tcp_input.c@ 15458

Last change on this file since 15458 was 15453, checked in by vboxsync, 16 years ago
slirp: removed the old 64-bit incompatible reassemble code
Property svn:eol-style set to `native`
File size: 65.5 KB

Line
1	/*
2	* Copyright (c) 1982, 1986, 1988, 1990, 1993, 1994
3	* The Regents of the University of California. All rights reserved.
4	*
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. All advertising materials mentioning features or use of this software
14	* must display the following acknowledgement:
15	* This product includes software developed by the University of
16	* California, Berkeley and its contributors.
17	* 4. Neither the name of the University nor the names of its contributors
18	* may be used to endorse or promote products derived from this software
19	* without specific prior written permission.
20	*
21	* THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22	* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23	* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24	* ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25	* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26	* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27	* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28	* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29	* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30	* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31	* SUCH DAMAGE.
32	*
33	* @(#)tcp_input.c 8.5 (Berkeley) 4/10/94
34	* tcp_input.c,v 1.10 1994/10/13 18:36:32 wollman Exp
35	*/
36
37	/*
38	* Changes and additions relating to SLiRP
39	* Copyright (c) 1995 Danny Gasparovski.
40	*
41	* Please read the file COPYRIGHT for the
42	* terms and conditions of the copyright.
43	*/
44
45	#include <slirp.h>
46	#include "ip_icmp.h"
47
48
49	#define TCP_PAWS_IDLE (24 * 24 * 60 * 60 * PR_SLOWHZ)
50
51	/* for modulo comparisons of timestamps */
52	#define TSTMP_LT(a,b) ((int)((a)-(b)) < 0)
53	#define TSTMP_GEQ(a,b) ((int)((a)-(b)) >= 0)
54
55	#ifndef TCP_ACK_HACK
56	#define DELAY_ACK(tp, ti) \
57	if (ti->ti_flags & TH_PUSH) \
58	tp->t_flags \|= TF_ACKNOW; \
59	else \
60	tp->t_flags \|= TF_DELACK;
61	#else /* !TCP_ACK_HACK */
62	#define DELAY_ACK(tp, ign) \
63	tp->t_flags \|= TF_DELACK;
64	#endif /* TCP_ACK_HACK */
65
66
67	/*
68	* deps: netinet/tcp_reass.c
69	* tcp_reass_maxqlen = 48 (deafault)
70	* tcp_reass_maxseg = nmbclusters/16 (nmbclusters = 1024 + maxusers * 64 from kern/kern_mbuf.c let's say 256)
71	*/
72	int
73	tcp_reass(PNATState pData, struct tcpcb tp, struct tcphdr th, int tlenp, struct mbuf m)
74	{
75	struct tseg_qent *q;
76	struct tseg_qent *p = NULL;
77	struct tseg_qent *nq;
78	struct tseg_qent *te = NULL;
79	struct socket *so = tp->t_socket;
80	int flags;
81
82	/*
83	* XXX: tcp_reass() is rather inefficient with its data structures
84	* and should be rewritten (see NetBSD for optimizations). While
85	* doing that it should move to its own file tcp_reass.c.
86	*/
87
88	/*
89	* Call with th==NULL after become established to
90	* force pre-ESTABLISHED data up to user socket.
91	*/
92	if (th == NULL)
93	goto present;
94
95	/*
96	* Limit the number of segments in the reassembly queue to prevent
97	* holding on to too many segments (and thus running out of mbufs).
98	* Make sure to let the missing segment through which caused this
99	* queue. Always keep one global queue entry spare to be able to
100	* process the missing segment.
101	*/
102	if (th->th_seq != tp->rcv_nxt &&
103	(tcp_reass_qsize + 1 >= tcp_reass_maxseg \|\|
104	tp->t_segqlen >= tcp_reass_maxqlen)) {
105	tcp_reass_overflows++;
106	tcpstat.tcps_rcvmemdrop++;
107	m_freem(pData, m);
108	*tlenp = 0;
109	return (0);
110	}
111
112	/*
113	* Allocate a new queue entry. If we can't, or hit the zone limit
114	* just drop the pkt.
115	*/
116	te = malloc(sizeof(struct tseg_qent));
117	if (te == NULL) {
118	tcpstat.tcps_rcvmemdrop++;
119	m_freem(pData, m);
120	*tlenp = 0;
121	return (0);
122	}
123	tp->t_segqlen++;
124	tcp_reass_qsize++;
125
126	/*
127	* Find a segment which begins after this one does.
128	*/
129	LIST_FOREACH(q, &tp->t_segq, tqe_q) {
130	if (SEQ_GT(q->tqe_th->th_seq, th->th_seq))
131	break;
132	p = q;
133	}
134
135	/*
136	* If there is a preceding segment, it may provide some of
137	* our data already. If so, drop the data from the incoming
138	* segment. If it provides all of our data, drop us.
139	*/
140	if (p != NULL) {
141	int i;
142	/* conversion to int (in i) handles seq wraparound */
143	i = p->tqe_th->th_seq + p->tqe_len - th->th_seq;
144	if (i > 0) {
145	if (i >= *tlenp) {
146	tcpstat.tcps_rcvduppack++;
147	tcpstat.tcps_rcvdupbyte += *tlenp;
148	m_freem(pData, m);
149	free(te);
150	tp->t_segqlen--;
151	tcp_reass_qsize--;
152	/*
153	* Try to present any queued data
154	* at the left window edge to the user.
155	* This is needed after the 3-WHS
156	* completes.
157	*/
158	goto present; /* ??? */
159	}
160	m_adj(m, i);
161	*tlenp -= i;
162	th->th_seq += i;
163	}
164	}
165	tcpstat.tcps_rcvoopack++;
166	tcpstat.tcps_rcvoobyte += *tlenp;
167
168	/*
169	* While we overlap succeeding segments trim them or,
170	* if they are completely covered, dequeue them.
171	*/
172	while (q) {
173	int i = (th->th_seq + *tlenp) - q->tqe_th->th_seq;
174	if (i <= 0)
175	break;
176	if (i < q->tqe_len) {
177	q->tqe_th->th_seq += i;
178	q->tqe_len -= i;
179	m_adj(q->tqe_m, i);
180	break;
181	}
182
183	nq = LIST_NEXT(q, tqe_q);
184	LIST_REMOVE(q, tqe_q);
185	m_freem(pData, q->tqe_m);
186	free(q);
187	tp->t_segqlen--;
188	tcp_reass_qsize--;
189	q = nq;
190	}
191
192	/* Insert the new segment queue entry into place. */
193	te->tqe_m = m;
194	te->tqe_th = th;
195	te->tqe_len = *tlenp;
196
197	if (p == NULL) {
198	LIST_INSERT_HEAD(&tp->t_segq, te, tqe_q);
199	} else {
200	LIST_INSERT_AFTER(p, te, tqe_q);
201	}
202
203	present:
204	/*
205	* Present data to user, advancing rcv_nxt through
206	* completed sequence space.
207	*/
208	if (!TCPS_HAVEESTABLISHED(tp->t_state))
209	return (0);
210	q = LIST_FIRST(&tp->t_segq);
211	if (!q \|\| q->tqe_th->th_seq != tp->rcv_nxt)
212	return (0);
213	do {
214	tp->rcv_nxt += q->tqe_len;
215	flags = q->tqe_th->th_flags & TH_FIN;
216	nq = LIST_NEXT(q, tqe_q);
217	LIST_REMOVE(q, tqe_q);
218	/* XXX: This place should be checked for the same code in
219	* original BSD code for Slirp and current BSD used SS_FCANTRCVMORE
220	*/
221	if (so->so_state & SS_FCANTSENDMORE)
222	m_freem(pData, q->tqe_m);
223	else
224	sbappend(pData, so, q->tqe_m);
225	free(q);
226	tp->t_segqlen--;
227	tcp_reass_qsize--;
228	q = nq;
229	} while (q && q->tqe_th->th_seq == tp->rcv_nxt);
230	return (flags);
231	}
232
233	/*
234	* TCP input routine, follows pages 65-76 of the
235	* protocol specification dated September, 1981 very closely.
236	*/
237	void
238	tcp_input(PNATState pData, register struct mbuf m, int iphlen, struct socket inso)
239	{
240	struct ip save_ip, *ip;
241	register struct tcpiphdr *ti;
242	caddr_t optp = NULL;
243	int optlen = 0;
244	int len, tlen, off;
245	register struct tcpcb *tp = 0;
246	register int tiflags;
247	struct socket *so = 0;
248	int todrop, acked, ourfinisacked, needoutput = 0;
249	/* int dropsocket = 0; */
250	int iss = 0;
251	u_long tiwin;
252	/* int ts_present = 0; */
253
254	DEBUG_CALL("tcp_input");
255	DEBUG_ARGS((dfd," m = %8lx iphlen = %2d inso = %lx\n",
256	(long )m, iphlen, (long )inso ));
257
258	/*
259	* If called with m == 0, then we're continuing the connect
260	*/
261	if (m == NULL) {
262	so = inso;
263
264	/* Re-set a few variables */
265	tp = sototcpcb(so);
266	m = so->so_m;
267	so->so_m = 0;
268	ti = so->so_ti;
269	tiwin = ti->ti_win;
270	tiflags = ti->ti_flags;
271
272	goto cont_conn;
273	}
274
275
276	tcpstat.tcps_rcvtotal++;
277	/*
278	* Get IP and TCP header together in first mbuf.
279	* Note: IP leaves IP header in first mbuf.
280	*/
281	ti = mtod(m, struct tcpiphdr *);
282	if (iphlen > sizeof(struct ip )) {
283	ip_stripoptions(m, (struct mbuf *)0);
284	iphlen=sizeof(struct ip );
285	}
286	/* XXX Check if too short */
287
288
289	/*
290	* Save a copy of the IP header in case we want restore it
291	* for sending an ICMP error message in response.
292	*/
293	ip=mtod(m, struct ip *);
294	save_ip = *ip;
295	save_ip.ip_len+= iphlen;
296
297	/*
298	* Checksum extended TCP header and data.
299	*/
300	tlen = ((struct ip *)ti)->ip_len;
301	memset(ti->ti_x1, 0, 9);
302	ti->ti_len = htons((u_int16_t)tlen);
303	len = sizeof(struct ip ) + tlen;
304	/* keep checksum for ICMP reply
305	* ti->ti_sum = cksum(m, len);
306	* if (ti->ti_sum) { */
307	if(cksum(m, len)) {
308	tcpstat.tcps_rcvbadsum++;
309	goto drop;
310	}
311
312	/*
313	* Check that TCP offset makes sense,
314	* pull out TCP options and adjust length. XXX
315	*/
316	off = ti->ti_off << 2;
317	if (off < sizeof (struct tcphdr) \|\| off > tlen) {
318	tcpstat.tcps_rcvbadoff++;
319	goto drop;
320	}
321	tlen -= off;
322	ti->ti_len = tlen;
323	if (off > sizeof (struct tcphdr)) {
324	optlen = off - sizeof (struct tcphdr);
325	optp = mtod(m, caddr_t) + sizeof (struct tcpiphdr);
326
327	/*
328	* Do quick retrieval of timestamp options ("options
329	* prediction?"). If timestamp is the only option and it's
330	* formatted as recommended in RFC 1323 appendix A, we
331	* quickly get the values now and not bother calling
332	* tcp_dooptions(), etc.
333	*/
334	/* if ((optlen == TCPOLEN_TSTAMP_APPA \|\|
335	* (optlen > TCPOLEN_TSTAMP_APPA &&
336	* optp[TCPOLEN_TSTAMP_APPA] == TCPOPT_EOL)) &&
337	* (u_int32_t )optp == htonl(TCPOPT_TSTAMP_HDR) &&
338	* (ti->ti_flags & TH_SYN) == 0) {
339	* ts_present = 1;
340	* ts_val = ntohl((u_int32_t )(optp + 4));
341	* ts_ecr = ntohl((u_int32_t )(optp + 8));
342	* optp = NULL; / * we've parsed the options * /
343	* }
344	*/
345	}
346	tiflags = ti->ti_flags;
347
348	/*
349	* Convert TCP protocol specific fields to host format.
350	*/
351	NTOHL(ti->ti_seq);
352	NTOHL(ti->ti_ack);
353	NTOHS(ti->ti_win);
354	NTOHS(ti->ti_urp);
355
356	/*
357	* Drop TCP, IP headers and TCP options.
358	*/
359	m->m_data += sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
360	m->m_len -= sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
361
362	/*
363	* Locate pcb for segment.
364	*/
365	findso:
366	so = tcp_last_so;
367	if (so->so_fport != ti->ti_dport \|\|
368	so->so_lport != ti->ti_sport \|\|
369	so->so_laddr.s_addr != ti->ti_src.s_addr \|\|
370	so->so_faddr.s_addr != ti->ti_dst.s_addr) {
371	so = solookup(&tcb, ti->ti_src, ti->ti_sport,
372	ti->ti_dst, ti->ti_dport);
373	if (so)
374	tcp_last_so = so;
375	++tcpstat.tcps_socachemiss;
376	}
377
378	/*
379	* If the state is CLOSED (i.e., TCB does not exist) then
380	* all data in the incoming segment is discarded.
381	* If the TCB exists but is in CLOSED state, it is embryonic,
382	* but should either do a listen or a connect soon.
383	*
384	* state == CLOSED means we've done socreate() but haven't
385	* attached it to a protocol yet...
386	*
387	* XXX If a TCB does not exist, and the TH_SYN flag is
388	* the only flag set, then create a session, mark it
389	* as if it was LISTENING, and continue...
390	*/
391	if (so == 0) {
392	if ((tiflags & (TH_SYN\|TH_FIN\|TH_RST\|TH_URG\|TH_ACK)) != TH_SYN)
393	goto dropwithreset;
394
395	if ((so = socreate()) == NULL)
396	goto dropwithreset;
397	if (tcp_attach(pData, so) < 0) {
398	free(so); /* Not sofree (if it failed, it's not insqued) */
399	goto dropwithreset;
400	}
401
402	sbreserve(&so->so_snd, tcp_sndspace);
403	sbreserve(&so->so_rcv, tcp_rcvspace);
404
405	/* tcp_last_so = so; / / XXX ? */
406	/* tp = sototcpcb(so); */
407
408	so->so_laddr = ti->ti_src;
409	so->so_lport = ti->ti_sport;
410	so->so_faddr = ti->ti_dst;
411	so->so_fport = ti->ti_dport;
412
413	if ((so->so_iptos = tcp_tos(so)) == 0)
414	so->so_iptos = ((struct ip *)ti)->ip_tos;
415
416	tp = sototcpcb(so);
417	tp->t_state = TCPS_LISTEN;
418	}
419
420	/*
421	* If this is a still-connecting socket, this probably
422	* a retransmit of the SYN. Whether it's a retransmit SYN
423	* or something else, we nuke it.
424	*/
425	if (so->so_state & SS_ISFCONNECTING)
426	goto drop;
427
428	tp = sototcpcb(so);
429
430	/* XXX Should never fail */
431	if (tp == 0)
432	goto dropwithreset;
433	if (tp->t_state == TCPS_CLOSED)
434	goto drop;
435
436	/* Unscale the window into a 32-bit value. */
437	/* if ((tiflags & TH_SYN) == 0)
438	* tiwin = ti->ti_win << tp->snd_scale;
439	* else
440	*/
441	tiwin = ti->ti_win;
442
443	/*
444	* Segment received on connection.
445	* Reset idle time and keep-alive timer.
446	*/
447	tp->t_idle = 0;
448	if (so_options)
449	tp->t_timer[TCPT_KEEP] = tcp_keepintvl;
450	else
451	tp->t_timer[TCPT_KEEP] = tcp_keepidle;
452
453	/*
454	* Process options if not in LISTEN state,
455	* else do it below (after getting remote address).
456	*/
457	if (optp && tp->t_state != TCPS_LISTEN)
458	tcp_dooptions(pData, tp, (u_char *)optp, optlen, ti);
459	/* , */
460	/* &ts_present, &ts_val, &ts_ecr); */
461
462	/*
463	* Header prediction: check for the two common cases
464	* of a uni-directional data xfer. If the packet has
465	* no control flags, is in-sequence, the window didn't
466	* change and we're not retransmitting, it's a
467	* candidate. If the length is zero and the ack moved
468	* forward, we're the sender side of the xfer. Just
469	* free the data acked & wake any higher level process
470	* that was blocked waiting for space. If the length
471	* is non-zero and the ack didn't move, we're the
472	* receiver side. If we're getting packets in-order
473	* (the reassembly queue is empty), add the data to
474	* the socket buffer and note that we need a delayed ack.
475	*
476	* XXX Some of these tests are not needed
477	* eg: the tiwin == tp->snd_wnd prevents many more
478	* predictions.. with no real advantage..
479	*/
480	if (tp->t_state == TCPS_ESTABLISHED &&
481	(tiflags & (TH_SYN\|TH_FIN\|TH_RST\|TH_URG\|TH_ACK)) == TH_ACK &&
482	/* (!ts_present \|\| TSTMP_GEQ(ts_val, tp->ts_recent)) && */
483	ti->ti_seq == tp->rcv_nxt &&
484	tiwin && tiwin == tp->snd_wnd &&
485	tp->snd_nxt == tp->snd_max) {
486	/*
487	* If last ACK falls within this segment's sequence numbers,
488	* record the timestamp.
489	*/
490	/* if (ts_present && SEQ_LEQ(ti->ti_seq, tp->last_ack_sent) &&
491	* SEQ_LT(tp->last_ack_sent, ti->ti_seq + ti->ti_len)) {
492	* tp->ts_recent_age = tcp_now;
493	* tp->ts_recent = ts_val;
494	* }
495	*/
496	if (ti->ti_len == 0) {
497	if (SEQ_GT(ti->ti_ack, tp->snd_una) &&
498	SEQ_LEQ(ti->ti_ack, tp->snd_max) &&
499	tp->snd_cwnd >= tp->snd_wnd) {
500	/*
501	* this is a pure ack for outstanding data.
502	*/
503	++tcpstat.tcps_predack;
504	/* if (ts_present)
505	* tcp_xmit_timer(tp, tcp_now-ts_ecr+1);
506	* else
507	*/ if (tp->t_rtt &&
508	SEQ_GT(ti->ti_ack, tp->t_rtseq))
509	tcp_xmit_timer(pData, tp, tp->t_rtt);
510	acked = ti->ti_ack - tp->snd_una;
511	tcpstat.tcps_rcvackpack++;
512	tcpstat.tcps_rcvackbyte += acked;
513	sbdrop(&so->so_snd, acked);
514	tp->snd_una = ti->ti_ack;
515	m_freem(pData, m);
516
517	/*
518	* If all outstanding data are acked, stop
519	* retransmit timer, otherwise restart timer
520	* using current (possibly backed-off) value.
521	* If process is waiting for space,
522	* wakeup/selwakeup/signal. If data
523	* are ready to send, let tcp_output
524	* decide between more output or persist.
525	*/
526	if (tp->snd_una == tp->snd_max)
527	tp->t_timer[TCPT_REXMT] = 0;
528	else if (tp->t_timer[TCPT_PERSIST] == 0)
529	tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
530
531	/*
532	* There's room in so_snd, sowwakup will read()
533	* from the socket if we can
534	*/
535	/* if (so->so_snd.sb_flags & SB_NOTIFY)
536	* sowwakeup(so);
537	*/
538	/*
539	* This is called because sowwakeup might have
540	* put data into so_snd. Since we don't so sowwakeup,
541	* we don't need this.. XXX???
542	*/
543	if (so->so_snd.sb_cc)
544	(void) tcp_output(pData, tp);
545
546	return;
547	}
548	} else if (ti->ti_ack == tp->snd_una &&
549	LIST_FIRST(&tp->t_segq) &&
550	ti->ti_len <= sbspace(&so->so_rcv)) {
551	/*
552	* this is a pure, in-sequence data packet
553	* with nothing on the reassembly queue and
554	* we have enough buffer space to take it.
555	*/
556	++tcpstat.tcps_preddat;
557	tp->rcv_nxt += ti->ti_len;
558	tcpstat.tcps_rcvpack++;
559	tcpstat.tcps_rcvbyte += ti->ti_len;
560	/*
561	* Add data to socket buffer.
562	*/
563	if (so->so_emu) {
564	if (tcp_emu(pData, so,m)) sbappend(pData, so, m);
565	} else
566	sbappend(pData, so, m);
567
568	/*
569	* XXX This is called when data arrives. Later, check
570	* if we can actually write() to the socket
571	* XXX Need to check? It's be NON_BLOCKING
572	*/
573	/* sorwakeup(so); */
574
575	/*
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.
579	*
580	* It is better to not delay acks at all to maximize
581	* TCP throughput. See RFC 2581.
582	*/
583	tp->t_flags \|= TF_ACKNOW;
584	tcp_output(pData, tp);
585	return;
586	}
587	} /* header prediction */
588	/*
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.
593	*/
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));
599	}
600
601	switch (tp->t_state) {
602
603	/*
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.
615	*/
616	case TCPS_LISTEN: {
617
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;
624
625	/*
626	* This has way too many gotos...
627	* But a bit of spaghetti code never hurt anybody :)
628	*/
629
630	if (so->so_emu & EMU_NOCONNECT) {
631	so->so_emu &= ~EMU_NOCONNECT;
632	goto cont_input;
633	}
634
635	if((tcp_fconnect(pData, so) == -1) && (errno != EINPROGRESS) && (errno != EWOULDBLOCK)) {
636	u_char code=ICMP_UNREACH_NET;
637	DEBUG_MISC((dfd," tcp fconnect errno = %d-%s\n",
638	errno,strerror(errno)));
639	if(errno == ECONNREFUSED) {
640	/* ACK the SYN, send RST to refuse the connection */
641	tcp_respond(pData, tp, ti, m, ti->ti_seq+1, (tcp_seq)0,
642	TH_RST\|TH_ACK);
643	} else {
644	if(errno == EHOSTUNREACH) code=ICMP_UNREACH_HOST;
645	HTONL(ti->ti_seq); /* restore tcp header */
646	HTONL(ti->ti_ack);
647	HTONS(ti->ti_win);
648	HTONS(ti->ti_urp);
649	m->m_data -= sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
650	m->m_len += sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
651	*ip=save_ip;
652	icmp_error(pData, m, ICMP_UNREACH,code, 0,strerror(errno));
653	}
654	tp = tcp_close(pData, tp);
655	m_free(pData, m);
656	} else {
657	/*
658	* Haven't connected yet, save the current mbuf
659	* and ti, and return
660	* XXX Some OS's don't tell us whether the connect()
661	* succeeded or not. So we must time it out.
662	*/
663	so->so_m = m;
664	so->so_ti = ti;
665	tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT;
666	tp->t_state = TCPS_SYN_RECEIVED;
667	}
668	return;
669
670	cont_conn:
671	/* m==NULL
672	* Check if the connect succeeded
673	*/
674	if (so->so_state & SS_NOFDREF) {
675	tp = tcp_close(pData, tp);
676	goto dropwithreset;
677	}
678	cont_input:
679	tcp_template(tp);
680
681	if (optp)
682	tcp_dooptions(pData, tp, (u_char *)optp, optlen, ti);
683	/* , */
684	/* &ts_present, &ts_val, &ts_ecr); */
685
686	if (iss)
687	tp->iss = iss;
688	else
689	tp->iss = tcp_iss;
690	tcp_iss += TCP_ISSINCR/2;
691	tp->irs = ti->ti_seq;
692	tcp_sendseqinit(tp);
693	tcp_rcvseqinit(tp);
694	tp->t_flags \|= TF_ACKNOW;
695	tp->t_state = TCPS_SYN_RECEIVED;
696	tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT;
697	tcpstat.tcps_accepts++;
698	goto trimthenstep6;
699	} /* case TCPS_LISTEN */
700
701	/*
702	* If the state is SYN_SENT:
703	* if seg contains an ACK, but not for our SYN, drop the input.
704	* if seg contains a RST, then drop the connection.
705	* if seg does not contain SYN, then drop it.
706	* Otherwise this is an acceptable SYN segment
707	* initialize tp->rcv_nxt and tp->irs
708	* if seg contains ack then advance tp->snd_una
709	* if SYN has been acked change to ESTABLISHED else SYN_RCVD state
710	* arrange for segment to be acked (eventually)
711	* continue processing rest of data/controls, beginning with URG
712	*/
713	case TCPS_SYN_SENT:
714	if ((tiflags & TH_ACK) &&
715	(SEQ_LEQ(ti->ti_ack, tp->iss) \|\|
716	SEQ_GT(ti->ti_ack, tp->snd_max)))
717	goto dropwithreset;
718
719	if (tiflags & TH_RST) {
720	if (tiflags & TH_ACK)
721	tp = tcp_drop(pData, tp,0); /* XXX Check t_softerror! */
722	goto drop;
723	}
724
725	if ((tiflags & TH_SYN) == 0)
726	goto drop;
727	if (tiflags & TH_ACK) {
728	tp->snd_una = ti->ti_ack;
729	if (SEQ_LT(tp->snd_nxt, tp->snd_una))
730	tp->snd_nxt = tp->snd_una;
731	}
732
733	tp->t_timer[TCPT_REXMT] = 0;
734	tp->irs = ti->ti_seq;
735	tcp_rcvseqinit(tp);
736	tp->t_flags \|= TF_ACKNOW;
737	if (tiflags & TH_ACK && SEQ_GT(tp->snd_una, tp->iss)) {
738	tcpstat.tcps_connects++;
739	soisfconnected(so);
740	tp->t_state = TCPS_ESTABLISHED;
741
742	/* Do window scaling on this connection? */
743	/* if ((tp->t_flags & (TF_RCVD_SCALE\|TF_REQ_SCALE)) ==
744	* (TF_RCVD_SCALE\|TF_REQ_SCALE)) {
745	* tp->snd_scale = tp->requested_s_scale;
746	* tp->rcv_scale = tp->request_r_scale;
747	* }
748	*/
749	(void) tcp_reass(pData, tp, (struct tcphdr )0, NULL, (struct mbuf )0);
750	/*
751	* if we didn't have to retransmit the SYN,
752	* use its rtt as our initial srtt & rtt var.
753	*/
754	if (tp->t_rtt)
755	tcp_xmit_timer(pData, tp, tp->t_rtt);
756	} else
757	tp->t_state = TCPS_SYN_RECEIVED;
758
759	trimthenstep6:
760	/*
761	* Advance ti->ti_seq to correspond to first data byte.
762	* If data, trim to stay within window,
763	* dropping FIN if necessary.
764	*/
765	ti->ti_seq++;
766	if (ti->ti_len > tp->rcv_wnd) {
767	todrop = ti->ti_len - tp->rcv_wnd;
768	m_adj(m, -todrop);
769	ti->ti_len = tp->rcv_wnd;
770	tiflags &= ~TH_FIN;
771	tcpstat.tcps_rcvpackafterwin++;
772	tcpstat.tcps_rcvbyteafterwin += todrop;
773	}
774	tp->snd_wl1 = ti->ti_seq - 1;
775	tp->rcv_up = ti->ti_seq;
776	goto step6;
777	} /* switch tp->t_state */
778	/*
779	* States other than LISTEN or SYN_SENT.
780	* First check timestamp, if present.
781	* Then check that at least some bytes of segment are within
782	* receive window. If segment begins before rcv_nxt,
783	* drop leading data (and SYN); if nothing left, just ack.
784	*
785	* RFC 1323 PAWS: If we have a timestamp reply on this segment
786	* and it's less than ts_recent, drop it.
787	*/
788	/* if (ts_present && (tiflags & TH_RST) == 0 && tp->ts_recent &&
789	* TSTMP_LT(ts_val, tp->ts_recent)) {
790	*
791	/ / Check to see if ts_recent is over 24 days old. */
792	/* if ((int)(tcp_now - tp->ts_recent_age) > TCP_PAWS_IDLE) {
793	/ /
794	* * Invalidate ts_recent. If this segment updates
795	* * ts_recent, the age will be reset later and ts_recent
796	* * will get a valid value. If it does not, setting
797	* * ts_recent to zero will at least satisfy the
798	* * requirement that zero be placed in the timestamp
799	* * echo reply when ts_recent isn't valid. The
800	* * age isn't reset until we get a valid ts_recent
801	* * because we don't want out-of-order segments to be
802	* * dropped when ts_recent is old.
803	* */
804	/* tp->ts_recent = 0;
805	* } else {
806	* tcpstat.tcps_rcvduppack++;
807	* tcpstat.tcps_rcvdupbyte += ti->ti_len;
808	* tcpstat.tcps_pawsdrop++;
809	* goto dropafterack;
810	* }
811	* }
812	*/
813
814	todrop = tp->rcv_nxt - ti->ti_seq;
815	if (todrop > 0) {
816	if (tiflags & TH_SYN) {
817	tiflags &= ~TH_SYN;
818	ti->ti_seq++;
819	if (ti->ti_urp > 1)
820	ti->ti_urp--;
821	else
822	tiflags &= ~TH_URG;
823	todrop--;
824	}
825	/*
826	* Following if statement from Stevens, vol. 2, p. 960.
827	*/
828	if (todrop > ti->ti_len
829	\|\| (todrop == ti->ti_len && (tiflags & TH_FIN) == 0)) {
830	/*
831	* Any valid FIN must be to the left of the window.
832	* At this point the FIN must be a duplicate or out
833	* of sequence; drop it.
834	*/
835	tiflags &= ~TH_FIN;
836
837	/*
838	* Send an ACK to resynchronize and drop any data.
839	* But keep on processing for RST or ACK.
840	*/
841	tp->t_flags \|= TF_ACKNOW;
842	todrop = ti->ti_len;
843	tcpstat.tcps_rcvduppack++;
844	tcpstat.tcps_rcvdupbyte += todrop;
845	} else {
846	tcpstat.tcps_rcvpartduppack++;
847	tcpstat.tcps_rcvpartdupbyte += todrop;
848	}
849	m_adj(m, todrop);
850	ti->ti_seq += todrop;
851	ti->ti_len -= todrop;
852	if (ti->ti_urp > todrop)
853	ti->ti_urp -= todrop;
854	else {
855	tiflags &= ~TH_URG;
856	ti->ti_urp = 0;
857	}
858	}
859	/*
860	* If new data are received on a connection after the
861	* user processes are gone, then RST the other end.
862	*/
863	if ((so->so_state & SS_NOFDREF) &&
864	tp->t_state > TCPS_CLOSE_WAIT && ti->ti_len) {
865	tp = tcp_close(pData, tp);
866	tcpstat.tcps_rcvafterclose++;
867	goto dropwithreset;
868	}
869
870	/*
871	* If segment ends after window, drop trailing data
872	* (and PUSH and FIN); if nothing left, just ACK.
873	*/
874	todrop = (ti->ti_seq+ti->ti_len) - (tp->rcv_nxt+tp->rcv_wnd);
875	if (todrop > 0) {
876	tcpstat.tcps_rcvpackafterwin++;
877	if (todrop >= ti->ti_len) {
878	tcpstat.tcps_rcvbyteafterwin += ti->ti_len;
879	/*
880	* If a new connection request is received
881	* while in TIME_WAIT, drop the old connection
882	* and start over if the sequence numbers
883	* are above the previous ones.
884	*/
885	if (tiflags & TH_SYN &&
886	tp->t_state == TCPS_TIME_WAIT &&
887	SEQ_GT(ti->ti_seq, tp->rcv_nxt)) {
888	iss = tp->rcv_nxt + TCP_ISSINCR;
889	tp = tcp_close(pData, tp);
890	goto findso;
891	}
892	/*
893	* If window is closed can only take segments at
894	* window edge, and have to drop data and PUSH from
895	* incoming segments. Continue processing, but
896	* remember to ack. Otherwise, drop segment
897	* and ack.
898	*/
899	if (tp->rcv_wnd == 0 && ti->ti_seq == tp->rcv_nxt) {
900	tp->t_flags \|= TF_ACKNOW;
901	tcpstat.tcps_rcvwinprobe++;
902	} else
903	goto dropafterack;
904	} else
905	tcpstat.tcps_rcvbyteafterwin += todrop;
906	m_adj(m, -todrop);
907	ti->ti_len -= todrop;
908	tiflags &= ~(TH_PUSH\|TH_FIN);
909	}
910
911	/*
912	* If last ACK falls within this segment's sequence numbers,
913	* record its timestamp.
914	*/
915	/* if (ts_present && SEQ_LEQ(ti->ti_seq, tp->last_ack_sent) &&
916	* SEQ_LT(tp->last_ack_sent, ti->ti_seq + ti->ti_len +
917	* ((tiflags & (TH_SYN\|TH_FIN)) != 0))) {
918	* tp->ts_recent_age = tcp_now;
919	* tp->ts_recent = ts_val;
920	* }
921	*/
922
923	/*
924	* If the RST bit is set examine the state:
925	* SYN_RECEIVED STATE:
926	* If passive open, return to LISTEN state.
927	* If active open, inform user that connection was refused.
928	* ESTABLISHED, FIN_WAIT_1, FIN_WAIT2, CLOSE_WAIT STATES:
929	* Inform user that connection was reset, and close tcb.
930	* CLOSING, LAST_ACK, TIME_WAIT STATES
931	* Close the tcb.
932	*/
933	if (tiflags&TH_RST) switch (tp->t_state) {
934
935	case TCPS_SYN_RECEIVED:
936	/* so->so_error = ECONNREFUSED; */
937	goto close;
938
939	case TCPS_ESTABLISHED:
940	case TCPS_FIN_WAIT_1:
941	case TCPS_FIN_WAIT_2:
942	case TCPS_CLOSE_WAIT:
943	/* so->so_error = ECONNRESET; */
944	close:
945	tp->t_state = TCPS_CLOSED;
946	tcpstat.tcps_drops++;
947	tp = tcp_close(pData, tp);
948	goto drop;
949
950	case TCPS_CLOSING:
951	case TCPS_LAST_ACK:
952	case TCPS_TIME_WAIT:
953	tp = tcp_close(pData, tp);
954	goto drop;
955	}
956
957	/*
958	* If a SYN is in the window, then this is an
959	* error and we send an RST and drop the connection.
960	*/
961	if (tiflags & TH_SYN) {
962	tp = tcp_drop(pData, tp,0);
963	goto dropwithreset;
964	}
965
966	/*
967	* If the ACK bit is off we drop the segment and return.
968	*/
969	if ((tiflags & TH_ACK) == 0) goto drop;
970
971	/*
972	* Ack processing.
973	*/
974	switch (tp->t_state) {
975	/*
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
979	*/
980	case TCPS_SYN_RECEIVED:
981
982	if (SEQ_GT(tp->snd_una, ti->ti_ack) \|\|
983	SEQ_GT(ti->ti_ack, tp->snd_max))
984	goto dropwithreset;
985	tcpstat.tcps_connects++;
986	tp->t_state = TCPS_ESTABLISHED;
987	/*
988	* The sent SYN is ack'ed with our sequence number +1
989	* The first data byte already in the buffer will get
990	* lost if no correction is made. This is only needed for
991	* SS_CTL since the buffer is empty otherwise.
992	* tp->snd_una++; or:
993	*/
994	tp->snd_una=ti->ti_ack;
995	soisfconnected(so);
996
997	/* Do window scaling? */
998	/* if ((tp->t_flags & (TF_RCVD_SCALE\|TF_REQ_SCALE)) ==
999	* (TF_RCVD_SCALE\|TF_REQ_SCALE)) {
1000	* tp->snd_scale = tp->requested_s_scale;
1001	* tp->rcv_scale = tp->request_r_scale;
1002	* }
1003	*/
1004	(void) tcp_reass(pData, tp, (struct tcphdr )0, (int )0, (struct mbuf *)0);
1005	tp->snd_wl1 = ti->ti_seq - 1;
1006	/* Avoid ack processing; snd_una==ti_ack => dup ack */
1007	goto synrx_to_est;
1008	/* fall into ... */
1009
1010	/*
1011	* In ESTABLISHED state: drop duplicate ACKs; ACK out of range
1012	* ACKs. If the ack is in the range
1013	* tp->snd_una < ti->ti_ack <= tp->snd_max
1014	* then advance tp->snd_una to ti->ti_ack and drop
1015	* data from the retransmission queue. If this ACK reflects
1016	* more up to date window information we update our window information.
1017	*/
1018	case TCPS_ESTABLISHED:
1019	case TCPS_FIN_WAIT_1:
1020	case TCPS_FIN_WAIT_2:
1021	case TCPS_CLOSE_WAIT:
1022	case TCPS_CLOSING:
1023	case TCPS_LAST_ACK:
1024	case TCPS_TIME_WAIT:
1025
1026	if (SEQ_LEQ(ti->ti_ack, tp->snd_una)) {
1027	if (ti->ti_len == 0 && tiwin == tp->snd_wnd) {
1028	tcpstat.tcps_rcvdupack++;
1029	DEBUG_MISC((dfd," dup ack m = %lx so = %lx \n",
1030	(long )m, (long )so));
1031	/*
1032	* If we have outstanding data (other than
1033	* a window probe), this is a completely
1034	* duplicate ack (ie, window info didn't
1035	* change), the ack is the biggest we've
1036	* seen and we've seen exactly our rexmt
1037	* threshold of them, assume a packet
1038	* has been dropped and retransmit it.
1039	* Kludge snd_nxt & the congestion
1040	* window so we send only this one
1041	* packet.
1042	*
1043	* We know we're losing at the current
1044	* window size so do congestion avoidance
1045	* (set ssthresh to half the current window
1046	* and pull our congestion window back to
1047	* the new ssthresh).
1048	*
1049	* Dup acks mean that packets have left the
1050	* network (they're now cached at the receiver)
1051	* so bump cwnd by the amount in the receiver
1052	* to keep a constant cwnd packets in the
1053	* network.
1054	*/
1055	if (tp->t_timer[TCPT_REXMT] == 0 \|\|
1056	ti->ti_ack != tp->snd_una)
1057	tp->t_dupacks = 0;
1058	else if (++tp->t_dupacks == tcprexmtthresh) {
1059	tcp_seq onxt = tp->snd_nxt;
1060	u_int win =
1061	min(tp->snd_wnd, tp->snd_cwnd) / 2 /
1062	tp->t_maxseg;
1063
1064	if (win < 2)
1065	win = 2;
1066	tp->snd_ssthresh = win * tp->t_maxseg;
1067	tp->t_timer[TCPT_REXMT] = 0;
1068	tp->t_rtt = 0;
1069	tp->snd_nxt = ti->ti_ack;
1070	tp->snd_cwnd = tp->t_maxseg;
1071	(void) tcp_output(pData, tp);
1072	tp->snd_cwnd = tp->snd_ssthresh +
1073	tp->t_maxseg * tp->t_dupacks;
1074	if (SEQ_GT(onxt, tp->snd_nxt))
1075	tp->snd_nxt = onxt;
1076	goto drop;
1077	} else if (tp->t_dupacks > tcprexmtthresh) {
1078	tp->snd_cwnd += tp->t_maxseg;
1079	(void) tcp_output(pData, tp);
1080	goto drop;
1081	}
1082	} else
1083	tp->t_dupacks = 0;
1084	break;
1085	}
1086	synrx_to_est:
1087	/*
1088	* If the congestion window was inflated to account
1089	* for the other side's cached packets, retract it.
1090	*/
1091	if (tp->t_dupacks > tcprexmtthresh &&
1092	tp->snd_cwnd > tp->snd_ssthresh)
1093	tp->snd_cwnd = tp->snd_ssthresh;
1094	tp->t_dupacks = 0;
1095	if (SEQ_GT(ti->ti_ack, tp->snd_max)) {
1096	tcpstat.tcps_rcvacktoomuch++;
1097	goto dropafterack;
1098	}
1099	acked = ti->ti_ack - tp->snd_una;
1100	tcpstat.tcps_rcvackpack++;
1101	tcpstat.tcps_rcvackbyte += acked;
1102
1103	/*
1104	* If we have a timestamp reply, update smoothed
1105	* round trip time. If no timestamp is present but
1106	* transmit timer is running and timed sequence
1107	* number was acked, update smoothed round trip time.
1108	* Since we now have an rtt measurement, cancel the
1109	* timer backoff (cf., Phil Karn's retransmit alg.).
1110	* Recompute the initial retransmit timer.
1111	*/
1112	/* if (ts_present)
1113	* tcp_xmit_timer(tp, tcp_now-ts_ecr+1);
1114	* else
1115	*/
1116	if (tp->t_rtt && SEQ_GT(ti->ti_ack, tp->t_rtseq))
1117	tcp_xmit_timer(pData, tp,tp->t_rtt);
1118
1119	/*
1120	* If all outstanding data is acked, stop retransmit
1121	* timer and remember to restart (more output or persist).
1122	* If there is more data to be acked, restart retransmit
1123	* timer, using current (possibly backed-off) value.
1124	*/
1125	if (ti->ti_ack == tp->snd_max) {
1126	tp->t_timer[TCPT_REXMT] = 0;
1127	needoutput = 1;
1128	} else if (tp->t_timer[TCPT_PERSIST] == 0)
1129	tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
1130	/*
1131	* When new data is acked, open the congestion window.
1132	* If the window gives us less than ssthresh packets
1133	* in flight, open exponentially (maxseg per packet).
1134	* Otherwise open linearly: maxseg per window
1135	* (maxseg^2 / cwnd per packet).
1136	*/
1137	{
1138	register u_int cw = tp->snd_cwnd;
1139	register u_int incr = tp->t_maxseg;
1140
1141	if (cw > tp->snd_ssthresh)
1142	incr = incr * incr / cw;
1143	tp->snd_cwnd = min(cw + incr, TCP_MAXWIN<<tp->snd_scale);
1144	}
1145	if (acked > so->so_snd.sb_cc) {
1146	tp->snd_wnd -= so->so_snd.sb_cc;
1147	sbdrop(&so->so_snd, (int )so->so_snd.sb_cc);
1148	ourfinisacked = 1;
1149	} else {
1150	sbdrop(&so->so_snd, acked);
1151	tp->snd_wnd -= acked;
1152	ourfinisacked = 0;
1153	}
1154	/*
1155	* XXX sowwakup is called when data is acked and there's room for
1156	* for more data... it should read() the socket
1157	*/
1158	/* if (so->so_snd.sb_flags & SB_NOTIFY)
1159	* sowwakeup(so);
1160	*/
1161	tp->snd_una = ti->ti_ack;
1162	if (SEQ_LT(tp->snd_nxt, tp->snd_una))
1163	tp->snd_nxt = tp->snd_una;
1164
1165	switch (tp->t_state) {
1166
1167	/*
1168	* In FIN_WAIT_1 STATE in addition to the processing
1169	* for the ESTABLISHED state if our FIN is now acknowledged
1170	* then enter FIN_WAIT_2.
1171	*/
1172	case TCPS_FIN_WAIT_1:
1173	if (ourfinisacked) {
1174	/*
1175	* If we can't receive any more
1176	* data, then closing user can proceed.
1177	* Starting the timer is contrary to the
1178	* specification, but if we don't get a FIN
1179	* we'll hang forever.
1180	*/
1181	if (so->so_state & SS_FCANTRCVMORE) {
1182	soisfdisconnected(so);
1183	tp->t_timer[TCPT_2MSL] = tcp_maxidle;
1184	}
1185	tp->t_state = TCPS_FIN_WAIT_2;
1186	}
1187	break;
1188
1189	/*
1190	* In CLOSING STATE in addition to the processing for
1191	* the ESTABLISHED state if the ACK acknowledges our FIN
1192	* then enter the TIME-WAIT state, otherwise ignore
1193	* the segment.
1194	*/
1195	case TCPS_CLOSING:
1196	if (ourfinisacked) {
1197	tp->t_state = TCPS_TIME_WAIT;
1198	tcp_canceltimers(tp);
1199	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1200	soisfdisconnected(so);
1201	}
1202	break;
1203
1204	/*
1205	* In LAST_ACK, we may still be waiting for data to drain
1206	* and/or to be acked, as well as for the ack of our FIN.
1207	* If our FIN is now acknowledged, delete the TCB,
1208	* enter the closed state and return.
1209	*/
1210	case TCPS_LAST_ACK:
1211	if (ourfinisacked) {
1212	tp = tcp_close(pData, tp);
1213	goto drop;
1214	}
1215	break;
1216
1217	/*
1218	* In TIME_WAIT state the only thing that should arrive
1219	* is a retransmission of the remote FIN. Acknowledge
1220	* it and restart the finack timer.
1221	*/
1222	case TCPS_TIME_WAIT:
1223	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1224	goto dropafterack;
1225	}
1226	} /* switch(tp->t_state) */
1227
1228	step6:
1229	/*
1230	* Update window information.
1231	* Don't look at window if no ACK: TAC's send garbage on first SYN.
1232	*/
1233	if ((tiflags & TH_ACK) &&
1234	(SEQ_LT(tp->snd_wl1, ti->ti_seq) \|\|
1235	(tp->snd_wl1 == ti->ti_seq && (SEQ_LT(tp->snd_wl2, ti->ti_ack) \|\|
1236	(tp->snd_wl2 == ti->ti_ack && tiwin > tp->snd_wnd))))) {
1237	/* keep track of pure window updates */
1238	if (ti->ti_len == 0 &&
1239	tp->snd_wl2 == ti->ti_ack && tiwin > tp->snd_wnd)
1240	tcpstat.tcps_rcvwinupd++;
1241	tp->snd_wnd = tiwin;
1242	tp->snd_wl1 = ti->ti_seq;
1243	tp->snd_wl2 = ti->ti_ack;
1244	if (tp->snd_wnd > tp->max_sndwnd)
1245	tp->max_sndwnd = tp->snd_wnd;
1246	needoutput = 1;
1247	}
1248
1249	/*
1250	* Process segments with URG.
1251	*/
1252	if ((tiflags & TH_URG) && ti->ti_urp &&
1253	TCPS_HAVERCVDFIN(tp->t_state) == 0) {
1254	/*
1255	* This is a kludge, but if we receive and accept
1256	* random urgent pointers, we'll crash in
1257	* soreceive. It's hard to imagine someone
1258	* actually wanting to send this much urgent data.
1259	*/
1260	if (ti->ti_urp + so->so_rcv.sb_cc > so->so_rcv.sb_datalen) {
1261	ti->ti_urp = 0;
1262	tiflags &= ~TH_URG;
1263	goto dodata;
1264	}
1265	/*
1266	* If this segment advances the known urgent pointer,
1267	* then mark the data stream. This should not happen
1268	* in CLOSE_WAIT, CLOSING, LAST_ACK or TIME_WAIT STATES since
1269	* a FIN has been received from the remote side.
1270	* In these states we ignore the URG.
1271	*
1272	* According to RFC961 (Assigned Protocols),
1273	* the urgent pointer points to the last octet
1274	* of urgent data. We continue, however,
1275	* to consider it to indicate the first octet
1276	* of data past the urgent section as the original
1277	* spec states (in one of two places).
1278	*/
1279	if (SEQ_GT(ti->ti_seq+ti->ti_urp, tp->rcv_up)) {
1280	tp->rcv_up = ti->ti_seq + ti->ti_urp;
1281	so->so_urgc = so->so_rcv.sb_cc +
1282	(tp->rcv_up - tp->rcv_nxt); /* -1; */
1283	tp->rcv_up = ti->ti_seq + ti->ti_urp;
1284
1285	}
1286	} else
1287	/*
1288	* If no out of band data is expected,
1289	* pull receive urgent pointer along
1290	* with the receive window.
1291	*/
1292	if (SEQ_GT(tp->rcv_nxt, tp->rcv_up))
1293	tp->rcv_up = tp->rcv_nxt;
1294	dodata:
1295
1296	/*
1297	* If this is a small packet, then ACK now - with Nagel
1298	* congestion avoidance sender won't send more until
1299	* he gets an ACK.
1300	*
1301	* See above.
1302	*/
1303	if (ti->ti_len && (unsigned)ti->ti_len <= 5 &&
1304	((struct tcpiphdr_2 *)ti)->first_char == (char)27) {
1305	tp->t_flags \|= TF_ACKNOW;
1306	}
1307
1308	/*
1309	* Process the segment text, merging it into the TCP sequencing queue,
1310	* and arranging for acknowledgment of receipt if necessary.
1311	* This process logically involves adjusting tp->rcv_wnd as data
1312	* is presented to the user (this happens in tcp_usrreq.c,
1313	* case PRU_RCVD). If a FIN has already been received on this
1314	* connection then we just ignore the text.
1315	*/
1316	if ((ti->ti_len \|\| (tiflags&TH_FIN)) &&
1317	TCPS_HAVERCVDFIN(tp->t_state) == 0) {
1318	if (ti->ti_seq == tp->rcv_nxt
1319	&& LIST_EMPTY(&tp->t_segq)
1320	&& tp->t_state == TCPS_ESTABLISHED) {
1321	DELAY_ACK(tp, ti); /* little bit different from BSD declaration see netinet/tcp_input.c */
1322	tp->rcv_nxt += tlen;
1323	tiflags = ti->ti_t.th_flags & TH_FIN;
1324	tcpstat.tcps_rcvpack++;
1325	tcpstat.tcps_rcvbyte += tlen;
1326	if (so->so_state & SS_FCANTRCVMORE)
1327	m_freem(pData, m);
1328	else
1329	sbappend(pData, so, m);
1330	}
1331	else {
1332	tiflags = tcp_reass(pData, tp, &ti->ti_t, &tlen, m);
1333	tiflags \|= TF_ACKNOW;
1334	}
1335	/*
1336	* Note the amount of data that peer has sent into
1337	* our window, in order to estimate the sender's
1338	* buffer size.
1339	*/
1340	len = so->so_rcv.sb_datalen - (tp->rcv_adv - tp->rcv_nxt);
1341	} else {
1342	m_free(pData, m);
1343	tiflags &= ~TH_FIN;
1344	}
1345
1346	/*
1347	* If FIN is received ACK the FIN and let the user know
1348	* that the connection is closing.
1349	*/
1350	if (tiflags & TH_FIN) {
1351	if (TCPS_HAVERCVDFIN(tp->t_state) == 0) {
1352	/*
1353	* If we receive a FIN we can't send more data,
1354	* set it SS_FDRAIN
1355	* Shutdown the socket if there is no rx data in the
1356	* buffer.
1357	* soread() is called on completion of shutdown() and
1358	* will got to TCPS_LAST_ACK, and use tcp_output()
1359	* to send the FIN.
1360	*/
1361	/* sofcantrcvmore(so); */
1362	sofwdrain(so);
1363
1364	tp->t_flags \|= TF_ACKNOW;
1365	tp->rcv_nxt++;
1366	}
1367	switch (tp->t_state) {
1368
1369	/*
1370	* In SYN_RECEIVED and ESTABLISHED STATES
1371	* enter the CLOSE_WAIT state.
1372	*/
1373	case TCPS_SYN_RECEIVED:
1374	case TCPS_ESTABLISHED:
1375	if(so->so_emu == EMU_CTL) /* no shutdown on socket */
1376	tp->t_state = TCPS_LAST_ACK;
1377	else
1378	tp->t_state = TCPS_CLOSE_WAIT;
1379	break;
1380
1381	/*
1382	* If still in FIN_WAIT_1 STATE FIN has not been acked so
1383	* enter the CLOSING state.
1384	*/
1385	case TCPS_FIN_WAIT_1:
1386	tp->t_state = TCPS_CLOSING;
1387	break;
1388
1389	/*
1390	* In FIN_WAIT_2 state enter the TIME_WAIT state,
1391	* starting the time-wait timer, turning off the other
1392	* standard timers.
1393	*/
1394	case TCPS_FIN_WAIT_2:
1395	tp->t_state = TCPS_TIME_WAIT;
1396	tcp_canceltimers(tp);
1397	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1398	soisfdisconnected(so);
1399	break;
1400
1401	/*
1402	* In TIME_WAIT state restart the 2 MSL time_wait timer.
1403	*/
1404	case TCPS_TIME_WAIT:
1405	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1406	break;
1407	}
1408	}
1409
1410	/*
1411	* Return any desired output.
1412	*/
1413	if (needoutput \|\| (tp->t_flags & TF_ACKNOW)) {
1414	(void) tcp_output(pData, tp);
1415	}
1416	return;
1417
1418	dropafterack:
1419	/*
1420	* Generate an ACK dropping incoming segment if it occupies
1421	* sequence space, where the ACK reflects our state.
1422	*/
1423	if (tiflags & TH_RST)
1424	goto drop;
1425	m_freem(pData, m);
1426	tp->t_flags \|= TF_ACKNOW;
1427	(void) tcp_output(pData, tp);
1428	return;
1429
1430	dropwithreset:
1431	/* reuses m if m!=NULL, m_free() unnecessary */
1432	if (tiflags & TH_ACK)
1433	tcp_respond(pData, tp, ti, m, (tcp_seq)0, ti->ti_ack, TH_RST);
1434	else {
1435	if (tiflags & TH_SYN) ti->ti_len++;
1436	tcp_respond(pData, tp, ti, m, ti->ti_seq+ti->ti_len, (tcp_seq)0,
1437	TH_RST\|TH_ACK);
1438	}
1439
1440	return;
1441
1442	drop:
1443	/*
1444	* Drop space held by incoming segment and return.
1445	*/
1446	m_free(pData, m);
1447
1448	return;
1449	}
1450
1451	/* , ts_present, ts_val, ts_ecr) */
1452	/* int *ts_present;
1453	* u_int32_t ts_val, ts_ecr;
1454	*/
1455	void
1456	tcp_dooptions(PNATState pData, struct tcpcb tp, u_char cp, int cnt, struct tcpiphdr *ti)
1457	{
1458	u_int16_t mss;
1459	int opt, optlen;
1460
1461	DEBUG_CALL("tcp_dooptions");
1462	DEBUG_ARGS((dfd," tp = %lx cnt=%i \n", (long )tp, cnt));
1463
1464	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1465	opt = cp[0];
1466	if (opt == TCPOPT_EOL)
1467	break;
1468	if (opt == TCPOPT_NOP)
1469	optlen = 1;
1470	else {
1471	optlen = cp[1];
1472	if (optlen <= 0)
1473	break;
1474	}
1475	switch (opt) {
1476
1477	default:
1478	continue;
1479
1480	case TCPOPT_MAXSEG:
1481	if (optlen != TCPOLEN_MAXSEG)
1482	continue;
1483	if (!(ti->ti_flags & TH_SYN))
1484	continue;
1485	memcpy((char ) &mss, (char ) cp + 2, sizeof(mss));
1486	NTOHS(mss);
1487	(void) tcp_mss(pData, tp, mss); /* sets t_maxseg */
1488	break;
1489
1490	/* case TCPOPT_WINDOW:
1491	* if (optlen != TCPOLEN_WINDOW)
1492	* continue;
1493	* if (!(ti->ti_flags & TH_SYN))
1494	* continue;
1495	* tp->t_flags \|= TF_RCVD_SCALE;
1496	* tp->requested_s_scale = min(cp[2], TCP_MAX_WINSHIFT);
1497	* break;
1498	*/
1499	/* case TCPOPT_TIMESTAMP:
1500	* if (optlen != TCPOLEN_TIMESTAMP)
1501	* continue;
1502	* *ts_present = 1;
1503	* memcpy((char ) ts_val, (char )cp + 2, sizeof(*ts_val));
1504	* NTOHL(*ts_val);
1505	* memcpy((char ) ts_ecr, (char )cp + 6, sizeof(*ts_ecr));
1506	* NTOHL(*ts_ecr);
1507	*
1508	/ /
1509	* * A timestamp received in a SYN makes
1510	* * it ok to send timestamp requests and replies.
1511	* */
1512	/* if (ti->ti_flags & TH_SYN) {
1513	* tp->t_flags \|= TF_RCVD_TSTMP;
1514	* tp->ts_recent = *ts_val;
1515	* tp->ts_recent_age = tcp_now;
1516	* }
1517	*/ break;
1518	}
1519	}
1520	}
1521
1522
1523	/*
1524	* Pull out of band byte out of a segment so
1525	* it doesn't appear in the user's data queue.
1526	* It is still reflected in the segment length for
1527	* sequencing purposes.
1528	*/
1529
1530	#ifdef notdef
1531
1532	void
1533	tcp_pulloutofband(so, ti, m)
1534	struct socket *so;
1535	struct tcpiphdr *ti;
1536	register struct mbuf *m;
1537	{
1538	int cnt = ti->ti_urp - 1;
1539
1540	while (cnt >= 0) {
1541	if (m->m_len > cnt) {
1542	char *cp = mtod(m, caddr_t) + cnt;
1543	struct tcpcb *tp = sototcpcb(so);
1544
1545	tp->t_iobc = *cp;
1546	tp->t_oobflags \|= TCPOOB_HAVEDATA;
1547	memcpy(sp, cp+1, (unsigned)(m->m_len - cnt - 1));
1548	m->m_len--;
1549	return;
1550	}
1551	cnt -= m->m_len;
1552	m = m->m_next; /* XXX WRONG! Fix it! */
1553	if (m == 0)
1554	break;
1555	}
1556	panic("tcp_pulloutofband");
1557	}
1558
1559	#endif /* notdef */
1560
1561	/*
1562	* Collect new round-trip time estimate
1563	* and update averages and current timeout.
1564	*/
1565
1566	void
1567	tcp_xmit_timer(PNATState pData, register struct tcpcb *tp, int rtt)
1568	{
1569	register short delta;
1570
1571	DEBUG_CALL("tcp_xmit_timer");
1572	DEBUG_ARG("tp = %lx", (long)tp);
1573	DEBUG_ARG("rtt = %d", rtt);
1574
1575	tcpstat.tcps_rttupdated++;
1576	if (tp->t_srtt != 0) {
1577	/*
1578	* srtt is stored as fixed point with 3 bits after the
1579	* binary point (i.e., scaled by 8). The following magic
1580	* is equivalent to the smoothing algorithm in rfc793 with
1581	* an alpha of .875 (srtt = rtt/8 + srtt*7/8 in fixed
1582	* point). Adjust rtt to origin 0.
1583	*/
1584	delta = rtt - 1 - (tp->t_srtt >> TCP_RTT_SHIFT);
1585	if ((tp->t_srtt += delta) <= 0)
1586	tp->t_srtt = 1;
1587	/*
1588	* We accumulate a smoothed rtt variance (actually, a
1589	* smoothed mean difference), then set the retransmit
1590	* timer to smoothed rtt + 4 times the smoothed variance.
1591	* rttvar is stored as fixed point with 2 bits after the
1592	* binary point (scaled by 4). The following is
1593	* equivalent to rfc793 smoothing with an alpha of .75
1594	* (rttvar = rttvar*3/4 + \|delta\| / 4). This replaces
1595	* rfc793's wired-in beta.
1596	*/
1597	if (delta < 0)
1598	delta = -delta;
1599	delta -= (tp->t_rttvar >> TCP_RTTVAR_SHIFT);
1600	if ((tp->t_rttvar += delta) <= 0)
1601	tp->t_rttvar = 1;
1602	} else {
1603	/*
1604	* No rtt measurement yet - use the unsmoothed rtt.
1605	* Set the variance to half the rtt (so our first
1606	* retransmit happens at 3*rtt).
1607	*/
1608	tp->t_srtt = rtt << TCP_RTT_SHIFT;
1609	tp->t_rttvar = rtt << (TCP_RTTVAR_SHIFT - 1);
1610	}
1611	tp->t_rtt = 0;
1612	tp->t_rxtshift = 0;
1613
1614	/*
1615	* the retransmit should happen at rtt + 4 * rttvar.
1616	* Because of the way we do the smoothing, srtt and rttvar
1617	* will each average +1/2 tick of bias. When we compute
1618	* the retransmit timer, we want 1/2 tick of rounding and
1619	* 1 extra tick because of +-1/2 tick uncertainty in the
1620	* firing of the timer. The bias will give us exactly the
1621	* 1.5 tick we need. But, because the bias is
1622	* statistical, we have to test that we don't drop below
1623	* the minimum feasible timer (which is 2 ticks).
1624	*/
1625	TCPT_RANGESET(tp->t_rxtcur, TCP_REXMTVAL(tp),
1626	(short)tp->t_rttmin, TCPTV_REXMTMAX); /* XXX */
1627
1628	/*
1629	* We received an ack for a packet that wasn't retransmitted;
1630	* it is probably safe to discard any error indications we've
1631	* received recently. This isn't quite right, but close enough
1632	* for now (a route might have failed after we sent a segment,
1633	* and the return path might not be symmetrical).
1634	*/
1635	tp->t_softerror = 0;
1636	}
1637
1638	/*
1639	* Determine a reasonable value for maxseg size.
1640	* If the route is known, check route for mtu.
1641	* If none, use an mss that can be handled on the outgoing
1642	* interface without forcing IP to fragment; if bigger than
1643	* an mbuf cluster (MCLBYTES), round down to nearest multiple of MCLBYTES
1644	* to utilize large mbufs. If no route is found, route has no mtu,
1645	* or the destination isn't local, use a default, hopefully conservative
1646	* size (usually 512 or the default IP max size, but no more than the mtu
1647	* of the interface), as we can't discover anything about intervening
1648	* gateways or networks. We also initialize the congestion/slow start
1649	* window to be a single segment if the destination isn't local.
1650	* While looking at the routing entry, we also initialize other path-dependent
1651	* parameters from pre-set or cached values in the routing entry.
1652	*/
1653
1654	int
1655	tcp_mss(PNATState pData, register struct tcpcb *tp, u_int offer)
1656	{
1657	struct socket *so = tp->t_socket;
1658	int mss;
1659
1660	DEBUG_CALL("tcp_mss");
1661	DEBUG_ARG("tp = %lx", (long)tp);
1662	DEBUG_ARG("offer = %d", offer);
1663
1664	mss = min(if_mtu, if_mru) - sizeof(struct tcpiphdr);
1665	if (offer)
1666	mss = min(mss, offer);
1667	mss = max(mss, 32);
1668	if (mss < tp->t_maxseg \|\| offer != 0)
1669	tp->t_maxseg = mss;
1670
1671	tp->snd_cwnd = mss;
1672
1673	sbreserve(&so->so_snd, tcp_sndspace+((tcp_sndspace%mss)?(mss-(tcp_sndspace%mss)):0));
1674	sbreserve(&so->so_rcv, tcp_rcvspace+((tcp_rcvspace%mss)?(mss-(tcp_rcvspace%mss)):0));
1675
1676	DEBUG_MISC((dfd, " returning mss = %d\n", mss));
1677
1678	return mss;
1679	}

Note: See TracBrowser for help on using the repository browser.

source: vbox/trunk/src/VBox/Devices/Network/slirp/tcp_input.c@ 15458

Download in other formats: