <sys/malloc.h> 41 #include <sys/mbuf.h> 42 #include <sys/socket.h> 43 #include <sys/socketvar.h> 44 #include <sys/protosw.h> 45 #include <sys/errno.h> 46 #include <sys/queue.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_maxidle; 66 #endif /* TUBA_INCLUDE */ 67 /* 68 * Fast timeout routine for processing delayed acks 69 */ 70 void 71 tcp_fasttimo() 72 { 73 register struct inpcb *inp; 74 register struct tcpcb *tp; 75 int s; 76 77 s = splnet(); 78 79 for (inp = tcb.lh_first; inp != NULL; inp = inp->inp_list.le_next) { 80 if ((tp = (struct tcpcb *)inp->inp_ppcb) && 81 (tp->t_flags & TF_DELACK)) { 82 tp->t_flags &= ~TF_DELACK; 83 tp->t_flags |= TF_ACKNOW; 84 tcpstat.tcps_delack++; 85 (void) tcp_output(tp); 86 } 87 } 88 splx(s); 89 } 90 91 /* 92 * Tcp protocol timeout routine called every 500 ms. 93 * Updates the timers in all active tcb's and 94 * causes finite state machine actions if timers expire. 95 */ 96 void 97 tcp_slowtimo() 98 { 99 register struct inpcb *ip, *ipnxt; 100 register struct tcpcb *tp; 101 register int i; 102 int s; 103 104 s = splnet(); 105 106 tcp_maxidle = TCPTV_KEEPCNT * tcp_keepintvl; 107 108 ip = tcb.lh_first; 109 if (ip == NULL) { 110 splx(s); 111 return; 112 } 113 /* 114 * Search through tcb's and update active timers. 115 */ 116 for (; ip != NULL; ip = ipnxt) { 117 ipnxt = ip->inp_list.le_next; 118 tp = intotcpcb(ip); 119 if (tp == 0) 120 continue; 121 for (i = 0; i < TCPT_NTIMERS; i++) { 122 if (tp->t_timer[i] && --tp->t_timer[i] == 0) { 123 (void) tcp_usrreq(tp->t_inpcb->inp_socket, 124 PRU_SLOWTIMO, (struct mbuf *)0, 125 (struct mbuf *)i, (struct mbuf *)0); 126 if (*ipnxt->inp_list.le_prev != ip) 127 goto tpgone; 128 } 129 } 130 tp->t_idle++; 131 tp->t_duration++; 132 if (tp->t_rtt) 133 tp->t_rtt++; 134 tpgone: 135 ; 136 } 137 tcp_iss += TCP_ISSINCR/PR_SLOWHZ; /* increment iss */ 138 #ifdef TCP_COMPAT_42 139 if ((int)tcp_iss < 0) 140 tcp_iss = 0; /* XXX */ 141 #endif 142 tcp_now++; /* for timestamps */ 143 splx(s); 144 } 145 #ifndef TUBA_INCLUDE 146 147 /* 148 * Cancel all timers for TCP tp. 149 */ 150 void 151 tcp_canceltimers(tp) 152 struct tcpcb *tp; 153 { 154 register int i; 155 156 for (i = 0; i < TCPT_NTIMERS; i++) 157 tp->t_timer[i] = 0; 158 } 159 160 int tcp_backoff[TCP_MAXRXTSHIFT + 1] = 161 { 1, 2, 4, 8, 16, 32, 64, 64, 64, 64, 64, 64, 64 }; 162 163 /* 164 * TCP timer processing. 165 */ 166 struct tcpcb * 167 tcp_timers(tp, timer) 168 register struct tcpcb *tp; 169 int timer; 170 { 171 register int rexmt; 172 173 switch (timer) { 174 175 /* 176 * 2 MSL timeout in shutdown went off. If we're closed but 177 * still waiting for peer to close and connection has been idle 178 * too long, or if 2MSL time is up from TIME_WAIT, delete connection 179 * control block. Otherwise, check again in a bit. 180 */ 181 case TCPT_2MSL: 182 if (tp->t_state != TCPS_TIME_WAIT && 183 tp->t_idle <= tcp_maxidle) 184 tp->t_timer[TCPT_2MSL] = tcp_keepintvl; 185 else 186 tp = tcp_close(tp); 187 break; 188 189 /* 190 * Retransmission timer went off. Message has not 191 * been acked within retransmit interval. Back off 192 * to a longer retransmit interval and retransmit one segment. 193 */ 194 case TCPT_REXMT: 195 if (++tp->t_rxtshift > TCP_MAXRXTSHIFT) { 196 tp->t_rxtshift = TCP_MAXRXTSHIFT; 197 tcpstat.tcps_timeoutdrop++; 198 tp = tcp_drop(tp, tp->t_softerror ? 199 tp->t_softerror : ETIMEDOUT); 200 break; 201 } 202 tcpstat.tcps_rexmttimeo++; 203 rexmt = TCP_REXMTVAL(tp) * tcp_backoff[tp->t_rxtshift]; 204 TCPT_RANGESET(tp->t_rxtcur, rexmt, 205 tp->t_rttmin, TCPTV_REXMTMAX); 206 tp->t_timer[TCPT_REXMT] = tp->t_rxtcur; 207 /* 208 * If losing, let the lower level know and try for 209 * a better route. Also, if we backed off this far, 210 * our srtt estimate is probably bogus. Clobber it 211 * so we'll take the next rtt measurement as our srtt; 212 * move the current srtt into rttvar to keep the current 213 * retransmit times until then. 214 */ 215 if (tp->t_rxtshift > TCP_MAXRXTSHIFT / 4) { 216 in_losing(tp->t_inpcb); 217 tp->t_rttvar += (tp->t_srtt >> TCP_RTT_SHIFT); 218 tp->t_srtt = 0; 219 } 220 tp->snd_nxt = tp->snd_una; 221 /* 222 * Force a segment to be sent. 223 */ 224 tp->t_flags |= TF_ACKNOW; 225 /* 226 * If timing a segment in this window, stop the timer. 227 */ 228 tp->t_rtt = 0; 229 /* 230 * Close the congestion window down to one segment 231 * (we'll open it by one segment for each ack we get). 232 * Since we probably have a window's worth of unacked 233 * data accumulated, this "slow start" keeps us from 234 * dumping all that data as back-to-back packets (which 235 * might overwhelm an intermediate gateway). 236 * 237 * There are two phases to the opening: Initially we 238 * open by one mss on each ack. This makes the window 239 * size increase exponentially with time. If the 240 * window is larger than the path can handle, this 241 * exponential growth results in dropped packet(s) 242 * almost immediately. To get more time between 243 * drops but still "push" the network to take advantage 244 * of improving conditions, we switch from exponential 245 * to linear window opening at some threshhold size. 246 * For a threshhold, we use half the current window 247 * size, truncated to a multiple of the mss. 248 * 249 * (the minimum cwnd that will give us exponential 250 * growth is 2 mss. We don't allow the threshhold 251 * to go below this.) 252 */ 253 { 254 u_int win = min(tp->snd_wnd, tp->snd_cwnd) / 2 / tp->t_maxseg; 255 if (win < 2) 256 win = 2; 257 tp->snd_cwnd = tp->t_maxseg; 258 tp->snd_ssthresh = win * tp->t_maxseg; 259 tp->t_dupacks = 0; 260 } 261 (void) tcp_output(tp); 262 break; 263 264 /* 265 * Persistance timer into zero window. 266 * Force a byte to be output, if possible. 267 */ 268 case TCPT_PERSIST: 269 tcpstat.tcps_persisttimeo++; 270 tcp_setpersist(tp); 271 tp->t_force = 1; 272 (void) tcp_output(tp); 273 tp->t_force = 0; 274 break; 275 276 /* 277 * Keep-alive timer went off; send something 278 * or drop connection if idle for too long. 279 */ 280 case TCPT_KEEP: 281 tcpstat.tcps_keeptimeo++; 282 if (tp->t_state < TCPS_ESTABLISHED) 283 goto dropit; 284 if (tp->t_inpcb->inp_socket->so_options & SO_KEEPALIVE && 285 tp->t_state <= TCPS_CLOSE_WAIT) { 286 if (tp->t_idle >= tcp_keepidle + tcp_maxidle) 287 goto dropit; 288 /* 289 * Send a packet designed to force a response 290 * if the peer is up and reachable: 291 * either an ACK if the connection is still alive, 292 * or an RST if the peer has closed the connection 293 * due to timeout or reboot. 294 * Using sequence number tp->snd_una-1 295 * causes the transmitted zero-length segment 296 * to lie outside the receive window; 297 * by the protocol spec, this requires the 298 * correspondent TCP to respond. 299 */ 300 tcpstat.tcps_keepprobe++; 301 #ifdef TCP_COMPAT_42 302 /* 303 * The keepalive packet must have nonzero length 304 * to get a 4.2 host to respond. 305 */ 306 tcp_respond(tp, tp->t_template, (struct mbuf *)NULL, 307 tp->rcv_nxt - 1, tp->snd_una - 1, 0); 308 #else 309 tcp_respond(tp, tp->t_template, (struct mbuf *)NULL, 310 tp->rcv_nxt, tp->snd_una - 1, 0); 311 #endif 312 tp->t_timer[TCPT_KEEP] = tcp_keepintvl; 313 } else 314 tp->t_timer[TCPT_KEEP] = tcp_keepidle; 315 break; 316 dropit: 317 tcpstat.tcps_keepdrops++; 318 tp = tcp_drop(tp, ETIMEDOUT); 319 break; 320 } 321 return (tp); 322 } 323 #endif /* TUBA_INCLUDE */ 324