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