/* * vtcp_input.c * * Derived from: * * Copyright (c) 1982, 1986 Regents of the University of California. * All rights reserved. * * Redistribution and use in source and binary forms are permitted * provided that this notice is preserved and that due credit is given * to the University of California at 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'' without express or implied warranty. * * @(#)tcp_input.c 7.13+ (Berkeley) 11/13/87 * * Modified for x-kernel v3.3 12/10/90 * Modifications Copyright (C) 1996,1990 Larry L. Peterson and Norman C. Hutchinson * * $Revision: 1.8 $ * $Date: 1998/05/13 02:37:08 $ */ #include "xkernel.h" #include "ip.h" #include "tcp_internal.h" #include "tcp_fsm.h" #include "tcp_seq.h" #include "tcp_timer.h" #include "tcp_var.h" #include "tcpip.h" #include "tcp_debug.h" #define CANTRCVMORE(so) (sototcpst(so)->closed & 2) extern int Vtcp_rcv_size; /* int tcpprintfs = 0; */ static int tcpcksum = 1; static struct tcpiphdr tcp_saveti; #ifdef __STDC__ static void tcp_dooptions(PSTATE *, struct tcpcb *, char *, int, struct tcphdr *); static void tcp_pulloutofband(Sessn, struct tcphdr *, Msg *); #else static void tcp_dooptions(); static void tcp_pulloutofband(); #endif /* * 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). * * Reassembled message, if one exists, is placed in demuxmsg * Vegas ACKS on every packet! * --------------------------- */ #define TCP_REASS_DELACK(tp, th, so, m, dMsg, flags) { \ if ((th)->th_seq == (tp)->rcv_nxt && \ (tp)->seg_next == (struct reass *)(tp) && \ (tp)->t_state == TCPS_ESTABLISHED) { \ (tp)->t_flags |= TF_DELACK; \ (tp)->rcv_nxt += msgLength(m); \ flags = (th)->th_flags & TH_FIN; \ ps->tcpstat.tcps_rcvpack++;\ ps->tcpstat.tcps_rcvbyte += msgLength(m);\ msgAssign((dMsg), (m)); \ } else { \ (flags) = tcp_reass((tp), (th), (so), (m), (dMsg)); \ (tp)->t_flags |= TF_ACKNOW; \ } \ } #define TCP_REASS_ACKNOW(tp, th, so, m, dMsg, flags) { \ if ((th)->th_seq == (tp)->rcv_nxt && \ (tp)->seg_next == (struct reass *)(tp) && \ (tp)->t_state == TCPS_ESTABLISHED) { \ /* (tp)->t_flags |= TF_DELACK; */\ (tp)->rcv_nxt += msgLength(m); \ flags = (th)->th_flags & TH_FIN; \ ps->tcpstat.tcps_rcvpack++;\ ps->tcpstat.tcps_rcvbyte += msgLength(m);\ msgAssign((dMsg), (m)); \ } else \ (flags) = tcp_reass((tp), (th), (so), (m), (dMsg)); \ (tp)->t_flags |= TF_ACKNOW; \ } int tcp_reass(tp, th, so, m, demuxmsg) register struct tcpcb *tp; register struct tcphdr *th; Sessn so; Msg *m, *demuxmsg; { register struct reass *q, *next; int flags; PSTATE *ps=(PSTATE *)so->myprotl->state; xTrace0(tcpp, TR_FUNCTIONAL_TRACE, "tcp_reass function entered"); /* * Call with th==0 after become established to * force pre-ESTABLISHED data up to user socket. */ if (th == 0) goto present; /* * Find a segment which begins after this one does. */ for (q = tp->seg_next; q != (struct reass *)tp; q = (struct reass *)q->next) if (SEQ_GT(q->th.th_seq, th->th_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 (q->prev != (struct reass *)tp) { register int i; q = q->prev; /* conversion to int (in i) handles seq wraparound */ i = q->th.th_seq + msgLength(&q->m) - th->th_seq; if (i > 0) { if (i >= msgLength(m)) { ps->tcpstat.tcps_rcvduppack++; ps->tcpstat.tcps_rcvdupbyte += msgLength(m); return(0); } msgDiscard(m, i); th->th_seq += i; } q = q->next; } ps->tcpstat.tcps_rcvoopack++; ps->tcpstat.tcps_rcvoobyte += msgLength(m); /* * While we overlap succeeding segments trim them or, * if they are completely covered, dequeue them. */ while (q != (struct reass *)tp) { register int i = (th->th_seq + msgLength(m)) - q->th.th_seq; if (i <= 0) break; if (i < msgLength(&q->m)) { q->th.th_seq += i; msgDiscard(&q->m, i); break; } next = q->next; remque(q); msgDestroy(&q->m); xFree((char *) q); q = next; } /* * Stick new segment in its place. */ { struct reass *new; new = (struct reass *)xMalloc(sizeof *new); new->th = *th; msgConstructCopy(&new->m, m); insque(new, q); /* print_reass(tp, "Inserted segment"); */ } present: /* * Present data to user, advancing rcv_nxt through * completed sequence space. */ if (TCPS_HAVERCVDSYN(tp->t_state) == 0) return (0); q = tp->seg_next; th = &q->th; if (q == (struct reass *)tp || th->th_seq != tp->rcv_nxt) return (0); if (tp->t_state == TCPS_SYN_RECEIVED && msgLength(&q->m)) return (0); do { tp->rcv_nxt += msgLength(&q->m); flags = th->th_flags & TH_FIN; next = q->next; remque(q); if (! CANTRCVMORE(so)) { msgJoin(demuxmsg, demuxmsg, &q->m); } msgDestroy(&q->m); xFree((char *)q); q = next; th = &q->th; } while (q != (struct reass *)tp && th->th_seq == tp->rcv_nxt); return (flags); } XkReturn vtcpPop(self, lls, m, hdr) Sessn self; Sessn lls; Msg *m; void *hdr; { return xDemux(xGetUp(self), self, m); } /* * TCP input routine, follows pages 65-76 of the * protocol specification dated September, 1981 very closely. */ XkReturn vtcp_input(self, transport_s, m) Protl self; Sessn transport_s; Msg *m; { struct inpcb *inp; char *options = 0; int option_len = 0; int len, tlen, off; struct tcpcb *tp = 0; register int tiflags; Sessn so = 0; Protl hlp = 0, hlpType = 0; struct tcpstate *tcpst; int todrop, acked, ourfinisacked, needoutput = 0; short ostate; int dropsocket = 0; int iss = 0; u_long tiwin; Msg demuxmsg; int demuxmsgDestroyed=0; struct tcphdr tHdr; IPpseudoHdr *pHdr; bool invoke_cantrcvmore = FALSE; int sendTime; int transmits; int currentTime; PSTATE *ps=(PSTATE *)self->state; void *buffer; u_short check_sum; ps->tcpstat.tcps_rcvtotal++; tlen = msgLength(m); pHdr = (IPpseudoHdr *)msgGetAttr(m, 0); xAssert(pHdr); /* LSB 8/9/94 */ /* * Get IP and TCP header together in stack part of message. * UNIX Note: IP leaves IP header in first mbuf. * x-kernel Note: IP attaches IP pseudoheader (in network byte * order) as the attribute of the message */ check_sum = inCkSum(m, (u_short *)pHdr, sizeof(IPpseudoHdr)); buffer = msgPop(m, sizeof(tHdr)); if (! buffer) { xTrace0(tcpp, 3, "vtcp_input: msgPop of header failed -- dropping"); return XK_FAILURE; } tcpHdrLoad(&tHdr, buffer, sizeof(tHdr), m); msgSetAttr(m, 0, 0, 0); xTrace4(tcpp, 3, "vtcp_input seq %d, dlen %d, f( %s ) to port (%d)", tHdr.th_seq, msgLength(m), tcpFlagStr(tHdr.th_flags), tHdr.th_dport); /* millisecond clock assumption */ currentTime = tcpGetTime(); DO_TRACE(ps, TCP_EVENT_IN, currentTime, msgLength(m), GET_TID(tHdr.th_sport,tHdr.th_dport)); DO_TRACE(ps, TCP_EVENT_DATA, tHdr.th_seq, tHdr.th_win>>4, 0); DO_TRACE(ps, TCP_EVENT_DATA, tHdr.th_ack, tHdr.th_flags, 0); msgConstructEmpty(&demuxmsg); /* * Check the checksum value (calculated in tcpHdrLoad) */ if (tcpcksum) { if (check_sum) { xTrace1(tcpp, TR_ERRORS, "vtcp_input: bad checksum (%x)", check_sum); ps->tcpstat.tcps_rcvbadsum++; #if BSD<=43 ps->tcpstat.tcps_badsum++; #endif DO_TRACE(ps, TCP_EVENT_BADSUM, 0, 0, 0); goto drop; } } xTrace1(tcpp, 4, "Received msg with sequence number %d", tHdr.th_seq); /* * Check that TCP offset makes sense, * pull out TCP options and adjust length. */ off = tHdr.th_off << 2; if (off < sizeof (struct tcphdr) || off > tlen) { xTrace2(tcpp, 1, "bad tcp off: src %x off %d", pHdr->src, off); ps->tcpstat.tcps_rcvbadoff++; #if BSD<=43 ps->tcpstat.tcps_badoff++; #endif goto drop; } tlen -= off; if ((option_len = off - sizeof (struct tcphdr)) > 0) { xTrace1(tcpp, 3, "vtcp_input: %d bytes of options", option_len); /* * Re-check the length, cause option_len increases the size * of the tcp header */ if (tlen < 0) { xTrace2(tcpp, 2, "vtcp_input: rcvd short optlen = %d, len = %d", option_len, tlen); ps->tcpstat.tcps_rcvshort++; msgDestroy(&demuxmsg); return XK_FAILURE; } /* * Put the options somewhere reasonable */ DO_TRACE(ps, TCP_EVENT_ROPT, option_len, 0, 0); options = xMalloc(option_len); buffer = msgPop(m, option_len); xAssert(buffer); tcpOptionsLoad(options, buffer, option_len, 0); } tiflags = tHdr.th_flags; /* * Locate pcb for segment. */ findpcb: { ActiveId activeId; PassiveId passiveId; PSTATE *pstate; pstate = (PSTATE *)self->state; bzero((char *)&activeId, sizeof(activeId)); activeId.localport = tHdr.th_dport; activeId.remoteport = tHdr.th_sport; activeId.remoteaddr = pHdr->src; xTrace3(tcpp, 4, "Looking for %d->%s.%d", activeId.localport, ipHostStr(&activeId.remoteaddr), activeId.remoteport); /* look in the active map */ if (mapResolve(pstate->activeMap, &activeId, (void **)&so) == XK_FAILURE) { /* * Look in the passive map */ Enable *e; passiveId = activeId.localport; xTrace1(tcpp, 4, "Looking for passive open on %d", passiveId); /* look in the map */ if (mapResolve(pstate->passiveMap, &passiveId, (void **)&e) == XK_FAILURE) { xTrace0(tcpp, 4, "No passive open object exists"); so = 0; hlp = 0; inp = 0; tp = 0; } else { hlp = e->hlp; hlpType = e->hlpType; Vtcp_rcv_size = (int)e->info; xTrace1(tcpp, 4, "Found openenable object: %lx", (u_long)so); inp = 0; tp = 0; } } else { inp = sototcpst(so)->inpcb; tp = sototcpst(so)->tcpcb; } } /* * 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 (!so && !hlp) { xTrace0(tcpp, 2, "vtcp_input: no so"); goto dropwithreset; } if (xIsProtl(hlp)) { so = vtcp_sonewconn(self, hlp, hlpType, &pHdr->src, &pHdr->dst, tHdr.th_sport, tHdr.th_dport); 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 = sotoinpcb(so); bcopy((char *)&pHdr->dst, (char *)&inp->inp_laddr, sizeof(pHdr->dst)); inp->inp_lport = tHdr.th_dport; tp = intotcpcb(inp); tp->t_state = TCPS_LISTEN; /* Compute proper scaling value from buffer space */ while (tp->request_r_scale < TCP_MAX_WINSHIFT && TCP_MAXWIN << tp->request_r_scale < sototcpst(so)->rcv_hiwat) tp->request_r_scale++; } xIfTrace(tcpp, 1) { ostate = tp->t_state; #if 0 tcp_saveti = *ti; #endif } tcpst = sototcpst(so); if (inp == 0) { xTrace0(tcpp, 2, "vtcp_input: no inp"); goto dropwithreset; } if (tp == 0) { xTrace0(tcpp, 2, "vtcp_input: no tp"); goto dropwithreset; } if (tp->t_state == TCPS_CLOSED) { xTrace0(tcpp, 2, "vtcp_input: TCPS_CLOSED"); goto drop; } /* Unscale the window into a 32-bit value. */ if ((tiflags & TH_SYN) == 0) tiwin = tHdr.th_win << tp->snd_scale; else tiwin = tHdr.th_win; /* * Proposed tcp_demux / tcp_pop split */ /* * Segment received on connection. * Reset idle time and keep-alive timer. */ tp->t_idle = 0; tp->t_timer[TCPT_KEEP] = TCPTV_KEEP; /* * Process options if not in LISTEN state, * else do it below (after getting remote address). */ if (options && tp->t_state != TCPS_LISTEN) { tcp_dooptions(ps, tp, options, option_len, &tHdr); options = 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. */ { int win; win = tcpst->rcv_space; if (win < 0) win = 0; tp->rcv_wnd = MAX(win, (int)(tp->rcv_adv - tp->rcv_nxt)); xTrace4(tcpp, 5, "New win (%x) = max(std (%x), adv (%x) - nxt (%x))", tp->rcv_wnd, win, tp->rcv_adv, tp->rcv_nxt); } xTrace2(tcpp, 4, "vtcp_input: tcpcb %lx switch %s", (u_long)tp, tcpstates[tp->t_state]); DO_TRACE(ps, TCP_EVENT_IN0, tp->rcv_nxt, tp->t_state, tcpst->tid); switch (tp->t_state) { /* * 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: * * Also initialize tp->snd_nxt to tp->iss+1 and tp->snd_una to tp->iss. * Fill in remote peer address fields if not previously specified. * Enter SYN_RECEIVED state, and process any other fields of this * segment in this state. */ case TCPS_LISTEN: { if (tiflags & TH_RST) goto drop; if (tiflags & TH_ACK) goto dropwithreset; if ((tiflags & TH_SYN) == 0) goto drop; if (in_broadcast(*(struct in_addr *)&pHdr->dst)) goto drop; xTrace0(tcpp, 3, "vtcp_input: LISTEN"); tp->t_template = tcp_template(tp); if (tp->t_template == 0) { tp = tcp_drop(tp, ENOBUFS); dropsocket = 0; /* socket is already gone */ goto drop; } if (options) { tcp_dooptions(ps, tp, options, option_len, &tHdr); } if (iss) tp->iss = iss; else tp->iss = ps->tcp_iss; ps->tcp_iss += TCP_ISSINCR/2; tp->irs = tHdr.th_seq; tcp_sendseqinit(tp); tp->v_cong_detect_begseq = tp->snd_nxt; /* Vegas var */ 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 */ ps->tcpstat.tcps_accepts++; goto trimthenstep6; } /* * If the state is SYN_SENT: * if seg contains an ACK, but not for our SYN, drop the input. * if seg contains a RST, then drop the connection. * if seg does not contain SYN, then drop it. * Otherwise this is an acceptable SYN segment * initialize tp->rcv_nxt and tp->irs * if seg contains ack then advance tp->snd_una * if SYN has been acked change to ESTABLISHED else SYN_RCVD state * arrange for segment to be acked (eventually) * continue processing rest of data/controls, beginning with URG */ case TCPS_SYN_SENT: if ((tiflags & TH_ACK) && (SEQ_LEQ(tHdr.th_ack, tp->iss) || SEQ_GT(tHdr.th_ack, tp->snd_max))) { xTrace0(tcpp, 4, "input state SYN_SENT -- dropping with reset"); xTrace3(tcpp, 4, " (ack==%d, iss==%d, snd_max==%d)", tHdr.th_ack, tp->iss, tp->snd_max); goto dropwithreset; } if (tiflags & TH_RST) { if (tiflags & TH_ACK) tp = tcp_drop(tp, ECONNREFUSED); goto drop; } if ((tiflags & TH_SYN) == 0) goto drop; if (tiflags & TH_ACK) { tp->snd_una = tHdr.th_ack; if (SEQ_LT(tp->snd_nxt, tp->snd_una)) tp->snd_nxt = tp->snd_una; } tp->t_timer[TCPT_REXMT] = 0; tp->irs = tHdr.th_seq; tcp_rcvseqinit(tp); tp->t_flags |= TF_ACKNOW; tp->v_cong_detect_begtime = currentTime; /* Vegas var */ if (tiflags & TH_ACK && SEQ_GT(tp->snd_una, tp->iss)) { ps->tcpstat.tcps_connects++; soisconnected(so); tp->t_state = TCPS_ESTABLISHED; if ((tp->t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) == (TF_RCVD_SCALE|TF_REQ_SCALE)) { tp->snd_scale = tp->requested_s_scale; tp->rcv_scale = tp->request_r_scale; } tp->t_maxseg = MIN(tp->t_maxseg, tcp_mss(tp)); (void) tcp_reass(tp, NULL, so, NULL, &demuxmsg); /* * if we didn't have to retransmit the SYN, * use its rtt as our initial srtt & rtt var. */ if (tp->t_rtt) { /* Initialize Vegas RTT variables */ if (tp->v_timeout == 1000 && tp->v_synsent != 0) { tp->v_rtt = currentTime - tp->v_synsent + 1; tp->v_sa = tp->v_rtt << 3; tp->v_sd = tp->v_rtt; tp->v_timeout = ((tp->v_sa >> 2) + tp->v_sd) >> 1; DO_TRACE(ps, TCP_EVENT_SET_TO, tp->v_timeout, tp->v_rtt, tHdr.th_dport); } 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); DO_TRACE(ps, TCP_EVENT_OTHER9, tp->t_rxtcur, 4, 0); tp->t_rtt = 0; } } else tp->t_state = TCPS_SYN_RECEIVED; trimthenstep6: /* * Advance ti->ti_seq to correspond to first data byte. * If data, trim to stay within window, * dropping FIN if necessary. */ tHdr.th_seq++; if (tlen > tp->rcv_wnd) { todrop = tlen - tp->rcv_wnd; xTrace2(tcpp, 5, "tlen (%d) > rcv_wnd (%d) ... truncating", tlen, tp->rcv_wnd); msgTruncate(m, todrop); tlen = tp->rcv_wnd; tiflags &= ~TH_FIN; ps->tcpstat.tcps_rcvpackafterwin++; ps->tcpstat.tcps_rcvbyteafterwin += todrop; } tp->snd_wl1 = tHdr.th_seq - 1; tp->rcv_up = tHdr.th_seq; goto step6; } /* * 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 - tHdr.th_seq; if (todrop > 0) { if (tiflags & TH_SYN) { tiflags &= ~TH_SYN; tHdr.th_seq++; if (tHdr.th_urp > 1) tHdr.th_urp--; else tiflags &= ~TH_URG; todrop--; } if (todrop > tlen || todrop == tlen && (tiflags & TH_FIN) == 0) { #ifdef TCP_COMPAT_42 /* * Don't toss RST in response to 4.2-style keepalive. */ if (tHdr.th_seq == tp->rcv_nxt - 1 && tiflags & TH_RST) goto do_rst; #endif ps->tcpstat.tcps_rcvduppack++; ps->tcpstat.tcps_rcvdupbyte += tlen; todrop = tlen; tiflags &= ~TH_FIN; tp->t_flags |= TF_ACKNOW; } else { ps->tcpstat.tcps_rcvpartduppack++; ps->tcpstat.tcps_rcvpartdupbyte += todrop; } xTrace1(tcpp, 5, "Discarding %d bytes from front of msg", todrop); msgDiscard(m, todrop); tHdr.th_seq += todrop; tlen -= todrop; if (tHdr.th_urp > todrop) tHdr.th_urp -= todrop; else { tiflags &= ~TH_URG; tHdr.th_urp = 0; } } /* * If new data is received on a connection after the * user processes are gone, then RST the other end. */ #define ISNOTREFERENCED(X) 0 if (ISNOTREFERENCED(so) && tp->t_state > TCPS_CLOSE_WAIT && tlen) { tp = tcp_destroy(tp); ps->tcpstat.tcps_rcvafterclose++; goto dropwithreset; } /* * If segment ends after window, drop trailing data * (and PUSH and FIN); if nothing left, just ACK. */ todrop = (tHdr.th_seq + tlen) - (tp->rcv_nxt + tp->rcv_wnd); if (todrop > 0) { ps->tcpstat.tcps_rcvpackafterwin++; if (todrop >= tlen) { ps->tcpstat.tcps_rcvbyteafterwin += tlen; /* * 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(tHdr.th_seq, tp->rcv_nxt)) { iss = tp->rcv_nxt + TCP_ISSINCR; (void) tcp_destroy(tp); goto findpcb; } /* * 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 && tHdr.th_seq == tp->rcv_nxt) { tp->t_flags |= TF_ACKNOW; ps->tcpstat.tcps_rcvwinprobe++; } else goto dropafterack; } else ps->tcpstat.tcps_rcvbyteafterwin += todrop; xTrace1(tcpp, 5, "segment ends after window -- truncating to %d", todrop); msgTruncate(m, todrop); tlen -= todrop; tiflags &= ~(TH_PUSH|TH_FIN); } #ifdef TCP_COMPAT_42 do_rst: #endif /* * If the RST bit is set examine the state: * SYN_RECEIVED STATE: * If passive open, return to LISTEN state. * If active open, inform user that connection was refused. * ESTABLISHED, FIN_WAIT_1, FIN_WAIT2, CLOSE_WAIT STATES: * Inform user that connection was reset, and close tcb. * CLOSING, LAST_ACK, TIME_WAIT STATES * Close the tcb. */ if (tiflags & TH_RST) switch (tp->t_state) { case TCPS_SYN_RECEIVED: tp = tcp_drop(tp, ECONNREFUSED); goto drop; case TCPS_ESTABLISHED: case TCPS_FIN_WAIT_1: case TCPS_FIN_WAIT_2: case TCPS_CLOSE_WAIT: tp = tcp_drop(tp, ECONNRESET); goto drop; case TCPS_CLOSING: case TCPS_LAST_ACK: case TCPS_TIME_WAIT: tp = tcp_destroy(tp); goto drop; } /* * If a SYN is in the window, then this is an * error and we send an RST and drop the connection. */ if (tiflags & TH_SYN) { tp = tcp_drop(tp, ECONNRESET); goto dropwithreset; } /* * If the ACK bit is off we drop the segment and return. */ if ((tiflags & TH_ACK) == 0) goto drop; /* * Ack processing. */ switch (tp->t_state) { /* * In SYN_RECEIVED state if the ack ACKs our SYN then enter * ESTABLISHED state and continue processing, otherwise * send an RST. */ case TCPS_SYN_RECEIVED: if (SEQ_GT(tp->snd_una, tHdr.th_ack) || SEQ_GT(tHdr.th_ack, tp->snd_max)) goto dropwithreset; ps->tcpstat.tcps_connects++; soisconnected(so); tp->t_state = TCPS_ESTABLISHED; /* Do window scaling? */ if ((tp->t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) == (TF_RCVD_SCALE|TF_REQ_SCALE)) { tp->snd_scale = tp->requested_s_scale; tp->rcv_scale = tp->request_r_scale; } tp->t_maxseg = MIN(tp->t_maxseg, tcp_mss(tp)); (void) tcp_reass(tp, NULL, so, NULL, &demuxmsg); tp->snd_wl1 = tHdr.th_seq - 1; /* fall into ... */ /* * In ESTABLISHED state: drop duplicate ACKs; ACK out of range * ACKs. If the ack is in the range * tp->snd_una < ti->ti_ack <= tp->snd_max * then advance tp->snd_una to ti->ti_ack and drop * data from the retransmission queue. If this ACK reflects * more up to date window information we update our window information. */ case TCPS_ESTABLISHED: case TCPS_FIN_WAIT_1: case TCPS_FIN_WAIT_2: case TCPS_CLOSE_WAIT: case TCPS_CLOSING: case TCPS_LAST_ACK: case TCPS_TIME_WAIT: if (SEQ_LEQ(tHdr.th_ack, tp->snd_una)) { if (tlen == 0 && tiwin == tp->snd_wnd) { ps->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. * * We know we're losing at the current * window size so do congestion avoidance * (set ssthresh to half the current window * and pull our congestion window back to * the new ssthresh). * * Dup acks mean that packets have left the * network (they're now cached at the receiver) * so bump cwnd by the amount in the receiver * to keep a constant cwnd packets in the * network. */ if (tp->t_timer[TCPT_REXMT] == 0 || tHdr.th_ack != tp->snd_una) tp->t_dupacks = 0; else { /* we have a duplicate ACK */ /* Vegas retransmit mods: * Find out if any segments timed out. Sbexpire * sets transmits and sendTime and returns * the number of bytes contained in all segments * which have timed out. Transmits is * the number of times the first segment that * needs to be retransmitted has been transmitted * already. SendTime is the time when the first * segment was last transmitted. */ len = sbexpire(tcpst->snd, currentTime-tp->v_timeout, tp->t_maxseg, &transmits, &sendTime); /* if ((tp->snd_nxt-tp->snd_una >= 2*tp->t_maxseg && */ if (( len >= tp->t_maxseg) || ++tp->t_dupacks == tp->t_rexmtthresh) { tcp_seq onxt=tp->snd_nxt; u_long win = MIN(tp->snd_wnd, tp->snd_cwnd); u_long old_cwnd=tp->snd_cwnd; /* we will check for timed out segments when * we receive the next two non-duplicate ACKS * (or less depending on how much is in transit). */ tp->v_worried = MIN(2, (tp->snd_nxt-tp->snd_una)/ tp->t_maxseg - 1); tp->snd_nxt = tHdr.th_ack; tp->snd_cwnd = tp->t_maxseg; if (tp->t_dupacks != tp->t_rexmtthresh) { ps->tcpstat.tcps_dupftimeo++; DO_TRACE(ps, TCP_EVENT_WSND, len, tp->v_worried, tHdr.th_dport); } /* Do exponential backoff if segment has * been sent more than once */ if (transmits > 1) { DO_TRACE(ps, TCP_EVENT_RTO, tp->v_timeout<<1, tp->v_timeout, tHdr.th_dport); tp->v_timeout <<= 1; /* if (tp->t_dupacks != tp->t_rexmtthresh) */ /* break; */ } else tp->v_timeout += (tp->v_timeout >> 3); /* If cwnd hasn't changed since the segment * was sent, then we need to decrease it */ if (tp->v_time_cwin_chg < sendTime) { win /= tp->t_maxseg; if (win <= 3) win = 2; else if (transmits > 1) win /= 2; else win -= (win >> 2); tp->v_newcwnd = win * tp->t_maxseg; tp->t_timer[TCPT_REXMT] = 0; tp->t_rtt = 0; (void) vtcp_output(tp); tp->v_time_cwin_chg = tcpGetTime(); tp->snd_cwnd = tp->v_newcwnd + tp->t_maxseg * tp->t_dupacks; if (len > 0) { DO_TRACE(ps, TCP_EVENT_RCWND, tp->snd_cwnd, TCP_NOTE_WORRIED, tHdr.th_dport); } else { DO_TRACE(ps, TCP_EVENT_RCWND, tp->snd_cwnd, TCP_NOTE_DUPACK, tHdr.th_dport); } } else { /* no need to change cwnd */ (void) vtcp_output(tp); tp->snd_cwnd = old_cwnd; } if (transmits == 1) tp->t_dupacks = tp->t_rexmtthresh; if (SEQ_GT(onxt, tp->snd_nxt)) tp->snd_nxt = onxt; goto drop; } else if (tp->t_dupacks > tp->t_rexmtthresh) { if (tp->snd_cwnd < (65535 - tp->t_maxseg)) tp->snd_cwnd += tp->t_maxseg; /* Toggle for exponential decrease of * sending rate after losses, instead of * waiting half a RTT before starting sending * again. Currently not used. */ /* if ((tp->v_send_toggle = !tp->v_send_toggle)) tp->v_use_sndwnd = 1; */ (void) vtcp_output(tp); goto drop; } } } else tp->t_dupacks = 0; break; } /* * If the congestion window was inflated to account * for the other side's cached packets, retract it. */ if (tp->t_dupacks > tp->t_rexmtthresh && tp->snd_cwnd > tp->v_newcwnd) { tp->snd_cwnd = tp->v_newcwnd; /* In Vegas ssthresh is only used during slow-start */ tp->snd_ssthresh = 2*tp->t_maxseg; } tp->t_dupacks = 0; if (SEQ_GT(tHdr.th_ack, tp->snd_max)) { ps->tcpstat.tcps_rcvacktoomuch++; goto dropafterack; } acked = tHdr.th_ack - tp->snd_una; ps->tcpstat.tcps_rcvackpack++; ps->tcpstat.tcps_rcvackbyte += acked; xTrace2(tcpp, 4, "Received ACK for byte %d (%d new bytes ACKED)", tHdr.th_ack, acked); DO_TRACE(ps, TCP_EVENT_IN1, tHdr.th_ack, acked, tcpst->tid); /* * 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(tHdr.th_ack, tp->t_rtseq)) { ps->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); DO_TRACE(ps, TCP_EVENT_OTHER9, tp->t_rxtcur, 4, 0); } /* * 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 (tHdr.th_ack == tp->snd_max) { tp->t_timer[TCPT_REXMT] = 0; needoutput = 1; } else if (tp->t_timer[TCPT_PERSIST] == 0) tp->t_timer[TCPT_REXMT] = tp->t_rxtcur; /* Reno did this: * 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). * * Vegas modifies it by adding congestion detection/avoidance. * So we may not only not increase the congestion window, * we may even decrease it before losses occur. * * Once per RTT we need to do the congestion avoidance * stuff, and the modified slow-start (only increasing * every other RTT). We need to add a fast path so as not * to do this is only one or two segments in transit * (LAN case)? */ if (tHdr.th_ack > tp->v_cong_detect_begseq) { int rtt, rttLen, sndEpoch, actual_rate, pred_rate, win; if (tp->v_cong_detect_cntRTT > 0) rtt = tp->v_cong_detect_sumRTT/tp->v_cong_detect_cntRTT; else rtt = currentTime - tp->v_cong_detect_begtime + 1; tp->v_cong_detect_sumRTT = tp->v_cong_detect_cntRTT = 0; rttLen = tp->snd_nxt - tp->v_cong_detect_begseq; /* If there was only one packet in transit, then * update baseRTT. */ if (rttLen <= tp->t_maxseg && rtt > 0) { tp->v_cong_detect_baseRtt = rtt; DO_TRACE(ps, TCP_EVENT_OTHER9, rtt, rttLen, tHdr.th_dport); } if (rtt > 0) { /* The following rates are in 100s bytes/sec. This assumes * that the time is given in milliseconds. */ actual_rate = rttLen*10/rtt; pred_rate = (tp->snd_nxt - tp->snd_una + MIN(tp->t_maxseg - acked, 0))*10 / tp->v_cong_detect_baseRtt; DO_TRACE(ps, TCP_EVENT_OTHER6, actual_rate/10, pred_rate/10, tHdr.th_dport); /* If we are currently doing slow-start */ if (tp->snd_cwnd < tp->snd_ssthresh) { /* Check if we are sending faster than available bandwidth */ if ((pred_rate - actual_rate)*tp->v_cong_detect_baseRtt > 10*tp->v_exp_inc_nseg*tp->t_maxseg) tp->v_exp_inc_isless = 1; else tp->v_exp_inc_isless = 0; /* sndEpoch is the interval of time during which * we sent data. During slow-start, we usually * don't send during the whole RTT, only during * part of the beginning. */ sndEpoch = tp->v_cong_detect_last_sendtime - tp->v_cong_detect_begtime; win = tp->snd_cwnd - tp->t_maxseg; DO_TRACE(ps, TCP_EVENT_OTHER8, sndEpoch, rtt, tHdr.th_dport); /* The following is executed if we are trying to * predict the available bandwidth during slow-start */ if (tp->v_cong_detect_predict_do && sndEpoch > 0) { if (tp->v_exp_inc_do) { if (tp->v_exp_inc_flag) { if (tp->snd_cwnd >= 8*tp->t_maxseg) { tp->snd_ssthresh = (win*rtt)/(2*sndEpoch); tp->v_cong_detect_predict_do = 0; } } else { if (tp->snd_cwnd >= 4*tp->t_maxseg) { tp->snd_ssthresh = (win*rtt)/(sndEpoch); tp->v_cong_detect_predict_do = 0; } } } else if (tp->snd_cwnd >= 8*tp->t_maxseg) { tp->snd_ssthresh = (win*rtt)/(2*sndEpoch); tp->v_cong_detect_predict_do = 0; } if (tp->snd_ssthresh <= tp->snd_cwnd) tp->snd_ssthresh = 2*tp->t_maxseg; } tp->v_exp_inc_flag = !tp->v_exp_inc_flag; DO_TRACE(ps, TCP_EVENT_OTHER5, tp->v_exp_inc_isless, tp->v_exp_inc_flag, tHdr.th_dport); } /* Check if we are sending too slowly (using less than * top_nseg buffers */ if ((pred_rate - actual_rate)*tp->v_cong_detect_baseRtt > 10*tp->v_cong_detect_top_nseg*tp->t_maxseg) tp->v_cong_detect_top_isless = 1; else tp->v_cong_detect_top_isless = 0; /* Check if we are sending to fast (using more than * bot_nseg worth of buffers */ if ((pred_rate - actual_rate)*tp->v_cong_detect_baseRtt > 10*tp->v_cong_detect_bot_nseg*tp->t_maxseg) tp->v_cong_detect_bot_isless = 1; else tp->v_cong_detect_bot_isless = 0; } /* rtt > 0 */ tp->v_cong_detect_begseq = tp->snd_nxt; tp->v_cong_detect_begtime = currentTime; DO_TRACE(ps, TCP_EVENT_OTHER5, tp->v_cong_detect_bot_isless, tp->v_cong_detect_top_isless, tHdr.th_dport); /* This next section decides by how much to increase * or decrease the congestion window when each ACK is * received. In Reno, it is increased by maxseg during * slow-start and by maxseg*(maxseg/cwnd) otherwise. * * In Vegas, the congestion avoidance mechanism (when * active) decides by how much. If it is active during * slow-start, then the increases only happen during * every other RTT, and then the increase is maxseg, * as long as we have not overran the available bw * (exp_inc_isless == 0). * If we are not doing slow-start right now, * then if we need to send faster (top_isless == 0) * we increase it by maxseg*(maxseg/cwnd), if we * need to send slower, then we decrease cwnd by * maxseg immediately and we don't change the cwnd * during the next RTT. */ if (tp->snd_cwnd < tp->snd_ssthresh) { /* Doing slow-start right now */ if (tp->v_exp_inc_do) { /* Doing congestion avoidance during slow-start */ if (!tp->v_exp_inc_flag) /* Not increasing during this RTT */ tp->v_incr = 0; else if (tp->v_exp_inc_isless) { /* Going faster than available bw, so slow down */ tp->snd_ssthresh = 2*tp->t_maxseg; tp->snd_cwnd -= (tp->snd_cwnd >> 3); tp->v_incr = 0; } else /* Increase by one maxseg with each ACK */ tp->v_incr = tp->t_maxseg; } else { /* Not doing congestion avoidance during slow-start */ tp->v_incr = tp->t_maxseg; } } /* end of doing slow-start right now */ else { /* not doing slow-start right now */ if (tp->v_cong_detect_do) { /* Doing congestion detection/avoidance */ if (tp->v_cong_detect_bot_isless) { /* Going too fast, so slow down */ tp->snd_cwnd -= tp->t_maxseg; if (tp->snd_cwnd < 2*tp->t_maxseg) tp->snd_cwnd = 2*tp->t_maxseg; tp->v_incr = 0; } else if (tp->v_cong_detect_top_isless) /* The current rate is fine */ tp->v_incr = 0; else /* We can go faster */ tp->v_incr = MAX(((int)tp->t_maxseg)*tp->t_maxseg / tp->snd_cwnd, 1); } else /* Not doing congestion detection/avoidance */ tp->v_incr = MAX(((int)tp->t_maxseg)*tp->t_maxseg / tp->snd_cwnd, 1); } /* end of not doing slow-start right now */ } /* End of code done once per RTT */ /* Since we set how much to increase only once per RTT, * we need to check during slow-start if we have * surpassed sstthresh before the RTT period is over. */ if (tp->v_incr == tp->t_maxseg && tp->snd_cwnd >= tp->snd_ssthresh) tp->v_incr = 0; xTrace4(tcpp, TR_MAJOR_EVENTS, "vtcp_input: congestion window=%d (MIN(%d+%d,65535<<%d))", tp->snd_cwnd + tp->v_incr, tp->snd_cwnd, tp->v_incr, tp->snd_scale); /* Increase congestion window unless we haven't been able * to keep up with the congestion window. */ if (tp->v_incr > 0 && (tp->snd_cwnd - (tp->snd_nxt - tp->snd_una)) <= 2*tp->t_maxseg) tp->snd_cwnd = MIN(tp->snd_cwnd + tp->v_incr, TCP_MAXWIN<snd_scale); /* sbdrop will drop all acked bytes from the send buffer. * sendTime is set to the time the last acked segment * was sent, transmits to the largest number of times * an acked segment was sent. */ if (acked > sblength(tcpst->snd)) { tp->snd_wnd -= sblength(tcpst->snd); sbdrop(tcpst->snd, sblength(tcpst->snd), &sendTime, &transmits); ourfinisacked = 1; } else { sbdrop(tcpst->snd, acked, &sendTime, &transmits); tp->snd_wnd -= acked; ourfinisacked = 0; } /* Do Vegas' RTT stuff. Note that Vegas ACKs every packet. * If running against a TCP which doesn't ACK every packet, * it may be a good idea to change sbdrop so sendTime * returns the time of the earliest segment sent. I'm not * sure if it is really needed, as Vegas only checks for * timeouts when ACKs are received (usually duplicate ACKs). */ if (( sendTime != 0) && (transmits == 1)) { int rtt, n; rtt = currentTime - sendTime + 1; tp->v_cong_detect_sumRTT += rtt; tp->v_cong_detect_cntRTT++; if (rtt > 0) { tp->v_rtt = rtt; /* The following is used fpr spike prevention */ tp->v_spike_timeInc = rtt*tp->t_maxseg/(tp->snd_cwnd+1); if (tp->v_rtt < tp->v_cong_detect_baseRtt) tp->v_cong_detect_baseRtt = tp->v_rtt; n = tp->v_rtt - (tp->v_sa >> 3); tp->v_sa += n; if (n <0) n = -n; n -= (tp->v_sd >> 2); tp->v_sd += n; tp->v_timeout = (((tp->v_sa >> 2) + tp->v_sd) >> 1); /* Since we use more accurate time, we need to make it a little bigger */ tp->v_timeout += (tp->v_timeout >> 4); DO_TRACE(ps, TCP_EVENT_SET_TO, tp->v_timeout, tp->v_rtt, tHdr.th_dport); } } /* End of Vegas' RTT stuff */ if (tcpst->waiting.waitCount) { tcpSemVAll(&tcpst->waiting); } #if 0 if ((so->so_snd.sb_flags & SB_WAIT) || so->so_snd.sb_sel) sowwakeup(so); #endif tp->snd_una = tHdr.th_ack; if (SEQ_LT(tp->snd_nxt, tp->snd_una)) tp->snd_nxt = tp->snd_una; /* If this is the first or second ACK after a retransmit, * check for timed out segments (Vegas). */ if (tp->v_worried > 0) { tp->v_worried--; /* Check if any segments have expired. If so, we * will retransmit one segment but there is no need * to change the congestion window since we already * did it. */ len = sbexpire(tcpst->snd, currentTime-tp->v_timeout, tp->t_maxseg, &transmits, &sendTime); if (len) { tcp_seq old_next=tp->snd_nxt; u_long old_cwnd=tp->snd_cwnd; ps->tcpstat.tcps_worryftimeo++; DO_TRACE(ps, TCP_EVENT_WSND, len, tp->v_worried, tHdr.th_dport); tp->t_dupacks = tp->t_rexmtthresh; tp->snd_nxt = tHdr.th_ack; tp->snd_cwnd = tp->t_maxseg; (void) vtcp_output(tp); if (SEQ_GT(old_next, tp->snd_nxt)) tp->snd_nxt = old_next; tp->snd_cwnd = old_cwnd; } else tp->v_worried = 0; } switch (tp->t_state) { /* * In FIN_WAIT_1 STATE in addition to the processing * for the ESTABLISHED state if our FIN is now acknowledged * then enter FIN_WAIT_2. */ case TCPS_FIN_WAIT_1: if (ourfinisacked) { /* * 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 (CANTRCVMORE(so)) { soisdisconnected(so); tp->t_timer[TCPT_2MSL] = TCPTV_MAXIDLE; } tp->t_state = TCPS_FIN_WAIT_2; } break; /* * In CLOSING STATE in addition to the processing for * the ESTABLISHED state if the ACK acknowledges our FIN * then enter the TIME-WAIT state, otherwise ignore * the segment. */ case TCPS_CLOSING: if (ourfinisacked) { tp->t_state = TCPS_TIME_WAIT; vtcp_canceltimers(tp); tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL; soisdisconnected(so); } break; /* * 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) { tp = tcp_destroy(tp); goto drop; } break; /* * In TIME_WAIT state the only thing that should arrive * is a retransmission of the remote FIN. Acknowledge * it and restart the finack timer. */ case TCPS_TIME_WAIT: tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL; goto dropafterack; } } step6: xTrace1(tcpp, 4, "vtcp_input: step6 on %lx", tp); /* * Update window information. * Don't look at window if no ACK: TAC's send garbage on first SYN. */ if ((tiflags & TH_ACK) && (SEQ_LT(tp->snd_wl1, tHdr.th_seq) || tp->snd_wl1 == tHdr.th_seq && (SEQ_LT(tp->snd_wl2, tHdr.th_ack) || tp->snd_wl2 == tHdr.th_ack && tiwin > tp->snd_wnd))) { /* keep track of pure window updates */ if (tlen == 0 && tp->snd_wl2 == tHdr.th_ack && tiwin > tp->snd_wnd) ps->tcpstat.tcps_rcvwinupd++; tp->snd_wnd = tiwin; tp->snd_wl1 = tHdr.th_seq; tp->snd_wl2 = tHdr.th_ack; if (tp->snd_wnd > tp->max_sndwnd) tp->max_sndwnd = tp->snd_wnd; needoutput = 1; } /* * Process segments with URG. */ if ((tiflags & TH_URG) && tHdr.th_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 (tHdr.th_urp > SB_MAX) { tHdr.th_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(tHdr.th_seq+tHdr.th_urp, tp->rcv_up)) { tp->rcv_up = tHdr.th_seq + tHdr.th_urp; sohasoutofband(so, (tcpst->rcv_hiwat - tcpst->rcv_space) + (tp->rcv_up - tp->rcv_nxt) - 1); tp->t_oobflags &= ~(TCPOOB_HAVEDATA | TCPOOB_HADDATA); } /* * Remove out of band data so doesn't get presented to user. * This can happen independent of advancing the URG pointer, * but if two URG's are pending at once, some out-of-band * data may creep in... ick. */ if (tHdr.th_urp <= tlen && (tcpst->flags & TCPST_OOBINLINE) == 0) tcp_pulloutofband(so, &tHdr, m); } 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 */ xTrace1(tcpp, 4, "vtcp_input: do data on %lx", (u_long)tp); /* * Process the segment text, merging it into the TCP sequencing queue, * and arranging for acknowledgment of receipt if necessary. * This process logically involves adjusting tp->rcv_wnd as data * is presented to the user (this happens in tcp_usrreq.c, * case PRU_RCVD). If a FIN has already been received on this * connection then we just ignore the text. */ if ((tlen || (tiflags&TH_FIN)) && TCPS_HAVERCVDFIN(tp->t_state) == 0) { xTrace1(tcpp, 5, "Calling macro reassemble with msg len %d", msgLength(m)); if (ps->delack) { TCP_REASS_DELACK(tp, &tHdr, so, m, &demuxmsg, tiflags); } else { TCP_REASS_ACKNOW(tp, &tHdr, so, m, &demuxmsg, tiflags); } xTrace1(tcpp, 5, "after reassembly demux msg has length %d", msgLength(m)); /* * Pass received message to user: */ { int mlen = msgLength(&demuxmsg); if (mlen != 0) { tcpst->rcv_space -= mlen; xPop(so, transport_s, &demuxmsg, (void *)0); } /* if */ msgDestroy(&demuxmsg); demuxmsgDestroyed = 1; } if (ps->delack) tp->t_flags |= TF_DELACK; /* * The following else causes every packet to be acked * twice, once on receipt and again when the buffer is * freed, causing a window update. DOGN. */ /* 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 = tcpst->rcv_hiwat - (tp->rcv_adv - tp->rcv_nxt); if (len > tp->max_rcvd) tp->max_rcvd = len; } else { tiflags &= ~TH_FIN; } /* * If FIN is received ACK the FIN and let the user know * that the connection is closing. */ if (tiflags & TH_FIN) { xTrace1(tcpp, 4, "vtcp_input: got fin on %lx", (u_long)tp); if (TCPS_HAVERCVDFIN(tp->t_state) == 0) { invoke_cantrcvmore = TRUE; tp->t_flags |= TF_ACKNOW; tp->rcv_nxt++; } switch (tp->t_state) { /* * In SYN_RECEIVED and ESTABLISHED STATES * enter the CLOSE_WAIT state. */ case TCPS_SYN_RECEIVED: case TCPS_ESTABLISHED: tp->t_state = TCPS_CLOSE_WAIT; break; /* * If still in FIN_WAIT_1 STATE FIN has not been acked so * enter the CLOSING state. */ case TCPS_FIN_WAIT_1: tp->t_state = TCPS_CLOSING; break; /* * In FIN_WAIT_2 state enter the TIME_WAIT state, * starting the time-wait timer, turning off the other * standard timers. */ case TCPS_FIN_WAIT_2: tp->t_state = TCPS_TIME_WAIT; vtcp_canceltimers(tp); tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL; soisdisconnected(so); break; /* * In TIME_WAIT state restart the 2 MSL time_wait timer. */ case TCPS_TIME_WAIT: tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL; break; } } xIfTrace(tcpp, 2) tcp_trace(TA_INPUT, ostate, tp, &tcp_saveti, 0); /* * Return any desired output. */ if (needoutput || (tp->t_flags & TF_ACKNOW)) { (void) vtcp_output(tp); } if (invoke_cantrcvmore) { socantrcvmore(so); } /* if */ if (!demuxmsgDestroyed) msgDestroy(&demuxmsg); return XK_SUCCESS; dropafterack: xTrace1(tcpp, 4, "vtcp_input: drop after ack %lx", (u_long)tp); DO_TRACE(ps, TCP_EVENT_DROPAA, tiflags, 0, 0); /* * Generate an ACK dropping incoming segment if it occupies * sequence space, where the ACK reflects our state. */ if (tiflags & TH_RST) goto drop; tp->t_flags |= TF_ACKNOW; (void) vtcp_output(tp); msgDestroy(&demuxmsg); return XK_SUCCESS; dropwithreset: xTrace1(tcpp, 4, "vtcp_input: drop with reset on %lx", (u_long)tp); DO_TRACE(ps, TCP_EVENT_DROPWR, tiflags, 0, 0); if (options) { (void) xFree(options); options = 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) || in_broadcast(*(struct in_addr *)&pHdr->dst)) goto drop; { IPpseudoHdr tmpPhdr; tmpPhdr.src = pHdr->dst; tmpPhdr.dst = pHdr->src; tmpPhdr.zero = 0; tmpPhdr.prot = pHdr->prot; if (tiflags & TH_ACK) tcp_respond(self, tp, &tHdr, &tmpPhdr, tHdr.th_ack, (tcp_seq)0, TH_RST); else { if (tiflags & TH_SYN) tlen++; tcp_respond(self, tp, &tHdr, &tmpPhdr, tHdr.th_seq + tlen, (tcp_seq)0, TH_RST|TH_ACK); } } /* destroy temporarily created socket */ if (dropsocket) soabort(so); msgDestroy(&demuxmsg); return XK_SUCCESS; drop: xTrace1(tcpp, 4, "vtcp_input: drop on %lx", (u_long)tp); DO_TRACE(ps, TCP_EVENT_DROP, tiflags, 0, 0); if (options) { xFree(options); options = 0; } /* * Drop space held by incoming segment and return. */ xIfTrace(tcpp, 2) tcp_trace(TA_DROP, ostate, tp, &tcp_saveti, 0); /* destroy temporarily created socket */ if (dropsocket) soabort(so); msgDestroy(&demuxmsg); return XK_SUCCESS; } static void tcp_dooptions(ps, tp, options, option_len, tHdr) PSTATE *ps; struct tcpcb *tp; char *options; int option_len; struct tcphdr *tHdr; { register u_char *cp; int opt, optlen, cnt; cp = (u_char *)options; cnt = option_len; for (; cnt > 0; cnt -= optlen, cp += optlen) { opt = cp[0]; if (opt == TCPOPT_EOL) break; if (opt == TCPOPT_NOP) optlen = 1; else { optlen = cp[1]; if (optlen <= 0) break; } switch (opt) { default: continue; case TCPOPT_MAXSEG: if (optlen != 4) continue; if (!(tHdr->th_flags & TH_SYN)) continue; tp->t_maxseg = *(u_short *)(cp + 2); tp->t_maxseg = ntohs((u_short)tp->t_maxseg); DO_TRACE(ps, TCP_EVENT_OTHER9, tp->t_maxseg, 0, 0); tp->t_maxseg = MIN(tp->t_maxseg, tcp_mss(tp)); break; case TCPOPT_WINDOW: if (optlen != TCPOLEN_WINDOW) continue; if (!(tHdr->th_flags & TH_SYN)) continue; tp->t_flags |= TF_RCVD_SCALE; tp->requested_s_scale = MIN(cp[2], TCP_MAX_WINSHIFT); DO_TRACE(ps, TCP_EVENT_OTHER9, tp->requested_s_scale, 2, 0); break; } } xFree(options); } /* * Pull out of band byte out of a segment so * it doesn't appear in the user's data queue. * It is still reflected in the segment length for * sequencing purposes. */ static void tcp_pulloutofband(so, th, m) Sessn so; struct tcphdr *th; Msg *m; { Msg firstPart; struct tcpcb *tp = sototcpcb(so); int cnt = th->th_urp - 1; int dataLen; MsgWalk mi; u_char *data; tp->t_oobflags |= TCPOOB_HAVEDATA; msgConstructEmpty(&firstPart); msgBreak(m, &firstPart, cnt); msgWalkInit(&mi, m); if ((data = msgWalkNext(&mi, &dataLen)) != 0) { xAssert(dataLen >= 1); tp->t_iobc = *(char *)data; } msgWalkDone(&mi); msgDiscard(m, 1); msgJoin(m, &firstPart, m); msgDestroy(&firstPart); } /* #ifdef XNETSIM extern int TcpPacket; #endif */ /* * 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. */ int tcp_mss(tp) register struct tcpcb *tp; { Sessn tcpSessn; int mss; tcpSessn = tcpcbtoso(tp); if (xControlSessn(xGetSessnDown(tcpSessn, 0), GETOPTPACKET, (char *)&mss, sizeof(int)) < sizeof(int)) { xTrace0(tcpp, 3, "tcp_mss: GETOPTPACKET control of lls failed"); mss = 512; } /* #ifdef XNETSIM if (TcpPacket > 0) mss = TcpPacket + sizeof(struct tcphdr); #endif */ xTrace1(tcpp, 3, "tcp_mss: GETOPTPACKET control of lls returned %d", mss); mss -= sizeof(struct tcphdr); tp->snd_cwnd = tp->t_slowstart * mss; return (mss); } void print_reass(tp, m) struct tcpcb *tp; char *m; { struct reass *s; printf("Printing reass\n"); printf("%s %x:", m, tp->rcv_nxt); for (s = tp->seg_next; s != (struct reass *)tp; s = s->next) { printf("[%x-%x) ", s->th.th_seq, msgLength(&s->m)); } printf("\n"); }