Annotation of qemu/slirp/tcp_timer.c, revision 1.1.1.3

1.1       root        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.
1.1.1.3 ! root       13:  * 3. Neither the name of the University nor the names of its contributors
1.1       root       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: 
1.1.1.2   root       35: #ifdef LOG_ENABLED
1.1       root       36: struct   tcpstat tcpstat;        /* tcp statistics */
1.1.1.2   root       37: #endif
                     38: 
1.1       root       39: u_int32_t        tcp_now;                /* for RFC 1323 timestamps */
                     40: 
1.1.1.2   root       41: static struct tcpcb *tcp_timers(register struct tcpcb *tp, int timer);
                     42: 
1.1       root       43: /*
                     44:  * Fast timeout routine for processing delayed acks
                     45:  */
                     46: void
                     47: tcp_fasttimo()
                     48: {
                     49:        register struct socket *so;
                     50:        register struct tcpcb *tp;
                     51: 
                     52:        DEBUG_CALL("tcp_fasttimo");
1.1.1.2   root       53: 
1.1       root       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;
1.1.1.2   root       61:                        STAT(tcpstat.tcps_delack++);
1.1       root       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()
                     73: {
                     74:        register struct socket *ip, *ipnxt;
                     75:        register struct tcpcb *tp;
                     76:        register int i;
                     77: 
                     78:        DEBUG_CALL("tcp_slowtimo");
1.1.1.2   root       79: 
1.1       root       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: #ifdef TCP_COMPAT_42
                    106:        if ((int)tcp_iss < 0)
                    107:                tcp_iss = 0;                            /* XXX */
                    108: #endif
                    109:        tcp_now++;                                      /* for timestamps */
                    110: }
                    111: 
                    112: /*
                    113:  * Cancel all timers for TCP tp.
                    114:  */
                    115: void
                    116: tcp_canceltimers(tp)
                    117:        struct tcpcb *tp;
                    118: {
                    119:        register int i;
                    120: 
                    121:        for (i = 0; i < TCPT_NTIMERS; i++)
                    122:                tp->t_timer[i] = 0;
                    123: }
                    124: 
1.1.1.2   root      125: const int tcp_backoff[TCP_MAXRXTSHIFT + 1] =
1.1       root      126:    { 1, 2, 4, 8, 16, 32, 64, 64, 64, 64, 64, 64, 64 };
                    127: 
                    128: /*
                    129:  * TCP timer processing.
                    130:  */
1.1.1.2   root      131: static struct tcpcb *
                    132: tcp_timers(register struct tcpcb *tp, int timer)
1.1       root      133: {
                    134:        register int rexmt;
1.1.1.2   root      135: 
1.1       root      136:        DEBUG_CALL("tcp_timers");
1.1.1.2   root      137: 
1.1       root      138:        switch (timer) {
                    139: 
                    140:        /*
                    141:         * 2 MSL timeout in shutdown went off.  If we're closed but
                    142:         * still waiting for peer to close and connection has been idle
                    143:         * too long, or if 2MSL time is up from TIME_WAIT, delete connection
                    144:         * control block.  Otherwise, check again in a bit.
                    145:         */
                    146:        case TCPT_2MSL:
                    147:                if (tp->t_state != TCPS_TIME_WAIT &&
1.1.1.2   root      148:                    tp->t_idle <= TCP_MAXIDLE)
                    149:                        tp->t_timer[TCPT_2MSL] = TCPTV_KEEPINTVL;
1.1       root      150:                else
                    151:                        tp = tcp_close(tp);
                    152:                break;
                    153: 
                    154:        /*
                    155:         * Retransmission timer went off.  Message has not
                    156:         * been acked within retransmit interval.  Back off
                    157:         * to a longer retransmit interval and retransmit one segment.
                    158:         */
                    159:        case TCPT_REXMT:
1.1.1.2   root      160: 
1.1       root      161:                /*
                    162:                 * XXXXX If a packet has timed out, then remove all the queued
                    163:                 * packets for that session.
                    164:                 */
1.1.1.2   root      165: 
1.1       root      166:                if (++tp->t_rxtshift > TCP_MAXRXTSHIFT) {
                    167:                        /*
                    168:                         * This is a hack to suit our terminal server here at the uni of canberra
                    169:                         * since they have trouble with zeroes... It usually lets them through
                    170:                         * unharmed, but under some conditions, it'll eat the zeros.  If we
                    171:                         * keep retransmitting it, it'll keep eating the zeroes, so we keep
                    172:                         * retransmitting, and eventually the connection dies...
                    173:                         * (this only happens on incoming data)
1.1.1.2   root      174:                         *
1.1       root      175:                         * So, if we were gonna drop the connection from too many retransmits,
                    176:                         * don't... instead halve the t_maxseg, which might break up the NULLs and
                    177:                         * let them through
1.1.1.2   root      178:                         *
1.1       root      179:                         * *sigh*
                    180:                         */
1.1.1.2   root      181: 
1.1       root      182:                        tp->t_maxseg >>= 1;
                    183:                        if (tp->t_maxseg < 32) {
                    184:                                /*
                    185:                                 * We tried our best, now the connection must die!
                    186:                                 */
                    187:                                tp->t_rxtshift = TCP_MAXRXTSHIFT;
1.1.1.2   root      188:                                STAT(tcpstat.tcps_timeoutdrop++);
1.1       root      189:                                tp = tcp_drop(tp, tp->t_softerror);
                    190:                                /* tp->t_softerror : ETIMEDOUT); */ /* XXX */
                    191:                                return (tp); /* XXX */
                    192:                        }
1.1.1.2   root      193: 
1.1       root      194:                        /*
                    195:                         * Set rxtshift to 6, which is still at the maximum
                    196:                         * backoff time
                    197:                         */
                    198:                        tp->t_rxtshift = 6;
                    199:                }
1.1.1.2   root      200:                STAT(tcpstat.tcps_rexmttimeo++);
1.1       root      201:                rexmt = TCP_REXMTVAL(tp) * tcp_backoff[tp->t_rxtshift];
                    202:                TCPT_RANGESET(tp->t_rxtcur, rexmt,
                    203:                    (short)tp->t_rttmin, TCPTV_REXMTMAX); /* XXX */
                    204:                tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
                    205:                /*
                    206:                 * If losing, let the lower level know and try for
                    207:                 * a better route.  Also, if we backed off this far,
                    208:                 * our srtt estimate is probably bogus.  Clobber it
                    209:                 * so we'll take the next rtt measurement as our srtt;
                    210:                 * move the current srtt into rttvar to keep the current
                    211:                 * retransmit times until then.
                    212:                 */
                    213:                if (tp->t_rxtshift > TCP_MAXRXTSHIFT / 4) {
                    214: /*                     in_losing(tp->t_inpcb); */
                    215:                        tp->t_rttvar += (tp->t_srtt >> TCP_RTT_SHIFT);
                    216:                        tp->t_srtt = 0;
                    217:                }
                    218:                tp->snd_nxt = tp->snd_una;
                    219:                /*
                    220:                 * If timing a segment in this window, stop the timer.
                    221:                 */
                    222:                tp->t_rtt = 0;
                    223:                /*
                    224:                 * Close the congestion window down to one segment
                    225:                 * (we'll open it by one segment for each ack we get).
                    226:                 * Since we probably have a window's worth of unacked
                    227:                 * data accumulated, this "slow start" keeps us from
                    228:                 * dumping all that data as back-to-back packets (which
                    229:                 * might overwhelm an intermediate gateway).
                    230:                 *
                    231:                 * There are two phases to the opening: Initially we
                    232:                 * open by one mss on each ack.  This makes the window
                    233:                 * size increase exponentially with time.  If the
                    234:                 * window is larger than the path can handle, this
                    235:                 * exponential growth results in dropped packet(s)
1.1.1.2   root      236:                 * almost immediately.  To get more time between
1.1       root      237:                 * drops but still "push" the network to take advantage
                    238:                 * of improving conditions, we switch from exponential
                    239:                 * to linear window opening at some threshold size.
                    240:                 * For a threshold, we use half the current window
                    241:                 * size, truncated to a multiple of the mss.
                    242:                 *
                    243:                 * (the minimum cwnd that will give us exponential
                    244:                 * growth is 2 mss.  We don't allow the threshold
                    245:                 * to go below this.)
                    246:                 */
                    247:                {
                    248:                u_int win = min(tp->snd_wnd, tp->snd_cwnd) / 2 / tp->t_maxseg;
                    249:                if (win < 2)
                    250:                        win = 2;
                    251:                tp->snd_cwnd = tp->t_maxseg;
                    252:                tp->snd_ssthresh = win * tp->t_maxseg;
                    253:                tp->t_dupacks = 0;
                    254:                }
                    255:                (void) tcp_output(tp);
                    256:                break;
                    257: 
                    258:        /*
                    259:         * Persistence timer into zero window.
                    260:         * Force a byte to be output, if possible.
                    261:         */
                    262:        case TCPT_PERSIST:
1.1.1.2   root      263:                STAT(tcpstat.tcps_persisttimeo++);
1.1       root      264:                tcp_setpersist(tp);
                    265:                tp->t_force = 1;
                    266:                (void) tcp_output(tp);
                    267:                tp->t_force = 0;
                    268:                break;
                    269: 
                    270:        /*
                    271:         * Keep-alive timer went off; send something
                    272:         * or drop connection if idle for too long.
                    273:         */
                    274:        case TCPT_KEEP:
1.1.1.2   root      275:                STAT(tcpstat.tcps_keeptimeo++);
1.1       root      276:                if (tp->t_state < TCPS_ESTABLISHED)
                    277:                        goto dropit;
                    278: 
                    279: /*             if (tp->t_socket->so_options & SO_KEEPALIVE && */
1.1.1.2   root      280:                if ((SO_OPTIONS) && tp->t_state <= TCPS_CLOSE_WAIT) {
                    281:                        if (tp->t_idle >= TCPTV_KEEP_IDLE + TCP_MAXIDLE)
1.1       root      282:                                goto dropit;
                    283:                        /*
                    284:                         * Send a packet designed to force a response
                    285:                         * if the peer is up and reachable:
                    286:                         * either an ACK if the connection is still alive,
                    287:                         * or an RST if the peer has closed the connection
                    288:                         * due to timeout or reboot.
                    289:                         * Using sequence number tp->snd_una-1
                    290:                         * causes the transmitted zero-length segment
                    291:                         * to lie outside the receive window;
                    292:                         * by the protocol spec, this requires the
                    293:                         * correspondent TCP to respond.
                    294:                         */
1.1.1.2   root      295:                        STAT(tcpstat.tcps_keepprobe++);
1.1       root      296: #ifdef TCP_COMPAT_42
                    297:                        /*
                    298:                         * The keepalive packet must have nonzero length
                    299:                         * to get a 4.2 host to respond.
                    300:                         */
                    301:                        tcp_respond(tp, &tp->t_template, (struct mbuf *)NULL,
                    302:                            tp->rcv_nxt - 1, tp->snd_una - 1, 0);
                    303: #else
                    304:                        tcp_respond(tp, &tp->t_template, (struct mbuf *)NULL,
                    305:                            tp->rcv_nxt, tp->snd_una - 1, 0);
                    306: #endif
1.1.1.2   root      307:                        tp->t_timer[TCPT_KEEP] = TCPTV_KEEPINTVL;
1.1       root      308:                } else
1.1.1.2   root      309:                        tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_IDLE;
1.1       root      310:                break;
                    311: 
                    312:        dropit:
1.1.1.2   root      313:                STAT(tcpstat.tcps_keepdrops++);
1.1       root      314:                tp = tcp_drop(tp, 0); /* ETIMEDOUT); */
                    315:                break;
                    316:        }
                    317: 
                    318:        return (tp);
                    319: }

unix.superglobalmegacorp.com