X-Git-Url: https://git.subgeniuskitty.com/unix-history/.git/blobdiff_plain/6e7edb25cc39d3c7a292ffbce29991ab528249e3..05db08ca40dcf52f15be693f25a96a593409f690:/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 f44bb78230..109b7902ac 100644 --- a/usr/src/sys/netinet/tcp_input.c +++ b/usr/src/sys/netinet/tcp_input.c @@ -1,33 +1,169 @@ -/* tcp_input.c 1.80 82/10/20 */ - -#include "../h/param.h" -#include "../h/systm.h" -#include "../h/mbuf.h" -#include "../h/protosw.h" -#include "../h/socket.h" -#include "../h/socketvar.h" -#include "../netinet/in.h" -#include "../net/route.h" -#include "../netinet/in_pcb.h" -#include "../netinet/in_systm.h" +/* + * Copyright (c) 1982, 1986 Regents of the University of California. + * All rights reserved. The Berkeley software License Agreement + * specifies the terms and conditions for redistribution. + * + * @(#)tcp_input.c 7.12 (Berkeley) %G% + */ + +#include "param.h" +#include "systm.h" +#include "mbuf.h" +#include "protosw.h" +#include "socket.h" +#include "socketvar.h" +#include "errno.h" + #include "../net/if.h" -#include "../netinet/ip.h" -#include "../netinet/ip_var.h" -#include "../netinet/tcp.h" -#include "../netinet/tcp_fsm.h" -#include "../netinet/tcp_seq.h" -#include "../netinet/tcp_timer.h" -#include "../netinet/tcp_var.h" -#include "../netinet/tcpip.h" -#include "../netinet/tcp_debug.h" -#include +#include "../net/route.h" + +#include "in.h" +#include "in_pcb.h" +#include "in_systm.h" +#include "ip.h" +#include "ip_var.h" +#include "tcp.h" +#include "tcp_fsm.h" +#include "tcp_seq.h" +#include "tcp_timer.h" +#include "tcp_var.h" +#include "tcpip.h" +#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. + */ +#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)->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)); \ +} + +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. @@ -43,10 +179,13 @@ tcp_input(m0) 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. @@ -57,7 +196,7 @@ tcp_input(m0) ip_stripoptions((struct ip *)ti, (struct mbuf *)0); if (m->m_off > MMAXOFF || 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 *); @@ -72,13 +211,11 @@ tcp_input(m0) ti->ti_next = ti->ti_prev = 0; ti->ti_x1 = 0; ti->ti_len = (u_short)tlen; -#if vax || pdp11 || ns16032 ti->ti_len = htons((u_short)ti->ti_len); -#endif if (ti->ti_sum = in_cksum(m, len)) { - tcpstat.tcps_badsum++; if (tcpprintfs) - printf("tcp cksum %x\n", ti->ti_sum); + printf("tcp sum: src %x\n", ti->ti_src); + tcpstat.tcps_rcvbadsum++; goto drop; } } @@ -89,18 +226,22 @@ tcp_input(m0) */ off = ti->ti_off << 2; if (off < sizeof (struct tcphdr) || off > tlen) { - tcpstat.tcps_badoff++; + if (tcpprintfs) + printf("tcp off: src %x off %d\n", ti->ti_src, off); + 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++; - goto drop; + 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); + om = m_get(M_DONTWAIT, MT_DATA); if (om == 0) goto drop; om->m_len = off - sizeof (struct tcphdr); @@ -114,13 +255,11 @@ 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_off += sizeof(struct tcpiphdr); + m->m_len -= sizeof(struct tcpiphdr); -#if vax || pdp11 || ns16032 /* * Convert TCP protocol specific fields to host format. */ @@ -128,11 +267,11 @@ tcp_input(m0) ti->ti_ack = ntohl(ti->ti_ack); ti->ti_win = ntohs(ti->ti_win); ti->ti_urp = ntohs(ti->ti_urp); -#endif /* * Locate pcb for segment. */ +findpcb: inp = in_pcblookup (&tcb, ti->ti_src, ti->ti_sport, ti->ti_dst, ti->ti_dport, INPLOOKUP_WILDCARD); @@ -140,12 +279,16 @@ 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; @@ -155,9 +298,22 @@ tcp_input(m0) so = sonewconn(so); if (so == 0) goto drop; + /* + * This is ugly, but .... + * + * Mark socket as temporary until we're + * committed to keeping it. The code at + * ``drop'' and ``dropwithreset'' check the + * flag dropsocket to see if the temporary + * socket created here should be discarded. + * We mark the socket as discardable until + * we're committed to it below in TCPS_LISTEN. + */ + dropsocket++; inp = (struct inpcb *)so->so_pcb; inp->inp_laddr = ti->ti_dst; inp->inp_lport = ti->ti_dport; + inp->inp_options = ip_srcroute(); tp = intotcpcb(inp); tp->t_state = TCPS_LISTEN; } @@ -170,20 +326,27 @@ tcp_input(m0) tp->t_timer[TCPT_KEEP] = TCPTV_KEEP; /* - * Process options. + * Process options if not in LISTEN state, + * else do it below (after getting remote address). */ - if (om) { - tcp_dooptions(tp, om); + if (om && tp->t_state != TCPS_LISTEN) { + tcp_dooptions(tp, om, ti); om = 0; } /* * 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) { @@ -191,6 +354,7 @@ 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: * @@ -200,44 +364,57 @@ tcp_input(m0) * segment in this state. */ case TCPS_LISTEN: { - struct mbuf *am = m_get(M_DONTWAIT); + struct mbuf *am; register struct sockaddr_in *sin; - if (am == 0) - goto drop; - am->m_len = sizeof (struct sockaddr_in); if (tiflags & TH_RST) goto drop; if (tiflags & TH_ACK) goto dropwithreset; if ((tiflags & TH_SYN) == 0) goto drop; + if (in_broadcast(ti->ti_dst)) + goto drop; + am = m_get(M_DONTWAIT, MT_SONAME); + if (am == NULL) + goto drop; + am->m_len = sizeof (struct sockaddr_in); sin = mtod(am, struct sockaddr_in *); sin->sin_family = AF_INET; sin->sin_addr = ti->ti_src; sin->sin_port = ti->ti_sport; laddr = inp->inp_laddr; - if (inp->inp_laddr.s_addr == 0) + if (inp->inp_laddr.s_addr == INADDR_ANY) inp->inp_laddr = ti->ti_dst; if (in_pcbconnect(inp, am)) { inp->inp_laddr = laddr; - m_free(am); + (void) m_free(am); goto drop; } - m_free(am); + (void) m_free(am); tp->t_template = tcp_template(tp); if (tp->t_template == 0) { - in_pcbdisconnect(inp); - inp->inp_laddr = laddr; - tp = 0; + tp = tcp_drop(tp, ENOBUFS); + dropsocket = 0; /* socket is already gone */ goto drop; } - tp->iss = tcp_iss; tcp_iss += TCP_ISSINCR/2; + if (om) { + tcp_dooptions(tp, om, ti); + om = 0; + } + 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; + dropsocket = 0; /* committed to socket */ + tcpstat.tcps_accepts++; goto trimthenstep6; } @@ -255,33 +432,45 @@ 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) { - if (tiflags & TH_ACK) { - tcp_drop(tp, ECONNREFUSED); - tp = 0; - } + if (tiflags & TH_ACK) + tp = tcp_drop(tp, ECONNREFUSED); goto drop; } 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)); (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: /* @@ -294,85 +483,132 @@ trimthenstep6: todrop = ti->ti_len - tp->rcv_wnd; m_adj(m, -todrop); 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; } /* * States other than LISTEN or SYN_SENT. * First check that at least some bytes of segment are within - * receive window. + * 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) { +#ifdef TCP_COMPAT_42 + /* + * Don't toss RST in response to 4.2-style keepalive. + */ + if (ti->ti_seq == tp->rcv_nxt - 1 && tiflags & TH_RST) + goto do_rst; +#endif + tcpstat.tcps_rcvduppack++; + tcpstat.tcps_rcvdupbyte += ti->ti_len; + todrop = ti->ti_len; + tiflags &= ~TH_FIN; + tp->t_flags |= TF_ACKNOW; + } 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 (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 new data is received on a connection after the + * user processes are gone, then RST the other end. */ - if (tp->rcv_nxt != ti->ti_seq) + if ((so->so_state & SS_NOFDREF) && + tp->t_state > TCPS_CLOSE_WAIT && ti->ti_len) { + tp = tcp_close(tp); + tcpstat.tcps_rcvafterclose++; + goto dropwithreset; + } + if (tp->rcv_nxt != ti->ti_seq) { + tcpstat.tcps_rcvpackafterwin++; + tcpstat.tcps_rcvbyteafterwin += ti->ti_len; goto dropafterack; + } if (ti->ti_len > 0) { + if (ti->ti_len == 1) + tcpstat.tcps_rcvwinprobe++; + else { + tcpstat.tcps_rcvpackafterwin++; + tcpstat.tcps_rcvbyteafterwin += ti->ti_len; + } m_adj(m, ti->ti_len); ti->ti_len = 0; - ti->ti_flags &= ~(TH_PUSH|TH_FIN); + tiflags &= ~(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--; - } - 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 (todrop >= 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 == 1) + tcpstat.tcps_rcvwinprobe++; + else { + tcpstat.tcps_rcvpackafterwin++; + tcpstat.tcps_rcvbyteafterwin += ti->ti_len; + } goto dropafterack; + } + tcpstat.tcps_rcvpackafterwin++; + tcpstat.tcps_rcvbyteafterwin += todrop; m_adj(m, -todrop); ti->ti_len -= todrop; - ti->ti_flags &= ~(TH_PUSH|TH_FIN); + tiflags &= ~(TH_PUSH|TH_FIN); } } - /* - * If a segment is received on a connection after the - * user processes are gone, then RST the other end. - */ - if (so->so_state & SS_NOFDREF) { - tcp_close(tp); - tp = 0; - goto dropwithreset; - } - +#ifdef TCP_COMPAT_42 +do_rst: +#endif /* * If the RST bit is set examine the state: * SYN_RECEIVED STATE: @@ -386,23 +622,20 @@ trimthenstep6: if (tiflags&TH_RST) switch (tp->t_state) { case TCPS_SYN_RECEIVED: - tcp_drop(tp, ECONNREFUSED); - tp = 0; + tp = tcp_drop(tp, ECONNREFUSED); goto drop; case TCPS_ESTABLISHED: case TCPS_FIN_WAIT_1: case TCPS_FIN_WAIT_2: case TCPS_CLOSE_WAIT: - tcp_drop(tp, ECONNRESET); - tp = 0; + tp = tcp_drop(tp, ECONNRESET); goto drop; case TCPS_CLOSING: case TCPS_LAST_ACK: case TCPS_TIME_WAIT: - tcp_close(tp); - tp = 0; + tp = tcp_close(tp); goto drop; } @@ -411,8 +644,7 @@ trimthenstep6: * error and we send an RST and drop the connection. */ if (tiflags & TH_SYN) { - tcp_drop(tp, ECONNRESET); - tp = 0; + tp = tcp_drop(tp, ECONNRESET); goto dropwithreset; } @@ -429,19 +661,17 @@ 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)); (void) tcp_reass(tp, (struct tcpiphdr *)0); tp->snd_wl1 = ti->ti_seq - 1; /* fall into ... */ @@ -461,44 +691,163 @@ 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; -/* printf("rtt %d srtt*100 now %d\n", tp->t_rtt, (int)(tp->t_srtt*100)); */ + 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. + */ + tp->t_rtt--; + delta = tp->t_rtt - (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_rtt = 1; - 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, 65535); /* XXX */ } if (acked > so->so_snd.sb_cc) { - sbdrop(&so->so_snd, so->so_snd.sb_cc); tp->snd_wnd -= 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); @@ -518,9 +867,14 @@ 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] = TCPTV_MAXIDLE; + } tp->t_state = TCPS_FIN_WAIT_2; } break; @@ -541,17 +895,17 @@ 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) { - tcp_close(tp); - tp = 0; + tp = tcp_close(tp); + goto drop; } - goto drop; + break; /* * In TIME_WAIT state the only thing that should arrive @@ -562,21 +916,27 @@ 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; } /* @@ -584,12 +944,30 @@ step6: */ if ((tiflags & TH_URG) && ti->ti_urp && TCPS_HAVERCVDFIN(tp->t_state) == 0) { + /* + * 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 + so->so_rcv.sb_cc > SB_MAX) { + ti->ti_urp = 0; /* XXX */ + tiflags &= ~TH_URG; /* XXX */ + goto dodata; /* XXX */ + } /* * If this segment advances the known urgent pointer, * then mark the data stream. This should not happen * 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; @@ -598,7 +976,7 @@ 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. @@ -606,9 +984,18 @@ 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 && + (so->so_options & SO_OOBINLINE) == 0) tcp_pulloutofband(so, ti); - } + } 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, @@ -620,11 +1007,19 @@ step6: */ if ((ti->ti_len || (tiflags&TH_FIN)) && TCPS_HAVERCVDFIN(tp->t_state) == 0) { - tiflags = tcp_reass(tp, ti); + TCP_REASS(tp, ti, m, so, tiflags); if (tcpnodelack == 0) tp->t_flags |= TF_DELACK; else tp->t_flags |= TF_ACKNOW; + /* + * 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; @@ -685,7 +1080,8 @@ step6: /* * Return any desired output. */ - (void) tcp_output(tp); + if (needoutput || (tp->t_flags & TF_ACKNOW)) + (void) tcp_output(tp); return; dropafterack: @@ -693,22 +1089,24 @@ 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: - if (om) + if (om) { (void) m_free(om); + om = 0; + } /* * 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) || in_broadcast(ti->ti_dst)) goto drop; if (tiflags & TH_ACK) tcp_respond(tp, ti, (tcp_seq)0, ti->ti_ack, TH_RST); @@ -718,21 +1116,30 @@ dropwithreset: tcp_respond(tp, ti, ti->ti_seq+ti->ti_len, (tcp_seq)0, TH_RST|TH_ACK); } + /* destroy temporarily created socket */ + if (dropsocket) + (void) soabort(so); return; drop: + if (om) + (void) m_free(om); /* * Drop space held by incoming segment and return. */ if (tp && (tp->t_inpcb->inp_socket->so_options & SO_DEBUG)) tcp_trace(TA_DROP, ostate, tp, &tcp_saveti, 0); m_freem(m); + /* destroy temporarily created socket */ + if (dropsocket) + (void) soabort(so); return; } -tcp_dooptions(tp, om) +tcp_dooptions(tp, om, ti) struct tcpcb *tp; struct mbuf *om; + struct tcpiphdr *ti; { register u_char *cp; int opt, optlen, cnt; @@ -745,8 +1152,11 @@ tcp_dooptions(tp, om) break; if (opt == TCPOPT_NOP) optlen = 1; - else + else { optlen = cp[1]; + if (optlen <= 0) + break; + } switch (opt) { default: @@ -755,10 +1165,11 @@ tcp_dooptions(tp, om) case TCPOPT_MAXSEG: if (optlen != 4) continue; + if (!(ti->ti_flags & TH_SYN)) + continue; tp->t_maxseg = *(u_short *)(cp + 2); -#if vax || pdp11 || ns16032 tp->t_maxseg = ntohs((u_short)tp->t_maxseg); -#endif + tp->t_maxseg = MIN(tp->t_maxseg, tcp_mss(tp)); break; } } @@ -799,108 +1210,56 @@ tcp_pulloutofband(so, ti) } /* - * Insert segment ti into reassembly queue of tcp with - * control block tp. Return TH_FIN if reassembly now includes - * a segment with FIN. + * 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 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, 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_reass(tp, ti) +tcp_mss(tp) 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; + struct route *ro; + struct ifnet *ifp; + int mss; + struct inpcb *inp; - /* - * 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; + inp = tp->t_inpcb; + ro = &inp->inp_route; + if ((ro->ro_rt == (struct rtentry *)0) || + (ifp = ro->ro_rt->rt_ifp) == (struct ifnet *)0) { + /* No route yet, so try to acquire one */ + if (inp->inp_faddr.s_addr != INADDR_ANY) { + ro->ro_dst.sa_family = AF_INET; + ((struct sockaddr_in *) &ro->ro_dst)->sin_addr = + inp->inp_faddr; + rtalloc(ro); } - q = (struct tcpiphdr *)(q->ti_next); + if ((ro->ro_rt == 0) || (ifp = ro->ro_rt->rt_ifp) == 0) + return (TCP_MSS); } - /* - * 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); + mss = ifp->if_mtu - sizeof(struct tcpiphdr); +#if (MCLBYTES & (MCLBYTES - 1)) == 0 + if (mss > MCLBYTES) + mss &= ~(MCLBYTES-1); +#else + if (mss > MCLBYTES) + mss = mss / MCLBYTES * MCLBYTES; +#endif + if (in_localaddr(inp->inp_faddr)) + return (mss); -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 { -SBCHECK(&so->so_rcv, "tcp_input before"); - sbappend(&so->so_rcv, m); -SBCHECK(&so->so_rcv, "tcp_input after"); - } - } while (ti != (struct tcpiphdr *)tp && ti->ti_seq == tp->rcv_nxt); - sorwakeup(so); - return (flags); -drop: - m_freem(dtom(ti)); - return (0); + mss = MIN(mss, TCP_MSS); + tp->snd_cwnd = mss; + return (mss); }