slirp/tcp_timer.c

   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  * tcp_timer.c,v 1.2 1994/08/02 07:49:10 davidg Exp
  35  */
  36
  37 #include <slirp.h>
  38
  39 int     tcp_keepidle = TCPTV_KEEP_IDLE;
  40 int     tcp_keepintvl = TCPTV_KEEPINTVL;
  41 int     tcp_maxidle;
  42 int     so_options = DO_KEEPALIVE;
  43
  44 #ifdef LOG_ENABLED
  45 struct   tcpstat tcpstat;        /* tcp statistics */
  46 #endif
  47
  48 u_int32_t        tcp_now;                /* for RFC 1323 timestamps */
  49
  50 /*
  51  * Fast timeout routine for processing delayed acks
  52  */
  53 void
  54 tcp_fasttimo()
  55 {
  56         register struct socket *so;
  57         register struct tcpcb *tp;
  58
  59         DEBUG_CALL("tcp_fasttimo");
  60
  61         so = tcb.so_next;
  62         if (so)
  63         for (; so != &tcb; so = so->so_next)
  64                 if ((tp = (struct tcpcb *)so->so_tcpcb) &&
  65                     (tp->t_flags & TF_DELACK)) {
  66                         tp->t_flags &= ~TF_DELACK;
  67                         tp->t_flags |= TF_ACKNOW;
  68                         STAT(tcpstat.tcps_delack++);
  69                         (void) tcp_output(tp);
  70                 }
  71 }
  72
  73 /*
  74  * Tcp protocol timeout routine called every 500 ms.
  75  * Updates the timers in all active tcb's and
  76  * causes finite state machine actions if timers expire.
  77  */
  78 void
  79 tcp_slowtimo()
  80 {
  81         register struct socket *ip, *ipnxt;
  82         register struct tcpcb *tp;
  83         register int i;
  84
  85         DEBUG_CALL("tcp_slowtimo");
  86
  87         tcp_maxidle = TCPTV_KEEPCNT * tcp_keepintvl;
  88         /*
  89          * Search through tcb's and update active timers.
  90          */
  91         ip = tcb.so_next;
  92         if (ip == 0)
  93            return;
  94         for (; ip != &tcb; ip = ipnxt) {
  95                 ipnxt = ip->so_next;
  96                 tp = sototcpcb(ip);
  97                 if (tp == 0)
  98                         continue;
  99                 for (i = 0; i < TCPT_NTIMERS; i++) {
 100                         if (tp->t_timer[i] && --tp->t_timer[i] == 0) {
 101                                 tcp_timers(tp,i);
 102                                 if (ipnxt->so_prev != ip)
 103                                         goto tpgone;
 104                         }
 105                 }
 106                 tp->t_idle++;
 107                 if (tp->t_rtt)
 108                    tp->t_rtt++;
 109 tpgone:
 110                 ;
 111         }
 112         tcp_iss += TCP_ISSINCR/PR_SLOWHZ;               /* increment iss */
 113 #ifdef TCP_COMPAT_42
 114         if ((int)tcp_iss < 0)
 115                 tcp_iss = 0;                            /* XXX */
 116 #endif
 117         tcp_now++;                                      /* for timestamps */
 118 }
 119
 120 /*
 121  * Cancel all timers for TCP tp.
 122  */
 123 void
 124 tcp_canceltimers(tp)
 125         struct tcpcb *tp;
 126 {
 127         register int i;
 128
 129         for (i = 0; i < TCPT_NTIMERS; i++)
 130                 tp->t_timer[i] = 0;
 131 }
 132
 133 int     tcp_backoff[TCP_MAXRXTSHIFT + 1] =
 134    { 1, 2, 4, 8, 16, 32, 64, 64, 64, 64, 64, 64, 64 };
 135
 136 /*
 137  * TCP timer processing.
 138  */
 139 struct tcpcb *
 140 tcp_timers(tp, timer)
 141         register struct tcpcb *tp;
 142         int timer;
 143 {
 144         register int rexmt;
 145
 146         DEBUG_CALL("tcp_timers");
 147
 148         switch (timer) {
 149
 150         /*
 151          * 2 MSL timeout in shutdown went off.  If we're closed but
 152          * still waiting for peer to close and connection has been idle
 153          * too long, or if 2MSL time is up from TIME_WAIT, delete connection
 154          * control block.  Otherwise, check again in a bit.
 155          */
 156         case TCPT_2MSL:
 157                 if (tp->t_state != TCPS_TIME_WAIT &&
 158                     tp->t_idle <= tcp_maxidle)
 159                         tp->t_timer[TCPT_2MSL] = tcp_keepintvl;
 160                 else
 161                         tp = tcp_close(tp);
 162                 break;
 163
 164         /*
 165          * Retransmission timer went off.  Message has not
 166          * been acked within retransmit interval.  Back off
 167          * to a longer retransmit interval and retransmit one segment.
 168          */
 169         case TCPT_REXMT:
 170
 171                 /*
 172                  * XXXXX If a packet has timed out, then remove all the queued
 173                  * packets for that session.
 174                  */
 175
 176                 if (++tp->t_rxtshift > TCP_MAXRXTSHIFT) {
 177                         /*
 178                          * This is a hack to suit our terminal server here at the uni of canberra
 179                          * since they have trouble with zeroes... It usually lets them through
 180                          * unharmed, but under some conditions, it'll eat the zeros.  If we
 181                          * keep retransmitting it, it'll keep eating the zeroes, so we keep
 182                          * retransmitting, and eventually the connection dies...
 183                          * (this only happens on incoming data)
 184                          *
 185                          * So, if we were gonna drop the connection from too many retransmits,
 186                          * don't... instead halve the t_maxseg, which might break up the NULLs and
 187                          * let them through
 188                          *
 189                          * *sigh*
 190                          */
 191
 192                         tp->t_maxseg >>= 1;
 193                         if (tp->t_maxseg < 32) {
 194                                 /*
 195                                  * We tried our best, now the connection must die!
 196                                  */
 197                                 tp->t_rxtshift = TCP_MAXRXTSHIFT;
 198                                 STAT(tcpstat.tcps_timeoutdrop++);
 199                                 tp = tcp_drop(tp, tp->t_softerror);
 200                                 /* tp->t_softerror : ETIMEDOUT); */ /* XXX */
 201                                 return (tp); /* XXX */
 202                         }
 203
 204                         /*
 205                          * Set rxtshift to 6, which is still at the maximum
 206                          * backoff time
 207                          */
 208                         tp->t_rxtshift = 6;
 209                 }
 210                 STAT(tcpstat.tcps_rexmttimeo++);
 211                 rexmt = TCP_REXMTVAL(tp) * tcp_backoff[tp->t_rxtshift];
 212                 TCPT_RANGESET(tp->t_rxtcur, rexmt,
 213                     (short)tp->t_rttmin, TCPTV_REXMTMAX); /* XXX */
 214                 tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
 215                 /*
 216                  * If losing, let the lower level know and try for
 217                  * a better route.  Also, if we backed off this far,
 218                  * our srtt estimate is probably bogus.  Clobber it
 219                  * so we'll take the next rtt measurement as our srtt;
 220                  * move the current srtt into rttvar to keep the current
 221                  * retransmit times until then.
 222                  */
 223                 if (tp->t_rxtshift > TCP_MAXRXTSHIFT / 4) {
 224 /*                      in_losing(tp->t_inpcb); */
 225                         tp->t_rttvar += (tp->t_srtt >> TCP_RTT_SHIFT);
 226                         tp->t_srtt = 0;
 227                 }
 228                 tp->snd_nxt = tp->snd_una;
 229                 /*
 230                  * If timing a segment in this window, stop the timer.
 231                  */
 232                 tp->t_rtt = 0;
 233                 /*
 234                  * Close the congestion window down to one segment
 235                  * (we'll open it by one segment for each ack we get).
 236                  * Since we probably have a window's worth of unacked
 237                  * data accumulated, this "slow start" keeps us from
 238                  * dumping all that data as back-to-back packets (which
 239                  * might overwhelm an intermediate gateway).
 240                  *
 241                  * There are two phases to the opening: Initially we
 242                  * open by one mss on each ack.  This makes the window
 243                  * size increase exponentially with time.  If the
 244                  * window is larger than the path can handle, this
 245                  * exponential growth results in dropped packet(s)
 246                  * almost immediately.  To get more time between
 247                  * drops but still "push" the network to take advantage
 248                  * of improving conditions, we switch from exponential
 249                  * to linear window opening at some threshold size.
 250                  * For a threshold, we use half the current window
 251                  * size, truncated to a multiple of the mss.
 252                  *
 253                  * (the minimum cwnd that will give us exponential
 254                  * growth is 2 mss.  We don't allow the threshold
 255                  * to go below this.)
 256                  */
 257                 {
 258                 u_int win = min(tp->snd_wnd, tp->snd_cwnd) / 2 / tp->t_maxseg;
 259                 if (win < 2)
 260                         win = 2;
 261                 tp->snd_cwnd = tp->t_maxseg;
 262                 tp->snd_ssthresh = win * tp->t_maxseg;
 263                 tp->t_dupacks = 0;
 264                 }
 265                 (void) tcp_output(tp);
 266                 break;
 267
 268         /*
 269          * Persistence timer into zero window.
 270          * Force a byte to be output, if possible.
 271          */
 272         case TCPT_PERSIST:
 273                 STAT(tcpstat.tcps_persisttimeo++);
 274                 tcp_setpersist(tp);
 275                 tp->t_force = 1;
 276                 (void) tcp_output(tp);
 277                 tp->t_force = 0;
 278                 break;
 279
 280         /*
 281          * Keep-alive timer went off; send something
 282          * or drop connection if idle for too long.
 283          */
 284         case TCPT_KEEP:
 285                 STAT(tcpstat.tcps_keeptimeo++);
 286                 if (tp->t_state < TCPS_ESTABLISHED)
 287                         goto dropit;
 288
 289 /*              if (tp->t_socket->so_options & SO_KEEPALIVE && */
 290                 if ((so_options) && tp->t_state <= TCPS_CLOSE_WAIT) {
 291                         if (tp->t_idle >= tcp_keepidle + tcp_maxidle)
 292                                 goto dropit;
 293                         /*
 294                          * Send a packet designed to force a response
 295                          * if the peer is up and reachable:
 296                          * either an ACK if the connection is still alive,
 297                          * or an RST if the peer has closed the connection
 298                          * due to timeout or reboot.
 299                          * Using sequence number tp->snd_una-1
 300                          * causes the transmitted zero-length segment
 301                          * to lie outside the receive window;
 302                          * by the protocol spec, this requires the
 303                          * correspondent TCP to respond.
 304                          */
 305                         STAT(tcpstat.tcps_keepprobe++);
 306 #ifdef TCP_COMPAT_42
 307                         /*
 308                          * The keepalive packet must have nonzero length
 309                          * to get a 4.2 host to respond.
 310                          */
 311                         tcp_respond(tp, &tp->t_template, (struct mbuf *)NULL,
 312                             tp->rcv_nxt - 1, tp->snd_una - 1, 0);
 313 #else
 314                         tcp_respond(tp, &tp->t_template, (struct mbuf *)NULL,
 315                             tp->rcv_nxt, tp->snd_una - 1, 0);
 316 #endif
 317                         tp->t_timer[TCPT_KEEP] = tcp_keepintvl;
 318                 } else
 319                         tp->t_timer[TCPT_KEEP] = tcp_keepidle;
 320                 break;
 321
 322         dropit:
 323                 STAT(tcpstat.tcps_keepdrops++);
 324                 tp = tcp_drop(tp, 0); /* ETIMEDOUT); */
 325                 break;
 326         }
 327
 328         return (tp);
 329 }