/* * Copyright (c) 1996, 1997 Berkeley Software Design, Inc. * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that this notice is retained, * the conditions in the following notices are met, and terms applying * to contributors in the following notices also apply to Berkeley * Software Design, Inc. * * BSDI tcp_subr.c,v 2.11 1997/01/16 14:06:35 karels Exp */ /* * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995 * The Regents of the University of California. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. All advertising materials mentioning features or use of this software * must display the following acknowledgement: * This product includes software developed by the University of * California, Berkeley and its contributors. * 4. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)tcp_subr.c 8.2 (Berkeley) 5/24/95 */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include /* patchable/settable parameters for tcp */ extern int tcp_mssdflt; extern int tcp_rttdflt; extern int tcp_do_rfc1323; extern struct inpcb *tcp_last_inpcb; /* * Tcp initialization */ void tcp_init() { tcp_iss = random(); /* wrong, but better than a constant */ tcb.inp_next = tcb.inp_prev = &tcb; if (max_protohdr < sizeof(struct tcpiphdr)) max_protohdr = sizeof(struct tcpiphdr); if (max_linkhdr + sizeof(struct tcpiphdr) > MHLEN) panic("tcp_init"); #ifdef FINE_GRAINED_TSTAMP { struct timeval curtime; microtime(&curtime); exact_time_at_slowtick = EXACT_TIME(&curtime); exact_time_at_fasttick = EXACT_TIME(&curtime); } #endif } /* * Create template to be used to send tcp packets on a connection. * Call after host entry created, allocates an mbuf and fills * in a skeletal tcp/ip header, minimizing the amount of work * necessary when the connection is used. */ struct tcpiphdr * tcp_template(tp) struct tcpcb *tp; { register struct inpcb *inp = tp->t_inpcb; register struct mbuf *m; register struct tcpiphdr *n; if ((n = tp->t_template) == 0) { m = m_get(M_DONTWAIT, MT_HEADER); if (m == NULL) return (0); m->m_len = sizeof (struct tcpiphdr); n = mtod(m, struct tcpiphdr *); } n->ti_next = n->ti_prev = 0; n->ti_x1 = 0; n->ti_pr = IPPROTO_TCP; n->ti_len = htons(sizeof (struct tcpiphdr) - sizeof (struct ip)); n->ti_src = inp->inp_laddr; n->ti_dst = inp->inp_faddr; n->ti_sport = inp->inp_lport; n->ti_dport = inp->inp_fport; n->ti_seq = 0; n->ti_ack = 0; n->ti_x2 = 0; n->ti_off = 5; n->ti_flags = 0; n->ti_win = 0; n->ti_sum = 0; n->ti_urp = 0; return (n); } /* * Send a single message to the TCP at address specified by * the given TCP/IP header. If m == 0, then we make a copy * of the tcpiphdr at ti and send directly to the addressed host. * This is used to force keep alive messages out using the TCP * template for a connection tp->t_template. If flags are given * then we send a message back to the TCP which originated the * segment ti, and discard the mbuf containing it and any other * attached mbufs. * * In any case the ack and sequence number of the transmitted * segment are as specified by the parameters. */ int tcp_respond(tp, ti, m, ack, seq, flags) struct tcpcb *tp; register struct tcpiphdr *ti; register struct mbuf *m; tcp_seq ack, seq; int flags; { register int tlen, optlen = 0; int win = 0; struct route *ro = 0; if (tp) { win = sbspace(&tp->t_inpcb->inp_socket->so_rcv); ro = &tp->t_inpcb->inp_route; } if (m == 0) { m = m_gethdr(M_DONTWAIT, MT_HEADER); if (m == NULL) return(ENOBUFS); #ifdef TCP_COMPAT_42 tlen = 1; #else tlen = 0; #endif m->m_data += max_linkhdr; *mtod(m, struct tcpiphdr *) = *ti; ti = mtod(m, struct tcpiphdr *); flags = TH_ACK; } else { m_freem(m->m_next); m->m_next = 0; m->m_data = (caddr_t)ti; m->m_len = sizeof (struct tcpiphdr); tlen = 0; #define xchg(a,b,type) { type t; t=a; a=b; b=t; } xchg(ti->ti_dst.s_addr, ti->ti_src.s_addr, u_long); xchg(ti->ti_dport, ti->ti_sport, u_short); #undef xchg } ti->ti_next = ti->ti_prev = 0; ti->ti_x1 = 0; ti->ti_seq = htonl(seq); ti->ti_ack = htonl(ack); ti->ti_x2 = 0; if ((flags & TH_SYN) == 0) { if (tp) ti->ti_win = htons((u_short) (win >> tp->rcv_scale)); else ti->ti_win = htons((u_short)win); ti->ti_off = sizeof (struct tcphdr) >> 2; tlen += sizeof (struct tcphdr); } else tlen += ti->ti_off << 2; ti->ti_len = htons((u_short)tlen); tlen += sizeof (struct ip); m->m_len = tlen; m->m_pkthdr.len = tlen; m->m_pkthdr.rcvif = (struct ifnet *) 0; ti->ti_flags = flags; ti->ti_urp = 0; ti->ti_sum = 0; ti->ti_sum = in_cksum(m, tlen); ((struct ip *)ti)->ip_len = tlen; ((struct ip *)ti)->ip_ttl = ip_defttl; return(ip_output(m, NULL, ro, 0, NULL)); } /* * Create a new TCP control block, making an * empty reassembly queue and hooking it to the argument * protocol control block. */ struct tcpcb * tcp_newtcpcb(inp) struct inpcb *inp; { register struct tcpcb *tp; tp = malloc(sizeof(*tp), M_PCB, M_NOWAIT); if (tp == NULL) return ((struct tcpcb *)0); bzero((char *) tp, sizeof(struct tcpcb)); tp->seg_next = tp->seg_prev = (struct tcpiphdr *)tp; tp->t_maxseg = tcp_mssdflt; tp->t_peermaxseg = tcp_mssdflt; tp->t_flags = tcp_do_rfc1323 ? (TF_USE_SCALE|TF_SEND_TSTMP) : 0; tp->t_xflags = 0; tp->t_dupacks = 0; tp->t_inpcb = inp; /* * Init srtt to TCPTV_SRTTBASE (0), so we can tell that we have no * rtt estimate. Set rttvar so that srtt + 4 * rttvar gives * reasonable initial retransmit time (tcp_rttdflt seconds). */ tp->t_srtt = TCPTV_SRTTBASE; #ifdef FINE_GRAINED_TSTAMP /* * XXX It is okay to set both these to zero because proper * initialization will be done in tcp_xmit_timer_exact() before these * values are used to set a timer. */ tp->t_srtt_exact = TCPTV_SRTTBASE_EXACT; tp->t_rttvar_exact = 0; tp->t_rxtcur_exact = tp->t_srtt_exact + (tp->t_rttvar_exact << 2); #endif #ifdef TCP_STANDARD tp->num_segs_since_ack = 0; #endif tp->t_rttvar = tcp_rttdflt * PR_SLOWHZ /* / 4 << 2 */; tp->t_rttmin = TCPTV_MIN; TCPT_RANGESET(tp->t_rxtcur, TCP_REXMTVAL(tp), TCPTV_MIN, TCPTV_REXMTMAX); tp->snd_cwnd = TCP_MAXWIN << TCP_MAX_WINSHIFT; tp->snd_ssthresh = TCP_MAXWIN << TCP_MAX_WINSHIFT; #if defined(TCP_STATS) && defined(TOMH) tp->last_cwnd = 0; #endif /* * XXX Only non-zero fields need to be initialized because of the bzero(). */ #ifdef ACC tp->delack = 2; /* delayed ack factor */ tp->max_delack = 0; /* max. delayed ack factor */ tp->peer_win_in_segs = 0; /* size of peer's window (in segs) */ tp->min_acks_per_win = 3; /* minimum # of acks per window */ #endif #ifdef LIMIT_BURST tp->maxburst = 0; tp->burst_pending = 0; #endif inp->inp_ip.ip_ttl = ip_defttl; inp->inp_ppcb = (caddr_t)tp; return (tp); } /* * Drop a TCP connection, reporting * the specified error. If connection is synchronized, * then send a RST to peer. */ struct tcpcb * tcp_drop(tp, errno) register struct tcpcb *tp; int errno; { struct socket *so = tp->t_inpcb->inp_socket; if (TCPS_HAVERCVDSYN(tp->t_state)) { tp->t_state = TCPS_CLOSED; (void) tcp_output(tp); tcpstat.tcps_drops++; } else tcpstat.tcps_conndrops++; if (errno == ETIMEDOUT && tp->t_softerror) errno = tp->t_softerror; so->so_error = errno; return (tcp_close(tp)); } /* * Close a TCP control block: * discard all space held by the tcp * discard internet protocol block * wake up any sleepers */ struct tcpcb * tcp_close(tp) register struct tcpcb *tp; { register struct tcpiphdr *t; struct inpcb *inp = tp->t_inpcb; struct socket *so = inp->inp_socket; register struct mbuf *m; #ifdef RTV_RTT register struct rtentry *rt; /* * If we sent enough data to get some meaningful characteristics, * save them in the routing entry. 'Enough' is arbitrarily * defined as the sendpipesize (default 4K) * 16. This would * give us 16 rtt samples assuming we only get one sample per * window (the usual case on a long haul net). 16 samples is * enough for the srtt filter to converge to within 5% of the correct * value; fewer samples and we could save a very bogus rtt. * * Don't update the default route's characteristics and don't * update anything that the user "locked". */ if (SEQ_LT(tp->iss + so->so_snd.sb_hiwat * 16, tp->snd_max) && (rt = inp->inp_route.ro_rt) && ((struct sockaddr_in *)rt_key(rt))->sin_addr.s_addr != INADDR_ANY) { register u_long i; if ((rt->rt_rmx.rmx_locks & RTV_RTT) == 0) { i = tp->t_srtt * (RTM_RTTUNIT / (PR_SLOWHZ * TCP_RTT_SCALE)); if (rt->rt_rmx.rmx_rtt && i) /* * filter this update to half the old & half * the new values, converting scale. * See route.h and tcp_var.h for a * description of the scaling constants. */ rt->rt_rmx.rmx_rtt = (rt->rt_rmx.rmx_rtt + i) / 2; else rt->rt_rmx.rmx_rtt = i; } if ((rt->rt_rmx.rmx_locks & RTV_RTTVAR) == 0) { i = tp->t_rttvar * (RTM_RTTUNIT / (PR_SLOWHZ * TCP_RTTVAR_SCALE)); if (rt->rt_rmx.rmx_rttvar && i) rt->rt_rmx.rmx_rttvar = (rt->rt_rmx.rmx_rttvar + i) / 2; else rt->rt_rmx.rmx_rttvar = i; } /* * update the pipelimit (ssthresh) if it has been updated * already or if a pipesize was specified & the threshhold * got below half the pipesize. I.e., wait for bad news * before we start updating, then update on both good * and bad news. */ if ((rt->rt_rmx.rmx_locks & RTV_SSTHRESH) == 0 && (i = tp->snd_ssthresh) && rt->rt_rmx.rmx_ssthresh || i < (rt->rt_rmx.rmx_sendpipe / 2)) { /* * convert the limit from user data bytes to * packets then to packet data bytes. */ i = (i + tp->t_maxseg / 2) / tp->t_maxseg; if (i < 2) i = 2; i *= (u_long)(tp->t_maxseg + sizeof (struct tcpiphdr)); if (rt->rt_rmx.rmx_ssthresh) rt->rt_rmx.rmx_ssthresh = (rt->rt_rmx.rmx_ssthresh + i) / 2; else rt->rt_rmx.rmx_ssthresh = i; } } #endif /* RTV_RTT */ /* free the reassembly queue, if any */ t = tp->seg_next; while (t != (struct tcpiphdr *)tp) { t = (struct tcpiphdr *)t->ti_next; m = REASS_MBUF((struct tcpiphdr *)t->ti_prev); remque(t->ti_prev); m_freem(m); } if (tp->t_template) (void) m_free(dtom(tp->t_template)); tcp_delack_done(tp); /* just in case */ if (tp->t_state == TCPS_TIME_WAIT) tcp_cancel2msl(inp, tp); free(tp, M_PCB); inp->inp_ppcb = 0; /* * Flush any remaining data. We might have data in the output queue * if connection was reset, and the socket might not be closed * for a while if user does not notice. */ sbdrop(&so->so_snd, (int)so->so_snd.sb_cc); soisdisconnected(so); /* clobber input pcb cache if we're closing the cached connection */ if (inp == tcp_last_inpcb) tcp_last_inpcb = &tcb; if (inp->inp_hlist.le_prev) /* XXX */ LIST_REMOVE(inp, inp_hlist); in_pcbdetach(inp); tcpstat.tcps_closed++; return ((struct tcpcb *)0); } void tcp_drain() { } /* * Notify a tcp user of an asynchronous error; * store error as soft error, but wake up user * (for now, won't do anything until can select for soft error). */ void tcp_notify(inp, error) struct inpcb *inp; int error; { register struct tcpcb *tp = (struct tcpcb *)inp->inp_ppcb; register struct socket *so = inp->inp_socket; /* * If we are hooked up. do not report errors directly, * but record them as soft errors in case we time out. * If connection hasn't completed, has retransmitted several times, * and receives a second error, give up now. This is better * than waiting a long time to establish a connection that * can never complete. */ #if 0 if (tp->t_state == TCPS_ESTABLISHED && (error == EHOSTUNREACH || error == ENETUNREACH || error == EHOSTDOWN)) { tp->t_softerror = error; return; } else #endif if (tp->t_state < TCPS_ESTABLISHED && tp->t_rxtshift >= 3 && tp->t_softerror) so->so_error = error; else tp->t_softerror = error; wakeup((caddr_t) &so->so_timeo); sorwakeup(so); sowwakeup(so); } void tcp_ctlinput(cmd, sa, ip) int cmd; struct sockaddr *sa; register struct ip *ip; { register struct tcphdr *th; extern struct in_addr zeroin_addr; extern u_char inetctlerrmap[]; void (*notify) __P((struct inpcb *, int)) = tcp_notify; int nmatch; if (cmd == PRC_QUENCH) notify = tcp_quench; else if (cmd == PRC_IFNEWADDR) in_mrejoin(&tcb); else if (cmd == PRC_MSGSIZE) notify = tcp_mtudisc; else if (!PRC_IS_REDIRECT(cmd) && ((unsigned)cmd > PRC_NCMDS || inetctlerrmap[cmd] == 0)) return; if (ip) { th = (struct tcphdr *)((caddr_t)ip + (ip->ip_hl << 2)); nmatch = in_pcbnotify(&tcb, sa, th->th_dport, ip->ip_src, th->th_sport, cmd, notify); if (nmatch == 0 && syn_cache_count && (inetctlerrmap[cmd] == EHOSTUNREACH || inetctlerrmap[cmd] == ENETUNREACH || inetctlerrmap[cmd] == EHOSTDOWN)) syn_cache_unreach(ip, th); } else (void) in_pcbnotify(&tcb, sa, 0, zeroin_addr, 0, cmd, notify); } /* * When a source quench is received, close congestion window * to one segment. We will gradually open it again as we proceed. */ void tcp_quench(inp, errno) struct inpcb *inp; int errno; { struct tcpcb *tp = intotcpcb(inp); if (tp->t_xflags & TXF_NO_QUENCH) return; if (tp) { tp->snd_cwnd = tp->t_maxseg; } } /* * When a Destination Unreachable/Fragmentation Needed and DF set * is received, this routine is called for every TCP connection that * has a problematic destination: * If current route isn't route-to-host then purge it and get new route; * icmp_mtudisc() has already created a new route for us to use * Change t_maxseg * Retransmit dropped segment * Note that some of these connections might NOT need to retransmit; * we accept that there may be a few spurious retransmissions. */ void tcp_mtudisc(inp, errno) struct inpcb *inp; int errno; { struct tcpcb *tp = intotcpcb(inp); struct route *ro = &(inp->inp_route); int usable_mtu; if (tp == NULL) return; /* * Note that we only re-route if we already had a route; * this should avoid creating a route if SO_DONTROUTE was set */ if (ro->ro_rt == NULL) return; /* * Make sure we have a route-to-host for this destination. */ if ((ro->ro_rt->rt_flags & RTF_HOST) == 0) { in_rtchange(inp, 0); rtcalloc(ro); if (ro->ro_rt == NULL) { printf("tcp_mtudisc: no new route?\n"); tcp_changemss(tp, TCP_MSS); return; } if ((ro->ro_rt->rt_flags & RTF_HOST) == 0) { printf("tcp_mtudisc: new route not to host?\n"); usable_mtu = ro->ro_rt->rt_ifp->if_mtu - sizeof(struct tcpiphdr); tcp_changemss(tp, min(usable_mtu, TCP_MSS)); return; } } usable_mtu = ro->ro_rt->rt_rmx.rmx_mtu - sizeof(struct tcpiphdr); tcp_changemss(tp, usable_mtu); /* * If there is more than 1 segment of unacknowledged data, * force a retransmission of the (probably) lost datagrams */ if (SEQ_GT(tp->snd_una, tp->snd_nxt + tp->t_maxseg)) { tp->snd_nxt = tp->snd_una; (void) tcp_output(tp); } } /* * Called when the MSS should change to reflect a new MTU value. */ void tcp_changemss(tp, usable_mtu) register struct tcpcb *tp; u_int usable_mtu; { u_int win; int newmaxseg; newmaxseg = tcp_maxseg(tp, min(usable_mtu, tp->t_peermaxseg)); if (tp->t_maxseg == newmaxseg) return; /* no change */ tp->t_maxseg = newmaxseg; /* * Readjust things that depend upon t_maxseg * (code copied from tcp_timer.c in the REXMT case) */ win = min(tp->snd_wnd, tp->snd_cwnd) / 2 / tp->t_maxseg; if (win < 2) win = 2; tp->snd_cwnd = tp->t_maxseg; tp->snd_ssthresh = win * tp->t_maxseg; tp->t_dupacks = 0; } /* * Check to see if our Path MTU information is out of date. * If so, try again to use a higher MTU and see if things have * improved. */ tcp_agepathmtu(inp, rt) struct inpcb *inp; register struct rtentry *rt; { unsigned int usable_mtu; extern unsigned int mtu_table[]; unsigned int *mtup, rt_mtu = rt->rt_rmx.rmx_mtu; time_t expire_at; /* Find the next higher MTU plateau */ for (mtup = mtu_table; *mtup; mtup++) { if (mtup[1] <= rt_mtu) break; } usable_mtu = *mtup; /* Don't try an MTU greater than the if_mtu! */ if (usable_mtu > rt->rt_ifp->if_mtu) usable_mtu = rt->rt_ifp->if_mtu; if (usable_mtu > rt_mtu) { expire_at = time.tv_sec + pmtu_probe; rt->rt_rmx.rmx_mtu = usable_mtu; usable_mtu -= sizeof(struct tcpiphdr); tcp_changemss(intotcpcb(inp), usable_mtu); } else { expire_at = time.tv_sec + pmtu_expire; } rt->rt_flags &= ~RTF_PROBEMTU; rt_timer_add(rt, expire_at, icmp_mtuexpire, NULL); }