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. Neither the name of the University nor the names of its contributors
  14  *    may be used to endorse or promote products derived from this software
  15  *    without specific prior written permission.
  16  *
  17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  27  * SUCH DAMAGE.
  28  *
  29  *      @(#)tcp_timer.c 8.1 (Berkeley) 6/10/93
  30  * tcp_timer.c,v 1.2 1994/08/02 07:49:10 davidg Exp
  31  */
  32
  33 #include <slirp.h>
  34
  35 u_int32_t        tcp_now;                /* for RFC 1323 timestamps */
  36
  37 static struct tcpcb *tcp_timers(register struct tcpcb *tp, int timer);
  38
  39 /*
  40  * Fast timeout routine for processing delayed acks
  41  */
  42 void
  43 tcp_fasttimo(void)
  44 {
  45         register struct socket *so;
  46         register struct tcpcb *tp;
  47
  48         DEBUG_CALL("tcp_fasttimo");
  49
  50         so = tcb.so_next;
  51         if (so)
  52         for (; so != &tcb; so = so->so_next)
  53                 if ((tp = (struct tcpcb *)so->so_tcpcb) &&
  54                     (tp->t_flags & TF_DELACK)) {
  55                         tp->t_flags &= ~TF_DELACK;
  56                         tp->t_flags |= TF_ACKNOW;
  57                         (void) tcp_output(tp);
  58                 }
  59 }
  60
  61 /*
  62  * Tcp protocol timeout routine called every 500 ms.
  63  * Updates the timers in all active tcb's and
  64  * causes finite state machine actions if timers expire.
  65  */
  66 void
  67 tcp_slowtimo(void)
  68 {
  69         register struct socket *ip, *ipnxt;
  70         register struct tcpcb *tp;
  71         register int i;
  72
  73         DEBUG_CALL("tcp_slowtimo");
  74
  75         /*
  76          * Search through tcb's and update active timers.
  77          */
  78         ip = tcb.so_next;
  79         if (ip == 0)
  80            return;
  81         for (; ip != &tcb; ip = ipnxt) {
  82                 ipnxt = ip->so_next;
  83                 tp = sototcpcb(ip);
  84                 if (tp == 0)
  85                         continue;
  86                 for (i = 0; i < TCPT_NTIMERS; i++) {
  87                         if (tp->t_timer[i] && --tp->t_timer[i] == 0) {
  88                                 tcp_timers(tp,i);
  89                                 if (ipnxt->so_prev != ip)
  90                                         goto tpgone;
  91                         }
  92                 }
  93                 tp->t_idle++;
  94                 if (tp->t_rtt)
  95                    tp->t_rtt++;
  96 tpgone:
  97                 ;
  98         }
  99         tcp_iss += TCP_ISSINCR/PR_SLOWHZ;               /* increment iss */
 100         tcp_now++;                                      /* for timestamps */
 101 }
 102
 103 /*
 104  * Cancel all timers for TCP tp.
 105  */
 106 void
 107 tcp_canceltimers(struct tcpcb *tp)
 108 {
 109         register int i;
 110
 111         for (i = 0; i < TCPT_NTIMERS; i++)
 112                 tp->t_timer[i] = 0;
 113 }
 114
 115 const int tcp_backoff[TCP_MAXRXTSHIFT + 1] =
 116    { 1, 2, 4, 8, 16, 32, 64, 64, 64, 64, 64, 64, 64 };
 117
 118 /*
 119  * TCP timer processing.
 120  */
 121 static struct tcpcb *
 122 tcp_timers(register struct tcpcb *tp, int timer)
 123 {
 124         register int rexmt;
 125
 126         DEBUG_CALL("tcp_timers");
 127
 128         switch (timer) {
 129
 130         /*
 131          * 2 MSL timeout in shutdown went off.  If we're closed but
 132          * still waiting for peer to close and connection has been idle
 133          * too long, or if 2MSL time is up from TIME_WAIT, delete connection
 134          * control block.  Otherwise, check again in a bit.
 135          */
 136         case TCPT_2MSL:
 137                 if (tp->t_state != TCPS_TIME_WAIT &&
 138                     tp->t_idle <= TCP_MAXIDLE)
 139                         tp->t_timer[TCPT_2MSL] = TCPTV_KEEPINTVL;
 140                 else
 141                         tp = tcp_close(tp);
 142                 break;
 143
 144         /*
 145          * Retransmission timer went off.  Message has not
 146          * been acked within retransmit interval.  Back off
 147          * to a longer retransmit interval and retransmit one segment.
 148          */
 149         case TCPT_REXMT:
 150
 151                 /*
 152                  * XXXXX If a packet has timed out, then remove all the queued
 153                  * packets for that session.
 154                  */
 155
 156                 if (++tp->t_rxtshift > TCP_MAXRXTSHIFT) {
 157                         /*
 158                          * This is a hack to suit our terminal server here at the uni of canberra
 159                          * since they have trouble with zeroes... It usually lets them through
 160                          * unharmed, but under some conditions, it'll eat the zeros.  If we
 161                          * keep retransmitting it, it'll keep eating the zeroes, so we keep
 162                          * retransmitting, and eventually the connection dies...
 163                          * (this only happens on incoming data)
 164                          *
 165                          * So, if we were gonna drop the connection from too many retransmits,
 166                          * don't... instead halve the t_maxseg, which might break up the NULLs and
 167                          * let them through
 168                          *
 169                          * *sigh*
 170                          */
 171
 172                         tp->t_maxseg >>= 1;
 173                         if (tp->t_maxseg < 32) {
 174                                 /*
 175                                  * We tried our best, now the connection must die!
 176                                  */
 177                                 tp->t_rxtshift = TCP_MAXRXTSHIFT;
 178                                 tp = tcp_drop(tp, tp->t_softerror);
 179                                 /* tp->t_softerror : ETIMEDOUT); */ /* XXX */
 180                                 return (tp); /* XXX */
 181                         }
 182
 183                         /*
 184                          * Set rxtshift to 6, which is still at the maximum
 185                          * backoff time
 186                          */
 187                         tp->t_rxtshift = 6;
 188                 }
 189                 rexmt = TCP_REXMTVAL(tp) * tcp_backoff[tp->t_rxtshift];
 190                 TCPT_RANGESET(tp->t_rxtcur, rexmt,
 191                     (short)tp->t_rttmin, TCPTV_REXMTMAX); /* XXX */
 192                 tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
 193                 /*
 194                  * If losing, let the lower level know and try for
 195                  * a better route.  Also, if we backed off this far,
 196                  * our srtt estimate is probably bogus.  Clobber it
 197                  * so we'll take the next rtt measurement as our srtt;
 198                  * move the current srtt into rttvar to keep the current
 199                  * retransmit times until then.
 200                  */
 201                 if (tp->t_rxtshift > TCP_MAXRXTSHIFT / 4) {
 202                         tp->t_rttvar += (tp->t_srtt >> TCP_RTT_SHIFT);
 203                         tp->t_srtt = 0;
 204                 }
 205                 tp->snd_nxt = tp->snd_una;
 206                 /*
 207                  * If timing a segment in this window, stop the timer.
 208                  */
 209                 tp->t_rtt = 0;
 210                 /*
 211                  * Close the congestion window down to one segment
 212                  * (we'll open it by one segment for each ack we get).
 213                  * Since we probably have a window's worth of unacked
 214                  * data accumulated, this "slow start" keeps us from
 215                  * dumping all that data as back-to-back packets (which
 216                  * might overwhelm an intermediate gateway).
 217                  *
 218                  * There are two phases to the opening: Initially we
 219                  * open by one mss on each ack.  This makes the window
 220                  * size increase exponentially with time.  If the
 221                  * window is larger than the path can handle, this
 222                  * exponential growth results in dropped packet(s)
 223                  * almost immediately.  To get more time between
 224                  * drops but still "push" the network to take advantage
 225                  * of improving conditions, we switch from exponential
 226                  * to linear window opening at some threshold size.
 227                  * For a threshold, we use half the current window
 228                  * size, truncated to a multiple of the mss.
 229                  *
 230                  * (the minimum cwnd that will give us exponential
 231                  * growth is 2 mss.  We don't allow the threshold
 232                  * to go below this.)
 233                  */
 234                 {
 235                 u_int win = min(tp->snd_wnd, tp->snd_cwnd) / 2 / tp->t_maxseg;
 236                 if (win < 2)
 237                         win = 2;
 238                 tp->snd_cwnd = tp->t_maxseg;
 239                 tp->snd_ssthresh = win * tp->t_maxseg;
 240                 tp->t_dupacks = 0;
 241                 }
 242                 (void) tcp_output(tp);
 243                 break;
 244
 245         /*
 246          * Persistence timer into zero window.
 247          * Force a byte to be output, if possible.
 248          */
 249         case TCPT_PERSIST:
 250                 tcp_setpersist(tp);
 251                 tp->t_force = 1;
 252                 (void) tcp_output(tp);
 253                 tp->t_force = 0;
 254                 break;
 255
 256         /*
 257          * Keep-alive timer went off; send something
 258          * or drop connection if idle for too long.
 259          */
 260         case TCPT_KEEP:
 261                 if (tp->t_state < TCPS_ESTABLISHED)
 262                         goto dropit;
 263
 264                 if ((SO_OPTIONS) && tp->t_state <= TCPS_CLOSE_WAIT) {
 265                         if (tp->t_idle >= TCPTV_KEEP_IDLE + TCP_MAXIDLE)
 266                                 goto dropit;
 267                         /*
 268                          * Send a packet designed to force a response
 269                          * if the peer is up and reachable:
 270                          * either an ACK if the connection is still alive,
 271                          * or an RST if the peer has closed the connection
 272                          * due to timeout or reboot.
 273                          * Using sequence number tp->snd_una-1
 274                          * causes the transmitted zero-length segment
 275                          * to lie outside the receive window;
 276                          * by the protocol spec, this requires the
 277                          * correspondent TCP to respond.
 278                          */
 279                         tcp_respond(tp, &tp->t_template, (struct mbuf *)NULL,
 280                             tp->rcv_nxt, tp->snd_una - 1, 0);
 281                         tp->t_timer[TCPT_KEEP] = TCPTV_KEEPINTVL;
 282                 } else
 283                         tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_IDLE;
 284                 break;
 285
 286         dropit:
 287                 tp = tcp_drop(tp, 0);
 288                 break;
 289         }
 290
 291         return (tp);
 292 }