tcp_input.c@ 15136

Last change on this file since 15136 was 14638, checked in by vboxsync, 16 years ago
NAT: some reformatting for better readability; merged VBOX_WITH_BSD_TCP_REASS with VBOX_WITH_BSD_REASS
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_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_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_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	#if !defined(VBOX_WITH_BSD_REASS)
473	ti->ti_next = ti->ti_prev = 0;
474	ti->ti_x1 = 0;
475	#else
476	memset(ti->ti_x1, 0, 9);
477	#endif
478	ti->ti_len = htons((u_int16_t)tlen);
479	len = sizeof(struct ip ) + tlen;
480	/* keep checksum for ICMP reply
481	* ti->ti_sum = cksum(m, len);
482	* if (ti->ti_sum) { */
483	if(cksum(m, len)) {
484	tcpstat.tcps_rcvbadsum++;
485	goto drop;
486	}
487
488	/*
489	* Check that TCP offset makes sense,
490	* pull out TCP options and adjust length. XXX
491	*/
492	off = ti->ti_off << 2;
493	if (off < sizeof (struct tcphdr) \|\| off > tlen) {
494	tcpstat.tcps_rcvbadoff++;
495	goto drop;
496	}
497	tlen -= off;
498	ti->ti_len = tlen;
499	if (off > sizeof (struct tcphdr)) {
500	optlen = off - sizeof (struct tcphdr);
501	optp = mtod(m, caddr_t) + sizeof (struct tcpiphdr);
502
503	/*
504	* Do quick retrieval of timestamp options ("options
505	* prediction?"). If timestamp is the only option and it's
506	* formatted as recommended in RFC 1323 appendix A, we
507	* quickly get the values now and not bother calling
508	* tcp_dooptions(), etc.
509	*/
510	/* if ((optlen == TCPOLEN_TSTAMP_APPA \|\|
511	* (optlen > TCPOLEN_TSTAMP_APPA &&
512	* optp[TCPOLEN_TSTAMP_APPA] == TCPOPT_EOL)) &&
513	* (u_int32_t )optp == htonl(TCPOPT_TSTAMP_HDR) &&
514	* (ti->ti_flags & TH_SYN) == 0) {
515	* ts_present = 1;
516	* ts_val = ntohl((u_int32_t )(optp + 4));
517	* ts_ecr = ntohl((u_int32_t )(optp + 8));
518	* optp = NULL; / * we've parsed the options * /
519	* }
520	*/
521	}
522	tiflags = ti->ti_flags;
523
524	/*
525	* Convert TCP protocol specific fields to host format.
526	*/
527	NTOHL(ti->ti_seq);
528	NTOHL(ti->ti_ack);
529	NTOHS(ti->ti_win);
530	NTOHS(ti->ti_urp);
531
532	/*
533	* Drop TCP, IP headers and TCP options.
534	*/
535	m->m_data += sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
536	m->m_len -= sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
537
538	/*
539	* Locate pcb for segment.
540	*/
541	findso:
542	so = tcp_last_so;
543	if (so->so_fport != ti->ti_dport \|\|
544	so->so_lport != ti->ti_sport \|\|
545	so->so_laddr.s_addr != ti->ti_src.s_addr \|\|
546	so->so_faddr.s_addr != ti->ti_dst.s_addr) {
547	so = solookup(&tcb, ti->ti_src, ti->ti_sport,
548	ti->ti_dst, ti->ti_dport);
549	if (so)
550	tcp_last_so = so;
551	++tcpstat.tcps_socachemiss;
552	}
553
554	/*
555	* If the state is CLOSED (i.e., TCB does not exist) then
556	* all data in the incoming segment is discarded.
557	* If the TCB exists but is in CLOSED state, it is embryonic,
558	* but should either do a listen or a connect soon.
559	*
560	* state == CLOSED means we've done socreate() but haven't
561	* attached it to a protocol yet...
562	*
563	* XXX If a TCB does not exist, and the TH_SYN flag is
564	* the only flag set, then create a session, mark it
565	* as if it was LISTENING, and continue...
566	*/
567	if (so == 0) {
568	if ((tiflags & (TH_SYN\|TH_FIN\|TH_RST\|TH_URG\|TH_ACK)) != TH_SYN)
569	goto dropwithreset;
570
571	if ((so = socreate()) == NULL)
572	goto dropwithreset;
573	if (tcp_attach(pData, so) < 0) {
574	free(so); /* Not sofree (if it failed, it's not insqued) */
575	goto dropwithreset;
576	}
577
578	sbreserve(&so->so_snd, tcp_sndspace);
579	sbreserve(&so->so_rcv, tcp_rcvspace);
580
581	/* tcp_last_so = so; / / XXX ? */
582	/* tp = sototcpcb(so); */
583
584	so->so_laddr = ti->ti_src;
585	so->so_lport = ti->ti_sport;
586	so->so_faddr = ti->ti_dst;
587	so->so_fport = ti->ti_dport;
588
589	if ((so->so_iptos = tcp_tos(so)) == 0)
590	so->so_iptos = ((struct ip *)ti)->ip_tos;
591
592	tp = sototcpcb(so);
593	tp->t_state = TCPS_LISTEN;
594	}
595
596	/*
597	* If this is a still-connecting socket, this probably
598	* a retransmit of the SYN. Whether it's a retransmit SYN
599	* or something else, we nuke it.
600	*/
601	if (so->so_state & SS_ISFCONNECTING)
602	goto drop;
603
604	tp = sototcpcb(so);
605
606	/* XXX Should never fail */
607	if (tp == 0)
608	goto dropwithreset;
609	if (tp->t_state == TCPS_CLOSED)
610	goto drop;
611
612	/* Unscale the window into a 32-bit value. */
613	/* if ((tiflags & TH_SYN) == 0)
614	* tiwin = ti->ti_win << tp->snd_scale;
615	* else
616	*/
617	tiwin = ti->ti_win;
618
619	/*
620	* Segment received on connection.
621	* Reset idle time and keep-alive timer.
622	*/
623	tp->t_idle = 0;
624	if (so_options)
625	tp->t_timer[TCPT_KEEP] = tcp_keepintvl;
626	else
627	tp->t_timer[TCPT_KEEP] = tcp_keepidle;
628
629	/*
630	* Process options if not in LISTEN state,
631	* else do it below (after getting remote address).
632	*/
633	if (optp && tp->t_state != TCPS_LISTEN)
634	tcp_dooptions(pData, tp, (u_char *)optp, optlen, ti);
635	/* , */
636	/* &ts_present, &ts_val, &ts_ecr); */
637
638	/*
639	* Header prediction: check for the two common cases
640	* of a uni-directional data xfer. If the packet has
641	* no control flags, is in-sequence, the window didn't
642	* change and we're not retransmitting, it's a
643	* candidate. If the length is zero and the ack moved
644	* forward, we're the sender side of the xfer. Just
645	* free the data acked & wake any higher level process
646	* that was blocked waiting for space. If the length
647	* is non-zero and the ack didn't move, we're the
648	* receiver side. If we're getting packets in-order
649	* (the reassembly queue is empty), add the data to
650	* the socket buffer and note that we need a delayed ack.
651	*
652	* XXX Some of these tests are not needed
653	* eg: the tiwin == tp->snd_wnd prevents many more
654	* predictions.. with no real advantage..
655	*/
656	if (tp->t_state == TCPS_ESTABLISHED &&
657	(tiflags & (TH_SYN\|TH_FIN\|TH_RST\|TH_URG\|TH_ACK)) == TH_ACK &&
658	/* (!ts_present \|\| TSTMP_GEQ(ts_val, tp->ts_recent)) && */
659	ti->ti_seq == tp->rcv_nxt &&
660	tiwin && tiwin == tp->snd_wnd &&
661	tp->snd_nxt == tp->snd_max) {
662	/*
663	* If last ACK falls within this segment's sequence numbers,
664	* record the timestamp.
665	*/
666	/* if (ts_present && SEQ_LEQ(ti->ti_seq, tp->last_ack_sent) &&
667	* SEQ_LT(tp->last_ack_sent, ti->ti_seq + ti->ti_len)) {
668	* tp->ts_recent_age = tcp_now;
669	* tp->ts_recent = ts_val;
670	* }
671	*/
672	if (ti->ti_len == 0) {
673	if (SEQ_GT(ti->ti_ack, tp->snd_una) &&
674	SEQ_LEQ(ti->ti_ack, tp->snd_max) &&
675	tp->snd_cwnd >= tp->snd_wnd) {
676	/*
677	* this is a pure ack for outstanding data.
678	*/
679	++tcpstat.tcps_predack;
680	/* if (ts_present)
681	* tcp_xmit_timer(tp, tcp_now-ts_ecr+1);
682	* else
683	*/ if (tp->t_rtt &&
684	SEQ_GT(ti->ti_ack, tp->t_rtseq))
685	tcp_xmit_timer(pData, tp, tp->t_rtt);
686	acked = ti->ti_ack - tp->snd_una;
687	tcpstat.tcps_rcvackpack++;
688	tcpstat.tcps_rcvackbyte += acked;
689	sbdrop(&so->so_snd, acked);
690	tp->snd_una = ti->ti_ack;
691	m_freem(pData, m);
692
693	/*
694	* If all outstanding data are acked, stop
695	* retransmit timer, otherwise restart timer
696	* using current (possibly backed-off) value.
697	* If process is waiting for space,
698	* wakeup/selwakeup/signal. If data
699	* are ready to send, let tcp_output
700	* decide between more output or persist.
701	*/
702	if (tp->snd_una == tp->snd_max)
703	tp->t_timer[TCPT_REXMT] = 0;
704	else if (tp->t_timer[TCPT_PERSIST] == 0)
705	tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
706
707	/*
708	* There's room in so_snd, sowwakup will read()
709	* from the socket if we can
710	*/
711	/* if (so->so_snd.sb_flags & SB_NOTIFY)
712	* sowwakeup(so);
713	*/
714	/*
715	* This is called because sowwakeup might have
716	* put data into so_snd. Since we don't so sowwakeup,
717	* we don't need this.. XXX???
718	*/
719	if (so->so_snd.sb_cc)
720	(void) tcp_output(pData, tp);
721
722	return;
723	}
724	} else if (ti->ti_ack == tp->snd_una &&
725	#ifndef VBOX_WITH_BSD_REASS
726	u32_to_ptr(pData, tp->seg_next, struct tcpcb *) == tp &&
727	#else /* VBOX_WITH_BSD_REASS */
728	LIST_FIRST(&tp->t_segq) &&
729	#endif /* VBOX_WITH_BSD_REASS */
730	ti->ti_len <= sbspace(&so->so_rcv)) {
731	/*
732	* this is a pure, in-sequence data packet
733	* with nothing on the reassembly queue and
734	* we have enough buffer space to take it.
735	*/
736	++tcpstat.tcps_preddat;
737	tp->rcv_nxt += ti->ti_len;
738	tcpstat.tcps_rcvpack++;
739	tcpstat.tcps_rcvbyte += ti->ti_len;
740	/*
741	* Add data to socket buffer.
742	*/
743	if (so->so_emu) {
744	if (tcp_emu(pData, so,m)) sbappend(pData, so, m);
745	} else
746	sbappend(pData, so, m);
747
748	/*
749	* XXX This is called when data arrives. Later, check
750	* if we can actually write() to the socket
751	* XXX Need to check? It's be NON_BLOCKING
752	*/
753	/* sorwakeup(so); */
754
755	/*
756	* If this is a short packet, then ACK now - with Nagel
757	* congestion avoidance sender won't send more until
758	* he gets an ACK.
759	*
760	* It is better to not delay acks at all to maximize
761	* TCP throughput. See RFC 2581.
762	*/
763	tp->t_flags \|= TF_ACKNOW;
764	tcp_output(pData, tp);
765	return;
766	}
767	} /* header prediction */
768	/*
769	* Calculate amount of space in receive window,
770	* and then do TCP input processing.
771	* Receive window is amount of space in rcv queue,
772	* but not less than advertised window.
773	*/
774	{ int win;
775	win = sbspace(&so->so_rcv);
776	if (win < 0)
777	win = 0;
778	tp->rcv_wnd = max(win, (int)(tp->rcv_adv - tp->rcv_nxt));
779	}
780
781	switch (tp->t_state) {
782
783	/*
784	* If the state is LISTEN then ignore segment if it contains an RST.
785	* If the segment contains an ACK then it is bad and send a RST.
786	* If it does not contain a SYN then it is not interesting; drop it.
787	* Don't bother responding if the destination was a broadcast.
788	* Otherwise initialize tp->rcv_nxt, and tp->irs, select an initial
789	* tp->iss, and send a segment:
790	* <SEQ=ISS><ACK=RCV_NXT><CTL=SYN,ACK>
791	* Also initialize tp->snd_nxt to tp->iss+1 and tp->snd_una to tp->iss.
792	* Fill in remote peer address fields if not previously specified.
793	* Enter SYN_RECEIVED state, and process any other fields of this
794	* segment in this state.
795	*/
796	case TCPS_LISTEN: {
797
798	if (tiflags & TH_RST)
799	goto drop;
800	if (tiflags & TH_ACK)
801	goto dropwithreset;
802	if ((tiflags & TH_SYN) == 0)
803	goto drop;
804
805	/*
806	* This has way too many gotos...
807	* But a bit of spaghetti code never hurt anybody :)
808	*/
809
810	if (so->so_emu & EMU_NOCONNECT) {
811	so->so_emu &= ~EMU_NOCONNECT;
812	goto cont_input;
813	}
814
815	if((tcp_fconnect(pData, so) == -1) && (errno != EINPROGRESS) && (errno != EWOULDBLOCK)) {
816	u_char code=ICMP_UNREACH_NET;
817	DEBUG_MISC((dfd," tcp fconnect errno = %d-%s\n",
818	errno,strerror(errno)));
819	if(errno == ECONNREFUSED) {
820	/* ACK the SYN, send RST to refuse the connection */
821	tcp_respond(pData, tp, ti, m, ti->ti_seq+1, (tcp_seq)0,
822	TH_RST\|TH_ACK);
823	} else {
824	if(errno == EHOSTUNREACH) code=ICMP_UNREACH_HOST;
825	HTONL(ti->ti_seq); /* restore tcp header */
826	HTONL(ti->ti_ack);
827	HTONS(ti->ti_win);
828	HTONS(ti->ti_urp);
829	m->m_data -= sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
830	m->m_len += sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
831	*ip=save_ip;
832	icmp_error(pData, m, ICMP_UNREACH,code, 0,strerror(errno));
833	}
834	tp = tcp_close(pData, tp);
835	m_free(pData, m);
836	} else {
837	/*
838	* Haven't connected yet, save the current mbuf
839	* and ti, and return
840	* XXX Some OS's don't tell us whether the connect()
841	* succeeded or not. So we must time it out.
842	*/
843	so->so_m = m;
844	so->so_ti = ti;
845	tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT;
846	tp->t_state = TCPS_SYN_RECEIVED;
847	}
848	return;
849
850	cont_conn:
851	/* m==NULL
852	* Check if the connect succeeded
853	*/
854	if (so->so_state & SS_NOFDREF) {
855	tp = tcp_close(pData, tp);
856	goto dropwithreset;
857	}
858	cont_input:
859	tcp_template(tp);
860
861	if (optp)
862	tcp_dooptions(pData, tp, (u_char *)optp, optlen, ti);
863	/* , */
864	/* &ts_present, &ts_val, &ts_ecr); */
865
866	if (iss)
867	tp->iss = iss;
868	else
869	tp->iss = tcp_iss;
870	tcp_iss += TCP_ISSINCR/2;
871	tp->irs = ti->ti_seq;
872	tcp_sendseqinit(tp);
873	tcp_rcvseqinit(tp);
874	tp->t_flags \|= TF_ACKNOW;
875	tp->t_state = TCPS_SYN_RECEIVED;
876	tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT;
877	tcpstat.tcps_accepts++;
878	goto trimthenstep6;
879	} /* case TCPS_LISTEN */
880
881	/*
882	* If the state is SYN_SENT:
883	* if seg contains an ACK, but not for our SYN, drop the input.
884	* if seg contains a RST, then drop the connection.
885	* if seg does not contain SYN, then drop it.
886	* Otherwise this is an acceptable SYN segment
887	* initialize tp->rcv_nxt and tp->irs
888	* if seg contains ack then advance tp->snd_una
889	* if SYN has been acked change to ESTABLISHED else SYN_RCVD state
890	* arrange for segment to be acked (eventually)
891	* continue processing rest of data/controls, beginning with URG
892	*/
893	case TCPS_SYN_SENT:
894	if ((tiflags & TH_ACK) &&
895	(SEQ_LEQ(ti->ti_ack, tp->iss) \|\|
896	SEQ_GT(ti->ti_ack, tp->snd_max)))
897	goto dropwithreset;
898
899	if (tiflags & TH_RST) {
900	if (tiflags & TH_ACK)
901	tp = tcp_drop(pData, tp,0); /* XXX Check t_softerror! */
902	goto drop;
903	}
904
905	if ((tiflags & TH_SYN) == 0)
906	goto drop;
907	if (tiflags & TH_ACK) {
908	tp->snd_una = ti->ti_ack;
909	if (SEQ_LT(tp->snd_nxt, tp->snd_una))
910	tp->snd_nxt = tp->snd_una;
911	}
912
913	tp->t_timer[TCPT_REXMT] = 0;
914	tp->irs = ti->ti_seq;
915	tcp_rcvseqinit(tp);
916	tp->t_flags \|= TF_ACKNOW;
917	if (tiflags & TH_ACK && SEQ_GT(tp->snd_una, tp->iss)) {
918	tcpstat.tcps_connects++;
919	soisfconnected(so);
920	tp->t_state = TCPS_ESTABLISHED;
921
922	/* Do window scaling on this connection? */
923	/* if ((tp->t_flags & (TF_RCVD_SCALE\|TF_REQ_SCALE)) ==
924	* (TF_RCVD_SCALE\|TF_REQ_SCALE)) {
925	* tp->snd_scale = tp->requested_s_scale;
926	* tp->rcv_scale = tp->request_r_scale;
927	* }
928	*/
929	#ifndef VBOX_WITH_BSD_REASS
930	(void) tcp_reass(pData, tp, (struct tcpiphdr *)0,
931	(struct mbuf *)0);
932	#else /* VBOX_WITH_BSD_REASS */
933	(void) tcp_reass(pData, tp, (struct tcphdr )0, NULL, (struct mbuf )0);
934	#endif /* VBOX_WITH_BSD_REASS */
935	/*
936	* if we didn't have to retransmit the SYN,
937	* use its rtt as our initial srtt & rtt var.
938	*/
939	if (tp->t_rtt)
940	tcp_xmit_timer(pData, tp, tp->t_rtt);
941	} else
942	tp->t_state = TCPS_SYN_RECEIVED;
943
944	trimthenstep6:
945	/*
946	* Advance ti->ti_seq to correspond to first data byte.
947	* If data, trim to stay within window,
948	* dropping FIN if necessary.
949	*/
950	ti->ti_seq++;
951	if (ti->ti_len > tp->rcv_wnd) {
952	todrop = ti->ti_len - tp->rcv_wnd;
953	m_adj(m, -todrop);
954	ti->ti_len = tp->rcv_wnd;
955	tiflags &= ~TH_FIN;
956	tcpstat.tcps_rcvpackafterwin++;
957	tcpstat.tcps_rcvbyteafterwin += todrop;
958	}
959	tp->snd_wl1 = ti->ti_seq - 1;
960	tp->rcv_up = ti->ti_seq;
961	goto step6;
962	} /* switch tp->t_state */
963	/*
964	* States other than LISTEN or SYN_SENT.
965	* First check timestamp, if present.
966	* Then check that at least some bytes of segment are within
967	* receive window. If segment begins before rcv_nxt,
968	* drop leading data (and SYN); if nothing left, just ack.
969	*
970	* RFC 1323 PAWS: If we have a timestamp reply on this segment
971	* and it's less than ts_recent, drop it.
972	*/
973	/* if (ts_present && (tiflags & TH_RST) == 0 && tp->ts_recent &&
974	* TSTMP_LT(ts_val, tp->ts_recent)) {
975	*
976	/ / Check to see if ts_recent is over 24 days old. */
977	/* if ((int)(tcp_now - tp->ts_recent_age) > TCP_PAWS_IDLE) {
978	/ /
979	* * Invalidate ts_recent. If this segment updates
980	* * ts_recent, the age will be reset later and ts_recent
981	* * will get a valid value. If it does not, setting
982	* * ts_recent to zero will at least satisfy the
983	* * requirement that zero be placed in the timestamp
984	* * echo reply when ts_recent isn't valid. The
985	* * age isn't reset until we get a valid ts_recent
986	* * because we don't want out-of-order segments to be
987	* * dropped when ts_recent is old.
988	* */
989	/* tp->ts_recent = 0;
990	* } else {
991	* tcpstat.tcps_rcvduppack++;
992	* tcpstat.tcps_rcvdupbyte += ti->ti_len;
993	* tcpstat.tcps_pawsdrop++;
994	* goto dropafterack;
995	* }
996	* }
997	*/
998
999	todrop = tp->rcv_nxt - ti->ti_seq;
1000	if (todrop > 0) {
1001	if (tiflags & TH_SYN) {
1002	tiflags &= ~TH_SYN;
1003	ti->ti_seq++;
1004	if (ti->ti_urp > 1)
1005	ti->ti_urp--;
1006	else
1007	tiflags &= ~TH_URG;
1008	todrop--;
1009	}
1010	/*
1011	* Following if statement from Stevens, vol. 2, p. 960.
1012	*/
1013	if (todrop > ti->ti_len
1014	\|\| (todrop == ti->ti_len && (tiflags & TH_FIN) == 0)) {
1015	/*
1016	* Any valid FIN must be to the left of the window.
1017	* At this point the FIN must be a duplicate or out
1018	* of sequence; drop it.
1019	*/
1020	tiflags &= ~TH_FIN;
1021
1022	/*
1023	* Send an ACK to resynchronize and drop any data.
1024	* But keep on processing for RST or ACK.
1025	*/
1026	tp->t_flags \|= TF_ACKNOW;
1027	todrop = ti->ti_len;
1028	tcpstat.tcps_rcvduppack++;
1029	tcpstat.tcps_rcvdupbyte += todrop;
1030	} else {
1031	tcpstat.tcps_rcvpartduppack++;
1032	tcpstat.tcps_rcvpartdupbyte += todrop;
1033	}
1034	m_adj(m, todrop);
1035	ti->ti_seq += todrop;
1036	ti->ti_len -= todrop;
1037	if (ti->ti_urp > todrop)
1038	ti->ti_urp -= todrop;
1039	else {
1040	tiflags &= ~TH_URG;
1041	ti->ti_urp = 0;
1042	}
1043	}
1044	/*
1045	* If new data are received on a connection after the
1046	* user processes are gone, then RST the other end.
1047	*/
1048	if ((so->so_state & SS_NOFDREF) &&
1049	tp->t_state > TCPS_CLOSE_WAIT && ti->ti_len) {
1050	tp = tcp_close(pData, tp);
1051	tcpstat.tcps_rcvafterclose++;
1052	goto dropwithreset;
1053	}
1054
1055	/*
1056	* If segment ends after window, drop trailing data
1057	* (and PUSH and FIN); if nothing left, just ACK.
1058	*/
1059	todrop = (ti->ti_seq+ti->ti_len) - (tp->rcv_nxt+tp->rcv_wnd);
1060	if (todrop > 0) {
1061	tcpstat.tcps_rcvpackafterwin++;
1062	if (todrop >= ti->ti_len) {
1063	tcpstat.tcps_rcvbyteafterwin += ti->ti_len;
1064	/*
1065	* If a new connection request is received
1066	* while in TIME_WAIT, drop the old connection
1067	* and start over if the sequence numbers
1068	* are above the previous ones.
1069	*/
1070	if (tiflags & TH_SYN &&
1071	tp->t_state == TCPS_TIME_WAIT &&
1072	SEQ_GT(ti->ti_seq, tp->rcv_nxt)) {
1073	iss = tp->rcv_nxt + TCP_ISSINCR;
1074	tp = tcp_close(pData, tp);
1075	goto findso;
1076	}
1077	/*
1078	* If window is closed can only take segments at
1079	* window edge, and have to drop data and PUSH from
1080	* incoming segments. Continue processing, but
1081	* remember to ack. Otherwise, drop segment
1082	* and ack.
1083	*/
1084	if (tp->rcv_wnd == 0 && ti->ti_seq == tp->rcv_nxt) {
1085	tp->t_flags \|= TF_ACKNOW;
1086	tcpstat.tcps_rcvwinprobe++;
1087	} else
1088	goto dropafterack;
1089	} else
1090	tcpstat.tcps_rcvbyteafterwin += todrop;
1091	m_adj(m, -todrop);
1092	ti->ti_len -= todrop;
1093	tiflags &= ~(TH_PUSH\|TH_FIN);
1094	}
1095
1096	/*
1097	* If last ACK falls within this segment's sequence numbers,
1098	* record its timestamp.
1099	*/
1100	/* if (ts_present && SEQ_LEQ(ti->ti_seq, tp->last_ack_sent) &&
1101	* SEQ_LT(tp->last_ack_sent, ti->ti_seq + ti->ti_len +
1102	* ((tiflags & (TH_SYN\|TH_FIN)) != 0))) {
1103	* tp->ts_recent_age = tcp_now;
1104	* tp->ts_recent = ts_val;
1105	* }
1106	*/
1107
1108	/*
1109	* If the RST bit is set examine the state:
1110	* SYN_RECEIVED STATE:
1111	* If passive open, return to LISTEN state.
1112	* If active open, inform user that connection was refused.
1113	* ESTABLISHED, FIN_WAIT_1, FIN_WAIT2, CLOSE_WAIT STATES:
1114	* Inform user that connection was reset, and close tcb.
1115	* CLOSING, LAST_ACK, TIME_WAIT STATES
1116	* Close the tcb.
1117	*/
1118	if (tiflags&TH_RST) switch (tp->t_state) {
1119
1120	case TCPS_SYN_RECEIVED:
1121	/* so->so_error = ECONNREFUSED; */
1122	goto close;
1123
1124	case TCPS_ESTABLISHED:
1125	case TCPS_FIN_WAIT_1:
1126	case TCPS_FIN_WAIT_2:
1127	case TCPS_CLOSE_WAIT:
1128	/* so->so_error = ECONNRESET; */
1129	close:
1130	tp->t_state = TCPS_CLOSED;
1131	tcpstat.tcps_drops++;
1132	tp = tcp_close(pData, tp);
1133	goto drop;
1134
1135	case TCPS_CLOSING:
1136	case TCPS_LAST_ACK:
1137	case TCPS_TIME_WAIT:
1138	tp = tcp_close(pData, tp);
1139	goto drop;
1140	}
1141
1142	/*
1143	* If a SYN is in the window, then this is an
1144	* error and we send an RST and drop the connection.
1145	*/
1146	if (tiflags & TH_SYN) {
1147	tp = tcp_drop(pData, tp,0);
1148	goto dropwithreset;
1149	}
1150
1151	/*
1152	* If the ACK bit is off we drop the segment and return.
1153	*/
1154	if ((tiflags & TH_ACK) == 0) goto drop;
1155
1156	/*
1157	* Ack processing.
1158	*/
1159	switch (tp->t_state) {
1160	/*
1161	* In SYN_RECEIVED state if the ack ACKs our SYN then enter
1162	* ESTABLISHED state and continue processing, otherwise
1163	* send an RST. una<=ack<=max
1164	*/
1165	case TCPS_SYN_RECEIVED:
1166
1167	if (SEQ_GT(tp->snd_una, ti->ti_ack) \|\|
1168	SEQ_GT(ti->ti_ack, tp->snd_max))
1169	goto dropwithreset;
1170	tcpstat.tcps_connects++;
1171	tp->t_state = TCPS_ESTABLISHED;
1172	/*
1173	* The sent SYN is ack'ed with our sequence number +1
1174	* The first data byte already in the buffer will get
1175	* lost if no correction is made. This is only needed for
1176	* SS_CTL since the buffer is empty otherwise.
1177	* tp->snd_una++; or:
1178	*/
1179	tp->snd_una=ti->ti_ack;
1180	soisfconnected(so);
1181
1182	/* Do window scaling? */
1183	/* if ((tp->t_flags & (TF_RCVD_SCALE\|TF_REQ_SCALE)) ==
1184	* (TF_RCVD_SCALE\|TF_REQ_SCALE)) {
1185	* tp->snd_scale = tp->requested_s_scale;
1186	* tp->rcv_scale = tp->request_r_scale;
1187	* }
1188	*/
1189	#ifndef VBOX_WITH_BSD_REASS
1190	(void) tcp_reass(pData, tp, (struct tcpiphdr )0, (struct mbuf )0);
1191	#else /* VBOX_WITH_BSD_REASS */
1192	(void) tcp_reass(pData, tp, (struct tcphdr )0, (int )0, (struct mbuf *)0);
1193	#endif /VBOX_WITH_BSD_REASS/
1194	tp->snd_wl1 = ti->ti_seq - 1;
1195	/* Avoid ack processing; snd_una==ti_ack => dup ack */
1196	goto synrx_to_est;
1197	/* fall into ... */
1198
1199	/*
1200	* In ESTABLISHED state: drop duplicate ACKs; ACK out of range
1201	* ACKs. If the ack is in the range
1202	* tp->snd_una < ti->ti_ack <= tp->snd_max
1203	* then advance tp->snd_una to ti->ti_ack and drop
1204	* data from the retransmission queue. If this ACK reflects
1205	* more up to date window information we update our window information.
1206	*/
1207	case TCPS_ESTABLISHED:
1208	case TCPS_FIN_WAIT_1:
1209	case TCPS_FIN_WAIT_2:
1210	case TCPS_CLOSE_WAIT:
1211	case TCPS_CLOSING:
1212	case TCPS_LAST_ACK:
1213	case TCPS_TIME_WAIT:
1214
1215	if (SEQ_LEQ(ti->ti_ack, tp->snd_una)) {
1216	if (ti->ti_len == 0 && tiwin == tp->snd_wnd) {
1217	tcpstat.tcps_rcvdupack++;
1218	DEBUG_MISC((dfd," dup ack m = %lx so = %lx \n",
1219	(long )m, (long )so));
1220	/*
1221	* If we have outstanding data (other than
1222	* a window probe), this is a completely
1223	* duplicate ack (ie, window info didn't
1224	* change), the ack is the biggest we've
1225	* seen and we've seen exactly our rexmt
1226	* threshold of them, assume a packet
1227	* has been dropped and retransmit it.
1228	* Kludge snd_nxt & the congestion
1229	* window so we send only this one
1230	* packet.
1231	*
1232	* We know we're losing at the current
1233	* window size so do congestion avoidance
1234	* (set ssthresh to half the current window
1235	* and pull our congestion window back to
1236	* the new ssthresh).
1237	*
1238	* Dup acks mean that packets have left the
1239	* network (they're now cached at the receiver)
1240	* so bump cwnd by the amount in the receiver
1241	* to keep a constant cwnd packets in the
1242	* network.
1243	*/
1244	if (tp->t_timer[TCPT_REXMT] == 0 \|\|
1245	ti->ti_ack != tp->snd_una)
1246	tp->t_dupacks = 0;
1247	else if (++tp->t_dupacks == tcprexmtthresh) {
1248	tcp_seq onxt = tp->snd_nxt;
1249	u_int win =
1250	min(tp->snd_wnd, tp->snd_cwnd) / 2 /
1251	tp->t_maxseg;
1252
1253	if (win < 2)
1254	win = 2;
1255	tp->snd_ssthresh = win * tp->t_maxseg;
1256	tp->t_timer[TCPT_REXMT] = 0;
1257	tp->t_rtt = 0;
1258	tp->snd_nxt = ti->ti_ack;
1259	tp->snd_cwnd = tp->t_maxseg;
1260	(void) tcp_output(pData, tp);
1261	tp->snd_cwnd = tp->snd_ssthresh +
1262	tp->t_maxseg * tp->t_dupacks;
1263	if (SEQ_GT(onxt, tp->snd_nxt))
1264	tp->snd_nxt = onxt;
1265	goto drop;
1266	} else if (tp->t_dupacks > tcprexmtthresh) {
1267	tp->snd_cwnd += tp->t_maxseg;
1268	(void) tcp_output(pData, tp);
1269	goto drop;
1270	}
1271	} else
1272	tp->t_dupacks = 0;
1273	break;
1274	}
1275	synrx_to_est:
1276	/*
1277	* If the congestion window was inflated to account
1278	* for the other side's cached packets, retract it.
1279	*/
1280	if (tp->t_dupacks > tcprexmtthresh &&
1281	tp->snd_cwnd > tp->snd_ssthresh)
1282	tp->snd_cwnd = tp->snd_ssthresh;
1283	tp->t_dupacks = 0;
1284	if (SEQ_GT(ti->ti_ack, tp->snd_max)) {
1285	tcpstat.tcps_rcvacktoomuch++;
1286	goto dropafterack;
1287	}
1288	acked = ti->ti_ack - tp->snd_una;
1289	tcpstat.tcps_rcvackpack++;
1290	tcpstat.tcps_rcvackbyte += acked;
1291
1292	/*
1293	* If we have a timestamp reply, update smoothed
1294	* round trip time. If no timestamp is present but
1295	* transmit timer is running and timed sequence
1296	* number was acked, update smoothed round trip time.
1297	* Since we now have an rtt measurement, cancel the
1298	* timer backoff (cf., Phil Karn's retransmit alg.).
1299	* Recompute the initial retransmit timer.
1300	*/
1301	/* if (ts_present)
1302	* tcp_xmit_timer(tp, tcp_now-ts_ecr+1);
1303	* else
1304	*/
1305	if (tp->t_rtt && SEQ_GT(ti->ti_ack, tp->t_rtseq))
1306	tcp_xmit_timer(pData, tp,tp->t_rtt);
1307
1308	/*
1309	* If all outstanding data is acked, stop retransmit
1310	* timer and remember to restart (more output or persist).
1311	* If there is more data to be acked, restart retransmit
1312	* timer, using current (possibly backed-off) value.
1313	*/
1314	if (ti->ti_ack == tp->snd_max) {
1315	tp->t_timer[TCPT_REXMT] = 0;
1316	needoutput = 1;
1317	} else if (tp->t_timer[TCPT_PERSIST] == 0)
1318	tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
1319	/*
1320	* When new data is acked, open the congestion window.
1321	* If the window gives us less than ssthresh packets
1322	* in flight, open exponentially (maxseg per packet).
1323	* Otherwise open linearly: maxseg per window
1324	* (maxseg^2 / cwnd per packet).
1325	*/
1326	{
1327	register u_int cw = tp->snd_cwnd;
1328	register u_int incr = tp->t_maxseg;
1329
1330	if (cw > tp->snd_ssthresh)
1331	incr = incr * incr / cw;
1332	tp->snd_cwnd = min(cw + incr, TCP_MAXWIN<<tp->snd_scale);
1333	}
1334	if (acked > so->so_snd.sb_cc) {
1335	tp->snd_wnd -= so->so_snd.sb_cc;
1336	sbdrop(&so->so_snd, (int )so->so_snd.sb_cc);
1337	ourfinisacked = 1;
1338	} else {
1339	sbdrop(&so->so_snd, acked);
1340	tp->snd_wnd -= acked;
1341	ourfinisacked = 0;
1342	}
1343	/*
1344	* XXX sowwakup is called when data is acked and there's room for
1345	* for more data... it should read() the socket
1346	*/
1347	/* if (so->so_snd.sb_flags & SB_NOTIFY)
1348	* sowwakeup(so);
1349	*/
1350	tp->snd_una = ti->ti_ack;
1351	if (SEQ_LT(tp->snd_nxt, tp->snd_una))
1352	tp->snd_nxt = tp->snd_una;
1353
1354	switch (tp->t_state) {
1355
1356	/*
1357	* In FIN_WAIT_1 STATE in addition to the processing
1358	* for the ESTABLISHED state if our FIN is now acknowledged
1359	* then enter FIN_WAIT_2.
1360	*/
1361	case TCPS_FIN_WAIT_1:
1362	if (ourfinisacked) {
1363	/*
1364	* If we can't receive any more
1365	* data, then closing user can proceed.
1366	* Starting the timer is contrary to the
1367	* specification, but if we don't get a FIN
1368	* we'll hang forever.
1369	*/
1370	if (so->so_state & SS_FCANTRCVMORE) {
1371	soisfdisconnected(so);
1372	tp->t_timer[TCPT_2MSL] = tcp_maxidle;
1373	}
1374	tp->t_state = TCPS_FIN_WAIT_2;
1375	}
1376	break;
1377
1378	/*
1379	* In CLOSING STATE in addition to the processing for
1380	* the ESTABLISHED state if the ACK acknowledges our FIN
1381	* then enter the TIME-WAIT state, otherwise ignore
1382	* the segment.
1383	*/
1384	case TCPS_CLOSING:
1385	if (ourfinisacked) {
1386	tp->t_state = TCPS_TIME_WAIT;
1387	tcp_canceltimers(tp);
1388	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1389	soisfdisconnected(so);
1390	}
1391	break;
1392
1393	/*
1394	* In LAST_ACK, we may still be waiting for data to drain
1395	* and/or to be acked, as well as for the ack of our FIN.
1396	* If our FIN is now acknowledged, delete the TCB,
1397	* enter the closed state and return.
1398	*/
1399	case TCPS_LAST_ACK:
1400	if (ourfinisacked) {
1401	tp = tcp_close(pData, tp);
1402	goto drop;
1403	}
1404	break;
1405
1406	/*
1407	* In TIME_WAIT state the only thing that should arrive
1408	* is a retransmission of the remote FIN. Acknowledge
1409	* it and restart the finack timer.
1410	*/
1411	case TCPS_TIME_WAIT:
1412	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1413	goto dropafterack;
1414	}
1415	} /* switch(tp->t_state) */
1416
1417	step6:
1418	/*
1419	* Update window information.
1420	* Don't look at window if no ACK: TAC's send garbage on first SYN.
1421	*/
1422	if ((tiflags & TH_ACK) &&
1423	(SEQ_LT(tp->snd_wl1, ti->ti_seq) \|\|
1424	(tp->snd_wl1 == ti->ti_seq && (SEQ_LT(tp->snd_wl2, ti->ti_ack) \|\|
1425	(tp->snd_wl2 == ti->ti_ack && tiwin > tp->snd_wnd))))) {
1426	/* keep track of pure window updates */
1427	if (ti->ti_len == 0 &&
1428	tp->snd_wl2 == ti->ti_ack && tiwin > tp->snd_wnd)
1429	tcpstat.tcps_rcvwinupd++;
1430	tp->snd_wnd = tiwin;
1431	tp->snd_wl1 = ti->ti_seq;
1432	tp->snd_wl2 = ti->ti_ack;
1433	if (tp->snd_wnd > tp->max_sndwnd)
1434	tp->max_sndwnd = tp->snd_wnd;
1435	needoutput = 1;
1436	}
1437
1438	/*
1439	* Process segments with URG.
1440	*/
1441	if ((tiflags & TH_URG) && ti->ti_urp &&
1442	TCPS_HAVERCVDFIN(tp->t_state) == 0) {
1443	/*
1444	* This is a kludge, but if we receive and accept
1445	* random urgent pointers, we'll crash in
1446	* soreceive. It's hard to imagine someone
1447	* actually wanting to send this much urgent data.
1448	*/
1449	if (ti->ti_urp + so->so_rcv.sb_cc > so->so_rcv.sb_datalen) {
1450	ti->ti_urp = 0;
1451	tiflags &= ~TH_URG;
1452	goto dodata;
1453	}
1454	/*
1455	* If this segment advances the known urgent pointer,
1456	* then mark the data stream. This should not happen
1457	* in CLOSE_WAIT, CLOSING, LAST_ACK or TIME_WAIT STATES since
1458	* a FIN has been received from the remote side.
1459	* In these states we ignore the URG.
1460	*
1461	* According to RFC961 (Assigned Protocols),
1462	* the urgent pointer points to the last octet
1463	* of urgent data. We continue, however,
1464	* to consider it to indicate the first octet
1465	* of data past the urgent section as the original
1466	* spec states (in one of two places).
1467	*/
1468	if (SEQ_GT(ti->ti_seq+ti->ti_urp, tp->rcv_up)) {
1469	tp->rcv_up = ti->ti_seq + ti->ti_urp;
1470	so->so_urgc = so->so_rcv.sb_cc +
1471	(tp->rcv_up - tp->rcv_nxt); /* -1; */
1472	tp->rcv_up = ti->ti_seq + ti->ti_urp;
1473
1474	}
1475	} else
1476	/*
1477	* If no out of band data is expected,
1478	* pull receive urgent pointer along
1479	* with the receive window.
1480	*/
1481	if (SEQ_GT(tp->rcv_nxt, tp->rcv_up))
1482	tp->rcv_up = tp->rcv_nxt;
1483	dodata:
1484
1485	/*
1486	* If this is a small packet, then ACK now - with Nagel
1487	* congestion avoidance sender won't send more until
1488	* he gets an ACK.
1489	*
1490	* See above.
1491	*/
1492	if (ti->ti_len && (unsigned)ti->ti_len <= 5 &&
1493	((struct tcpiphdr_2 *)ti)->first_char == (char)27) {
1494	tp->t_flags \|= TF_ACKNOW;
1495	}
1496
1497	/*
1498	* Process the segment text, merging it into the TCP sequencing queue,
1499	* and arranging for acknowledgment of receipt if necessary.
1500	* This process logically involves adjusting tp->rcv_wnd as data
1501	* is presented to the user (this happens in tcp_usrreq.c,
1502	* case PRU_RCVD). If a FIN has already been received on this
1503	* connection then we just ignore the text.
1504	*/
1505	if ((ti->ti_len \|\| (tiflags&TH_FIN)) &&
1506	TCPS_HAVERCVDFIN(tp->t_state) == 0) {
1507	#ifndef VBOX_WITH_BSD_REASS
1508	TCP_REASS(pData, tp, ti, m, so, tiflags);
1509	#else /* VBOX_WITH_BSD_REASS */
1510	if (ti->ti_seq == tp->rcv_nxt
1511	&& LIST_EMPTY(&tp->t_segq)
1512	&& tp->t_state == TCPS_ESTABLISHED) {
1513	DELAY_ACK(tp, ti); /* little bit different from BSD declaration see netinet/tcp_input.c */
1514	tp->rcv_nxt += tlen;
1515	tiflags = ti->ti_t.th_flags & TH_FIN;
1516	tcpstat.tcps_rcvpack++;
1517	tcpstat.tcps_rcvbyte += tlen;
1518	if (so->so_state & SS_FCANTRCVMORE)
1519	m_freem(pData, m);
1520	else
1521	sbappend(pData, so, m);
1522	}
1523	else {
1524	tiflags = tcp_reass(pData, tp, &ti->ti_t, &tlen, m);
1525	tiflags \|= TF_ACKNOW;
1526	}
1527	#endif /* VBOX_WITH_BSD_REASS */
1528	/*
1529	* Note the amount of data that peer has sent into
1530	* our window, in order to estimate the sender's
1531	* buffer size.
1532	*/
1533	len = so->so_rcv.sb_datalen - (tp->rcv_adv - tp->rcv_nxt);
1534	} else {
1535	m_free(pData, m);
1536	tiflags &= ~TH_FIN;
1537	}
1538
1539	/*
1540	* If FIN is received ACK the FIN and let the user know
1541	* that the connection is closing.
1542	*/
1543	if (tiflags & TH_FIN) {
1544	if (TCPS_HAVERCVDFIN(tp->t_state) == 0) {
1545	/*
1546	* If we receive a FIN we can't send more data,
1547	* set it SS_FDRAIN
1548	* Shutdown the socket if there is no rx data in the
1549	* buffer.
1550	* soread() is called on completion of shutdown() and
1551	* will got to TCPS_LAST_ACK, and use tcp_output()
1552	* to send the FIN.
1553	*/
1554	/* sofcantrcvmore(so); */
1555	sofwdrain(so);
1556
1557	tp->t_flags \|= TF_ACKNOW;
1558	tp->rcv_nxt++;
1559	}
1560	switch (tp->t_state) {
1561
1562	/*
1563	* In SYN_RECEIVED and ESTABLISHED STATES
1564	* enter the CLOSE_WAIT state.
1565	*/
1566	case TCPS_SYN_RECEIVED:
1567	case TCPS_ESTABLISHED:
1568	if(so->so_emu == EMU_CTL) /* no shutdown on socket */
1569	tp->t_state = TCPS_LAST_ACK;
1570	else
1571	tp->t_state = TCPS_CLOSE_WAIT;
1572	break;
1573
1574	/*
1575	* If still in FIN_WAIT_1 STATE FIN has not been acked so
1576	* enter the CLOSING state.
1577	*/
1578	case TCPS_FIN_WAIT_1:
1579	tp->t_state = TCPS_CLOSING;
1580	break;
1581
1582	/*
1583	* In FIN_WAIT_2 state enter the TIME_WAIT state,
1584	* starting the time-wait timer, turning off the other
1585	* standard timers.
1586	*/
1587	case TCPS_FIN_WAIT_2:
1588	tp->t_state = TCPS_TIME_WAIT;
1589	tcp_canceltimers(tp);
1590	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1591	soisfdisconnected(so);
1592	break;
1593
1594	/*
1595	* In TIME_WAIT state restart the 2 MSL time_wait timer.
1596	*/
1597	case TCPS_TIME_WAIT:
1598	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1599	break;
1600	}
1601	}
1602
1603	/*
1604	* Return any desired output.
1605	*/
1606	if (needoutput \|\| (tp->t_flags & TF_ACKNOW)) {
1607	(void) tcp_output(pData, tp);
1608	}
1609	return;
1610
1611	dropafterack:
1612	/*
1613	* Generate an ACK dropping incoming segment if it occupies
1614	* sequence space, where the ACK reflects our state.
1615	*/
1616	if (tiflags & TH_RST)
1617	goto drop;
1618	m_freem(pData, m);
1619	tp->t_flags \|= TF_ACKNOW;
1620	(void) tcp_output(pData, tp);
1621	return;
1622
1623	dropwithreset:
1624	/* reuses m if m!=NULL, m_free() unnecessary */
1625	if (tiflags & TH_ACK)
1626	tcp_respond(pData, tp, ti, m, (tcp_seq)0, ti->ti_ack, TH_RST);
1627	else {
1628	if (tiflags & TH_SYN) ti->ti_len++;
1629	tcp_respond(pData, tp, ti, m, ti->ti_seq+ti->ti_len, (tcp_seq)0,
1630	TH_RST\|TH_ACK);
1631	}
1632
1633	return;
1634
1635	drop:
1636	/*
1637	* Drop space held by incoming segment and return.
1638	*/
1639	m_free(pData, m);
1640
1641	return;
1642	}
1643
1644	/* , ts_present, ts_val, ts_ecr) */
1645	/* int *ts_present;
1646	* u_int32_t ts_val, ts_ecr;
1647	*/
1648	void
1649	tcp_dooptions(PNATState pData, struct tcpcb tp, u_char cp, int cnt, struct tcpiphdr *ti)
1650	{
1651	u_int16_t mss;
1652	int opt, optlen;
1653
1654	DEBUG_CALL("tcp_dooptions");
1655	DEBUG_ARGS((dfd," tp = %lx cnt=%i \n", (long )tp, cnt));
1656
1657	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1658	opt = cp[0];
1659	if (opt == TCPOPT_EOL)
1660	break;
1661	if (opt == TCPOPT_NOP)
1662	optlen = 1;
1663	else {
1664	optlen = cp[1];
1665	if (optlen <= 0)
1666	break;
1667	}
1668	switch (opt) {
1669
1670	default:
1671	continue;
1672
1673	case TCPOPT_MAXSEG:
1674	if (optlen != TCPOLEN_MAXSEG)
1675	continue;
1676	if (!(ti->ti_flags & TH_SYN))
1677	continue;
1678	memcpy((char ) &mss, (char ) cp + 2, sizeof(mss));
1679	NTOHS(mss);
1680	(void) tcp_mss(pData, tp, mss); /* sets t_maxseg */
1681	break;
1682
1683	/* case TCPOPT_WINDOW:
1684	* if (optlen != TCPOLEN_WINDOW)
1685	* continue;
1686	* if (!(ti->ti_flags & TH_SYN))
1687	* continue;
1688	* tp->t_flags \|= TF_RCVD_SCALE;
1689	* tp->requested_s_scale = min(cp[2], TCP_MAX_WINSHIFT);
1690	* break;
1691	*/
1692	/* case TCPOPT_TIMESTAMP:
1693	* if (optlen != TCPOLEN_TIMESTAMP)
1694	* continue;
1695	* *ts_present = 1;
1696	* memcpy((char ) ts_val, (char )cp + 2, sizeof(*ts_val));
1697	* NTOHL(*ts_val);
1698	* memcpy((char ) ts_ecr, (char )cp + 6, sizeof(*ts_ecr));
1699	* NTOHL(*ts_ecr);
1700	*
1701	/ /
1702	* * A timestamp received in a SYN makes
1703	* * it ok to send timestamp requests and replies.
1704	* */
1705	/* if (ti->ti_flags & TH_SYN) {
1706	* tp->t_flags \|= TF_RCVD_TSTMP;
1707	* tp->ts_recent = *ts_val;
1708	* tp->ts_recent_age = tcp_now;
1709	* }
1710	*/ break;
1711	}
1712	}
1713	}
1714
1715
1716	/*
1717	* Pull out of band byte out of a segment so
1718	* it doesn't appear in the user's data queue.
1719	* It is still reflected in the segment length for
1720	* sequencing purposes.
1721	*/
1722
1723	#ifdef notdef
1724
1725	void
1726	tcp_pulloutofband(so, ti, m)
1727	struct socket *so;
1728	struct tcpiphdr *ti;
1729	register struct mbuf *m;
1730	{
1731	int cnt = ti->ti_urp - 1;
1732
1733	while (cnt >= 0) {
1734	if (m->m_len > cnt) {
1735	char *cp = mtod(m, caddr_t) + cnt;
1736	struct tcpcb *tp = sototcpcb(so);
1737
1738	tp->t_iobc = *cp;
1739	tp->t_oobflags \|= TCPOOB_HAVEDATA;
1740	memcpy(sp, cp+1, (unsigned)(m->m_len - cnt - 1));
1741	m->m_len--;
1742	return;
1743	}
1744	cnt -= m->m_len;
1745	m = m->m_next; /* XXX WRONG! Fix it! */
1746	if (m == 0)
1747	break;
1748	}
1749	panic("tcp_pulloutofband");
1750	}
1751
1752	#endif /* notdef */
1753
1754	/*
1755	* Collect new round-trip time estimate
1756	* and update averages and current timeout.
1757	*/
1758
1759	void
1760	tcp_xmit_timer(PNATState pData, register struct tcpcb *tp, int rtt)
1761	{
1762	register short delta;
1763
1764	DEBUG_CALL("tcp_xmit_timer");
1765	DEBUG_ARG("tp = %lx", (long)tp);
1766	DEBUG_ARG("rtt = %d", rtt);
1767
1768	tcpstat.tcps_rttupdated++;
1769	if (tp->t_srtt != 0) {
1770	/*
1771	* srtt is stored as fixed point with 3 bits after the
1772	* binary point (i.e., scaled by 8). The following magic
1773	* is equivalent to the smoothing algorithm in rfc793 with
1774	* an alpha of .875 (srtt = rtt/8 + srtt*7/8 in fixed
1775	* point). Adjust rtt to origin 0.
1776	*/
1777	delta = rtt - 1 - (tp->t_srtt >> TCP_RTT_SHIFT);
1778	if ((tp->t_srtt += delta) <= 0)
1779	tp->t_srtt = 1;
1780	/*
1781	* We accumulate a smoothed rtt variance (actually, a
1782	* smoothed mean difference), then set the retransmit
1783	* timer to smoothed rtt + 4 times the smoothed variance.
1784	* rttvar is stored as fixed point with 2 bits after the
1785	* binary point (scaled by 4). The following is
1786	* equivalent to rfc793 smoothing with an alpha of .75
1787	* (rttvar = rttvar*3/4 + \|delta\| / 4). This replaces
1788	* rfc793's wired-in beta.
1789	*/
1790	if (delta < 0)
1791	delta = -delta;
1792	delta -= (tp->t_rttvar >> TCP_RTTVAR_SHIFT);
1793	if ((tp->t_rttvar += delta) <= 0)
1794	tp->t_rttvar = 1;
1795	} else {
1796	/*
1797	* No rtt measurement yet - use the unsmoothed rtt.
1798	* Set the variance to half the rtt (so our first
1799	* retransmit happens at 3*rtt).
1800	*/
1801	tp->t_srtt = rtt << TCP_RTT_SHIFT;
1802	tp->t_rttvar = rtt << (TCP_RTTVAR_SHIFT - 1);
1803	}
1804	tp->t_rtt = 0;
1805	tp->t_rxtshift = 0;
1806
1807	/*
1808	* the retransmit should happen at rtt + 4 * rttvar.
1809	* Because of the way we do the smoothing, srtt and rttvar
1810	* will each average +1/2 tick of bias. When we compute
1811	* the retransmit timer, we want 1/2 tick of rounding and
1812	* 1 extra tick because of +-1/2 tick uncertainty in the
1813	* firing of the timer. The bias will give us exactly the
1814	* 1.5 tick we need. But, because the bias is
1815	* statistical, we have to test that we don't drop below
1816	* the minimum feasible timer (which is 2 ticks).
1817	*/
1818	TCPT_RANGESET(tp->t_rxtcur, TCP_REXMTVAL(tp),
1819	(short)tp->t_rttmin, TCPTV_REXMTMAX); /* XXX */
1820
1821	/*
1822	* We received an ack for a packet that wasn't retransmitted;
1823	* it is probably safe to discard any error indications we've
1824	* received recently. This isn't quite right, but close enough
1825	* for now (a route might have failed after we sent a segment,
1826	* and the return path might not be symmetrical).
1827	*/
1828	tp->t_softerror = 0;
1829	}
1830
1831	/*
1832	* Determine a reasonable value for maxseg size.
1833	* If the route is known, check route for mtu.
1834	* If none, use an mss that can be handled on the outgoing
1835	* interface without forcing IP to fragment; if bigger than
1836	* an mbuf cluster (MCLBYTES), round down to nearest multiple of MCLBYTES
1837	* to utilize large mbufs. If no route is found, route has no mtu,
1838	* or the destination isn't local, use a default, hopefully conservative
1839	* size (usually 512 or the default IP max size, but no more than the mtu
1840	* of the interface), as we can't discover anything about intervening
1841	* gateways or networks. We also initialize the congestion/slow start
1842	* window to be a single segment if the destination isn't local.
1843	* While looking at the routing entry, we also initialize other path-dependent
1844	* parameters from pre-set or cached values in the routing entry.
1845	*/
1846
1847	int
1848	tcp_mss(PNATState pData, register struct tcpcb *tp, u_int offer)
1849	{
1850	struct socket *so = tp->t_socket;
1851	int mss;
1852
1853	DEBUG_CALL("tcp_mss");
1854	DEBUG_ARG("tp = %lx", (long)tp);
1855	DEBUG_ARG("offer = %d", offer);
1856
1857	mss = min(if_mtu, if_mru) - sizeof(struct tcpiphdr);
1858	if (offer)
1859	mss = min(mss, offer);
1860	mss = max(mss, 32);
1861	if (mss < tp->t_maxseg \|\| offer != 0)
1862	tp->t_maxseg = mss;
1863
1864	tp->snd_cwnd = mss;
1865
1866	sbreserve(&so->so_snd, tcp_sndspace+((tcp_sndspace%mss)?(mss-(tcp_sndspace%mss)):0));
1867	sbreserve(&so->so_rcv, tcp_rcvspace+((tcp_rcvspace%mss)?(mss-(tcp_rcvspace%mss)):0));
1868
1869	DEBUG_MISC((dfd, " returning mss = %d\n", mss));
1870
1871	return mss;
1872	}

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

source: vbox/trunk/src/VBox/Devices/Network/slirp/tcp_input.c@ 15136

Download in other formats: