globals: Allow global properties to be optional
[qemu/kevin.git] / slirp / tcp_input.c
blob09bdf9b4828eace1cc254f6bab519bfa67b89a1a
1 /*
2 * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1994
3 * The Regents of the University of California. All rights reserved.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 3. Neither the name of the University nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27 * SUCH DAMAGE.
29 * @(#)tcp_input.c 8.5 (Berkeley) 4/10/94
30 * tcp_input.c,v 1.10 1994/10/13 18:36:32 wollman Exp
34 * Changes and additions relating to SLiRP
35 * Copyright (c) 1995 Danny Gasparovski.
37 * Please read the file COPYRIGHT for the
38 * terms and conditions of the copyright.
41 #include "qemu/osdep.h"
42 #include "slirp.h"
43 #include "ip_icmp.h"
45 #define TCPREXMTTHRESH 3
47 #define TCP_PAWS_IDLE (24 * 24 * 60 * 60 * PR_SLOWHZ)
49 /* for modulo comparisons of timestamps */
50 #define TSTMP_LT(a,b) ((int)((a)-(b)) < 0)
51 #define TSTMP_GEQ(a,b) ((int)((a)-(b)) >= 0)
54 * Insert segment ti into reassembly queue of tcp with
55 * control block tp. Return TH_FIN if reassembly now includes
56 * a segment with FIN. The macro form does the common case inline
57 * (segment is the next to be received on an established connection,
58 * and the queue is empty), avoiding linkage into and removal
59 * from the queue and repetition of various conversions.
60 * Set DELACK for segments received in order, but ack immediately
61 * when segments are out of order (so fast retransmit can work).
63 #ifdef TCP_ACK_HACK
64 #define TCP_REASS(tp, ti, m, so, flags) {\
65 if ((ti)->ti_seq == (tp)->rcv_nxt && \
66 tcpfrag_list_empty(tp) && \
67 (tp)->t_state == TCPS_ESTABLISHED) {\
68 if (ti->ti_flags & TH_PUSH) \
69 tp->t_flags |= TF_ACKNOW; \
70 else \
71 tp->t_flags |= TF_DELACK; \
72 (tp)->rcv_nxt += (ti)->ti_len; \
73 flags = (ti)->ti_flags & TH_FIN; \
74 if (so->so_emu) { \
75 if (tcp_emu((so),(m))) sbappend((so), (m)); \
76 } else \
77 sbappend((so), (m)); \
78 } else {\
79 (flags) = tcp_reass((tp), (ti), (m)); \
80 tp->t_flags |= TF_ACKNOW; \
81 } \
83 #else
84 #define TCP_REASS(tp, ti, m, so, flags) { \
85 if ((ti)->ti_seq == (tp)->rcv_nxt && \
86 tcpfrag_list_empty(tp) && \
87 (tp)->t_state == TCPS_ESTABLISHED) { \
88 tp->t_flags |= TF_DELACK; \
89 (tp)->rcv_nxt += (ti)->ti_len; \
90 flags = (ti)->ti_flags & TH_FIN; \
91 if (so->so_emu) { \
92 if (tcp_emu((so),(m))) sbappend(so, (m)); \
93 } else \
94 sbappend((so), (m)); \
95 } else { \
96 (flags) = tcp_reass((tp), (ti), (m)); \
97 tp->t_flags |= TF_ACKNOW; \
98 } \
100 #endif
101 static void tcp_dooptions(struct tcpcb *tp, u_char *cp, int cnt,
102 struct tcpiphdr *ti);
103 static void tcp_xmit_timer(register struct tcpcb *tp, int rtt);
105 static int
106 tcp_reass(register struct tcpcb *tp, register struct tcpiphdr *ti,
107 struct mbuf *m)
109 register struct tcpiphdr *q;
110 struct socket *so = tp->t_socket;
111 int flags;
114 * Call with ti==NULL after become established to
115 * force pre-ESTABLISHED data up to user socket.
117 if (ti == NULL)
118 goto present;
121 * Find a segment which begins after this one does.
123 for (q = tcpfrag_list_first(tp); !tcpfrag_list_end(q, tp);
124 q = tcpiphdr_next(q))
125 if (SEQ_GT(q->ti_seq, ti->ti_seq))
126 break;
129 * If there is a preceding segment, it may provide some of
130 * our data already. If so, drop the data from the incoming
131 * segment. If it provides all of our data, drop us.
133 if (!tcpfrag_list_end(tcpiphdr_prev(q), tp)) {
134 register int i;
135 q = tcpiphdr_prev(q);
136 /* conversion to int (in i) handles seq wraparound */
137 i = q->ti_seq + q->ti_len - ti->ti_seq;
138 if (i > 0) {
139 if (i >= ti->ti_len) {
140 m_free(m);
142 * Try to present any queued data
143 * at the left window edge to the user.
144 * This is needed after the 3-WHS
145 * completes.
147 goto present; /* ??? */
149 m_adj(m, i);
150 ti->ti_len -= i;
151 ti->ti_seq += i;
153 q = tcpiphdr_next(q);
155 ti->ti_mbuf = m;
158 * While we overlap succeeding segments trim them or,
159 * if they are completely covered, dequeue them.
161 while (!tcpfrag_list_end(q, tp)) {
162 register int i = (ti->ti_seq + ti->ti_len) - q->ti_seq;
163 if (i <= 0)
164 break;
165 if (i < q->ti_len) {
166 q->ti_seq += i;
167 q->ti_len -= i;
168 m_adj(q->ti_mbuf, i);
169 break;
171 q = tcpiphdr_next(q);
172 m = tcpiphdr_prev(q)->ti_mbuf;
173 remque(tcpiphdr2qlink(tcpiphdr_prev(q)));
174 m_free(m);
178 * Stick new segment in its place.
180 insque(tcpiphdr2qlink(ti), tcpiphdr2qlink(tcpiphdr_prev(q)));
182 present:
184 * Present data to user, advancing rcv_nxt through
185 * completed sequence space.
187 if (!TCPS_HAVEESTABLISHED(tp->t_state))
188 return (0);
189 ti = tcpfrag_list_first(tp);
190 if (tcpfrag_list_end(ti, tp) || ti->ti_seq != tp->rcv_nxt)
191 return (0);
192 if (tp->t_state == TCPS_SYN_RECEIVED && ti->ti_len)
193 return (0);
194 do {
195 tp->rcv_nxt += ti->ti_len;
196 flags = ti->ti_flags & TH_FIN;
197 remque(tcpiphdr2qlink(ti));
198 m = ti->ti_mbuf;
199 ti = tcpiphdr_next(ti);
200 if (so->so_state & SS_FCANTSENDMORE)
201 m_free(m);
202 else {
203 if (so->so_emu) {
204 if (tcp_emu(so,m)) sbappend(so, m);
205 } else
206 sbappend(so, m);
208 } while (ti != (struct tcpiphdr *)tp && ti->ti_seq == tp->rcv_nxt);
209 return (flags);
213 * TCP input routine, follows pages 65-76 of the
214 * protocol specification dated September, 1981 very closely.
216 void
217 tcp_input(struct mbuf *m, int iphlen, struct socket *inso, unsigned short af)
219 struct ip save_ip, *ip;
220 struct ip6 save_ip6, *ip6;
221 register struct tcpiphdr *ti;
222 caddr_t optp = NULL;
223 int optlen = 0;
224 int len, tlen, off;
225 register struct tcpcb *tp = NULL;
226 register int tiflags;
227 struct socket *so = NULL;
228 int todrop, acked, ourfinisacked, needoutput = 0;
229 int iss = 0;
230 u_long tiwin;
231 int ret;
232 struct sockaddr_storage lhost, fhost;
233 struct sockaddr_in *lhost4, *fhost4;
234 struct sockaddr_in6 *lhost6, *fhost6;
235 struct ex_list *ex_ptr;
236 Slirp *slirp;
238 DEBUG_CALL("tcp_input");
239 DEBUG_ARGS((dfd, " m = %p iphlen = %2d inso = %p\n",
240 m, iphlen, inso));
243 * If called with m == 0, then we're continuing the connect
245 if (m == NULL) {
246 so = inso;
247 slirp = so->slirp;
249 /* Re-set a few variables */
250 tp = sototcpcb(so);
251 m = so->so_m;
252 so->so_m = NULL;
253 ti = so->so_ti;
254 tiwin = ti->ti_win;
255 tiflags = ti->ti_flags;
257 goto cont_conn;
259 slirp = m->slirp;
261 ip = mtod(m, struct ip *);
262 ip6 = mtod(m, struct ip6 *);
264 switch (af) {
265 case AF_INET:
266 if (iphlen > sizeof(struct ip)) {
267 ip_stripoptions(m, (struct mbuf *)0);
268 iphlen = sizeof(struct ip);
270 /* XXX Check if too short */
274 * Save a copy of the IP header in case we want restore it
275 * for sending an ICMP error message in response.
277 save_ip = *ip;
278 save_ip.ip_len += iphlen;
281 * Get IP and TCP header together in first mbuf.
282 * Note: IP leaves IP header in first mbuf.
284 m->m_data -= sizeof(struct tcpiphdr) - sizeof(struct ip)
285 - sizeof(struct tcphdr);
286 m->m_len += sizeof(struct tcpiphdr) - sizeof(struct ip)
287 - sizeof(struct tcphdr);
288 ti = mtod(m, struct tcpiphdr *);
291 * Checksum extended TCP header and data.
293 tlen = ip->ip_len;
294 tcpiphdr2qlink(ti)->next = tcpiphdr2qlink(ti)->prev = NULL;
295 memset(&ti->ih_mbuf, 0 , sizeof(struct mbuf_ptr));
296 memset(&ti->ti, 0, sizeof(ti->ti));
297 ti->ti_x0 = 0;
298 ti->ti_src = save_ip.ip_src;
299 ti->ti_dst = save_ip.ip_dst;
300 ti->ti_pr = save_ip.ip_p;
301 ti->ti_len = htons((uint16_t)tlen);
302 break;
304 case AF_INET6:
306 * Save a copy of the IP header in case we want restore it
307 * for sending an ICMP error message in response.
309 save_ip6 = *ip6;
311 * Get IP and TCP header together in first mbuf.
312 * Note: IP leaves IP header in first mbuf.
314 m->m_data -= sizeof(struct tcpiphdr) - (sizeof(struct ip6)
315 + sizeof(struct tcphdr));
316 m->m_len += sizeof(struct tcpiphdr) - (sizeof(struct ip6)
317 + sizeof(struct tcphdr));
318 ti = mtod(m, struct tcpiphdr *);
320 tlen = ip6->ip_pl;
321 tcpiphdr2qlink(ti)->next = tcpiphdr2qlink(ti)->prev = NULL;
322 memset(&ti->ih_mbuf, 0 , sizeof(struct mbuf_ptr));
323 memset(&ti->ti, 0, sizeof(ti->ti));
324 ti->ti_x0 = 0;
325 ti->ti_src6 = save_ip6.ip_src;
326 ti->ti_dst6 = save_ip6.ip_dst;
327 ti->ti_nh6 = save_ip6.ip_nh;
328 ti->ti_len = htons((uint16_t)tlen);
329 break;
331 default:
332 g_assert_not_reached();
335 len = ((sizeof(struct tcpiphdr) - sizeof(struct tcphdr)) + tlen);
336 if (cksum(m, len)) {
337 goto drop;
341 * Check that TCP offset makes sense,
342 * pull out TCP options and adjust length. XXX
344 off = ti->ti_off << 2;
345 if (off < sizeof (struct tcphdr) || off > tlen) {
346 goto drop;
348 tlen -= off;
349 ti->ti_len = tlen;
350 if (off > sizeof (struct tcphdr)) {
351 optlen = off - sizeof (struct tcphdr);
352 optp = mtod(m, caddr_t) + sizeof (struct tcpiphdr);
354 tiflags = ti->ti_flags;
357 * Convert TCP protocol specific fields to host format.
359 NTOHL(ti->ti_seq);
360 NTOHL(ti->ti_ack);
361 NTOHS(ti->ti_win);
362 NTOHS(ti->ti_urp);
365 * Drop TCP, IP headers and TCP options.
367 m->m_data += sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
368 m->m_len -= sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
371 * Locate pcb for segment.
373 findso:
374 lhost.ss_family = af;
375 fhost.ss_family = af;
376 switch (af) {
377 case AF_INET:
378 lhost4 = (struct sockaddr_in *) &lhost;
379 lhost4->sin_addr = ti->ti_src;
380 lhost4->sin_port = ti->ti_sport;
381 fhost4 = (struct sockaddr_in *) &fhost;
382 fhost4->sin_addr = ti->ti_dst;
383 fhost4->sin_port = ti->ti_dport;
384 break;
385 case AF_INET6:
386 lhost6 = (struct sockaddr_in6 *) &lhost;
387 lhost6->sin6_addr = ti->ti_src6;
388 lhost6->sin6_port = ti->ti_sport;
389 fhost6 = (struct sockaddr_in6 *) &fhost;
390 fhost6->sin6_addr = ti->ti_dst6;
391 fhost6->sin6_port = ti->ti_dport;
392 break;
393 default:
394 g_assert_not_reached();
397 so = solookup(&slirp->tcp_last_so, &slirp->tcb, &lhost, &fhost);
400 * If the state is CLOSED (i.e., TCB does not exist) then
401 * all data in the incoming segment is discarded.
402 * If the TCB exists but is in CLOSED state, it is embryonic,
403 * but should either do a listen or a connect soon.
405 * state == CLOSED means we've done socreate() but haven't
406 * attached it to a protocol yet...
408 * XXX If a TCB does not exist, and the TH_SYN flag is
409 * the only flag set, then create a session, mark it
410 * as if it was LISTENING, and continue...
412 if (so == NULL) {
413 if (slirp->restricted) {
414 /* Any hostfwds will have an existing socket, so we only get here
415 * for non-hostfwd connections. These should be dropped, unless it
416 * happens to be a guestfwd.
418 for (ex_ptr = slirp->exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next) {
419 if (ex_ptr->ex_fport == ti->ti_dport &&
420 ti->ti_dst.s_addr == ex_ptr->ex_addr.s_addr) {
421 break;
424 if (!ex_ptr) {
425 goto dropwithreset;
429 if ((tiflags & (TH_SYN|TH_FIN|TH_RST|TH_URG|TH_ACK)) != TH_SYN)
430 goto dropwithreset;
432 so = socreate(slirp);
433 if (tcp_attach(so) < 0) {
434 g_free(so); /* Not sofree (if it failed, it's not insqued) */
435 goto dropwithreset;
438 sbreserve(&so->so_snd, TCP_SNDSPACE);
439 sbreserve(&so->so_rcv, TCP_RCVSPACE);
441 so->lhost.ss = lhost;
442 so->fhost.ss = fhost;
444 so->so_iptos = tcp_tos(so);
445 if (so->so_iptos == 0) {
446 switch (af) {
447 case AF_INET:
448 so->so_iptos = ((struct ip *)ti)->ip_tos;
449 break;
450 case AF_INET6:
451 break;
452 default:
453 g_assert_not_reached();
457 tp = sototcpcb(so);
458 tp->t_state = TCPS_LISTEN;
462 * If this is a still-connecting socket, this probably
463 * a retransmit of the SYN. Whether it's a retransmit SYN
464 * or something else, we nuke it.
466 if (so->so_state & SS_ISFCONNECTING)
467 goto drop;
469 tp = sototcpcb(so);
471 /* XXX Should never fail */
472 if (tp == NULL)
473 goto dropwithreset;
474 if (tp->t_state == TCPS_CLOSED)
475 goto drop;
477 tiwin = ti->ti_win;
480 * Segment received on connection.
481 * Reset idle time and keep-alive timer.
483 tp->t_idle = 0;
484 if (SO_OPTIONS)
485 tp->t_timer[TCPT_KEEP] = TCPTV_KEEPINTVL;
486 else
487 tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_IDLE;
490 * Process options if not in LISTEN state,
491 * else do it below (after getting remote address).
493 if (optp && tp->t_state != TCPS_LISTEN)
494 tcp_dooptions(tp, (u_char *)optp, optlen, ti);
497 * Header prediction: check for the two common cases
498 * of a uni-directional data xfer. If the packet has
499 * no control flags, is in-sequence, the window didn't
500 * change and we're not retransmitting, it's a
501 * candidate. If the length is zero and the ack moved
502 * forward, we're the sender side of the xfer. Just
503 * free the data acked & wake any higher level process
504 * that was blocked waiting for space. If the length
505 * is non-zero and the ack didn't move, we're the
506 * receiver side. If we're getting packets in-order
507 * (the reassembly queue is empty), add the data to
508 * the socket buffer and note that we need a delayed ack.
510 * XXX Some of these tests are not needed
511 * eg: the tiwin == tp->snd_wnd prevents many more
512 * predictions.. with no *real* advantage..
514 if (tp->t_state == TCPS_ESTABLISHED &&
515 (tiflags & (TH_SYN|TH_FIN|TH_RST|TH_URG|TH_ACK)) == TH_ACK &&
516 ti->ti_seq == tp->rcv_nxt &&
517 tiwin && tiwin == tp->snd_wnd &&
518 tp->snd_nxt == tp->snd_max) {
519 if (ti->ti_len == 0) {
520 if (SEQ_GT(ti->ti_ack, tp->snd_una) &&
521 SEQ_LEQ(ti->ti_ack, tp->snd_max) &&
522 tp->snd_cwnd >= tp->snd_wnd) {
524 * this is a pure ack for outstanding data.
526 if (tp->t_rtt &&
527 SEQ_GT(ti->ti_ack, tp->t_rtseq))
528 tcp_xmit_timer(tp, tp->t_rtt);
529 acked = ti->ti_ack - tp->snd_una;
530 sbdrop(&so->so_snd, acked);
531 tp->snd_una = ti->ti_ack;
532 m_free(m);
535 * If all outstanding data are acked, stop
536 * retransmit timer, otherwise restart timer
537 * using current (possibly backed-off) value.
538 * If process is waiting for space,
539 * wakeup/selwakeup/signal. If data
540 * are ready to send, let tcp_output
541 * decide between more output or persist.
543 if (tp->snd_una == tp->snd_max)
544 tp->t_timer[TCPT_REXMT] = 0;
545 else if (tp->t_timer[TCPT_PERSIST] == 0)
546 tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
549 * This is called because sowwakeup might have
550 * put data into so_snd. Since we don't so sowwakeup,
551 * we don't need this.. XXX???
553 if (so->so_snd.sb_cc)
554 (void) tcp_output(tp);
556 return;
558 } else if (ti->ti_ack == tp->snd_una &&
559 tcpfrag_list_empty(tp) &&
560 ti->ti_len <= sbspace(&so->so_rcv)) {
562 * this is a pure, in-sequence data packet
563 * with nothing on the reassembly queue and
564 * we have enough buffer space to take it.
566 tp->rcv_nxt += ti->ti_len;
568 * Add data to socket buffer.
570 if (so->so_emu) {
571 if (tcp_emu(so,m)) sbappend(so, m);
572 } else
573 sbappend(so, m);
576 * If this is a short packet, then ACK now - with Nagel
577 * congestion avoidance sender won't send more until
578 * he gets an ACK.
580 * It is better to not delay acks at all to maximize
581 * TCP throughput. See RFC 2581.
583 tp->t_flags |= TF_ACKNOW;
584 tcp_output(tp);
585 return;
587 } /* header prediction */
589 * Calculate amount of space in receive window,
590 * and then do TCP input processing.
591 * Receive window is amount of space in rcv queue,
592 * but not less than advertised window.
594 { int win;
595 win = sbspace(&so->so_rcv);
596 if (win < 0)
597 win = 0;
598 tp->rcv_wnd = MAX(win, (int)(tp->rcv_adv - tp->rcv_nxt));
601 switch (tp->t_state) {
604 * If the state is LISTEN then ignore segment if it contains an RST.
605 * If the segment contains an ACK then it is bad and send a RST.
606 * If it does not contain a SYN then it is not interesting; drop it.
607 * Don't bother responding if the destination was a broadcast.
608 * Otherwise initialize tp->rcv_nxt, and tp->irs, select an initial
609 * tp->iss, and send a segment:
610 * <SEQ=ISS><ACK=RCV_NXT><CTL=SYN,ACK>
611 * Also initialize tp->snd_nxt to tp->iss+1 and tp->snd_una to tp->iss.
612 * Fill in remote peer address fields if not previously specified.
613 * Enter SYN_RECEIVED state, and process any other fields of this
614 * segment in this state.
616 case TCPS_LISTEN: {
618 if (tiflags & TH_RST)
619 goto drop;
620 if (tiflags & TH_ACK)
621 goto dropwithreset;
622 if ((tiflags & TH_SYN) == 0)
623 goto drop;
626 * This has way too many gotos...
627 * But a bit of spaghetti code never hurt anybody :)
631 * If this is destined for the control address, then flag to
632 * tcp_ctl once connected, otherwise connect
634 if (af == AF_INET &&
635 (so->so_faddr.s_addr & slirp->vnetwork_mask.s_addr) ==
636 slirp->vnetwork_addr.s_addr) {
637 if (so->so_faddr.s_addr != slirp->vhost_addr.s_addr &&
638 so->so_faddr.s_addr != slirp->vnameserver_addr.s_addr) {
639 /* May be an add exec */
640 for (ex_ptr = slirp->exec_list; ex_ptr;
641 ex_ptr = ex_ptr->ex_next) {
642 if(ex_ptr->ex_fport == so->so_fport &&
643 so->so_faddr.s_addr == ex_ptr->ex_addr.s_addr) {
644 so->so_state |= SS_CTL;
645 break;
648 if (so->so_state & SS_CTL) {
649 goto cont_input;
652 /* CTL_ALIAS: Do nothing, tcp_fconnect will be called on it */
655 if (so->so_emu & EMU_NOCONNECT) {
656 so->so_emu &= ~EMU_NOCONNECT;
657 goto cont_input;
660 if ((tcp_fconnect(so, so->so_ffamily) == -1) &&
661 (errno != EAGAIN) &&
662 (errno != EINPROGRESS) && (errno != EWOULDBLOCK)
664 uint8_t code;
665 DEBUG_MISC((dfd, " tcp fconnect errno = %d-%s\n",
666 errno,strerror(errno)));
667 if(errno == ECONNREFUSED) {
668 /* ACK the SYN, send RST to refuse the connection */
669 tcp_respond(tp, ti, m, ti->ti_seq + 1, (tcp_seq) 0,
670 TH_RST | TH_ACK, af);
671 } else {
672 switch (af) {
673 case AF_INET:
674 code = ICMP_UNREACH_NET;
675 if (errno == EHOSTUNREACH) {
676 code = ICMP_UNREACH_HOST;
678 break;
679 case AF_INET6:
680 code = ICMP6_UNREACH_NO_ROUTE;
681 if (errno == EHOSTUNREACH) {
682 code = ICMP6_UNREACH_ADDRESS;
684 break;
685 default:
686 g_assert_not_reached();
688 HTONL(ti->ti_seq); /* restore tcp header */
689 HTONL(ti->ti_ack);
690 HTONS(ti->ti_win);
691 HTONS(ti->ti_urp);
692 m->m_data -= sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
693 m->m_len += sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
694 switch (af) {
695 case AF_INET:
696 m->m_data += sizeof(struct tcpiphdr) - sizeof(struct ip)
697 - sizeof(struct tcphdr);
698 m->m_len -= sizeof(struct tcpiphdr) - sizeof(struct ip)
699 - sizeof(struct tcphdr);
700 *ip = save_ip;
701 icmp_send_error(m, ICMP_UNREACH, code, 0, strerror(errno));
702 break;
703 case AF_INET6:
704 m->m_data += sizeof(struct tcpiphdr) - (sizeof(struct ip6)
705 + sizeof(struct tcphdr));
706 m->m_len -= sizeof(struct tcpiphdr) - (sizeof(struct ip6)
707 + sizeof(struct tcphdr));
708 *ip6 = save_ip6;
709 icmp6_send_error(m, ICMP6_UNREACH, code);
710 break;
711 default:
712 g_assert_not_reached();
715 tcp_close(tp);
716 m_free(m);
717 } else {
719 * Haven't connected yet, save the current mbuf
720 * and ti, and return
721 * XXX Some OS's don't tell us whether the connect()
722 * succeeded or not. So we must time it out.
724 so->so_m = m;
725 so->so_ti = ti;
726 tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT;
727 tp->t_state = TCPS_SYN_RECEIVED;
729 * Initialize receive sequence numbers now so that we can send a
730 * valid RST if the remote end rejects our connection.
732 tp->irs = ti->ti_seq;
733 tcp_rcvseqinit(tp);
734 tcp_template(tp);
736 return;
738 cont_conn:
739 /* m==NULL
740 * Check if the connect succeeded
742 if (so->so_state & SS_NOFDREF) {
743 tp = tcp_close(tp);
744 goto dropwithreset;
746 cont_input:
747 tcp_template(tp);
749 if (optp)
750 tcp_dooptions(tp, (u_char *)optp, optlen, ti);
752 if (iss)
753 tp->iss = iss;
754 else
755 tp->iss = slirp->tcp_iss;
756 slirp->tcp_iss += TCP_ISSINCR/2;
757 tp->irs = ti->ti_seq;
758 tcp_sendseqinit(tp);
759 tcp_rcvseqinit(tp);
760 tp->t_flags |= TF_ACKNOW;
761 tp->t_state = TCPS_SYN_RECEIVED;
762 tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT;
763 goto trimthenstep6;
764 } /* case TCPS_LISTEN */
767 * If the state is SYN_SENT:
768 * if seg contains an ACK, but not for our SYN, drop the input.
769 * if seg contains a RST, then drop the connection.
770 * if seg does not contain SYN, then drop it.
771 * Otherwise this is an acceptable SYN segment
772 * initialize tp->rcv_nxt and tp->irs
773 * if seg contains ack then advance tp->snd_una
774 * if SYN has been acked change to ESTABLISHED else SYN_RCVD state
775 * arrange for segment to be acked (eventually)
776 * continue processing rest of data/controls, beginning with URG
778 case TCPS_SYN_SENT:
779 if ((tiflags & TH_ACK) &&
780 (SEQ_LEQ(ti->ti_ack, tp->iss) ||
781 SEQ_GT(ti->ti_ack, tp->snd_max)))
782 goto dropwithreset;
784 if (tiflags & TH_RST) {
785 if (tiflags & TH_ACK) {
786 tcp_drop(tp, 0); /* XXX Check t_softerror! */
788 goto drop;
791 if ((tiflags & TH_SYN) == 0)
792 goto drop;
793 if (tiflags & TH_ACK) {
794 tp->snd_una = ti->ti_ack;
795 if (SEQ_LT(tp->snd_nxt, tp->snd_una))
796 tp->snd_nxt = tp->snd_una;
799 tp->t_timer[TCPT_REXMT] = 0;
800 tp->irs = ti->ti_seq;
801 tcp_rcvseqinit(tp);
802 tp->t_flags |= TF_ACKNOW;
803 if (tiflags & TH_ACK && SEQ_GT(tp->snd_una, tp->iss)) {
804 soisfconnected(so);
805 tp->t_state = TCPS_ESTABLISHED;
807 (void) tcp_reass(tp, (struct tcpiphdr *)0,
808 (struct mbuf *)0);
810 * if we didn't have to retransmit the SYN,
811 * use its rtt as our initial srtt & rtt var.
813 if (tp->t_rtt)
814 tcp_xmit_timer(tp, tp->t_rtt);
815 } else
816 tp->t_state = TCPS_SYN_RECEIVED;
818 trimthenstep6:
820 * Advance ti->ti_seq to correspond to first data byte.
821 * If data, trim to stay within window,
822 * dropping FIN if necessary.
824 ti->ti_seq++;
825 if (ti->ti_len > tp->rcv_wnd) {
826 todrop = ti->ti_len - tp->rcv_wnd;
827 m_adj(m, -todrop);
828 ti->ti_len = tp->rcv_wnd;
829 tiflags &= ~TH_FIN;
831 tp->snd_wl1 = ti->ti_seq - 1;
832 tp->rcv_up = ti->ti_seq;
833 goto step6;
834 } /* switch tp->t_state */
836 * States other than LISTEN or SYN_SENT.
837 * Check that at least some bytes of segment are within
838 * receive window. If segment begins before rcv_nxt,
839 * drop leading data (and SYN); if nothing left, just ack.
841 todrop = tp->rcv_nxt - ti->ti_seq;
842 if (todrop > 0) {
843 if (tiflags & TH_SYN) {
844 tiflags &= ~TH_SYN;
845 ti->ti_seq++;
846 if (ti->ti_urp > 1)
847 ti->ti_urp--;
848 else
849 tiflags &= ~TH_URG;
850 todrop--;
853 * Following if statement from Stevens, vol. 2, p. 960.
855 if (todrop > ti->ti_len
856 || (todrop == ti->ti_len && (tiflags & TH_FIN) == 0)) {
858 * Any valid FIN must be to the left of the window.
859 * At this point the FIN must be a duplicate or out
860 * of sequence; drop it.
862 tiflags &= ~TH_FIN;
865 * Send an ACK to resynchronize and drop any data.
866 * But keep on processing for RST or ACK.
868 tp->t_flags |= TF_ACKNOW;
869 todrop = ti->ti_len;
871 m_adj(m, todrop);
872 ti->ti_seq += todrop;
873 ti->ti_len -= todrop;
874 if (ti->ti_urp > todrop)
875 ti->ti_urp -= todrop;
876 else {
877 tiflags &= ~TH_URG;
878 ti->ti_urp = 0;
882 * If new data are received on a connection after the
883 * user processes are gone, then RST the other end.
885 if ((so->so_state & SS_NOFDREF) &&
886 tp->t_state > TCPS_CLOSE_WAIT && ti->ti_len) {
887 tp = tcp_close(tp);
888 goto dropwithreset;
892 * If segment ends after window, drop trailing data
893 * (and PUSH and FIN); if nothing left, just ACK.
895 todrop = (ti->ti_seq+ti->ti_len) - (tp->rcv_nxt+tp->rcv_wnd);
896 if (todrop > 0) {
897 if (todrop >= ti->ti_len) {
899 * If a new connection request is received
900 * while in TIME_WAIT, drop the old connection
901 * and start over if the sequence numbers
902 * are above the previous ones.
904 if (tiflags & TH_SYN &&
905 tp->t_state == TCPS_TIME_WAIT &&
906 SEQ_GT(ti->ti_seq, tp->rcv_nxt)) {
907 iss = tp->rcv_nxt + TCP_ISSINCR;
908 tp = tcp_close(tp);
909 goto findso;
912 * If window is closed can only take segments at
913 * window edge, and have to drop data and PUSH from
914 * incoming segments. Continue processing, but
915 * remember to ack. Otherwise, drop segment
916 * and ack.
918 if (tp->rcv_wnd == 0 && ti->ti_seq == tp->rcv_nxt) {
919 tp->t_flags |= TF_ACKNOW;
920 } else {
921 goto dropafterack;
924 m_adj(m, -todrop);
925 ti->ti_len -= todrop;
926 tiflags &= ~(TH_PUSH|TH_FIN);
930 * If the RST bit is set examine the state:
931 * SYN_RECEIVED STATE:
932 * If passive open, return to LISTEN state.
933 * If active open, inform user that connection was refused.
934 * ESTABLISHED, FIN_WAIT_1, FIN_WAIT2, CLOSE_WAIT STATES:
935 * Inform user that connection was reset, and close tcb.
936 * CLOSING, LAST_ACK, TIME_WAIT STATES
937 * Close the tcb.
939 if (tiflags&TH_RST) switch (tp->t_state) {
941 case TCPS_SYN_RECEIVED:
942 case TCPS_ESTABLISHED:
943 case TCPS_FIN_WAIT_1:
944 case TCPS_FIN_WAIT_2:
945 case TCPS_CLOSE_WAIT:
946 tp->t_state = TCPS_CLOSED;
947 tcp_close(tp);
948 goto drop;
950 case TCPS_CLOSING:
951 case TCPS_LAST_ACK:
952 case TCPS_TIME_WAIT:
953 tcp_close(tp);
954 goto drop;
958 * If a SYN is in the window, then this is an
959 * error and we send an RST and drop the connection.
961 if (tiflags & TH_SYN) {
962 tp = tcp_drop(tp,0);
963 goto dropwithreset;
967 * If the ACK bit is off we drop the segment and return.
969 if ((tiflags & TH_ACK) == 0) goto drop;
972 * Ack processing.
974 switch (tp->t_state) {
976 * In SYN_RECEIVED state if the ack ACKs our SYN then enter
977 * ESTABLISHED state and continue processing, otherwise
978 * send an RST. una<=ack<=max
980 case TCPS_SYN_RECEIVED:
982 if (SEQ_GT(tp->snd_una, ti->ti_ack) ||
983 SEQ_GT(ti->ti_ack, tp->snd_max))
984 goto dropwithreset;
985 tp->t_state = TCPS_ESTABLISHED;
987 * The sent SYN is ack'ed with our sequence number +1
988 * The first data byte already in the buffer will get
989 * lost if no correction is made. This is only needed for
990 * SS_CTL since the buffer is empty otherwise.
991 * tp->snd_una++; or:
993 tp->snd_una=ti->ti_ack;
994 if (so->so_state & SS_CTL) {
995 /* So tcp_ctl reports the right state */
996 ret = tcp_ctl(so);
997 if (ret == 1) {
998 soisfconnected(so);
999 so->so_state &= ~SS_CTL; /* success XXX */
1000 } else if (ret == 2) {
1001 so->so_state &= SS_PERSISTENT_MASK;
1002 so->so_state |= SS_NOFDREF; /* CTL_CMD */
1003 } else {
1004 needoutput = 1;
1005 tp->t_state = TCPS_FIN_WAIT_1;
1007 } else {
1008 soisfconnected(so);
1011 (void) tcp_reass(tp, (struct tcpiphdr *)0, (struct mbuf *)0);
1012 tp->snd_wl1 = ti->ti_seq - 1;
1013 /* Avoid ack processing; snd_una==ti_ack => dup ack */
1014 goto synrx_to_est;
1015 /* fall into ... */
1018 * In ESTABLISHED state: drop duplicate ACKs; ACK out of range
1019 * ACKs. If the ack is in the range
1020 * tp->snd_una < ti->ti_ack <= tp->snd_max
1021 * then advance tp->snd_una to ti->ti_ack and drop
1022 * data from the retransmission queue. If this ACK reflects
1023 * more up to date window information we update our window information.
1025 case TCPS_ESTABLISHED:
1026 case TCPS_FIN_WAIT_1:
1027 case TCPS_FIN_WAIT_2:
1028 case TCPS_CLOSE_WAIT:
1029 case TCPS_CLOSING:
1030 case TCPS_LAST_ACK:
1031 case TCPS_TIME_WAIT:
1033 if (SEQ_LEQ(ti->ti_ack, tp->snd_una)) {
1034 if (ti->ti_len == 0 && tiwin == tp->snd_wnd) {
1035 DEBUG_MISC((dfd, " dup ack m = %p so = %p\n",
1036 m, so));
1038 * If we have outstanding data (other than
1039 * a window probe), this is a completely
1040 * duplicate ack (ie, window info didn't
1041 * change), the ack is the biggest we've
1042 * seen and we've seen exactly our rexmt
1043 * threshold of them, assume a packet
1044 * has been dropped and retransmit it.
1045 * Kludge snd_nxt & the congestion
1046 * window so we send only this one
1047 * packet.
1049 * We know we're losing at the current
1050 * window size so do congestion avoidance
1051 * (set ssthresh to half the current window
1052 * and pull our congestion window back to
1053 * the new ssthresh).
1055 * Dup acks mean that packets have left the
1056 * network (they're now cached at the receiver)
1057 * so bump cwnd by the amount in the receiver
1058 * to keep a constant cwnd packets in the
1059 * network.
1061 if (tp->t_timer[TCPT_REXMT] == 0 ||
1062 ti->ti_ack != tp->snd_una)
1063 tp->t_dupacks = 0;
1064 else if (++tp->t_dupacks == TCPREXMTTHRESH) {
1065 tcp_seq onxt = tp->snd_nxt;
1066 u_int win =
1067 MIN(tp->snd_wnd, tp->snd_cwnd) /
1068 2 / tp->t_maxseg;
1070 if (win < 2)
1071 win = 2;
1072 tp->snd_ssthresh = win * tp->t_maxseg;
1073 tp->t_timer[TCPT_REXMT] = 0;
1074 tp->t_rtt = 0;
1075 tp->snd_nxt = ti->ti_ack;
1076 tp->snd_cwnd = tp->t_maxseg;
1077 (void) tcp_output(tp);
1078 tp->snd_cwnd = tp->snd_ssthresh +
1079 tp->t_maxseg * tp->t_dupacks;
1080 if (SEQ_GT(onxt, tp->snd_nxt))
1081 tp->snd_nxt = onxt;
1082 goto drop;
1083 } else if (tp->t_dupacks > TCPREXMTTHRESH) {
1084 tp->snd_cwnd += tp->t_maxseg;
1085 (void) tcp_output(tp);
1086 goto drop;
1088 } else
1089 tp->t_dupacks = 0;
1090 break;
1092 synrx_to_est:
1094 * If the congestion window was inflated to account
1095 * for the other side's cached packets, retract it.
1097 if (tp->t_dupacks > TCPREXMTTHRESH &&
1098 tp->snd_cwnd > tp->snd_ssthresh)
1099 tp->snd_cwnd = tp->snd_ssthresh;
1100 tp->t_dupacks = 0;
1101 if (SEQ_GT(ti->ti_ack, tp->snd_max)) {
1102 goto dropafterack;
1104 acked = ti->ti_ack - tp->snd_una;
1107 * If transmit timer is running and timed sequence
1108 * number was acked, update smoothed round trip time.
1109 * Since we now have an rtt measurement, cancel the
1110 * timer backoff (cf., Phil Karn's retransmit alg.).
1111 * Recompute the initial retransmit timer.
1113 if (tp->t_rtt && SEQ_GT(ti->ti_ack, tp->t_rtseq))
1114 tcp_xmit_timer(tp,tp->t_rtt);
1117 * If all outstanding data is acked, stop retransmit
1118 * timer and remember to restart (more output or persist).
1119 * If there is more data to be acked, restart retransmit
1120 * timer, using current (possibly backed-off) value.
1122 if (ti->ti_ack == tp->snd_max) {
1123 tp->t_timer[TCPT_REXMT] = 0;
1124 needoutput = 1;
1125 } else if (tp->t_timer[TCPT_PERSIST] == 0)
1126 tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
1128 * When new data is acked, open the congestion window.
1129 * If the window gives us less than ssthresh packets
1130 * in flight, open exponentially (maxseg per packet).
1131 * Otherwise open linearly: maxseg per window
1132 * (maxseg^2 / cwnd per packet).
1135 register u_int cw = tp->snd_cwnd;
1136 register u_int incr = tp->t_maxseg;
1138 if (cw > tp->snd_ssthresh)
1139 incr = incr * incr / cw;
1140 tp->snd_cwnd = MIN(cw + incr, TCP_MAXWIN << tp->snd_scale);
1142 if (acked > so->so_snd.sb_cc) {
1143 tp->snd_wnd -= so->so_snd.sb_cc;
1144 sbdrop(&so->so_snd, (int )so->so_snd.sb_cc);
1145 ourfinisacked = 1;
1146 } else {
1147 sbdrop(&so->so_snd, acked);
1148 tp->snd_wnd -= acked;
1149 ourfinisacked = 0;
1151 tp->snd_una = ti->ti_ack;
1152 if (SEQ_LT(tp->snd_nxt, tp->snd_una))
1153 tp->snd_nxt = tp->snd_una;
1155 switch (tp->t_state) {
1158 * In FIN_WAIT_1 STATE in addition to the processing
1159 * for the ESTABLISHED state if our FIN is now acknowledged
1160 * then enter FIN_WAIT_2.
1162 case TCPS_FIN_WAIT_1:
1163 if (ourfinisacked) {
1165 * If we can't receive any more
1166 * data, then closing user can proceed.
1167 * Starting the timer is contrary to the
1168 * specification, but if we don't get a FIN
1169 * we'll hang forever.
1171 if (so->so_state & SS_FCANTRCVMORE) {
1172 tp->t_timer[TCPT_2MSL] = TCP_MAXIDLE;
1174 tp->t_state = TCPS_FIN_WAIT_2;
1176 break;
1179 * In CLOSING STATE in addition to the processing for
1180 * the ESTABLISHED state if the ACK acknowledges our FIN
1181 * then enter the TIME-WAIT state, otherwise ignore
1182 * the segment.
1184 case TCPS_CLOSING:
1185 if (ourfinisacked) {
1186 tp->t_state = TCPS_TIME_WAIT;
1187 tcp_canceltimers(tp);
1188 tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1190 break;
1193 * In LAST_ACK, we may still be waiting for data to drain
1194 * and/or to be acked, as well as for the ack of our FIN.
1195 * If our FIN is now acknowledged, delete the TCB,
1196 * enter the closed state and return.
1198 case TCPS_LAST_ACK:
1199 if (ourfinisacked) {
1200 tcp_close(tp);
1201 goto drop;
1203 break;
1206 * In TIME_WAIT state the only thing that should arrive
1207 * is a retransmission of the remote FIN. Acknowledge
1208 * it and restart the finack timer.
1210 case TCPS_TIME_WAIT:
1211 tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1212 goto dropafterack;
1214 } /* switch(tp->t_state) */
1216 step6:
1218 * Update window information.
1219 * Don't look at window if no ACK: TAC's send garbage on first SYN.
1221 if ((tiflags & TH_ACK) &&
1222 (SEQ_LT(tp->snd_wl1, ti->ti_seq) ||
1223 (tp->snd_wl1 == ti->ti_seq && (SEQ_LT(tp->snd_wl2, ti->ti_ack) ||
1224 (tp->snd_wl2 == ti->ti_ack && tiwin > tp->snd_wnd))))) {
1225 tp->snd_wnd = tiwin;
1226 tp->snd_wl1 = ti->ti_seq;
1227 tp->snd_wl2 = ti->ti_ack;
1228 if (tp->snd_wnd > tp->max_sndwnd)
1229 tp->max_sndwnd = tp->snd_wnd;
1230 needoutput = 1;
1234 * Process segments with URG.
1236 if ((tiflags & TH_URG) && ti->ti_urp &&
1237 TCPS_HAVERCVDFIN(tp->t_state) == 0) {
1239 * This is a kludge, but if we receive and accept
1240 * random urgent pointers, we'll crash in
1241 * soreceive. It's hard to imagine someone
1242 * actually wanting to send this much urgent data.
1244 if (ti->ti_urp + so->so_rcv.sb_cc > so->so_rcv.sb_datalen) {
1245 ti->ti_urp = 0;
1246 tiflags &= ~TH_URG;
1247 goto dodata;
1250 * If this segment advances the known urgent pointer,
1251 * then mark the data stream. This should not happen
1252 * in CLOSE_WAIT, CLOSING, LAST_ACK or TIME_WAIT STATES since
1253 * a FIN has been received from the remote side.
1254 * In these states we ignore the URG.
1256 * According to RFC961 (Assigned Protocols),
1257 * the urgent pointer points to the last octet
1258 * of urgent data. We continue, however,
1259 * to consider it to indicate the first octet
1260 * of data past the urgent section as the original
1261 * spec states (in one of two places).
1263 if (SEQ_GT(ti->ti_seq+ti->ti_urp, tp->rcv_up)) {
1264 tp->rcv_up = ti->ti_seq + ti->ti_urp;
1265 so->so_urgc = so->so_rcv.sb_cc +
1266 (tp->rcv_up - tp->rcv_nxt); /* -1; */
1267 tp->rcv_up = ti->ti_seq + ti->ti_urp;
1270 } else
1272 * If no out of band data is expected,
1273 * pull receive urgent pointer along
1274 * with the receive window.
1276 if (SEQ_GT(tp->rcv_nxt, tp->rcv_up))
1277 tp->rcv_up = tp->rcv_nxt;
1278 dodata:
1281 * If this is a small packet, then ACK now - with Nagel
1282 * congestion avoidance sender won't send more until
1283 * he gets an ACK.
1285 if (ti->ti_len && (unsigned)ti->ti_len <= 5 &&
1286 ((struct tcpiphdr_2 *)ti)->first_char == (char)27) {
1287 tp->t_flags |= TF_ACKNOW;
1291 * Process the segment text, merging it into the TCP sequencing queue,
1292 * and arranging for acknowledgment of receipt if necessary.
1293 * This process logically involves adjusting tp->rcv_wnd as data
1294 * is presented to the user (this happens in tcp_usrreq.c,
1295 * case PRU_RCVD). If a FIN has already been received on this
1296 * connection then we just ignore the text.
1298 if ((ti->ti_len || (tiflags&TH_FIN)) &&
1299 TCPS_HAVERCVDFIN(tp->t_state) == 0) {
1300 TCP_REASS(tp, ti, m, so, tiflags);
1301 } else {
1302 m_free(m);
1303 tiflags &= ~TH_FIN;
1307 * If FIN is received ACK the FIN and let the user know
1308 * that the connection is closing.
1310 if (tiflags & TH_FIN) {
1311 if (TCPS_HAVERCVDFIN(tp->t_state) == 0) {
1313 * If we receive a FIN we can't send more data,
1314 * set it SS_FDRAIN
1315 * Shutdown the socket if there is no rx data in the
1316 * buffer.
1317 * soread() is called on completion of shutdown() and
1318 * will got to TCPS_LAST_ACK, and use tcp_output()
1319 * to send the FIN.
1321 sofwdrain(so);
1323 tp->t_flags |= TF_ACKNOW;
1324 tp->rcv_nxt++;
1326 switch (tp->t_state) {
1329 * In SYN_RECEIVED and ESTABLISHED STATES
1330 * enter the CLOSE_WAIT state.
1332 case TCPS_SYN_RECEIVED:
1333 case TCPS_ESTABLISHED:
1334 if(so->so_emu == EMU_CTL) /* no shutdown on socket */
1335 tp->t_state = TCPS_LAST_ACK;
1336 else
1337 tp->t_state = TCPS_CLOSE_WAIT;
1338 break;
1341 * If still in FIN_WAIT_1 STATE FIN has not been acked so
1342 * enter the CLOSING state.
1344 case TCPS_FIN_WAIT_1:
1345 tp->t_state = TCPS_CLOSING;
1346 break;
1349 * In FIN_WAIT_2 state enter the TIME_WAIT state,
1350 * starting the time-wait timer, turning off the other
1351 * standard timers.
1353 case TCPS_FIN_WAIT_2:
1354 tp->t_state = TCPS_TIME_WAIT;
1355 tcp_canceltimers(tp);
1356 tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1357 break;
1360 * In TIME_WAIT state restart the 2 MSL time_wait timer.
1362 case TCPS_TIME_WAIT:
1363 tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1364 break;
1369 * Return any desired output.
1371 if (needoutput || (tp->t_flags & TF_ACKNOW)) {
1372 (void) tcp_output(tp);
1374 return;
1376 dropafterack:
1378 * Generate an ACK dropping incoming segment if it occupies
1379 * sequence space, where the ACK reflects our state.
1381 if (tiflags & TH_RST)
1382 goto drop;
1383 m_free(m);
1384 tp->t_flags |= TF_ACKNOW;
1385 (void) tcp_output(tp);
1386 return;
1388 dropwithreset:
1389 /* reuses m if m!=NULL, m_free() unnecessary */
1390 if (tiflags & TH_ACK)
1391 tcp_respond(tp, ti, m, (tcp_seq)0, ti->ti_ack, TH_RST, af);
1392 else {
1393 if (tiflags & TH_SYN) ti->ti_len++;
1394 tcp_respond(tp, ti, m, ti->ti_seq + ti->ti_len, (tcp_seq) 0,
1395 TH_RST | TH_ACK, af);
1398 return;
1400 drop:
1402 * Drop space held by incoming segment and return.
1404 m_free(m);
1407 static void
1408 tcp_dooptions(struct tcpcb *tp, u_char *cp, int cnt, struct tcpiphdr *ti)
1410 uint16_t mss;
1411 int opt, optlen;
1413 DEBUG_CALL("tcp_dooptions");
1414 DEBUG_ARGS((dfd, " tp = %p cnt=%i\n", tp, cnt));
1416 for (; cnt > 0; cnt -= optlen, cp += optlen) {
1417 opt = cp[0];
1418 if (opt == TCPOPT_EOL)
1419 break;
1420 if (opt == TCPOPT_NOP)
1421 optlen = 1;
1422 else {
1423 optlen = cp[1];
1424 if (optlen <= 0)
1425 break;
1427 switch (opt) {
1429 default:
1430 continue;
1432 case TCPOPT_MAXSEG:
1433 if (optlen != TCPOLEN_MAXSEG)
1434 continue;
1435 if (!(ti->ti_flags & TH_SYN))
1436 continue;
1437 memcpy((char *) &mss, (char *) cp + 2, sizeof(mss));
1438 NTOHS(mss);
1439 (void) tcp_mss(tp, mss); /* sets t_maxseg */
1440 break;
1447 * Pull out of band byte out of a segment so
1448 * it doesn't appear in the user's data queue.
1449 * It is still reflected in the segment length for
1450 * sequencing purposes.
1453 #ifdef notdef
1455 void
1456 tcp_pulloutofband(so, ti, m)
1457 struct socket *so;
1458 struct tcpiphdr *ti;
1459 register struct mbuf *m;
1461 int cnt = ti->ti_urp - 1;
1463 while (cnt >= 0) {
1464 if (m->m_len > cnt) {
1465 char *cp = mtod(m, caddr_t) + cnt;
1466 struct tcpcb *tp = sototcpcb(so);
1468 tp->t_iobc = *cp;
1469 tp->t_oobflags |= TCPOOB_HAVEDATA;
1470 memcpy(sp, cp+1, (unsigned)(m->m_len - cnt - 1));
1471 m->m_len--;
1472 return;
1474 cnt -= m->m_len;
1475 m = m->m_next; /* XXX WRONG! Fix it! */
1476 if (m == 0)
1477 break;
1479 panic("tcp_pulloutofband");
1482 #endif /* notdef */
1485 * Collect new round-trip time estimate
1486 * and update averages and current timeout.
1489 static void
1490 tcp_xmit_timer(register struct tcpcb *tp, int rtt)
1492 register short delta;
1494 DEBUG_CALL("tcp_xmit_timer");
1495 DEBUG_ARG("tp = %p", tp);
1496 DEBUG_ARG("rtt = %d", rtt);
1498 if (tp->t_srtt != 0) {
1500 * srtt is stored as fixed point with 3 bits after the
1501 * binary point (i.e., scaled by 8). The following magic
1502 * is equivalent to the smoothing algorithm in rfc793 with
1503 * an alpha of .875 (srtt = rtt/8 + srtt*7/8 in fixed
1504 * point). Adjust rtt to origin 0.
1506 delta = rtt - 1 - (tp->t_srtt >> TCP_RTT_SHIFT);
1507 if ((tp->t_srtt += delta) <= 0)
1508 tp->t_srtt = 1;
1510 * We accumulate a smoothed rtt variance (actually, a
1511 * smoothed mean difference), then set the retransmit
1512 * timer to smoothed rtt + 4 times the smoothed variance.
1513 * rttvar is stored as fixed point with 2 bits after the
1514 * binary point (scaled by 4). The following is
1515 * equivalent to rfc793 smoothing with an alpha of .75
1516 * (rttvar = rttvar*3/4 + |delta| / 4). This replaces
1517 * rfc793's wired-in beta.
1519 if (delta < 0)
1520 delta = -delta;
1521 delta -= (tp->t_rttvar >> TCP_RTTVAR_SHIFT);
1522 if ((tp->t_rttvar += delta) <= 0)
1523 tp->t_rttvar = 1;
1524 } else {
1526 * No rtt measurement yet - use the unsmoothed rtt.
1527 * Set the variance to half the rtt (so our first
1528 * retransmit happens at 3*rtt).
1530 tp->t_srtt = rtt << TCP_RTT_SHIFT;
1531 tp->t_rttvar = rtt << (TCP_RTTVAR_SHIFT - 1);
1533 tp->t_rtt = 0;
1534 tp->t_rxtshift = 0;
1537 * the retransmit should happen at rtt + 4 * rttvar.
1538 * Because of the way we do the smoothing, srtt and rttvar
1539 * will each average +1/2 tick of bias. When we compute
1540 * the retransmit timer, we want 1/2 tick of rounding and
1541 * 1 extra tick because of +-1/2 tick uncertainty in the
1542 * firing of the timer. The bias will give us exactly the
1543 * 1.5 tick we need. But, because the bias is
1544 * statistical, we have to test that we don't drop below
1545 * the minimum feasible timer (which is 2 ticks).
1547 TCPT_RANGESET(tp->t_rxtcur, TCP_REXMTVAL(tp),
1548 (short)tp->t_rttmin, TCPTV_REXMTMAX); /* XXX */
1551 * We received an ack for a packet that wasn't retransmitted;
1552 * it is probably safe to discard any error indications we've
1553 * received recently. This isn't quite right, but close enough
1554 * for now (a route might have failed after we sent a segment,
1555 * and the return path might not be symmetrical).
1557 tp->t_softerror = 0;
1561 * Determine a reasonable value for maxseg size.
1562 * If the route is known, check route for mtu.
1563 * If none, use an mss that can be handled on the outgoing
1564 * interface without forcing IP to fragment; if bigger than
1565 * an mbuf cluster (MCLBYTES), round down to nearest multiple of MCLBYTES
1566 * to utilize large mbufs. If no route is found, route has no mtu,
1567 * or the destination isn't local, use a default, hopefully conservative
1568 * size (usually 512 or the default IP max size, but no more than the mtu
1569 * of the interface), as we can't discover anything about intervening
1570 * gateways or networks. We also initialize the congestion/slow start
1571 * window to be a single segment if the destination isn't local.
1572 * While looking at the routing entry, we also initialize other path-dependent
1573 * parameters from pre-set or cached values in the routing entry.
1577 tcp_mss(struct tcpcb *tp, u_int offer)
1579 struct socket *so = tp->t_socket;
1580 int mss;
1582 DEBUG_CALL("tcp_mss");
1583 DEBUG_ARG("tp = %p", tp);
1584 DEBUG_ARG("offer = %d", offer);
1586 switch (so->so_ffamily) {
1587 case AF_INET:
1588 mss = MIN(IF_MTU, IF_MRU) - sizeof(struct tcphdr)
1589 - sizeof(struct ip);
1590 break;
1591 case AF_INET6:
1592 mss = MIN(IF_MTU, IF_MRU) - sizeof(struct tcphdr)
1593 - sizeof(struct ip6);
1594 break;
1595 default:
1596 g_assert_not_reached();
1599 if (offer)
1600 mss = MIN(mss, offer);
1601 mss = MAX(mss, 32);
1602 if (mss < tp->t_maxseg || offer != 0)
1603 tp->t_maxseg = mss;
1605 tp->snd_cwnd = mss;
1607 sbreserve(&so->so_snd, TCP_SNDSPACE + ((TCP_SNDSPACE % mss) ?
1608 (mss - (TCP_SNDSPACE % mss)) :
1609 0));
1610 sbreserve(&so->so_rcv, TCP_RCVSPACE + ((TCP_RCVSPACE % mss) ?
1611 (mss - (TCP_RCVSPACE % mss)) :
1612 0));
1614 DEBUG_MISC((dfd, " returning mss = %d\n", mss));
1616 return mss;