tcp_input.c@ 40623

Last change on this file since 40623 was 40423, checked in by vboxsync, 13 years ago
NAT: warnings [-Wunused-macros]
Property svn:eol-style set to `native` Property svn:keywords set to `Author Date Id Revision`
File size: 65.6 KB

Line
1	/* $Id: tcp_input.c 40423 2012-03-11 03:22:22Z vboxsync $ */
2	/** @file
3	* NAT - TCP input.
4	*/
5
6	/*
7	* Copyright (C) 2006-2010 Oracle Corporation
8	*
9	* This file is part of VirtualBox Open Source Edition (OSE), as
10	* available from http://www.virtualbox.org. This file is free software;
11	* you can redistribute it and/or modify it under the terms of the GNU
12	* General Public License (GPL) as published by the Free Software
13	* Foundation, in version 2 as it comes in the "COPYING" file of the
14	* VirtualBox OSE distribution. VirtualBox OSE is distributed in the
15	* hope that it will be useful, but WITHOUT ANY WARRANTY of any kind.
16	*/
17
18	/*
19	* This code is based on:
20	*
21	* Copyright (c) 1982, 1986, 1988, 1990, 1993, 1994
22	* The Regents of the University of California. All rights reserved.
23	*
24	* Redistribution and use in source and binary forms, with or without
25	* modification, are permitted provided that the following conditions
26	* are met:
27	* 1. Redistributions of source code must retain the above copyright
28	* notice, this list of conditions and the following disclaimer.
29	* 2. Redistributions in binary form must reproduce the above copyright
30	* notice, this list of conditions and the following disclaimer in the
31	* documentation and/or other materials provided with the distribution.
32	* 3. All advertising materials mentioning features or use of this software
33	* must display the following acknowledgement:
34	* This product includes software developed by the University of
35	* California, Berkeley and its contributors.
36	* 4. Neither the name of the University nor the names of its contributors
37	* may be used to endorse or promote products derived from this software
38	* without specific prior written permission.
39	*
40	* THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
41	* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
42	* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
43	* ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
44	* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
45	* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
46	* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
47	* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
48	* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
49	* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
50	* SUCH DAMAGE.
51	*
52	* @(#)tcp_input.c 8.5 (Berkeley) 4/10/94
53	* tcp_input.c,v 1.10 1994/10/13 18:36:32 wollman Exp
54	*/
55
56	/*
57	* Changes and additions relating to SLiRP
58	* Copyright (c) 1995 Danny Gasparovski.
59	*
60	* Please read the file COPYRIGHT for the
61	* terms and conditions of the copyright.
62	*/
63
64	#include <slirp.h>
65	#include "ip_icmp.h"
66
67
68	#if 0 /* code using this macroses is commented out */
69	# define TCP_PAWS_IDLE (24 * 24 * 60 * 60 * PR_SLOWHZ)
70
71	/* for modulo comparisons of timestamps */
72	# define TSTMP_LT(a, b) ((int)((a)-(b)) < 0)
73	# define TSTMP_GEQ(a, b) ((int)((a)-(b)) >= 0)
74	#endif
75
76	#ifndef TCP_ACK_HACK
77	#define DELAY_ACK(tp, ti) \
78	if (ti->ti_flags & TH_PUSH) \
79	tp->t_flags \|= TF_ACKNOW; \
80	else \
81	tp->t_flags \|= TF_DELACK;
82	#else /* !TCP_ACK_HACK */
83	#define DELAY_ACK(tp, ign) \
84	tp->t_flags \|= TF_DELACK;
85	#endif /* TCP_ACK_HACK */
86
87
88	/*
89	* deps: netinet/tcp_reass.c
90	* tcp_reass_maxqlen = 48 (deafault)
91	* tcp_reass_maxseg = nmbclusters/16 (nmbclusters = 1024 + maxusers * 64 from kern/kern_mbuf.c let's say 256)
92	*/
93	int
94	tcp_reass(PNATState pData, struct tcpcb tp, struct tcphdr th, int tlenp, struct mbuf m)
95	{
96	struct tseg_qent *q;
97	struct tseg_qent *p = NULL;
98	struct tseg_qent *nq;
99	struct tseg_qent *te = NULL;
100	struct socket *so = tp->t_socket;
101	int flags;
102	STAM_PROFILE_START(&pData->StatTCP_reassamble, tcp_reassamble);
103	LogFlowFunc(("ENTER: pData:%p, tp:%R[tcpcb793], th:%p, tlenp:%p, m:%p\n", pData, tp, th, tlenp, m));
104
105	/*
106	* XXX: tcp_reass() is rather inefficient with its data structures
107	* and should be rewritten (see NetBSD for optimizations). While
108	* doing that it should move to its own file tcp_reass.c.
109	*/
110
111	/*
112	* Call with th==NULL after become established to
113	* force pre-ESTABLISHED data up to user socket.
114	*/
115	if (th == NULL)
116	{
117	LogFlowFunc(("%d -> present\n", __LINE__));
118	goto present;
119	}
120
121	/*
122	* Limit the number of segments in the reassembly queue to prevent
123	* holding on to too many segments (and thus running out of mbufs).
124	* Make sure to let the missing segment through which caused this
125	* queue. Always keep one global queue entry spare to be able to
126	* process the missing segment.
127	*/
128	if ( th->th_seq != tp->rcv_nxt
129	&& ( tcp_reass_qsize + 1 >= tcp_reass_maxseg
130	\|\| tp->t_segqlen >= tcp_reass_maxqlen))
131	{
132	tcp_reass_overflows++;
133	tcpstat.tcps_rcvmemdrop++;
134	m_freem(pData, m);
135	*tlenp = 0;
136	STAM_PROFILE_STOP(&pData->StatTCP_reassamble, tcp_reassamble);
137	LogFlowFuncLeave();
138	return (0);
139	}
140
141	/*
142	* Allocate a new queue entry. If we can't, or hit the zone limit
143	* just drop the pkt.
144	*/
145	te = RTMemAlloc(sizeof(struct tseg_qent));
146	if (te == NULL)
147	{
148	tcpstat.tcps_rcvmemdrop++;
149	m_freem(pData, m);
150	*tlenp = 0;
151	STAM_PROFILE_STOP(&pData->StatTCP_reassamble, tcp_reassamble);
152	LogFlowFuncLeave();
153	return (0);
154	}
155	tp->t_segqlen++;
156	tcp_reass_qsize++;
157
158	/*
159	* Find a segment which begins after this one does.
160	*/
161	LIST_FOREACH(q, &tp->t_segq, tqe_q)
162	{
163	if (SEQ_GT(q->tqe_th->th_seq, th->th_seq))
164	break;
165	p = q;
166	}
167
168	/*
169	* If there is a preceding segment, it may provide some of
170	* our data already. If so, drop the data from the incoming
171	* segment. If it provides all of our data, drop us.
172	*/
173	if (p != NULL)
174	{
175	int i;
176	/* conversion to int (in i) handles seq wraparound */
177	i = p->tqe_th->th_seq + p->tqe_len - th->th_seq;
178	if (i > 0)
179	{
180	if (i >= *tlenp)
181	{
182	tcpstat.tcps_rcvduppack++;
183	tcpstat.tcps_rcvdupbyte += *tlenp;
184	m_freem(pData, m);
185	RTMemFree(te);
186	tp->t_segqlen--;
187	tcp_reass_qsize--;
188	/*
189	* Try to present any queued data
190	* at the left window edge to the user.
191	* This is needed after the 3-WHS
192	* completes.
193	*/
194	LogFlowFunc(("%d -> present\n", __LINE__));
195	goto present; /* ??? */
196	}
197	m_adj(m, i);
198	*tlenp -= i;
199	th->th_seq += i;
200	}
201	}
202	tcpstat.tcps_rcvoopack++;
203	tcpstat.tcps_rcvoobyte += *tlenp;
204
205	/*
206	* While we overlap succeeding segments trim them or,
207	* if they are completely covered, dequeue them.
208	*/
209	while (q)
210	{
211	int i = (th->th_seq + *tlenp) - q->tqe_th->th_seq;
212	if (i <= 0)
213	break;
214	if (i < q->tqe_len)
215	{
216	q->tqe_th->th_seq += i;
217	q->tqe_len -= i;
218	m_adj(q->tqe_m, i);
219	break;
220	}
221
222	nq = LIST_NEXT(q, tqe_q);
223	LIST_REMOVE(q, tqe_q);
224	m_freem(pData, q->tqe_m);
225	RTMemFree(q);
226	tp->t_segqlen--;
227	tcp_reass_qsize--;
228	q = nq;
229	}
230
231	/* Insert the new segment queue entry into place. */
232	te->tqe_m = m;
233	te->tqe_th = th;
234	te->tqe_len = *tlenp;
235
236	if (p == NULL)
237	{
238	LIST_INSERT_HEAD(&tp->t_segq, te, tqe_q);
239	}
240	else
241	{
242	LIST_INSERT_AFTER(p, te, tqe_q);
243	}
244
245	present:
246	/*
247	* Present data to user, advancing rcv_nxt through
248	* completed sequence space.
249	*/
250	if (!TCPS_HAVEESTABLISHED(tp->t_state))
251	{
252	STAM_PROFILE_STOP(&pData->StatTCP_reassamble, tcp_reassamble);
253	return (0);
254	}
255	q = LIST_FIRST(&tp->t_segq);
256	if (!q \|\| q->tqe_th->th_seq != tp->rcv_nxt)
257	{
258	STAM_PROFILE_STOP(&pData->StatTCP_reassamble, tcp_reassamble);
259	return (0);
260	}
261	do
262	{
263	tp->rcv_nxt += q->tqe_len;
264	flags = q->tqe_th->th_flags & TH_FIN;
265	nq = LIST_NEXT(q, tqe_q);
266	LIST_REMOVE(q, tqe_q);
267	/* XXX: This place should be checked for the same code in
268	* original BSD code for Slirp and current BSD used SS_FCANTRCVMORE
269	*/
270	if (so->so_state & SS_FCANTSENDMORE)
271	m_freem(pData, q->tqe_m);
272	else
273	sbappend(pData, so, q->tqe_m);
274	RTMemFree(q);
275	tp->t_segqlen--;
276	tcp_reass_qsize--;
277	q = nq;
278	}
279	while (q && q->tqe_th->th_seq == tp->rcv_nxt);
280
281	STAM_PROFILE_STOP(&pData->StatTCP_reassamble, tcp_reassamble);
282	return flags;
283	}
284
285	/*
286	* TCP input routine, follows pages 65-76 of the
287	* protocol specification dated September, 1981 very closely.
288	*/
289	void
290	tcp_input(PNATState pData, register struct mbuf m, int iphlen, struct socket inso)
291	{
292	struct ip save_ip, *ip;
293	register struct tcpiphdr *ti;
294	caddr_t optp = NULL;
295	int optlen = 0;
296	int len, tlen, off;
297	register struct tcpcb *tp = 0;
298	register int tiflags;
299	struct socket *so = 0;
300	int todrop, acked, ourfinisacked, needoutput = 0;
301	/* int dropsocket = 0; */
302	int iss = 0;
303	u_long tiwin;
304	/* int ts_present = 0; */
305	STAM_PROFILE_START(&pData->StatTCP_input, counter_input);
306
307	LogFlow(("tcp_input: m = %8lx, iphlen = %2d, inso = %R[natsock]\n",
308	(long)m, iphlen, inso));
309
310	if (inso != NULL)
311	{
312	QSOCKET_LOCK(tcb);
313	SOCKET_LOCK(inso);
314	QSOCKET_UNLOCK(tcb);
315	}
316	/*
317	* If called with m == 0, then we're continuing the connect
318	*/
319	if (m == NULL)
320	{
321	so = inso;
322	Log4(("NAT: tcp_input: %R[natsock]\n", so));
323	/* Re-set a few variables */
324	tp = sototcpcb(so);
325	m = so->so_m;
326
327	so->so_m = 0;
328	ti = so->so_ti;
329
330	/** @todo (vvl) clarify why it might happens */
331	if (ti == NULL)
332	{
333	LogRel(("NAT: ti is null. can't do any reseting connection actions\n"));
334	/* mbuf should be cleared in sofree called from tcp_close */
335	tcp_close(pData, tp);
336	STAM_PROFILE_STOP(&pData->StatTCP_input, counter_input);
337	LogFlowFuncLeave();
338	return;
339	}
340
341	tiwin = ti->ti_win;
342	tiflags = ti->ti_flags;
343
344	LogFlowFunc(("%d -> cont_conn\n", __LINE__));
345	goto cont_conn;
346	}
347
348	tcpstat.tcps_rcvtotal++;
349	/*
350	* Get IP and TCP header together in first mbuf.
351	* Note: IP leaves IP header in first mbuf.
352	*/
353	ti = mtod(m, struct tcpiphdr *);
354	if (iphlen > sizeof(struct ip))
355	{
356	ip_stripoptions(m, (struct mbuf *)0);
357	iphlen = sizeof(struct ip);
358	}
359	/* XXX Check if too short */
360
361
362	/*
363	* Save a copy of the IP header in case we want restore it
364	* for sending an ICMP error message in response.
365	*/
366	ip = mtod(m, struct ip *);
367	/*
368	* (vvl) ip_input substracts IP header length from ip->ip_len value.
369	* here we do the test the same as input method of UDP protocol.
370	*/
371	Assert((ip->ip_len + iphlen == m_length(m, NULL)));
372	save_ip = *ip;
373	save_ip.ip_len+= iphlen;
374
375	/*
376	* Checksum extended TCP header and data.
377	*/
378	tlen = ((struct ip *)ti)->ip_len;
379	memset(ti->ti_x1, 0, 9);
380	ti->ti_len = RT_H2N_U16((u_int16_t)tlen);
381	len = sizeof(struct ip) + tlen;
382	/* keep checksum for ICMP reply
383	* ti->ti_sum = cksum(m, len);
384	* if (ti->ti_sum) { */
385	if (cksum(m, len))
386	{
387	tcpstat.tcps_rcvbadsum++;
388	LogFlowFunc(("%d -> drop\n", __LINE__));
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)
398	\|\| off > tlen)
399	{
400	tcpstat.tcps_rcvbadoff++;
401	LogFlowFunc(("%d -> drop\n", __LINE__));
402	goto drop;
403	}
404	tlen -= off;
405	ti->ti_len = tlen;
406	if (off > sizeof (struct tcphdr))
407	{
408	optlen = off - sizeof (struct tcphdr);
409	optp = mtod(m, caddr_t) + sizeof (struct tcpiphdr);
410
411	/*
412	* Do quick retrieval of timestamp options ("options
413	* prediction?"). If timestamp is the only option and it's
414	* formatted as recommended in RFC 1323 appendix A, we
415	* quickly get the values now and not bother calling
416	* tcp_dooptions(), etc.
417	*/
418	#if 0
419	if (( optlen == TCPOLEN_TSTAMP_APPA
420	\|\| ( optlen > TCPOLEN_TSTAMP_APPA
421	&& optp[TCPOLEN_TSTAMP_APPA] == TCPOPT_EOL)) &&
422	(u_int32_t )optp == RT_H2N_U32_C(TCPOPT_TSTAMP_HDR) &&
423	(ti->ti_flags & TH_SYN) == 0)
424	{
425	ts_present = 1;
426	ts_val = RT_N2H_U32((u_int32_t )(optp + 4));
427	ts_ecr = RT_N2H_U32((u_int32_t )(optp + 8));
428	optp = NULL; / * we have parsed the options * /
429	}
430	#endif
431	}
432	tiflags = ti->ti_flags;
433
434	/*
435	* Convert TCP protocol specific fields to host format.
436	*/
437	NTOHL(ti->ti_seq);
438	NTOHL(ti->ti_ack);
439	NTOHS(ti->ti_win);
440	NTOHS(ti->ti_urp);
441
442	/*
443	* Drop TCP, IP headers and TCP options.
444	*/
445	m->m_data += sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
446	m->m_len -= sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
447
448	/*
449	* Locate pcb for segment.
450	*/
451	findso:
452	LogFlowFunc(("(enter) findso: %R[natsock]\n", so));
453	if (so != NULL && so != &tcb)
454	SOCKET_UNLOCK(so);
455	QSOCKET_LOCK(tcb);
456	so = tcp_last_so;
457	if ( so->so_fport != ti->ti_dport
458	\|\| so->so_lport != ti->ti_sport
459	\|\| so->so_laddr.s_addr != ti->ti_src.s_addr
460	\|\| so->so_faddr.s_addr != ti->ti_dst.s_addr)
461	{
462	#ifdef VBOX_WITH_SLIRP_MT
463	struct socket *sonxt;
464	#endif
465	QSOCKET_UNLOCK(tcb);
466	/* @todo fix SOLOOKUP macrodefinition to be usable here */
467	#ifndef VBOX_WITH_SLIRP_MT
468	so = solookup(&tcb, ti->ti_src, ti->ti_sport,
469	ti->ti_dst, ti->ti_dport);
470	#else
471	so = NULL;
472	QSOCKET_FOREACH(so, sonxt, tcp)
473	/* { */
474	if ( so->so_lport == ti->ti_sport
475	&& so->so_laddr.s_addr == ti->ti_src.s_addr
476	&& so->so_faddr.s_addr == ti->ti_dst.s_addr
477	&& so->so_fport == ti->ti_dport
478	&& so->so_deleted != 1)
479	{
480	break; /* so is locked here */
481	}
482	LOOP_LABEL(tcp, so, sonxt);
483	}
484	if (so == &tcb) {
485	so = NULL;
486	}
487	#endif
488	if (so)
489	{
490	tcp_last_so = so;
491	}
492	++tcpstat.tcps_socachemiss;
493	}
494	else
495	{
496	SOCKET_LOCK(so);
497	QSOCKET_UNLOCK(tcb);
498	}
499	LogFlowFunc(("(leave) findso: %R[natsock]\n", so));
500
501	/*
502	* If the state is CLOSED (i.e., TCB does not exist) then
503	* all data in the incoming segment is discarded.
504	* If the TCB exists but is in CLOSED state, it is embryonic,
505	* but should either do a listen or a connect soon.
506	*
507	* state == CLOSED means we've done socreate() but haven't
508	* attached it to a protocol yet...
509	*
510	* XXX If a TCB does not exist, and the TH_SYN flag is
511	* the only flag set, then create a session, mark it
512	* as if it was LISTENING, and continue...
513	*/
514	if (so == 0)
515	{
516	if ((tiflags & (TH_SYN\|TH_FIN\|TH_RST\|TH_URG\|TH_ACK)) != TH_SYN)
517	{
518	LogFlowFunc(("%d -> dropwithreset\n", __LINE__));
519	goto dropwithreset;
520	}
521
522	if ((so = socreate()) == NULL)
523	{
524	LogFlowFunc(("%d -> dropwithreset\n", __LINE__));
525	goto dropwithreset;
526	}
527	if (tcp_attach(pData, so) < 0)
528	{
529	RTMemFree(so); /* Not sofree (if it failed, it's not insqued) */
530	LogFlowFunc(("%d -> dropwithreset\n", __LINE__));
531	goto dropwithreset;
532	}
533	SOCKET_LOCK(so);
534	#ifndef VBOX_WITH_SLIRP_BSD_SBUF
535	sbreserve(pData, &so->so_snd, tcp_sndspace);
536	sbreserve(pData, &so->so_rcv, tcp_rcvspace);
537	#else
538	sbuf_new(&so->so_snd, NULL, tcp_sndspace, SBUF_AUTOEXTEND);
539	sbuf_new(&so->so_rcv, NULL, tcp_rcvspace, SBUF_AUTOEXTEND);
540	#endif
541
542	/* tcp_last_so = so; / / XXX ? */
543	/* tp = sototcpcb(so); */
544
545	so->so_laddr = ti->ti_src;
546	so->so_lport = ti->ti_sport;
547	so->so_faddr = ti->ti_dst;
548	so->so_fport = ti->ti_dport;
549
550	so->so_iptos = ((struct ip *)ti)->ip_tos;
551
552	tp = sototcpcb(so);
553	TCP_STATE_SWITCH_TO(tp, TCPS_LISTEN);
554	}
555
556	/*
557	* If this is a still-connecting socket, this probably
558	* a retransmit of the SYN. Whether it's a retransmit SYN
559	* or something else, we nuke it.
560	*/
561	if (so->so_state & SS_ISFCONNECTING)
562	{
563	LogFlowFunc(("%d -> drop\n", __LINE__));
564	goto drop;
565	}
566
567	tp = sototcpcb(so);
568
569	/* XXX Should never fail */
570	if (tp == 0)
571	{
572	LogFlowFunc(("%d -> dropwithreset\n", __LINE__));
573	goto dropwithreset;
574	}
575	if (tp->t_state == TCPS_CLOSED)
576	{
577	LogFlowFunc(("%d -> drop\n", __LINE__));
578	goto drop;
579	}
580
581	/* Unscale the window into a 32-bit value. */
582	/* if ((tiflags & TH_SYN) == 0)
583	* tiwin = ti->ti_win << tp->snd_scale;
584	* else
585	*/
586	tiwin = ti->ti_win;
587
588	/*
589	* Segment received on connection.
590	* Reset idle time and keep-alive timer.
591	*/
592	tp->t_idle = 0;
593	if (so_options)
594	tp->t_timer[TCPT_KEEP] = tcp_keepintvl;
595	else
596	tp->t_timer[TCPT_KEEP] = tcp_keepidle;
597
598	/*
599	* Process options if not in LISTEN state,
600	* else do it below (after getting remote address).
601	*/
602	if (optp && tp->t_state != TCPS_LISTEN)
603	tcp_dooptions(pData, tp, (u_char *)optp, optlen, ti);
604	/* , */
605	/* &ts_present, &ts_val, &ts_ecr); */
606
607	/*
608	* Header prediction: check for the two common cases
609	* of a uni-directional data xfer. If the packet has
610	* no control flags, is in-sequence, the window didn't
611	* change and we're not retransmitting, it's a
612	* candidate. If the length is zero and the ack moved
613	* forward, we're the sender side of the xfer. Just
614	* free the data acked & wake any higher level process
615	* that was blocked waiting for space. If the length
616	* is non-zero and the ack didn't move, we're the
617	* receiver side. If we're getting packets in-order
618	* (the reassembly queue is empty), add the data to
619	* the socket buffer and note that we need a delayed ack.
620	*
621	* XXX Some of these tests are not needed
622	* eg: the tiwin == tp->snd_wnd prevents many more
623	* predictions.. with no real advantage..
624	*/
625	if ( tp->t_state == TCPS_ESTABLISHED
626	&& (tiflags & (TH_SYN\|TH_FIN\|TH_RST\|TH_URG\|TH_ACK)) == TH_ACK
627	/* && (!ts_present \|\| TSTMP_GEQ(ts_val, tp->ts_recent)) */
628	&& ti->ti_seq == tp->rcv_nxt
629	&& tiwin && tiwin == tp->snd_wnd
630	&& tp->snd_nxt == tp->snd_max)
631	{
632	/*
633	* If last ACK falls within this segment's sequence numbers,
634	* record the timestamp.
635	*/
636	#if 0
637	if (ts_present && SEQ_LEQ(ti->ti_seq, tp->last_ack_sent) &&
638	SEQ_LT(tp->last_ack_sent, ti->ti_seq + ti->ti_len))
639	{
640	tp->ts_recent_age = tcp_now;
641	tp->ts_recent = ts_val;
642	}
643	#endif
644
645	if (ti->ti_len == 0)
646	{
647	if ( SEQ_GT(ti->ti_ack, tp->snd_una)
648	&& SEQ_LEQ(ti->ti_ack, tp->snd_max)
649	&& tp->snd_cwnd >= tp->snd_wnd)
650	{
651	/*
652	* this is a pure ack for outstanding data.
653	*/
654	++tcpstat.tcps_predack;
655	#if 0
656	if (ts_present)
657	tcp_xmit_timer(tp, tcp_now-ts_ecr+1);
658	else
659	#endif
660	if ( tp->t_rtt
661	&& SEQ_GT(ti->ti_ack, tp->t_rtseq))
662	tcp_xmit_timer(pData, tp, tp->t_rtt);
663	acked = ti->ti_ack - tp->snd_una;
664	tcpstat.tcps_rcvackpack++;
665	tcpstat.tcps_rcvackbyte += acked;
666	#ifndef VBOX_WITH_SLIRP_BSD_SBUF
667	sbdrop(&so->so_snd, acked);
668	#else
669	if (sbuf_len(&so->so_snd) < acked)
670	/* drop all what sbuf have */
671	sbuf_setpos(&so->so_snd, 0);
672	else
673	sbuf_setpos(&so->so_snd, sbuf_len(&so->so_snd) - acked);
674	#endif
675	tp->snd_una = ti->ti_ack;
676	m_freem(pData, m);
677
678	/*
679	* If all outstanding data are acked, stop
680	* retransmit timer, otherwise restart timer
681	* using current (possibly backed-off) value.
682	* If process is waiting for space,
683	* wakeup/selwakeup/signal. If data
684	* are ready to send, let tcp_output
685	* decide between more output or persist.
686	*/
687	if (tp->snd_una == tp->snd_max)
688	tp->t_timer[TCPT_REXMT] = 0;
689	else if (tp->t_timer[TCPT_PERSIST] == 0)
690	tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
691
692	/*
693	* There's room in so_snd, sowwakup will read()
694	* from the socket if we can
695	*/
696	#if 0
697	if (so->so_snd.sb_flags & SB_NOTIFY)
698	sowwakeup(so);
699	#endif
700	/*
701	* This is called because sowwakeup might have
702	* put data into so_snd. Since we don't so sowwakeup,
703	* we don't need this.. XXX???
704	*/
705	if (SBUF_LEN(&so->so_snd))
706	(void) tcp_output(pData, tp);
707
708	SOCKET_UNLOCK(so);
709	STAM_PROFILE_STOP(&pData->StatTCP_input, counter_input);
710	return;
711	}
712	}
713	else if ( ti->ti_ack == tp->snd_una
714	&& LIST_FIRST(&tp->t_segq)
715	&& ti->ti_len <= sbspace(&so->so_rcv))
716	{
717	/*
718	* this is a pure, in-sequence data packet
719	* with nothing on the reassembly queue and
720	* we have enough buffer space to take it.
721	*/
722	++tcpstat.tcps_preddat;
723	tp->rcv_nxt += ti->ti_len;
724	tcpstat.tcps_rcvpack++;
725	tcpstat.tcps_rcvbyte += ti->ti_len;
726	/*
727	* Add data to socket buffer.
728	*/
729	sbappend(pData, so, m);
730
731	/*
732	* XXX This is called when data arrives. Later, check
733	* if we can actually write() to the socket
734	* XXX Need to check? It's be NON_BLOCKING
735	*/
736	/* sorwakeup(so); */
737
738	/*
739	* If this is a short packet, then ACK now - with Nagel
740	* congestion avoidance sender won't send more until
741	* he gets an ACK.
742	*
743	* It is better to not delay acks at all to maximize
744	* TCP throughput. See RFC 2581.
745	*/
746	tp->t_flags \|= TF_ACKNOW;
747	tcp_output(pData, tp);
748	SOCKET_UNLOCK(so);
749	STAM_PROFILE_STOP(&pData->StatTCP_input, counter_input);
750	return;
751	}
752	} /* header prediction */
753	/*
754	* Calculate amount of space in receive window,
755	* and then do TCP input processing.
756	* Receive window is amount of space in rcv queue,
757	* but not less than advertised window.
758	*/
759	{
760	int win;
761	win = sbspace(&so->so_rcv);
762	if (win < 0)
763	win = 0;
764	tp->rcv_wnd = max(win, (int)(tp->rcv_adv - tp->rcv_nxt));
765	}
766
767	switch (tp->t_state)
768	{
769	/*
770	* If the state is LISTEN then ignore segment if it contains an RST.
771	* If the segment contains an ACK then it is bad and send a RST.
772	* If it does not contain a SYN then it is not interesting; drop it.
773	* Don't bother responding if the destination was a broadcast.
774	* Otherwise initialize tp->rcv_nxt, and tp->irs, select an initial
775	* tp->iss, and send a segment:
776	* <SEQ=ISS><ACK=RCV_NXT><CTL=SYN,ACK>
777	* Also initialize tp->snd_nxt to tp->iss+1 and tp->snd_una to tp->iss.
778	* Fill in remote peer address fields if not previously specified.
779	* Enter SYN_RECEIVED state, and process any other fields of this
780	* segment in this state.
781	*/
782	case TCPS_LISTEN:
783	{
784	if (tiflags & TH_RST)
785	{
786	LogFlowFunc(("%d -> drop\n", __LINE__));
787	goto drop;
788	}
789	if (tiflags & TH_ACK)
790	{
791	LogFlowFunc(("%d -> dropwithreset\n", __LINE__));
792	goto dropwithreset;
793	}
794	if ((tiflags & TH_SYN) == 0)
795	{
796	LogFlowFunc(("%d -> drop\n", __LINE__));
797	goto drop;
798	}
799
800	/*
801	* This has way too many gotos...
802	* But a bit of spaghetti code never hurt anybody :)
803	*/
804	if ( (tcp_fconnect(pData, so) == -1)
805	&& errno != EINPROGRESS
806	&& errno != EWOULDBLOCK)
807	{
808	u_char code = ICMP_UNREACH_NET;
809	Log2((" tcp fconnect errno = %d (%s)\n", errno, strerror(errno)));
810	if (errno == ECONNREFUSED)
811	{
812	/* ACK the SYN, send RST to refuse the connection */
813	tcp_respond(pData, tp, ti, m, ti->ti_seq+1, (tcp_seq)0,
814	TH_RST\|TH_ACK);
815	}
816	else
817	{
818	if (errno == EHOSTUNREACH)
819	code = ICMP_UNREACH_HOST;
820	HTONL(ti->ti_seq); /* restore tcp header */
821	HTONL(ti->ti_ack);
822	HTONS(ti->ti_win);
823	HTONS(ti->ti_urp);
824	m->m_data -= sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
825	m->m_len += sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
826	*ip = save_ip;
827	icmp_error(pData, m, ICMP_UNREACH, code, 0, strerror(errno));
828	tp->t_socket->so_m = NULL;
829	}
830	tp = tcp_close(pData, tp);
831	}
832	else
833	{
834	/*
835	* Haven't connected yet, save the current mbuf
836	* and ti, and return
837	* XXX Some OS's don't tell us whether the connect()
838	* succeeded or not. So we must time it out.
839	*/
840	so->so_m = m;
841	so->so_ti = ti;
842	tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT;
843	TCP_STATE_SWITCH_TO(tp, TCPS_SYN_RECEIVED);
844	}
845	SOCKET_UNLOCK(so);
846	STAM_PROFILE_STOP(&pData->StatTCP_input, counter_input);
847	LogFlowFuncLeave();
848	return;
849
850	cont_conn:
851	/* m==NULL
852	* Check if the connect succeeded
853	*/
854	LogFlowFunc(("cont_conn:\n"));
855	if (so->so_state & SS_NOFDREF)
856	{
857	tp = tcp_close(pData, tp);
858	LogFlowFunc(("%d -> dropwithreset\n", __LINE__));
859	goto dropwithreset;
860	}
861
862	tcp_template(tp);
863
864	if (optp)
865	tcp_dooptions(pData, tp, (u_char *)optp, optlen, ti);
866
867	if (iss)
868	tp->iss = iss;
869	else
870	tp->iss = tcp_iss;
871	tcp_iss += TCP_ISSINCR/2;
872	tp->irs = ti->ti_seq;
873	tcp_sendseqinit(tp);
874	tcp_rcvseqinit(tp);
875	tp->t_flags \|= TF_ACKNOW;
876	TCP_STATE_SWITCH_TO(tp, TCPS_SYN_RECEIVED);
877	tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT;
878	tcpstat.tcps_accepts++;
879	LogFlowFunc(("%d -> trimthenstep6\n", __LINE__));
880	goto trimthenstep6;
881	} /* case TCPS_LISTEN */
882
883	/*
884	* If the state is SYN_SENT:
885	* if seg contains an ACK, but not for our SYN, drop the input.
886	* if seg contains a RST, then drop the connection.
887	* if seg does not contain SYN, then drop it.
888	* Otherwise this is an acceptable SYN segment
889	* initialize tp->rcv_nxt and tp->irs
890	* if seg contains ack then advance tp->snd_una
891	* if SYN has been acked change to ESTABLISHED else SYN_RCVD state
892	* arrange for segment to be acked (eventually)
893	* continue processing rest of data/controls, beginning with URG
894	*/
895	case TCPS_SYN_SENT:
896	if ( (tiflags & TH_ACK)
897	&& ( SEQ_LEQ(ti->ti_ack, tp->iss)
898	\|\| SEQ_GT(ti->ti_ack, tp->snd_max)))
899	{
900	LogFlowFunc(("%d -> dropwithreset\n", __LINE__));
901	goto dropwithreset;
902	}
903
904	if (tiflags & TH_RST)
905	{
906	if (tiflags & TH_ACK)
907	tp = tcp_drop(pData, tp, 0); /* XXX Check t_softerror! */
908	LogFlowFunc(("%d -> drop\n", __LINE__));
909	goto drop;
910	}
911
912	if ((tiflags & TH_SYN) == 0)
913	{
914	LogFlowFunc(("%d -> drop\n", __LINE__));
915	goto drop;
916	}
917	if (tiflags & TH_ACK)
918	{
919	tp->snd_una = ti->ti_ack;
920	if (SEQ_LT(tp->snd_nxt, tp->snd_una))
921	tp->snd_nxt = tp->snd_una;
922	}
923
924	tp->t_timer[TCPT_REXMT] = 0;
925	tp->irs = ti->ti_seq;
926	tcp_rcvseqinit(tp);
927	tp->t_flags \|= TF_ACKNOW;
928	if (tiflags & TH_ACK && SEQ_GT(tp->snd_una, tp->iss))
929	{
930	tcpstat.tcps_connects++;
931	soisfconnected(so);
932	TCP_STATE_SWITCH_TO(tp, TCPS_ESTABLISHED);
933
934	/* Do window scaling on this connection? */
935	#if 0
936	if (( tp->t_flags & (TF_RCVD_SCALE\|TF_REQ_SCALE))
937	== (TF_RCVD_SCALE\|TF_REQ_SCALE))
938	{
939	tp->snd_scale = tp->requested_s_scale;
940	tp->rcv_scale = tp->request_r_scale;
941	}
942	#endif
943	(void) tcp_reass(pData, tp, (struct tcphdr )0, NULL, (struct mbuf )0);
944	/*
945	* if we didn't have to retransmit the SYN,
946	* use its rtt as our initial srtt & rtt var.
947	*/
948	if (tp->t_rtt)
949	tcp_xmit_timer(pData, tp, tp->t_rtt);
950	}
951	else
952	TCP_STATE_SWITCH_TO(tp, TCPS_SYN_RECEIVED);
953
954	trimthenstep6:
955	LogFlowFunc(("trimthenstep6:\n"));
956	/*
957	* Advance ti->ti_seq to correspond to first data byte.
958	* If data, trim to stay within window,
959	* dropping FIN if necessary.
960	*/
961	ti->ti_seq++;
962	if (ti->ti_len > tp->rcv_wnd)
963	{
964	todrop = ti->ti_len - tp->rcv_wnd;
965	m_adj(m, -todrop);
966	ti->ti_len = tp->rcv_wnd;
967	tiflags &= ~TH_FIN;
968	tcpstat.tcps_rcvpackafterwin++;
969	tcpstat.tcps_rcvbyteafterwin += todrop;
970	}
971	tp->snd_wl1 = ti->ti_seq - 1;
972	tp->rcv_up = ti->ti_seq;
973	LogFlowFunc(("%d -> step6\n", __LINE__));
974	goto step6;
975	} /* switch tp->t_state */
976	/*
977	* States other than LISTEN or SYN_SENT.
978	* First check timestamp, if present.
979	* Then check that at least some bytes of segment are within
980	* receive window. If segment begins before rcv_nxt,
981	* drop leading data (and SYN); if nothing left, just ack.
982	*
983	* RFC 1323 PAWS: If we have a timestamp reply on this segment
984	* and it's less than ts_recent, drop it.
985	*/
986	#if 0
987	if ( ts_present
988	&& (tiflags & TH_RST) == 0
989	&& tp->ts_recent
990	&& TSTMP_LT(ts_val, tp->ts_recent))
991	{
992	/* Check to see if ts_recent is over 24 days old. */
993	if ((int)(tcp_now - tp->ts_recent_age) > TCP_PAWS_IDLE)
994	{
995	/*
996	* Invalidate ts_recent. If this segment updates
997	* ts_recent, the age will be reset later and ts_recent
998	* will get a valid value. If it does not, setting
999	* ts_recent to zero will at least satisfy the
1000	* requirement that zero be placed in the timestamp
1001	* echo reply when ts_recent isn't valid. The
1002	* age isn't reset until we get a valid ts_recent
1003	* because we don't want out-of-order segments to be
1004	* dropped when ts_recent is old.
1005	*/
1006	tp->ts_recent = 0;
1007	}
1008	else
1009	{
1010	tcpstat.tcps_rcvduppack++;
1011	tcpstat.tcps_rcvdupbyte += ti->ti_len;
1012	tcpstat.tcps_pawsdrop++;
1013	goto dropafterack;
1014	}
1015	}
1016	#endif
1017
1018	todrop = tp->rcv_nxt - ti->ti_seq;
1019	if (todrop > 0)
1020	{
1021	if (tiflags & TH_SYN)
1022	{
1023	tiflags &= ~TH_SYN;
1024	ti->ti_seq++;
1025	if (ti->ti_urp > 1)
1026	ti->ti_urp--;
1027	else
1028	tiflags &= ~TH_URG;
1029	todrop--;
1030	}
1031	/*
1032	* Following if statement from Stevens, vol. 2, p. 960.
1033	*/
1034	if ( todrop > ti->ti_len
1035	\|\| ( todrop == ti->ti_len
1036	&& (tiflags & TH_FIN) == 0))
1037	{
1038	/*
1039	* Any valid FIN must be to the left of the window.
1040	* At this point the FIN must be a duplicate or out
1041	* of sequence; drop it.
1042	*/
1043	tiflags &= ~TH_FIN;
1044
1045	/*
1046	* Send an ACK to resynchronize and drop any data.
1047	* But keep on processing for RST or ACK.
1048	*/
1049	tp->t_flags \|= TF_ACKNOW;
1050	todrop = ti->ti_len;
1051	tcpstat.tcps_rcvduppack++;
1052	tcpstat.tcps_rcvdupbyte += todrop;
1053	}
1054	else
1055	{
1056	tcpstat.tcps_rcvpartduppack++;
1057	tcpstat.tcps_rcvpartdupbyte += todrop;
1058	}
1059	m_adj(m, todrop);
1060	ti->ti_seq += todrop;
1061	ti->ti_len -= todrop;
1062	if (ti->ti_urp > todrop)
1063	ti->ti_urp -= todrop;
1064	else
1065	{
1066	tiflags &= ~TH_URG;
1067	ti->ti_urp = 0;
1068	}
1069	}
1070	/*
1071	* If new data are received on a connection after the
1072	* user processes are gone, then RST the other end.
1073	*/
1074	if ( (so->so_state & SS_NOFDREF)
1075	&& tp->t_state > TCPS_CLOSE_WAIT && ti->ti_len)
1076	{
1077	tp = tcp_close(pData, tp);
1078	tcpstat.tcps_rcvafterclose++;
1079	LogFlowFunc(("%d -> dropwithreset\n", __LINE__));
1080	goto dropwithreset;
1081	}
1082
1083	/*
1084	* If segment ends after window, drop trailing data
1085	* (and PUSH and FIN); if nothing left, just ACK.
1086	*/
1087	todrop = (ti->ti_seq+ti->ti_len) - (tp->rcv_nxt+tp->rcv_wnd);
1088	if (todrop > 0)
1089	{
1090	tcpstat.tcps_rcvpackafterwin++;
1091	if (todrop >= ti->ti_len)
1092	{
1093	tcpstat.tcps_rcvbyteafterwin += ti->ti_len;
1094	/*
1095	* If a new connection request is received
1096	* while in TIME_WAIT, drop the old connection
1097	* and start over if the sequence numbers
1098	* are above the previous ones.
1099	*/
1100	if ( tiflags & TH_SYN
1101	&& tp->t_state == TCPS_TIME_WAIT
1102	&& SEQ_GT(ti->ti_seq, tp->rcv_nxt))
1103	{
1104	iss = tp->rcv_nxt + TCP_ISSINCR;
1105	tp = tcp_close(pData, tp);
1106	SOCKET_UNLOCK(tp->t_socket);
1107	LogFlowFunc(("%d -> findso\n", __LINE__));
1108	goto findso;
1109	}
1110	/*
1111	* If window is closed can only take segments at
1112	* window edge, and have to drop data and PUSH from
1113	* incoming segments. Continue processing, but
1114	* remember to ack. Otherwise, drop segment
1115	* and ack.
1116	*/
1117	if (tp->rcv_wnd == 0 && ti->ti_seq == tp->rcv_nxt)
1118	{
1119	tp->t_flags \|= TF_ACKNOW;
1120	tcpstat.tcps_rcvwinprobe++;
1121	}
1122	else
1123	{
1124	LogFlowFunc(("%d -> dropafterack\n", __LINE__));
1125	goto dropafterack;
1126	}
1127	}
1128	else
1129	tcpstat.tcps_rcvbyteafterwin += todrop;
1130	m_adj(m, -todrop);
1131	ti->ti_len -= todrop;
1132	tiflags &= ~(TH_PUSH\|TH_FIN);
1133	}
1134
1135	/*
1136	* If last ACK falls within this segment's sequence numbers,
1137	* record its timestamp.
1138	*/
1139	#if 0
1140	if ( ts_present
1141	&& SEQ_LEQ(ti->ti_seq, tp->last_ack_sent)
1142	&& SEQ_LT(tp->last_ack_sent, ti->ti_seq + ti->ti_len + ((tiflags & (TH_SYN\|TH_FIN)) != 0)))
1143	{
1144	tp->ts_recent_age = tcp_now;
1145	tp->ts_recent = ts_val;
1146	}
1147	#endif
1148
1149	/*
1150	* If the RST bit is set examine the state:
1151	* SYN_RECEIVED STATE:
1152	* If passive open, return to LISTEN state.
1153	* If active open, inform user that connection was refused.
1154	* ESTABLISHED, FIN_WAIT_1, FIN_WAIT2, CLOSE_WAIT STATES:
1155	* Inform user that connection was reset, and close tcb.
1156	* CLOSING, LAST_ACK, TIME_WAIT STATES
1157	* Close the tcb.
1158	*/
1159	if (tiflags&TH_RST)
1160	switch (tp->t_state)
1161	{
1162	case TCPS_SYN_RECEIVED:
1163	/* so->so_error = ECONNREFUSED; */
1164	LogFlowFunc(("%d -> close\n", __LINE__));
1165	goto close;
1166
1167	case TCPS_ESTABLISHED:
1168	case TCPS_FIN_WAIT_1:
1169	case TCPS_FIN_WAIT_2:
1170	case TCPS_CLOSE_WAIT:
1171	/* so->so_error = ECONNRESET; */
1172	close:
1173	LogFlowFunc(("close:\n"));
1174	TCP_STATE_SWITCH_TO(tp, TCPS_CLOSED);
1175	tcpstat.tcps_drops++;
1176	tp = tcp_close(pData, tp);
1177	LogFlowFunc(("%d -> drop\n", __LINE__));
1178	goto drop;
1179
1180	case TCPS_CLOSING:
1181	case TCPS_LAST_ACK:
1182	case TCPS_TIME_WAIT:
1183	tp = tcp_close(pData, tp);
1184	LogFlowFunc(("%d -> drop\n", __LINE__));
1185	goto drop;
1186	}
1187
1188	/*
1189	* If a SYN is in the window, then this is an
1190	* error and we send an RST and drop the connection.
1191	*/
1192	if (tiflags & TH_SYN)
1193	{
1194	tp = tcp_drop(pData, tp, 0);
1195	LogFlowFunc(("%d -> dropwithreset\n", __LINE__));
1196	goto dropwithreset;
1197	}
1198
1199	/*
1200	* If the ACK bit is off we drop the segment and return.
1201	*/
1202	if ((tiflags & TH_ACK) == 0)
1203	{
1204	LogFlowFunc(("%d -> drop\n", __LINE__));
1205	goto drop;
1206	}
1207
1208	/*
1209	* Ack processing.
1210	*/
1211	switch (tp->t_state)
1212	{
1213	/*
1214	* In SYN_RECEIVED state if the ack ACKs our SYN then enter
1215	* ESTABLISHED state and continue processing, otherwise
1216	* send an RST. una<=ack<=max
1217	*/
1218	case TCPS_SYN_RECEIVED:
1219	LogFlowFunc(("%d -> TCPS_SYN_RECEIVED\n", __LINE__));
1220	if ( SEQ_GT(tp->snd_una, ti->ti_ack)
1221	\|\| SEQ_GT(ti->ti_ack, tp->snd_max))
1222	goto dropwithreset;
1223	tcpstat.tcps_connects++;
1224	TCP_STATE_SWITCH_TO(tp, TCPS_ESTABLISHED);
1225	/*
1226	* The sent SYN is ack'ed with our sequence number +1
1227	* The first data byte already in the buffer will get
1228	* lost if no correction is made. This is only needed for
1229	* SS_CTL since the buffer is empty otherwise.
1230	* tp->snd_una++; or:
1231	*/
1232	tp->snd_una = ti->ti_ack;
1233	soisfconnected(so);
1234
1235	/* Do window scaling? */
1236	#if 0
1237	if ( (tp->t_flags & (TF_RCVD_SCALE\|TF_REQ_SCALE))
1238	== (TF_RCVD_SCALE\|TF_REQ_SCALE))
1239	{
1240	tp->snd_scale = tp->requested_s_scale;
1241	tp->rcv_scale = tp->request_r_scale;
1242	}
1243	#endif
1244	(void) tcp_reass(pData, tp, (struct tcphdr )0, (int )0, (struct mbuf *)0);
1245	tp->snd_wl1 = ti->ti_seq - 1;
1246	/* Avoid ack processing; snd_una==ti_ack => dup ack */
1247	LogFlowFunc(("%d -> synrx_to_est\n", __LINE__));
1248	goto synrx_to_est;
1249	/* fall into ... */
1250
1251	/*
1252	* In ESTABLISHED state: drop duplicate ACKs; ACK out of range
1253	* ACKs. If the ack is in the range
1254	* tp->snd_una < ti->ti_ack <= tp->snd_max
1255	* then advance tp->snd_una to ti->ti_ack and drop
1256	* data from the retransmission queue. If this ACK reflects
1257	* more up to date window information we update our window information.
1258	*/
1259	case TCPS_ESTABLISHED:
1260	case TCPS_FIN_WAIT_1:
1261	case TCPS_FIN_WAIT_2:
1262	case TCPS_CLOSE_WAIT:
1263	case TCPS_CLOSING:
1264	case TCPS_LAST_ACK:
1265	case TCPS_TIME_WAIT:
1266	LogFlowFunc(("%d -> TCPS_ESTABLISHED\|TCPS_FIN_WAIT_1\|TCPS_FIN_WAIT_2\|TCPS_CLOSE_WAIT\|"
1267	"TCPS_CLOSING\|TCPS_LAST_ACK\|TCPS_TIME_WAIT\n", __LINE__));
1268	if (SEQ_LEQ(ti->ti_ack, tp->snd_una))
1269	{
1270	if (ti->ti_len == 0 && tiwin == tp->snd_wnd)
1271	{
1272	tcpstat.tcps_rcvdupack++;
1273	Log2((" dup ack m = %lx, so = %lx\n", (long)m, (long)so));
1274	/*
1275	* If we have outstanding data (other than
1276	* a window probe), this is a completely
1277	* duplicate ack (ie, window info didn't
1278	* change), the ack is the biggest we've
1279	* seen and we've seen exactly our rexmt
1280	* threshold of them, assume a packet
1281	* has been dropped and retransmit it.
1282	* Kludge snd_nxt & the congestion
1283	* window so we send only this one
1284	* packet.
1285	*
1286	* We know we're losing at the current
1287	* window size so do congestion avoidance
1288	* (set ssthresh to half the current window
1289	* and pull our congestion window back to
1290	* the new ssthresh).
1291	*
1292	* Dup acks mean that packets have left the
1293	* network (they're now cached at the receiver)
1294	* so bump cwnd by the amount in the receiver
1295	* to keep a constant cwnd packets in the
1296	* network.
1297	*/
1298	if ( tp->t_timer[TCPT_REXMT] == 0
1299	\|\| ti->ti_ack != tp->snd_una)
1300	tp->t_dupacks = 0;
1301	else if (++tp->t_dupacks == tcprexmtthresh)
1302	{
1303	tcp_seq onxt = tp->snd_nxt;
1304	u_int win = min(tp->snd_wnd, tp->snd_cwnd) / 2 / tp->t_maxseg;
1305	if (win < 2)
1306	win = 2;
1307	tp->snd_ssthresh = win * tp->t_maxseg;
1308	tp->t_timer[TCPT_REXMT] = 0;
1309	tp->t_rtt = 0;
1310	tp->snd_nxt = ti->ti_ack;
1311	tp->snd_cwnd = tp->t_maxseg;
1312	(void) tcp_output(pData, tp);
1313	tp->snd_cwnd = tp->snd_ssthresh +
1314	tp->t_maxseg * tp->t_dupacks;
1315	if (SEQ_GT(onxt, tp->snd_nxt))
1316	tp->snd_nxt = onxt;
1317	LogFlowFunc(("%d -> drop\n", __LINE__));
1318	goto drop;
1319	}
1320	else if (tp->t_dupacks > tcprexmtthresh)
1321	{
1322	tp->snd_cwnd += tp->t_maxseg;
1323	(void) tcp_output(pData, tp);
1324	LogFlowFunc(("%d -> drop\n", __LINE__));
1325	goto drop;
1326	}
1327	}
1328	else
1329	tp->t_dupacks = 0;
1330	break;
1331	}
1332	synrx_to_est:
1333	LogFlowFunc(("synrx_to_est:\n"));
1334	/*
1335	* If the congestion window was inflated to account
1336	* for the other side's cached packets, retract it.
1337	*/
1338	if ( tp->t_dupacks > tcprexmtthresh
1339	&& tp->snd_cwnd > tp->snd_ssthresh)
1340	tp->snd_cwnd = tp->snd_ssthresh;
1341	tp->t_dupacks = 0;
1342	if (SEQ_GT(ti->ti_ack, tp->snd_max))
1343	{
1344	tcpstat.tcps_rcvacktoomuch++;
1345	LogFlowFunc(("%d -> dropafterack\n", __LINE__));
1346	goto dropafterack;
1347	}
1348	acked = ti->ti_ack - tp->snd_una;
1349	tcpstat.tcps_rcvackpack++;
1350	tcpstat.tcps_rcvackbyte += acked;
1351
1352	/*
1353	* If we have a timestamp reply, update smoothed
1354	* round trip time. If no timestamp is present but
1355	* transmit timer is running and timed sequence
1356	* number was acked, update smoothed round trip time.
1357	* Since we now have an rtt measurement, cancel the
1358	* timer backoff (cf., Phil Karn's retransmit alg.).
1359	* Recompute the initial retransmit timer.
1360	*/
1361	#if 0
1362	if (ts_present)
1363	tcp_xmit_timer(tp, tcp_now-ts_ecr+1);
1364	else
1365	#endif
1366	if (tp->t_rtt && SEQ_GT(ti->ti_ack, tp->t_rtseq))
1367	tcp_xmit_timer(pData, tp, tp->t_rtt);
1368
1369	/*
1370	* If all outstanding data is acked, stop retransmit
1371	* timer and remember to restart (more output or persist).
1372	* If there is more data to be acked, restart retransmit
1373	* timer, using current (possibly backed-off) value.
1374	*/
1375	if (ti->ti_ack == tp->snd_max)
1376	{
1377	tp->t_timer[TCPT_REXMT] = 0;
1378	needoutput = 1;
1379	}
1380	else if (tp->t_timer[TCPT_PERSIST] == 0)
1381	tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
1382	/*
1383	* When new data is acked, open the congestion window.
1384	* If the window gives us less than ssthresh packets
1385	* in flight, open exponentially (maxseg per packet).
1386	* Otherwise open linearly: maxseg per window
1387	* (maxseg^2 / cwnd per packet).
1388	*/
1389	{
1390	register u_int cw = tp->snd_cwnd;
1391	register u_int incr = tp->t_maxseg;
1392
1393	if (cw > tp->snd_ssthresh)
1394	incr = incr * incr / cw;
1395	tp->snd_cwnd = min(cw + incr, TCP_MAXWIN<<tp->snd_scale);
1396	}
1397	if (acked > SBUF_LEN(&so->so_snd))
1398	{
1399	tp->snd_wnd -= SBUF_LEN(&so->so_snd);
1400	#ifndef VBOX_WITH_SLIRP_BSD_SBUF
1401	sbdrop(&so->so_snd, (int)so->so_snd.sb_cc);
1402	#else
1403	sbuf_clear(&so->so_snd);
1404	#endif
1405	ourfinisacked = 1;
1406	}
1407	else
1408	{
1409	#ifndef VBOX_WITH_SLIRP_BSD_SBUF
1410	sbdrop(&so->so_snd, acked);
1411	#else
1412	sbuf_setpos(&so->so_snd, sbuf_len(&so->so_snd) - acked);
1413	#endif
1414	tp->snd_wnd -= acked;
1415	ourfinisacked = 0;
1416	}
1417	/*
1418	* XXX sowwakup is called when data is acked and there's room for
1419	* for more data... it should read() the socket
1420	*/
1421	#if 0
1422	if (so->so_snd.sb_flags & SB_NOTIFY)
1423	sowwakeup(so);
1424	#endif
1425	tp->snd_una = ti->ti_ack;
1426	if (SEQ_LT(tp->snd_nxt, tp->snd_una))
1427	tp->snd_nxt = tp->snd_una;
1428
1429	switch (tp->t_state)
1430	{
1431	/*
1432	* In FIN_WAIT_1 STATE in addition to the processing
1433	* for the ESTABLISHED state if our FIN is now acknowledged
1434	* then enter FIN_WAIT_2.
1435	*/
1436	case TCPS_FIN_WAIT_1:
1437	if (ourfinisacked)
1438	{
1439	/*
1440	* If we can't receive any more
1441	* data, then closing user can proceed.
1442	* Starting the timer is contrary to the
1443	* specification, but if we don't get a FIN
1444	* we'll hang forever.
1445	*/
1446	if (so->so_state & SS_FCANTRCVMORE)
1447	{
1448	soisfdisconnected(so);
1449	tp->t_timer[TCPT_2MSL] = tcp_maxidle;
1450	}
1451	TCP_STATE_SWITCH_TO(tp, TCPS_FIN_WAIT_2);
1452	}
1453	break;
1454
1455	/*
1456	* In CLOSING STATE in addition to the processing for
1457	* the ESTABLISHED state if the ACK acknowledges our FIN
1458	* then enter the TIME-WAIT state, otherwise ignore
1459	* the segment.
1460	*/
1461	case TCPS_CLOSING:
1462	if (ourfinisacked)
1463	{
1464	TCP_STATE_SWITCH_TO(tp, TCPS_TIME_WAIT);
1465	tcp_canceltimers(tp);
1466	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1467	soisfdisconnected(so);
1468	}
1469	break;
1470
1471	/*
1472	* In LAST_ACK, we may still be waiting for data to drain
1473	* and/or to be acked, as well as for the ack of our FIN.
1474	* If our FIN is now acknowledged, delete the TCB,
1475	* enter the closed state and return.
1476	*/
1477	case TCPS_LAST_ACK:
1478	if (ourfinisacked)
1479	{
1480	tp = tcp_close(pData, tp);
1481	LogFlowFunc(("%d -> drop\n", __LINE__));
1482	goto drop;
1483	}
1484	break;
1485
1486	/*
1487	* In TIME_WAIT state the only thing that should arrive
1488	* is a retransmission of the remote FIN. Acknowledge
1489	* it and restart the finack timer.
1490	*/
1491	case TCPS_TIME_WAIT:
1492	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1493	LogFlowFunc(("%d -> dropafterack\n", __LINE__));
1494	goto dropafterack;
1495	}
1496	} /* switch(tp->t_state) */
1497
1498	step6:
1499	LogFlowFunc(("step6:\n"));
1500	/*
1501	* Update window information.
1502	* Don't look at window if no ACK: TAC's send garbage on first SYN.
1503	*/
1504	if ( (tiflags & TH_ACK)
1505	&& ( SEQ_LT(tp->snd_wl1, ti->ti_seq)
1506	\|\| ( tp->snd_wl1 == ti->ti_seq
1507	&& ( SEQ_LT(tp->snd_wl2, ti->ti_ack)
1508	\|\| ( tp->snd_wl2 == ti->ti_ack
1509	&& tiwin > tp->snd_wnd)))))
1510	{
1511	/* keep track of pure window updates */
1512	if ( ti->ti_len == 0
1513	&& tp->snd_wl2 == ti->ti_ack
1514	&& tiwin > tp->snd_wnd)
1515	tcpstat.tcps_rcvwinupd++;
1516	tp->snd_wnd = tiwin;
1517	tp->snd_wl1 = ti->ti_seq;
1518	tp->snd_wl2 = ti->ti_ack;
1519	if (tp->snd_wnd > tp->max_sndwnd)
1520	tp->max_sndwnd = tp->snd_wnd;
1521	needoutput = 1;
1522	}
1523
1524	/*
1525	* Process segments with URG.
1526	*/
1527	if ((tiflags & TH_URG) && ti->ti_urp &&
1528	TCPS_HAVERCVDFIN(tp->t_state) == 0)
1529	{
1530	/* BSD's sbufs are auto extent so we shouldn't worry here */
1531	#ifndef VBOX_WITH_SLIRP_BSD_SBUF
1532	/*
1533	* This is a kludge, but if we receive and accept
1534	* random urgent pointers, we'll crash in
1535	* soreceive. It's hard to imagine someone
1536	* actually wanting to send this much urgent data.
1537	*/
1538	if (ti->ti_urp + so->so_rcv.sb_cc > so->so_rcv.sb_datalen)
1539	{
1540	ti->ti_urp = 0;
1541	tiflags &= ~TH_URG;
1542	LogFlowFunc(("%d -> dodata\n", __LINE__));
1543	goto dodata;
1544	}
1545	#endif
1546	/*
1547	* If this segment advances the known urgent pointer,
1548	* then mark the data stream. This should not happen
1549	* in CLOSE_WAIT, CLOSING, LAST_ACK or TIME_WAIT STATES since
1550	* a FIN has been received from the remote side.
1551	* In these states we ignore the URG.
1552	*
1553	* According to RFC961 (Assigned Protocols),
1554	* the urgent pointer points to the last octet
1555	* of urgent data. We continue, however,
1556	* to consider it to indicate the first octet
1557	* of data past the urgent section as the original
1558	* spec states (in one of two places).
1559	*/
1560	if (SEQ_GT(ti->ti_seq+ti->ti_urp, tp->rcv_up))
1561	{
1562	tp->rcv_up = ti->ti_seq + ti->ti_urp;
1563	so->so_urgc = SBUF_LEN(&so->so_rcv) +
1564	(tp->rcv_up - tp->rcv_nxt); /* -1; */
1565	tp->rcv_up = ti->ti_seq + ti->ti_urp;
1566	}
1567	}
1568	else
1569	/*
1570	* If no out of band data is expected,
1571	* pull receive urgent pointer along
1572	* with the receive window.
1573	*/
1574	if (SEQ_GT(tp->rcv_nxt, tp->rcv_up))
1575	tp->rcv_up = tp->rcv_nxt;
1576	dodata:
1577	LogFlowFunc(("dodata:\n"));
1578
1579	/*
1580	* If this is a small packet, then ACK now - with Nagel
1581	* congestion avoidance sender won't send more until
1582	* he gets an ACK.
1583	*
1584	* See above.
1585	*/
1586	if ( ti->ti_len
1587	&& (unsigned)ti->ti_len <= 5
1588	&& ((struct tcpiphdr_2 *)ti)->first_char == (char)27)
1589	{
1590	tp->t_flags \|= TF_ACKNOW;
1591	}
1592
1593	/*
1594	* Process the segment text, merging it into the TCP sequencing queue,
1595	* and arranging for acknowledgment of receipt if necessary.
1596	* This process logically involves adjusting tp->rcv_wnd as data
1597	* is presented to the user (this happens in tcp_usrreq.c,
1598	* case PRU_RCVD). If a FIN has already been received on this
1599	* connection then we just ignore the text.
1600	*/
1601	if ( (ti->ti_len \|\| (tiflags&TH_FIN))
1602	&& TCPS_HAVERCVDFIN(tp->t_state) == 0)
1603	{
1604	if ( ti->ti_seq == tp->rcv_nxt
1605	&& LIST_EMPTY(&tp->t_segq)
1606	&& tp->t_state == TCPS_ESTABLISHED)
1607	{
1608	DELAY_ACK(tp, ti); /* little bit different from BSD declaration see netinet/tcp_input.c */
1609	tp->rcv_nxt += tlen;
1610	tiflags = ti->ti_t.th_flags & TH_FIN;
1611	tcpstat.tcps_rcvpack++;
1612	tcpstat.tcps_rcvbyte += tlen;
1613	if (so->so_state & SS_FCANTRCVMORE)
1614	m_freem(pData, m);
1615	else
1616	sbappend(pData, so, m);
1617	}
1618	else
1619	{
1620	tiflags = tcp_reass(pData, tp, &ti->ti_t, &tlen, m);
1621	tiflags \|= TF_ACKNOW;
1622	}
1623	/*
1624	* Note the amount of data that peer has sent into
1625	* our window, in order to estimate the sender's
1626	* buffer size.
1627	*/
1628	len = SBUF_SIZE(&so->so_rcv) - (tp->rcv_adv - tp->rcv_nxt);
1629	}
1630	else
1631	{
1632	m_freem(pData, m);
1633	tiflags &= ~TH_FIN;
1634	}
1635
1636	/*
1637	* If FIN is received ACK the FIN and let the user know
1638	* that the connection is closing.
1639	*/
1640	if (tiflags & TH_FIN)
1641	{
1642	if (TCPS_HAVERCVDFIN(tp->t_state) == 0)
1643	{
1644	/*
1645	* If we receive a FIN we can't send more data,
1646	* set it SS_FDRAIN
1647	* Shutdown the socket if there is no rx data in the
1648	* buffer.
1649	* soread() is called on completion of shutdown() and
1650	* will got to TCPS_LAST_ACK, and use tcp_output()
1651	* to send the FIN.
1652	*/
1653	/* sofcantrcvmore(so); */
1654	sofwdrain(so);
1655
1656	tp->t_flags \|= TF_ACKNOW;
1657	tp->rcv_nxt++;
1658	}
1659	switch (tp->t_state)
1660	{
1661	/*
1662	* In SYN_RECEIVED and ESTABLISHED STATES
1663	* enter the CLOSE_WAIT state.
1664	*/
1665	case TCPS_SYN_RECEIVED:
1666	case TCPS_ESTABLISHED:
1667	TCP_STATE_SWITCH_TO(tp, TCPS_CLOSE_WAIT);
1668	break;
1669
1670	/*
1671	* If still in FIN_WAIT_1 STATE FIN has not been acked so
1672	* enter the CLOSING state.
1673	*/
1674	case TCPS_FIN_WAIT_1:
1675	TCP_STATE_SWITCH_TO(tp, TCPS_CLOSING);
1676	break;
1677
1678	/*
1679	* In FIN_WAIT_2 state enter the TIME_WAIT state,
1680	* starting the time-wait timer, turning off the other
1681	* standard timers.
1682	*/
1683	case TCPS_FIN_WAIT_2:
1684	TCP_STATE_SWITCH_TO(tp, TCPS_TIME_WAIT);
1685	tcp_canceltimers(tp);
1686	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1687	soisfdisconnected(so);
1688	break;
1689
1690	/*
1691	* In TIME_WAIT state restart the 2 MSL time_wait timer.
1692	*/
1693	case TCPS_TIME_WAIT:
1694	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1695	break;
1696	}
1697	}
1698
1699	/*
1700	* Return any desired output.
1701	*/
1702	if (needoutput \|\| (tp->t_flags & TF_ACKNOW))
1703	tcp_output(pData, tp);
1704
1705	SOCKET_UNLOCK(so);
1706	STAM_PROFILE_STOP(&pData->StatTCP_input, counter_input);
1707	LogFlowFuncLeave();
1708	return;
1709
1710	dropafterack:
1711	LogFlowFunc(("dropafterack:\n"));
1712	/*
1713	* Generate an ACK dropping incoming segment if it occupies
1714	* sequence space, where the ACK reflects our state.
1715	*/
1716	if (tiflags & TH_RST)
1717	{
1718	LogFlowFunc(("%d -> drop\n", __LINE__));
1719	goto drop;
1720	}
1721	m_freem(pData, m);
1722	tp->t_flags \|= TF_ACKNOW;
1723	(void) tcp_output(pData, tp);
1724	SOCKET_UNLOCK(so);
1725	STAM_PROFILE_STOP(&pData->StatTCP_input, counter_input);
1726	LogFlowFuncLeave();
1727	return;
1728
1729	dropwithreset:
1730	LogFlowFunc(("dropwithreset:\n"));
1731	/* reuses m if m!=NULL, m_free() unnecessary */
1732	if (tiflags & TH_ACK)
1733	tcp_respond(pData, tp, ti, m, (tcp_seq)0, ti->ti_ack, TH_RST);
1734	else
1735	{
1736	if (tiflags & TH_SYN)
1737	ti->ti_len++;
1738	tcp_respond(pData, tp, ti, m, ti->ti_seq+ti->ti_len, (tcp_seq)0,
1739	TH_RST\|TH_ACK);
1740	}
1741
1742	if (so != &tcb)
1743	SOCKET_UNLOCK(so);
1744	STAM_PROFILE_STOP(&pData->StatTCP_input, counter_input);
1745	LogFlowFuncLeave();
1746	return;
1747
1748	drop:
1749	LogFlowFunc(("drop:\n"));
1750	/*
1751	* Drop space held by incoming segment and return.
1752	*/
1753	m_freem(pData, m);
1754
1755	#ifdef VBOX_WITH_SLIRP_MT
1756	if (RTCritSectIsOwned(&so->so_mutex))
1757	{
1758	SOCKET_UNLOCK(so);
1759	}
1760	#endif
1761
1762	STAM_PROFILE_STOP(&pData->StatTCP_input, counter_input);
1763	LogFlowFuncLeave();
1764	return;
1765	}
1766
1767	void
1768	tcp_dooptions(PNATState pData, struct tcpcb tp, u_char cp, int cnt, struct tcpiphdr *ti)
1769	{
1770	u_int16_t mss;
1771	int opt, optlen;
1772
1773	LogFlowFunc(("tcp_dooptions: tp = %R[tcpcb793], cnt=%i\n", tp, cnt));
1774
1775	for (; cnt > 0; cnt -= optlen, cp += optlen)
1776	{
1777	opt = cp[0];
1778	if (opt == TCPOPT_EOL)
1779	break;
1780	if (opt == TCPOPT_NOP)
1781	optlen = 1;
1782	else
1783	{
1784	optlen = cp[1];
1785	if (optlen <= 0)
1786	break;
1787	}
1788	switch (opt)
1789	{
1790	default:
1791	continue;
1792
1793	case TCPOPT_MAXSEG:
1794	if (optlen != TCPOLEN_MAXSEG)
1795	continue;
1796	if (!(ti->ti_flags & TH_SYN))
1797	continue;
1798	memcpy((char ) &mss, (char ) cp + 2, sizeof(mss));
1799	NTOHS(mss);
1800	(void) tcp_mss(pData, tp, mss); /* sets t_maxseg */
1801	break;
1802
1803	#if 0
1804	case TCPOPT_WINDOW:
1805	if (optlen != TCPOLEN_WINDOW)
1806	continue;
1807	if (!(ti->ti_flags & TH_SYN))
1808	continue;
1809	tp->t_flags \|= TF_RCVD_SCALE;
1810	tp->requested_s_scale = min(cp[2], TCP_MAX_WINSHIFT);
1811	break;
1812
1813	case TCPOPT_TIMESTAMP:
1814	if (optlen != TCPOLEN_TIMESTAMP)
1815	continue;
1816	*ts_present = 1;
1817	memcpy((char ) ts_val, (char )cp + 2, sizeof(*ts_val));
1818	NTOHL(*ts_val);
1819	memcpy((char ) ts_ecr, (char )cp + 6, sizeof(*ts_ecr));
1820	NTOHL(*ts_ecr);
1821
1822	/*
1823	* A timestamp received in a SYN makes
1824	* it ok to send timestamp requests and replies.
1825	*/
1826	if (ti->ti_flags & TH_SYN)
1827	{
1828	tp->t_flags \|= TF_RCVD_TSTMP;
1829	tp->ts_recent = *ts_val;
1830	tp->ts_recent_age = tcp_now;
1831	}
1832	break;
1833	#endif
1834	}
1835	}
1836	}
1837
1838
1839	/*
1840	* Pull out of band byte out of a segment so
1841	* it doesn't appear in the user's data queue.
1842	* It is still reflected in the segment length for
1843	* sequencing purposes.
1844	*/
1845
1846	#if 0
1847	void
1848	tcp_pulloutofband(struct socket so, struct tcpiphdr ti, struct mbuf *m)
1849	{
1850	int cnt = ti->ti_urp - 1;
1851
1852	while (cnt >= 0)
1853	{
1854	if (m->m_len > cnt)
1855	{
1856	char *cp = mtod(m, caddr_t) + cnt;
1857	struct tcpcb *tp = sototcpcb(so);
1858
1859	tp->t_iobc = *cp;
1860	tp->t_oobflags \|= TCPOOB_HAVEDATA;
1861	memcpy(sp, cp+1, (unsigned)(m->m_len - cnt - 1));
1862	m->m_len--;
1863	return;
1864	}
1865	cnt -= m->m_len;
1866	m = m->m_next; /* XXX WRONG! Fix it! */
1867	if (m == 0)
1868	break;
1869	}
1870	panic("tcp_pulloutofband");
1871	}
1872	#endif
1873
1874	/*
1875	* Collect new round-trip time estimate
1876	* and update averages and current timeout.
1877	*/
1878
1879	void
1880	tcp_xmit_timer(PNATState pData, register struct tcpcb *tp, int rtt)
1881	{
1882	register short delta;
1883
1884	LogFlowFunc(("ENTER: tcp_xmit_timer: tp = %R[tcpcb793] rtt = %d\n", tp, rtt));
1885
1886	tcpstat.tcps_rttupdated++;
1887	if (tp->t_srtt != 0)
1888	{
1889	/*
1890	* srtt is stored as fixed point with 3 bits after the
1891	* binary point (i.e., scaled by 8). The following magic
1892	* is equivalent to the smoothing algorithm in rfc793 with
1893	* an alpha of .875 (srtt = rtt/8 + srtt*7/8 in fixed
1894	* point). Adjust rtt to origin 0.
1895	*/
1896	delta = rtt - 1 - (tp->t_srtt >> TCP_RTT_SHIFT);
1897	if ((tp->t_srtt += delta) <= 0)
1898	tp->t_srtt = 1;
1899	/*
1900	* We accumulate a smoothed rtt variance (actually, a
1901	* smoothed mean difference), then set the retransmit
1902	* timer to smoothed rtt + 4 times the smoothed variance.
1903	* rttvar is stored as fixed point with 2 bits after the
1904	* binary point (scaled by 4). The following is
1905	* equivalent to rfc793 smoothing with an alpha of .75
1906	* (rttvar = rttvar*3/4 + \|delta\| / 4). This replaces
1907	* rfc793's wired-in beta.
1908	*/
1909	if (delta < 0)
1910	delta = -delta;
1911	delta -= (tp->t_rttvar >> TCP_RTTVAR_SHIFT);
1912	if ((tp->t_rttvar += delta) <= 0)
1913	tp->t_rttvar = 1;
1914	}
1915	else
1916	{
1917	/*
1918	* No rtt measurement yet - use the unsmoothed rtt.
1919	* Set the variance to half the rtt (so our first
1920	* retransmit happens at 3*rtt).
1921	*/
1922	tp->t_srtt = rtt << TCP_RTT_SHIFT;
1923	tp->t_rttvar = rtt << (TCP_RTTVAR_SHIFT - 1);
1924	}
1925	tp->t_rtt = 0;
1926	tp->t_rxtshift = 0;
1927
1928	/*
1929	* the retransmit should happen at rtt + 4 * rttvar.
1930	* Because of the way we do the smoothing, srtt and rttvar
1931	* will each average +1/2 tick of bias. When we compute
1932	* the retransmit timer, we want 1/2 tick of rounding and
1933	* 1 extra tick because of +-1/2 tick uncertainty in the
1934	* firing of the timer. The bias will give us exactly the
1935	* 1.5 tick we need. But, because the bias is
1936	* statistical, we have to test that we don't drop below
1937	* the minimum feasible timer (which is 2 ticks).
1938	*/
1939	TCPT_RANGESET(tp->t_rxtcur, TCP_REXMTVAL(tp),
1940	(short)tp->t_rttmin, TCPTV_REXMTMAX); /* XXX */
1941
1942	/*
1943	* We received an ack for a packet that wasn't retransmitted;
1944	* it is probably safe to discard any error indications we've
1945	* received recently. This isn't quite right, but close enough
1946	* for now (a route might have failed after we sent a segment,
1947	* and the return path might not be symmetrical).
1948	*/
1949	tp->t_softerror = 0;
1950	}
1951
1952	/*
1953	* Determine a reasonable value for maxseg size.
1954	* If the route is known, check route for mtu.
1955	* If none, use an mss that can be handled on the outgoing
1956	* interface without forcing IP to fragment; if bigger than
1957	* an mbuf cluster (MCLBYTES), round down to nearest multiple of MCLBYTES
1958	* to utilize large mbufs. If no route is found, route has no mtu,
1959	* or the destination isn't local, use a default, hopefully conservative
1960	* size (usually 512 or the default IP max size, but no more than the mtu
1961	* of the interface), as we can't discover anything about intervening
1962	* gateways or networks. We also initialize the congestion/slow start
1963	* window to be a single segment if the destination isn't local.
1964	* While looking at the routing entry, we also initialize other path-dependent
1965	* parameters from pre-set or cached values in the routing entry.
1966	*/
1967
1968	int
1969	tcp_mss(PNATState pData, register struct tcpcb *tp, u_int offer)
1970	{
1971	struct socket *so = tp->t_socket;
1972	int mss;
1973
1974	LogFlowFunc(("ENTER: tcp_mss: tp = %R[tcpcb793], offer = %d\n", tp, offer));
1975
1976	mss = min(if_mtu, if_mru) - sizeof(struct tcpiphdr);
1977	if (offer)
1978	mss = min(mss, offer);
1979	mss = max(mss, 32);
1980	if (mss < tp->t_maxseg \|\| offer != 0)
1981	tp->t_maxseg = mss;
1982
1983	tp->snd_cwnd = mss;
1984
1985	#ifndef VBOX_WITH_SLIRP_BSD_SBUF
1986	sbreserve(pData, &so->so_snd, tcp_sndspace+((tcp_sndspace%mss)?(mss-(tcp_sndspace%mss)):0));
1987	sbreserve(pData, &so->so_rcv, tcp_rcvspace+((tcp_rcvspace%mss)?(mss-(tcp_rcvspace%mss)):0));
1988	#else
1989	sbuf_new(&so->so_snd, NULL, tcp_sndspace+((tcp_sndspace%mss)?(mss-(tcp_sndspace%mss)):0), SBUF_AUTOEXTEND);
1990	sbuf_new(&so->so_rcv, NULL, tcp_rcvspace+((tcp_rcvspace%mss)?(mss-(tcp_rcvspace%mss)):0), SBUF_AUTOEXTEND);
1991	#endif
1992
1993	Log2((" returning mss = %d\n", mss));
1994
1995	return mss;
1996	}

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

source: vbox/trunk/src/VBox/Devices/Network/slirp/tcp_input.c@ 40623

Download in other formats: