X-Git-Url: https://git.subgeniuskitty.com/unix-history/.git/blobdiff_plain/9957814917a14bb854b73d38067c1d8d4cf8cfaf..7985f2abad4958a0f6ef93ad1df40487ba6ff5a8:/usr/src/sys/netinet/tcp_input.c diff --git a/usr/src/sys/netinet/tcp_input.c b/usr/src/sys/netinet/tcp_input.c index 36464baa34..a54439c66b 100644 --- a/usr/src/sys/netinet/tcp_input.c +++ b/usr/src/sys/netinet/tcp_input.c @@ -1,7 +1,25 @@ -/* tcp_input.c 6.6 84/10/19 */ +/* + * Copyright (c) 1982, 1986, 1988 Regents of the University of California. + * All rights reserved. + * + * Redistribution and use in source and binary forms are permitted + * provided that the above copyright notice and this paragraph are + * duplicated in all such forms and that any documentation, + * advertising materials, and other materials related to such + * distribution and use acknowledge that the software was developed + * by the University of California, Berkeley. The name of the + * University may not be used to endorse or promote products derived + * from this software without specific prior written permission. + * THIS SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY EXPRESS OR + * IMPLIED WARRANTIES, INCLUDING, WITHOUT LIMITATION, THE IMPLIED + * WARRANTIES OF MERCHANTIBILITY AND FITNESS FOR A PARTICULAR PURPOSE. + * + * @(#)tcp_input.c 7.20 (Berkeley) %G% + */ #include "param.h" #include "systm.h" +#include "malloc.h" #include "mbuf.h" #include "protosw.h" #include "socket.h" @@ -25,42 +43,174 @@ #include "tcp_debug.h" int tcpprintfs = 0; -int tcpcksum = 1; +int tcprexmtthresh = 3; struct tcpiphdr tcp_saveti; -extern tcpnodelack; struct tcpcb *tcp_newtcpcb(); + +/* + * Insert segment ti into reassembly queue of tcp with + * control block tp. Return TH_FIN if reassembly now includes + * a segment with FIN. The macro form does the common case inline + * (segment is the next to be received on an established connection, + * and the queue is empty), avoiding linkage into and removal + * from the queue and repetition of various conversions. + * Set DELACK for segments received in order, but ack immediately + * when segments are out of order (so fast retransmit can work). + */ +#define TCP_REASS(tp, ti, m, so, flags) { \ + if ((ti)->ti_seq == (tp)->rcv_nxt && \ + (tp)->seg_next == (struct tcpiphdr *)(tp) && \ + (tp)->t_state == TCPS_ESTABLISHED) { \ + tp->t_flags |= TF_DELACK; \ + (tp)->rcv_nxt += (ti)->ti_len; \ + flags = (ti)->ti_flags & TH_FIN; \ + tcpstat.tcps_rcvpack++;\ + tcpstat.tcps_rcvbyte += (ti)->ti_len;\ + sbappend(&(so)->so_rcv, (m)); \ + sorwakeup(so); \ + } else { \ + (flags) = tcp_reass((tp), (ti)); \ + tp->t_flags |= TF_ACKNOW; \ + } \ +} + +tcp_reass(tp, ti) + register struct tcpcb *tp; + register struct tcpiphdr *ti; +{ + register struct tcpiphdr *q; + struct socket *so = tp->t_inpcb->inp_socket; + struct mbuf *m; + int flags; + + /* + * Call with ti==0 after become established to + * force pre-ESTABLISHED data up to user socket. + */ + if (ti == 0) + goto present; + + /* + * Find a segment which begins after this one does. + */ + for (q = tp->seg_next; q != (struct tcpiphdr *)tp; + q = (struct tcpiphdr *)q->ti_next) + if (SEQ_GT(q->ti_seq, ti->ti_seq)) + break; + + /* + * If there is a preceding segment, it may provide some of + * our data already. If so, drop the data from the incoming + * segment. If it provides all of our data, drop us. + */ + if ((struct tcpiphdr *)q->ti_prev != (struct tcpiphdr *)tp) { + register int i; + q = (struct tcpiphdr *)q->ti_prev; + /* conversion to int (in i) handles seq wraparound */ + i = q->ti_seq + q->ti_len - ti->ti_seq; + if (i > 0) { + if (i >= ti->ti_len) { + tcpstat.tcps_rcvduppack++; + tcpstat.tcps_rcvdupbyte += ti->ti_len; + goto drop; + } + m_adj(dtom(ti), i); + ti->ti_len -= i; + ti->ti_seq += i; + } + q = (struct tcpiphdr *)(q->ti_next); + } + tcpstat.tcps_rcvoopack++; + tcpstat.tcps_rcvoobyte += ti->ti_len; + + /* + * While we overlap succeeding segments trim them or, + * if they are completely covered, dequeue them. + */ + while (q != (struct tcpiphdr *)tp) { + register int i = (ti->ti_seq + ti->ti_len) - q->ti_seq; + if (i <= 0) + break; + if (i < q->ti_len) { + q->ti_seq += i; + q->ti_len -= i; + m_adj(dtom(q), i); + break; + } + q = (struct tcpiphdr *)q->ti_next; + m = dtom(q->ti_prev); + remque(q->ti_prev); + m_freem(m); + } + + /* + * Stick new segment in its place. + */ + insque(ti, q->ti_prev); + +present: + /* + * Present data to user, advancing rcv_nxt through + * completed sequence space. + */ + if (TCPS_HAVERCVDSYN(tp->t_state) == 0) + return (0); + ti = tp->seg_next; + if (ti == (struct tcpiphdr *)tp || ti->ti_seq != tp->rcv_nxt) + return (0); + if (tp->t_state == TCPS_SYN_RECEIVED && ti->ti_len) + return (0); + do { + tp->rcv_nxt += ti->ti_len; + flags = ti->ti_flags & TH_FIN; + remque(ti); + m = dtom(ti); + ti = (struct tcpiphdr *)ti->ti_next; + if (so->so_state & SS_CANTRCVMORE) + m_freem(m); + else + sbappend(&so->so_rcv, m); + } while (ti != (struct tcpiphdr *)tp && ti->ti_seq == tp->rcv_nxt); + sorwakeup(so); + return (flags); +drop: + m_freem(dtom(ti)); + return (0); +} + /* * TCP input routine, follows pages 65-76 of the * protocol specification dated September, 1981 very closely. */ -tcp_input(m0) - struct mbuf *m0; +tcp_input(m, iphlen) + register struct mbuf *m; + int iphlen; { register struct tcpiphdr *ti; struct inpcb *inp; - register struct mbuf *m; struct mbuf *om = 0; int len, tlen, off; register struct tcpcb *tp = 0; register int tiflags; struct socket *so; - int todrop, acked; + int todrop, acked, ourfinisacked, needoutput = 0; short ostate; struct in_addr laddr; int dropsocket = 0; + int iss = 0; + tcpstat.tcps_rcvtotal++; /* * Get IP and TCP header together in first mbuf. * Note: IP leaves IP header in first mbuf. */ - m = m0; ti = mtod(m, struct tcpiphdr *); - if (((struct ip *)ti)->ip_hl > (sizeof (struct ip) >> 2)) - ip_stripoptions((struct ip *)ti, (struct mbuf *)0); - if (m->m_off > MMAXOFF || m->m_len < sizeof (struct tcpiphdr)) { + if (iphlen > sizeof (struct ip)) + ip_stripoptions(m, (struct mbuf *)0); + if (m->m_flags & M_EXT || m->m_len < sizeof (struct tcpiphdr)) { if ((m = m_pullup(m, sizeof (struct tcpiphdr))) == 0) { - tcpstat.tcps_hdrops++; + tcpstat.tcps_rcvshort++; return; } ti = mtod(m, struct tcpiphdr *); @@ -71,17 +221,15 @@ tcp_input(m0) */ tlen = ((struct ip *)ti)->ip_len; len = sizeof (struct ip) + tlen; - if (tcpcksum) { - ti->ti_next = ti->ti_prev = 0; - ti->ti_x1 = 0; - ti->ti_len = (u_short)tlen; - ti->ti_len = htons((u_short)ti->ti_len); - if (ti->ti_sum = in_cksum(m, len)) { - if (tcpprintfs) - printf("tcp sum: src %x\n", ti->ti_src); - tcpstat.tcps_badsum++; - goto drop; - } + ti->ti_next = ti->ti_prev = 0; + ti->ti_x1 = 0; + ti->ti_len = (u_short)tlen; + ti->ti_len = htons((u_short)ti->ti_len); + if (ti->ti_sum = in_cksum(m, len)) { + if (tcpprintfs) + printf("tcp sum: src %x\n", ti->ti_src); + tcpstat.tcps_rcvbadsum++; + goto drop; } /* @@ -92,17 +240,19 @@ tcp_input(m0) if (off < sizeof (struct tcphdr) || off > tlen) { if (tcpprintfs) printf("tcp off: src %x off %d\n", ti->ti_src, off); - tcpstat.tcps_badoff++; + tcpstat.tcps_rcvbadoff++; goto drop; } tlen -= off; ti->ti_len = tlen; if (off > sizeof (struct tcphdr)) { - if ((m = m_pullup(m, sizeof (struct ip) + off)) == 0) { - tcpstat.tcps_hdrops++; - return; + if (m->m_len < sizeof(struct ip) + off) { + if ((m = m_pullup(m, sizeof (struct ip) + off)) == 0) { + tcpstat.tcps_rcvshort++; + return; + } + ti = mtod(m, struct tcpiphdr *); } - ti = mtod(m, struct tcpiphdr *); om = m_get(M_DONTWAIT, MT_DATA); if (om == 0) goto drop; @@ -110,6 +260,7 @@ tcp_input(m0) { caddr_t op = mtod(m, caddr_t) + sizeof (struct tcpiphdr); bcopy(op, mtod(om, caddr_t), (unsigned)om->m_len); m->m_len -= om->m_len; + m->m_pkthdr.len -= om->m_len; bcopy(op+om->m_len, op, (unsigned)(m->m_len-sizeof (struct tcpiphdr))); } @@ -117,11 +268,11 @@ tcp_input(m0) tiflags = ti->ti_flags; /* - * Drop TCP and IP headers. + * Drop TCP and IP headers; TCP options were dropped above. */ - off += sizeof (struct ip); - m->m_off += off; - m->m_len -= off; + m->m_data += sizeof(struct tcpiphdr); + m->m_len -= sizeof(struct tcpiphdr); + m->m_pkthdr.len -= sizeof(struct tcpiphdr); /* * Convert TCP protocol specific fields to host format. @@ -134,6 +285,7 @@ tcp_input(m0) /* * Locate pcb for segment. */ +findpcb: inp = in_pcblookup (&tcb, ti->ti_src, ti->ti_sport, ti->ti_dst, ti->ti_dport, INPLOOKUP_WILDCARD); @@ -141,12 +293,16 @@ tcp_input(m0) /* * If the state is CLOSED (i.e., TCB does not exist) then * all data in the incoming segment is discarded. + * If the TCB exists but is in CLOSED state, it is embryonic, + * but should either do a listen or a connect soon. */ if (inp == 0) goto dropwithreset; tp = intotcpcb(inp); if (tp == 0) goto dropwithreset; + if (tp->t_state == TCPS_CLOSED) + goto drop; so = inp->inp_socket; if (so->so_options & SO_DEBUG) { ostate = tp->t_state; @@ -171,6 +327,9 @@ tcp_input(m0) inp = (struct inpcb *)so->so_pcb; inp->inp_laddr = ti->ti_dst; inp->inp_lport = ti->ti_dport; +#if BSD>=43 + inp->inp_options = ip_srcroute(); +#endif tp = intotcpcb(inp); tp->t_state = TCPS_LISTEN; } @@ -180,7 +339,7 @@ tcp_input(m0) * Reset idle time and keep-alive timer. */ tp->t_idle = 0; - tp->t_timer[TCPT_KEEP] = TCPTV_KEEP; + tp->t_timer[TCPT_KEEP] = tcp_keepidle; /* * Process options if not in LISTEN state, @@ -194,10 +353,16 @@ tcp_input(m0) /* * Calculate amount of space in receive window, * and then do TCP input processing. + * Receive window is amount of space in rcv queue, + * but not less than advertised window. */ - tp->rcv_wnd = sbspace(&so->so_rcv); - if (tp->rcv_wnd < 0) - tp->rcv_wnd = 0; + { int win; + + win = sbspace(&so->so_rcv); + if (win < 0) + win = 0; + tp->rcv_wnd = max(win, (int)(tp->rcv_adv - tp->rcv_nxt)); + } switch (tp->t_state) { @@ -205,6 +370,7 @@ tcp_input(m0) * If the state is LISTEN then ignore segment if it contains an RST. * If the segment contains an ACK then it is bad and send a RST. * If it does not contain a SYN then it is not interesting; drop it. + * Don't bother responding if the destination was a broadcast. * Otherwise initialize tp->rcv_nxt, and tp->irs, select an initial * tp->iss, and send a segment: * @@ -223,6 +389,8 @@ tcp_input(m0) goto dropwithreset; if ((tiflags & TH_SYN) == 0) goto drop; + if (m->m_flags & M_BCAST) + goto drop; am = m_get(M_DONTWAIT, MT_SONAME); if (am == NULL) goto drop; @@ -242,23 +410,27 @@ tcp_input(m0) (void) m_free(am); tp->t_template = tcp_template(tp); if (tp->t_template == 0) { - in_pcbdisconnect(inp); + tp = tcp_drop(tp, ENOBUFS); dropsocket = 0; /* socket is already gone */ - inp->inp_laddr = laddr; - tp = 0; goto drop; } if (om) { tcp_dooptions(tp, om, ti); om = 0; } - tp->iss = tcp_iss; tcp_iss += TCP_ISSINCR/2; + if (iss) + tp->iss = iss; + else + tp->iss = tcp_iss; + tcp_iss += TCP_ISSINCR/2; tp->irs = ti->ti_seq; tcp_sendseqinit(tp); tcp_rcvseqinit(tp); + tp->t_flags |= TF_ACKNOW; tp->t_state = TCPS_SYN_RECEIVED; - tp->t_timer[TCPT_KEEP] = TCPTV_KEEP; + tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT; dropsocket = 0; /* committed to socket */ + tcpstat.tcps_accepts++; goto trimthenstep6; } @@ -276,8 +448,7 @@ tcp_input(m0) */ case TCPS_SYN_SENT: if ((tiflags & TH_ACK) && -/* this should be SEQ_LT; is SEQ_LEQ for BBN vax TCP only */ - (SEQ_LT(ti->ti_ack, tp->iss) || + (SEQ_LEQ(ti->ti_ack, tp->iss) || SEQ_GT(ti->ti_ack, tp->snd_max))) goto dropwithreset; if (tiflags & TH_RST) { @@ -287,21 +458,35 @@ tcp_input(m0) } if ((tiflags & TH_SYN) == 0) goto drop; - tp->snd_una = ti->ti_ack; - if (SEQ_LT(tp->snd_nxt, tp->snd_una)) - tp->snd_nxt = tp->snd_una; + if (tiflags & TH_ACK) { + tp->snd_una = ti->ti_ack; + if (SEQ_LT(tp->snd_nxt, tp->snd_una)) + tp->snd_nxt = tp->snd_una; + } tp->t_timer[TCPT_REXMT] = 0; tp->irs = ti->ti_seq; tcp_rcvseqinit(tp); tp->t_flags |= TF_ACKNOW; - if (SEQ_GT(tp->snd_una, tp->iss)) { + if (tiflags & TH_ACK && SEQ_GT(tp->snd_una, tp->iss)) { + tcpstat.tcps_connects++; soisconnected(so); tp->t_state = TCPS_ESTABLISHED; - tp->t_maxseg = MIN(tp->t_maxseg, tcp_mss(tp)); + tp->t_maxseg = min(tp->t_maxseg, tcp_mss(tp)); (void) tcp_reass(tp, (struct tcpiphdr *)0); + /* + * if we didn't have to retransmit the SYN, + * use its rtt as our initial srtt & rtt var. + */ + if (tp->t_rtt) { + tp->t_srtt = tp->t_rtt << 3; + tp->t_rttvar = tp->t_rtt << 1; + TCPT_RANGESET(tp->t_rxtcur, + ((tp->t_srtt >> 2) + tp->t_rttvar) >> 1, + TCPTV_MIN, TCPTV_REXMTMAX); + tp->t_rtt = 0; + } } else tp->t_state = TCPS_SYN_RECEIVED; - goto trimthenstep6; trimthenstep6: /* @@ -312,85 +497,143 @@ trimthenstep6: ti->ti_seq++; if (ti->ti_len > tp->rcv_wnd) { todrop = ti->ti_len - tp->rcv_wnd; +#if BSD>=43 m_adj(m, -todrop); +#else + /* XXX work around 4.2 m_adj bug */ + if (m->m_len) { + m_adj(m, -todrop); + } else { + /* skip tcp/ip header in first mbuf */ + m_adj(m->m_next, -todrop); + } +#endif ti->ti_len = tp->rcv_wnd; - ti->ti_flags &= ~TH_FIN; + tiflags &= ~TH_FIN; + tcpstat.tcps_rcvpackafterwin++; + tcpstat.tcps_rcvbyteafterwin += todrop; } tp->snd_wl1 = ti->ti_seq - 1; + tp->rcv_up = ti->ti_seq; goto step6; } /* - * If data is received on a connection after the + * States other than LISTEN or SYN_SENT. + * First check that at least some bytes of segment are within + * receive window. If segment begins before rcv_nxt, + * drop leading data (and SYN); if nothing left, just ack. + */ + todrop = tp->rcv_nxt - ti->ti_seq; + if (todrop > 0) { + if (tiflags & TH_SYN) { + tiflags &= ~TH_SYN; + ti->ti_seq++; + if (ti->ti_urp > 1) + ti->ti_urp--; + else + tiflags &= ~TH_URG; + todrop--; + } + if (todrop > ti->ti_len || + todrop == ti->ti_len && (tiflags&TH_FIN) == 0) { + tcpstat.tcps_rcvduppack++; + tcpstat.tcps_rcvdupbyte += ti->ti_len; + /* + * If segment is just one to the left of the window, + * check two special cases: + * 1. Don't toss RST in response to 4.2-style keepalive. + * 2. If the only thing to drop is a FIN, we can drop + * it, but check the ACK or we will get into FIN + * wars if our FINs crossed (both CLOSING). + * In either case, send ACK to resynchronize, + * but keep on processing for RST or ACK. + */ + if ((tiflags & TH_FIN && todrop == ti->ti_len + 1) +#ifdef TCP_COMPAT_42 + || (tiflags & TH_RST && ti->ti_seq == tp->rcv_nxt - 1) +#endif + ) { + todrop = ti->ti_len; + tiflags &= ~TH_FIN; + tp->t_flags |= TF_ACKNOW; + } else + goto dropafterack; + } else { + tcpstat.tcps_rcvpartduppack++; + tcpstat.tcps_rcvpartdupbyte += todrop; + } + m_adj(m, todrop); + ti->ti_seq += todrop; + ti->ti_len -= todrop; + if (ti->ti_urp > todrop) + ti->ti_urp -= todrop; + else { + tiflags &= ~TH_URG; + ti->ti_urp = 0; + } + } + + /* + * If new data are received on a connection after the * user processes are gone, then RST the other end. */ - if ((so->so_state & SS_NOFDREF) && tp->t_state > TCPS_CLOSE_WAIT && - ti->ti_len) { + if ((so->so_state & SS_NOFDREF) && + tp->t_state > TCPS_CLOSE_WAIT && ti->ti_len) { tp = tcp_close(tp); + tcpstat.tcps_rcvafterclose++; goto dropwithreset; } /* - * States other than LISTEN or SYN_SENT. - * First check that at least some bytes of segment are within - * receive window. + * If segment ends after window, drop trailing data + * (and PUSH and FIN); if nothing left, just ACK. */ - if (tp->rcv_wnd == 0) { - /* - * If window is closed can only take segments at - * window edge, and have to drop data and PUSH from - * incoming segments. - */ - if (tp->rcv_nxt != ti->ti_seq) - goto dropafterack; - if (ti->ti_len > 0) { - m_adj(m, ti->ti_len); - ti->ti_len = 0; - ti->ti_flags &= ~(TH_PUSH|TH_FIN); - } - } else { - /* - * If segment begins before rcv_nxt, drop leading - * data (and SYN); if nothing left, just ack. - */ - todrop = tp->rcv_nxt - ti->ti_seq; - if (todrop > 0) { - if (tiflags & TH_SYN) { - tiflags &= ~TH_SYN; - ti->ti_flags &= ~TH_SYN; - ti->ti_seq++; - if (ti->ti_urp > 1) - ti->ti_urp--; - else - tiflags &= ~TH_URG; - todrop--; + todrop = (ti->ti_seq+ti->ti_len) - (tp->rcv_nxt+tp->rcv_wnd); + if (todrop > 0) { + tcpstat.tcps_rcvpackafterwin++; + if (todrop >= ti->ti_len) { + tcpstat.tcps_rcvbyteafterwin += ti->ti_len; + /* + * If a new connection request is received + * while in TIME_WAIT, drop the old connection + * and start over if the sequence numbers + * are above the previous ones. + */ + if (tiflags & TH_SYN && + tp->t_state == TCPS_TIME_WAIT && + SEQ_GT(ti->ti_seq, tp->rcv_nxt)) { + iss = tp->rcv_nxt + TCP_ISSINCR; + (void) tcp_close(tp); + goto findpcb; } - if (todrop > ti->ti_len || - todrop == ti->ti_len && (tiflags&TH_FIN) == 0) - goto dropafterack; - m_adj(m, todrop); - ti->ti_seq += todrop; - ti->ti_len -= todrop; - if (ti->ti_urp > todrop) - ti->ti_urp -= todrop; - else { - tiflags &= ~TH_URG; - ti->ti_flags &= ~TH_URG; - ti->ti_urp = 0; - } - } - /* - * If segment ends after window, drop trailing data - * (and PUSH and FIN); if nothing left, just ACK. - */ - todrop = (ti->ti_seq+ti->ti_len) - (tp->rcv_nxt+tp->rcv_wnd); - if (todrop > 0) { - if (todrop >= ti->ti_len) + /* + * If window is closed can only take segments at + * window edge, and have to drop data and PUSH from + * incoming segments. Continue processing, but + * remember to ack. Otherwise, drop segment + * and ack. + */ + if (tp->rcv_wnd == 0 && ti->ti_seq == tp->rcv_nxt) { + tp->t_flags |= TF_ACKNOW; + tcpstat.tcps_rcvwinprobe++; + } else goto dropafterack; + } else + tcpstat.tcps_rcvbyteafterwin += todrop; +#if BSD>=43 + m_adj(m, -todrop); +#else + /* XXX work around m_adj bug */ + if (m->m_len) { m_adj(m, -todrop); - ti->ti_len -= todrop; - ti->ti_flags &= ~(TH_PUSH|TH_FIN); + } else { + /* skip tcp/ip header in first mbuf */ + m_adj(m->m_next, -todrop); } +#endif + ti->ti_len -= todrop; + tiflags &= ~(TH_PUSH|TH_FIN); } /* @@ -406,14 +649,18 @@ trimthenstep6: if (tiflags&TH_RST) switch (tp->t_state) { case TCPS_SYN_RECEIVED: - tp = tcp_drop(tp, ECONNREFUSED); - goto drop; + so->so_error = ECONNREFUSED; + goto close; case TCPS_ESTABLISHED: case TCPS_FIN_WAIT_1: case TCPS_FIN_WAIT_2: case TCPS_CLOSE_WAIT: - tp = tcp_drop(tp, ECONNRESET); + so->so_error = ECONNRESET; + close: + tp->t_state = TCPS_CLOSED; + tcpstat.tcps_drops++; + tp = tcp_close(tp); goto drop; case TCPS_CLOSING: @@ -445,20 +692,17 @@ trimthenstep6: /* * In SYN_RECEIVED state if the ack ACKs our SYN then enter - * ESTABLISHED state and continue processing, othewise + * ESTABLISHED state and continue processing, otherwise * send an RST. */ case TCPS_SYN_RECEIVED: if (SEQ_GT(tp->snd_una, ti->ti_ack) || SEQ_GT(ti->ti_ack, tp->snd_max)) goto dropwithreset; - tp->snd_una++; /* SYN acked */ - if (SEQ_LT(tp->snd_nxt, tp->snd_una)) - tp->snd_nxt = tp->snd_una; - tp->t_timer[TCPT_REXMT] = 0; + tcpstat.tcps_connects++; soisconnected(so); tp->t_state = TCPS_ESTABLISHED; - tp->t_maxseg = MIN(tp->t_maxseg, tcp_mss(tp)); + tp->t_maxseg = min(tp->t_maxseg, tcp_mss(tp)); (void) tcp_reass(tp, (struct tcpiphdr *)0); tp->snd_wl1 = ti->ti_seq - 1; /* fall into ... */ @@ -478,45 +722,164 @@ trimthenstep6: case TCPS_CLOSING: case TCPS_LAST_ACK: case TCPS_TIME_WAIT: -#define ourfinisacked (acked > 0) - if (SEQ_LEQ(ti->ti_ack, tp->snd_una)) + if (SEQ_LEQ(ti->ti_ack, tp->snd_una)) { + if (ti->ti_len == 0 && ti->ti_win == tp->snd_wnd) { + tcpstat.tcps_rcvdupack++; + /* + * If we have outstanding data (not a + * window probe), this is a completely + * duplicate ack (ie, window info didn't + * change), the ack is the biggest we've + * seen and we've seen exactly our rexmt + * threshhold of them, assume a packet + * has been dropped and retransmit it. + * Kludge snd_nxt & the congestion + * window so we send only this one + * packet. If this packet fills the + * only hole in the receiver's seq. + * space, the next real ack will fully + * open our window. This means we + * have to do the usual slow-start to + * not overwhelm an intermediate gateway + * with a burst of packets. Leave + * here with the congestion window set + * to allow 2 packets on the next real + * ack and the exp-to-linear thresh + * set for half the current window + * size (since we know we're losing at + * the current window size). + */ + if (tp->t_timer[TCPT_REXMT] == 0 || + ti->ti_ack != tp->snd_una) + tp->t_dupacks = 0; + else if (++tp->t_dupacks == tcprexmtthresh) { + tcp_seq onxt = tp->snd_nxt; + u_int win = + min(tp->snd_wnd, tp->snd_cwnd) / 2 / + tp->t_maxseg; + + if (win < 2) + win = 2; + tp->snd_ssthresh = win * tp->t_maxseg; + + tp->t_timer[TCPT_REXMT] = 0; + tp->t_rtt = 0; + tp->snd_nxt = ti->ti_ack; + tp->snd_cwnd = tp->t_maxseg; + (void) tcp_output(tp); + + if (SEQ_GT(onxt, tp->snd_nxt)) + tp->snd_nxt = onxt; + goto drop; + } + } else + tp->t_dupacks = 0; break; - if (SEQ_GT(ti->ti_ack, tp->snd_max)) + } + tp->t_dupacks = 0; + if (SEQ_GT(ti->ti_ack, tp->snd_max)) { + tcpstat.tcps_rcvacktoomuch++; goto dropafterack; + } acked = ti->ti_ack - tp->snd_una; + tcpstat.tcps_rcvackpack++; + tcpstat.tcps_rcvackbyte += acked; /* * If transmit timer is running and timed sequence * number was acked, update smoothed round trip time. + * Since we now have an rtt measurement, cancel the + * timer backoff (cf., Phil Karn's retransmit alg.). + * Recompute the initial retransmit timer. */ if (tp->t_rtt && SEQ_GT(ti->ti_ack, tp->t_rtseq)) { - if (tp->t_srtt == 0) - tp->t_srtt = tp->t_rtt; - else - tp->t_srtt = - tcp_alpha * tp->t_srtt + - (1 - tcp_alpha) * tp->t_rtt; + tcpstat.tcps_rttupdated++; + if (tp->t_srtt != 0) { + register short delta; + + /* + * srtt is stored as fixed point with 3 bits + * after the binary point (i.e., scaled by 8). + * The following magic is equivalent + * to the smoothing algorithm in rfc793 + * with an alpha of .875 + * (srtt = rtt/8 + srtt*7/8 in fixed point). + * Adjust t_rtt to origin 0. + */ + delta = tp->t_rtt - 1 - (tp->t_srtt >> 3); + if ((tp->t_srtt += delta) <= 0) + tp->t_srtt = 1; + /* + * We accumulate a smoothed rtt variance + * (actually, a smoothed mean difference), + * then set the retransmit timer to smoothed + * rtt + 2 times the smoothed variance. + * rttvar is stored as fixed point + * with 2 bits after the binary point + * (scaled by 4). The following is equivalent + * to rfc793 smoothing with an alpha of .75 + * (rttvar = rttvar*3/4 + |delta| / 4). + * This replaces rfc793's wired-in beta. + */ + if (delta < 0) + delta = -delta; + delta -= (tp->t_rttvar >> 2); + if ((tp->t_rttvar += delta) <= 0) + tp->t_rttvar = 1; + } else { + /* + * No rtt measurement yet - use the + * unsmoothed rtt. Set the variance + * to half the rtt (so our first + * retransmit happens at 2*rtt) + */ + tp->t_srtt = tp->t_rtt << 3; + tp->t_rttvar = tp->t_rtt << 1; + } tp->t_rtt = 0; + tp->t_rxtshift = 0; + TCPT_RANGESET(tp->t_rxtcur, + ((tp->t_srtt >> 2) + tp->t_rttvar) >> 1, + TCPTV_MIN, TCPTV_REXMTMAX); } - if (ti->ti_ack == tp->snd_max) + /* + * If all outstanding data is acked, stop retransmit + * timer and remember to restart (more output or persist). + * If there is more data to be acked, restart retransmit + * timer, using current (possibly backed-off) value. + */ + if (ti->ti_ack == tp->snd_max) { tp->t_timer[TCPT_REXMT] = 0; - else { - TCPT_RANGESET(tp->t_timer[TCPT_REXMT], - tcp_beta * tp->t_srtt, TCPTV_MIN, TCPTV_MAX); - tp->t_rxtshift = 0; + needoutput = 1; + } else if (tp->t_timer[TCPT_PERSIST] == 0) + tp->t_timer[TCPT_REXMT] = tp->t_rxtcur; + /* + * When new data is acked, open the congestion window. + * If the window gives us less than ssthresh packets + * in flight, open exponentially (maxseg per packet). + * Otherwise open linearly (maxseg per window, + * or maxseg^2 / cwnd per packet). + */ + { + u_int incr = tp->t_maxseg; + + if (tp->snd_cwnd > tp->snd_ssthresh) + incr = max(incr * incr / tp->snd_cwnd, 1); + + tp->snd_cwnd = min(tp->snd_cwnd + incr, USHRT_MAX); /* XXX */ } if (acked > so->so_snd.sb_cc) { tp->snd_wnd -= so->so_snd.sb_cc; - sbdrop(&so->so_snd, so->so_snd.sb_cc); + sbdrop(&so->so_snd, (int)so->so_snd.sb_cc); + ourfinisacked = 1; } else { sbdrop(&so->so_snd, acked); tp->snd_wnd -= acked; - acked = 0; + ourfinisacked = 0; } - if ((so->so_snd.sb_flags & SB_WAIT) || so->so_snd.sb_sel) - sowwakeup(so); + sowwakeup(so); tp->snd_una = ti->ti_ack; if (SEQ_LT(tp->snd_nxt, tp->snd_una)) tp->snd_nxt = tp->snd_una; @@ -533,15 +896,15 @@ trimthenstep6: /* * If we can't receive any more * data, then closing user can proceed. + * Starting the timer is contrary to the + * specification, but if we don't get a FIN + * we'll hang forever. */ - if (so->so_state & SS_CANTRCVMORE) + if (so->so_state & SS_CANTRCVMORE) { soisdisconnected(so); + tp->t_timer[TCPT_2MSL] = tcp_maxidle; + } tp->t_state = TCPS_FIN_WAIT_2; - /* - * This is contrary to the specification, - * but if we haven't gotten our FIN in - * 5 minutes, it's not forthcoming. - */ } break; @@ -561,15 +924,17 @@ trimthenstep6: break; /* - * The only thing that can arrive in LAST_ACK state - * is an acknowledgment of our FIN. If our FIN is now - * acknowledged, delete the TCB, enter the closed state - * and return. + * In LAST_ACK, we may still be waiting for data to drain + * and/or to be acked, as well as for the ack of our FIN. + * If our FIN is now acknowledged, delete the TCB, + * enter the closed state and return. */ case TCPS_LAST_ACK: - if (ourfinisacked) + if (ourfinisacked) { tp = tcp_close(tp); - goto drop; + goto drop; + } + break; /* * In TIME_WAIT state the only thing that should arrive @@ -580,21 +945,27 @@ trimthenstep6: tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL; goto dropafterack; } -#undef ourfinisacked } step6: /* * Update window information. + * Don't look at window if no ACK: TAC's send garbage on first SYN. */ - if (SEQ_LT(tp->snd_wl1, ti->ti_seq) || tp->snd_wl1 == ti->ti_seq && + if ((tiflags & TH_ACK) && + (SEQ_LT(tp->snd_wl1, ti->ti_seq) || tp->snd_wl1 == ti->ti_seq && (SEQ_LT(tp->snd_wl2, ti->ti_ack) || - tp->snd_wl2 == ti->ti_ack && ti->ti_win > tp->snd_wnd)) { + tp->snd_wl2 == ti->ti_ack && ti->ti_win > tp->snd_wnd))) { + /* keep track of pure window updates */ + if (ti->ti_len == 0 && + tp->snd_wl2 == ti->ti_ack && ti->ti_win > tp->snd_wnd) + tcpstat.tcps_rcvwinupd++; tp->snd_wnd = ti->ti_win; tp->snd_wl1 = ti->ti_seq; tp->snd_wl2 = ti->ti_ack; - if (tp->snd_wnd != 0) - tp->t_timer[TCPT_PERSIST] = 0; + if (tp->snd_wnd > tp->max_sndwnd) + tp->max_sndwnd = tp->snd_wnd; + needoutput = 1; } /* @@ -603,16 +974,15 @@ step6: if ((tiflags & TH_URG) && ti->ti_urp && TCPS_HAVERCVDFIN(tp->t_state) == 0) { /* - * This is a kludge, but if we receive accept + * This is a kludge, but if we receive and accept * random urgent pointers, we'll crash in * soreceive. It's hard to imagine someone * actually wanting to send this much urgent data. */ - if (ti->ti_urp > tp->rcv_wnd + 1) { /* XXX */ + if (ti->ti_urp + so->so_rcv.sb_cc > SB_MAX) { ti->ti_urp = 0; /* XXX */ tiflags &= ~TH_URG; /* XXX */ - ti->ti_flags &= ~TH_URG; /* XXX */ - goto badurp; /* XXX */ + goto dodata; /* XXX */ } /* * If this segment advances the known urgent pointer, @@ -620,6 +990,13 @@ step6: * in CLOSE_WAIT, CLOSING, LAST_ACK or TIME_WAIT STATES since * a FIN has been received from the remote side. * In these states we ignore the URG. + * + * According to RFC961 (Assigned Protocols), + * the urgent pointer points to the last octet + * of urgent data. We continue, however, + * to consider it to indicate the first octet + * of data past the urgent section + * as the original spec states. */ if (SEQ_GT(ti->ti_seq+ti->ti_urp, tp->rcv_up)) { tp->rcv_up = ti->ti_seq + ti->ti_urp; @@ -628,7 +1005,7 @@ step6: if (so->so_oobmark == 0) so->so_state |= SS_RCVATMARK; sohasoutofband(so); - tp->t_oobflags &= ~TCPOOB_HAVEDATA; + tp->t_oobflags &= ~(TCPOOB_HAVEDATA | TCPOOB_HADDATA); } /* * Remove out of band data so doesn't get presented to user. @@ -636,10 +1013,21 @@ step6: * but if two URG's are pending at once, some out-of-band * data may creep in... ick. */ - if (ti->ti_urp <= ti->ti_len) + if (ti->ti_urp <= ti->ti_len +#ifdef SO_OOBINLINE + && (so->so_options & SO_OOBINLINE) == 0 +#endif + ) tcp_pulloutofband(so, ti); - } -badurp: /* XXX */ + } else + /* + * If no out of band data is expected, + * pull receive urgent pointer along + * with the receive window. + */ + if (SEQ_GT(tp->rcv_nxt, tp->rcv_up)) + tp->rcv_up = tp->rcv_nxt; +dodata: /* XXX */ /* * Process the segment text, merging it into the TCP sequencing queue, @@ -651,11 +1039,15 @@ badurp: /* XXX */ */ if ((ti->ti_len || (tiflags&TH_FIN)) && TCPS_HAVERCVDFIN(tp->t_state) == 0) { - tiflags = tcp_reass(tp, ti); - if (tcpnodelack == 0) - tp->t_flags |= TF_DELACK; - else - tp->t_flags |= TF_ACKNOW; + TCP_REASS(tp, ti, m, so, tiflags); + /* + * Note the amount of data that peer has sent into + * our window, in order to estimate the sender's + * buffer size. + */ + len = so->so_rcv.sb_hiwat - (tp->rcv_adv - tp->rcv_nxt); + if (len > tp->max_rcvd) + tp->max_rcvd = len; } else { m_freem(m); tiflags &= ~TH_FIN; @@ -716,7 +1108,8 @@ badurp: /* XXX */ /* * Return any desired output. */ - (void) tcp_output(tp); + if (needoutput || (tp->t_flags & TF_ACKNOW)) + (void) tcp_output(tp); return; dropafterack: @@ -724,12 +1117,11 @@ dropafterack: * Generate an ACK dropping incoming segment if it occupies * sequence space, where the ACK reflects our state. */ - if ((tiflags&TH_RST) || - tlen == 0 && (tiflags&(TH_SYN|TH_FIN)) == 0) + if (tiflags & TH_RST) goto drop; - if (tp->t_inpcb->inp_socket->so_options & SO_DEBUG) - tcp_trace(TA_RESPOND, ostate, tp, &tcp_saveti, 0); - tcp_respond(tp, ti, tp->rcv_nxt, tp->snd_nxt, TH_ACK); + m_freem(m); + tp->t_flags |= TF_ACKNOW; + (void) tcp_output(tp); return; dropwithreset: @@ -740,15 +1132,16 @@ dropwithreset: /* * Generate a RST, dropping incoming segment. * Make ACK acceptable to originator of segment. + * Don't bother to respond if destination was broadcast. */ - if (tiflags & TH_RST) + if ((tiflags & TH_RST) || m->m_flags & M_BCAST) goto drop; if (tiflags & TH_ACK) - tcp_respond(tp, ti, (tcp_seq)0, ti->ti_ack, TH_RST); + tcp_respond(tp, ti, m, (tcp_seq)0, ti->ti_ack, TH_RST); else { if (tiflags & TH_SYN) ti->ti_len++; - tcp_respond(tp, ti, ti->ti_seq+ti->ti_len, (tcp_seq)0, + tcp_respond(tp, ti, m, ti->ti_seq+ti->ti_len, (tcp_seq)0, TH_RST|TH_ACK); } /* destroy temporarily created socket */ @@ -804,7 +1197,7 @@ tcp_dooptions(tp, om, ti) continue; tp->t_maxseg = *(u_short *)(cp + 2); tp->t_maxseg = ntohs((u_short)tp->t_maxseg); - tp->t_maxseg = MIN(tp->t_maxseg, tcp_mss(tp)); + tp->t_maxseg = min(tp->t_maxseg, tcp_mss(tp)); break; } } @@ -844,124 +1237,24 @@ tcp_pulloutofband(so, ti) panic("tcp_pulloutofband"); } -/* - * Insert segment ti into reassembly queue of tcp with - * control block tp. Return TH_FIN if reassembly now includes - * a segment with FIN. - */ -tcp_reass(tp, ti) - register struct tcpcb *tp; - register struct tcpiphdr *ti; -{ - register struct tcpiphdr *q; - struct socket *so = tp->t_inpcb->inp_socket; - struct mbuf *m; - int flags; - - /* - * Call with ti==0 after become established to - * force pre-ESTABLISHED data up to user socket. - */ - if (ti == 0) - goto present; - - /* - * Find a segment which begins after this one does. - */ - for (q = tp->seg_next; q != (struct tcpiphdr *)tp; - q = (struct tcpiphdr *)q->ti_next) - if (SEQ_GT(q->ti_seq, ti->ti_seq)) - break; - - /* - * If there is a preceding segment, it may provide some of - * our data already. If so, drop the data from the incoming - * segment. If it provides all of our data, drop us. - */ - if ((struct tcpiphdr *)q->ti_prev != (struct tcpiphdr *)tp) { - register int i; - q = (struct tcpiphdr *)q->ti_prev; - /* conversion to int (in i) handles seq wraparound */ - i = q->ti_seq + q->ti_len - ti->ti_seq; - if (i > 0) { - if (i >= ti->ti_len) - goto drop; - m_adj(dtom(ti), i); - ti->ti_len -= i; - ti->ti_seq += i; - } - q = (struct tcpiphdr *)(q->ti_next); - } - - /* - * While we overlap succeeding segments trim them or, - * if they are completely covered, dequeue them. - */ - while (q != (struct tcpiphdr *)tp) { - register int i = (ti->ti_seq + ti->ti_len) - q->ti_seq; - if (i <= 0) - break; - if (i < q->ti_len) { - q->ti_seq += i; - q->ti_len -= i; - m_adj(dtom(q), i); - break; - } - q = (struct tcpiphdr *)q->ti_next; - m = dtom(q->ti_prev); - remque(q->ti_prev); - m_freem(m); - } - - /* - * Stick new segment in its place. - */ - insque(ti, q->ti_prev); - -present: - /* - * Present data to user, advancing rcv_nxt through - * completed sequence space. - */ - if (TCPS_HAVERCVDSYN(tp->t_state) == 0) - return (0); - ti = tp->seg_next; - if (ti == (struct tcpiphdr *)tp || ti->ti_seq != tp->rcv_nxt) - return (0); - if (tp->t_state == TCPS_SYN_RECEIVED && ti->ti_len) - return (0); - do { - tp->rcv_nxt += ti->ti_len; - flags = ti->ti_flags & TH_FIN; - remque(ti); - m = dtom(ti); - ti = (struct tcpiphdr *)ti->ti_next; - if (so->so_state & SS_CANTRCVMORE) - m_freem(m); - else - sbappend(&so->so_rcv, m); - } while (ti != (struct tcpiphdr *)tp && ti->ti_seq == tp->rcv_nxt); - sorwakeup(so); - return (flags); -drop: - m_freem(dtom(ti)); - return (0); -} - /* * Determine a reasonable value for maxseg size. * If the route is known, use one that can be handled * on the given interface without forcing IP to fragment. - * If bigger than a page (CLSIZE), round down to nearest pagesize - * to utilize pagesize mbufs. + * If bigger than an mbuf cluster (MCLBYTES), round down to nearest size + * to utilize large mbufs. * If interface pointer is unavailable, or the destination isn't local, - * use a conservative size (512 or the default IP max size), + * use a conservative size (512 or the default IP max size, but no more + * than the mtu of the interface through which we route), * as we can't discover anything about intervening gateways or networks. + * We also initialize the congestion/slow start window to be a single + * segment if the destination isn't local; this information should + * probably all be saved with the routing entry at the transport level. * * This is ugly, and doesn't belong at this level, but has to happen somehow. */ tcp_mss(tp) -register struct tcpcb *tp; + register struct tcpcb *tp; { struct route *ro; struct ifnet *ifp; @@ -980,18 +1273,52 @@ register struct tcpcb *tp; rtalloc(ro); } if ((ro->ro_rt == 0) || (ifp = ro->ro_rt->rt_ifp) == 0) - return (MIN(IP_MSS - sizeof(struct tcpiphdr), 512)); + return (TCP_MSS); } mss = ifp->if_mtu - sizeof(struct tcpiphdr); -#if (CLBYTES & (CLBYTES - 1)) == 0 - if (mss > CLBYTES) - mss &= ~(CLBYTES-1); +#if (MCLBYTES & (MCLBYTES - 1)) == 0 + if (mss > MCLBYTES) + mss &= ~(MCLBYTES-1); #else - if (mss > CLBYTES) - mss = mss / CLBYTES * CLBYTES; + if (mss > MCLBYTES) + mss = mss / MCLBYTES * MCLBYTES; #endif - if (in_localaddr(tp->t_inpcb->inp_faddr)) - return(mss); - return (MIN(mss, MIN(IP_MSS - sizeof(struct tcpiphdr), 512))); + if (in_localaddr(inp->inp_faddr)) + return (mss); + + mss = min(mss, TCP_MSS); + tp->snd_cwnd = mss; + return (mss); +} + +#if BSD<43 +/* XXX this belongs in netinet/in.c */ +in_localaddr(in) + struct in_addr in; +{ + register u_long i = ntohl(in.s_addr); + register struct ifnet *ifp; + register struct sockaddr_in *sin; + register u_long mask; + + if (IN_CLASSA(i)) + mask = IN_CLASSA_NET; + else if (IN_CLASSB(i)) + mask = IN_CLASSB_NET; + else if (IN_CLASSC(i)) + mask = IN_CLASSC_NET; + else + return (0); + + i &= mask; + for (ifp = ifnet; ifp; ifp = ifp->if_next) { + if (ifp->if_addr.sa_family != AF_INET) + continue; + sin = (struct sockaddr_in *)&ifp->if_addr; + if ((sin->sin_addr.s_addr & mask) == i) + return (1); + } + return (0); } +#endif