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