tcp_input.c@ 1048

Last change on this file since 1048 was 1048, checked in by vboxsync, 18 years ago
slirp insque/remque fixes for amd64
Property svn:eol-style set to `native`
File size: 54.0 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	#ifndef VBOX
49	struct socket tcb;
50
51	int tcprexmtthresh = 3;
52	struct socket *tcp_last_so = &tcb;
53
54	tcp_seq tcp_iss; /* tcp initial send seq # */
55	#endif /* !VBOX */
56
57	#define TCP_PAWS_IDLE (24 * 24 * 60 * 60 * PR_SLOWHZ)
58
59	/* for modulo comparisons of timestamps */
60	#define TSTMP_LT(a,b) ((int)((a)-(b)) < 0)
61	#define TSTMP_GEQ(a,b) ((int)((a)-(b)) >= 0)
62
63	/*
64	* Insert segment ti into reassembly queue of tcp with
65	* control block tp. Return TH_FIN if reassembly now includes
66	* a segment with FIN. The macro form does the common case inline
67	* (segment is the next to be received on an established connection,
68	* and the queue is empty), avoiding linkage into and removal
69	* from the queue and repetition of various conversions.
70	* Set DELACK for segments received in order, but ack immediately
71	* when segments are out of order (so fast retransmit can work).
72	*/
73	#ifdef VBOX
74	#ifdef TCP_ACK_HACK
75	#define TCP_REASS(pData, tp, ti, m, so, flags) {\
76	if ((ti)->ti_seq == (tp)->rcv_nxt && \
77	u32_to_ptr((pData), (tp)->seg_next, struct tcpcb *) == (tp) && \
78	(tp)->t_state == TCPS_ESTABLISHED) {\
79	if (ti->ti_flags & TH_PUSH) \
80	tp->t_flags \|= TF_ACKNOW; \
81	else \
82	tp->t_flags \|= TF_DELACK; \
83	(tp)->rcv_nxt += (ti)->ti_len; \
84	flags = (ti)->ti_flags & TH_FIN; \
85	tcpstat.tcps_rcvpack++;\
86	tcpstat.tcps_rcvbyte += (ti)->ti_len;\
87	if (so->so_emu) { \
88	if (tcp_emu((pData), (so),(m))) sbappend((pData), (so), (m)); \
89	} else \
90	sbappend((pData), (so), (m)); \
91	/* sorwakeup(so); */ \
92	} else {\
93	(flags) = tcp_reass((pData), (tp), (ti), (m)); \
94	tp->t_flags \|= TF_ACKNOW; \
95	} \
96	}
97	#else
98	#define TCP_REASS(pData, tp, ti, m, so, flags) { \
99	if ((ti)->ti_seq == (tp)->rcv_nxt && \
100	u32_to_ptr((pData), (tp)->seg_next, struct tcpcb *) == (tp) && \
101	(tp)->t_state == TCPS_ESTABLISHED) { \
102	tp->t_flags \|= TF_DELACK; \
103	(tp)->rcv_nxt += (ti)->ti_len; \
104	flags = (ti)->ti_flags & TH_FIN; \
105	tcpstat.tcps_rcvpack++;\
106	tcpstat.tcps_rcvbyte += (ti)->ti_len;\
107	if (so->so_emu) { \
108	if (tcp_emu((pData), (so),(m))) sbappend((pData), (so), (m)); \
109	} else \
110	sbappend((pData), (so), (m)); \
111	/* sorwakeup(so); */ \
112	} else { \
113	(flags) = tcp_reass((pData), (tp), (ti), (m)); \
114	tp->t_flags \|= TF_ACKNOW; \
115	} \
116	}
117	#endif
118	#else /* !VBOX */
119	#ifdef TCP_ACK_HACK
120	#define TCP_REASS(tp, ti, m, so, flags) {\
121	if ((ti)->ti_seq == (tp)->rcv_nxt && \
122	u32_to_ptr((tp)->seg_next, struct tcpcb *) == (tp) && \
123	(tp)->t_state == TCPS_ESTABLISHED) {\
124	if (ti->ti_flags & TH_PUSH) \
125	tp->t_flags \|= TF_ACKNOW; \
126	else \
127	tp->t_flags \|= TF_DELACK; \
128	(tp)->rcv_nxt += (ti)->ti_len; \
129	flags = (ti)->ti_flags & TH_FIN; \
130	tcpstat.tcps_rcvpack++;\
131	tcpstat.tcps_rcvbyte += (ti)->ti_len;\
132	if (so->so_emu) { \
133	if (tcp_emu((so),(m))) sbappend((so), (m)); \
134	} else \
135	sbappend((so), (m)); \
136	/* sorwakeup(so); */ \
137	} else {\
138	(flags) = tcp_reass((tp), (ti), (m)); \
139	tp->t_flags \|= TF_ACKNOW; \
140	} \
141	}
142	#else
143	#define TCP_REASS(tp, ti, m, so, flags) { \
144	if ((ti)->ti_seq == (tp)->rcv_nxt && \
145	u32_to_ptr((tp)->seg_next, struct tcpcb *) == (tp) && \
146	(tp)->t_state == TCPS_ESTABLISHED) { \
147	tp->t_flags \|= TF_DELACK; \
148	(tp)->rcv_nxt += (ti)->ti_len; \
149	flags = (ti)->ti_flags & TH_FIN; \
150	tcpstat.tcps_rcvpack++;\
151	tcpstat.tcps_rcvbyte += (ti)->ti_len;\
152	if (so->so_emu) { \
153	if (tcp_emu((so),(m))) sbappend(so, (m)); \
154	} else \
155	sbappend((so), (m)); \
156	/* sorwakeup(so); */ \
157	} else { \
158	(flags) = tcp_reass((tp), (ti), (m)); \
159	tp->t_flags \|= TF_ACKNOW; \
160	} \
161	}
162	#endif
163	#endif /* !VBOX */
164
165	int
166	#ifdef VBOX
167	tcp_reass(PNATState pData, register struct tcpcb tp, register struct tcpiphdr ti, struct mbuf *m)
168	#else /* !VBOX */
169	tcp_reass(tp, ti, m)
170	register struct tcpcb *tp;
171	register struct tcpiphdr *ti;
172	struct mbuf *m;
173	#endif /* !VBOX */
174	{
175	register struct tcpiphdr *q;
176	struct socket *so = tp->t_socket;
177	int flags;
178
179	/*
180	* Call with ti==0 after become established to
181	* force pre-ESTABLISHED data up to user socket.
182	*/
183	if (ti == 0)
184	goto present;
185
186	/*
187	* Find a segment which begins after this one does.
188	*/
189	for (q = u32_to_ptr(pData, tp->seg_next, struct tcpiphdr ); q != (struct tcpiphdr )tp;
190	q = u32_to_ptr(pData, q->ti_next, struct tcpiphdr *))
191	if (SEQ_GT(q->ti_seq, ti->ti_seq))
192	break;
193
194	/*
195	* If there is a preceding segment, it may provide some of
196	* our data already. If so, drop the data from the incoming
197	* segment. If it provides all of our data, drop us.
198	*/
199	if (u32_to_ptr(pData, q->ti_prev, struct tcpiphdr ) != (struct tcpiphdr )tp) {
200	register int i;
201	q = u32_to_ptr(pData, q->ti_prev, struct tcpiphdr *);
202	/* conversion to int (in i) handles seq wraparound */
203	i = q->ti_seq + q->ti_len - ti->ti_seq;
204	if (i > 0) {
205	if (i >= ti->ti_len) {
206	tcpstat.tcps_rcvduppack++;
207	tcpstat.tcps_rcvdupbyte += ti->ti_len;
208	#ifdef VBOX
209	m_freem(pData, m);
210	#else /* !VBOX */
211	m_freem(m);
212	#endif /* !VBOX */
213	/*
214	* Try to present any queued data
215	* at the left window edge to the user.
216	* This is needed after the 3-WHS
217	* completes.
218	*/
219	goto present; /* ??? */
220	}
221	m_adj(m, i);
222	ti->ti_len -= i;
223	ti->ti_seq += i;
224	}
225	q = u32_to_ptr(pData, q->ti_next, struct tcpiphdr *);
226	}
227	tcpstat.tcps_rcvoopack++;
228	tcpstat.tcps_rcvoobyte += ti->ti_len;
229	REASS_MBUF_SET(ti, m); /* XXX */
230
231	/*
232	* While we overlap succeeding segments trim them or,
233	* if they are completely covered, dequeue them.
234	*/
235	while (q != (struct tcpiphdr *)tp) {
236	register int i = (ti->ti_seq + ti->ti_len) - q->ti_seq;
237	if (i <= 0)
238	break;
239	if (i < q->ti_len) {
240	q->ti_seq += i;
241	q->ti_len -= i;
242	m_adj(REASS_MBUF_GET(q), i);
243	break;
244	}
245	q = u32_to_ptr(pData, q->ti_next, struct tcpiphdr *);
246	m = REASS_MBUF_GET(u32_to_ptr(pData, q->ti_prev, struct tcpiphdr *));
247	remque_32(pData, u32_to_ptr(pData, q->ti_prev, struct tcpiphdr *));
248	#ifdef VBOX
249	m_freem(pData, m);
250	#else /* !VBOX */
251	m_freem(m);
252	#endif /* !VBOX */
253	}
254
255	/*
256	* Stick new segment in its place.
257	*/
258	insque_32(pData, ti, u32_to_ptr(pData, q->ti_prev, struct tcpiphdr *));
259
260	present:
261	/*
262	* Present data to user, advancing rcv_nxt through
263	* completed sequence space.
264	*/
265	if (!TCPS_HAVEESTABLISHED(tp->t_state))
266	return (0);
267	ti = u32_to_ptr(pData, tp->seg_next, struct tcpiphdr *);
268	if (ti == (struct tcpiphdr *)tp \|\| ti->ti_seq != tp->rcv_nxt)
269	return (0);
270	if (tp->t_state == TCPS_SYN_RECEIVED && ti->ti_len)
271	return (0);
272	do {
273	tp->rcv_nxt += ti->ti_len;
274	flags = ti->ti_flags & TH_FIN;
275	remque_32(pData, ti);
276	m = REASS_MBUF_GET(ti); /* XXX */
277	ti = u32_to_ptr(pData, ti->ti_next, struct tcpiphdr *);
278	/* if (so->so_state & SS_FCANTRCVMORE) */
279	if (so->so_state & SS_FCANTSENDMORE)
280	#ifdef VBOX
281	m_freem(pData, m);
282	#else /* !VBOX */
283	m_freem(m);
284	#endif /* !VBOX */
285	else {
286	if (so->so_emu) {
287	#ifdef VBOX
288	if (tcp_emu(pData, so,m)) sbappend(pData, so, m);
289	#else /* !VBOX */
290	if (tcp_emu(so,m)) sbappend(so, m);
291	#endif /* !VBOX */
292	} else
293	#ifdef VBOX
294	sbappend(pData, so, m);
295	#else /* !VBOX */
296	sbappend(so, m);
297	#endif /* !VBOX */
298	}
299	} while (ti != (struct tcpiphdr *)tp && ti->ti_seq == tp->rcv_nxt);
300	/* sorwakeup(so); */
301	return (flags);
302	}
303
304	/*
305	* TCP input routine, follows pages 65-76 of the
306	* protocol specification dated September, 1981 very closely.
307	*/
308	void
309	#ifdef VBOX
310	tcp_input(PNATState pData, register struct mbuf m, int iphlen, struct socket inso)
311	#else /* !VBOX */
312	tcp_input(m, iphlen, inso)
313	register struct mbuf *m;
314	int iphlen;
315	struct socket *inso;
316	#endif /* !VBOX */
317	{
318	struct ip save_ip, *ip;
319	register struct tcpiphdr *ti;
320	caddr_t optp = NULL;
321	int optlen = 0;
322	int len, tlen, off;
323	register struct tcpcb *tp = 0;
324	register int tiflags;
325	struct socket *so = 0;
326	int todrop, acked, ourfinisacked, needoutput = 0;
327	/* int dropsocket = 0; */
328	int iss = 0;
329	u_long tiwin;
330	int ret;
331	/* int ts_present = 0; */
332
333	DEBUG_CALL("tcp_input");
334	DEBUG_ARGS((dfd," m = %8lx iphlen = %2d inso = %lx\n",
335	(long )m, iphlen, (long )inso ));
336
337	/*
338	* If called with m == 0, then we're continuing the connect
339	*/
340	if (m == NULL) {
341	so = inso;
342
343	/* Re-set a few variables */
344	tp = sototcpcb(so);
345	m = so->so_m;
346	so->so_m = 0;
347	ti = so->so_ti;
348	tiwin = ti->ti_win;
349	tiflags = ti->ti_flags;
350
351	goto cont_conn;
352	}
353
354
355	tcpstat.tcps_rcvtotal++;
356	/*
357	* Get IP and TCP header together in first mbuf.
358	* Note: IP leaves IP header in first mbuf.
359	*/
360	ti = mtod(m, struct tcpiphdr *);
361	if (iphlen > sizeof(struct ip )) {
362	ip_stripoptions(m, (struct mbuf *)0);
363	iphlen=sizeof(struct ip );
364	}
365	/* XXX Check if too short */
366
367
368	/*
369	* Save a copy of the IP header in case we want restore it
370	* for sending an ICMP error message in response.
371	*/
372	ip=mtod(m, struct ip *);
373	save_ip = *ip;
374	save_ip.ip_len+= iphlen;
375
376	/*
377	* Checksum extended TCP header and data.
378	*/
379	tlen = ((struct ip *)ti)->ip_len;
380	ti->ti_next = ti->ti_prev = 0;
381	ti->ti_x1 = 0;
382	ti->ti_len = htons((u_int16_t)tlen);
383	len = sizeof(struct ip ) + tlen;
384	/* keep checksum for ICMP reply
385	* ti->ti_sum = cksum(m, len);
386	* if (ti->ti_sum) { */
387	if(cksum(m, len)) {
388	tcpstat.tcps_rcvbadsum++;
389	goto drop;
390	}
391
392	/*
393	* Check that TCP offset makes sense,
394	* pull out TCP options and adjust length. XXX
395	*/
396	off = ti->ti_off << 2;
397	if (off < sizeof (struct tcphdr) \|\| off > tlen) {
398	tcpstat.tcps_rcvbadoff++;
399	goto drop;
400	}
401	tlen -= off;
402	ti->ti_len = tlen;
403	if (off > sizeof (struct tcphdr)) {
404	optlen = off - sizeof (struct tcphdr);
405	optp = mtod(m, caddr_t) + sizeof (struct tcpiphdr);
406
407	/*
408	* Do quick retrieval of timestamp options ("options
409	* prediction?"). If timestamp is the only option and it's
410	* formatted as recommended in RFC 1323 appendix A, we
411	* quickly get the values now and not bother calling
412	* tcp_dooptions(), etc.
413	*/
414	/* if ((optlen == TCPOLEN_TSTAMP_APPA \|\|
415	* (optlen > TCPOLEN_TSTAMP_APPA &&
416	* optp[TCPOLEN_TSTAMP_APPA] == TCPOPT_EOL)) &&
417	* (u_int32_t )optp == htonl(TCPOPT_TSTAMP_HDR) &&
418	* (ti->ti_flags & TH_SYN) == 0) {
419	* ts_present = 1;
420	* ts_val = ntohl((u_int32_t )(optp + 4));
421	* ts_ecr = ntohl((u_int32_t )(optp + 8));
422	* optp = NULL; / * we've parsed the options * /
423	* }
424	*/
425	}
426	tiflags = ti->ti_flags;
427
428	/*
429	* Convert TCP protocol specific fields to host format.
430	*/
431	NTOHL(ti->ti_seq);
432	NTOHL(ti->ti_ack);
433	NTOHS(ti->ti_win);
434	NTOHS(ti->ti_urp);
435
436	/*
437	* Drop TCP, IP headers and TCP options.
438	*/
439	m->m_data += sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
440	m->m_len -= sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
441
442	/*
443	* Locate pcb for segment.
444	*/
445	findso:
446	so = tcp_last_so;
447	if (so->so_fport != ti->ti_dport \|\|
448	so->so_lport != ti->ti_sport \|\|
449	so->so_laddr.s_addr != ti->ti_src.s_addr \|\|
450	so->so_faddr.s_addr != ti->ti_dst.s_addr) {
451	so = solookup(&tcb, ti->ti_src, ti->ti_sport,
452	ti->ti_dst, ti->ti_dport);
453	if (so)
454	tcp_last_so = so;
455	++tcpstat.tcps_socachemiss;
456	}
457
458	/*
459	* If the state is CLOSED (i.e., TCB does not exist) then
460	* all data in the incoming segment is discarded.
461	* If the TCB exists but is in CLOSED state, it is embryonic,
462	* but should either do a listen or a connect soon.
463	*
464	* state == CLOSED means we've done socreate() but haven't
465	* attached it to a protocol yet...
466	*
467	* XXX If a TCB does not exist, and the TH_SYN flag is
468	* the only flag set, then create a session, mark it
469	* as if it was LISTENING, and continue...
470	*/
471	if (so == 0) {
472	if ((tiflags & (TH_SYN\|TH_FIN\|TH_RST\|TH_URG\|TH_ACK)) != TH_SYN)
473	goto dropwithreset;
474
475	if ((so = socreate()) == NULL)
476	goto dropwithreset;
477	#ifdef VBOX
478	if (tcp_attach(pData, so) < 0) {
479	#else /* !VBOX */
480	if (tcp_attach(so) < 0) {
481	#endif /* !VBOX */
482	free(so); /* Not sofree (if it failed, it's not insqued) */
483	goto dropwithreset;
484	}
485
486	sbreserve(&so->so_snd, tcp_sndspace);
487	sbreserve(&so->so_rcv, tcp_rcvspace);
488
489	/* tcp_last_so = so; / / XXX ? */
490	/* tp = sototcpcb(so); */
491
492	so->so_laddr = ti->ti_src;
493	so->so_lport = ti->ti_sport;
494	so->so_faddr = ti->ti_dst;
495	so->so_fport = ti->ti_dport;
496
497	if ((so->so_iptos = tcp_tos(so)) == 0)
498	so->so_iptos = ((struct ip *)ti)->ip_tos;
499
500	tp = sototcpcb(so);
501	tp->t_state = TCPS_LISTEN;
502	}
503
504	/*
505	* If this is a still-connecting socket, this probably
506	* a retransmit of the SYN. Whether it's a retransmit SYN
507	* or something else, we nuke it.
508	*/
509	if (so->so_state & SS_ISFCONNECTING)
510	goto drop;
511
512	tp = sototcpcb(so);
513
514	/* XXX Should never fail */
515	if (tp == 0)
516	goto dropwithreset;
517	if (tp->t_state == TCPS_CLOSED)
518	goto drop;
519
520	/* Unscale the window into a 32-bit value. */
521	/* if ((tiflags & TH_SYN) == 0)
522	* tiwin = ti->ti_win << tp->snd_scale;
523	* else
524	*/
525	tiwin = ti->ti_win;
526
527	/*
528	* Segment received on connection.
529	* Reset idle time and keep-alive timer.
530	*/
531	tp->t_idle = 0;
532	if (so_options)
533	tp->t_timer[TCPT_KEEP] = tcp_keepintvl;
534	else
535	tp->t_timer[TCPT_KEEP] = tcp_keepidle;
536
537	/*
538	* Process options if not in LISTEN state,
539	* else do it below (after getting remote address).
540	*/
541	if (optp && tp->t_state != TCPS_LISTEN)
542	#ifdef VBOX
543	tcp_dooptions(pData, tp, (u_char *)optp, optlen, ti);
544	#else /* !VBOX */
545	tcp_dooptions(tp, (u_char *)optp, optlen, ti);
546	#endif /* !VBOX */
547	/* , */
548	/* &ts_present, &ts_val, &ts_ecr); */
549
550	/*
551	* Header prediction: check for the two common cases
552	* of a uni-directional data xfer. If the packet has
553	* no control flags, is in-sequence, the window didn't
554	* change and we're not retransmitting, it's a
555	* candidate. If the length is zero and the ack moved
556	* forward, we're the sender side of the xfer. Just
557	* free the data acked & wake any higher level process
558	* that was blocked waiting for space. If the length
559	* is non-zero and the ack didn't move, we're the
560	* receiver side. If we're getting packets in-order
561	* (the reassembly queue is empty), add the data to
562	* the socket buffer and note that we need a delayed ack.
563	*
564	* XXX Some of these tests are not needed
565	* eg: the tiwin == tp->snd_wnd prevents many more
566	* predictions.. with no real advantage..
567	*/
568	if (tp->t_state == TCPS_ESTABLISHED &&
569	(tiflags & (TH_SYN\|TH_FIN\|TH_RST\|TH_URG\|TH_ACK)) == TH_ACK &&
570	/* (!ts_present \|\| TSTMP_GEQ(ts_val, tp->ts_recent)) && */
571	ti->ti_seq == tp->rcv_nxt &&
572	tiwin && tiwin == tp->snd_wnd &&
573	tp->snd_nxt == tp->snd_max) {
574	/*
575	* If last ACK falls within this segment's sequence numbers,
576	* record the timestamp.
577	*/
578	/* if (ts_present && SEQ_LEQ(ti->ti_seq, tp->last_ack_sent) &&
579	* SEQ_LT(tp->last_ack_sent, ti->ti_seq + ti->ti_len)) {
580	* tp->ts_recent_age = tcp_now;
581	* tp->ts_recent = ts_val;
582	* }
583	*/
584	if (ti->ti_len == 0) {
585	if (SEQ_GT(ti->ti_ack, tp->snd_una) &&
586	SEQ_LEQ(ti->ti_ack, tp->snd_max) &&
587	tp->snd_cwnd >= tp->snd_wnd) {
588	/*
589	* this is a pure ack for outstanding data.
590	*/
591	++tcpstat.tcps_predack;
592	/* if (ts_present)
593	* tcp_xmit_timer(tp, tcp_now-ts_ecr+1);
594	* else
595	*/ if (tp->t_rtt &&
596	SEQ_GT(ti->ti_ack, tp->t_rtseq))
597	#ifdef VBOX
598	tcp_xmit_timer(pData, tp, tp->t_rtt);
599	#else /* !VBOX */
600	tcp_xmit_timer(tp, tp->t_rtt);
601	#endif /* !VBOX */
602	acked = ti->ti_ack - tp->snd_una;
603	tcpstat.tcps_rcvackpack++;
604	tcpstat.tcps_rcvackbyte += acked;
605	sbdrop(&so->so_snd, acked);
606	tp->snd_una = ti->ti_ack;
607	#ifdef VBOX
608	m_freem(pData, m);
609	#else /* !VBOX */
610	m_freem(m);
611	#endif /* !VBOX */
612
613	/*
614	* If all outstanding data are acked, stop
615	* retransmit timer, otherwise restart timer
616	* using current (possibly backed-off) value.
617	* If process is waiting for space,
618	* wakeup/selwakeup/signal. If data
619	* are ready to send, let tcp_output
620	* decide between more output or persist.
621	*/
622	if (tp->snd_una == tp->snd_max)
623	tp->t_timer[TCPT_REXMT] = 0;
624	else if (tp->t_timer[TCPT_PERSIST] == 0)
625	tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
626
627	/*
628	* There's room in so_snd, sowwakup will read()
629	* from the socket if we can
630	*/
631	/* if (so->so_snd.sb_flags & SB_NOTIFY)
632	* sowwakeup(so);
633	*/
634	/*
635	* This is called because sowwakeup might have
636	* put data into so_snd. Since we don't so sowwakeup,
637	* we don't need this.. XXX???
638	*/
639	if (so->so_snd.sb_cc)
640	#ifdef VBOX
641	(void) tcp_output(pData, tp);
642	#else /* !VBOX */
643	(void) tcp_output(tp);
644	#endif /* !VBOX */
645
646	return;
647	}
648	} else if (ti->ti_ack == tp->snd_una &&
649	u32_to_ptr(pData, tp->seg_next, struct tcpcb *) == tp &&
650	ti->ti_len <= sbspace(&so->so_rcv)) {
651	/*
652	* this is a pure, in-sequence data packet
653	* with nothing on the reassembly queue and
654	* we have enough buffer space to take it.
655	*/
656	++tcpstat.tcps_preddat;
657	tp->rcv_nxt += ti->ti_len;
658	tcpstat.tcps_rcvpack++;
659	tcpstat.tcps_rcvbyte += ti->ti_len;
660	/*
661	* Add data to socket buffer.
662	*/
663	if (so->so_emu) {
664	#ifdef VBOX
665	if (tcp_emu(pData, so,m)) sbappend(pData, so, m);
666	#else /* !VBOX */
667	if (tcp_emu(so,m)) sbappend(so, m);
668	#endif /* !VBOX */
669	} else
670	#ifdef VBOX
671	sbappend(pData, so, m);
672	#else /* !VBOX */
673	sbappend(so, m);
674	#endif /* !VBOX */
675
676	/*
677	* XXX This is called when data arrives. Later, check
678	* if we can actually write() to the socket
679	* XXX Need to check? It's be NON_BLOCKING
680	*/
681	/* sorwakeup(so); */
682
683	/*
684	* If this is a short packet, then ACK now - with Nagel
685	* congestion avoidance sender won't send more until
686	* he gets an ACK.
687	*
688	* It is better to not delay acks at all to maximize
689	* TCP throughput. See RFC 2581.
690	*/
691	tp->t_flags \|= TF_ACKNOW;
692	#ifdef VBOX
693	tcp_output(pData, tp);
694	#else /* !VBOX */
695	tcp_output(tp);
696	#endif /* !VBOX */
697	return;
698	}
699	} /* header prediction */
700	/*
701	* Calculate amount of space in receive window,
702	* and then do TCP input processing.
703	* Receive window is amount of space in rcv queue,
704	* but not less than advertised window.
705	*/
706	{ int win;
707	win = sbspace(&so->so_rcv);
708	if (win < 0)
709	win = 0;
710	tp->rcv_wnd = max(win, (int)(tp->rcv_adv - tp->rcv_nxt));
711	}
712
713	switch (tp->t_state) {
714
715	/*
716	* If the state is LISTEN then ignore segment if it contains an RST.
717	* If the segment contains an ACK then it is bad and send a RST.
718	* If it does not contain a SYN then it is not interesting; drop it.
719	* Don't bother responding if the destination was a broadcast.
720	* Otherwise initialize tp->rcv_nxt, and tp->irs, select an initial
721	* tp->iss, and send a segment:
722	* <SEQ=ISS><ACK=RCV_NXT><CTL=SYN,ACK>
723	* Also initialize tp->snd_nxt to tp->iss+1 and tp->snd_una to tp->iss.
724	* Fill in remote peer address fields if not previously specified.
725	* Enter SYN_RECEIVED state, and process any other fields of this
726	* segment in this state.
727	*/
728	case TCPS_LISTEN: {
729
730	if (tiflags & TH_RST)
731	goto drop;
732	if (tiflags & TH_ACK)
733	goto dropwithreset;
734	if ((tiflags & TH_SYN) == 0)
735	goto drop;
736
737	/*
738	* This has way too many gotos...
739	* But a bit of spaghetti code never hurt anybody :)
740	*/
741
742	/*
743	* If this is destined for the control address, then flag to
744	* tcp_ctl once connected, otherwise connect
745	*/
746	if ((so->so_faddr.s_addr&htonl(0xffffff00)) == special_addr.s_addr) {
747	int lastbyte=ntohl(so->so_faddr.s_addr) & 0xff;
748	if (lastbyte!=CTL_ALIAS && lastbyte!=CTL_DNS) {
749	#if 0
750	if(lastbyte==CTL_CMD \|\| lastbyte==CTL_EXEC) {
751	/* Command or exec adress */
752	so->so_state \|= SS_CTL;
753	} else
754	#endif
755	{
756	/* May be an add exec */
757	struct ex_list *ex_ptr;
758	for(ex_ptr = exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next) {
759	if(ex_ptr->ex_fport == so->so_fport &&
760	lastbyte == ex_ptr->ex_addr) {
761	so->so_state \|= SS_CTL;
762	break;
763	}
764	}
765	}
766	if(so->so_state & SS_CTL) goto cont_input;
767	}
768	/* CTL_ALIAS: Do nothing, tcp_fconnect will be called on it */
769	}
770
771	if (so->so_emu & EMU_NOCONNECT) {
772	so->so_emu &= ~EMU_NOCONNECT;
773	goto cont_input;
774	}
775
776	#ifdef VBOX
777	if((tcp_fconnect(pData, so) == -1) && (errno != EINPROGRESS) && (errno != EWOULDBLOCK)) {
778	#else /* !VBOX */
779	if((tcp_fconnect(so) == -1) && (errno != EINPROGRESS) && (errno != EWOULDBLOCK)) {
780	#endif /* !VBOX */
781	u_char code=ICMP_UNREACH_NET;
782	DEBUG_MISC((dfd," tcp fconnect errno = %d-%s\n",
783	errno,strerror(errno)));
784	if(errno == ECONNREFUSED) {
785	/* ACK the SYN, send RST to refuse the connection */
786	#ifdef VBOX
787	tcp_respond(pData, tp, ti, m, ti->ti_seq+1, (tcp_seq)0,
788	TH_RST\|TH_ACK);
789	#else /* !VBOX */
790	tcp_respond(tp, ti, m, ti->ti_seq+1, (tcp_seq)0,
791	TH_RST\|TH_ACK);
792	#endif /* !VBOX */
793	} else {
794	if(errno == EHOSTUNREACH) code=ICMP_UNREACH_HOST;
795	HTONL(ti->ti_seq); /* restore tcp header */
796	HTONL(ti->ti_ack);
797	HTONS(ti->ti_win);
798	HTONS(ti->ti_urp);
799	m->m_data -= sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
800	m->m_len += sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
801	*ip=save_ip;
802	#ifdef VBOX
803	icmp_error(pData, m, ICMP_UNREACH,code, 0,strerror(errno));
804	#else /* !VBOX */
805	icmp_error(m, ICMP_UNREACH,code, 0,strerror(errno));
806	#endif /* !VBOX */
807	}
808	#ifdef VBOX
809	tp = tcp_close(pData, tp);
810	m_free(pData, m);
811	#else /* !VBOX */
812	tp = tcp_close(tp);
813	m_free(m);
814	#endif /* !VBOX */
815	} else {
816	/*
817	* Haven't connected yet, save the current mbuf
818	* and ti, and return
819	* XXX Some OS's don't tell us whether the connect()
820	* succeeded or not. So we must time it out.
821	*/
822	so->so_m = m;
823	so->so_ti = ti;
824	tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT;
825	tp->t_state = TCPS_SYN_RECEIVED;
826	}
827	return;
828
829	cont_conn:
830	/* m==NULL
831	* Check if the connect succeeded
832	*/
833	if (so->so_state & SS_NOFDREF) {
834	#ifdef VBOX
835	tp = tcp_close(pData, tp);
836	#else /* !VBOX */
837	tp = tcp_close(tp);
838	#endif /* !VBOX */
839	goto dropwithreset;
840	}
841	cont_input:
842	tcp_template(tp);
843
844	if (optp)
845	#ifdef VBOX
846	tcp_dooptions(pData, tp, (u_char *)optp, optlen, ti);
847	#else /* !VBOX */
848	tcp_dooptions(tp, (u_char *)optp, optlen, ti);
849	#endif /* !VBOX */
850	/* , */
851	/* &ts_present, &ts_val, &ts_ecr); */
852
853	if (iss)
854	tp->iss = iss;
855	else
856	tp->iss = tcp_iss;
857	tcp_iss += TCP_ISSINCR/2;
858	tp->irs = ti->ti_seq;
859	tcp_sendseqinit(tp);
860	tcp_rcvseqinit(tp);
861	tp->t_flags \|= TF_ACKNOW;
862	tp->t_state = TCPS_SYN_RECEIVED;
863	tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT;
864	tcpstat.tcps_accepts++;
865	goto trimthenstep6;
866	} /* case TCPS_LISTEN */
867
868	/*
869	* If the state is SYN_SENT:
870	* if seg contains an ACK, but not for our SYN, drop the input.
871	* if seg contains a RST, then drop the connection.
872	* if seg does not contain SYN, then drop it.
873	* Otherwise this is an acceptable SYN segment
874	* initialize tp->rcv_nxt and tp->irs
875	* if seg contains ack then advance tp->snd_una
876	* if SYN has been acked change to ESTABLISHED else SYN_RCVD state
877	* arrange for segment to be acked (eventually)
878	* continue processing rest of data/controls, beginning with URG
879	*/
880	case TCPS_SYN_SENT:
881	if ((tiflags & TH_ACK) &&
882	(SEQ_LEQ(ti->ti_ack, tp->iss) \|\|
883	SEQ_GT(ti->ti_ack, tp->snd_max)))
884	goto dropwithreset;
885
886	if (tiflags & TH_RST) {
887	if (tiflags & TH_ACK)
888	#ifdef VBOX
889	tp = tcp_drop(pData, tp,0); /* XXX Check t_softerror! */
890	#else /* !VBOX */
891	tp = tcp_drop(tp,0); /* XXX Check t_softerror! */
892	#endif /* !VBOX */
893	goto drop;
894	}
895
896	if ((tiflags & TH_SYN) == 0)
897	goto drop;
898	if (tiflags & TH_ACK) {
899	tp->snd_una = ti->ti_ack;
900	if (SEQ_LT(tp->snd_nxt, tp->snd_una))
901	tp->snd_nxt = tp->snd_una;
902	}
903
904	tp->t_timer[TCPT_REXMT] = 0;
905	tp->irs = ti->ti_seq;
906	tcp_rcvseqinit(tp);
907	tp->t_flags \|= TF_ACKNOW;
908	if (tiflags & TH_ACK && SEQ_GT(tp->snd_una, tp->iss)) {
909	tcpstat.tcps_connects++;
910	soisfconnected(so);
911	tp->t_state = TCPS_ESTABLISHED;
912
913	/* Do window scaling on this connection? */
914	/* if ((tp->t_flags & (TF_RCVD_SCALE\|TF_REQ_SCALE)) ==
915	* (TF_RCVD_SCALE\|TF_REQ_SCALE)) {
916	* tp->snd_scale = tp->requested_s_scale;
917	* tp->rcv_scale = tp->request_r_scale;
918	* }
919	*/
920	#ifdef VBOX
921	(void) tcp_reass(pData, tp, (struct tcpiphdr *)0,
922	(struct mbuf *)0);
923	#else /* !VBOX */
924	(void) tcp_reass(tp, (struct tcpiphdr *)0,
925	(struct mbuf *)0);
926	#endif /* !VBOX */
927	/*
928	* if we didn't have to retransmit the SYN,
929	* use its rtt as our initial srtt & rtt var.
930	*/
931	if (tp->t_rtt)
932	#ifdef VBOX
933	tcp_xmit_timer(pData, tp, tp->t_rtt);
934	#else /* !VBOX */
935	tcp_xmit_timer(tp, tp->t_rtt);
936	#endif /* !VBOX */
937	} else
938	tp->t_state = TCPS_SYN_RECEIVED;
939
940	trimthenstep6:
941	/*
942	* Advance ti->ti_seq to correspond to first data byte.
943	* If data, trim to stay within window,
944	* dropping FIN if necessary.
945	*/
946	ti->ti_seq++;
947	if (ti->ti_len > tp->rcv_wnd) {
948	todrop = ti->ti_len - tp->rcv_wnd;
949	m_adj(m, -todrop);
950	ti->ti_len = tp->rcv_wnd;
951	tiflags &= ~TH_FIN;
952	tcpstat.tcps_rcvpackafterwin++;
953	tcpstat.tcps_rcvbyteafterwin += todrop;
954	}
955	tp->snd_wl1 = ti->ti_seq - 1;
956	tp->rcv_up = ti->ti_seq;
957	goto step6;
958	} /* switch tp->t_state */
959	/*
960	* States other than LISTEN or SYN_SENT.
961	* First check timestamp, if present.
962	* Then check that at least some bytes of segment are within
963	* receive window. If segment begins before rcv_nxt,
964	* drop leading data (and SYN); if nothing left, just ack.
965	*
966	* RFC 1323 PAWS: If we have a timestamp reply on this segment
967	* and it's less than ts_recent, drop it.
968	*/
969	/* if (ts_present && (tiflags & TH_RST) == 0 && tp->ts_recent &&
970	* TSTMP_LT(ts_val, tp->ts_recent)) {
971	*
972	/ / Check to see if ts_recent is over 24 days old. */
973	/* if ((int)(tcp_now - tp->ts_recent_age) > TCP_PAWS_IDLE) {
974	/ /
975	* * Invalidate ts_recent. If this segment updates
976	* * ts_recent, the age will be reset later and ts_recent
977	* * will get a valid value. If it does not, setting
978	* * ts_recent to zero will at least satisfy the
979	* * requirement that zero be placed in the timestamp
980	* * echo reply when ts_recent isn't valid. The
981	* * age isn't reset until we get a valid ts_recent
982	* * because we don't want out-of-order segments to be
983	* * dropped when ts_recent is old.
984	* */
985	/* tp->ts_recent = 0;
986	* } else {
987	* tcpstat.tcps_rcvduppack++;
988	* tcpstat.tcps_rcvdupbyte += ti->ti_len;
989	* tcpstat.tcps_pawsdrop++;
990	* goto dropafterack;
991	* }
992	* }
993	*/
994
995	todrop = tp->rcv_nxt - ti->ti_seq;
996	if (todrop > 0) {
997	if (tiflags & TH_SYN) {
998	tiflags &= ~TH_SYN;
999	ti->ti_seq++;
1000	if (ti->ti_urp > 1)
1001	ti->ti_urp--;
1002	else
1003	tiflags &= ~TH_URG;
1004	todrop--;
1005	}
1006	/*
1007	* Following if statement from Stevens, vol. 2, p. 960.
1008	*/
1009	if (todrop > ti->ti_len
1010	\|\| (todrop == ti->ti_len && (tiflags & TH_FIN) == 0)) {
1011	/*
1012	* Any valid FIN must be to the left of the window.
1013	* At this point the FIN must be a duplicate or out
1014	* of sequence; drop it.
1015	*/
1016	tiflags &= ~TH_FIN;
1017
1018	/*
1019	* Send an ACK to resynchronize and drop any data.
1020	* But keep on processing for RST or ACK.
1021	*/
1022	tp->t_flags \|= TF_ACKNOW;
1023	todrop = ti->ti_len;
1024	tcpstat.tcps_rcvduppack++;
1025	tcpstat.tcps_rcvdupbyte += todrop;
1026	} else {
1027	tcpstat.tcps_rcvpartduppack++;
1028	tcpstat.tcps_rcvpartdupbyte += todrop;
1029	}
1030	m_adj(m, todrop);
1031	ti->ti_seq += todrop;
1032	ti->ti_len -= todrop;
1033	if (ti->ti_urp > todrop)
1034	ti->ti_urp -= todrop;
1035	else {
1036	tiflags &= ~TH_URG;
1037	ti->ti_urp = 0;
1038	}
1039	}
1040	/*
1041	* If new data are received on a connection after the
1042	* user processes are gone, then RST the other end.
1043	*/
1044	if ((so->so_state & SS_NOFDREF) &&
1045	tp->t_state > TCPS_CLOSE_WAIT && ti->ti_len) {
1046	#ifdef VBOX
1047	tp = tcp_close(pData, tp);
1048	#else /* !VBOX */
1049	tp = tcp_close(tp);
1050	#endif /* !VBOX */
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	#ifdef VBOX
1075	tp = tcp_close(pData, tp);
1076	#else /* !VBOX */
1077	tp = tcp_close(tp);
1078	#endif /* !VBOX */
1079	goto findso;
1080	}
1081	/*
1082	* If window is closed can only take segments at
1083	* window edge, and have to drop data and PUSH from
1084	* incoming segments. Continue processing, but
1085	* remember to ack. Otherwise, drop segment
1086	* and ack.
1087	*/
1088	if (tp->rcv_wnd == 0 && ti->ti_seq == tp->rcv_nxt) {
1089	tp->t_flags \|= TF_ACKNOW;
1090	tcpstat.tcps_rcvwinprobe++;
1091	} else
1092	goto dropafterack;
1093	} else
1094	tcpstat.tcps_rcvbyteafterwin += todrop;
1095	m_adj(m, -todrop);
1096	ti->ti_len -= todrop;
1097	tiflags &= ~(TH_PUSH\|TH_FIN);
1098	}
1099
1100	/*
1101	* If last ACK falls within this segment's sequence numbers,
1102	* record its timestamp.
1103	*/
1104	/* if (ts_present && SEQ_LEQ(ti->ti_seq, tp->last_ack_sent) &&
1105	* SEQ_LT(tp->last_ack_sent, ti->ti_seq + ti->ti_len +
1106	* ((tiflags & (TH_SYN\|TH_FIN)) != 0))) {
1107	* tp->ts_recent_age = tcp_now;
1108	* tp->ts_recent = ts_val;
1109	* }
1110	*/
1111
1112	/*
1113	* If the RST bit is set examine the state:
1114	* SYN_RECEIVED STATE:
1115	* If passive open, return to LISTEN state.
1116	* If active open, inform user that connection was refused.
1117	* ESTABLISHED, FIN_WAIT_1, FIN_WAIT2, CLOSE_WAIT STATES:
1118	* Inform user that connection was reset, and close tcb.
1119	* CLOSING, LAST_ACK, TIME_WAIT STATES
1120	* Close the tcb.
1121	*/
1122	if (tiflags&TH_RST) switch (tp->t_state) {
1123
1124	case TCPS_SYN_RECEIVED:
1125	/* so->so_error = ECONNREFUSED; */
1126	goto close;
1127
1128	case TCPS_ESTABLISHED:
1129	case TCPS_FIN_WAIT_1:
1130	case TCPS_FIN_WAIT_2:
1131	case TCPS_CLOSE_WAIT:
1132	/* so->so_error = ECONNRESET; */
1133	close:
1134	tp->t_state = TCPS_CLOSED;
1135	tcpstat.tcps_drops++;
1136	#ifdef VBOX
1137	tp = tcp_close(pData, tp);
1138	#else /* !VBOX */
1139	tp = tcp_close(tp);
1140	#endif /* !VBOX */
1141	goto drop;
1142
1143	case TCPS_CLOSING:
1144	case TCPS_LAST_ACK:
1145	case TCPS_TIME_WAIT:
1146	#ifdef VBOX
1147	tp = tcp_close(pData, tp);
1148	#else /* !VBOX */
1149	tp = tcp_close(tp);
1150	#endif /* !VBOX */
1151	goto drop;
1152	}
1153
1154	/*
1155	* If a SYN is in the window, then this is an
1156	* error and we send an RST and drop the connection.
1157	*/
1158	if (tiflags & TH_SYN) {
1159	#ifdef VBOX
1160	tp = tcp_drop(pData, tp,0);
1161	#else /* !VBOX */
1162	tp = tcp_drop(tp,0);
1163	#endif /* !VBOX */
1164	goto dropwithreset;
1165	}
1166
1167	/*
1168	* If the ACK bit is off we drop the segment and return.
1169	*/
1170	if ((tiflags & TH_ACK) == 0) goto drop;
1171
1172	/*
1173	* Ack processing.
1174	*/
1175	switch (tp->t_state) {
1176	/*
1177	* In SYN_RECEIVED state if the ack ACKs our SYN then enter
1178	* ESTABLISHED state and continue processing, otherwise
1179	* send an RST. una<=ack<=max
1180	*/
1181	case TCPS_SYN_RECEIVED:
1182
1183	if (SEQ_GT(tp->snd_una, ti->ti_ack) \|\|
1184	SEQ_GT(ti->ti_ack, tp->snd_max))
1185	goto dropwithreset;
1186	tcpstat.tcps_connects++;
1187	tp->t_state = TCPS_ESTABLISHED;
1188	/*
1189	* The sent SYN is ack'ed with our sequence number +1
1190	* The first data byte already in the buffer will get
1191	* lost if no correction is made. This is only needed for
1192	* SS_CTL since the buffer is empty otherwise.
1193	* tp->snd_una++; or:
1194	*/
1195	tp->snd_una=ti->ti_ack;
1196	if (so->so_state & SS_CTL) {
1197	/* So tcp_ctl reports the right state */
1198	#ifdef VBOX
1199	ret = tcp_ctl(pData, so);
1200	#else /* !VBOX */
1201	ret = tcp_ctl(so);
1202	#endif /* !VBOX */
1203	if (ret == 1) {
1204	soisfconnected(so);
1205	so->so_state &= ~SS_CTL; /* success XXX */
1206	} else if (ret == 2) {
1207	so->so_state = SS_NOFDREF; /* CTL_CMD */
1208	} else {
1209	needoutput = 1;
1210	tp->t_state = TCPS_FIN_WAIT_1;
1211	}
1212	} else {
1213	soisfconnected(so);
1214	}
1215
1216	/* Do window scaling? */
1217	/* if ((tp->t_flags & (TF_RCVD_SCALE\|TF_REQ_SCALE)) ==
1218	* (TF_RCVD_SCALE\|TF_REQ_SCALE)) {
1219	* tp->snd_scale = tp->requested_s_scale;
1220	* tp->rcv_scale = tp->request_r_scale;
1221	* }
1222	*/
1223	#ifdef VBOX
1224	(void) tcp_reass(pData, tp, (struct tcpiphdr )0, (struct mbuf )0);
1225	#else /* !VBOX */
1226	(void) tcp_reass(tp, (struct tcpiphdr )0, (struct mbuf )0);
1227	#endif /* !VBOX */
1228	tp->snd_wl1 = ti->ti_seq - 1;
1229	/* Avoid ack processing; snd_una==ti_ack => dup ack */
1230	goto synrx_to_est;
1231	/* fall into ... */
1232
1233	/*
1234	* In ESTABLISHED state: drop duplicate ACKs; ACK out of range
1235	* ACKs. If the ack is in the range
1236	* tp->snd_una < ti->ti_ack <= tp->snd_max
1237	* then advance tp->snd_una to ti->ti_ack and drop
1238	* data from the retransmission queue. If this ACK reflects
1239	* more up to date window information we update our window information.
1240	*/
1241	case TCPS_ESTABLISHED:
1242	case TCPS_FIN_WAIT_1:
1243	case TCPS_FIN_WAIT_2:
1244	case TCPS_CLOSE_WAIT:
1245	case TCPS_CLOSING:
1246	case TCPS_LAST_ACK:
1247	case TCPS_TIME_WAIT:
1248
1249	if (SEQ_LEQ(ti->ti_ack, tp->snd_una)) {
1250	if (ti->ti_len == 0 && tiwin == tp->snd_wnd) {
1251	tcpstat.tcps_rcvdupack++;
1252	DEBUG_MISC((dfd," dup ack m = %lx so = %lx \n",
1253	(long )m, (long )so));
1254	/*
1255	* If we have outstanding data (other than
1256	* a window probe), this is a completely
1257	* duplicate ack (ie, window info didn't
1258	* change), the ack is the biggest we've
1259	* seen and we've seen exactly our rexmt
1260	* threshold of them, assume a packet
1261	* has been dropped and retransmit it.
1262	* Kludge snd_nxt & the congestion
1263	* window so we send only this one
1264	* packet.
1265	*
1266	* We know we're losing at the current
1267	* window size so do congestion avoidance
1268	* (set ssthresh to half the current window
1269	* and pull our congestion window back to
1270	* the new ssthresh).
1271	*
1272	* Dup acks mean that packets have left the
1273	* network (they're now cached at the receiver)
1274	* so bump cwnd by the amount in the receiver
1275	* to keep a constant cwnd packets in the
1276	* network.
1277	*/
1278	if (tp->t_timer[TCPT_REXMT] == 0 \|\|
1279	ti->ti_ack != tp->snd_una)
1280	tp->t_dupacks = 0;
1281	else if (++tp->t_dupacks == tcprexmtthresh) {
1282	tcp_seq onxt = tp->snd_nxt;
1283	u_int win =
1284	min(tp->snd_wnd, tp->snd_cwnd) / 2 /
1285	tp->t_maxseg;
1286
1287	if (win < 2)
1288	win = 2;
1289	tp->snd_ssthresh = win * tp->t_maxseg;
1290	tp->t_timer[TCPT_REXMT] = 0;
1291	tp->t_rtt = 0;
1292	tp->snd_nxt = ti->ti_ack;
1293	tp->snd_cwnd = tp->t_maxseg;
1294	#ifdef VBOX
1295	(void) tcp_output(pData, tp);
1296	#else /* !VBOX */
1297	(void) tcp_output(tp);
1298	#endif /* !VBOX */
1299	tp->snd_cwnd = tp->snd_ssthresh +
1300	tp->t_maxseg * tp->t_dupacks;
1301	if (SEQ_GT(onxt, tp->snd_nxt))
1302	tp->snd_nxt = onxt;
1303	goto drop;
1304	} else if (tp->t_dupacks > tcprexmtthresh) {
1305	tp->snd_cwnd += tp->t_maxseg;
1306	#ifdef VBOX
1307	(void) tcp_output(pData, tp);
1308	#else /* !VBOX */
1309	(void) tcp_output(tp);
1310	#endif /* !VBOX */
1311	goto drop;
1312	}
1313	} else
1314	tp->t_dupacks = 0;
1315	break;
1316	}
1317	synrx_to_est:
1318	/*
1319	* If the congestion window was inflated to account
1320	* for the other side's cached packets, retract it.
1321	*/
1322	if (tp->t_dupacks > tcprexmtthresh &&
1323	tp->snd_cwnd > tp->snd_ssthresh)
1324	tp->snd_cwnd = tp->snd_ssthresh;
1325	tp->t_dupacks = 0;
1326	if (SEQ_GT(ti->ti_ack, tp->snd_max)) {
1327	tcpstat.tcps_rcvacktoomuch++;
1328	goto dropafterack;
1329	}
1330	acked = ti->ti_ack - tp->snd_una;
1331	tcpstat.tcps_rcvackpack++;
1332	tcpstat.tcps_rcvackbyte += acked;
1333
1334	/*
1335	* If we have a timestamp reply, update smoothed
1336	* round trip time. If no timestamp is present but
1337	* transmit timer is running and timed sequence
1338	* number was acked, update smoothed round trip time.
1339	* Since we now have an rtt measurement, cancel the
1340	* timer backoff (cf., Phil Karn's retransmit alg.).
1341	* Recompute the initial retransmit timer.
1342	*/
1343	/* if (ts_present)
1344	* tcp_xmit_timer(tp, tcp_now-ts_ecr+1);
1345	* else
1346	*/
1347	if (tp->t_rtt && SEQ_GT(ti->ti_ack, tp->t_rtseq))
1348	#ifdef VBOX
1349	tcp_xmit_timer(pData, tp,tp->t_rtt);
1350	#else /* !VBOX */
1351	tcp_xmit_timer(tp,tp->t_rtt);
1352	#endif /* !VBOX */
1353
1354	/*
1355	* If all outstanding data is acked, stop retransmit
1356	* timer and remember to restart (more output or persist).
1357	* If there is more data to be acked, restart retransmit
1358	* timer, using current (possibly backed-off) value.
1359	*/
1360	if (ti->ti_ack == tp->snd_max) {
1361	tp->t_timer[TCPT_REXMT] = 0;
1362	needoutput = 1;
1363	} else if (tp->t_timer[TCPT_PERSIST] == 0)
1364	tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
1365	/*
1366	* When new data is acked, open the congestion window.
1367	* If the window gives us less than ssthresh packets
1368	* in flight, open exponentially (maxseg per packet).
1369	* Otherwise open linearly: maxseg per window
1370	* (maxseg^2 / cwnd per packet).
1371	*/
1372	{
1373	register u_int cw = tp->snd_cwnd;
1374	register u_int incr = tp->t_maxseg;
1375
1376	if (cw > tp->snd_ssthresh)
1377	incr = incr * incr / cw;
1378	tp->snd_cwnd = min(cw + incr, TCP_MAXWIN<<tp->snd_scale);
1379	}
1380	if (acked > so->so_snd.sb_cc) {
1381	tp->snd_wnd -= so->so_snd.sb_cc;
1382	sbdrop(&so->so_snd, (int )so->so_snd.sb_cc);
1383	ourfinisacked = 1;
1384	} else {
1385	sbdrop(&so->so_snd, acked);
1386	tp->snd_wnd -= acked;
1387	ourfinisacked = 0;
1388	}
1389	/*
1390	* XXX sowwakup is called when data is acked and there's room for
1391	* for more data... it should read() the socket
1392	*/
1393	/* if (so->so_snd.sb_flags & SB_NOTIFY)
1394	* sowwakeup(so);
1395	*/
1396	tp->snd_una = ti->ti_ack;
1397	if (SEQ_LT(tp->snd_nxt, tp->snd_una))
1398	tp->snd_nxt = tp->snd_una;
1399
1400	switch (tp->t_state) {
1401
1402	/*
1403	* In FIN_WAIT_1 STATE in addition to the processing
1404	* for the ESTABLISHED state if our FIN is now acknowledged
1405	* then enter FIN_WAIT_2.
1406	*/
1407	case TCPS_FIN_WAIT_1:
1408	if (ourfinisacked) {
1409	/*
1410	* If we can't receive any more
1411	* data, then closing user can proceed.
1412	* Starting the timer is contrary to the
1413	* specification, but if we don't get a FIN
1414	* we'll hang forever.
1415	*/
1416	if (so->so_state & SS_FCANTRCVMORE) {
1417	soisfdisconnected(so);
1418	tp->t_timer[TCPT_2MSL] = tcp_maxidle;
1419	}
1420	tp->t_state = TCPS_FIN_WAIT_2;
1421	}
1422	break;
1423
1424	/*
1425	* In CLOSING STATE in addition to the processing for
1426	* the ESTABLISHED state if the ACK acknowledges our FIN
1427	* then enter the TIME-WAIT state, otherwise ignore
1428	* the segment.
1429	*/
1430	case TCPS_CLOSING:
1431	if (ourfinisacked) {
1432	tp->t_state = TCPS_TIME_WAIT;
1433	tcp_canceltimers(tp);
1434	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1435	soisfdisconnected(so);
1436	}
1437	break;
1438
1439	/*
1440	* In LAST_ACK, we may still be waiting for data to drain
1441	* and/or to be acked, as well as for the ack of our FIN.
1442	* If our FIN is now acknowledged, delete the TCB,
1443	* enter the closed state and return.
1444	*/
1445	case TCPS_LAST_ACK:
1446	if (ourfinisacked) {
1447	#ifdef VBOX
1448	tp = tcp_close(pData, tp);
1449	#else /* !VBOX */
1450	tp = tcp_close(tp);
1451	#endif /* !VBOX */
1452	goto drop;
1453	}
1454	break;
1455
1456	/*
1457	* In TIME_WAIT state the only thing that should arrive
1458	* is a retransmission of the remote FIN. Acknowledge
1459	* it and restart the finack timer.
1460	*/
1461	case TCPS_TIME_WAIT:
1462	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1463	goto dropafterack;
1464	}
1465	} /* switch(tp->t_state) */
1466
1467	step6:
1468	/*
1469	* Update window information.
1470	* Don't look at window if no ACK: TAC's send garbage on first SYN.
1471	*/
1472	if ((tiflags & TH_ACK) &&
1473	(SEQ_LT(tp->snd_wl1, ti->ti_seq) \|\|
1474	(tp->snd_wl1 == ti->ti_seq && (SEQ_LT(tp->snd_wl2, ti->ti_ack) \|\|
1475	(tp->snd_wl2 == ti->ti_ack && tiwin > tp->snd_wnd))))) {
1476	/* keep track of pure window updates */
1477	if (ti->ti_len == 0 &&
1478	tp->snd_wl2 == ti->ti_ack && tiwin > tp->snd_wnd)
1479	tcpstat.tcps_rcvwinupd++;
1480	tp->snd_wnd = tiwin;
1481	tp->snd_wl1 = ti->ti_seq;
1482	tp->snd_wl2 = ti->ti_ack;
1483	if (tp->snd_wnd > tp->max_sndwnd)
1484	tp->max_sndwnd = tp->snd_wnd;
1485	needoutput = 1;
1486	}
1487
1488	/*
1489	* Process segments with URG.
1490	*/
1491	if ((tiflags & TH_URG) && ti->ti_urp &&
1492	TCPS_HAVERCVDFIN(tp->t_state) == 0) {
1493	/*
1494	* This is a kludge, but if we receive and accept
1495	* random urgent pointers, we'll crash in
1496	* soreceive. It's hard to imagine someone
1497	* actually wanting to send this much urgent data.
1498	*/
1499	if (ti->ti_urp + so->so_rcv.sb_cc > so->so_rcv.sb_datalen) {
1500	ti->ti_urp = 0;
1501	tiflags &= ~TH_URG;
1502	goto dodata;
1503	}
1504	/*
1505	* If this segment advances the known urgent pointer,
1506	* then mark the data stream. This should not happen
1507	* in CLOSE_WAIT, CLOSING, LAST_ACK or TIME_WAIT STATES since
1508	* a FIN has been received from the remote side.
1509	* In these states we ignore the URG.
1510	*
1511	* According to RFC961 (Assigned Protocols),
1512	* the urgent pointer points to the last octet
1513	* of urgent data. We continue, however,
1514	* to consider it to indicate the first octet
1515	* of data past the urgent section as the original
1516	* spec states (in one of two places).
1517	*/
1518	if (SEQ_GT(ti->ti_seq+ti->ti_urp, tp->rcv_up)) {
1519	tp->rcv_up = ti->ti_seq + ti->ti_urp;
1520	so->so_urgc = so->so_rcv.sb_cc +
1521	(tp->rcv_up - tp->rcv_nxt); /* -1; */
1522	tp->rcv_up = ti->ti_seq + ti->ti_urp;
1523
1524	}
1525	} else
1526	/*
1527	* If no out of band data is expected,
1528	* pull receive urgent pointer along
1529	* with the receive window.
1530	*/
1531	if (SEQ_GT(tp->rcv_nxt, tp->rcv_up))
1532	tp->rcv_up = tp->rcv_nxt;
1533	dodata:
1534
1535	/*
1536	* Process the segment text, merging it into the TCP sequencing queue,
1537	* and arranging for acknowledgment of receipt if necessary.
1538	* This process logically involves adjusting tp->rcv_wnd as data
1539	* is presented to the user (this happens in tcp_usrreq.c,
1540	* case PRU_RCVD). If a FIN has already been received on this
1541	* connection then we just ignore the text.
1542	*/
1543	if ((ti->ti_len \|\| (tiflags&TH_FIN)) &&
1544	TCPS_HAVERCVDFIN(tp->t_state) == 0) {
1545	#ifdef VBOX
1546	TCP_REASS(pData, tp, ti, m, so, tiflags);
1547	#else /* !VBOX */
1548	TCP_REASS(tp, ti, m, so, tiflags);
1549	#endif /* !VBOX */
1550	/*
1551	* Note the amount of data that peer has sent into
1552	* our window, in order to estimate the sender's
1553	* buffer size.
1554	*/
1555	len = so->so_rcv.sb_datalen - (tp->rcv_adv - tp->rcv_nxt);
1556	} else {
1557	#ifdef VBOX
1558	m_free(pData, m);
1559	#else /* !VBOX */
1560	m_free(m);
1561	#endif /* !VBOX */
1562	tiflags &= ~TH_FIN;
1563	}
1564
1565	/*
1566	* If FIN is received ACK the FIN and let the user know
1567	* that the connection is closing.
1568	*/
1569	if (tiflags & TH_FIN) {
1570	if (TCPS_HAVERCVDFIN(tp->t_state) == 0) {
1571	/*
1572	* If we receive a FIN we can't send more data,
1573	* set it SS_FDRAIN
1574	* Shutdown the socket if there is no rx data in the
1575	* buffer.
1576	* soread() is called on completion of shutdown() and
1577	* will got to TCPS_LAST_ACK, and use tcp_output()
1578	* to send the FIN.
1579	*/
1580	/* sofcantrcvmore(so); */
1581	sofwdrain(so);
1582
1583	tp->t_flags \|= TF_ACKNOW;
1584	tp->rcv_nxt++;
1585	}
1586	switch (tp->t_state) {
1587
1588	/*
1589	* In SYN_RECEIVED and ESTABLISHED STATES
1590	* enter the CLOSE_WAIT state.
1591	*/
1592	case TCPS_SYN_RECEIVED:
1593	case TCPS_ESTABLISHED:
1594	if(so->so_emu == EMU_CTL) /* no shutdown on socket */
1595	tp->t_state = TCPS_LAST_ACK;
1596	else
1597	tp->t_state = TCPS_CLOSE_WAIT;
1598	break;
1599
1600	/*
1601	* If still in FIN_WAIT_1 STATE FIN has not been acked so
1602	* enter the CLOSING state.
1603	*/
1604	case TCPS_FIN_WAIT_1:
1605	tp->t_state = TCPS_CLOSING;
1606	break;
1607
1608	/*
1609	* In FIN_WAIT_2 state enter the TIME_WAIT state,
1610	* starting the time-wait timer, turning off the other
1611	* standard timers.
1612	*/
1613	case TCPS_FIN_WAIT_2:
1614	tp->t_state = TCPS_TIME_WAIT;
1615	tcp_canceltimers(tp);
1616	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1617	soisfdisconnected(so);
1618	break;
1619
1620	/*
1621	* In TIME_WAIT state restart the 2 MSL time_wait timer.
1622	*/
1623	case TCPS_TIME_WAIT:
1624	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1625	break;
1626	}
1627	}
1628
1629	/*
1630	* If this is a small packet, then ACK now - with Nagel
1631	* congestion avoidance sender won't send more until
1632	* he gets an ACK.
1633	*
1634	* See above.
1635	*/
1636	/* if (ti->ti_len && (unsigned)ti->ti_len < tp->t_maxseg) {
1637	*/
1638	/* if ((ti->ti_len && (unsigned)ti->ti_len < tp->t_maxseg &&
1639	* (so->so_iptos & IPTOS_LOWDELAY) == 0) \|\|
1640	* ((so->so_iptos & IPTOS_LOWDELAY) &&
1641	* ((struct tcpiphdr_2 *)ti)->first_char == (char)27)) {
1642	*/
1643	if (ti->ti_len && (unsigned)ti->ti_len <= 5 &&
1644	((struct tcpiphdr_2 *)ti)->first_char == (char)27) {
1645	tp->t_flags \|= TF_ACKNOW;
1646	}
1647
1648	/*
1649	* Return any desired output.
1650	*/
1651	if (needoutput \|\| (tp->t_flags & TF_ACKNOW)) {
1652	#ifdef VBOX
1653	(void) tcp_output(pData, tp);
1654	#else /* !VBOX */
1655	(void) tcp_output(tp);
1656	#endif /* !VBOX */
1657	}
1658	return;
1659
1660	dropafterack:
1661	/*
1662	* Generate an ACK dropping incoming segment if it occupies
1663	* sequence space, where the ACK reflects our state.
1664	*/
1665	if (tiflags & TH_RST)
1666	goto drop;
1667	#ifdef VBOX
1668	m_freem(pData, m);
1669	#else /* !VBOX */
1670	m_freem(m);
1671	#endif /* !VBOX */
1672	tp->t_flags \|= TF_ACKNOW;
1673	#ifdef VBOX
1674	(void) tcp_output(pData, tp);
1675	#else /* !VBOX */
1676	(void) tcp_output(tp);
1677	#endif /* !VBOX */
1678	return;
1679
1680	dropwithreset:
1681	/* reuses m if m!=NULL, m_free() unnecessary */
1682	if (tiflags & TH_ACK)
1683	#ifdef VBOX
1684	tcp_respond(pData, tp, ti, m, (tcp_seq)0, ti->ti_ack, TH_RST);
1685	#else /* !VBOX */
1686	tcp_respond(tp, ti, m, (tcp_seq)0, ti->ti_ack, TH_RST);
1687	#endif /* !VBOX */
1688	else {
1689	if (tiflags & TH_SYN) ti->ti_len++;
1690	#ifdef VBOX
1691	tcp_respond(pData, tp, ti, m, ti->ti_seq+ti->ti_len, (tcp_seq)0,
1692	TH_RST\|TH_ACK);
1693	#else /* !VBOX */
1694	tcp_respond(tp, ti, m, ti->ti_seq+ti->ti_len, (tcp_seq)0,
1695	TH_RST\|TH_ACK);
1696	#endif /* !VBOX */
1697	}
1698
1699	return;
1700
1701	drop:
1702	/*
1703	* Drop space held by incoming segment and return.
1704	*/
1705	#ifdef VBOX
1706	m_free(pData, m);
1707	#else /* !VBOX */
1708	m_free(m);
1709	#endif /* !VBOX */
1710
1711	return;
1712	}
1713
1714	/* , ts_present, ts_val, ts_ecr) */
1715	/* int *ts_present;
1716	* u_int32_t ts_val, ts_ecr;
1717	*/
1718	void
1719	#ifdef VBOX
1720	tcp_dooptions(PNATState pData, struct tcpcb tp, u_char cp, int cnt, struct tcpiphdr *ti)
1721	#else /* !VBOX */
1722	tcp_dooptions(tp, cp, cnt, ti)
1723	struct tcpcb *tp;
1724	u_char *cp;
1725	int cnt;
1726	struct tcpiphdr *ti;
1727	#endif /* !VBOX */
1728	{
1729	u_int16_t mss;
1730	int opt, optlen;
1731
1732	DEBUG_CALL("tcp_dooptions");
1733	DEBUG_ARGS((dfd," tp = %lx cnt=%i \n", (long )tp, cnt));
1734
1735	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1736	opt = cp[0];
1737	if (opt == TCPOPT_EOL)
1738	break;
1739	if (opt == TCPOPT_NOP)
1740	optlen = 1;
1741	else {
1742	optlen = cp[1];
1743	if (optlen <= 0)
1744	break;
1745	}
1746	switch (opt) {
1747
1748	default:
1749	continue;
1750
1751	case TCPOPT_MAXSEG:
1752	if (optlen != TCPOLEN_MAXSEG)
1753	continue;
1754	if (!(ti->ti_flags & TH_SYN))
1755	continue;
1756	memcpy((char ) &mss, (char ) cp + 2, sizeof(mss));
1757	NTOHS(mss);
1758	#ifdef VBOX
1759	(void) tcp_mss(pData, tp, mss); /* sets t_maxseg */
1760	#else /* !VBOX */
1761	(void) tcp_mss(tp, mss); /* sets t_maxseg */
1762	#endif /* !VBOX */
1763	break;
1764
1765	/* case TCPOPT_WINDOW:
1766	* if (optlen != TCPOLEN_WINDOW)
1767	* continue;
1768	* if (!(ti->ti_flags & TH_SYN))
1769	* continue;
1770	* tp->t_flags \|= TF_RCVD_SCALE;
1771	* tp->requested_s_scale = min(cp[2], TCP_MAX_WINSHIFT);
1772	* break;
1773	*/
1774	/* case TCPOPT_TIMESTAMP:
1775	* if (optlen != TCPOLEN_TIMESTAMP)
1776	* continue;
1777	* *ts_present = 1;
1778	* memcpy((char ) ts_val, (char )cp + 2, sizeof(*ts_val));
1779	* NTOHL(*ts_val);
1780	* memcpy((char ) ts_ecr, (char )cp + 6, sizeof(*ts_ecr));
1781	* NTOHL(*ts_ecr);
1782	*
1783	/ /
1784	* * A timestamp received in a SYN makes
1785	* * it ok to send timestamp requests and replies.
1786	* */
1787	/* if (ti->ti_flags & TH_SYN) {
1788	* tp->t_flags \|= TF_RCVD_TSTMP;
1789	* tp->ts_recent = *ts_val;
1790	* tp->ts_recent_age = tcp_now;
1791	* }
1792	*/ break;
1793	}
1794	}
1795	}
1796
1797
1798	/*
1799	* Pull out of band byte out of a segment so
1800	* it doesn't appear in the user's data queue.
1801	* It is still reflected in the segment length for
1802	* sequencing purposes.
1803	*/
1804
1805	#ifdef notdef
1806
1807	void
1808	tcp_pulloutofband(so, ti, m)
1809	struct socket *so;
1810	struct tcpiphdr *ti;
1811	register struct mbuf *m;
1812	{
1813	int cnt = ti->ti_urp - 1;
1814
1815	while (cnt >= 0) {
1816	if (m->m_len > cnt) {
1817	char *cp = mtod(m, caddr_t) + cnt;
1818	struct tcpcb *tp = sototcpcb(so);
1819
1820	tp->t_iobc = *cp;
1821	tp->t_oobflags \|= TCPOOB_HAVEDATA;
1822	memcpy(sp, cp+1, (unsigned)(m->m_len - cnt - 1));
1823	m->m_len--;
1824	return;
1825	}
1826	cnt -= m->m_len;
1827	m = m->m_next; /* XXX WRONG! Fix it! */
1828	if (m == 0)
1829	break;
1830	}
1831	panic("tcp_pulloutofband");
1832	}
1833
1834	#endif /* notdef */
1835
1836	/*
1837	* Collect new round-trip time estimate
1838	* and update averages and current timeout.
1839	*/
1840
1841	void
1842	#ifdef VBOX
1843	tcp_xmit_timer(PNATState pData, register struct tcpcb *tp, int rtt)
1844	#else /* !VBOX */
1845	tcp_xmit_timer(tp, rtt)
1846	register struct tcpcb *tp;
1847	int rtt;
1848	#endif /* !VBOX */
1849	{
1850	register short delta;
1851
1852	DEBUG_CALL("tcp_xmit_timer");
1853	DEBUG_ARG("tp = %lx", (long)tp);
1854	DEBUG_ARG("rtt = %d", rtt);
1855
1856	tcpstat.tcps_rttupdated++;
1857	if (tp->t_srtt != 0) {
1858	/*
1859	* srtt is stored as fixed point with 3 bits after the
1860	* binary point (i.e., scaled by 8). The following magic
1861	* is equivalent to the smoothing algorithm in rfc793 with
1862	* an alpha of .875 (srtt = rtt/8 + srtt*7/8 in fixed
1863	* point). Adjust rtt to origin 0.
1864	*/
1865	delta = rtt - 1 - (tp->t_srtt >> TCP_RTT_SHIFT);
1866	if ((tp->t_srtt += delta) <= 0)
1867	tp->t_srtt = 1;
1868	/*
1869	* We accumulate a smoothed rtt variance (actually, a
1870	* smoothed mean difference), then set the retransmit
1871	* timer to smoothed rtt + 4 times the smoothed variance.
1872	* rttvar is stored as fixed point with 2 bits after the
1873	* binary point (scaled by 4). The following is
1874	* equivalent to rfc793 smoothing with an alpha of .75
1875	* (rttvar = rttvar*3/4 + \|delta\| / 4). This replaces
1876	* rfc793's wired-in beta.
1877	*/
1878	if (delta < 0)
1879	delta = -delta;
1880	delta -= (tp->t_rttvar >> TCP_RTTVAR_SHIFT);
1881	if ((tp->t_rttvar += delta) <= 0)
1882	tp->t_rttvar = 1;
1883	} else {
1884	/*
1885	* No rtt measurement yet - use the unsmoothed rtt.
1886	* Set the variance to half the rtt (so our first
1887	* retransmit happens at 3*rtt).
1888	*/
1889	tp->t_srtt = rtt << TCP_RTT_SHIFT;
1890	tp->t_rttvar = rtt << (TCP_RTTVAR_SHIFT - 1);
1891	}
1892	tp->t_rtt = 0;
1893	tp->t_rxtshift = 0;
1894
1895	/*
1896	* the retransmit should happen at rtt + 4 * rttvar.
1897	* Because of the way we do the smoothing, srtt and rttvar
1898	* will each average +1/2 tick of bias. When we compute
1899	* the retransmit timer, we want 1/2 tick of rounding and
1900	* 1 extra tick because of +-1/2 tick uncertainty in the
1901	* firing of the timer. The bias will give us exactly the
1902	* 1.5 tick we need. But, because the bias is
1903	* statistical, we have to test that we don't drop below
1904	* the minimum feasible timer (which is 2 ticks).
1905	*/
1906	TCPT_RANGESET(tp->t_rxtcur, TCP_REXMTVAL(tp),
1907	(short)tp->t_rttmin, TCPTV_REXMTMAX); /* XXX */
1908
1909	/*
1910	* We received an ack for a packet that wasn't retransmitted;
1911	* it is probably safe to discard any error indications we've
1912	* received recently. This isn't quite right, but close enough
1913	* for now (a route might have failed after we sent a segment,
1914	* and the return path might not be symmetrical).
1915	*/
1916	tp->t_softerror = 0;
1917	}
1918
1919	/*
1920	* Determine a reasonable value for maxseg size.
1921	* If the route is known, check route for mtu.
1922	* If none, use an mss that can be handled on the outgoing
1923	* interface without forcing IP to fragment; if bigger than
1924	* an mbuf cluster (MCLBYTES), round down to nearest multiple of MCLBYTES
1925	* to utilize large mbufs. If no route is found, route has no mtu,
1926	* or the destination isn't local, use a default, hopefully conservative
1927	* size (usually 512 or the default IP max size, but no more than the mtu
1928	* of the interface), as we can't discover anything about intervening
1929	* gateways or networks. We also initialize the congestion/slow start
1930	* window to be a single segment if the destination isn't local.
1931	* While looking at the routing entry, we also initialize other path-dependent
1932	* parameters from pre-set or cached values in the routing entry.
1933	*/
1934
1935	int
1936	#ifdef VBOX
1937	tcp_mss(PNATState pData, register struct tcpcb *tp, u_int offer)
1938	#else /* !VBOX */
1939	tcp_mss(tp, offer)
1940	register struct tcpcb *tp;
1941	u_int offer;
1942	#endif /* !VBOX */
1943	{
1944	struct socket *so = tp->t_socket;
1945	int mss;
1946
1947	DEBUG_CALL("tcp_mss");
1948	DEBUG_ARG("tp = %lx", (long)tp);
1949	DEBUG_ARG("offer = %d", offer);
1950
1951	mss = min(if_mtu, if_mru) - sizeof(struct tcpiphdr);
1952	if (offer)
1953	mss = min(mss, offer);
1954	mss = max(mss, 32);
1955	if (mss < tp->t_maxseg \|\| offer != 0)
1956	tp->t_maxseg = mss;
1957
1958	tp->snd_cwnd = mss;
1959
1960	sbreserve(&so->so_snd, tcp_sndspace+((tcp_sndspace%mss)?(mss-(tcp_sndspace%mss)):0));
1961	sbreserve(&so->so_rcv, tcp_rcvspace+((tcp_rcvspace%mss)?(mss-(tcp_rcvspace%mss)):0));
1962
1963	DEBUG_MISC((dfd, " returning mss = %d\n", mss));
1964
1965	return mss;
1966	}

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

source: vbox/trunk/src/VBox/Devices/Network/slirp/tcp_input.c@ 1048

Download in other formats: