tcp_input.c@ 14470

最後變更在這個檔案從14470是 14470,由 vboxsync 提交於 16 年前
slirp: remove tabs
屬性 svn:eol-style 設為 `native`
檔案大小: 73.5 KB

行
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 ret;
424	/* int ts_present = 0; */
425	int mbuf_freed = 0;
426
427	DEBUG_CALL("tcp_input");
428	DEBUG_ARGS((dfd," m = %8lx iphlen = %2d inso = %lx\n",
429	(long )m, iphlen, (long )inso ));
430
431	/*
432	* If called with m == 0, then we're continuing the connect
433	*/
434	if (m == NULL) {
435	so = inso;
436
437	/* Re-set a few variables */
438	tp = sototcpcb(so);
439	m = so->so_m;
440	so->so_m = 0;
441	ti = so->so_ti;
442	tiwin = ti->ti_win;
443	tiflags = ti->ti_flags;
444
445	goto cont_conn;
446	}
447
448
449	tcpstat.tcps_rcvtotal++;
450	/*
451	* Get IP and TCP header together in first mbuf.
452	* Note: IP leaves IP header in first mbuf.
453	*/
454	ti = mtod(m, struct tcpiphdr *);
455	if (iphlen > sizeof(struct ip )) {
456	ip_stripoptions(m, (struct mbuf *)0);
457	iphlen=sizeof(struct ip );
458	}
459	/* XXX Check if too short */
460
461
462	/*
463	* Save a copy of the IP header in case we want restore it
464	* for sending an ICMP error message in response.
465	*/
466	ip=mtod(m, struct ip *);
467	save_ip = *ip;
468	save_ip.ip_len+= iphlen;
469
470	/*
471	* Checksum extended TCP header and data.
472	*/
473	tlen = ((struct ip *)ti)->ip_len;
474	ti->ti_next = ti->ti_prev = 0;
475	ti->ti_x1 = 0;
476	ti->ti_len = htons((u_int16_t)tlen);
477	len = sizeof(struct ip ) + tlen;
478	/* keep checksum for ICMP reply
479	* ti->ti_sum = cksum(m, len);
480	* if (ti->ti_sum) { */
481	if(cksum(m, len)) {
482	tcpstat.tcps_rcvbadsum++;
483	goto drop;
484	}
485
486	/*
487	* Check that TCP offset makes sense,
488	* pull out TCP options and adjust length. XXX
489	*/
490	off = ti->ti_off << 2;
491	if (off < sizeof (struct tcphdr) \|\| off > tlen) {
492	tcpstat.tcps_rcvbadoff++;
493	goto drop;
494	}
495	tlen -= off;
496	ti->ti_len = tlen;
497	if (off > sizeof (struct tcphdr)) {
498	optlen = off - sizeof (struct tcphdr);
499	optp = mtod(m, caddr_t) + sizeof (struct tcpiphdr);
500
501	/*
502	* Do quick retrieval of timestamp options ("options
503	* prediction?"). If timestamp is the only option and it's
504	* formatted as recommended in RFC 1323 appendix A, we
505	* quickly get the values now and not bother calling
506	* tcp_dooptions(), etc.
507	*/
508	/* if ((optlen == TCPOLEN_TSTAMP_APPA \|\|
509	* (optlen > TCPOLEN_TSTAMP_APPA &&
510	* optp[TCPOLEN_TSTAMP_APPA] == TCPOPT_EOL)) &&
511	* (u_int32_t )optp == htonl(TCPOPT_TSTAMP_HDR) &&
512	* (ti->ti_flags & TH_SYN) == 0) {
513	* ts_present = 1;
514	* ts_val = ntohl((u_int32_t )(optp + 4));
515	* ts_ecr = ntohl((u_int32_t )(optp + 8));
516	* optp = NULL; / * we've parsed the options * /
517	* }
518	*/
519	}
520	tiflags = ti->ti_flags;
521
522	/*
523	* Convert TCP protocol specific fields to host format.
524	*/
525	NTOHL(ti->ti_seq);
526	NTOHL(ti->ti_ack);
527	NTOHS(ti->ti_win);
528	NTOHS(ti->ti_urp);
529
530	/*
531	* Drop TCP, IP headers and TCP options.
532	*/
533	m->m_data += sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
534	m->m_len -= sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
535
536	/*
537	* Locate pcb for segment.
538	*/
539	findso:
540	so = tcp_last_so;
541	if (so->so_fport != ti->ti_dport \|\|
542	so->so_lport != ti->ti_sport \|\|
543	so->so_laddr.s_addr != ti->ti_src.s_addr \|\|
544	so->so_faddr.s_addr != ti->ti_dst.s_addr) {
545	so = solookup(&tcb, ti->ti_src, ti->ti_sport,
546	ti->ti_dst, ti->ti_dport);
547	if (so)
548	tcp_last_so = so;
549	++tcpstat.tcps_socachemiss;
550	}
551
552	/*
553	* If the state is CLOSED (i.e., TCB does not exist) then
554	* all data in the incoming segment is discarded.
555	* If the TCB exists but is in CLOSED state, it is embryonic,
556	* but should either do a listen or a connect soon.
557	*
558	* state == CLOSED means we've done socreate() but haven't
559	* attached it to a protocol yet...
560	*
561	* XXX If a TCB does not exist, and the TH_SYN flag is
562	* the only flag set, then create a session, mark it
563	* as if it was LISTENING, and continue...
564	*/
565	if (so == 0) {
566	if ((tiflags & (TH_SYN\|TH_FIN\|TH_RST\|TH_URG\|TH_ACK)) != TH_SYN)
567	goto dropwithreset;
568
569	if ((so = socreate()) == NULL)
570	goto dropwithreset;
571	if (tcp_attach(pData, so) < 0) {
572	free(so); /* Not sofree (if it failed, it's not insqued) */
573	goto dropwithreset;
574	}
575
576	sbreserve(&so->so_snd, tcp_sndspace);
577	sbreserve(&so->so_rcv, tcp_rcvspace);
578
579	/* tcp_last_so = so; / / XXX ? */
580	/* tp = sototcpcb(so); */
581
582	so->so_laddr = ti->ti_src;
583	so->so_lport = ti->ti_sport;
584	so->so_faddr = ti->ti_dst;
585	so->so_fport = ti->ti_dport;
586
587	if ((so->so_iptos = tcp_tos(so)) == 0)
588	so->so_iptos = ((struct ip *)ti)->ip_tos;
589
590	tp = sototcpcb(so);
591	tp->t_state = TCPS_LISTEN;
592	}
593
594	/*
595	* If this is a still-connecting socket, this probably
596	* a retransmit of the SYN. Whether it's a retransmit SYN
597	* or something else, we nuke it.
598	*/
599	if (so->so_state & SS_ISFCONNECTING)
600	goto drop;
601
602	tp = sototcpcb(so);
603
604	/* XXX Should never fail */
605	if (tp == 0)
606	goto dropwithreset;
607	if (tp->t_state == TCPS_CLOSED)
608	goto drop;
609
610	/* Unscale the window into a 32-bit value. */
611	/* if ((tiflags & TH_SYN) == 0)
612	* tiwin = ti->ti_win << tp->snd_scale;
613	* else
614	*/
615	tiwin = ti->ti_win;
616
617	/*
618	* Segment received on connection.
619	* Reset idle time and keep-alive timer.
620	*/
621	tp->t_idle = 0;
622	if (so_options)
623	tp->t_timer[TCPT_KEEP] = tcp_keepintvl;
624	else
625	tp->t_timer[TCPT_KEEP] = tcp_keepidle;
626
627	/*
628	* Process options if not in LISTEN state,
629	* else do it below (after getting remote address).
630	*/
631	if (optp && tp->t_state != TCPS_LISTEN)
632	tcp_dooptions(pData, tp, (u_char *)optp, optlen, ti);
633	/* , */
634	/* &ts_present, &ts_val, &ts_ecr); */
635
636	/*
637	* Header prediction: check for the two common cases
638	* of a uni-directional data xfer. If the packet has
639	* no control flags, is in-sequence, the window didn't
640	* change and we're not retransmitting, it's a
641	* candidate. If the length is zero and the ack moved
642	* forward, we're the sender side of the xfer. Just
643	* free the data acked & wake any higher level process
644	* that was blocked waiting for space. If the length
645	* is non-zero and the ack didn't move, we're the
646	* receiver side. If we're getting packets in-order
647	* (the reassembly queue is empty), add the data to
648	* the socket buffer and note that we need a delayed ack.
649	*
650	* XXX Some of these tests are not needed
651	* eg: the tiwin == tp->snd_wnd prevents many more
652	* predictions.. with no real advantage..
653	*/
654	if (tp->t_state == TCPS_ESTABLISHED &&
655	(tiflags & (TH_SYN\|TH_FIN\|TH_RST\|TH_URG\|TH_ACK)) == TH_ACK &&
656	/* (!ts_present \|\| TSTMP_GEQ(ts_val, tp->ts_recent)) && */
657	ti->ti_seq == tp->rcv_nxt &&
658	tiwin && tiwin == tp->snd_wnd &&
659	tp->snd_nxt == tp->snd_max) {
660	/*
661	* If last ACK falls within this segment's sequence numbers,
662	* record the timestamp.
663	*/
664	/* if (ts_present && SEQ_LEQ(ti->ti_seq, tp->last_ack_sent) &&
665	* SEQ_LT(tp->last_ack_sent, ti->ti_seq + ti->ti_len)) {
666	* tp->ts_recent_age = tcp_now;
667	* tp->ts_recent = ts_val;
668	* }
669	*/
670	if (ti->ti_len == 0) {
671	if (SEQ_GT(ti->ti_ack, tp->snd_una) &&
672	SEQ_LEQ(ti->ti_ack, tp->snd_max) &&
673	tp->snd_cwnd >= tp->snd_wnd) {
674	/*
675	* this is a pure ack for outstanding data.
676	*/
677	++tcpstat.tcps_predack;
678	/* if (ts_present)
679	* tcp_xmit_timer(tp, tcp_now-ts_ecr+1);
680	* else
681	*/ if (tp->t_rtt &&
682	SEQ_GT(ti->ti_ack, tp->t_rtseq))
683	tcp_xmit_timer(pData, tp, tp->t_rtt);
684	acked = ti->ti_ack - tp->snd_una;
685	tcpstat.tcps_rcvackpack++;
686	tcpstat.tcps_rcvackbyte += acked;
687	sbdrop(&so->so_snd, acked);
688	tp->snd_una = ti->ti_ack;
689	m_freem(pData, m);
690
691	/*
692	* If all outstanding data are acked, stop
693	* retransmit timer, otherwise restart timer
694	* using current (possibly backed-off) value.
695	* If process is waiting for space,
696	* wakeup/selwakeup/signal. If data
697	* are ready to send, let tcp_output
698	* decide between more output or persist.
699	*/
700	if (tp->snd_una == tp->snd_max)
701	tp->t_timer[TCPT_REXMT] = 0;
702	else if (tp->t_timer[TCPT_PERSIST] == 0)
703	tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
704
705	/*
706	* There's room in so_snd, sowwakup will read()
707	* from the socket if we can
708	*/
709	/* if (so->so_snd.sb_flags & SB_NOTIFY)
710	* sowwakeup(so);
711	*/
712	/*
713	* This is called because sowwakeup might have
714	* put data into so_snd. Since we don't so sowwakeup,
715	* we don't need this.. XXX???
716	*/
717	if (so->so_snd.sb_cc)
718	(void) tcp_output(pData, tp);
719
720	return;
721	}
722	} else if (ti->ti_ack == tp->snd_una &&
723	#ifndef VBOX_WITH_BSD_TCP_REASS
724	u32_to_ptr(pData, tp->seg_next, struct tcpcb *) == tp &&
725	#else /* VBOX_WITH_BSD_TCP_REASS */
726	LIST_FIRST(&tp->t_segq) &&
727	#endif /* VBOX_WITH_BSD_TCP_REASS */
728	ti->ti_len <= sbspace(&so->so_rcv)) {
729	/*
730	* this is a pure, in-sequence data packet
731	* with nothing on the reassembly queue and
732	* we have enough buffer space to take it.
733	*/
734	++tcpstat.tcps_preddat;
735	tp->rcv_nxt += ti->ti_len;
736	tcpstat.tcps_rcvpack++;
737	tcpstat.tcps_rcvbyte += ti->ti_len;
738	/*
739	* Add data to socket buffer.
740	*/
741	if (so->so_emu) {
742	if (tcp_emu(pData, so,m)) sbappend(pData, so, m);
743	} else
744	sbappend(pData, so, m);
745
746	/*
747	* XXX This is called when data arrives. Later, check
748	* if we can actually write() to the socket
749	* XXX Need to check? It's be NON_BLOCKING
750	*/
751	/* sorwakeup(so); */
752
753	/*
754	* If this is a short packet, then ACK now - with Nagel
755	* congestion avoidance sender won't send more until
756	* he gets an ACK.
757	*
758	* It is better to not delay acks at all to maximize
759	* TCP throughput. See RFC 2581.
760	*/
761	tp->t_flags \|= TF_ACKNOW;
762	tcp_output(pData, tp);
763	return;
764	}
765	} /* header prediction */
766	/*
767	* Calculate amount of space in receive window,
768	* and then do TCP input processing.
769	* Receive window is amount of space in rcv queue,
770	* but not less than advertised window.
771	*/
772	{ int win;
773	win = sbspace(&so->so_rcv);
774	if (win < 0)
775	win = 0;
776	tp->rcv_wnd = max(win, (int)(tp->rcv_adv - tp->rcv_nxt));
777	}
778
779	switch (tp->t_state) {
780
781	/*
782	* If the state is LISTEN then ignore segment if it contains an RST.
783	* If the segment contains an ACK then it is bad and send a RST.
784	* If it does not contain a SYN then it is not interesting; drop it.
785	* Don't bother responding if the destination was a broadcast.
786	* Otherwise initialize tp->rcv_nxt, and tp->irs, select an initial
787	* tp->iss, and send a segment:
788	* <SEQ=ISS><ACK=RCV_NXT><CTL=SYN,ACK>
789	* Also initialize tp->snd_nxt to tp->iss+1 and tp->snd_una to tp->iss.
790	* Fill in remote peer address fields if not previously specified.
791	* Enter SYN_RECEIVED state, and process any other fields of this
792	* segment in this state.
793	*/
794	case TCPS_LISTEN: {
795
796	if (tiflags & TH_RST)
797	goto drop;
798	if (tiflags & TH_ACK)
799	goto dropwithreset;
800	if ((tiflags & TH_SYN) == 0)
801	goto drop;
802
803	/*
804	* This has way too many gotos...
805	* But a bit of spaghetti code never hurt anybody :)
806	*/
807
808	if (so->so_emu & EMU_NOCONNECT) {
809	so->so_emu &= ~EMU_NOCONNECT;
810	goto cont_input;
811	}
812
813	if((tcp_fconnect(pData, so) == -1) && (errno != EINPROGRESS) && (errno != EWOULDBLOCK)) {
814	u_char code=ICMP_UNREACH_NET;
815	DEBUG_MISC((dfd," tcp fconnect errno = %d-%s\n",
816	errno,strerror(errno)));
817	if(errno == ECONNREFUSED) {
818	/* ACK the SYN, send RST to refuse the connection */
819	tcp_respond(pData, tp, ti, m, ti->ti_seq+1, (tcp_seq)0,
820	TH_RST\|TH_ACK);
821	} else {
822	if(errno == EHOSTUNREACH) code=ICMP_UNREACH_HOST;
823	HTONL(ti->ti_seq); /* restore tcp header */
824	HTONL(ti->ti_ack);
825	HTONS(ti->ti_win);
826	HTONS(ti->ti_urp);
827	m->m_data -= sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
828	m->m_len += sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
829	*ip=save_ip;
830	icmp_error(pData, m, ICMP_UNREACH,code, 0,strerror(errno));
831	}
832	tp = tcp_close(pData, tp);
833	m_free(pData, m);
834	} else {
835	/*
836	* Haven't connected yet, save the current mbuf
837	* and ti, and return
838	* XXX Some OS's don't tell us whether the connect()
839	* succeeded or not. So we must time it out.
840	*/
841	so->so_m = m;
842	so->so_ti = ti;
843	tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT;
844	tp->t_state = TCPS_SYN_RECEIVED;
845	}
846	return;
847
848	cont_conn:
849	/* m==NULL
850	* Check if the connect succeeded
851	*/
852	if (so->so_state & SS_NOFDREF) {
853	tp = tcp_close(pData, tp);
854	goto dropwithreset;
855	}
856	cont_input:
857	tcp_template(tp);
858
859	if (optp)
860	tcp_dooptions(pData, tp, (u_char *)optp, optlen, ti);
861	/* , */
862	/* &ts_present, &ts_val, &ts_ecr); */
863
864	if (iss)
865	tp->iss = iss;
866	else
867	tp->iss = tcp_iss;
868	tcp_iss += TCP_ISSINCR/2;
869	tp->irs = ti->ti_seq;
870	tcp_sendseqinit(tp);
871	tcp_rcvseqinit(tp);
872	tp->t_flags \|= TF_ACKNOW;
873	tp->t_state = TCPS_SYN_RECEIVED;
874	tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT;
875	tcpstat.tcps_accepts++;
876	goto trimthenstep6;
877	} /* case TCPS_LISTEN */
878
879	/*
880	* If the state is SYN_SENT:
881	* if seg contains an ACK, but not for our SYN, drop the input.
882	* if seg contains a RST, then drop the connection.
883	* if seg does not contain SYN, then drop it.
884	* Otherwise this is an acceptable SYN segment
885	* initialize tp->rcv_nxt and tp->irs
886	* if seg contains ack then advance tp->snd_una
887	* if SYN has been acked change to ESTABLISHED else SYN_RCVD state
888	* arrange for segment to be acked (eventually)
889	* continue processing rest of data/controls, beginning with URG
890	*/
891	case TCPS_SYN_SENT:
892	if ((tiflags & TH_ACK) &&
893	(SEQ_LEQ(ti->ti_ack, tp->iss) \|\|
894	SEQ_GT(ti->ti_ack, tp->snd_max)))
895	goto dropwithreset;
896
897	if (tiflags & TH_RST) {
898	if (tiflags & TH_ACK)
899	tp = tcp_drop(pData, tp,0); /* XXX Check t_softerror! */
900	goto drop;
901	}
902
903	if ((tiflags & TH_SYN) == 0)
904	goto drop;
905	if (tiflags & TH_ACK) {
906	tp->snd_una = ti->ti_ack;
907	if (SEQ_LT(tp->snd_nxt, tp->snd_una))
908	tp->snd_nxt = tp->snd_una;
909	}
910
911	tp->t_timer[TCPT_REXMT] = 0;
912	tp->irs = ti->ti_seq;
913	tcp_rcvseqinit(tp);
914	tp->t_flags \|= TF_ACKNOW;
915	if (tiflags & TH_ACK && SEQ_GT(tp->snd_una, tp->iss)) {
916	tcpstat.tcps_connects++;
917	soisfconnected(so);
918	tp->t_state = TCPS_ESTABLISHED;
919
920	/* Do window scaling on this connection? */
921	/* if ((tp->t_flags & (TF_RCVD_SCALE\|TF_REQ_SCALE)) ==
922	* (TF_RCVD_SCALE\|TF_REQ_SCALE)) {
923	* tp->snd_scale = tp->requested_s_scale;
924	* tp->rcv_scale = tp->request_r_scale;
925	* }
926	*/
927	#ifndef VBOX_WITH_BSD_TCP_REASS
928	(void) tcp_reass(pData, tp, (struct tcpiphdr *)0,
929	(struct mbuf *)0);
930	#else /* VBOX_WITH_BSD_TCP_REASS */
931	(void) tcp_reass(pData, tp, (struct tcphdr )0, NULL, (struct mbuf )0);
932	#endif /* VBOX_WITH_BSD_TCP_REASS */
933	/*
934	* if we didn't have to retransmit the SYN,
935	* use its rtt as our initial srtt & rtt var.
936	*/
937	if (tp->t_rtt)
938	tcp_xmit_timer(pData, tp, tp->t_rtt);
939	} else
940	tp->t_state = TCPS_SYN_RECEIVED;
941
942	trimthenstep6:
943	/*
944	* Advance ti->ti_seq to correspond to first data byte.
945	* If data, trim to stay within window,
946	* dropping FIN if necessary.
947	*/
948	ti->ti_seq++;
949	if (ti->ti_len > tp->rcv_wnd) {
950	todrop = ti->ti_len - tp->rcv_wnd;
951	m_adj(m, -todrop);
952	ti->ti_len = tp->rcv_wnd;
953	tiflags &= ~TH_FIN;
954	tcpstat.tcps_rcvpackafterwin++;
955	tcpstat.tcps_rcvbyteafterwin += todrop;
956	}
957	tp->snd_wl1 = ti->ti_seq - 1;
958	tp->rcv_up = ti->ti_seq;
959	goto step6;
960	} /* switch tp->t_state */
961	/*
962	* States other than LISTEN or SYN_SENT.
963	* First check timestamp, if present.
964	* Then check that at least some bytes of segment are within
965	* receive window. If segment begins before rcv_nxt,
966	* drop leading data (and SYN); if nothing left, just ack.
967	*
968	* RFC 1323 PAWS: If we have a timestamp reply on this segment
969	* and it's less than ts_recent, drop it.
970	*/
971	/* if (ts_present && (tiflags & TH_RST) == 0 && tp->ts_recent &&
972	* TSTMP_LT(ts_val, tp->ts_recent)) {
973	*
974	/ / Check to see if ts_recent is over 24 days old. */
975	/* if ((int)(tcp_now - tp->ts_recent_age) > TCP_PAWS_IDLE) {
976	/ /
977	* * Invalidate ts_recent. If this segment updates
978	* * ts_recent, the age will be reset later and ts_recent
979	* * will get a valid value. If it does not, setting
980	* * ts_recent to zero will at least satisfy the
981	* * requirement that zero be placed in the timestamp
982	* * echo reply when ts_recent isn't valid. The
983	* * age isn't reset until we get a valid ts_recent
984	* * because we don't want out-of-order segments to be
985	* * dropped when ts_recent is old.
986	* */
987	/* tp->ts_recent = 0;
988	* } else {
989	* tcpstat.tcps_rcvduppack++;
990	* tcpstat.tcps_rcvdupbyte += ti->ti_len;
991	* tcpstat.tcps_pawsdrop++;
992	* goto dropafterack;
993	* }
994	* }
995	*/
996
997	todrop = tp->rcv_nxt - ti->ti_seq;
998	if (todrop > 0) {
999	if (tiflags & TH_SYN) {
1000	tiflags &= ~TH_SYN;
1001	ti->ti_seq++;
1002	if (ti->ti_urp > 1)
1003	ti->ti_urp--;
1004	else
1005	tiflags &= ~TH_URG;
1006	todrop--;
1007	}
1008	/*
1009	* Following if statement from Stevens, vol. 2, p. 960.
1010	*/
1011	if (todrop > ti->ti_len
1012	\|\| (todrop == ti->ti_len && (tiflags & TH_FIN) == 0)) {
1013	/*
1014	* Any valid FIN must be to the left of the window.
1015	* At this point the FIN must be a duplicate or out
1016	* of sequence; drop it.
1017	*/
1018	tiflags &= ~TH_FIN;
1019
1020	/*
1021	* Send an ACK to resynchronize and drop any data.
1022	* But keep on processing for RST or ACK.
1023	*/
1024	tp->t_flags \|= TF_ACKNOW;
1025	todrop = ti->ti_len;
1026	tcpstat.tcps_rcvduppack++;
1027	tcpstat.tcps_rcvdupbyte += todrop;
1028	} else {
1029	tcpstat.tcps_rcvpartduppack++;
1030	tcpstat.tcps_rcvpartdupbyte += todrop;
1031	}
1032	m_adj(m, todrop);
1033	ti->ti_seq += todrop;
1034	ti->ti_len -= todrop;
1035	if (ti->ti_urp > todrop)
1036	ti->ti_urp -= todrop;
1037	else {
1038	tiflags &= ~TH_URG;
1039	ti->ti_urp = 0;
1040	}
1041	}
1042	/*
1043	* If new data are received on a connection after the
1044	* user processes are gone, then RST the other end.
1045	*/
1046	if ((so->so_state & SS_NOFDREF) &&
1047	tp->t_state > TCPS_CLOSE_WAIT && ti->ti_len) {
1048	tp = tcp_close(pData, tp);
1049	tcpstat.tcps_rcvafterclose++;
1050	goto dropwithreset;
1051	}
1052
1053	/*
1054	* If segment ends after window, drop trailing data
1055	* (and PUSH and FIN); if nothing left, just ACK.
1056	*/
1057	todrop = (ti->ti_seq+ti->ti_len) - (tp->rcv_nxt+tp->rcv_wnd);
1058	if (todrop > 0) {
1059	tcpstat.tcps_rcvpackafterwin++;
1060	if (todrop >= ti->ti_len) {
1061	tcpstat.tcps_rcvbyteafterwin += ti->ti_len;
1062	/*
1063	* If a new connection request is received
1064	* while in TIME_WAIT, drop the old connection
1065	* and start over if the sequence numbers
1066	* are above the previous ones.
1067	*/
1068	if (tiflags & TH_SYN &&
1069	tp->t_state == TCPS_TIME_WAIT &&
1070	SEQ_GT(ti->ti_seq, tp->rcv_nxt)) {
1071	iss = tp->rcv_nxt + TCP_ISSINCR;
1072	tp = tcp_close(pData, tp);
1073	goto findso;
1074	}
1075	/*
1076	* If window is closed can only take segments at
1077	* window edge, and have to drop data and PUSH from
1078	* incoming segments. Continue processing, but
1079	* remember to ack. Otherwise, drop segment
1080	* and ack.
1081	*/
1082	if (tp->rcv_wnd == 0 && ti->ti_seq == tp->rcv_nxt) {
1083	tp->t_flags \|= TF_ACKNOW;
1084	tcpstat.tcps_rcvwinprobe++;
1085	} else
1086	goto dropafterack;
1087	} else
1088	tcpstat.tcps_rcvbyteafterwin += todrop;
1089	m_adj(m, -todrop);
1090	ti->ti_len -= todrop;
1091	tiflags &= ~(TH_PUSH\|TH_FIN);
1092	}
1093
1094	/*
1095	* If last ACK falls within this segment's sequence numbers,
1096	* record its timestamp.
1097	*/
1098	/* if (ts_present && SEQ_LEQ(ti->ti_seq, tp->last_ack_sent) &&
1099	* SEQ_LT(tp->last_ack_sent, ti->ti_seq + ti->ti_len +
1100	* ((tiflags & (TH_SYN\|TH_FIN)) != 0))) {
1101	* tp->ts_recent_age = tcp_now;
1102	* tp->ts_recent = ts_val;
1103	* }
1104	*/
1105
1106	/*
1107	* If the RST bit is set examine the state:
1108	* SYN_RECEIVED STATE:
1109	* If passive open, return to LISTEN state.
1110	* If active open, inform user that connection was refused.
1111	* ESTABLISHED, FIN_WAIT_1, FIN_WAIT2, CLOSE_WAIT STATES:
1112	* Inform user that connection was reset, and close tcb.
1113	* CLOSING, LAST_ACK, TIME_WAIT STATES
1114	* Close the tcb.
1115	*/
1116	if (tiflags&TH_RST) switch (tp->t_state) {
1117
1118	case TCPS_SYN_RECEIVED:
1119	/* so->so_error = ECONNREFUSED; */
1120	goto close;
1121
1122	case TCPS_ESTABLISHED:
1123	case TCPS_FIN_WAIT_1:
1124	case TCPS_FIN_WAIT_2:
1125	case TCPS_CLOSE_WAIT:
1126	/* so->so_error = ECONNRESET; */
1127	close:
1128	tp->t_state = TCPS_CLOSED;
1129	tcpstat.tcps_drops++;
1130	tp = tcp_close(pData, tp);
1131	goto drop;
1132
1133	case TCPS_CLOSING:
1134	case TCPS_LAST_ACK:
1135	case TCPS_TIME_WAIT:
1136	tp = tcp_close(pData, tp);
1137	goto drop;
1138	}
1139
1140	/*
1141	* If a SYN is in the window, then this is an
1142	* error and we send an RST and drop the connection.
1143	*/
1144	if (tiflags & TH_SYN) {
1145	tp = tcp_drop(pData, tp,0);
1146	goto dropwithreset;
1147	}
1148
1149	/*
1150	* If the ACK bit is off we drop the segment and return.
1151	*/
1152	if ((tiflags & TH_ACK) == 0) goto drop;
1153
1154	/*
1155	* Ack processing.
1156	*/
1157	switch (tp->t_state) {
1158	/*
1159	* In SYN_RECEIVED state if the ack ACKs our SYN then enter
1160	* ESTABLISHED state and continue processing, otherwise
1161	* send an RST. una<=ack<=max
1162	*/
1163	case TCPS_SYN_RECEIVED:
1164
1165	if (SEQ_GT(tp->snd_una, ti->ti_ack) \|\|
1166	SEQ_GT(ti->ti_ack, tp->snd_max))
1167	goto dropwithreset;
1168	tcpstat.tcps_connects++;
1169	tp->t_state = TCPS_ESTABLISHED;
1170	/*
1171	* The sent SYN is ack'ed with our sequence number +1
1172	* The first data byte already in the buffer will get
1173	* lost if no correction is made. This is only needed for
1174	* SS_CTL since the buffer is empty otherwise.
1175	* tp->snd_una++; or:
1176	*/
1177	tp->snd_una=ti->ti_ack;
1178	soisfconnected(so);
1179
1180	/* Do window scaling? */
1181	/* if ((tp->t_flags & (TF_RCVD_SCALE\|TF_REQ_SCALE)) ==
1182	* (TF_RCVD_SCALE\|TF_REQ_SCALE)) {
1183	* tp->snd_scale = tp->requested_s_scale;
1184	* tp->rcv_scale = tp->request_r_scale;
1185	* }
1186	*/
1187	#ifndef VBOX_WITH_BSD_TCP_REASS
1188	(void) tcp_reass(pData, tp, (struct tcpiphdr )0, (struct mbuf )0);
1189	#else /* VBOX_WITH_BSD_TCP_REASS */
1190	(void) tcp_reass(pData, tp, (struct tcphdr )0, (int )0, (struct mbuf *)0);
1191	#endif /VBOX_WITH_BSD_TCP_REASS/
1192	tp->snd_wl1 = ti->ti_seq - 1;
1193	/* Avoid ack processing; snd_una==ti_ack => dup ack */
1194	goto synrx_to_est;
1195	/* fall into ... */
1196
1197	/*
1198	* In ESTABLISHED state: drop duplicate ACKs; ACK out of range
1199	* ACKs. If the ack is in the range
1200	* tp->snd_una < ti->ti_ack <= tp->snd_max
1201	* then advance tp->snd_una to ti->ti_ack and drop
1202	* data from the retransmission queue. If this ACK reflects
1203	* more up to date window information we update our window information.
1204	*/
1205	case TCPS_ESTABLISHED:
1206	case TCPS_FIN_WAIT_1:
1207	case TCPS_FIN_WAIT_2:
1208	case TCPS_CLOSE_WAIT:
1209	case TCPS_CLOSING:
1210	case TCPS_LAST_ACK:
1211	case TCPS_TIME_WAIT:
1212
1213	if (SEQ_LEQ(ti->ti_ack, tp->snd_una)) {
1214	if (ti->ti_len == 0 && tiwin == tp->snd_wnd) {
1215	tcpstat.tcps_rcvdupack++;
1216	DEBUG_MISC((dfd," dup ack m = %lx so = %lx \n",
1217	(long )m, (long )so));
1218	/*
1219	* If we have outstanding data (other than
1220	* a window probe), this is a completely
1221	* duplicate ack (ie, window info didn't
1222	* change), the ack is the biggest we've
1223	* seen and we've seen exactly our rexmt
1224	* threshold of them, assume a packet
1225	* has been dropped and retransmit it.
1226	* Kludge snd_nxt & the congestion
1227	* window so we send only this one
1228	* packet.
1229	*
1230	* We know we're losing at the current
1231	* window size so do congestion avoidance
1232	* (set ssthresh to half the current window
1233	* and pull our congestion window back to
1234	* the new ssthresh).
1235	*
1236	* Dup acks mean that packets have left the
1237	* network (they're now cached at the receiver)
1238	* so bump cwnd by the amount in the receiver
1239	* to keep a constant cwnd packets in the
1240	* network.
1241	*/
1242	if (tp->t_timer[TCPT_REXMT] == 0 \|\|
1243	ti->ti_ack != tp->snd_una)
1244	tp->t_dupacks = 0;
1245	else if (++tp->t_dupacks == tcprexmtthresh) {
1246	tcp_seq onxt = tp->snd_nxt;
1247	u_int win =
1248	min(tp->snd_wnd, tp->snd_cwnd) / 2 /
1249	tp->t_maxseg;
1250
1251	if (win < 2)
1252	win = 2;
1253	tp->snd_ssthresh = win * tp->t_maxseg;
1254	tp->t_timer[TCPT_REXMT] = 0;
1255	tp->t_rtt = 0;
1256	tp->snd_nxt = ti->ti_ack;
1257	tp->snd_cwnd = tp->t_maxseg;
1258	(void) tcp_output(pData, tp);
1259	tp->snd_cwnd = tp->snd_ssthresh +
1260	tp->t_maxseg * tp->t_dupacks;
1261	if (SEQ_GT(onxt, tp->snd_nxt))
1262	tp->snd_nxt = onxt;
1263	goto drop;
1264	} else if (tp->t_dupacks > tcprexmtthresh) {
1265	tp->snd_cwnd += tp->t_maxseg;
1266	(void) tcp_output(pData, tp);
1267	goto drop;
1268	}
1269	} else
1270	tp->t_dupacks = 0;
1271	break;
1272	}
1273	synrx_to_est:
1274	/*
1275	* If the congestion window was inflated to account
1276	* for the other side's cached packets, retract it.
1277	*/
1278	if (tp->t_dupacks > tcprexmtthresh &&
1279	tp->snd_cwnd > tp->snd_ssthresh)
1280	tp->snd_cwnd = tp->snd_ssthresh;
1281	tp->t_dupacks = 0;
1282	if (SEQ_GT(ti->ti_ack, tp->snd_max)) {
1283	tcpstat.tcps_rcvacktoomuch++;
1284	goto dropafterack;
1285	}
1286	acked = ti->ti_ack - tp->snd_una;
1287	tcpstat.tcps_rcvackpack++;
1288	tcpstat.tcps_rcvackbyte += acked;
1289
1290	/*
1291	* If we have a timestamp reply, update smoothed
1292	* round trip time. If no timestamp is present but
1293	* transmit timer is running and timed sequence
1294	* number was acked, update smoothed round trip time.
1295	* Since we now have an rtt measurement, cancel the
1296	* timer backoff (cf., Phil Karn's retransmit alg.).
1297	* Recompute the initial retransmit timer.
1298	*/
1299	/* if (ts_present)
1300	* tcp_xmit_timer(tp, tcp_now-ts_ecr+1);
1301	* else
1302	*/
1303	if (tp->t_rtt && SEQ_GT(ti->ti_ack, tp->t_rtseq))
1304	tcp_xmit_timer(pData, tp,tp->t_rtt);
1305
1306	/*
1307	* If all outstanding data is acked, stop retransmit
1308	* timer and remember to restart (more output or persist).
1309	* If there is more data to be acked, restart retransmit
1310	* timer, using current (possibly backed-off) value.
1311	*/
1312	if (ti->ti_ack == tp->snd_max) {
1313	tp->t_timer[TCPT_REXMT] = 0;
1314	needoutput = 1;
1315	} else if (tp->t_timer[TCPT_PERSIST] == 0)
1316	tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
1317	/*
1318	* When new data is acked, open the congestion window.
1319	* If the window gives us less than ssthresh packets
1320	* in flight, open exponentially (maxseg per packet).
1321	* Otherwise open linearly: maxseg per window
1322	* (maxseg^2 / cwnd per packet).
1323	*/
1324	{
1325	register u_int cw = tp->snd_cwnd;
1326	register u_int incr = tp->t_maxseg;
1327
1328	if (cw > tp->snd_ssthresh)
1329	incr = incr * incr / cw;
1330	tp->snd_cwnd = min(cw + incr, TCP_MAXWIN<<tp->snd_scale);
1331	}
1332	if (acked > so->so_snd.sb_cc) {
1333	tp->snd_wnd -= so->so_snd.sb_cc;
1334	sbdrop(&so->so_snd, (int )so->so_snd.sb_cc);
1335	ourfinisacked = 1;
1336	} else {
1337	sbdrop(&so->so_snd, acked);
1338	tp->snd_wnd -= acked;
1339	ourfinisacked = 0;
1340	}
1341	/*
1342	* XXX sowwakup is called when data is acked and there's room for
1343	* for more data... it should read() the socket
1344	*/
1345	/* if (so->so_snd.sb_flags & SB_NOTIFY)
1346	* sowwakeup(so);
1347	*/
1348	tp->snd_una = ti->ti_ack;
1349	if (SEQ_LT(tp->snd_nxt, tp->snd_una))
1350	tp->snd_nxt = tp->snd_una;
1351
1352	switch (tp->t_state) {
1353
1354	/*
1355	* In FIN_WAIT_1 STATE in addition to the processing
1356	* for the ESTABLISHED state if our FIN is now acknowledged
1357	* then enter FIN_WAIT_2.
1358	*/
1359	case TCPS_FIN_WAIT_1:
1360	if (ourfinisacked) {
1361	/*
1362	* If we can't receive any more
1363	* data, then closing user can proceed.
1364	* Starting the timer is contrary to the
1365	* specification, but if we don't get a FIN
1366	* we'll hang forever.
1367	*/
1368	if (so->so_state & SS_FCANTRCVMORE) {
1369	soisfdisconnected(so);
1370	tp->t_timer[TCPT_2MSL] = tcp_maxidle;
1371	}
1372	tp->t_state = TCPS_FIN_WAIT_2;
1373	}
1374	break;
1375
1376	/*
1377	* In CLOSING STATE in addition to the processing for
1378	* the ESTABLISHED state if the ACK acknowledges our FIN
1379	* then enter the TIME-WAIT state, otherwise ignore
1380	* the segment.
1381	*/
1382	case TCPS_CLOSING:
1383	if (ourfinisacked) {
1384	tp->t_state = TCPS_TIME_WAIT;
1385	tcp_canceltimers(tp);
1386	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1387	soisfdisconnected(so);
1388	}
1389	break;
1390
1391	/*
1392	* In LAST_ACK, we may still be waiting for data to drain
1393	* and/or to be acked, as well as for the ack of our FIN.
1394	* If our FIN is now acknowledged, delete the TCB,
1395	* enter the closed state and return.
1396	*/
1397	case TCPS_LAST_ACK:
1398	if (ourfinisacked) {
1399	tp = tcp_close(pData, tp);
1400	goto drop;
1401	}
1402	break;
1403
1404	/*
1405	* In TIME_WAIT state the only thing that should arrive
1406	* is a retransmission of the remote FIN. Acknowledge
1407	* it and restart the finack timer.
1408	*/
1409	case TCPS_TIME_WAIT:
1410	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1411	goto dropafterack;
1412	}
1413	} /* switch(tp->t_state) */
1414
1415	step6:
1416	/*
1417	* Update window information.
1418	* Don't look at window if no ACK: TAC's send garbage on first SYN.
1419	*/
1420	if ((tiflags & TH_ACK) &&
1421	(SEQ_LT(tp->snd_wl1, ti->ti_seq) \|\|
1422	(tp->snd_wl1 == ti->ti_seq && (SEQ_LT(tp->snd_wl2, ti->ti_ack) \|\|
1423	(tp->snd_wl2 == ti->ti_ack && tiwin > tp->snd_wnd))))) {
1424	/* keep track of pure window updates */
1425	if (ti->ti_len == 0 &&
1426	tp->snd_wl2 == ti->ti_ack && tiwin > tp->snd_wnd)
1427	tcpstat.tcps_rcvwinupd++;
1428	tp->snd_wnd = tiwin;
1429	tp->snd_wl1 = ti->ti_seq;
1430	tp->snd_wl2 = ti->ti_ack;
1431	if (tp->snd_wnd > tp->max_sndwnd)
1432	tp->max_sndwnd = tp->snd_wnd;
1433	needoutput = 1;
1434	}
1435
1436	/*
1437	* Process segments with URG.
1438	*/
1439	if ((tiflags & TH_URG) && ti->ti_urp &&
1440	TCPS_HAVERCVDFIN(tp->t_state) == 0) {
1441	/*
1442	* This is a kludge, but if we receive and accept
1443	* random urgent pointers, we'll crash in
1444	* soreceive. It's hard to imagine someone
1445	* actually wanting to send this much urgent data.
1446	*/
1447	if (ti->ti_urp + so->so_rcv.sb_cc > so->so_rcv.sb_datalen) {
1448	ti->ti_urp = 0;
1449	tiflags &= ~TH_URG;
1450	goto dodata;
1451	}
1452	/*
1453	* If this segment advances the known urgent pointer,
1454	* then mark the data stream. This should not happen
1455	* in CLOSE_WAIT, CLOSING, LAST_ACK or TIME_WAIT STATES since
1456	* a FIN has been received from the remote side.
1457	* In these states we ignore the URG.
1458	*
1459	* According to RFC961 (Assigned Protocols),
1460	* the urgent pointer points to the last octet
1461	* of urgent data. We continue, however,
1462	* to consider it to indicate the first octet
1463	* of data past the urgent section as the original
1464	* spec states (in one of two places).
1465	*/
1466	if (SEQ_GT(ti->ti_seq+ti->ti_urp, tp->rcv_up)) {
1467	tp->rcv_up = ti->ti_seq + ti->ti_urp;
1468	so->so_urgc = so->so_rcv.sb_cc +
1469	(tp->rcv_up - tp->rcv_nxt); /* -1; */
1470	tp->rcv_up = ti->ti_seq + ti->ti_urp;
1471
1472	}
1473	} else
1474	/*
1475	* If no out of band data is expected,
1476	* pull receive urgent pointer along
1477	* with the receive window.
1478	*/
1479	if (SEQ_GT(tp->rcv_nxt, tp->rcv_up))
1480	tp->rcv_up = tp->rcv_nxt;
1481	dodata:
1482
1483	/*
1484	* Process the segment text, merging it into the TCP sequencing queue,
1485	* and arranging for acknowledgment of receipt if necessary.
1486	* This process logically involves adjusting tp->rcv_wnd as data
1487	* is presented to the user (this happens in tcp_usrreq.c,
1488	* case PRU_RCVD). If a FIN has already been received on this
1489	* connection then we just ignore the text.
1490	*/
1491	if ((ti->ti_len \|\| (tiflags&TH_FIN)) &&
1492	TCPS_HAVERCVDFIN(tp->t_state) == 0) {
1493	#ifndef VBOX_WITH_BSD_TCP_REASS
1494	TCP_REASS(pData, tp, ti, m, so, tiflags);
1495	#else /* VBOX_WITH_BSD_TCP_REASS */
1496	if (ti->ti_seq == tp->rcv_nxt
1497	&& LIST_EMPTY(&tp->t_segq)
1498	&& tp->t_state == TCPS_ESTABLISHED) {
1499	DELAY_ACK(tp, ti); /* little bit different from BSD declaration see netinet/tcp_input.c */
1500	tp->rcv_nxt += tlen;
1501	tiflags = ti->ti_t.th_flags & TH_FIN;
1502	tcpstat.tcps_rcvpack++;
1503	tcpstat.tcps_rcvbyte += tlen;
1504	if (so->so_state & SS_FCANTRCVMORE)
1505	m_freem(pData, m);
1506	else
1507	sbappend(pData, so, m);
1508	}
1509	else {
1510	tiflags = tcp_reass(pData, tp, &ti->ti_t, &tlen, m);
1511	tiflags \|= TF_ACKNOW;
1512	}
1513	#endif /* VBOX_WITH_BSD_TCP_REASS */
1514	/*
1515	* Note the amount of data that peer has sent into
1516	* our window, in order to estimate the sender's
1517	* buffer size.
1518	*/
1519	len = so->so_rcv.sb_datalen - (tp->rcv_adv - tp->rcv_nxt);
1520	} else {
1521	mbuf_freed = 1; /* The mbuf must be freed, but only when its content is not needed anymore. */
1522	tiflags &= ~TH_FIN;
1523	}
1524
1525	/*
1526	* If FIN is received ACK the FIN and let the user know
1527	* that the connection is closing.
1528	*/
1529	if (tiflags & TH_FIN) {
1530	if (TCPS_HAVERCVDFIN(tp->t_state) == 0) {
1531	/*
1532	* If we receive a FIN we can't send more data,
1533	* set it SS_FDRAIN
1534	* Shutdown the socket if there is no rx data in the
1535	* buffer.
1536	* soread() is called on completion of shutdown() and
1537	* will got to TCPS_LAST_ACK, and use tcp_output()
1538	* to send the FIN.
1539	*/
1540	/* sofcantrcvmore(so); */
1541	sofwdrain(so);
1542
1543	tp->t_flags \|= TF_ACKNOW;
1544	tp->rcv_nxt++;
1545	}
1546	switch (tp->t_state) {
1547
1548	/*
1549	* In SYN_RECEIVED and ESTABLISHED STATES
1550	* enter the CLOSE_WAIT state.
1551	*/
1552	case TCPS_SYN_RECEIVED:
1553	case TCPS_ESTABLISHED:
1554	if(so->so_emu == EMU_CTL) /* no shutdown on socket */
1555	tp->t_state = TCPS_LAST_ACK;
1556	else
1557	tp->t_state = TCPS_CLOSE_WAIT;
1558	break;
1559
1560	/*
1561	* If still in FIN_WAIT_1 STATE FIN has not been acked so
1562	* enter the CLOSING state.
1563	*/
1564	case TCPS_FIN_WAIT_1:
1565	tp->t_state = TCPS_CLOSING;
1566	break;
1567
1568	/*
1569	* In FIN_WAIT_2 state enter the TIME_WAIT state,
1570	* starting the time-wait timer, turning off the other
1571	* standard timers.
1572	*/
1573	case TCPS_FIN_WAIT_2:
1574	tp->t_state = TCPS_TIME_WAIT;
1575	tcp_canceltimers(tp);
1576	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1577	soisfdisconnected(so);
1578	break;
1579
1580	/*
1581	* In TIME_WAIT state restart the 2 MSL time_wait timer.
1582	*/
1583	case TCPS_TIME_WAIT:
1584	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1585	break;
1586	}
1587	}
1588
1589	/*
1590	* If this is a small packet, then ACK now - with Nagel
1591	* congestion avoidance sender won't send more until
1592	* he gets an ACK.
1593	*
1594	* See above.
1595	*/
1596	/* if (ti->ti_len && (unsigned)ti->ti_len < tp->t_maxseg) {
1597	*/
1598	/* if ((ti->ti_len && (unsigned)ti->ti_len < tp->t_maxseg &&
1599	* (so->so_iptos & IPTOS_LOWDELAY) == 0) \|\|
1600	* ((so->so_iptos & IPTOS_LOWDELAY) &&
1601	* ((struct tcpiphdr_2 *)ti)->first_char == (char)27)) {
1602	*/
1603	if (ti->ti_len && (unsigned)ti->ti_len <= 5 &&
1604	((struct tcpiphdr_2 *)ti)->first_char == (char)27) {
1605	tp->t_flags \|= TF_ACKNOW;
1606	}
1607
1608	if (mbuf_freed) {
1609	m_free(pData, m);
1610	}
1611	/*
1612	* Return any desired output.
1613	*/
1614	if (needoutput \|\| (tp->t_flags & TF_ACKNOW)) {
1615	(void) tcp_output(pData, tp);
1616	}
1617	return;
1618
1619	dropafterack:
1620	/*
1621	* Generate an ACK dropping incoming segment if it occupies
1622	* sequence space, where the ACK reflects our state.
1623	*/
1624	if (tiflags & TH_RST)
1625	goto drop;
1626	m_freem(pData, m);
1627	tp->t_flags \|= TF_ACKNOW;
1628	(void) tcp_output(pData, tp);
1629	return;
1630
1631	dropwithreset:
1632	/* reuses m if m!=NULL, m_free() unnecessary */
1633	if (tiflags & TH_ACK)
1634	tcp_respond(pData, tp, ti, m, (tcp_seq)0, ti->ti_ack, TH_RST);
1635	else {
1636	if (tiflags & TH_SYN) ti->ti_len++;
1637	tcp_respond(pData, tp, ti, m, ti->ti_seq+ti->ti_len, (tcp_seq)0,
1638	TH_RST\|TH_ACK);
1639	}
1640
1641	return;
1642
1643	drop:
1644	/*
1645	* Drop space held by incoming segment and return.
1646	*/
1647	m_free(pData, m);
1648
1649	return;
1650	}
1651
1652	/* , ts_present, ts_val, ts_ecr) */
1653	/* int *ts_present;
1654	* u_int32_t ts_val, ts_ecr;
1655	*/
1656	void
1657	tcp_dooptions(PNATState pData, struct tcpcb tp, u_char cp, int cnt, struct tcpiphdr *ti)
1658	{
1659	u_int16_t mss;
1660	int opt, optlen;
1661
1662	DEBUG_CALL("tcp_dooptions");
1663	DEBUG_ARGS((dfd," tp = %lx cnt=%i \n", (long )tp, cnt));
1664
1665	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1666	opt = cp[0];
1667	if (opt == TCPOPT_EOL)
1668	break;
1669	if (opt == TCPOPT_NOP)
1670	optlen = 1;
1671	else {
1672	optlen = cp[1];
1673	if (optlen <= 0)
1674	break;
1675	}
1676	switch (opt) {
1677
1678	default:
1679	continue;
1680
1681	case TCPOPT_MAXSEG:
1682	if (optlen != TCPOLEN_MAXSEG)
1683	continue;
1684	if (!(ti->ti_flags & TH_SYN))
1685	continue;
1686	memcpy((char ) &mss, (char ) cp + 2, sizeof(mss));
1687	NTOHS(mss);
1688	(void) tcp_mss(pData, tp, mss); /* sets t_maxseg */
1689	break;
1690
1691	/* case TCPOPT_WINDOW:
1692	* if (optlen != TCPOLEN_WINDOW)
1693	* continue;
1694	* if (!(ti->ti_flags & TH_SYN))
1695	* continue;
1696	* tp->t_flags \|= TF_RCVD_SCALE;
1697	* tp->requested_s_scale = min(cp[2], TCP_MAX_WINSHIFT);
1698	* break;
1699	*/
1700	/* case TCPOPT_TIMESTAMP:
1701	* if (optlen != TCPOLEN_TIMESTAMP)
1702	* continue;
1703	* *ts_present = 1;
1704	* memcpy((char ) ts_val, (char )cp + 2, sizeof(*ts_val));
1705	* NTOHL(*ts_val);
1706	* memcpy((char ) ts_ecr, (char )cp + 6, sizeof(*ts_ecr));
1707	* NTOHL(*ts_ecr);
1708	*
1709	/ /
1710	* * A timestamp received in a SYN makes
1711	* * it ok to send timestamp requests and replies.
1712	* */
1713	/* if (ti->ti_flags & TH_SYN) {
1714	* tp->t_flags \|= TF_RCVD_TSTMP;
1715	* tp->ts_recent = *ts_val;
1716	* tp->ts_recent_age = tcp_now;
1717	* }
1718	*/ break;
1719	}
1720	}
1721	}
1722
1723
1724	/*
1725	* Pull out of band byte out of a segment so
1726	* it doesn't appear in the user's data queue.
1727	* It is still reflected in the segment length for
1728	* sequencing purposes.
1729	*/
1730
1731	#ifdef notdef
1732
1733	void
1734	tcp_pulloutofband(so, ti, m)
1735	struct socket *so;
1736	struct tcpiphdr *ti;
1737	register struct mbuf *m;
1738	{
1739	int cnt = ti->ti_urp - 1;
1740
1741	while (cnt >= 0) {
1742	if (m->m_len > cnt) {
1743	char *cp = mtod(m, caddr_t) + cnt;
1744	struct tcpcb *tp = sototcpcb(so);
1745
1746	tp->t_iobc = *cp;
1747	tp->t_oobflags \|= TCPOOB_HAVEDATA;
1748	memcpy(sp, cp+1, (unsigned)(m->m_len - cnt - 1));
1749	m->m_len--;
1750	return;
1751	}
1752	cnt -= m->m_len;
1753	m = m->m_next; /* XXX WRONG! Fix it! */
1754	if (m == 0)
1755	break;
1756	}
1757	panic("tcp_pulloutofband");
1758	}
1759
1760	#endif /* notdef */
1761
1762	/*
1763	* Collect new round-trip time estimate
1764	* and update averages and current timeout.
1765	*/
1766
1767	void
1768	tcp_xmit_timer(PNATState pData, register struct tcpcb *tp, int rtt)
1769	{
1770	register short delta;
1771
1772	DEBUG_CALL("tcp_xmit_timer");
1773	DEBUG_ARG("tp = %lx", (long)tp);
1774	DEBUG_ARG("rtt = %d", rtt);
1775
1776	tcpstat.tcps_rttupdated++;
1777	if (tp->t_srtt != 0) {
1778	/*
1779	* srtt is stored as fixed point with 3 bits after the
1780	* binary point (i.e., scaled by 8). The following magic
1781	* is equivalent to the smoothing algorithm in rfc793 with
1782	* an alpha of .875 (srtt = rtt/8 + srtt*7/8 in fixed
1783	* point). Adjust rtt to origin 0.
1784	*/
1785	delta = rtt - 1 - (tp->t_srtt >> TCP_RTT_SHIFT);
1786	if ((tp->t_srtt += delta) <= 0)
1787	tp->t_srtt = 1;
1788	/*
1789	* We accumulate a smoothed rtt variance (actually, a
1790	* smoothed mean difference), then set the retransmit
1791	* timer to smoothed rtt + 4 times the smoothed variance.
1792	* rttvar is stored as fixed point with 2 bits after the
1793	* binary point (scaled by 4). The following is
1794	* equivalent to rfc793 smoothing with an alpha of .75
1795	* (rttvar = rttvar*3/4 + \|delta\| / 4). This replaces
1796	* rfc793's wired-in beta.
1797	*/
1798	if (delta < 0)
1799	delta = -delta;
1800	delta -= (tp->t_rttvar >> TCP_RTTVAR_SHIFT);
1801	if ((tp->t_rttvar += delta) <= 0)
1802	tp->t_rttvar = 1;
1803	} else {
1804	/*
1805	* No rtt measurement yet - use the unsmoothed rtt.
1806	* Set the variance to half the rtt (so our first
1807	* retransmit happens at 3*rtt).
1808	*/
1809	tp->t_srtt = rtt << TCP_RTT_SHIFT;
1810	tp->t_rttvar = rtt << (TCP_RTTVAR_SHIFT - 1);
1811	}
1812	tp->t_rtt = 0;
1813	tp->t_rxtshift = 0;
1814
1815	/*
1816	* the retransmit should happen at rtt + 4 * rttvar.
1817	* Because of the way we do the smoothing, srtt and rttvar
1818	* will each average +1/2 tick of bias. When we compute
1819	* the retransmit timer, we want 1/2 tick of rounding and
1820	* 1 extra tick because of +-1/2 tick uncertainty in the
1821	* firing of the timer. The bias will give us exactly the
1822	* 1.5 tick we need. But, because the bias is
1823	* statistical, we have to test that we don't drop below
1824	* the minimum feasible timer (which is 2 ticks).
1825	*/
1826	TCPT_RANGESET(tp->t_rxtcur, TCP_REXMTVAL(tp),
1827	(short)tp->t_rttmin, TCPTV_REXMTMAX); /* XXX */
1828
1829	/*
1830	* We received an ack for a packet that wasn't retransmitted;
1831	* it is probably safe to discard any error indications we've
1832	* received recently. This isn't quite right, but close enough
1833	* for now (a route might have failed after we sent a segment,
1834	* and the return path might not be symmetrical).
1835	*/
1836	tp->t_softerror = 0;
1837	}
1838
1839	/*
1840	* Determine a reasonable value for maxseg size.
1841	* If the route is known, check route for mtu.
1842	* If none, use an mss that can be handled on the outgoing
1843	* interface without forcing IP to fragment; if bigger than
1844	* an mbuf cluster (MCLBYTES), round down to nearest multiple of MCLBYTES
1845	* to utilize large mbufs. If no route is found, route has no mtu,
1846	* or the destination isn't local, use a default, hopefully conservative
1847	* size (usually 512 or the default IP max size, but no more than the mtu
1848	* of the interface), as we can't discover anything about intervening
1849	* gateways or networks. We also initialize the congestion/slow start
1850	* window to be a single segment if the destination isn't local.
1851	* While looking at the routing entry, we also initialize other path-dependent
1852	* parameters from pre-set or cached values in the routing entry.
1853	*/
1854
1855	int
1856	tcp_mss(PNATState pData, register struct tcpcb *tp, u_int offer)
1857	{
1858	struct socket *so = tp->t_socket;
1859	int mss;
1860
1861	DEBUG_CALL("tcp_mss");
1862	DEBUG_ARG("tp = %lx", (long)tp);
1863	DEBUG_ARG("offer = %d", offer);
1864
1865	mss = min(if_mtu, if_mru) - sizeof(struct tcpiphdr);
1866	if (offer)
1867	mss = min(mss, offer);
1868	mss = max(mss, 32);
1869	if (mss < tp->t_maxseg \|\| offer != 0)
1870	tp->t_maxseg = mss;
1871
1872	tp->snd_cwnd = mss;
1873
1874	sbreserve(&so->so_snd, tcp_sndspace+((tcp_sndspace%mss)?(mss-(tcp_sndspace%mss)):0));
1875	sbreserve(&so->so_rcv, tcp_rcvspace+((tcp_rcvspace%mss)?(mss-(tcp_rcvspace%mss)):0));
1876
1877	DEBUG_MISC((dfd, " returning mss = %d\n", mss));
1878
1879	return mss;
1880	}

注意: 瀏覽 TracBrowser 來幫助您使用儲存庫瀏覽器

source: vbox/trunk/src/VBox/Devices/Network/slirp/tcp_input.c@ 14470

以其他格式下載: