tcp_input.c@ 14577

Last change on this file since 14577 was 14577, checked in by vboxsync, 16 years ago
slirp: attempt to fix memory corruption
Property svn:eol-style set to `native`
File size: 73.0 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
425	DEBUG_CALL("tcp_input");
426	DEBUG_ARGS((dfd," m = %8lx iphlen = %2d inso = %lx\n",
427	(long )m, iphlen, (long )inso ));
428
429	/*
430	* If called with m == 0, then we're continuing the connect
431	*/
432	if (m == NULL) {
433	so = inso;
434
435	/* Re-set a few variables */
436	tp = sototcpcb(so);
437	m = so->so_m;
438	so->so_m = 0;
439	ti = so->so_ti;
440	tiwin = ti->ti_win;
441	tiflags = ti->ti_flags;
442
443	goto cont_conn;
444	}
445
446
447	tcpstat.tcps_rcvtotal++;
448	/*
449	* Get IP and TCP header together in first mbuf.
450	* Note: IP leaves IP header in first mbuf.
451	*/
452	ti = mtod(m, struct tcpiphdr *);
453	if (iphlen > sizeof(struct ip )) {
454	ip_stripoptions(m, (struct mbuf *)0);
455	iphlen=sizeof(struct ip );
456	}
457	/* XXX Check if too short */
458
459
460	/*
461	* Save a copy of the IP header in case we want restore it
462	* for sending an ICMP error message in response.
463	*/
464	ip=mtod(m, struct ip *);
465	save_ip = *ip;
466	save_ip.ip_len+= iphlen;
467
468	/*
469	* Checksum extended TCP header and data.
470	*/
471	tlen = ((struct ip *)ti)->ip_len;
472	ti->ti_next = ti->ti_prev = 0;
473	ti->ti_x1 = 0;
474	ti->ti_len = htons((u_int16_t)tlen);
475	len = sizeof(struct ip ) + tlen;
476	/* keep checksum for ICMP reply
477	* ti->ti_sum = cksum(m, len);
478	* if (ti->ti_sum) { */
479	if(cksum(m, len)) {
480	tcpstat.tcps_rcvbadsum++;
481	goto drop;
482	}
483
484	/*
485	* Check that TCP offset makes sense,
486	* pull out TCP options and adjust length. XXX
487	*/
488	off = ti->ti_off << 2;
489	if (off < sizeof (struct tcphdr) \|\| off > tlen) {
490	tcpstat.tcps_rcvbadoff++;
491	goto drop;
492	}
493	tlen -= off;
494	ti->ti_len = tlen;
495	if (off > sizeof (struct tcphdr)) {
496	optlen = off - sizeof (struct tcphdr);
497	optp = mtod(m, caddr_t) + sizeof (struct tcpiphdr);
498
499	/*
500	* Do quick retrieval of timestamp options ("options
501	* prediction?"). If timestamp is the only option and it's
502	* formatted as recommended in RFC 1323 appendix A, we
503	* quickly get the values now and not bother calling
504	* tcp_dooptions(), etc.
505	*/
506	/* if ((optlen == TCPOLEN_TSTAMP_APPA \|\|
507	* (optlen > TCPOLEN_TSTAMP_APPA &&
508	* optp[TCPOLEN_TSTAMP_APPA] == TCPOPT_EOL)) &&
509	* (u_int32_t )optp == htonl(TCPOPT_TSTAMP_HDR) &&
510	* (ti->ti_flags & TH_SYN) == 0) {
511	* ts_present = 1;
512	* ts_val = ntohl((u_int32_t )(optp + 4));
513	* ts_ecr = ntohl((u_int32_t )(optp + 8));
514	* optp = NULL; / * we've parsed the options * /
515	* }
516	*/
517	}
518	tiflags = ti->ti_flags;
519
520	/*
521	* Convert TCP protocol specific fields to host format.
522	*/
523	NTOHL(ti->ti_seq);
524	NTOHL(ti->ti_ack);
525	NTOHS(ti->ti_win);
526	NTOHS(ti->ti_urp);
527
528	/*
529	* Drop TCP, IP headers and TCP options.
530	*/
531	m->m_data += sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
532	m->m_len -= sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
533
534	/*
535	* Locate pcb for segment.
536	*/
537	findso:
538	so = tcp_last_so;
539	if (so->so_fport != ti->ti_dport \|\|
540	so->so_lport != ti->ti_sport \|\|
541	so->so_laddr.s_addr != ti->ti_src.s_addr \|\|
542	so->so_faddr.s_addr != ti->ti_dst.s_addr) {
543	so = solookup(&tcb, ti->ti_src, ti->ti_sport,
544	ti->ti_dst, ti->ti_dport);
545	if (so)
546	tcp_last_so = so;
547	++tcpstat.tcps_socachemiss;
548	}
549
550	/*
551	* If the state is CLOSED (i.e., TCB does not exist) then
552	* all data in the incoming segment is discarded.
553	* If the TCB exists but is in CLOSED state, it is embryonic,
554	* but should either do a listen or a connect soon.
555	*
556	* state == CLOSED means we've done socreate() but haven't
557	* attached it to a protocol yet...
558	*
559	* XXX If a TCB does not exist, and the TH_SYN flag is
560	* the only flag set, then create a session, mark it
561	* as if it was LISTENING, and continue...
562	*/
563	if (so == 0) {
564	if ((tiflags & (TH_SYN\|TH_FIN\|TH_RST\|TH_URG\|TH_ACK)) != TH_SYN)
565	goto dropwithreset;
566
567	if ((so = socreate()) == NULL)
568	goto dropwithreset;
569	if (tcp_attach(pData, so) < 0) {
570	free(so); /* Not sofree (if it failed, it's not insqued) */
571	goto dropwithreset;
572	}
573
574	sbreserve(&so->so_snd, tcp_sndspace);
575	sbreserve(&so->so_rcv, tcp_rcvspace);
576
577	/* tcp_last_so = so; / / XXX ? */
578	/* tp = sototcpcb(so); */
579
580	so->so_laddr = ti->ti_src;
581	so->so_lport = ti->ti_sport;
582	so->so_faddr = ti->ti_dst;
583	so->so_fport = ti->ti_dport;
584
585	if ((so->so_iptos = tcp_tos(so)) == 0)
586	so->so_iptos = ((struct ip *)ti)->ip_tos;
587
588	tp = sototcpcb(so);
589	tp->t_state = TCPS_LISTEN;
590	}
591
592	/*
593	* If this is a still-connecting socket, this probably
594	* a retransmit of the SYN. Whether it's a retransmit SYN
595	* or something else, we nuke it.
596	*/
597	if (so->so_state & SS_ISFCONNECTING)
598	goto drop;
599
600	tp = sototcpcb(so);
601
602	/* XXX Should never fail */
603	if (tp == 0)
604	goto dropwithreset;
605	if (tp->t_state == TCPS_CLOSED)
606	goto drop;
607
608	/* Unscale the window into a 32-bit value. */
609	/* if ((tiflags & TH_SYN) == 0)
610	* tiwin = ti->ti_win << tp->snd_scale;
611	* else
612	*/
613	tiwin = ti->ti_win;
614
615	/*
616	* Segment received on connection.
617	* Reset idle time and keep-alive timer.
618	*/
619	tp->t_idle = 0;
620	if (so_options)
621	tp->t_timer[TCPT_KEEP] = tcp_keepintvl;
622	else
623	tp->t_timer[TCPT_KEEP] = tcp_keepidle;
624
625	/*
626	* Process options if not in LISTEN state,
627	* else do it below (after getting remote address).
628	*/
629	if (optp && tp->t_state != TCPS_LISTEN)
630	tcp_dooptions(pData, tp, (u_char *)optp, optlen, ti);
631	/* , */
632	/* &ts_present, &ts_val, &ts_ecr); */
633
634	/*
635	* Header prediction: check for the two common cases
636	* of a uni-directional data xfer. If the packet has
637	* no control flags, is in-sequence, the window didn't
638	* change and we're not retransmitting, it's a
639	* candidate. If the length is zero and the ack moved
640	* forward, we're the sender side of the xfer. Just
641	* free the data acked & wake any higher level process
642	* that was blocked waiting for space. If the length
643	* is non-zero and the ack didn't move, we're the
644	* receiver side. If we're getting packets in-order
645	* (the reassembly queue is empty), add the data to
646	* the socket buffer and note that we need a delayed ack.
647	*
648	* XXX Some of these tests are not needed
649	* eg: the tiwin == tp->snd_wnd prevents many more
650	* predictions.. with no real advantage..
651	*/
652	if (tp->t_state == TCPS_ESTABLISHED &&
653	(tiflags & (TH_SYN\|TH_FIN\|TH_RST\|TH_URG\|TH_ACK)) == TH_ACK &&
654	/* (!ts_present \|\| TSTMP_GEQ(ts_val, tp->ts_recent)) && */
655	ti->ti_seq == tp->rcv_nxt &&
656	tiwin && tiwin == tp->snd_wnd &&
657	tp->snd_nxt == tp->snd_max) {
658	/*
659	* If last ACK falls within this segment's sequence numbers,
660	* record the timestamp.
661	*/
662	/* if (ts_present && SEQ_LEQ(ti->ti_seq, tp->last_ack_sent) &&
663	* SEQ_LT(tp->last_ack_sent, ti->ti_seq + ti->ti_len)) {
664	* tp->ts_recent_age = tcp_now;
665	* tp->ts_recent = ts_val;
666	* }
667	*/
668	if (ti->ti_len == 0) {
669	if (SEQ_GT(ti->ti_ack, tp->snd_una) &&
670	SEQ_LEQ(ti->ti_ack, tp->snd_max) &&
671	tp->snd_cwnd >= tp->snd_wnd) {
672	/*
673	* this is a pure ack for outstanding data.
674	*/
675	++tcpstat.tcps_predack;
676	/* if (ts_present)
677	* tcp_xmit_timer(tp, tcp_now-ts_ecr+1);
678	* else
679	*/ if (tp->t_rtt &&
680	SEQ_GT(ti->ti_ack, tp->t_rtseq))
681	tcp_xmit_timer(pData, tp, tp->t_rtt);
682	acked = ti->ti_ack - tp->snd_una;
683	tcpstat.tcps_rcvackpack++;
684	tcpstat.tcps_rcvackbyte += acked;
685	sbdrop(&so->so_snd, acked);
686	tp->snd_una = ti->ti_ack;
687	m_freem(pData, m);
688
689	/*
690	* If all outstanding data are acked, stop
691	* retransmit timer, otherwise restart timer
692	* using current (possibly backed-off) value.
693	* If process is waiting for space,
694	* wakeup/selwakeup/signal. If data
695	* are ready to send, let tcp_output
696	* decide between more output or persist.
697	*/
698	if (tp->snd_una == tp->snd_max)
699	tp->t_timer[TCPT_REXMT] = 0;
700	else if (tp->t_timer[TCPT_PERSIST] == 0)
701	tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
702
703	/*
704	* There's room in so_snd, sowwakup will read()
705	* from the socket if we can
706	*/
707	/* if (so->so_snd.sb_flags & SB_NOTIFY)
708	* sowwakeup(so);
709	*/
710	/*
711	* This is called because sowwakeup might have
712	* put data into so_snd. Since we don't so sowwakeup,
713	* we don't need this.. XXX???
714	*/
715	if (so->so_snd.sb_cc)
716	(void) tcp_output(pData, tp);
717
718	return;
719	}
720	} else if (ti->ti_ack == tp->snd_una &&
721	#ifndef VBOX_WITH_BSD_TCP_REASS
722	u32_to_ptr(pData, tp->seg_next, struct tcpcb *) == tp &&
723	#else /* VBOX_WITH_BSD_TCP_REASS */
724	LIST_FIRST(&tp->t_segq) &&
725	#endif /* VBOX_WITH_BSD_TCP_REASS */
726	ti->ti_len <= sbspace(&so->so_rcv)) {
727	/*
728	* this is a pure, in-sequence data packet
729	* with nothing on the reassembly queue and
730	* we have enough buffer space to take it.
731	*/
732	++tcpstat.tcps_preddat;
733	tp->rcv_nxt += ti->ti_len;
734	tcpstat.tcps_rcvpack++;
735	tcpstat.tcps_rcvbyte += ti->ti_len;
736	/*
737	* Add data to socket buffer.
738	*/
739	if (so->so_emu) {
740	if (tcp_emu(pData, so,m)) sbappend(pData, so, m);
741	} else
742	sbappend(pData, so, m);
743
744	/*
745	* XXX This is called when data arrives. Later, check
746	* if we can actually write() to the socket
747	* XXX Need to check? It's be NON_BLOCKING
748	*/
749	/* sorwakeup(so); */
750
751	/*
752	* If this is a short packet, then ACK now - with Nagel
753	* congestion avoidance sender won't send more until
754	* he gets an ACK.
755	*
756	* It is better to not delay acks at all to maximize
757	* TCP throughput. See RFC 2581.
758	*/
759	tp->t_flags \|= TF_ACKNOW;
760	tcp_output(pData, tp);
761	return;
762	}
763	} /* header prediction */
764	/*
765	* Calculate amount of space in receive window,
766	* and then do TCP input processing.
767	* Receive window is amount of space in rcv queue,
768	* but not less than advertised window.
769	*/
770	{ int win;
771	win = sbspace(&so->so_rcv);
772	if (win < 0)
773	win = 0;
774	tp->rcv_wnd = max(win, (int)(tp->rcv_adv - tp->rcv_nxt));
775	}
776
777	switch (tp->t_state) {
778
779	/*
780	* If the state is LISTEN then ignore segment if it contains an RST.
781	* If the segment contains an ACK then it is bad and send a RST.
782	* If it does not contain a SYN then it is not interesting; drop it.
783	* Don't bother responding if the destination was a broadcast.
784	* Otherwise initialize tp->rcv_nxt, and tp->irs, select an initial
785	* tp->iss, and send a segment:
786	* <SEQ=ISS><ACK=RCV_NXT><CTL=SYN,ACK>
787	* Also initialize tp->snd_nxt to tp->iss+1 and tp->snd_una to tp->iss.
788	* Fill in remote peer address fields if not previously specified.
789	* Enter SYN_RECEIVED state, and process any other fields of this
790	* segment in this state.
791	*/
792	case TCPS_LISTEN: {
793
794	if (tiflags & TH_RST)
795	goto drop;
796	if (tiflags & TH_ACK)
797	goto dropwithreset;
798	if ((tiflags & TH_SYN) == 0)
799	goto drop;
800
801	/*
802	* This has way too many gotos...
803	* But a bit of spaghetti code never hurt anybody :)
804	*/
805
806	if (so->so_emu & EMU_NOCONNECT) {
807	so->so_emu &= ~EMU_NOCONNECT;
808	goto cont_input;
809	}
810
811	if((tcp_fconnect(pData, so) == -1) && (errno != EINPROGRESS) && (errno != EWOULDBLOCK)) {
812	u_char code=ICMP_UNREACH_NET;
813	DEBUG_MISC((dfd," tcp fconnect errno = %d-%s\n",
814	errno,strerror(errno)));
815	if(errno == ECONNREFUSED) {
816	/* ACK the SYN, send RST to refuse the connection */
817	tcp_respond(pData, tp, ti, m, ti->ti_seq+1, (tcp_seq)0,
818	TH_RST\|TH_ACK);
819	} else {
820	if(errno == EHOSTUNREACH) code=ICMP_UNREACH_HOST;
821	HTONL(ti->ti_seq); /* restore tcp header */
822	HTONL(ti->ti_ack);
823	HTONS(ti->ti_win);
824	HTONS(ti->ti_urp);
825	m->m_data -= sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
826	m->m_len += sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
827	*ip=save_ip;
828	icmp_error(pData, m, ICMP_UNREACH,code, 0,strerror(errno));
829	}
830	tp = tcp_close(pData, tp);
831	m_free(pData, m);
832	} else {
833	/*
834	* Haven't connected yet, save the current mbuf
835	* and ti, and return
836	* XXX Some OS's don't tell us whether the connect()
837	* succeeded or not. So we must time it out.
838	*/
839	so->so_m = m;
840	so->so_ti = ti;
841	tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT;
842	tp->t_state = TCPS_SYN_RECEIVED;
843	}
844	return;
845
846	cont_conn:
847	/* m==NULL
848	* Check if the connect succeeded
849	*/
850	if (so->so_state & SS_NOFDREF) {
851	tp = tcp_close(pData, tp);
852	goto dropwithreset;
853	}
854	cont_input:
855	tcp_template(tp);
856
857	if (optp)
858	tcp_dooptions(pData, tp, (u_char *)optp, optlen, ti);
859	/* , */
860	/* &ts_present, &ts_val, &ts_ecr); */
861
862	if (iss)
863	tp->iss = iss;
864	else
865	tp->iss = tcp_iss;
866	tcp_iss += TCP_ISSINCR/2;
867	tp->irs = ti->ti_seq;
868	tcp_sendseqinit(tp);
869	tcp_rcvseqinit(tp);
870	tp->t_flags \|= TF_ACKNOW;
871	tp->t_state = TCPS_SYN_RECEIVED;
872	tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT;
873	tcpstat.tcps_accepts++;
874	goto trimthenstep6;
875	} /* case TCPS_LISTEN */
876
877	/*
878	* If the state is SYN_SENT:
879	* if seg contains an ACK, but not for our SYN, drop the input.
880	* if seg contains a RST, then drop the connection.
881	* if seg does not contain SYN, then drop it.
882	* Otherwise this is an acceptable SYN segment
883	* initialize tp->rcv_nxt and tp->irs
884	* if seg contains ack then advance tp->snd_una
885	* if SYN has been acked change to ESTABLISHED else SYN_RCVD state
886	* arrange for segment to be acked (eventually)
887	* continue processing rest of data/controls, beginning with URG
888	*/
889	case TCPS_SYN_SENT:
890	if ((tiflags & TH_ACK) &&
891	(SEQ_LEQ(ti->ti_ack, tp->iss) \|\|
892	SEQ_GT(ti->ti_ack, tp->snd_max)))
893	goto dropwithreset;
894
895	if (tiflags & TH_RST) {
896	if (tiflags & TH_ACK)
897	tp = tcp_drop(pData, tp,0); /* XXX Check t_softerror! */
898	goto drop;
899	}
900
901	if ((tiflags & TH_SYN) == 0)
902	goto drop;
903	if (tiflags & TH_ACK) {
904	tp->snd_una = ti->ti_ack;
905	if (SEQ_LT(tp->snd_nxt, tp->snd_una))
906	tp->snd_nxt = tp->snd_una;
907	}
908
909	tp->t_timer[TCPT_REXMT] = 0;
910	tp->irs = ti->ti_seq;
911	tcp_rcvseqinit(tp);
912	tp->t_flags \|= TF_ACKNOW;
913	if (tiflags & TH_ACK && SEQ_GT(tp->snd_una, tp->iss)) {
914	tcpstat.tcps_connects++;
915	soisfconnected(so);
916	tp->t_state = TCPS_ESTABLISHED;
917
918	/* Do window scaling on this connection? */
919	/* if ((tp->t_flags & (TF_RCVD_SCALE\|TF_REQ_SCALE)) ==
920	* (TF_RCVD_SCALE\|TF_REQ_SCALE)) {
921	* tp->snd_scale = tp->requested_s_scale;
922	* tp->rcv_scale = tp->request_r_scale;
923	* }
924	*/
925	#ifndef VBOX_WITH_BSD_TCP_REASS
926	(void) tcp_reass(pData, tp, (struct tcpiphdr *)0,
927	(struct mbuf *)0);
928	#else /* VBOX_WITH_BSD_TCP_REASS */
929	(void) tcp_reass(pData, tp, (struct tcphdr )0, NULL, (struct mbuf )0);
930	#endif /* VBOX_WITH_BSD_TCP_REASS */
931	/*
932	* if we didn't have to retransmit the SYN,
933	* use its rtt as our initial srtt & rtt var.
934	*/
935	if (tp->t_rtt)
936	tcp_xmit_timer(pData, tp, tp->t_rtt);
937	} else
938	tp->t_state = TCPS_SYN_RECEIVED;
939
940	trimthenstep6:
941	/*
942	* Advance ti->ti_seq to correspond to first data byte.
943	* If data, trim to stay within window,
944	* dropping FIN if necessary.
945	*/
946	ti->ti_seq++;
947	if (ti->ti_len > tp->rcv_wnd) {
948	todrop = ti->ti_len - tp->rcv_wnd;
949	m_adj(m, -todrop);
950	ti->ti_len = tp->rcv_wnd;
951	tiflags &= ~TH_FIN;
952	tcpstat.tcps_rcvpackafterwin++;
953	tcpstat.tcps_rcvbyteafterwin += todrop;
954	}
955	tp->snd_wl1 = ti->ti_seq - 1;
956	tp->rcv_up = ti->ti_seq;
957	goto step6;
958	} /* switch tp->t_state */
959	/*
960	* States other than LISTEN or SYN_SENT.
961	* First check timestamp, if present.
962	* Then check that at least some bytes of segment are within
963	* receive window. If segment begins before rcv_nxt,
964	* drop leading data (and SYN); if nothing left, just ack.
965	*
966	* RFC 1323 PAWS: If we have a timestamp reply on this segment
967	* and it's less than ts_recent, drop it.
968	*/
969	/* if (ts_present && (tiflags & TH_RST) == 0 && tp->ts_recent &&
970	* TSTMP_LT(ts_val, tp->ts_recent)) {
971	*
972	/ / Check to see if ts_recent is over 24 days old. */
973	/* if ((int)(tcp_now - tp->ts_recent_age) > TCP_PAWS_IDLE) {
974	/ /
975	* * Invalidate ts_recent. If this segment updates
976	* * ts_recent, the age will be reset later and ts_recent
977	* * will get a valid value. If it does not, setting
978	* * ts_recent to zero will at least satisfy the
979	* * requirement that zero be placed in the timestamp
980	* * echo reply when ts_recent isn't valid. The
981	* * age isn't reset until we get a valid ts_recent
982	* * because we don't want out-of-order segments to be
983	* * dropped when ts_recent is old.
984	* */
985	/* tp->ts_recent = 0;
986	* } else {
987	* tcpstat.tcps_rcvduppack++;
988	* tcpstat.tcps_rcvdupbyte += ti->ti_len;
989	* tcpstat.tcps_pawsdrop++;
990	* goto dropafterack;
991	* }
992	* }
993	*/
994
995	todrop = tp->rcv_nxt - ti->ti_seq;
996	if (todrop > 0) {
997	if (tiflags & TH_SYN) {
998	tiflags &= ~TH_SYN;
999	ti->ti_seq++;
1000	if (ti->ti_urp > 1)
1001	ti->ti_urp--;
1002	else
1003	tiflags &= ~TH_URG;
1004	todrop--;
1005	}
1006	/*
1007	* Following if statement from Stevens, vol. 2, p. 960.
1008	*/
1009	if (todrop > ti->ti_len
1010	\|\| (todrop == ti->ti_len && (tiflags & TH_FIN) == 0)) {
1011	/*
1012	* Any valid FIN must be to the left of the window.
1013	* At this point the FIN must be a duplicate or out
1014	* of sequence; drop it.
1015	*/
1016	tiflags &= ~TH_FIN;
1017
1018	/*
1019	* Send an ACK to resynchronize and drop any data.
1020	* But keep on processing for RST or ACK.
1021	*/
1022	tp->t_flags \|= TF_ACKNOW;
1023	todrop = ti->ti_len;
1024	tcpstat.tcps_rcvduppack++;
1025	tcpstat.tcps_rcvdupbyte += todrop;
1026	} else {
1027	tcpstat.tcps_rcvpartduppack++;
1028	tcpstat.tcps_rcvpartdupbyte += todrop;
1029	}
1030	m_adj(m, todrop);
1031	ti->ti_seq += todrop;
1032	ti->ti_len -= todrop;
1033	if (ti->ti_urp > todrop)
1034	ti->ti_urp -= todrop;
1035	else {
1036	tiflags &= ~TH_URG;
1037	ti->ti_urp = 0;
1038	}
1039	}
1040	/*
1041	* If new data are received on a connection after the
1042	* user processes are gone, then RST the other end.
1043	*/
1044	if ((so->so_state & SS_NOFDREF) &&
1045	tp->t_state > TCPS_CLOSE_WAIT && ti->ti_len) {
1046	tp = tcp_close(pData, tp);
1047	tcpstat.tcps_rcvafterclose++;
1048	goto dropwithreset;
1049	}
1050
1051	/*
1052	* If segment ends after window, drop trailing data
1053	* (and PUSH and FIN); if nothing left, just ACK.
1054	*/
1055	todrop = (ti->ti_seq+ti->ti_len) - (tp->rcv_nxt+tp->rcv_wnd);
1056	if (todrop > 0) {
1057	tcpstat.tcps_rcvpackafterwin++;
1058	if (todrop >= ti->ti_len) {
1059	tcpstat.tcps_rcvbyteafterwin += ti->ti_len;
1060	/*
1061	* If a new connection request is received
1062	* while in TIME_WAIT, drop the old connection
1063	* and start over if the sequence numbers
1064	* are above the previous ones.
1065	*/
1066	if (tiflags & TH_SYN &&
1067	tp->t_state == TCPS_TIME_WAIT &&
1068	SEQ_GT(ti->ti_seq, tp->rcv_nxt)) {
1069	iss = tp->rcv_nxt + TCP_ISSINCR;
1070	tp = tcp_close(pData, tp);
1071	goto findso;
1072	}
1073	/*
1074	* If window is closed can only take segments at
1075	* window edge, and have to drop data and PUSH from
1076	* incoming segments. Continue processing, but
1077	* remember to ack. Otherwise, drop segment
1078	* and ack.
1079	*/
1080	if (tp->rcv_wnd == 0 && ti->ti_seq == tp->rcv_nxt) {
1081	tp->t_flags \|= TF_ACKNOW;
1082	tcpstat.tcps_rcvwinprobe++;
1083	} else
1084	goto dropafterack;
1085	} else
1086	tcpstat.tcps_rcvbyteafterwin += todrop;
1087	m_adj(m, -todrop);
1088	ti->ti_len -= todrop;
1089	tiflags &= ~(TH_PUSH\|TH_FIN);
1090	}
1091
1092	/*
1093	* If last ACK falls within this segment's sequence numbers,
1094	* record its timestamp.
1095	*/
1096	/* if (ts_present && SEQ_LEQ(ti->ti_seq, tp->last_ack_sent) &&
1097	* SEQ_LT(tp->last_ack_sent, ti->ti_seq + ti->ti_len +
1098	* ((tiflags & (TH_SYN\|TH_FIN)) != 0))) {
1099	* tp->ts_recent_age = tcp_now;
1100	* tp->ts_recent = ts_val;
1101	* }
1102	*/
1103
1104	/*
1105	* If the RST bit is set examine the state:
1106	* SYN_RECEIVED STATE:
1107	* If passive open, return to LISTEN state.
1108	* If active open, inform user that connection was refused.
1109	* ESTABLISHED, FIN_WAIT_1, FIN_WAIT2, CLOSE_WAIT STATES:
1110	* Inform user that connection was reset, and close tcb.
1111	* CLOSING, LAST_ACK, TIME_WAIT STATES
1112	* Close the tcb.
1113	*/
1114	if (tiflags&TH_RST) switch (tp->t_state) {
1115
1116	case TCPS_SYN_RECEIVED:
1117	/* so->so_error = ECONNREFUSED; */
1118	goto close;
1119
1120	case TCPS_ESTABLISHED:
1121	case TCPS_FIN_WAIT_1:
1122	case TCPS_FIN_WAIT_2:
1123	case TCPS_CLOSE_WAIT:
1124	/* so->so_error = ECONNRESET; */
1125	close:
1126	tp->t_state = TCPS_CLOSED;
1127	tcpstat.tcps_drops++;
1128	tp = tcp_close(pData, tp);
1129	goto drop;
1130
1131	case TCPS_CLOSING:
1132	case TCPS_LAST_ACK:
1133	case TCPS_TIME_WAIT:
1134	tp = tcp_close(pData, tp);
1135	goto drop;
1136	}
1137
1138	/*
1139	* If a SYN is in the window, then this is an
1140	* error and we send an RST and drop the connection.
1141	*/
1142	if (tiflags & TH_SYN) {
1143	tp = tcp_drop(pData, tp,0);
1144	goto dropwithreset;
1145	}
1146
1147	/*
1148	* If the ACK bit is off we drop the segment and return.
1149	*/
1150	if ((tiflags & TH_ACK) == 0) goto drop;
1151
1152	/*
1153	* Ack processing.
1154	*/
1155	switch (tp->t_state) {
1156	/*
1157	* In SYN_RECEIVED state if the ack ACKs our SYN then enter
1158	* ESTABLISHED state and continue processing, otherwise
1159	* send an RST. una<=ack<=max
1160	*/
1161	case TCPS_SYN_RECEIVED:
1162
1163	if (SEQ_GT(tp->snd_una, ti->ti_ack) \|\|
1164	SEQ_GT(ti->ti_ack, tp->snd_max))
1165	goto dropwithreset;
1166	tcpstat.tcps_connects++;
1167	tp->t_state = TCPS_ESTABLISHED;
1168	/*
1169	* The sent SYN is ack'ed with our sequence number +1
1170	* The first data byte already in the buffer will get
1171	* lost if no correction is made. This is only needed for
1172	* SS_CTL since the buffer is empty otherwise.
1173	* tp->snd_una++; or:
1174	*/
1175	tp->snd_una=ti->ti_ack;
1176	soisfconnected(so);
1177
1178	/* Do window scaling? */
1179	/* if ((tp->t_flags & (TF_RCVD_SCALE\|TF_REQ_SCALE)) ==
1180	* (TF_RCVD_SCALE\|TF_REQ_SCALE)) {
1181	* tp->snd_scale = tp->requested_s_scale;
1182	* tp->rcv_scale = tp->request_r_scale;
1183	* }
1184	*/
1185	#ifndef VBOX_WITH_BSD_TCP_REASS
1186	(void) tcp_reass(pData, tp, (struct tcpiphdr )0, (struct mbuf )0);
1187	#else /* VBOX_WITH_BSD_TCP_REASS */
1188	(void) tcp_reass(pData, tp, (struct tcphdr )0, (int )0, (struct mbuf *)0);
1189	#endif /VBOX_WITH_BSD_TCP_REASS/
1190	tp->snd_wl1 = ti->ti_seq - 1;
1191	/* Avoid ack processing; snd_una==ti_ack => dup ack */
1192	goto synrx_to_est;
1193	/* fall into ... */
1194
1195	/*
1196	* In ESTABLISHED state: drop duplicate ACKs; ACK out of range
1197	* ACKs. If the ack is in the range
1198	* tp->snd_una < ti->ti_ack <= tp->snd_max
1199	* then advance tp->snd_una to ti->ti_ack and drop
1200	* data from the retransmission queue. If this ACK reflects
1201	* more up to date window information we update our window information.
1202	*/
1203	case TCPS_ESTABLISHED:
1204	case TCPS_FIN_WAIT_1:
1205	case TCPS_FIN_WAIT_2:
1206	case TCPS_CLOSE_WAIT:
1207	case TCPS_CLOSING:
1208	case TCPS_LAST_ACK:
1209	case TCPS_TIME_WAIT:
1210
1211	if (SEQ_LEQ(ti->ti_ack, tp->snd_una)) {
1212	if (ti->ti_len == 0 && tiwin == tp->snd_wnd) {
1213	tcpstat.tcps_rcvdupack++;
1214	DEBUG_MISC((dfd," dup ack m = %lx so = %lx \n",
1215	(long )m, (long )so));
1216	/*
1217	* If we have outstanding data (other than
1218	* a window probe), this is a completely
1219	* duplicate ack (ie, window info didn't
1220	* change), the ack is the biggest we've
1221	* seen and we've seen exactly our rexmt
1222	* threshold of them, assume a packet
1223	* has been dropped and retransmit it.
1224	* Kludge snd_nxt & the congestion
1225	* window so we send only this one
1226	* packet.
1227	*
1228	* We know we're losing at the current
1229	* window size so do congestion avoidance
1230	* (set ssthresh to half the current window
1231	* and pull our congestion window back to
1232	* the new ssthresh).
1233	*
1234	* Dup acks mean that packets have left the
1235	* network (they're now cached at the receiver)
1236	* so bump cwnd by the amount in the receiver
1237	* to keep a constant cwnd packets in the
1238	* network.
1239	*/
1240	if (tp->t_timer[TCPT_REXMT] == 0 \|\|
1241	ti->ti_ack != tp->snd_una)
1242	tp->t_dupacks = 0;
1243	else if (++tp->t_dupacks == tcprexmtthresh) {
1244	tcp_seq onxt = tp->snd_nxt;
1245	u_int win =
1246	min(tp->snd_wnd, tp->snd_cwnd) / 2 /
1247	tp->t_maxseg;
1248
1249	if (win < 2)
1250	win = 2;
1251	tp->snd_ssthresh = win * tp->t_maxseg;
1252	tp->t_timer[TCPT_REXMT] = 0;
1253	tp->t_rtt = 0;
1254	tp->snd_nxt = ti->ti_ack;
1255	tp->snd_cwnd = tp->t_maxseg;
1256	(void) tcp_output(pData, tp);
1257	tp->snd_cwnd = tp->snd_ssthresh +
1258	tp->t_maxseg * tp->t_dupacks;
1259	if (SEQ_GT(onxt, tp->snd_nxt))
1260	tp->snd_nxt = onxt;
1261	goto drop;
1262	} else if (tp->t_dupacks > tcprexmtthresh) {
1263	tp->snd_cwnd += tp->t_maxseg;
1264	(void) tcp_output(pData, tp);
1265	goto drop;
1266	}
1267	} else
1268	tp->t_dupacks = 0;
1269	break;
1270	}
1271	synrx_to_est:
1272	/*
1273	* If the congestion window was inflated to account
1274	* for the other side's cached packets, retract it.
1275	*/
1276	if (tp->t_dupacks > tcprexmtthresh &&
1277	tp->snd_cwnd > tp->snd_ssthresh)
1278	tp->snd_cwnd = tp->snd_ssthresh;
1279	tp->t_dupacks = 0;
1280	if (SEQ_GT(ti->ti_ack, tp->snd_max)) {
1281	tcpstat.tcps_rcvacktoomuch++;
1282	goto dropafterack;
1283	}
1284	acked = ti->ti_ack - tp->snd_una;
1285	tcpstat.tcps_rcvackpack++;
1286	tcpstat.tcps_rcvackbyte += acked;
1287
1288	/*
1289	* If we have a timestamp reply, update smoothed
1290	* round trip time. If no timestamp is present but
1291	* transmit timer is running and timed sequence
1292	* number was acked, update smoothed round trip time.
1293	* Since we now have an rtt measurement, cancel the
1294	* timer backoff (cf., Phil Karn's retransmit alg.).
1295	* Recompute the initial retransmit timer.
1296	*/
1297	/* if (ts_present)
1298	* tcp_xmit_timer(tp, tcp_now-ts_ecr+1);
1299	* else
1300	*/
1301	if (tp->t_rtt && SEQ_GT(ti->ti_ack, tp->t_rtseq))
1302	tcp_xmit_timer(pData, tp,tp->t_rtt);
1303
1304	/*
1305	* If all outstanding data is acked, stop retransmit
1306	* timer and remember to restart (more output or persist).
1307	* If there is more data to be acked, restart retransmit
1308	* timer, using current (possibly backed-off) value.
1309	*/
1310	if (ti->ti_ack == tp->snd_max) {
1311	tp->t_timer[TCPT_REXMT] = 0;
1312	needoutput = 1;
1313	} else if (tp->t_timer[TCPT_PERSIST] == 0)
1314	tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
1315	/*
1316	* When new data is acked, open the congestion window.
1317	* If the window gives us less than ssthresh packets
1318	* in flight, open exponentially (maxseg per packet).
1319	* Otherwise open linearly: maxseg per window
1320	* (maxseg^2 / cwnd per packet).
1321	*/
1322	{
1323	register u_int cw = tp->snd_cwnd;
1324	register u_int incr = tp->t_maxseg;
1325
1326	if (cw > tp->snd_ssthresh)
1327	incr = incr * incr / cw;
1328	tp->snd_cwnd = min(cw + incr, TCP_MAXWIN<<tp->snd_scale);
1329	}
1330	if (acked > so->so_snd.sb_cc) {
1331	tp->snd_wnd -= so->so_snd.sb_cc;
1332	sbdrop(&so->so_snd, (int )so->so_snd.sb_cc);
1333	ourfinisacked = 1;
1334	} else {
1335	sbdrop(&so->so_snd, acked);
1336	tp->snd_wnd -= acked;
1337	ourfinisacked = 0;
1338	}
1339	/*
1340	* XXX sowwakup is called when data is acked and there's room for
1341	* for more data... it should read() the socket
1342	*/
1343	/* if (so->so_snd.sb_flags & SB_NOTIFY)
1344	* sowwakeup(so);
1345	*/
1346	tp->snd_una = ti->ti_ack;
1347	if (SEQ_LT(tp->snd_nxt, tp->snd_una))
1348	tp->snd_nxt = tp->snd_una;
1349
1350	switch (tp->t_state) {
1351
1352	/*
1353	* In FIN_WAIT_1 STATE in addition to the processing
1354	* for the ESTABLISHED state if our FIN is now acknowledged
1355	* then enter FIN_WAIT_2.
1356	*/
1357	case TCPS_FIN_WAIT_1:
1358	if (ourfinisacked) {
1359	/*
1360	* If we can't receive any more
1361	* data, then closing user can proceed.
1362	* Starting the timer is contrary to the
1363	* specification, but if we don't get a FIN
1364	* we'll hang forever.
1365	*/
1366	if (so->so_state & SS_FCANTRCVMORE) {
1367	soisfdisconnected(so);
1368	tp->t_timer[TCPT_2MSL] = tcp_maxidle;
1369	}
1370	tp->t_state = TCPS_FIN_WAIT_2;
1371	}
1372	break;
1373
1374	/*
1375	* In CLOSING STATE in addition to the processing for
1376	* the ESTABLISHED state if the ACK acknowledges our FIN
1377	* then enter the TIME-WAIT state, otherwise ignore
1378	* the segment.
1379	*/
1380	case TCPS_CLOSING:
1381	if (ourfinisacked) {
1382	tp->t_state = TCPS_TIME_WAIT;
1383	tcp_canceltimers(tp);
1384	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1385	soisfdisconnected(so);
1386	}
1387	break;
1388
1389	/*
1390	* In LAST_ACK, we may still be waiting for data to drain
1391	* and/or to be acked, as well as for the ack of our FIN.
1392	* If our FIN is now acknowledged, delete the TCB,
1393	* enter the closed state and return.
1394	*/
1395	case TCPS_LAST_ACK:
1396	if (ourfinisacked) {
1397	tp = tcp_close(pData, tp);
1398	goto drop;
1399	}
1400	break;
1401
1402	/*
1403	* In TIME_WAIT state the only thing that should arrive
1404	* is a retransmission of the remote FIN. Acknowledge
1405	* it and restart the finack timer.
1406	*/
1407	case TCPS_TIME_WAIT:
1408	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1409	goto dropafterack;
1410	}
1411	} /* switch(tp->t_state) */
1412
1413	step6:
1414	/*
1415	* Update window information.
1416	* Don't look at window if no ACK: TAC's send garbage on first SYN.
1417	*/
1418	if ((tiflags & TH_ACK) &&
1419	(SEQ_LT(tp->snd_wl1, ti->ti_seq) \|\|
1420	(tp->snd_wl1 == ti->ti_seq && (SEQ_LT(tp->snd_wl2, ti->ti_ack) \|\|
1421	(tp->snd_wl2 == ti->ti_ack && tiwin > tp->snd_wnd))))) {
1422	/* keep track of pure window updates */
1423	if (ti->ti_len == 0 &&
1424	tp->snd_wl2 == ti->ti_ack && tiwin > tp->snd_wnd)
1425	tcpstat.tcps_rcvwinupd++;
1426	tp->snd_wnd = tiwin;
1427	tp->snd_wl1 = ti->ti_seq;
1428	tp->snd_wl2 = ti->ti_ack;
1429	if (tp->snd_wnd > tp->max_sndwnd)
1430	tp->max_sndwnd = tp->snd_wnd;
1431	needoutput = 1;
1432	}
1433
1434	/*
1435	* Process segments with URG.
1436	*/
1437	if ((tiflags & TH_URG) && ti->ti_urp &&
1438	TCPS_HAVERCVDFIN(tp->t_state) == 0) {
1439	/*
1440	* This is a kludge, but if we receive and accept
1441	* random urgent pointers, we'll crash in
1442	* soreceive. It's hard to imagine someone
1443	* actually wanting to send this much urgent data.
1444	*/
1445	if (ti->ti_urp + so->so_rcv.sb_cc > so->so_rcv.sb_datalen) {
1446	ti->ti_urp = 0;
1447	tiflags &= ~TH_URG;
1448	goto dodata;
1449	}
1450	/*
1451	* If this segment advances the known urgent pointer,
1452	* then mark the data stream. This should not happen
1453	* in CLOSE_WAIT, CLOSING, LAST_ACK or TIME_WAIT STATES since
1454	* a FIN has been received from the remote side.
1455	* In these states we ignore the URG.
1456	*
1457	* According to RFC961 (Assigned Protocols),
1458	* the urgent pointer points to the last octet
1459	* of urgent data. We continue, however,
1460	* to consider it to indicate the first octet
1461	* of data past the urgent section as the original
1462	* spec states (in one of two places).
1463	*/
1464	if (SEQ_GT(ti->ti_seq+ti->ti_urp, tp->rcv_up)) {
1465	tp->rcv_up = ti->ti_seq + ti->ti_urp;
1466	so->so_urgc = so->so_rcv.sb_cc +
1467	(tp->rcv_up - tp->rcv_nxt); /* -1; */
1468	tp->rcv_up = ti->ti_seq + ti->ti_urp;
1469
1470	}
1471	} else
1472	/*
1473	* If no out of band data is expected,
1474	* pull receive urgent pointer along
1475	* with the receive window.
1476	*/
1477	if (SEQ_GT(tp->rcv_nxt, tp->rcv_up))
1478	tp->rcv_up = tp->rcv_nxt;
1479	dodata:
1480
1481	/*
1482	* If this is a small packet, then ACK now - with Nagel
1483	* congestion avoidance sender won't send more until
1484	* he gets an ACK.
1485	*
1486	* See above.
1487	*/
1488	if (ti->ti_len && (unsigned)ti->ti_len <= 5 &&
1489	((struct tcpiphdr_2 *)ti)->first_char == (char)27) {
1490	tp->t_flags \|= TF_ACKNOW;
1491	}
1492
1493	/*
1494	* Process the segment text, merging it into the TCP sequencing queue,
1495	* and arranging for acknowledgment of receipt if necessary.
1496	* This process logically involves adjusting tp->rcv_wnd as data
1497	* is presented to the user (this happens in tcp_usrreq.c,
1498	* case PRU_RCVD). If a FIN has already been received on this
1499	* connection then we just ignore the text.
1500	*/
1501	if ((ti->ti_len \|\| (tiflags&TH_FIN)) &&
1502	TCPS_HAVERCVDFIN(tp->t_state) == 0) {
1503	#ifndef VBOX_WITH_BSD_TCP_REASS
1504	TCP_REASS(pData, tp, ti, m, so, tiflags);
1505	#else /* VBOX_WITH_BSD_TCP_REASS */
1506	if (ti->ti_seq == tp->rcv_nxt
1507	&& LIST_EMPTY(&tp->t_segq)
1508	&& tp->t_state == TCPS_ESTABLISHED) {
1509	DELAY_ACK(tp, ti); /* little bit different from BSD declaration see netinet/tcp_input.c */
1510	tp->rcv_nxt += tlen;
1511	tiflags = ti->ti_t.th_flags & TH_FIN;
1512	tcpstat.tcps_rcvpack++;
1513	tcpstat.tcps_rcvbyte += tlen;
1514	if (so->so_state & SS_FCANTRCVMORE)
1515	m_freem(pData, m);
1516	else
1517	sbappend(pData, so, m);
1518	}
1519	else {
1520	tiflags = tcp_reass(pData, tp, &ti->ti_t, &tlen, m);
1521	tiflags \|= TF_ACKNOW;
1522	}
1523	#endif /* VBOX_WITH_BSD_TCP_REASS */
1524	/*
1525	* Note the amount of data that peer has sent into
1526	* our window, in order to estimate the sender's
1527	* buffer size.
1528	*/
1529	len = so->so_rcv.sb_datalen - (tp->rcv_adv - tp->rcv_nxt);
1530	} else {
1531	m_free(pData, m);
1532	tiflags &= ~TH_FIN;
1533	}
1534
1535	/*
1536	* If FIN is received ACK the FIN and let the user know
1537	* that the connection is closing.
1538	*/
1539	if (tiflags & TH_FIN) {
1540	if (TCPS_HAVERCVDFIN(tp->t_state) == 0) {
1541	/*
1542	* If we receive a FIN we can't send more data,
1543	* set it SS_FDRAIN
1544	* Shutdown the socket if there is no rx data in the
1545	* buffer.
1546	* soread() is called on completion of shutdown() and
1547	* will got to TCPS_LAST_ACK, and use tcp_output()
1548	* to send the FIN.
1549	*/
1550	/* sofcantrcvmore(so); */
1551	sofwdrain(so);
1552
1553	tp->t_flags \|= TF_ACKNOW;
1554	tp->rcv_nxt++;
1555	}
1556	switch (tp->t_state) {
1557
1558	/*
1559	* In SYN_RECEIVED and ESTABLISHED STATES
1560	* enter the CLOSE_WAIT state.
1561	*/
1562	case TCPS_SYN_RECEIVED:
1563	case TCPS_ESTABLISHED:
1564	if(so->so_emu == EMU_CTL) /* no shutdown on socket */
1565	tp->t_state = TCPS_LAST_ACK;
1566	else
1567	tp->t_state = TCPS_CLOSE_WAIT;
1568	break;
1569
1570	/*
1571	* If still in FIN_WAIT_1 STATE FIN has not been acked so
1572	* enter the CLOSING state.
1573	*/
1574	case TCPS_FIN_WAIT_1:
1575	tp->t_state = TCPS_CLOSING;
1576	break;
1577
1578	/*
1579	* In FIN_WAIT_2 state enter the TIME_WAIT state,
1580	* starting the time-wait timer, turning off the other
1581	* standard timers.
1582	*/
1583	case TCPS_FIN_WAIT_2:
1584	tp->t_state = TCPS_TIME_WAIT;
1585	tcp_canceltimers(tp);
1586	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1587	soisfdisconnected(so);
1588	break;
1589
1590	/*
1591	* In TIME_WAIT state restart the 2 MSL time_wait timer.
1592	*/
1593	case TCPS_TIME_WAIT:
1594	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1595	break;
1596	}
1597	}
1598
1599	/*
1600	* Return any desired output.
1601	*/
1602	if (needoutput \|\| (tp->t_flags & TF_ACKNOW)) {
1603	(void) tcp_output(pData, tp);
1604	}
1605	return;
1606
1607	dropafterack:
1608	/*
1609	* Generate an ACK dropping incoming segment if it occupies
1610	* sequence space, where the ACK reflects our state.
1611	*/
1612	if (tiflags & TH_RST)
1613	goto drop;
1614	m_freem(pData, m);
1615	tp->t_flags \|= TF_ACKNOW;
1616	(void) tcp_output(pData, tp);
1617	return;
1618
1619	dropwithreset:
1620	/* reuses m if m!=NULL, m_free() unnecessary */
1621	if (tiflags & TH_ACK)
1622	tcp_respond(pData, tp, ti, m, (tcp_seq)0, ti->ti_ack, TH_RST);
1623	else {
1624	if (tiflags & TH_SYN) ti->ti_len++;
1625	tcp_respond(pData, tp, ti, m, ti->ti_seq+ti->ti_len, (tcp_seq)0,
1626	TH_RST\|TH_ACK);
1627	}
1628
1629	return;
1630
1631	drop:
1632	/*
1633	* Drop space held by incoming segment and return.
1634	*/
1635	m_free(pData, m);
1636
1637	return;
1638	}
1639
1640	/* , ts_present, ts_val, ts_ecr) */
1641	/* int *ts_present;
1642	* u_int32_t ts_val, ts_ecr;
1643	*/
1644	void
1645	tcp_dooptions(PNATState pData, struct tcpcb tp, u_char cp, int cnt, struct tcpiphdr *ti)
1646	{
1647	u_int16_t mss;
1648	int opt, optlen;
1649
1650	DEBUG_CALL("tcp_dooptions");
1651	DEBUG_ARGS((dfd," tp = %lx cnt=%i \n", (long )tp, cnt));
1652
1653	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1654	opt = cp[0];
1655	if (opt == TCPOPT_EOL)
1656	break;
1657	if (opt == TCPOPT_NOP)
1658	optlen = 1;
1659	else {
1660	optlen = cp[1];
1661	if (optlen <= 0)
1662	break;
1663	}
1664	switch (opt) {
1665
1666	default:
1667	continue;
1668
1669	case TCPOPT_MAXSEG:
1670	if (optlen != TCPOLEN_MAXSEG)
1671	continue;
1672	if (!(ti->ti_flags & TH_SYN))
1673	continue;
1674	memcpy((char ) &mss, (char ) cp + 2, sizeof(mss));
1675	NTOHS(mss);
1676	(void) tcp_mss(pData, tp, mss); /* sets t_maxseg */
1677	break;
1678
1679	/* case TCPOPT_WINDOW:
1680	* if (optlen != TCPOLEN_WINDOW)
1681	* continue;
1682	* if (!(ti->ti_flags & TH_SYN))
1683	* continue;
1684	* tp->t_flags \|= TF_RCVD_SCALE;
1685	* tp->requested_s_scale = min(cp[2], TCP_MAX_WINSHIFT);
1686	* break;
1687	*/
1688	/* case TCPOPT_TIMESTAMP:
1689	* if (optlen != TCPOLEN_TIMESTAMP)
1690	* continue;
1691	* *ts_present = 1;
1692	* memcpy((char ) ts_val, (char )cp + 2, sizeof(*ts_val));
1693	* NTOHL(*ts_val);
1694	* memcpy((char ) ts_ecr, (char )cp + 6, sizeof(*ts_ecr));
1695	* NTOHL(*ts_ecr);
1696	*
1697	/ /
1698	* * A timestamp received in a SYN makes
1699	* * it ok to send timestamp requests and replies.
1700	* */
1701	/* if (ti->ti_flags & TH_SYN) {
1702	* tp->t_flags \|= TF_RCVD_TSTMP;
1703	* tp->ts_recent = *ts_val;
1704	* tp->ts_recent_age = tcp_now;
1705	* }
1706	*/ break;
1707	}
1708	}
1709	}
1710
1711
1712	/*
1713	* Pull out of band byte out of a segment so
1714	* it doesn't appear in the user's data queue.
1715	* It is still reflected in the segment length for
1716	* sequencing purposes.
1717	*/
1718
1719	#ifdef notdef
1720
1721	void
1722	tcp_pulloutofband(so, ti, m)
1723	struct socket *so;
1724	struct tcpiphdr *ti;
1725	register struct mbuf *m;
1726	{
1727	int cnt = ti->ti_urp - 1;
1728
1729	while (cnt >= 0) {
1730	if (m->m_len > cnt) {
1731	char *cp = mtod(m, caddr_t) + cnt;
1732	struct tcpcb *tp = sototcpcb(so);
1733
1734	tp->t_iobc = *cp;
1735	tp->t_oobflags \|= TCPOOB_HAVEDATA;
1736	memcpy(sp, cp+1, (unsigned)(m->m_len - cnt - 1));
1737	m->m_len--;
1738	return;
1739	}
1740	cnt -= m->m_len;
1741	m = m->m_next; /* XXX WRONG! Fix it! */
1742	if (m == 0)
1743	break;
1744	}
1745	panic("tcp_pulloutofband");
1746	}
1747
1748	#endif /* notdef */
1749
1750	/*
1751	* Collect new round-trip time estimate
1752	* and update averages and current timeout.
1753	*/
1754
1755	void
1756	tcp_xmit_timer(PNATState pData, register struct tcpcb *tp, int rtt)
1757	{
1758	register short delta;
1759
1760	DEBUG_CALL("tcp_xmit_timer");
1761	DEBUG_ARG("tp = %lx", (long)tp);
1762	DEBUG_ARG("rtt = %d", rtt);
1763
1764	tcpstat.tcps_rttupdated++;
1765	if (tp->t_srtt != 0) {
1766	/*
1767	* srtt is stored as fixed point with 3 bits after the
1768	* binary point (i.e., scaled by 8). The following magic
1769	* is equivalent to the smoothing algorithm in rfc793 with
1770	* an alpha of .875 (srtt = rtt/8 + srtt*7/8 in fixed
1771	* point). Adjust rtt to origin 0.
1772	*/
1773	delta = rtt - 1 - (tp->t_srtt >> TCP_RTT_SHIFT);
1774	if ((tp->t_srtt += delta) <= 0)
1775	tp->t_srtt = 1;
1776	/*
1777	* We accumulate a smoothed rtt variance (actually, a
1778	* smoothed mean difference), then set the retransmit
1779	* timer to smoothed rtt + 4 times the smoothed variance.
1780	* rttvar is stored as fixed point with 2 bits after the
1781	* binary point (scaled by 4). The following is
1782	* equivalent to rfc793 smoothing with an alpha of .75
1783	* (rttvar = rttvar*3/4 + \|delta\| / 4). This replaces
1784	* rfc793's wired-in beta.
1785	*/
1786	if (delta < 0)
1787	delta = -delta;
1788	delta -= (tp->t_rttvar >> TCP_RTTVAR_SHIFT);
1789	if ((tp->t_rttvar += delta) <= 0)
1790	tp->t_rttvar = 1;
1791	} else {
1792	/*
1793	* No rtt measurement yet - use the unsmoothed rtt.
1794	* Set the variance to half the rtt (so our first
1795	* retransmit happens at 3*rtt).
1796	*/
1797	tp->t_srtt = rtt << TCP_RTT_SHIFT;
1798	tp->t_rttvar = rtt << (TCP_RTTVAR_SHIFT - 1);
1799	}
1800	tp->t_rtt = 0;
1801	tp->t_rxtshift = 0;
1802
1803	/*
1804	* the retransmit should happen at rtt + 4 * rttvar.
1805	* Because of the way we do the smoothing, srtt and rttvar
1806	* will each average +1/2 tick of bias. When we compute
1807	* the retransmit timer, we want 1/2 tick of rounding and
1808	* 1 extra tick because of +-1/2 tick uncertainty in the
1809	* firing of the timer. The bias will give us exactly the
1810	* 1.5 tick we need. But, because the bias is
1811	* statistical, we have to test that we don't drop below
1812	* the minimum feasible timer (which is 2 ticks).
1813	*/
1814	TCPT_RANGESET(tp->t_rxtcur, TCP_REXMTVAL(tp),
1815	(short)tp->t_rttmin, TCPTV_REXMTMAX); /* XXX */
1816
1817	/*
1818	* We received an ack for a packet that wasn't retransmitted;
1819	* it is probably safe to discard any error indications we've
1820	* received recently. This isn't quite right, but close enough
1821	* for now (a route might have failed after we sent a segment,
1822	* and the return path might not be symmetrical).
1823	*/
1824	tp->t_softerror = 0;
1825	}
1826
1827	/*
1828	* Determine a reasonable value for maxseg size.
1829	* If the route is known, check route for mtu.
1830	* If none, use an mss that can be handled on the outgoing
1831	* interface without forcing IP to fragment; if bigger than
1832	* an mbuf cluster (MCLBYTES), round down to nearest multiple of MCLBYTES
1833	* to utilize large mbufs. If no route is found, route has no mtu,
1834	* or the destination isn't local, use a default, hopefully conservative
1835	* size (usually 512 or the default IP max size, but no more than the mtu
1836	* of the interface), as we can't discover anything about intervening
1837	* gateways or networks. We also initialize the congestion/slow start
1838	* window to be a single segment if the destination isn't local.
1839	* While looking at the routing entry, we also initialize other path-dependent
1840	* parameters from pre-set or cached values in the routing entry.
1841	*/
1842
1843	int
1844	tcp_mss(PNATState pData, register struct tcpcb *tp, u_int offer)
1845	{
1846	struct socket *so = tp->t_socket;
1847	int mss;
1848
1849	DEBUG_CALL("tcp_mss");
1850	DEBUG_ARG("tp = %lx", (long)tp);
1851	DEBUG_ARG("offer = %d", offer);
1852
1853	mss = min(if_mtu, if_mru) - sizeof(struct tcpiphdr);
1854	if (offer)
1855	mss = min(mss, offer);
1856	mss = max(mss, 32);
1857	if (mss < tp->t_maxseg \|\| offer != 0)
1858	tp->t_maxseg = mss;
1859
1860	tp->snd_cwnd = mss;
1861
1862	sbreserve(&so->so_snd, tcp_sndspace+((tcp_sndspace%mss)?(mss-(tcp_sndspace%mss)):0));
1863	sbreserve(&so->so_rcv, tcp_rcvspace+((tcp_rcvspace%mss)?(mss-(tcp_rcvspace%mss)):0));
1864
1865	DEBUG_MISC((dfd, " returning mss = %d\n", mss));
1866
1867	return mss;
1868	}

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

source: vbox/trunk/src/VBox/Devices/Network/slirp/tcp_input.c@ 14577

Download in other formats: