tcp_input.c@ 14529

Last change on this file since 14529 was 14476, checked in by vboxsync, 16 years ago
slirp: more dead code
Property svn:eol-style set to `native`
File size: 73.5 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 VBOX_WITH_BSD_TCP_REASS
56	/*
57	* Insert segment ti into reassembly queue of tcp with
58	* control block tp. Return TH_FIN if reassembly now includes
59	* a segment with FIN. The macro form does the common case inline
60	* (segment is the next to be received on an established connection,
61	* and the queue is empty), avoiding linkage into and removal
62	* from the queue and repetition of various conversions.
63	* Set DELACK for segments received in order, but ack immediately
64	* when segments are out of order (so fast retransmit can work).
65	*/
66	#ifdef TCP_ACK_HACK
67	#define TCP_REASS(pData, tp, ti, m, so, flags) {\
68	if ((ti)->ti_seq == (tp)->rcv_nxt && \
69	u32_to_ptr((pData), (tp)->seg_next, struct tcpcb *) == (tp) && \
70	(tp)->t_state == TCPS_ESTABLISHED) {\
71	if (ti->ti_flags & TH_PUSH) \
72	tp->t_flags \|= TF_ACKNOW; \
73	else \
74	tp->t_flags \|= TF_DELACK; \
75	(tp)->rcv_nxt += (ti)->ti_len; \
76	flags = (ti)->ti_flags & TH_FIN; \
77	tcpstat.tcps_rcvpack++;\
78	tcpstat.tcps_rcvbyte += (ti)->ti_len;\
79	if (so->so_emu) { \
80	if (tcp_emu((pData), (so),(m))) sbappend((pData), (so), (m)); \
81	} else \
82	sbappend((pData), (so), (m)); \
83	/* sorwakeup(so); */ \
84	} else {\
85	(flags) = tcp_reass((pData), (tp), (ti), (m)); \
86	tp->t_flags \|= TF_ACKNOW; \
87	} \
88	}
89	#else
90	#define TCP_REASS(pData, tp, ti, m, so, flags) { \
91	if ((ti)->ti_seq == (tp)->rcv_nxt && \
92	u32_to_ptr((pData), (tp)->seg_next, struct tcpcb *) == (tp) && \
93	(tp)->t_state == TCPS_ESTABLISHED) { \
94	tp->t_flags \|= TF_DELACK; \
95	(tp)->rcv_nxt += (ti)->ti_len; \
96	flags = (ti)->ti_flags & TH_FIN; \
97	tcpstat.tcps_rcvpack++;\
98	tcpstat.tcps_rcvbyte += (ti)->ti_len;\
99	if (so->so_emu) { \
100	if (tcp_emu((pData), (so),(m))) sbappend((pData), (so), (m)); \
101	} else \
102	sbappend((pData), (so), (m)); \
103	/* sorwakeup(so); */ \
104	} else { \
105	(flags) = tcp_reass((pData), (tp), (ti), (m)); \
106	tp->t_flags \|= TF_ACKNOW; \
107	} \
108	}
109	#endif
110
111	int
112	tcp_reass(PNATState pData, register struct tcpcb tp, register struct tcpiphdr ti, struct mbuf *m)
113	{
114	register struct tcpiphdr *q;
115	struct socket *so = tp->t_socket;
116	int flags;
117
118	/*
119	* Call with ti==0 after become established to
120	* force pre-ESTABLISHED data up to user socket.
121	*/
122	if (ti == 0)
123	goto present;
124
125	/*
126	* Find a segment which begins after this one does.
127	*/
128	for (q = u32_to_ptr(pData, tp->seg_next, struct tcpiphdr ); q != (struct tcpiphdr )tp;
129	q = u32_to_ptr(pData, q->ti_next, struct tcpiphdr *))
130	if (SEQ_GT(q->ti_seq, ti->ti_seq))
131	break;
132
133	/*
134	* If there is a preceding segment, it may provide some of
135	* our data already. If so, drop the data from the incoming
136	* segment. If it provides all of our data, drop us.
137	*/
138	if (u32_to_ptr(pData, q->ti_prev, struct tcpiphdr ) != (struct tcpiphdr )tp) {
139	register int i;
140	q = u32_to_ptr(pData, q->ti_prev, struct tcpiphdr *);
141	/* conversion to int (in i) handles seq wraparound */
142	i = q->ti_seq + q->ti_len - ti->ti_seq;
143	if (i > 0) {
144	if (i >= ti->ti_len) {
145	tcpstat.tcps_rcvduppack++;
146	tcpstat.tcps_rcvdupbyte += ti->ti_len;
147	m_freem(pData, m);
148	/*
149	* Try to present any queued data
150	* at the left window edge to the user.
151	* This is needed after the 3-WHS
152	* completes.
153	*/
154	goto present; /* ??? */
155	}
156	m_adj(m, i);
157	ti->ti_len -= i;
158	ti->ti_seq += i;
159	}
160	q = u32_to_ptr(pData, q->ti_next, struct tcpiphdr *);
161	}
162	tcpstat.tcps_rcvoopack++;
163	tcpstat.tcps_rcvoobyte += ti->ti_len;
164	REASS_MBUF_SET(ti, m); /* XXX */
165
166	/*
167	* While we overlap succeeding segments trim them or,
168	* if they are completely covered, dequeue them.
169	*/
170	while (q != (struct tcpiphdr *)tp) {
171	register int i = (ti->ti_seq + ti->ti_len) - q->ti_seq;
172	if (i <= 0)
173	break;
174	if (i < q->ti_len) {
175	q->ti_seq += i;
176	q->ti_len -= i;
177	m_adj(REASS_MBUF_GET(q), i);
178	break;
179	}
180	q = u32_to_ptr(pData, q->ti_next, struct tcpiphdr *);
181	m = REASS_MBUF_GET(u32_to_ptr(pData, q->ti_prev, struct tcpiphdr *));
182	remque_32(pData, u32_to_ptr(pData, q->ti_prev, struct tcpiphdr *));
183	m_freem(pData, m);
184	}
185
186	/*
187	* Stick new segment in its place.
188	*/
189	insque_32(pData, ti, u32_to_ptr(pData, q->ti_prev, struct tcpiphdr *));
190
191	present:
192	/*
193	* Present data to user, advancing rcv_nxt through
194	* completed sequence space.
195	*/
196	if (!TCPS_HAVEESTABLISHED(tp->t_state))
197	return (0);
198	ti = u32_to_ptr(pData, tp->seg_next, struct tcpiphdr *);
199	if (ti == (struct tcpiphdr *)tp \|\| ti->ti_seq != tp->rcv_nxt)
200	return (0);
201	if (tp->t_state == TCPS_SYN_RECEIVED && ti->ti_len)
202	return (0);
203	do {
204	tp->rcv_nxt += ti->ti_len;
205	flags = ti->ti_flags & TH_FIN;
206	remque_32(pData, ti);
207	m = REASS_MBUF_GET(ti); /* XXX */
208	ti = u32_to_ptr(pData, ti->ti_next, struct tcpiphdr *);
209	/* if (so->so_state & SS_FCANTRCVMORE) */
210	if (so->so_state & SS_FCANTSENDMORE)
211	m_freem(pData, m);
212	else {
213	if (so->so_emu) {
214	if (tcp_emu(pData, so,m)) sbappend(pData, so, m);
215	} else
216	sbappend(pData, so, m);
217	}
218	} while (ti != (struct tcpiphdr *)tp && ti->ti_seq == tp->rcv_nxt);
219	/* sorwakeup(so); */
220	return (flags);
221	}
222
223	#else /* VBOX_WITH_BSD_TCP_REASS */
224
225	#ifndef TCP_ACK_HACK
226	#define DELAY_ACK(tp, ti) \
227	if (ti->ti_flags & TH_PUSH) \
228	tp->t_flags \|= TF_ACKNOW; \
229	else \
230	tp->t_flags \|= TF_DELACK;
231	#else /* !TCP_ACK_HACK */
232	#define DELAY_ACK(tp, ign) \
233	tp->t_flags \|= TF_DELACK;
234	#endif /* TCP_ACK_HACK */
235
236
237	/*
238	* deps: netinet/tcp_reass.c
239	* tcp_reass_maxqlen = 48 (deafault)
240	* tcp_reass_maxseg = nmbclusters/16 (nmbclusters = 1024 + maxusers * 64 from kern/kern_mbuf.c let's say 256)
241	*/
242	int
243	tcp_reass(PNATState pData, struct tcpcb tp, struct tcphdr th, int tlenp, struct mbuf m)
244	{
245	struct tseg_qent *q;
246	struct tseg_qent *p = NULL;
247	struct tseg_qent *nq;
248	struct tseg_qent *te = NULL;
249	struct socket *so = tp->t_socket;
250	int flags;
251
252	/*
253	* XXX: tcp_reass() is rather inefficient with its data structures
254	* and should be rewritten (see NetBSD for optimizations). While
255	* doing that it should move to its own file tcp_reass.c.
256	*/
257
258	/*
259	* Call with th==NULL after become established to
260	* force pre-ESTABLISHED data up to user socket.
261	*/
262	if (th == NULL)
263	goto present;
264
265	/*
266	* Limit the number of segments in the reassembly queue to prevent
267	* holding on to too many segments (and thus running out of mbufs).
268	* Make sure to let the missing segment through which caused this
269	* queue. Always keep one global queue entry spare to be able to
270	* process the missing segment.
271	*/
272	if (th->th_seq != tp->rcv_nxt &&
273	(tcp_reass_qsize + 1 >= tcp_reass_maxseg \|\|
274	tp->t_segqlen >= tcp_reass_maxqlen)) {
275	tcp_reass_overflows++;
276	tcpstat.tcps_rcvmemdrop++;
277	m_freem(pData, m);
278	*tlenp = 0;
279	return (0);
280	}
281
282	/*
283	* Allocate a new queue entry. If we can't, or hit the zone limit
284	* just drop the pkt.
285	*/
286	te = malloc(sizeof(struct tseg_qent));
287	if (te == NULL) {
288	tcpstat.tcps_rcvmemdrop++;
289	m_freem(pData, m);
290	*tlenp = 0;
291	return (0);
292	}
293	tp->t_segqlen++;
294	tcp_reass_qsize++;
295
296	/*
297	* Find a segment which begins after this one does.
298	*/
299	LIST_FOREACH(q, &tp->t_segq, tqe_q) {
300	if (SEQ_GT(q->tqe_th->th_seq, th->th_seq))
301	break;
302	p = q;
303	}
304
305	/*
306	* If there is a preceding segment, it may provide some of
307	* our data already. If so, drop the data from the incoming
308	* segment. If it provides all of our data, drop us.
309	*/
310	if (p != NULL) {
311	int i;
312	/* conversion to int (in i) handles seq wraparound */
313	i = p->tqe_th->th_seq + p->tqe_len - th->th_seq;
314	if (i > 0) {
315	if (i >= *tlenp) {
316	tcpstat.tcps_rcvduppack++;
317	tcpstat.tcps_rcvdupbyte += *tlenp;
318	m_freem(pData, m);
319	free(te);
320	tp->t_segqlen--;
321	tcp_reass_qsize--;
322	/*
323	* Try to present any queued data
324	* at the left window edge to the user.
325	* This is needed after the 3-WHS
326	* completes.
327	*/
328	goto present; /* ??? */
329	}
330	m_adj(m, i);
331	*tlenp -= i;
332	th->th_seq += i;
333	}
334	}
335	tcpstat.tcps_rcvoopack++;
336	tcpstat.tcps_rcvoobyte += *tlenp;
337
338	/*
339	* While we overlap succeeding segments trim them or,
340	* if they are completely covered, dequeue them.
341	*/
342	while (q) {
343	int i = (th->th_seq + *tlenp) - q->tqe_th->th_seq;
344	if (i <= 0)
345	break;
346	if (i < q->tqe_len) {
347	q->tqe_th->th_seq += i;
348	q->tqe_len -= i;
349	m_adj(q->tqe_m, i);
350	break;
351	}
352
353	nq = LIST_NEXT(q, tqe_q);
354	LIST_REMOVE(q, tqe_q);
355	m_freem(pData, q->tqe_m);
356	free(q);
357	tp->t_segqlen--;
358	tcp_reass_qsize--;
359	q = nq;
360	}
361
362	/* Insert the new segment queue entry into place. */
363	te->tqe_m = m;
364	te->tqe_th = th;
365	te->tqe_len = *tlenp;
366
367	if (p == NULL) {
368	LIST_INSERT_HEAD(&tp->t_segq, te, tqe_q);
369	} else {
370	LIST_INSERT_AFTER(p, te, tqe_q);
371	}
372
373	present:
374	/*
375	* Present data to user, advancing rcv_nxt through
376	* completed sequence space.
377	*/
378	if (!TCPS_HAVEESTABLISHED(tp->t_state))
379	return (0);
380	q = LIST_FIRST(&tp->t_segq);
381	if (!q \|\| q->tqe_th->th_seq != tp->rcv_nxt)
382	return (0);
383	do {
384	tp->rcv_nxt += q->tqe_len;
385	flags = q->tqe_th->th_flags & TH_FIN;
386	nq = LIST_NEXT(q, tqe_q);
387	LIST_REMOVE(q, tqe_q);
388	/* XXX: This place should be checked for the same code in
389	* original BSD code for Slirp and current BSD used SS_FCANTRCVMORE
390	*/
391	if (so->so_state & SS_FCANTSENDMORE)
392	m_freem(pData, q->tqe_m);
393	else
394	sbappend(pData, so, q->tqe_m);
395	free(q);
396	tp->t_segqlen--;
397	tcp_reass_qsize--;
398	q = nq;
399	} while (q && q->tqe_th->th_seq == tp->rcv_nxt);
400	return (flags);
401	}
402	#endif /* VBOX_WITH_BSD_TCP_REASS */
403
404	/*
405	* TCP input routine, follows pages 65-76 of the
406	* protocol specification dated September, 1981 very closely.
407	*/
408	void
409	tcp_input(PNATState pData, register struct mbuf m, int iphlen, struct socket inso)
410	{
411	struct ip save_ip, *ip;
412	register struct tcpiphdr *ti;
413	caddr_t optp = NULL;
414	int optlen = 0;
415	int len, tlen, off;
416	register struct tcpcb *tp = 0;
417	register int tiflags;
418	struct socket *so = 0;
419	int todrop, acked, ourfinisacked, needoutput = 0;
420	/* int dropsocket = 0; */
421	int iss = 0;
422	u_long tiwin;
423	/* int ts_present = 0; */
424	int mbuf_freed = 0;
425
426	DEBUG_CALL("tcp_input");
427	DEBUG_ARGS((dfd," m = %8lx iphlen = %2d inso = %lx\n",
428	(long )m, iphlen, (long )inso ));
429
430	/*
431	* If called with m == 0, then we're continuing the connect
432	*/
433	if (m == NULL) {
434	so = inso;
435
436	/* Re-set a few variables */
437	tp = sototcpcb(so);
438	m = so->so_m;
439	so->so_m = 0;
440	ti = so->so_ti;
441	tiwin = ti->ti_win;
442	tiflags = ti->ti_flags;
443
444	goto cont_conn;
445	}
446
447
448	tcpstat.tcps_rcvtotal++;
449	/*
450	* Get IP and TCP header together in first mbuf.
451	* Note: IP leaves IP header in first mbuf.
452	*/
453	ti = mtod(m, struct tcpiphdr *);
454	if (iphlen > sizeof(struct ip )) {
455	ip_stripoptions(m, (struct mbuf *)0);
456	iphlen=sizeof(struct ip );
457	}
458	/* XXX Check if too short */
459
460
461	/*
462	* Save a copy of the IP header in case we want restore it
463	* for sending an ICMP error message in response.
464	*/
465	ip=mtod(m, struct ip *);
466	save_ip = *ip;
467	save_ip.ip_len+= iphlen;
468
469	/*
470	* Checksum extended TCP header and data.
471	*/
472	tlen = ((struct ip *)ti)->ip_len;
473	ti->ti_next = ti->ti_prev = 0;
474	ti->ti_x1 = 0;
475	ti->ti_len = htons((u_int16_t)tlen);
476	len = sizeof(struct ip ) + tlen;
477	/* keep checksum for ICMP reply
478	* ti->ti_sum = cksum(m, len);
479	* if (ti->ti_sum) { */
480	if(cksum(m, len)) {
481	tcpstat.tcps_rcvbadsum++;
482	goto drop;
483	}
484
485	/*
486	* Check that TCP offset makes sense,
487	* pull out TCP options and adjust length. XXX
488	*/
489	off = ti->ti_off << 2;
490	if (off < sizeof (struct tcphdr) \|\| off > tlen) {
491	tcpstat.tcps_rcvbadoff++;
492	goto drop;
493	}
494	tlen -= off;
495	ti->ti_len = tlen;
496	if (off > sizeof (struct tcphdr)) {
497	optlen = off - sizeof (struct tcphdr);
498	optp = mtod(m, caddr_t) + sizeof (struct tcpiphdr);
499
500	/*
501	* Do quick retrieval of timestamp options ("options
502	* prediction?"). If timestamp is the only option and it's
503	* formatted as recommended in RFC 1323 appendix A, we
504	* quickly get the values now and not bother calling
505	* tcp_dooptions(), etc.
506	*/
507	/* if ((optlen == TCPOLEN_TSTAMP_APPA \|\|
508	* (optlen > TCPOLEN_TSTAMP_APPA &&
509	* optp[TCPOLEN_TSTAMP_APPA] == TCPOPT_EOL)) &&
510	* (u_int32_t )optp == htonl(TCPOPT_TSTAMP_HDR) &&
511	* (ti->ti_flags & TH_SYN) == 0) {
512	* ts_present = 1;
513	* ts_val = ntohl((u_int32_t )(optp + 4));
514	* ts_ecr = ntohl((u_int32_t )(optp + 8));
515	* optp = NULL; / * we've parsed the options * /
516	* }
517	*/
518	}
519	tiflags = ti->ti_flags;
520
521	/*
522	* Convert TCP protocol specific fields to host format.
523	*/
524	NTOHL(ti->ti_seq);
525	NTOHL(ti->ti_ack);
526	NTOHS(ti->ti_win);
527	NTOHS(ti->ti_urp);
528
529	/*
530	* Drop TCP, IP headers and TCP options.
531	*/
532	m->m_data += sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
533	m->m_len -= sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
534
535	/*
536	* Locate pcb for segment.
537	*/
538	findso:
539	so = tcp_last_so;
540	if (so->so_fport != ti->ti_dport \|\|
541	so->so_lport != ti->ti_sport \|\|
542	so->so_laddr.s_addr != ti->ti_src.s_addr \|\|
543	so->so_faddr.s_addr != ti->ti_dst.s_addr) {
544	so = solookup(&tcb, ti->ti_src, ti->ti_sport,
545	ti->ti_dst, ti->ti_dport);
546	if (so)
547	tcp_last_so = so;
548	++tcpstat.tcps_socachemiss;
549	}
550
551	/*
552	* If the state is CLOSED (i.e., TCB does not exist) then
553	* all data in the incoming segment is discarded.
554	* If the TCB exists but is in CLOSED state, it is embryonic,
555	* but should either do a listen or a connect soon.
556	*
557	* state == CLOSED means we've done socreate() but haven't
558	* attached it to a protocol yet...
559	*
560	* XXX If a TCB does not exist, and the TH_SYN flag is
561	* the only flag set, then create a session, mark it
562	* as if it was LISTENING, and continue...
563	*/
564	if (so == 0) {
565	if ((tiflags & (TH_SYN\|TH_FIN\|TH_RST\|TH_URG\|TH_ACK)) != TH_SYN)
566	goto dropwithreset;
567
568	if ((so = socreate()) == NULL)
569	goto dropwithreset;
570	if (tcp_attach(pData, so) < 0) {
571	free(so); /* Not sofree (if it failed, it's not insqued) */
572	goto dropwithreset;
573	}
574
575	sbreserve(&so->so_snd, tcp_sndspace);
576	sbreserve(&so->so_rcv, tcp_rcvspace);
577
578	/* tcp_last_so = so; / / XXX ? */
579	/* tp = sototcpcb(so); */
580
581	so->so_laddr = ti->ti_src;
582	so->so_lport = ti->ti_sport;
583	so->so_faddr = ti->ti_dst;
584	so->so_fport = ti->ti_dport;
585
586	if ((so->so_iptos = tcp_tos(so)) == 0)
587	so->so_iptos = ((struct ip *)ti)->ip_tos;
588
589	tp = sototcpcb(so);
590	tp->t_state = TCPS_LISTEN;
591	}
592
593	/*
594	* If this is a still-connecting socket, this probably
595	* a retransmit of the SYN. Whether it's a retransmit SYN
596	* or something else, we nuke it.
597	*/
598	if (so->so_state & SS_ISFCONNECTING)
599	goto drop;
600
601	tp = sototcpcb(so);
602
603	/* XXX Should never fail */
604	if (tp == 0)
605	goto dropwithreset;
606	if (tp->t_state == TCPS_CLOSED)
607	goto drop;
608
609	/* Unscale the window into a 32-bit value. */
610	/* if ((tiflags & TH_SYN) == 0)
611	* tiwin = ti->ti_win << tp->snd_scale;
612	* else
613	*/
614	tiwin = ti->ti_win;
615
616	/*
617	* Segment received on connection.
618	* Reset idle time and keep-alive timer.
619	*/
620	tp->t_idle = 0;
621	if (so_options)
622	tp->t_timer[TCPT_KEEP] = tcp_keepintvl;
623	else
624	tp->t_timer[TCPT_KEEP] = tcp_keepidle;
625
626	/*
627	* Process options if not in LISTEN state,
628	* else do it below (after getting remote address).
629	*/
630	if (optp && tp->t_state != TCPS_LISTEN)
631	tcp_dooptions(pData, tp, (u_char *)optp, optlen, ti);
632	/* , */
633	/* &ts_present, &ts_val, &ts_ecr); */
634
635	/*
636	* Header prediction: check for the two common cases
637	* of a uni-directional data xfer. If the packet has
638	* no control flags, is in-sequence, the window didn't
639	* change and we're not retransmitting, it's a
640	* candidate. If the length is zero and the ack moved
641	* forward, we're the sender side of the xfer. Just
642	* free the data acked & wake any higher level process
643	* that was blocked waiting for space. If the length
644	* is non-zero and the ack didn't move, we're the
645	* receiver side. If we're getting packets in-order
646	* (the reassembly queue is empty), add the data to
647	* the socket buffer and note that we need a delayed ack.
648	*
649	* XXX Some of these tests are not needed
650	* eg: the tiwin == tp->snd_wnd prevents many more
651	* predictions.. with no real advantage..
652	*/
653	if (tp->t_state == TCPS_ESTABLISHED &&
654	(tiflags & (TH_SYN\|TH_FIN\|TH_RST\|TH_URG\|TH_ACK)) == TH_ACK &&
655	/* (!ts_present \|\| TSTMP_GEQ(ts_val, tp->ts_recent)) && */
656	ti->ti_seq == tp->rcv_nxt &&
657	tiwin && tiwin == tp->snd_wnd &&
658	tp->snd_nxt == tp->snd_max) {
659	/*
660	* If last ACK falls within this segment's sequence numbers,
661	* record the timestamp.
662	*/
663	/* if (ts_present && SEQ_LEQ(ti->ti_seq, tp->last_ack_sent) &&
664	* SEQ_LT(tp->last_ack_sent, ti->ti_seq + ti->ti_len)) {
665	* tp->ts_recent_age = tcp_now;
666	* tp->ts_recent = ts_val;
667	* }
668	*/
669	if (ti->ti_len == 0) {
670	if (SEQ_GT(ti->ti_ack, tp->snd_una) &&
671	SEQ_LEQ(ti->ti_ack, tp->snd_max) &&
672	tp->snd_cwnd >= tp->snd_wnd) {
673	/*
674	* this is a pure ack for outstanding data.
675	*/
676	++tcpstat.tcps_predack;
677	/* if (ts_present)
678	* tcp_xmit_timer(tp, tcp_now-ts_ecr+1);
679	* else
680	*/ if (tp->t_rtt &&
681	SEQ_GT(ti->ti_ack, tp->t_rtseq))
682	tcp_xmit_timer(pData, tp, tp->t_rtt);
683	acked = ti->ti_ack - tp->snd_una;
684	tcpstat.tcps_rcvackpack++;
685	tcpstat.tcps_rcvackbyte += acked;
686	sbdrop(&so->so_snd, acked);
687	tp->snd_una = ti->ti_ack;
688	m_freem(pData, m);
689
690	/*
691	* If all outstanding data are acked, stop
692	* retransmit timer, otherwise restart timer
693	* using current (possibly backed-off) value.
694	* If process is waiting for space,
695	* wakeup/selwakeup/signal. If data
696	* are ready to send, let tcp_output
697	* decide between more output or persist.
698	*/
699	if (tp->snd_una == tp->snd_max)
700	tp->t_timer[TCPT_REXMT] = 0;
701	else if (tp->t_timer[TCPT_PERSIST] == 0)
702	tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
703
704	/*
705	* There's room in so_snd, sowwakup will read()
706	* from the socket if we can
707	*/
708	/* if (so->so_snd.sb_flags & SB_NOTIFY)
709	* sowwakeup(so);
710	*/
711	/*
712	* This is called because sowwakeup might have
713	* put data into so_snd. Since we don't so sowwakeup,
714	* we don't need this.. XXX???
715	*/
716	if (so->so_snd.sb_cc)
717	(void) tcp_output(pData, tp);
718
719	return;
720	}
721	} else if (ti->ti_ack == tp->snd_una &&
722	#ifndef VBOX_WITH_BSD_TCP_REASS
723	u32_to_ptr(pData, tp->seg_next, struct tcpcb *) == tp &&
724	#else /* VBOX_WITH_BSD_TCP_REASS */
725	LIST_FIRST(&tp->t_segq) &&
726	#endif /* VBOX_WITH_BSD_TCP_REASS */
727	ti->ti_len <= sbspace(&so->so_rcv)) {
728	/*
729	* this is a pure, in-sequence data packet
730	* with nothing on the reassembly queue and
731	* we have enough buffer space to take it.
732	*/
733	++tcpstat.tcps_preddat;
734	tp->rcv_nxt += ti->ti_len;
735	tcpstat.tcps_rcvpack++;
736	tcpstat.tcps_rcvbyte += ti->ti_len;
737	/*
738	* Add data to socket buffer.
739	*/
740	if (so->so_emu) {
741	if (tcp_emu(pData, so,m)) sbappend(pData, so, m);
742	} else
743	sbappend(pData, so, m);
744
745	/*
746	* XXX This is called when data arrives. Later, check
747	* if we can actually write() to the socket
748	* XXX Need to check? It's be NON_BLOCKING
749	*/
750	/* sorwakeup(so); */
751
752	/*
753	* If this is a short packet, then ACK now - with Nagel
754	* congestion avoidance sender won't send more until
755	* he gets an ACK.
756	*
757	* It is better to not delay acks at all to maximize
758	* TCP throughput. See RFC 2581.
759	*/
760	tp->t_flags \|= TF_ACKNOW;
761	tcp_output(pData, tp);
762	return;
763	}
764	} /* header prediction */
765	/*
766	* Calculate amount of space in receive window,
767	* and then do TCP input processing.
768	* Receive window is amount of space in rcv queue,
769	* but not less than advertised window.
770	*/
771	{ int win;
772	win = sbspace(&so->so_rcv);
773	if (win < 0)
774	win = 0;
775	tp->rcv_wnd = max(win, (int)(tp->rcv_adv - tp->rcv_nxt));
776	}
777
778	switch (tp->t_state) {
779
780	/*
781	* If the state is LISTEN then ignore segment if it contains an RST.
782	* If the segment contains an ACK then it is bad and send a RST.
783	* If it does not contain a SYN then it is not interesting; drop it.
784	* Don't bother responding if the destination was a broadcast.
785	* Otherwise initialize tp->rcv_nxt, and tp->irs, select an initial
786	* tp->iss, and send a segment:
787	* <SEQ=ISS><ACK=RCV_NXT><CTL=SYN,ACK>
788	* Also initialize tp->snd_nxt to tp->iss+1 and tp->snd_una to tp->iss.
789	* Fill in remote peer address fields if not previously specified.
790	* Enter SYN_RECEIVED state, and process any other fields of this
791	* segment in this state.
792	*/
793	case TCPS_LISTEN: {
794
795	if (tiflags & TH_RST)
796	goto drop;
797	if (tiflags & TH_ACK)
798	goto dropwithreset;
799	if ((tiflags & TH_SYN) == 0)
800	goto drop;
801
802	/*
803	* This has way too many gotos...
804	* But a bit of spaghetti code never hurt anybody :)
805	*/
806
807	if (so->so_emu & EMU_NOCONNECT) {
808	so->so_emu &= ~EMU_NOCONNECT;
809	goto cont_input;
810	}
811
812	if((tcp_fconnect(pData, so) == -1) && (errno != EINPROGRESS) && (errno != EWOULDBLOCK)) {
813	u_char code=ICMP_UNREACH_NET;
814	DEBUG_MISC((dfd," tcp fconnect errno = %d-%s\n",
815	errno,strerror(errno)));
816	if(errno == ECONNREFUSED) {
817	/* ACK the SYN, send RST to refuse the connection */
818	tcp_respond(pData, tp, ti, m, ti->ti_seq+1, (tcp_seq)0,
819	TH_RST\|TH_ACK);
820	} else {
821	if(errno == EHOSTUNREACH) code=ICMP_UNREACH_HOST;
822	HTONL(ti->ti_seq); /* restore tcp header */
823	HTONL(ti->ti_ack);
824	HTONS(ti->ti_win);
825	HTONS(ti->ti_urp);
826	m->m_data -= sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
827	m->m_len += sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
828	*ip=save_ip;
829	icmp_error(pData, m, ICMP_UNREACH,code, 0,strerror(errno));
830	}
831	tp = tcp_close(pData, tp);
832	m_free(pData, m);
833	} else {
834	/*
835	* Haven't connected yet, save the current mbuf
836	* and ti, and return
837	* XXX Some OS's don't tell us whether the connect()
838	* succeeded or not. So we must time it out.
839	*/
840	so->so_m = m;
841	so->so_ti = ti;
842	tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT;
843	tp->t_state = TCPS_SYN_RECEIVED;
844	}
845	return;
846
847	cont_conn:
848	/* m==NULL
849	* Check if the connect succeeded
850	*/
851	if (so->so_state & SS_NOFDREF) {
852	tp = tcp_close(pData, tp);
853	goto dropwithreset;
854	}
855	cont_input:
856	tcp_template(tp);
857
858	if (optp)
859	tcp_dooptions(pData, tp, (u_char *)optp, optlen, ti);
860	/* , */
861	/* &ts_present, &ts_val, &ts_ecr); */
862
863	if (iss)
864	tp->iss = iss;
865	else
866	tp->iss = tcp_iss;
867	tcp_iss += TCP_ISSINCR/2;
868	tp->irs = ti->ti_seq;
869	tcp_sendseqinit(tp);
870	tcp_rcvseqinit(tp);
871	tp->t_flags \|= TF_ACKNOW;
872	tp->t_state = TCPS_SYN_RECEIVED;
873	tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT;
874	tcpstat.tcps_accepts++;
875	goto trimthenstep6;
876	} /* case TCPS_LISTEN */
877
878	/*
879	* If the state is SYN_SENT:
880	* if seg contains an ACK, but not for our SYN, drop the input.
881	* if seg contains a RST, then drop the connection.
882	* if seg does not contain SYN, then drop it.
883	* Otherwise this is an acceptable SYN segment
884	* initialize tp->rcv_nxt and tp->irs
885	* if seg contains ack then advance tp->snd_una
886	* if SYN has been acked change to ESTABLISHED else SYN_RCVD state
887	* arrange for segment to be acked (eventually)
888	* continue processing rest of data/controls, beginning with URG
889	*/
890	case TCPS_SYN_SENT:
891	if ((tiflags & TH_ACK) &&
892	(SEQ_LEQ(ti->ti_ack, tp->iss) \|\|
893	SEQ_GT(ti->ti_ack, tp->snd_max)))
894	goto dropwithreset;
895
896	if (tiflags & TH_RST) {
897	if (tiflags & TH_ACK)
898	tp = tcp_drop(pData, tp,0); /* XXX Check t_softerror! */
899	goto drop;
900	}
901
902	if ((tiflags & TH_SYN) == 0)
903	goto drop;
904	if (tiflags & TH_ACK) {
905	tp->snd_una = ti->ti_ack;
906	if (SEQ_LT(tp->snd_nxt, tp->snd_una))
907	tp->snd_nxt = tp->snd_una;
908	}
909
910	tp->t_timer[TCPT_REXMT] = 0;
911	tp->irs = ti->ti_seq;
912	tcp_rcvseqinit(tp);
913	tp->t_flags \|= TF_ACKNOW;
914	if (tiflags & TH_ACK && SEQ_GT(tp->snd_una, tp->iss)) {
915	tcpstat.tcps_connects++;
916	soisfconnected(so);
917	tp->t_state = TCPS_ESTABLISHED;
918
919	/* Do window scaling on this connection? */
920	/* if ((tp->t_flags & (TF_RCVD_SCALE\|TF_REQ_SCALE)) ==
921	* (TF_RCVD_SCALE\|TF_REQ_SCALE)) {
922	* tp->snd_scale = tp->requested_s_scale;
923	* tp->rcv_scale = tp->request_r_scale;
924	* }
925	*/
926	#ifndef VBOX_WITH_BSD_TCP_REASS
927	(void) tcp_reass(pData, tp, (struct tcpiphdr *)0,
928	(struct mbuf *)0);
929	#else /* VBOX_WITH_BSD_TCP_REASS */
930	(void) tcp_reass(pData, tp, (struct tcphdr )0, NULL, (struct mbuf )0);
931	#endif /* VBOX_WITH_BSD_TCP_REASS */
932	/*
933	* if we didn't have to retransmit the SYN,
934	* use its rtt as our initial srtt & rtt var.
935	*/
936	if (tp->t_rtt)
937	tcp_xmit_timer(pData, tp, tp->t_rtt);
938	} else
939	tp->t_state = TCPS_SYN_RECEIVED;
940
941	trimthenstep6:
942	/*
943	* Advance ti->ti_seq to correspond to first data byte.
944	* If data, trim to stay within window,
945	* dropping FIN if necessary.
946	*/
947	ti->ti_seq++;
948	if (ti->ti_len > tp->rcv_wnd) {
949	todrop = ti->ti_len - tp->rcv_wnd;
950	m_adj(m, -todrop);
951	ti->ti_len = tp->rcv_wnd;
952	tiflags &= ~TH_FIN;
953	tcpstat.tcps_rcvpackafterwin++;
954	tcpstat.tcps_rcvbyteafterwin += todrop;
955	}
956	tp->snd_wl1 = ti->ti_seq - 1;
957	tp->rcv_up = ti->ti_seq;
958	goto step6;
959	} /* switch tp->t_state */
960	/*
961	* States other than LISTEN or SYN_SENT.
962	* First check timestamp, if present.
963	* Then check that at least some bytes of segment are within
964	* receive window. If segment begins before rcv_nxt,
965	* drop leading data (and SYN); if nothing left, just ack.
966	*
967	* RFC 1323 PAWS: If we have a timestamp reply on this segment
968	* and it's less than ts_recent, drop it.
969	*/
970	/* if (ts_present && (tiflags & TH_RST) == 0 && tp->ts_recent &&
971	* TSTMP_LT(ts_val, tp->ts_recent)) {
972	*
973	/ / Check to see if ts_recent is over 24 days old. */
974	/* if ((int)(tcp_now - tp->ts_recent_age) > TCP_PAWS_IDLE) {
975	/ /
976	* * Invalidate ts_recent. If this segment updates
977	* * ts_recent, the age will be reset later and ts_recent
978	* * will get a valid value. If it does not, setting
979	* * ts_recent to zero will at least satisfy the
980	* * requirement that zero be placed in the timestamp
981	* * echo reply when ts_recent isn't valid. The
982	* * age isn't reset until we get a valid ts_recent
983	* * because we don't want out-of-order segments to be
984	* * dropped when ts_recent is old.
985	* */
986	/* tp->ts_recent = 0;
987	* } else {
988	* tcpstat.tcps_rcvduppack++;
989	* tcpstat.tcps_rcvdupbyte += ti->ti_len;
990	* tcpstat.tcps_pawsdrop++;
991	* goto dropafterack;
992	* }
993	* }
994	*/
995
996	todrop = tp->rcv_nxt - ti->ti_seq;
997	if (todrop > 0) {
998	if (tiflags & TH_SYN) {
999	tiflags &= ~TH_SYN;
1000	ti->ti_seq++;
1001	if (ti->ti_urp > 1)
1002	ti->ti_urp--;
1003	else
1004	tiflags &= ~TH_URG;
1005	todrop--;
1006	}
1007	/*
1008	* Following if statement from Stevens, vol. 2, p. 960.
1009	*/
1010	if (todrop > ti->ti_len
1011	\|\| (todrop == ti->ti_len && (tiflags & TH_FIN) == 0)) {
1012	/*
1013	* Any valid FIN must be to the left of the window.
1014	* At this point the FIN must be a duplicate or out
1015	* of sequence; drop it.
1016	*/
1017	tiflags &= ~TH_FIN;
1018
1019	/*
1020	* Send an ACK to resynchronize and drop any data.
1021	* But keep on processing for RST or ACK.
1022	*/
1023	tp->t_flags \|= TF_ACKNOW;
1024	todrop = ti->ti_len;
1025	tcpstat.tcps_rcvduppack++;
1026	tcpstat.tcps_rcvdupbyte += todrop;
1027	} else {
1028	tcpstat.tcps_rcvpartduppack++;
1029	tcpstat.tcps_rcvpartdupbyte += todrop;
1030	}
1031	m_adj(m, todrop);
1032	ti->ti_seq += todrop;
1033	ti->ti_len -= todrop;
1034	if (ti->ti_urp > todrop)
1035	ti->ti_urp -= todrop;
1036	else {
1037	tiflags &= ~TH_URG;
1038	ti->ti_urp = 0;
1039	}
1040	}
1041	/*
1042	* If new data are received on a connection after the
1043	* user processes are gone, then RST the other end.
1044	*/
1045	if ((so->so_state & SS_NOFDREF) &&
1046	tp->t_state > TCPS_CLOSE_WAIT && ti->ti_len) {
1047	tp = tcp_close(pData, tp);
1048	tcpstat.tcps_rcvafterclose++;
1049	goto dropwithreset;
1050	}
1051
1052	/*
1053	* If segment ends after window, drop trailing data
1054	* (and PUSH and FIN); if nothing left, just ACK.
1055	*/
1056	todrop = (ti->ti_seq+ti->ti_len) - (tp->rcv_nxt+tp->rcv_wnd);
1057	if (todrop > 0) {
1058	tcpstat.tcps_rcvpackafterwin++;
1059	if (todrop >= ti->ti_len) {
1060	tcpstat.tcps_rcvbyteafterwin += ti->ti_len;
1061	/*
1062	* If a new connection request is received
1063	* while in TIME_WAIT, drop the old connection
1064	* and start over if the sequence numbers
1065	* are above the previous ones.
1066	*/
1067	if (tiflags & TH_SYN &&
1068	tp->t_state == TCPS_TIME_WAIT &&
1069	SEQ_GT(ti->ti_seq, tp->rcv_nxt)) {
1070	iss = tp->rcv_nxt + TCP_ISSINCR;
1071	tp = tcp_close(pData, tp);
1072	goto findso;
1073	}
1074	/*
1075	* If window is closed can only take segments at
1076	* window edge, and have to drop data and PUSH from
1077	* incoming segments. Continue processing, but
1078	* remember to ack. Otherwise, drop segment
1079	* and ack.
1080	*/
1081	if (tp->rcv_wnd == 0 && ti->ti_seq == tp->rcv_nxt) {
1082	tp->t_flags \|= TF_ACKNOW;
1083	tcpstat.tcps_rcvwinprobe++;
1084	} else
1085	goto dropafterack;
1086	} else
1087	tcpstat.tcps_rcvbyteafterwin += todrop;
1088	m_adj(m, -todrop);
1089	ti->ti_len -= todrop;
1090	tiflags &= ~(TH_PUSH\|TH_FIN);
1091	}
1092
1093	/*
1094	* If last ACK falls within this segment's sequence numbers,
1095	* record its timestamp.
1096	*/
1097	/* if (ts_present && SEQ_LEQ(ti->ti_seq, tp->last_ack_sent) &&
1098	* SEQ_LT(tp->last_ack_sent, ti->ti_seq + ti->ti_len +
1099	* ((tiflags & (TH_SYN\|TH_FIN)) != 0))) {
1100	* tp->ts_recent_age = tcp_now;
1101	* tp->ts_recent = ts_val;
1102	* }
1103	*/
1104
1105	/*
1106	* If the RST bit is set examine the state:
1107	* SYN_RECEIVED STATE:
1108	* If passive open, return to LISTEN state.
1109	* If active open, inform user that connection was refused.
1110	* ESTABLISHED, FIN_WAIT_1, FIN_WAIT2, CLOSE_WAIT STATES:
1111	* Inform user that connection was reset, and close tcb.
1112	* CLOSING, LAST_ACK, TIME_WAIT STATES
1113	* Close the tcb.
1114	*/
1115	if (tiflags&TH_RST) switch (tp->t_state) {
1116
1117	case TCPS_SYN_RECEIVED:
1118	/* so->so_error = ECONNREFUSED; */
1119	goto close;
1120
1121	case TCPS_ESTABLISHED:
1122	case TCPS_FIN_WAIT_1:
1123	case TCPS_FIN_WAIT_2:
1124	case TCPS_CLOSE_WAIT:
1125	/* so->so_error = ECONNRESET; */
1126	close:
1127	tp->t_state = TCPS_CLOSED;
1128	tcpstat.tcps_drops++;
1129	tp = tcp_close(pData, tp);
1130	goto drop;
1131
1132	case TCPS_CLOSING:
1133	case TCPS_LAST_ACK:
1134	case TCPS_TIME_WAIT:
1135	tp = tcp_close(pData, tp);
1136	goto drop;
1137	}
1138
1139	/*
1140	* If a SYN is in the window, then this is an
1141	* error and we send an RST and drop the connection.
1142	*/
1143	if (tiflags & TH_SYN) {
1144	tp = tcp_drop(pData, tp,0);
1145	goto dropwithreset;
1146	}
1147
1148	/*
1149	* If the ACK bit is off we drop the segment and return.
1150	*/
1151	if ((tiflags & TH_ACK) == 0) goto drop;
1152
1153	/*
1154	* Ack processing.
1155	*/
1156	switch (tp->t_state) {
1157	/*
1158	* In SYN_RECEIVED state if the ack ACKs our SYN then enter
1159	* ESTABLISHED state and continue processing, otherwise
1160	* send an RST. una<=ack<=max
1161	*/
1162	case TCPS_SYN_RECEIVED:
1163
1164	if (SEQ_GT(tp->snd_una, ti->ti_ack) \|\|
1165	SEQ_GT(ti->ti_ack, tp->snd_max))
1166	goto dropwithreset;
1167	tcpstat.tcps_connects++;
1168	tp->t_state = TCPS_ESTABLISHED;
1169	/*
1170	* The sent SYN is ack'ed with our sequence number +1
1171	* The first data byte already in the buffer will get
1172	* lost if no correction is made. This is only needed for
1173	* SS_CTL since the buffer is empty otherwise.
1174	* tp->snd_una++; or:
1175	*/
1176	tp->snd_una=ti->ti_ack;
1177	soisfconnected(so);
1178
1179	/* Do window scaling? */
1180	/* if ((tp->t_flags & (TF_RCVD_SCALE\|TF_REQ_SCALE)) ==
1181	* (TF_RCVD_SCALE\|TF_REQ_SCALE)) {
1182	* tp->snd_scale = tp->requested_s_scale;
1183	* tp->rcv_scale = tp->request_r_scale;
1184	* }
1185	*/
1186	#ifndef VBOX_WITH_BSD_TCP_REASS
1187	(void) tcp_reass(pData, tp, (struct tcpiphdr )0, (struct mbuf )0);
1188	#else /* VBOX_WITH_BSD_TCP_REASS */
1189	(void) tcp_reass(pData, tp, (struct tcphdr )0, (int )0, (struct mbuf *)0);
1190	#endif /VBOX_WITH_BSD_TCP_REASS/
1191	tp->snd_wl1 = ti->ti_seq - 1;
1192	/* Avoid ack processing; snd_una==ti_ack => dup ack */
1193	goto synrx_to_est;
1194	/* fall into ... */
1195
1196	/*
1197	* In ESTABLISHED state: drop duplicate ACKs; ACK out of range
1198	* ACKs. If the ack is in the range
1199	* tp->snd_una < ti->ti_ack <= tp->snd_max
1200	* then advance tp->snd_una to ti->ti_ack and drop
1201	* data from the retransmission queue. If this ACK reflects
1202	* more up to date window information we update our window information.
1203	*/
1204	case TCPS_ESTABLISHED:
1205	case TCPS_FIN_WAIT_1:
1206	case TCPS_FIN_WAIT_2:
1207	case TCPS_CLOSE_WAIT:
1208	case TCPS_CLOSING:
1209	case TCPS_LAST_ACK:
1210	case TCPS_TIME_WAIT:
1211
1212	if (SEQ_LEQ(ti->ti_ack, tp->snd_una)) {
1213	if (ti->ti_len == 0 && tiwin == tp->snd_wnd) {
1214	tcpstat.tcps_rcvdupack++;
1215	DEBUG_MISC((dfd," dup ack m = %lx so = %lx \n",
1216	(long )m, (long )so));
1217	/*
1218	* If we have outstanding data (other than
1219	* a window probe), this is a completely
1220	* duplicate ack (ie, window info didn't
1221	* change), the ack is the biggest we've
1222	* seen and we've seen exactly our rexmt
1223	* threshold of them, assume a packet
1224	* has been dropped and retransmit it.
1225	* Kludge snd_nxt & the congestion
1226	* window so we send only this one
1227	* packet.
1228	*
1229	* We know we're losing at the current
1230	* window size so do congestion avoidance
1231	* (set ssthresh to half the current window
1232	* and pull our congestion window back to
1233	* the new ssthresh).
1234	*
1235	* Dup acks mean that packets have left the
1236	* network (they're now cached at the receiver)
1237	* so bump cwnd by the amount in the receiver
1238	* to keep a constant cwnd packets in the
1239	* network.
1240	*/
1241	if (tp->t_timer[TCPT_REXMT] == 0 \|\|
1242	ti->ti_ack != tp->snd_una)
1243	tp->t_dupacks = 0;
1244	else if (++tp->t_dupacks == tcprexmtthresh) {
1245	tcp_seq onxt = tp->snd_nxt;
1246	u_int win =
1247	min(tp->snd_wnd, tp->snd_cwnd) / 2 /
1248	tp->t_maxseg;
1249
1250	if (win < 2)
1251	win = 2;
1252	tp->snd_ssthresh = win * tp->t_maxseg;
1253	tp->t_timer[TCPT_REXMT] = 0;
1254	tp->t_rtt = 0;
1255	tp->snd_nxt = ti->ti_ack;
1256	tp->snd_cwnd = tp->t_maxseg;
1257	(void) tcp_output(pData, tp);
1258	tp->snd_cwnd = tp->snd_ssthresh +
1259	tp->t_maxseg * tp->t_dupacks;
1260	if (SEQ_GT(onxt, tp->snd_nxt))
1261	tp->snd_nxt = onxt;
1262	goto drop;
1263	} else if (tp->t_dupacks > tcprexmtthresh) {
1264	tp->snd_cwnd += tp->t_maxseg;
1265	(void) tcp_output(pData, tp);
1266	goto drop;
1267	}
1268	} else
1269	tp->t_dupacks = 0;
1270	break;
1271	}
1272	synrx_to_est:
1273	/*
1274	* If the congestion window was inflated to account
1275	* for the other side's cached packets, retract it.
1276	*/
1277	if (tp->t_dupacks > tcprexmtthresh &&
1278	tp->snd_cwnd > tp->snd_ssthresh)
1279	tp->snd_cwnd = tp->snd_ssthresh;
1280	tp->t_dupacks = 0;
1281	if (SEQ_GT(ti->ti_ack, tp->snd_max)) {
1282	tcpstat.tcps_rcvacktoomuch++;
1283	goto dropafterack;
1284	}
1285	acked = ti->ti_ack - tp->snd_una;
1286	tcpstat.tcps_rcvackpack++;
1287	tcpstat.tcps_rcvackbyte += acked;
1288
1289	/*
1290	* If we have a timestamp reply, update smoothed
1291	* round trip time. If no timestamp is present but
1292	* transmit timer is running and timed sequence
1293	* number was acked, update smoothed round trip time.
1294	* Since we now have an rtt measurement, cancel the
1295	* timer backoff (cf., Phil Karn's retransmit alg.).
1296	* Recompute the initial retransmit timer.
1297	*/
1298	/* if (ts_present)
1299	* tcp_xmit_timer(tp, tcp_now-ts_ecr+1);
1300	* else
1301	*/
1302	if (tp->t_rtt && SEQ_GT(ti->ti_ack, tp->t_rtseq))
1303	tcp_xmit_timer(pData, tp,tp->t_rtt);
1304
1305	/*
1306	* If all outstanding data is acked, stop retransmit
1307	* timer and remember to restart (more output or persist).
1308	* If there is more data to be acked, restart retransmit
1309	* timer, using current (possibly backed-off) value.
1310	*/
1311	if (ti->ti_ack == tp->snd_max) {
1312	tp->t_timer[TCPT_REXMT] = 0;
1313	needoutput = 1;
1314	} else if (tp->t_timer[TCPT_PERSIST] == 0)
1315	tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
1316	/*
1317	* When new data is acked, open the congestion window.
1318	* If the window gives us less than ssthresh packets
1319	* in flight, open exponentially (maxseg per packet).
1320	* Otherwise open linearly: maxseg per window
1321	* (maxseg^2 / cwnd per packet).
1322	*/
1323	{
1324	register u_int cw = tp->snd_cwnd;
1325	register u_int incr = tp->t_maxseg;
1326
1327	if (cw > tp->snd_ssthresh)
1328	incr = incr * incr / cw;
1329	tp->snd_cwnd = min(cw + incr, TCP_MAXWIN<<tp->snd_scale);
1330	}
1331	if (acked > so->so_snd.sb_cc) {
1332	tp->snd_wnd -= so->so_snd.sb_cc;
1333	sbdrop(&so->so_snd, (int )so->so_snd.sb_cc);
1334	ourfinisacked = 1;
1335	} else {
1336	sbdrop(&so->so_snd, acked);
1337	tp->snd_wnd -= acked;
1338	ourfinisacked = 0;
1339	}
1340	/*
1341	* XXX sowwakup is called when data is acked and there's room for
1342	* for more data... it should read() the socket
1343	*/
1344	/* if (so->so_snd.sb_flags & SB_NOTIFY)
1345	* sowwakeup(so);
1346	*/
1347	tp->snd_una = ti->ti_ack;
1348	if (SEQ_LT(tp->snd_nxt, tp->snd_una))
1349	tp->snd_nxt = tp->snd_una;
1350
1351	switch (tp->t_state) {
1352
1353	/*
1354	* In FIN_WAIT_1 STATE in addition to the processing
1355	* for the ESTABLISHED state if our FIN is now acknowledged
1356	* then enter FIN_WAIT_2.
1357	*/
1358	case TCPS_FIN_WAIT_1:
1359	if (ourfinisacked) {
1360	/*
1361	* If we can't receive any more
1362	* data, then closing user can proceed.
1363	* Starting the timer is contrary to the
1364	* specification, but if we don't get a FIN
1365	* we'll hang forever.
1366	*/
1367	if (so->so_state & SS_FCANTRCVMORE) {
1368	soisfdisconnected(so);
1369	tp->t_timer[TCPT_2MSL] = tcp_maxidle;
1370	}
1371	tp->t_state = TCPS_FIN_WAIT_2;
1372	}
1373	break;
1374
1375	/*
1376	* In CLOSING STATE in addition to the processing for
1377	* the ESTABLISHED state if the ACK acknowledges our FIN
1378	* then enter the TIME-WAIT state, otherwise ignore
1379	* the segment.
1380	*/
1381	case TCPS_CLOSING:
1382	if (ourfinisacked) {
1383	tp->t_state = TCPS_TIME_WAIT;
1384	tcp_canceltimers(tp);
1385	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1386	soisfdisconnected(so);
1387	}
1388	break;
1389
1390	/*
1391	* In LAST_ACK, we may still be waiting for data to drain
1392	* and/or to be acked, as well as for the ack of our FIN.
1393	* If our FIN is now acknowledged, delete the TCB,
1394	* enter the closed state and return.
1395	*/
1396	case TCPS_LAST_ACK:
1397	if (ourfinisacked) {
1398	tp = tcp_close(pData, tp);
1399	goto drop;
1400	}
1401	break;
1402
1403	/*
1404	* In TIME_WAIT state the only thing that should arrive
1405	* is a retransmission of the remote FIN. Acknowledge
1406	* it and restart the finack timer.
1407	*/
1408	case TCPS_TIME_WAIT:
1409	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1410	goto dropafterack;
1411	}
1412	} /* switch(tp->t_state) */
1413
1414	step6:
1415	/*
1416	* Update window information.
1417	* Don't look at window if no ACK: TAC's send garbage on first SYN.
1418	*/
1419	if ((tiflags & TH_ACK) &&
1420	(SEQ_LT(tp->snd_wl1, ti->ti_seq) \|\|
1421	(tp->snd_wl1 == ti->ti_seq && (SEQ_LT(tp->snd_wl2, ti->ti_ack) \|\|
1422	(tp->snd_wl2 == ti->ti_ack && tiwin > tp->snd_wnd))))) {
1423	/* keep track of pure window updates */
1424	if (ti->ti_len == 0 &&
1425	tp->snd_wl2 == ti->ti_ack && tiwin > tp->snd_wnd)
1426	tcpstat.tcps_rcvwinupd++;
1427	tp->snd_wnd = tiwin;
1428	tp->snd_wl1 = ti->ti_seq;
1429	tp->snd_wl2 = ti->ti_ack;
1430	if (tp->snd_wnd > tp->max_sndwnd)
1431	tp->max_sndwnd = tp->snd_wnd;
1432	needoutput = 1;
1433	}
1434
1435	/*
1436	* Process segments with URG.
1437	*/
1438	if ((tiflags & TH_URG) && ti->ti_urp &&
1439	TCPS_HAVERCVDFIN(tp->t_state) == 0) {
1440	/*
1441	* This is a kludge, but if we receive and accept
1442	* random urgent pointers, we'll crash in
1443	* soreceive. It's hard to imagine someone
1444	* actually wanting to send this much urgent data.
1445	*/
1446	if (ti->ti_urp + so->so_rcv.sb_cc > so->so_rcv.sb_datalen) {
1447	ti->ti_urp = 0;
1448	tiflags &= ~TH_URG;
1449	goto dodata;
1450	}
1451	/*
1452	* If this segment advances the known urgent pointer,
1453	* then mark the data stream. This should not happen
1454	* in CLOSE_WAIT, CLOSING, LAST_ACK or TIME_WAIT STATES since
1455	* a FIN has been received from the remote side.
1456	* In these states we ignore the URG.
1457	*
1458	* According to RFC961 (Assigned Protocols),
1459	* the urgent pointer points to the last octet
1460	* of urgent data. We continue, however,
1461	* to consider it to indicate the first octet
1462	* of data past the urgent section as the original
1463	* spec states (in one of two places).
1464	*/
1465	if (SEQ_GT(ti->ti_seq+ti->ti_urp, tp->rcv_up)) {
1466	tp->rcv_up = ti->ti_seq + ti->ti_urp;
1467	so->so_urgc = so->so_rcv.sb_cc +
1468	(tp->rcv_up - tp->rcv_nxt); /* -1; */
1469	tp->rcv_up = ti->ti_seq + ti->ti_urp;
1470
1471	}
1472	} else
1473	/*
1474	* If no out of band data is expected,
1475	* pull receive urgent pointer along
1476	* with the receive window.
1477	*/
1478	if (SEQ_GT(tp->rcv_nxt, tp->rcv_up))
1479	tp->rcv_up = tp->rcv_nxt;
1480	dodata:
1481
1482	/*
1483	* Process the segment text, merging it into the TCP sequencing queue,
1484	* and arranging for acknowledgment of receipt if necessary.
1485	* This process logically involves adjusting tp->rcv_wnd as data
1486	* is presented to the user (this happens in tcp_usrreq.c,
1487	* case PRU_RCVD). If a FIN has already been received on this
1488	* connection then we just ignore the text.
1489	*/
1490	if ((ti->ti_len \|\| (tiflags&TH_FIN)) &&
1491	TCPS_HAVERCVDFIN(tp->t_state) == 0) {
1492	#ifndef VBOX_WITH_BSD_TCP_REASS
1493	TCP_REASS(pData, tp, ti, m, so, tiflags);
1494	#else /* VBOX_WITH_BSD_TCP_REASS */
1495	if (ti->ti_seq == tp->rcv_nxt
1496	&& LIST_EMPTY(&tp->t_segq)
1497	&& tp->t_state == TCPS_ESTABLISHED) {
1498	DELAY_ACK(tp, ti); /* little bit different from BSD declaration see netinet/tcp_input.c */
1499	tp->rcv_nxt += tlen;
1500	tiflags = ti->ti_t.th_flags & TH_FIN;
1501	tcpstat.tcps_rcvpack++;
1502	tcpstat.tcps_rcvbyte += tlen;
1503	if (so->so_state & SS_FCANTRCVMORE)
1504	m_freem(pData, m);
1505	else
1506	sbappend(pData, so, m);
1507	}
1508	else {
1509	tiflags = tcp_reass(pData, tp, &ti->ti_t, &tlen, m);
1510	tiflags \|= TF_ACKNOW;
1511	}
1512	#endif /* VBOX_WITH_BSD_TCP_REASS */
1513	/*
1514	* Note the amount of data that peer has sent into
1515	* our window, in order to estimate the sender's
1516	* buffer size.
1517	*/
1518	len = so->so_rcv.sb_datalen - (tp->rcv_adv - tp->rcv_nxt);
1519	} else {
1520	mbuf_freed = 1; /* The mbuf must be freed, but only when its content is not needed anymore. */
1521	tiflags &= ~TH_FIN;
1522	}
1523
1524	/*
1525	* If FIN is received ACK the FIN and let the user know
1526	* that the connection is closing.
1527	*/
1528	if (tiflags & TH_FIN) {
1529	if (TCPS_HAVERCVDFIN(tp->t_state) == 0) {
1530	/*
1531	* If we receive a FIN we can't send more data,
1532	* set it SS_FDRAIN
1533	* Shutdown the socket if there is no rx data in the
1534	* buffer.
1535	* soread() is called on completion of shutdown() and
1536	* will got to TCPS_LAST_ACK, and use tcp_output()
1537	* to send the FIN.
1538	*/
1539	/* sofcantrcvmore(so); */
1540	sofwdrain(so);
1541
1542	tp->t_flags \|= TF_ACKNOW;
1543	tp->rcv_nxt++;
1544	}
1545	switch (tp->t_state) {
1546
1547	/*
1548	* In SYN_RECEIVED and ESTABLISHED STATES
1549	* enter the CLOSE_WAIT state.
1550	*/
1551	case TCPS_SYN_RECEIVED:
1552	case TCPS_ESTABLISHED:
1553	if(so->so_emu == EMU_CTL) /* no shutdown on socket */
1554	tp->t_state = TCPS_LAST_ACK;
1555	else
1556	tp->t_state = TCPS_CLOSE_WAIT;
1557	break;
1558
1559	/*
1560	* If still in FIN_WAIT_1 STATE FIN has not been acked so
1561	* enter the CLOSING state.
1562	*/
1563	case TCPS_FIN_WAIT_1:
1564	tp->t_state = TCPS_CLOSING;
1565	break;
1566
1567	/*
1568	* In FIN_WAIT_2 state enter the TIME_WAIT state,
1569	* starting the time-wait timer, turning off the other
1570	* standard timers.
1571	*/
1572	case TCPS_FIN_WAIT_2:
1573	tp->t_state = TCPS_TIME_WAIT;
1574	tcp_canceltimers(tp);
1575	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1576	soisfdisconnected(so);
1577	break;
1578
1579	/*
1580	* In TIME_WAIT state restart the 2 MSL time_wait timer.
1581	*/
1582	case TCPS_TIME_WAIT:
1583	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1584	break;
1585	}
1586	}
1587
1588	/*
1589	* If this is a small packet, then ACK now - with Nagel
1590	* congestion avoidance sender won't send more until
1591	* he gets an ACK.
1592	*
1593	* See above.
1594	*/
1595	/* if (ti->ti_len && (unsigned)ti->ti_len < tp->t_maxseg) {
1596	*/
1597	/* if ((ti->ti_len && (unsigned)ti->ti_len < tp->t_maxseg &&
1598	* (so->so_iptos & IPTOS_LOWDELAY) == 0) \|\|
1599	* ((so->so_iptos & IPTOS_LOWDELAY) &&
1600	* ((struct tcpiphdr_2 *)ti)->first_char == (char)27)) {
1601	*/
1602	if (ti->ti_len && (unsigned)ti->ti_len <= 5 &&
1603	((struct tcpiphdr_2 *)ti)->first_char == (char)27) {
1604	tp->t_flags \|= TF_ACKNOW;
1605	}
1606
1607	if (mbuf_freed) {
1608	m_free(pData, m);
1609	}
1610	/*
1611	* Return any desired output.
1612	*/
1613	if (needoutput \|\| (tp->t_flags & TF_ACKNOW)) {
1614	(void) tcp_output(pData, tp);
1615	}
1616	return;
1617
1618	dropafterack:
1619	/*
1620	* Generate an ACK dropping incoming segment if it occupies
1621	* sequence space, where the ACK reflects our state.
1622	*/
1623	if (tiflags & TH_RST)
1624	goto drop;
1625	m_freem(pData, m);
1626	tp->t_flags \|= TF_ACKNOW;
1627	(void) tcp_output(pData, tp);
1628	return;
1629
1630	dropwithreset:
1631	/* reuses m if m!=NULL, m_free() unnecessary */
1632	if (tiflags & TH_ACK)
1633	tcp_respond(pData, tp, ti, m, (tcp_seq)0, ti->ti_ack, TH_RST);
1634	else {
1635	if (tiflags & TH_SYN) ti->ti_len++;
1636	tcp_respond(pData, tp, ti, m, ti->ti_seq+ti->ti_len, (tcp_seq)0,
1637	TH_RST\|TH_ACK);
1638	}
1639
1640	return;
1641
1642	drop:
1643	/*
1644	* Drop space held by incoming segment and return.
1645	*/
1646	m_free(pData, m);
1647
1648	return;
1649	}
1650
1651	/* , ts_present, ts_val, ts_ecr) */
1652	/* int *ts_present;
1653	* u_int32_t ts_val, ts_ecr;
1654	*/
1655	void
1656	tcp_dooptions(PNATState pData, struct tcpcb tp, u_char cp, int cnt, struct tcpiphdr *ti)
1657	{
1658	u_int16_t mss;
1659	int opt, optlen;
1660
1661	DEBUG_CALL("tcp_dooptions");
1662	DEBUG_ARGS((dfd," tp = %lx cnt=%i \n", (long )tp, cnt));
1663
1664	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1665	opt = cp[0];
1666	if (opt == TCPOPT_EOL)
1667	break;
1668	if (opt == TCPOPT_NOP)
1669	optlen = 1;
1670	else {
1671	optlen = cp[1];
1672	if (optlen <= 0)
1673	break;
1674	}
1675	switch (opt) {
1676
1677	default:
1678	continue;
1679
1680	case TCPOPT_MAXSEG:
1681	if (optlen != TCPOLEN_MAXSEG)
1682	continue;
1683	if (!(ti->ti_flags & TH_SYN))
1684	continue;
1685	memcpy((char ) &mss, (char ) cp + 2, sizeof(mss));
1686	NTOHS(mss);
1687	(void) tcp_mss(pData, tp, mss); /* sets t_maxseg */
1688	break;
1689
1690	/* case TCPOPT_WINDOW:
1691	* if (optlen != TCPOLEN_WINDOW)
1692	* continue;
1693	* if (!(ti->ti_flags & TH_SYN))
1694	* continue;
1695	* tp->t_flags \|= TF_RCVD_SCALE;
1696	* tp->requested_s_scale = min(cp[2], TCP_MAX_WINSHIFT);
1697	* break;
1698	*/
1699	/* case TCPOPT_TIMESTAMP:
1700	* if (optlen != TCPOLEN_TIMESTAMP)
1701	* continue;
1702	* *ts_present = 1;
1703	* memcpy((char ) ts_val, (char )cp + 2, sizeof(*ts_val));
1704	* NTOHL(*ts_val);
1705	* memcpy((char ) ts_ecr, (char )cp + 6, sizeof(*ts_ecr));
1706	* NTOHL(*ts_ecr);
1707	*
1708	/ /
1709	* * A timestamp received in a SYN makes
1710	* * it ok to send timestamp requests and replies.
1711	* */
1712	/* if (ti->ti_flags & TH_SYN) {
1713	* tp->t_flags \|= TF_RCVD_TSTMP;
1714	* tp->ts_recent = *ts_val;
1715	* tp->ts_recent_age = tcp_now;
1716	* }
1717	*/ break;
1718	}
1719	}
1720	}
1721
1722
1723	/*
1724	* Pull out of band byte out of a segment so
1725	* it doesn't appear in the user's data queue.
1726	* It is still reflected in the segment length for
1727	* sequencing purposes.
1728	*/
1729
1730	#ifdef notdef
1731
1732	void
1733	tcp_pulloutofband(so, ti, m)
1734	struct socket *so;
1735	struct tcpiphdr *ti;
1736	register struct mbuf *m;
1737	{
1738	int cnt = ti->ti_urp - 1;
1739
1740	while (cnt >= 0) {
1741	if (m->m_len > cnt) {
1742	char *cp = mtod(m, caddr_t) + cnt;
1743	struct tcpcb *tp = sototcpcb(so);
1744
1745	tp->t_iobc = *cp;
1746	tp->t_oobflags \|= TCPOOB_HAVEDATA;
1747	memcpy(sp, cp+1, (unsigned)(m->m_len - cnt - 1));
1748	m->m_len--;
1749	return;
1750	}
1751	cnt -= m->m_len;
1752	m = m->m_next; /* XXX WRONG! Fix it! */
1753	if (m == 0)
1754	break;
1755	}
1756	panic("tcp_pulloutofband");
1757	}
1758
1759	#endif /* notdef */
1760
1761	/*
1762	* Collect new round-trip time estimate
1763	* and update averages and current timeout.
1764	*/
1765
1766	void
1767	tcp_xmit_timer(PNATState pData, register struct tcpcb *tp, int rtt)
1768	{
1769	register short delta;
1770
1771	DEBUG_CALL("tcp_xmit_timer");
1772	DEBUG_ARG("tp = %lx", (long)tp);
1773	DEBUG_ARG("rtt = %d", rtt);
1774
1775	tcpstat.tcps_rttupdated++;
1776	if (tp->t_srtt != 0) {
1777	/*
1778	* srtt is stored as fixed point with 3 bits after the
1779	* binary point (i.e., scaled by 8). The following magic
1780	* is equivalent to the smoothing algorithm in rfc793 with
1781	* an alpha of .875 (srtt = rtt/8 + srtt*7/8 in fixed
1782	* point). Adjust rtt to origin 0.
1783	*/
1784	delta = rtt - 1 - (tp->t_srtt >> TCP_RTT_SHIFT);
1785	if ((tp->t_srtt += delta) <= 0)
1786	tp->t_srtt = 1;
1787	/*
1788	* We accumulate a smoothed rtt variance (actually, a
1789	* smoothed mean difference), then set the retransmit
1790	* timer to smoothed rtt + 4 times the smoothed variance.
1791	* rttvar is stored as fixed point with 2 bits after the
1792	* binary point (scaled by 4). The following is
1793	* equivalent to rfc793 smoothing with an alpha of .75
1794	* (rttvar = rttvar*3/4 + \|delta\| / 4). This replaces
1795	* rfc793's wired-in beta.
1796	*/
1797	if (delta < 0)
1798	delta = -delta;
1799	delta -= (tp->t_rttvar >> TCP_RTTVAR_SHIFT);
1800	if ((tp->t_rttvar += delta) <= 0)
1801	tp->t_rttvar = 1;
1802	} else {
1803	/*
1804	* No rtt measurement yet - use the unsmoothed rtt.
1805	* Set the variance to half the rtt (so our first
1806	* retransmit happens at 3*rtt).
1807	*/
1808	tp->t_srtt = rtt << TCP_RTT_SHIFT;
1809	tp->t_rttvar = rtt << (TCP_RTTVAR_SHIFT - 1);
1810	}
1811	tp->t_rtt = 0;
1812	tp->t_rxtshift = 0;
1813
1814	/*
1815	* the retransmit should happen at rtt + 4 * rttvar.
1816	* Because of the way we do the smoothing, srtt and rttvar
1817	* will each average +1/2 tick of bias. When we compute
1818	* the retransmit timer, we want 1/2 tick of rounding and
1819	* 1 extra tick because of +-1/2 tick uncertainty in the
1820	* firing of the timer. The bias will give us exactly the
1821	* 1.5 tick we need. But, because the bias is
1822	* statistical, we have to test that we don't drop below
1823	* the minimum feasible timer (which is 2 ticks).
1824	*/
1825	TCPT_RANGESET(tp->t_rxtcur, TCP_REXMTVAL(tp),
1826	(short)tp->t_rttmin, TCPTV_REXMTMAX); /* XXX */
1827
1828	/*
1829	* We received an ack for a packet that wasn't retransmitted;
1830	* it is probably safe to discard any error indications we've
1831	* received recently. This isn't quite right, but close enough
1832	* for now (a route might have failed after we sent a segment,
1833	* and the return path might not be symmetrical).
1834	*/
1835	tp->t_softerror = 0;
1836	}
1837
1838	/*
1839	* Determine a reasonable value for maxseg size.
1840	* If the route is known, check route for mtu.
1841	* If none, use an mss that can be handled on the outgoing
1842	* interface without forcing IP to fragment; if bigger than
1843	* an mbuf cluster (MCLBYTES), round down to nearest multiple of MCLBYTES
1844	* to utilize large mbufs. If no route is found, route has no mtu,
1845	* or the destination isn't local, use a default, hopefully conservative
1846	* size (usually 512 or the default IP max size, but no more than the mtu
1847	* of the interface), as we can't discover anything about intervening
1848	* gateways or networks. We also initialize the congestion/slow start
1849	* window to be a single segment if the destination isn't local.
1850	* While looking at the routing entry, we also initialize other path-dependent
1851	* parameters from pre-set or cached values in the routing entry.
1852	*/
1853
1854	int
1855	tcp_mss(PNATState pData, register struct tcpcb *tp, u_int offer)
1856	{
1857	struct socket *so = tp->t_socket;
1858	int mss;
1859
1860	DEBUG_CALL("tcp_mss");
1861	DEBUG_ARG("tp = %lx", (long)tp);
1862	DEBUG_ARG("offer = %d", offer);
1863
1864	mss = min(if_mtu, if_mru) - sizeof(struct tcpiphdr);
1865	if (offer)
1866	mss = min(mss, offer);
1867	mss = max(mss, 32);
1868	if (mss < tp->t_maxseg \|\| offer != 0)
1869	tp->t_maxseg = mss;
1870
1871	tp->snd_cwnd = mss;
1872
1873	sbreserve(&so->so_snd, tcp_sndspace+((tcp_sndspace%mss)?(mss-(tcp_sndspace%mss)):0));
1874	sbreserve(&so->so_rcv, tcp_rcvspace+((tcp_rcvspace%mss)?(mss-(tcp_rcvspace%mss)):0));
1875
1876	DEBUG_MISC((dfd, " returning mss = %d\n", mss));
1877
1878	return mss;
1879	}

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

source: vbox/trunk/src/VBox/Devices/Network/slirp/tcp_input.c@ 14529

Download in other formats: