Annotation of 43BSDReno/sys/netinet/tcp_subr.c, revision 1.1.1.1

1.1       root        1: /*
                      2:  * Copyright (c) 1982, 1986, 1988, 1990 Regents of the University of California.
                      3:  * All rights reserved.
                      4:  *
                      5:  * Redistribution is only permitted until one year after the first shipment
                      6:  * of 4.4BSD by the Regents.  Otherwise, redistribution and use in source and
                      7:  * binary forms are permitted provided that: (1) source distributions retain
                      8:  * this entire copyright notice and comment, and (2) distributions including
                      9:  * binaries display the following acknowledgement:  This product includes
                     10:  * software developed by the University of California, Berkeley and its
                     11:  * contributors'' in the documentation or other materials provided with the
                     12:  * distribution and in all advertising materials mentioning features or use
                     13:  * of this software.  Neither the name of the University nor the names of
                     14:  * its contributors may be used to endorse or promote products derived from
                     15:  * this software without specific prior written permission.
                     16:  * THIS SOFTWARE IS PROVIDED AS IS'' AND WITHOUT ANY EXPRESS OR IMPLIED
                     17:  * WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED WARRANTIES OF
                     18:  * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE.
                     19:  *
                     20:  *     @(#)tcp_subr.c  7.19 (Berkeley) 7/25/90
                     21:  */
                     22: 
                     23: #include "param.h"
                     24: #include "systm.h"
                     25: #include "malloc.h"
                     26: #include "mbuf.h"
                     27: #include "socket.h"
                     28: #include "socketvar.h"
                     29: #include "protosw.h"
                     30: #include "errno.h"
                     31: 
                     32: #include "../net/route.h"
                     33: #include "../net/if.h"
                     34: 
                     35: #include "in.h"
                     36: #include "in_systm.h"
                     37: #include "ip.h"
                     38: #include "in_pcb.h"
                     39: #include "ip_var.h"
                     40: #include "ip_icmp.h"
                     41: #include "tcp.h"
                     42: #include "tcp_fsm.h"
                     43: #include "tcp_seq.h"
                     44: #include "tcp_timer.h"
                     45: #include "tcp_var.h"
                     46: #include "tcpip.h"
                     47: 
                     48: /* patchable/settable parameters for tcp */
                     49: int    tcp_ttl = TCP_TTL;
                     50: int    tcp_mssdflt = TCP_MSS;
                     51: int    tcp_rttdflt = TCPTV_SRTTDFLT / PR_SLOWHZ;
                     52: 
                     53: extern struct inpcb *tcp_last_inpcb;
                     54: 
                     55: /*
                     56:  * Tcp initialization
                     57:  */
                     58: tcp_init()
                     59: {
                     60: 
                     61:        tcp_iss = 1;            /* wrong */
                     62:        tcb.inp_next = tcb.inp_prev = &tcb;
                     63:        if (max_protohdr < sizeof(struct tcpiphdr))
                     64:                max_protohdr = sizeof(struct tcpiphdr);
                     65:        if (max_linkhdr + sizeof(struct tcpiphdr) > MHLEN)
                     66:                panic("tcp_init");
                     67: }
                     68: 
                     69: /*
                     70:  * Create template to be used to send tcp packets on a connection.
                     71:  * Call after host entry created, allocates an mbuf and fills
                     72:  * in a skeletal tcp/ip header, minimizing the amount of work
                     73:  * necessary when the connection is used.
                     74:  */
                     75: struct tcpiphdr *
                     76: tcp_template(tp)
                     77:        struct tcpcb *tp;
                     78: {
                     79:        register struct inpcb *inp = tp->t_inpcb;
                     80:        register struct mbuf *m;
                     81:        register struct tcpiphdr *n;
                     82: 
                     83:        if ((n = tp->t_template) == 0) {
                     84:                m = m_get(M_DONTWAIT, MT_HEADER);
                     85:                if (m == NULL)
                     86:                        return (0);
                     87:                m->m_len = sizeof (struct tcpiphdr);
                     88:                n = mtod(m, struct tcpiphdr *);
                     89:        }
                     90:        n->ti_next = n->ti_prev = 0;
                     91:        n->ti_x1 = 0;
                     92:        n->ti_pr = IPPROTO_TCP;
                     93:        n->ti_len = htons(sizeof (struct tcpiphdr) - sizeof (struct ip));
                     94:        n->ti_src = inp->inp_laddr;
                     95:        n->ti_dst = inp->inp_faddr;
                     96:        n->ti_sport = inp->inp_lport;
                     97:        n->ti_dport = inp->inp_fport;
                     98:        n->ti_seq = 0;
                     99:        n->ti_ack = 0;
                    100:        n->ti_x2 = 0;
                    101:        n->ti_off = 5;
                    102:        n->ti_flags = 0;
                    103:        n->ti_win = 0;
                    104:        n->ti_sum = 0;
                    105:        n->ti_urp = 0;
                    106:        return (n);
                    107: }
                    108: 
                    109: /*
                    110:  * Send a single message to the TCP at address specified by
                    111:  * the given TCP/IP header.  If m == 0, then we make a copy
                    112:  * of the tcpiphdr at ti and send directly to the addressed host.
                    113:  * This is used to force keep alive messages out using the TCP
                    114:  * template for a connection tp->t_template.  If flags are given
                    115:  * then we send a message back to the TCP which originated the
                    116:  * segment ti, and discard the mbuf containing it and any other
                    117:  * attached mbufs.
                    118:  *
                    119:  * In any case the ack and sequence number of the transmitted
                    120:  * segment are as specified by the parameters.
                    121:  */
                    122: tcp_respond(tp, ti, m, ack, seq, flags)
                    123:        struct tcpcb *tp;
                    124:        register struct tcpiphdr *ti;
                    125:        register struct mbuf *m;
                    126:        tcp_seq ack, seq;
                    127:        int flags;
                    128: {
                    129:        register int tlen;
                    130:        int win = 0;
                    131:        struct route *ro = 0;
                    132: 
                    133:        if (tp) {
                    134:                win = sbspace(&tp->t_inpcb->inp_socket->so_rcv);
                    135:                ro = &tp->t_inpcb->inp_route;
                    136:        }
                    137:        if (m == 0) {
                    138:                m = m_gethdr(M_DONTWAIT, MT_HEADER);
                    139:                if (m == NULL)
                    140:                        return;
                    141: #ifdef TCP_COMPAT_42
                    142:                tlen = 1;
                    143: #else
                    144:                tlen = 0;
                    145: #endif
                    146:                m->m_data += max_linkhdr;
                    147:                *mtod(m, struct tcpiphdr *) = *ti;
                    148:                ti = mtod(m, struct tcpiphdr *);
                    149:                flags = TH_ACK;
                    150:        } else {
                    151:                m_freem(m->m_next);
                    152:                m->m_next = 0;
                    153:                m->m_data = (caddr_t)ti;
                    154:                m->m_len = sizeof (struct tcpiphdr);
                    155:                tlen = 0;
                    156: #define xchg(a,b,type) { type t; t=a; a=b; b=t; }
                    157:                xchg(ti->ti_dst.s_addr, ti->ti_src.s_addr, u_long);
                    158:                xchg(ti->ti_dport, ti->ti_sport, u_short);
                    159: #undef xchg
                    160:        }
                    161:        ti->ti_len = htons((u_short)(sizeof (struct tcphdr) + tlen));
                    162:        tlen += sizeof (struct tcpiphdr);
                    163:        m->m_len = tlen;
                    164:        m->m_pkthdr.len = tlen;
                    165:        m->m_pkthdr.rcvif = (struct ifnet *) 0;
                    166:        ti->ti_next = ti->ti_prev = 0;
                    167:        ti->ti_x1 = 0;
                    168:        ti->ti_seq = htonl(seq);
                    169:        ti->ti_ack = htonl(ack);
                    170:        ti->ti_x2 = 0;
                    171:        ti->ti_off = sizeof (struct tcphdr) >> 2;
                    172:        ti->ti_flags = flags;
                    173:        ti->ti_win = htons((u_short)win);
                    174:        ti->ti_urp = 0;
                    175:        ti->ti_sum = in_cksum(m, tlen);
                    176:        ((struct ip *)ti)->ip_len = tlen;
                    177:        ((struct ip *)ti)->ip_ttl = tcp_ttl;
                    178:        (void) ip_output(m, (struct mbuf *)0, ro, 0);
                    179: }
                    180: 
                    181: /*
                    182:  * Create a new TCP control block, making an
                    183:  * empty reassembly queue and hooking it to the argument
                    184:  * protocol control block.
                    185:  */
                    186: struct tcpcb *
                    187: tcp_newtcpcb(inp)
                    188:        struct inpcb *inp;
                    189: {
                    190:        struct mbuf *m = m_getclr(M_DONTWAIT, MT_PCB);
                    191:        register struct tcpcb *tp;
                    192: 
                    193:        if (m == NULL)
                    194:                return ((struct tcpcb *)0);
                    195:        tp = mtod(m, struct tcpcb *);
                    196:        tp->seg_next = tp->seg_prev = (struct tcpiphdr *)tp;
                    197:        tp->t_maxseg = tcp_mssdflt;
                    198: 
                    199:        tp->t_flags = 0;                /* sends options! */
                    200:        tp->t_inpcb = inp;
                    201:        /*
                    202:         * Init srtt to TCPTV_SRTTBASE (0), so we can tell that we have no
                    203:         * rtt estimate.  Set rttvar so that srtt + 2 * rttvar gives
                    204:         * reasonable initial retransmit time.
                    205:         */
                    206:        tp->t_srtt = TCPTV_SRTTBASE;
                    207:        tp->t_rttvar = tcp_rttdflt * PR_SLOWHZ << 2;
                    208:        tp->t_rttmin = TCPTV_MIN;
                    209:        TCPT_RANGESET(tp->t_rxtcur, 
                    210:            ((TCPTV_SRTTBASE >> 2) + (TCPTV_SRTTDFLT << 2)) >> 1,
                    211:            TCPTV_MIN, TCPTV_REXMTMAX);
                    212:        tp->snd_cwnd = TCP_MAXWIN;
                    213:        tp->snd_ssthresh = TCP_MAXWIN;
                    214:        inp->inp_ip.ip_ttl = tcp_ttl;
                    215:        inp->inp_ppcb = (caddr_t)tp;
                    216:        return (tp);
                    217: }
                    218: 
                    219: /*
                    220:  * Drop a TCP connection, reporting
                    221:  * the specified error.  If connection is synchronized,
                    222:  * then send a RST to peer.
                    223:  */
                    224: struct tcpcb *
                    225: tcp_drop(tp, errno)
                    226:        register struct tcpcb *tp;
                    227:        int errno;
                    228: {
                    229:        struct socket *so = tp->t_inpcb->inp_socket;
                    230: 
                    231:        if (TCPS_HAVERCVDSYN(tp->t_state)) {
                    232:                tp->t_state = TCPS_CLOSED;
                    233:                (void) tcp_output(tp);
                    234:                tcpstat.tcps_drops++;
                    235:        } else
                    236:                tcpstat.tcps_conndrops++;
                    237:        if (errno == ETIMEDOUT && tp->t_softerror)
                    238:                errno = tp->t_softerror;
                    239:        so->so_error = errno;
                    240:        return (tcp_close(tp));
                    241: }
                    242: 
                    243: /*
                    244:  * Close a TCP control block:
                    245:  *     discard all space held by the tcp
                    246:  *     discard internet protocol block
                    247:  *     wake up any sleepers
                    248:  */
                    249: struct tcpcb *
                    250: tcp_close(tp)
                    251:        register struct tcpcb *tp;
                    252: {
                    253:        register struct tcpiphdr *t;
                    254:        struct inpcb *inp = tp->t_inpcb;
                    255:        struct socket *so = inp->inp_socket;
                    256:        register struct mbuf *m;
                    257: #ifdef RTV_RTT
                    258:        register struct rtentry *rt;
                    259: 
                    260:        /*
                    261:         * If we sent enough data to get some meaningful characteristics,
                    262:         * save them in the routing entry.  'Enough' is arbitrarily 
                    263:         * defined as 4K (default tcp_sendspace) * 16.  This would
                    264:         * give us 16 rtt samples assuming we only get one sample per
                    265:         * window (the usual case on a long haul net).  16 samples is
                    266:         * enough for the srtt filter to converge to within 5% of the correct
                    267:         * value; fewer samples and we could save a very bogus rtt.
                    268:         *
                    269:         * Don't update the default route's characteristics and don't
                    270:         * update anything that the user "locked".
                    271:         */
                    272:        if (SEQ_LT(tp->iss+(4096*16), tp->snd_max) &&
                    273:            (rt = inp->inp_route.ro_rt) &&
                    274:            ((struct sockaddr_in *) rt_key(rt))->sin_addr.s_addr !=
                    275:            INADDR_ANY) {
                    276:                register u_long i;
                    277: 
                    278:                if ((rt->rt_rmx.rmx_locks & RTV_RTT) == 0) {
                    279:                        i = tp->t_srtt *
                    280:                            (RTM_RTTUNIT / (PR_SLOWHZ * TCP_RTT_SCALE));
                    281:                        if (rt->rt_rmx.rmx_rtt && i)
                    282:                                /*
                    283:                                 * filter this update to half the old & half
                    284:                                 * the new values, converting scale.
                    285:                                 * See route.h and tcp_var.h for a
                    286:                                 * description of the scaling constants.
                    287:                                 */
                    288:                                rt->rt_rmx.rmx_rtt =
                    289:                                    (rt->rt_rmx.rmx_rtt + i) / 2;
                    290:                        else
                    291:                                rt->rt_rmx.rmx_rtt = i;
                    292:                }
                    293:                if ((rt->rt_rmx.rmx_locks & RTV_RTTVAR) == 0) {
                    294:                        i = tp->t_rttvar *
                    295:                            (RTM_RTTUNIT / (PR_SLOWHZ * TCP_RTTVAR_SCALE));
                    296:                        if (rt->rt_rmx.rmx_rttvar && i)
                    297:                                rt->rt_rmx.rmx_rttvar =
                    298:                                    (rt->rt_rmx.rmx_rttvar + i) / 2;
                    299:                        else
                    300:                                rt->rt_rmx.rmx_rttvar = i;
                    301:                }
                    302:                /*
                    303:                 * update the pipelimit (ssthresh) if it has been updated
                    304:                 * already or if a pipesize was specified & the threshhold
                    305:                 * got below half the pipesize.  I.e., wait for bad news
                    306:                 * before we start updating, then update on both good
                    307:                 * and bad news.
                    308:                 */
                    309:                if ((rt->rt_rmx.rmx_locks & RTV_SSTHRESH) == 0 &&
                    310:                    (i = tp->snd_ssthresh) && rt->rt_rmx.rmx_ssthresh ||
                    311:                    i < (rt->rt_rmx.rmx_sendpipe / 2)) {
                    312:                        /*
                    313:                         * convert the limit from user data bytes to
                    314:                         * packets then to packet data bytes.
                    315:                         */
                    316:                        i = (i + tp->t_maxseg / 2) / tp->t_maxseg;
                    317:                        if (i < 2)
                    318:                                i = 2;
                    319:                        i *= (u_long)(tp->t_maxseg + sizeof (struct tcpiphdr));
                    320:                        if (rt->rt_rmx.rmx_ssthresh)
                    321:                                rt->rt_rmx.rmx_ssthresh =
                    322:                                    (rt->rt_rmx.rmx_ssthresh + i) / 2;
                    323:                        else
                    324:                                rt->rt_rmx.rmx_ssthresh = i;
                    325:                }
                    326:        }
                    327: #endif RTV_RTT
                    328:        /* free the reassembly queue, if any */
                    329:        t = tp->seg_next;
                    330:        while (t != (struct tcpiphdr *)tp) {
                    331:                t = (struct tcpiphdr *)t->ti_next;
                    332:                m = REASS_MBUF((struct tcpiphdr *)t->ti_prev);
                    333:                remque(t->ti_prev);
                    334:                m_freem(m);
                    335:        }
                    336:        if (tp->t_template)
                    337:                (void) m_free(dtom(tp->t_template));
                    338:        (void) m_free(dtom(tp));
                    339:        inp->inp_ppcb = 0;
                    340:        soisdisconnected(so);
                    341:        /* clobber input pcb cache if we're closing the cached connection */
                    342:        if (inp == tcp_last_inpcb)
                    343:                tcp_last_inpcb = &tcb;
                    344:        in_pcbdetach(inp);
                    345:        tcpstat.tcps_closed++;
                    346:        return ((struct tcpcb *)0);
                    347: }
                    348: 
                    349: tcp_drain()
                    350: {
                    351: 
                    352: }
                    353: 
                    354: /*
                    355:  * Notify a tcp user of an asynchronous error;
                    356:  * store error as soft error, but wake up user
                    357:  * (for now, won't do anything until can select for soft error).
                    358:  */
                    359: tcp_notify(inp, error)
                    360:        register struct inpcb *inp;
                    361:        int error;
                    362: {
                    363: 
                    364:        ((struct tcpcb *)inp->inp_ppcb)->t_softerror = error;
                    365:        wakeup((caddr_t) &inp->inp_socket->so_timeo);
                    366:        sorwakeup(inp->inp_socket);
                    367:        sowwakeup(inp->inp_socket);
                    368: }
                    369: 
                    370: tcp_ctlinput(cmd, sa, ip)
                    371:        int cmd;
                    372:        struct sockaddr *sa;
                    373:        register struct ip *ip;
                    374: {
                    375:        register struct tcphdr *th;
                    376:        extern struct in_addr zeroin_addr;
                    377:        extern u_char inetctlerrmap[];
                    378:        int (*notify)() = tcp_notify, tcp_quench();
                    379: 
                    380:        if (cmd == PRC_QUENCH)
                    381:                notify = tcp_quench;
                    382:        else if ((unsigned)cmd > PRC_NCMDS || inetctlerrmap[cmd] == 0)
                    383:                return;
                    384:        if (ip) {
                    385:                th = (struct tcphdr *)((caddr_t)ip + (ip->ip_hl << 2));
                    386:                in_pcbnotify(&tcb, sa, th->th_dport, ip->ip_src, th->th_sport,
                    387:                        cmd, notify);
                    388:        } else
                    389:                in_pcbnotify(&tcb, sa, 0, zeroin_addr, 0, cmd, notify);
                    390: }
                    391: 
                    392: /*
                    393:  * When a source quench is received, close congestion window
                    394:  * to one segment.  We will gradually open it again as we proceed.
                    395:  */
                    396: tcp_quench(inp)
                    397:        struct inpcb *inp;
                    398: {
                    399:        struct tcpcb *tp = intotcpcb(inp);
                    400: 
                    401:        if (tp)
                    402:                tp->snd_cwnd = tp->t_maxseg;
                    403: }

unix.superglobalmegacorp.com

This archive runs on limited infrastructure. Preserving old code on modern bandwidth. Automated agents are requested to crawl responsibly.