| 1 | /* |
| 2 | * Copyright (c) 1982, 1986, 1988, 1990 Regents of the University of California. |
| 3 | * 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 | * from: @(#)tcp_timer.c 7.18 (Berkeley) 6/28/90 |
| 34 | * $Id: tcp_timer.c,v 1.3 1993/11/25 01:35:18 wollman Exp $ |
| 35 | */ |
| 36 | |
| 37 | #include "param.h" |
| 38 | #include "systm.h" |
| 39 | #include "malloc.h" |
| 40 | #include "mbuf.h" |
| 41 | #include "socket.h" |
| 42 | #include "socketvar.h" |
| 43 | #include "protosw.h" |
| 44 | #include "errno.h" |
| 45 | |
| 46 | #include "../net/if.h" |
| 47 | #include "../net/route.h" |
| 48 | |
| 49 | #include "in.h" |
| 50 | #include "in_systm.h" |
| 51 | #include "ip.h" |
| 52 | #include "in_pcb.h" |
| 53 | #include "ip_var.h" |
| 54 | #include "tcp.h" |
| 55 | #include "tcp_fsm.h" |
| 56 | #include "tcp_seq.h" |
| 57 | #include "tcp_timer.h" |
| 58 | #include "tcp_var.h" |
| 59 | #include "tcpip.h" |
| 60 | |
| 61 | /* |
| 62 | * Fast timeout routine for processing delayed acks |
| 63 | */ |
| 64 | void |
| 65 | tcp_fasttimo() |
| 66 | { |
| 67 | register struct inpcb *inp; |
| 68 | register struct tcpcb *tp; |
| 69 | int s = splnet(); |
| 70 | |
| 71 | inp = tcb.inp_next; |
| 72 | if (inp) |
| 73 | for (; inp != &tcb; inp = inp->inp_next) |
| 74 | if ((tp = (struct tcpcb *)inp->inp_ppcb) && |
| 75 | (tp->t_flags & TF_DELACK)) { |
| 76 | tp->t_flags &= ~TF_DELACK; |
| 77 | tp->t_flags |= TF_ACKNOW; |
| 78 | tcpstat.tcps_delack++; |
| 79 | (void) tcp_output(tp); |
| 80 | } |
| 81 | splx(s); |
| 82 | } |
| 83 | |
| 84 | /* |
| 85 | * Tcp protocol timeout routine called every 500 ms. |
| 86 | * Updates the timers in all active tcb's and |
| 87 | * causes finite state machine actions if timers expire. |
| 88 | */ |
| 89 | void |
| 90 | tcp_slowtimo() |
| 91 | { |
| 92 | register struct inpcb *ip, *ipnxt; |
| 93 | register struct tcpcb *tp; |
| 94 | int s = splnet(); |
| 95 | register int i; |
| 96 | |
| 97 | tcp_maxidle = TCPTV_KEEPCNT * tcp_keepintvl; |
| 98 | /* |
| 99 | * Search through tcb's and update active timers. |
| 100 | */ |
| 101 | ip = tcb.inp_next; |
| 102 | if (ip == 0) { |
| 103 | splx(s); |
| 104 | return; |
| 105 | } |
| 106 | for (; ip != &tcb; ip = ipnxt) { |
| 107 | ipnxt = ip->inp_next; |
| 108 | tp = intotcpcb(ip); |
| 109 | if (tp == 0) |
| 110 | continue; |
| 111 | for (i = 0; i < TCPT_NTIMERS; i++) { |
| 112 | if (tp->t_timer[i] && --tp->t_timer[i] == 0) { |
| 113 | (void) tcp_usrreq(tp->t_inpcb->inp_socket, |
| 114 | PRU_SLOWTIMO, (struct mbuf *)0, |
| 115 | (struct mbuf *)i, (struct mbuf *)0, |
| 116 | (struct mbuf *)0); |
| 117 | if (ipnxt->inp_prev != ip) |
| 118 | goto tpgone; |
| 119 | } |
| 120 | } |
| 121 | tp->t_idle++; |
| 122 | if (tp->t_rtt) |
| 123 | tp->t_rtt++; |
| 124 | tpgone: |
| 125 | ; |
| 126 | } |
| 127 | tcp_iss += TCP_ISSINCR/PR_SLOWHZ; /* increment iss */ |
| 128 | #ifdef TCP_COMPAT_42 |
| 129 | if ((int)tcp_iss < 0) |
| 130 | tcp_iss = 0; /* XXX */ |
| 131 | #endif |
| 132 | splx(s); |
| 133 | } |
| 134 | |
| 135 | /* |
| 136 | * Cancel all timers for TCP tp. |
| 137 | */ |
| 138 | void |
| 139 | tcp_canceltimers(tp) |
| 140 | struct tcpcb *tp; |
| 141 | { |
| 142 | register int i; |
| 143 | |
| 144 | for (i = 0; i < TCPT_NTIMERS; i++) |
| 145 | tp->t_timer[i] = 0; |
| 146 | } |
| 147 | |
| 148 | int tcp_backoff[TCP_MAXRXTSHIFT + 1] = |
| 149 | { 1, 2, 4, 8, 16, 32, 64, 64, 64, 64, 64, 64, 64 }; |
| 150 | |
| 151 | /* |
| 152 | * TCP timer processing. |
| 153 | */ |
| 154 | struct tcpcb * |
| 155 | tcp_timers(tp, timer) |
| 156 | register struct tcpcb *tp; |
| 157 | int timer; |
| 158 | { |
| 159 | register int rexmt; |
| 160 | |
| 161 | switch (timer) { |
| 162 | |
| 163 | /* |
| 164 | * 2 MSL timeout in shutdown went off. If we're closed but |
| 165 | * still waiting for peer to close and connection has been idle |
| 166 | * too long, or if 2MSL time is up from TIME_WAIT, delete connection |
| 167 | * control block. Otherwise, check again in a bit. |
| 168 | */ |
| 169 | case TCPT_2MSL: |
| 170 | if (tp->t_state != TCPS_TIME_WAIT && |
| 171 | tp->t_idle <= tcp_maxidle) |
| 172 | tp->t_timer[TCPT_2MSL] = tcp_keepintvl; |
| 173 | else |
| 174 | tp = tcp_close(tp); |
| 175 | break; |
| 176 | |
| 177 | /* |
| 178 | * Retransmission timer went off. Message has not |
| 179 | * been acked within retransmit interval. Back off |
| 180 | * to a longer retransmit interval and retransmit one segment. |
| 181 | */ |
| 182 | case TCPT_REXMT: |
| 183 | if (++tp->t_rxtshift > TCP_MAXRXTSHIFT) { |
| 184 | tp->t_rxtshift = TCP_MAXRXTSHIFT; |
| 185 | tcpstat.tcps_timeoutdrop++; |
| 186 | tp = tcp_drop(tp, tp->t_softerror ? |
| 187 | tp->t_softerror : ETIMEDOUT); |
| 188 | break; |
| 189 | } |
| 190 | tcpstat.tcps_rexmttimeo++; |
| 191 | rexmt = TCP_REXMTVAL(tp) * tcp_backoff[tp->t_rxtshift]; |
| 192 | TCPT_RANGESET(tp->t_rxtcur, rexmt, |
| 193 | tp->t_rttmin, TCPTV_REXMTMAX); |
| 194 | tp->t_timer[TCPT_REXMT] = tp->t_rxtcur; |
| 195 | /* |
| 196 | * If losing, let the lower level know and try for |
| 197 | * a better route. Also, if we backed off this far, |
| 198 | * our srtt estimate is probably bogus. Clobber it |
| 199 | * so we'll take the next rtt measurement as our srtt; |
| 200 | * move the current srtt into rttvar to keep the current |
| 201 | * retransmit times until then. |
| 202 | */ |
| 203 | if (tp->t_rxtshift > TCP_MAXRXTSHIFT / 4) { |
| 204 | in_losing(tp->t_inpcb); |
| 205 | tp->t_rttvar += (tp->t_srtt >> TCP_RTT_SHIFT); |
| 206 | tp->t_srtt = 0; |
| 207 | } |
| 208 | tp->snd_nxt = tp->snd_una; |
| 209 | /* |
| 210 | * If timing a segment in this window, stop the timer. |
| 211 | */ |
| 212 | tp->t_rtt = 0; |
| 213 | /* |
| 214 | * Close the congestion window down to one segment |
| 215 | * (we'll open it by one segment for each ack we get). |
| 216 | * Since we probably have a window's worth of unacked |
| 217 | * data accumulated, this "slow start" keeps us from |
| 218 | * dumping all that data as back-to-back packets (which |
| 219 | * might overwhelm an intermediate gateway). |
| 220 | * |
| 221 | * There are two phases to the opening: Initially we |
| 222 | * open by one mss on each ack. This makes the window |
| 223 | * size increase exponentially with time. If the |
| 224 | * window is larger than the path can handle, this |
| 225 | * exponential growth results in dropped packet(s) |
| 226 | * almost immediately. To get more time between |
| 227 | * drops but still "push" the network to take advantage |
| 228 | * of improving conditions, we switch from exponential |
| 229 | * to linear window opening at some threshhold size. |
| 230 | * For a threshhold, we use half the current window |
| 231 | * size, truncated to a multiple of the mss. |
| 232 | * |
| 233 | * (the minimum cwnd that will give us exponential |
| 234 | * growth is 2 mss. We don't allow the threshhold |
| 235 | * to go below this.) |
| 236 | */ |
| 237 | { |
| 238 | u_int win = min(tp->snd_wnd, tp->snd_cwnd) / 2 / tp->t_maxseg; |
| 239 | if (win < 2) |
| 240 | win = 2; |
| 241 | tp->snd_cwnd = tp->t_maxseg; |
| 242 | tp->snd_ssthresh = win * tp->t_maxseg; |
| 243 | tp->t_dupacks = 0; |
| 244 | } |
| 245 | (void) tcp_output(tp); |
| 246 | break; |
| 247 | |
| 248 | /* |
| 249 | * Persistance timer into zero window. |
| 250 | * Force a byte to be output, if possible. |
| 251 | */ |
| 252 | case TCPT_PERSIST: |
| 253 | tcpstat.tcps_persisttimeo++; |
| 254 | tcp_setpersist(tp); |
| 255 | tp->t_force = 1; |
| 256 | (void) tcp_output(tp); |
| 257 | tp->t_force = 0; |
| 258 | break; |
| 259 | |
| 260 | /* |
| 261 | * Keep-alive timer went off; send something |
| 262 | * or drop connection if idle for too long. |
| 263 | */ |
| 264 | case TCPT_KEEP: |
| 265 | tcpstat.tcps_keeptimeo++; |
| 266 | if (tp->t_state < TCPS_ESTABLISHED) |
| 267 | goto dropit; |
| 268 | if (tp->t_inpcb->inp_socket->so_options & SO_KEEPALIVE && |
| 269 | tp->t_state <= TCPS_CLOSE_WAIT) { |
| 270 | if (tp->t_idle >= tcp_keepidle + tcp_maxidle) |
| 271 | goto dropit; |
| 272 | /* |
| 273 | * Send a packet designed to force a response |
| 274 | * if the peer is up and reachable: |
| 275 | * either an ACK if the connection is still alive, |
| 276 | * or an RST if the peer has closed the connection |
| 277 | * due to timeout or reboot. |
| 278 | * Using sequence number tp->snd_una-1 |
| 279 | * causes the transmitted zero-length segment |
| 280 | * to lie outside the receive window; |
| 281 | * by the protocol spec, this requires the |
| 282 | * correspondent TCP to respond. |
| 283 | */ |
| 284 | tcpstat.tcps_keepprobe++; |
| 285 | #ifdef TCP_COMPAT_42 |
| 286 | /* |
| 287 | * The keepalive packet must have nonzero length |
| 288 | * to get a 4.2 host to respond. |
| 289 | */ |
| 290 | tcp_respond(tp, tp->t_template, (struct mbuf *)NULL, |
| 291 | tp->rcv_nxt - 1, tp->snd_una - 1, 0); |
| 292 | #else |
| 293 | tcp_respond(tp, tp->t_template, (struct mbuf *)NULL, |
| 294 | tp->rcv_nxt, tp->snd_una - 1, 0); |
| 295 | #endif |
| 296 | tp->t_timer[TCPT_KEEP] = tcp_keepintvl; |
| 297 | } else |
| 298 | tp->t_timer[TCPT_KEEP] = tcp_keepidle; |
| 299 | break; |
| 300 | dropit: |
| 301 | tcpstat.tcps_keepdrops++; |
| 302 | tp = tcp_drop(tp, ETIMEDOUT); |
| 303 | break; |
| 304 | } |
| 305 | return (tp); |
| 306 | } |