tcp_input.c@ 37910

Last change on this file since 37910 was 37910, checked in by vboxsync, 13 years ago
NAT: findso: better logging.
Property svn:eol-style set to `native` Property svn:keywords set to `Author Date Id Revision`
File size: 65.3 KB

Line
1	/* $Id: tcp_input.c 37910 2011-07-13 11:18:39Z 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	#define TCP_PAWS_IDLE (24 * 24 * 60 * 60 * PR_SLOWHZ)
69
70	/* for modulo comparisons of timestamps */
71	#define TSTMP_LT(a, b) ((int)((a)-(b)) < 0)
72	#define TSTMP_GEQ(a, b) ((int)((a)-(b)) >= 0)
73
74	#ifndef TCP_ACK_HACK
75	#define DELAY_ACK(tp, ti) \
76	if (ti->ti_flags & TH_PUSH) \
77	tp->t_flags \|= TF_ACKNOW; \
78	else \
79	tp->t_flags \|= TF_DELACK;
80	#else /* !TCP_ACK_HACK */
81	#define DELAY_ACK(tp, ign) \
82	tp->t_flags \|= TF_DELACK;
83	#endif /* TCP_ACK_HACK */
84
85
86	/*
87	* deps: netinet/tcp_reass.c
88	* tcp_reass_maxqlen = 48 (deafault)
89	* tcp_reass_maxseg = nmbclusters/16 (nmbclusters = 1024 + maxusers * 64 from kern/kern_mbuf.c let's say 256)
90	*/
91	int
92	tcp_reass(PNATState pData, struct tcpcb tp, struct tcphdr th, int tlenp, struct mbuf m)
93	{
94	struct tseg_qent *q;
95	struct tseg_qent *p = NULL;
96	struct tseg_qent *nq;
97	struct tseg_qent *te = NULL;
98	struct socket *so = tp->t_socket;
99	int flags;
100	STAM_PROFILE_START(&pData->StatTCP_reassamble, tcp_reassamble);
101	LogFlowFuncEnter();
102	LogFlowFunc(("pData:%p, tp:%p, th:%p, tlenp:%p, m:%p\n", pData, tp, th, tlenp, m));
103
104	/*
105	* XXX: tcp_reass() is rather inefficient with its data structures
106	* and should be rewritten (see NetBSD for optimizations). While
107	* doing that it should move to its own file tcp_reass.c.
108	*/
109
110	/*
111	* Call with th==NULL after become established to
112	* force pre-ESTABLISHED data up to user socket.
113	*/
114	if (th == NULL)
115	{
116	LogFlowFunc(("%d -> present\n", __LINE__));
117	goto present;
118	}
119
120	/*
121	* Limit the number of segments in the reassembly queue to prevent
122	* holding on to too many segments (and thus running out of mbufs).
123	* Make sure to let the missing segment through which caused this
124	* queue. Always keep one global queue entry spare to be able to
125	* process the missing segment.
126	*/
127	if ( th->th_seq != tp->rcv_nxt
128	&& ( tcp_reass_qsize + 1 >= tcp_reass_maxseg
129	\|\| tp->t_segqlen >= tcp_reass_maxqlen))
130	{
131	tcp_reass_overflows++;
132	tcpstat.tcps_rcvmemdrop++;
133	m_freem(pData, m);
134	*tlenp = 0;
135	STAM_PROFILE_STOP(&pData->StatTCP_reassamble, tcp_reassamble);
136	LogFlowFuncLeave();
137	return (0);
138	}
139
140	/*
141	* Allocate a new queue entry. If we can't, or hit the zone limit
142	* just drop the pkt.
143	*/
144	te = RTMemAlloc(sizeof(struct tseg_qent));
145	if (te == NULL)
146	{
147	tcpstat.tcps_rcvmemdrop++;
148	m_freem(pData, m);
149	*tlenp = 0;
150	STAM_PROFILE_STOP(&pData->StatTCP_reassamble, tcp_reassamble);
151	LogFlowFuncLeave();
152	return (0);
153	}
154	tp->t_segqlen++;
155	tcp_reass_qsize++;
156
157	/*
158	* Find a segment which begins after this one does.
159	*/
160	LIST_FOREACH(q, &tp->t_segq, tqe_q)
161	{
162	if (SEQ_GT(q->tqe_th->th_seq, th->th_seq))
163	break;
164	p = q;
165	}
166
167	/*
168	* If there is a preceding segment, it may provide some of
169	* our data already. If so, drop the data from the incoming
170	* segment. If it provides all of our data, drop us.
171	*/
172	if (p != NULL)
173	{
174	int i;
175	/* conversion to int (in i) handles seq wraparound */
176	i = p->tqe_th->th_seq + p->tqe_len - th->th_seq;
177	if (i > 0)
178	{
179	if (i >= *tlenp)
180	{
181	tcpstat.tcps_rcvduppack++;
182	tcpstat.tcps_rcvdupbyte += *tlenp;
183	m_freem(pData, m);
184	RTMemFree(te);
185	tp->t_segqlen--;
186	tcp_reass_qsize--;
187	/*
188	* Try to present any queued data
189	* at the left window edge to the user.
190	* This is needed after the 3-WHS
191	* completes.
192	*/
193	LogFlowFunc(("%d -> present\n", __LINE__));
194	goto present; /* ??? */
195	}
196	m_adj(m, i);
197	*tlenp -= i;
198	th->th_seq += i;
199	}
200	}
201	tcpstat.tcps_rcvoopack++;
202	tcpstat.tcps_rcvoobyte += *tlenp;
203
204	/*
205	* While we overlap succeeding segments trim them or,
206	* if they are completely covered, dequeue them.
207	*/
208	while (q)
209	{
210	int i = (th->th_seq + *tlenp) - q->tqe_th->th_seq;
211	if (i <= 0)
212	break;
213	if (i < q->tqe_len)
214	{
215	q->tqe_th->th_seq += i;
216	q->tqe_len -= i;
217	m_adj(q->tqe_m, i);
218	break;
219	}
220
221	nq = LIST_NEXT(q, tqe_q);
222	LIST_REMOVE(q, tqe_q);
223	m_freem(pData, q->tqe_m);
224	RTMemFree(q);
225	tp->t_segqlen--;
226	tcp_reass_qsize--;
227	q = nq;
228	}
229
230	/* Insert the new segment queue entry into place. */
231	te->tqe_m = m;
232	te->tqe_th = th;
233	te->tqe_len = *tlenp;
234
235	if (p == NULL)
236	{
237	LIST_INSERT_HEAD(&tp->t_segq, te, tqe_q);
238	}
239	else
240	{
241	LIST_INSERT_AFTER(p, te, tqe_q);
242	}
243
244	present:
245	/*
246	* Present data to user, advancing rcv_nxt through
247	* completed sequence space.
248	*/
249	if (!TCPS_HAVEESTABLISHED(tp->t_state))
250	{
251	STAM_PROFILE_STOP(&pData->StatTCP_reassamble, tcp_reassamble);
252	return (0);
253	}
254	q = LIST_FIRST(&tp->t_segq);
255	if (!q \|\| q->tqe_th->th_seq != tp->rcv_nxt)
256	{
257	STAM_PROFILE_STOP(&pData->StatTCP_reassamble, tcp_reassamble);
258	return (0);
259	}
260	do
261	{
262	tp->rcv_nxt += q->tqe_len;
263	flags = q->tqe_th->th_flags & TH_FIN;
264	nq = LIST_NEXT(q, tqe_q);
265	LIST_REMOVE(q, tqe_q);
266	/* XXX: This place should be checked for the same code in
267	* original BSD code for Slirp and current BSD used SS_FCANTRCVMORE
268	*/
269	if (so->so_state & SS_FCANTSENDMORE)
270	m_freem(pData, q->tqe_m);
271	else
272	sbappend(pData, so, q->tqe_m);
273	RTMemFree(q);
274	tp->t_segqlen--;
275	tcp_reass_qsize--;
276	q = nq;
277	}
278	while (q && q->tqe_th->th_seq == tp->rcv_nxt);
279
280	STAM_PROFILE_STOP(&pData->StatTCP_reassamble, tcp_reassamble);
281	return flags;
282	}
283
284	/*
285	* TCP input routine, follows pages 65-76 of the
286	* protocol specification dated September, 1981 very closely.
287	*/
288	void
289	tcp_input(PNATState pData, register struct mbuf m, int iphlen, struct socket inso)
290	{
291	struct ip save_ip, *ip;
292	register struct tcpiphdr *ti;
293	caddr_t optp = NULL;
294	int optlen = 0;
295	int len, tlen, off;
296	register struct tcpcb *tp = 0;
297	register int tiflags;
298	struct socket *so = 0;
299	int todrop, acked, ourfinisacked, needoutput = 0;
300	/* int dropsocket = 0; */
301	int iss = 0;
302	u_long tiwin;
303	/* int ts_present = 0; */
304	STAM_PROFILE_START(&pData->StatTCP_input, counter_input);
305
306	LogFlow(("tcp_input: m = %8lx, iphlen = %2d, inso = %lx\n",
307	(long)m, iphlen, (long)inso));
308
309	if (inso != NULL)
310	{
311	QSOCKET_LOCK(tcb);
312	SOCKET_LOCK(inso);
313	QSOCKET_UNLOCK(tcb);
314	}
315	/*
316	* If called with m == 0, then we're continuing the connect
317	*/
318	if (m == NULL)
319	{
320	so = inso;
321	Log4(("NAT: tcp_input: %R[natsock]\n", so));
322	/* Re-set a few variables */
323	tp = sototcpcb(so);
324	m = so->so_m;
325
326	so->so_m = 0;
327	ti = so->so_ti;
328
329	/** @todo (vvl) clarify why it might happens */
330	if (ti == NULL)
331	{
332	LogRel(("NAT: ti is null. can't do any reseting connection actions\n"));
333	/* mbuf should be cleared in sofree called from tcp_close */
334	tcp_close(pData, tp);
335	STAM_PROFILE_STOP(&pData->StatTCP_input, counter_input);
336	LogFlowFuncLeave();
337	return;
338	}
339
340	tiwin = ti->ti_win;
341	tiflags = ti->ti_flags;
342
343	LogFlowFunc(("%d -> cont_conn\n", __LINE__));
344	goto cont_conn;
345	}
346
347	tcpstat.tcps_rcvtotal++;
348	/*
349	* Get IP and TCP header together in first mbuf.
350	* Note: IP leaves IP header in first mbuf.
351	*/
352	ti = mtod(m, struct tcpiphdr *);
353	if (iphlen > sizeof(struct ip))
354	{
355	ip_stripoptions(m, (struct mbuf *)0);
356	iphlen = sizeof(struct ip);
357	}
358	/* XXX Check if too short */
359
360
361	/*
362	* Save a copy of the IP header in case we want restore it
363	* for sending an ICMP error message in response.
364	*/
365	ip = mtod(m, struct ip *);
366	/*
367	* (vvl) ip_input substracts IP header length from ip->ip_len value.
368	* here we do the test the same as input method of UDP protocol.
369	*/
370	Assert((ip->ip_len + iphlen == m_length(m, NULL)));
371	save_ip = *ip;
372	save_ip.ip_len+= iphlen;
373
374	/*
375	* Checksum extended TCP header and data.
376	*/
377	tlen = ((struct ip *)ti)->ip_len;
378	memset(ti->ti_x1, 0, 9);
379	ti->ti_len = RT_H2N_U16((u_int16_t)tlen);
380	len = sizeof(struct ip) + tlen;
381	/* keep checksum for ICMP reply
382	* ti->ti_sum = cksum(m, len);
383	* if (ti->ti_sum) { */
384	if (cksum(m, len))
385	{
386	tcpstat.tcps_rcvbadsum++;
387	LogFlowFunc(("%d -> drop\n", __LINE__));
388	goto drop;
389	}
390
391	/*
392	* Check that TCP offset makes sense,
393	* pull out TCP options and adjust length. XXX
394	*/
395	off = ti->ti_off << 2;
396	if ( off < sizeof (struct tcphdr)
397	\|\| off > tlen)
398	{
399	tcpstat.tcps_rcvbadoff++;
400	LogFlowFunc(("%d -> drop\n", __LINE__));
401	goto drop;
402	}
403	tlen -= off;
404	ti->ti_len = tlen;
405	if (off > sizeof (struct tcphdr))
406	{
407	optlen = off - sizeof (struct tcphdr);
408	optp = mtod(m, caddr_t) + sizeof (struct tcpiphdr);
409
410	/*
411	* Do quick retrieval of timestamp options ("options
412	* prediction?"). If timestamp is the only option and it's
413	* formatted as recommended in RFC 1323 appendix A, we
414	* quickly get the values now and not bother calling
415	* tcp_dooptions(), etc.
416	*/
417	#if 0
418	if (( optlen == TCPOLEN_TSTAMP_APPA
419	\|\| ( optlen > TCPOLEN_TSTAMP_APPA
420	&& optp[TCPOLEN_TSTAMP_APPA] == TCPOPT_EOL)) &&
421	(u_int32_t )optp == RT_H2N_U32_C(TCPOPT_TSTAMP_HDR) &&
422	(ti->ti_flags & TH_SYN) == 0)
423	{
424	ts_present = 1;
425	ts_val = RT_N2H_U32((u_int32_t )(optp + 4));
426	ts_ecr = RT_N2H_U32((u_int32_t )(optp + 8));
427	optp = NULL; / * we have parsed the options * /
428	}
429	#endif
430	}
431	tiflags = ti->ti_flags;
432
433	/*
434	* Convert TCP protocol specific fields to host format.
435	*/
436	NTOHL(ti->ti_seq);
437	NTOHL(ti->ti_ack);
438	NTOHS(ti->ti_win);
439	NTOHS(ti->ti_urp);
440
441	/*
442	* Drop TCP, IP headers and TCP options.
443	*/
444	m->m_data += sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
445	m->m_len -= sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
446
447	/*
448	* Locate pcb for segment.
449	*/
450	findso:
451	LogFlowFunc(("(enter) findso: %R[natsock]\n", so));
452	if (so != NULL && so != &tcb)
453	SOCKET_UNLOCK(so);
454	QSOCKET_LOCK(tcb);
455	so = tcp_last_so;
456	if ( so->so_fport != ti->ti_dport
457	\|\| so->so_lport != ti->ti_sport
458	\|\| so->so_laddr.s_addr != ti->ti_src.s_addr
459	\|\| so->so_faddr.s_addr != ti->ti_dst.s_addr)
460	{
461	#ifdef VBOX_WITH_SLIRP_MT
462	struct socket *sonxt;
463	#endif
464	QSOCKET_UNLOCK(tcb);
465	/* @todo fix SOLOOKUP macrodefinition to be usable here */
466	#ifndef VBOX_WITH_SLIRP_MT
467	so = solookup(&tcb, ti->ti_src, ti->ti_sport,
468	ti->ti_dst, ti->ti_dport);
469	#else
470	so = NULL;
471	QSOCKET_FOREACH(so, sonxt, tcp)
472	/* { */
473	if ( so->so_lport == ti->ti_sport
474	&& so->so_laddr.s_addr == ti->ti_src.s_addr
475	&& so->so_faddr.s_addr == ti->ti_dst.s_addr
476	&& so->so_fport == ti->ti_dport
477	&& so->so_deleted != 1)
478	{
479	break; /* so is locked here */
480	}
481	LOOP_LABEL(tcp, so, sonxt);
482	}
483	if (so == &tcb) {
484	so = NULL;
485	}
486	#endif
487	if (so)
488	{
489	tcp_last_so = so;
490	}
491	++tcpstat.tcps_socachemiss;
492	}
493	else
494	{
495	SOCKET_LOCK(so);
496	QSOCKET_UNLOCK(tcb);
497	}
498	LogFlowFunc(("(leave) findso: %R[natsock]\n", so));
499
500	/*
501	* If the state is CLOSED (i.e., TCB does not exist) then
502	* all data in the incoming segment is discarded.
503	* If the TCB exists but is in CLOSED state, it is embryonic,
504	* but should either do a listen or a connect soon.
505	*
506	* state == CLOSED means we've done socreate() but haven't
507	* attached it to a protocol yet...
508	*
509	* XXX If a TCB does not exist, and the TH_SYN flag is
510	* the only flag set, then create a session, mark it
511	* as if it was LISTENING, and continue...
512	*/
513	if (so == 0)
514	{
515	if ((tiflags & (TH_SYN\|TH_FIN\|TH_RST\|TH_URG\|TH_ACK)) != TH_SYN)
516	{
517	LogFlowFunc(("%d -> dropwithreset\n", __LINE__));
518	goto dropwithreset;
519	}
520
521	if ((so = socreate()) == NULL)
522	{
523	LogFlowFunc(("%d -> dropwithreset\n", __LINE__));
524	goto dropwithreset;
525	}
526	if (tcp_attach(pData, so) < 0)
527	{
528	RTMemFree(so); /* Not sofree (if it failed, it's not insqued) */
529	LogFlowFunc(("%d -> dropwithreset\n", __LINE__));
530	goto dropwithreset;
531	}
532	SOCKET_LOCK(so);
533	#ifndef VBOX_WITH_SLIRP_BSD_SBUF
534	sbreserve(pData, &so->so_snd, tcp_sndspace);
535	sbreserve(pData, &so->so_rcv, tcp_rcvspace);
536	#else
537	sbuf_new(&so->so_snd, NULL, tcp_sndspace, SBUF_AUTOEXTEND);
538	sbuf_new(&so->so_rcv, NULL, tcp_rcvspace, SBUF_AUTOEXTEND);
539	#endif
540
541	/* tcp_last_so = so; / / XXX ? */
542	/* tp = sototcpcb(so); */
543
544	so->so_laddr = ti->ti_src;
545	so->so_lport = ti->ti_sport;
546	so->so_faddr = ti->ti_dst;
547	so->so_fport = ti->ti_dport;
548
549	so->so_iptos = ((struct ip *)ti)->ip_tos;
550
551	tp = sototcpcb(so);
552	TCP_STATE_SWITCH_TO(tp, TCPS_LISTEN);
553	}
554
555	/*
556	* If this is a still-connecting socket, this probably
557	* a retransmit of the SYN. Whether it's a retransmit SYN
558	* or something else, we nuke it.
559	*/
560	if (so->so_state & SS_ISFCONNECTING)
561	{
562	LogFlowFunc(("%d -> drop\n", __LINE__));
563	goto drop;
564	}
565
566	tp = sototcpcb(so);
567
568	/* XXX Should never fail */
569	if (tp == 0)
570	{
571	LogFlowFunc(("%d -> dropwithreset\n", __LINE__));
572	goto dropwithreset;
573	}
574	if (tp->t_state == TCPS_CLOSED)
575	{
576	LogFlowFunc(("%d -> drop\n", __LINE__));
577	goto drop;
578	}
579
580	/* Unscale the window into a 32-bit value. */
581	/* if ((tiflags & TH_SYN) == 0)
582	* tiwin = ti->ti_win << tp->snd_scale;
583	* else
584	*/
585	tiwin = ti->ti_win;
586
587	/*
588	* Segment received on connection.
589	* Reset idle time and keep-alive timer.
590	*/
591	tp->t_idle = 0;
592	if (so_options)
593	tp->t_timer[TCPT_KEEP] = tcp_keepintvl;
594	else
595	tp->t_timer[TCPT_KEEP] = tcp_keepidle;
596
597	/*
598	* Process options if not in LISTEN state,
599	* else do it below (after getting remote address).
600	*/
601	if (optp && tp->t_state != TCPS_LISTEN)
602	tcp_dooptions(pData, tp, (u_char *)optp, optlen, ti);
603	/* , */
604	/* &ts_present, &ts_val, &ts_ecr); */
605
606	/*
607	* Header prediction: check for the two common cases
608	* of a uni-directional data xfer. If the packet has
609	* no control flags, is in-sequence, the window didn't
610	* change and we're not retransmitting, it's a
611	* candidate. If the length is zero and the ack moved
612	* forward, we're the sender side of the xfer. Just
613	* free the data acked & wake any higher level process
614	* that was blocked waiting for space. If the length
615	* is non-zero and the ack didn't move, we're the
616	* receiver side. If we're getting packets in-order
617	* (the reassembly queue is empty), add the data to
618	* the socket buffer and note that we need a delayed ack.
619	*
620	* XXX Some of these tests are not needed
621	* eg: the tiwin == tp->snd_wnd prevents many more
622	* predictions.. with no real advantage..
623	*/
624	if ( tp->t_state == TCPS_ESTABLISHED
625	&& (tiflags & (TH_SYN\|TH_FIN\|TH_RST\|TH_URG\|TH_ACK)) == TH_ACK
626	/* && (!ts_present \|\| TSTMP_GEQ(ts_val, tp->ts_recent)) */
627	&& ti->ti_seq == tp->rcv_nxt
628	&& tiwin && tiwin == tp->snd_wnd
629	&& tp->snd_nxt == tp->snd_max)
630	{
631	/*
632	* If last ACK falls within this segment's sequence numbers,
633	* record the timestamp.
634	*/
635	#if 0
636	if (ts_present && SEQ_LEQ(ti->ti_seq, tp->last_ack_sent) &&
637	SEQ_LT(tp->last_ack_sent, ti->ti_seq + ti->ti_len))
638	{
639	tp->ts_recent_age = tcp_now;
640	tp->ts_recent = ts_val;
641	}
642	#endif
643
644	if (ti->ti_len == 0)
645	{
646	if ( SEQ_GT(ti->ti_ack, tp->snd_una)
647	&& SEQ_LEQ(ti->ti_ack, tp->snd_max)
648	&& tp->snd_cwnd >= tp->snd_wnd)
649	{
650	/*
651	* this is a pure ack for outstanding data.
652	*/
653	++tcpstat.tcps_predack;
654	#if 0
655	if (ts_present)
656	tcp_xmit_timer(tp, tcp_now-ts_ecr+1);
657	else
658	#endif
659	if ( tp->t_rtt
660	&& SEQ_GT(ti->ti_ack, tp->t_rtseq))
661	tcp_xmit_timer(pData, tp, tp->t_rtt);
662	acked = ti->ti_ack - tp->snd_una;
663	tcpstat.tcps_rcvackpack++;
664	tcpstat.tcps_rcvackbyte += acked;
665	#ifndef VBOX_WITH_SLIRP_BSD_SBUF
666	sbdrop(&so->so_snd, acked);
667	#else
668	if (sbuf_len(&so->so_snd) < acked)
669	/* drop all what sbuf have */
670	sbuf_setpos(&so->so_snd, 0);
671	else
672	sbuf_setpos(&so->so_snd, sbuf_len(&so->so_snd) - acked);
673	#endif
674	tp->snd_una = ti->ti_ack;
675	m_freem(pData, m);
676
677	/*
678	* If all outstanding data are acked, stop
679	* retransmit timer, otherwise restart timer
680	* using current (possibly backed-off) value.
681	* If process is waiting for space,
682	* wakeup/selwakeup/signal. If data
683	* are ready to send, let tcp_output
684	* decide between more output or persist.
685	*/
686	if (tp->snd_una == tp->snd_max)
687	tp->t_timer[TCPT_REXMT] = 0;
688	else if (tp->t_timer[TCPT_PERSIST] == 0)
689	tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
690
691	/*
692	* There's room in so_snd, sowwakup will read()
693	* from the socket if we can
694	*/
695	#if 0
696	if (so->so_snd.sb_flags & SB_NOTIFY)
697	sowwakeup(so);
698	#endif
699	/*
700	* This is called because sowwakeup might have
701	* put data into so_snd. Since we don't so sowwakeup,
702	* we don't need this.. XXX???
703	*/
704	if (SBUF_LEN(&so->so_snd))
705	(void) tcp_output(pData, tp);
706
707	SOCKET_UNLOCK(so);
708	STAM_PROFILE_STOP(&pData->StatTCP_input, counter_input);
709	return;
710	}
711	}
712	else if ( ti->ti_ack == tp->snd_una
713	&& LIST_FIRST(&tp->t_segq)
714	&& ti->ti_len <= sbspace(&so->so_rcv))
715	{
716	/*
717	* this is a pure, in-sequence data packet
718	* with nothing on the reassembly queue and
719	* we have enough buffer space to take it.
720	*/
721	++tcpstat.tcps_preddat;
722	tp->rcv_nxt += ti->ti_len;
723	tcpstat.tcps_rcvpack++;
724	tcpstat.tcps_rcvbyte += ti->ti_len;
725	/*
726	* Add data to socket buffer.
727	*/
728	sbappend(pData, so, m);
729
730	/*
731	* XXX This is called when data arrives. Later, check
732	* if we can actually write() to the socket
733	* XXX Need to check? It's be NON_BLOCKING
734	*/
735	/* sorwakeup(so); */
736
737	/*
738	* If this is a short packet, then ACK now - with Nagel
739	* congestion avoidance sender won't send more until
740	* he gets an ACK.
741	*
742	* It is better to not delay acks at all to maximize
743	* TCP throughput. See RFC 2581.
744	*/
745	tp->t_flags \|= TF_ACKNOW;
746	tcp_output(pData, tp);
747	SOCKET_UNLOCK(so);
748	STAM_PROFILE_STOP(&pData->StatTCP_input, counter_input);
749	return;
750	}
751	} /* header prediction */
752	/*
753	* Calculate amount of space in receive window,
754	* and then do TCP input processing.
755	* Receive window is amount of space in rcv queue,
756	* but not less than advertised window.
757	*/
758	{
759	int win;
760	win = sbspace(&so->so_rcv);
761	if (win < 0)
762	win = 0;
763	tp->rcv_wnd = max(win, (int)(tp->rcv_adv - tp->rcv_nxt));
764	}
765
766	switch (tp->t_state)
767	{
768	/*
769	* If the state is LISTEN then ignore segment if it contains an RST.
770	* If the segment contains an ACK then it is bad and send a RST.
771	* If it does not contain a SYN then it is not interesting; drop it.
772	* Don't bother responding if the destination was a broadcast.
773	* Otherwise initialize tp->rcv_nxt, and tp->irs, select an initial
774	* tp->iss, and send a segment:
775	* <SEQ=ISS><ACK=RCV_NXT><CTL=SYN,ACK>
776	* Also initialize tp->snd_nxt to tp->iss+1 and tp->snd_una to tp->iss.
777	* Fill in remote peer address fields if not previously specified.
778	* Enter SYN_RECEIVED state, and process any other fields of this
779	* segment in this state.
780	*/
781	case TCPS_LISTEN:
782	{
783	if (tiflags & TH_RST)
784	{
785	LogFlowFunc(("%d -> drop\n", __LINE__));
786	goto drop;
787	}
788	if (tiflags & TH_ACK)
789	{
790	LogFlowFunc(("%d -> dropwithreset\n", __LINE__));
791	goto dropwithreset;
792	}
793	if ((tiflags & TH_SYN) == 0)
794	{
795	LogFlowFunc(("%d -> drop\n", __LINE__));
796	goto drop;
797	}
798
799	/*
800	* This has way too many gotos...
801	* But a bit of spaghetti code never hurt anybody :)
802	*/
803	if ( (tcp_fconnect(pData, so) == -1)
804	&& errno != EINPROGRESS
805	&& errno != EWOULDBLOCK)
806	{
807	u_char code = ICMP_UNREACH_NET;
808	Log2((" tcp fconnect errno = %d (%s)\n", errno, strerror(errno)));
809	if (errno == ECONNREFUSED)
810	{
811	/* ACK the SYN, send RST to refuse the connection */
812	tcp_respond(pData, tp, ti, m, ti->ti_seq+1, (tcp_seq)0,
813	TH_RST\|TH_ACK);
814	}
815	else
816	{
817	if (errno == EHOSTUNREACH)
818	code = ICMP_UNREACH_HOST;
819	HTONL(ti->ti_seq); /* restore tcp header */
820	HTONL(ti->ti_ack);
821	HTONS(ti->ti_win);
822	HTONS(ti->ti_urp);
823	m->m_data -= sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
824	m->m_len += sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
825	*ip = save_ip;
826	icmp_error(pData, m, ICMP_UNREACH, code, 0, strerror(errno));
827	tp->t_socket->so_m = NULL;
828	}
829	tp = tcp_close(pData, tp);
830	}
831	else
832	{
833	/*
834	* Haven't connected yet, save the current mbuf
835	* and ti, and return
836	* XXX Some OS's don't tell us whether the connect()
837	* succeeded or not. So we must time it out.
838	*/
839	so->so_m = m;
840	so->so_ti = ti;
841	tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT;
842	TCP_STATE_SWITCH_TO(tp, TCPS_SYN_RECEIVED);
843	}
844	SOCKET_UNLOCK(so);
845	STAM_PROFILE_STOP(&pData->StatTCP_input, counter_input);
846	LogFlowFuncLeave();
847	return;
848
849	cont_conn:
850	/* m==NULL
851	* Check if the connect succeeded
852	*/
853	LogFlowFunc(("cont_conn:\n"));
854	if (so->so_state & SS_NOFDREF)
855	{
856	tp = tcp_close(pData, tp);
857	LogFlowFunc(("%d -> dropwithreset\n", __LINE__));
858	goto dropwithreset;
859	}
860	cont_input:
861	LogFlowFunc(("cont_input:\n"));
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	if ( SEQ_GT(tp->snd_una, ti->ti_ack)
1220	\|\| SEQ_GT(ti->ti_ack, tp->snd_max))
1221	goto dropwithreset;
1222	tcpstat.tcps_connects++;
1223	TCP_STATE_SWITCH_TO(tp, TCPS_ESTABLISHED);
1224	/*
1225	* The sent SYN is ack'ed with our sequence number +1
1226	* The first data byte already in the buffer will get
1227	* lost if no correction is made. This is only needed for
1228	* SS_CTL since the buffer is empty otherwise.
1229	* tp->snd_una++; or:
1230	*/
1231	tp->snd_una = ti->ti_ack;
1232	soisfconnected(so);
1233
1234	/* Do window scaling? */
1235	#if 0
1236	if ( (tp->t_flags & (TF_RCVD_SCALE\|TF_REQ_SCALE))
1237	== (TF_RCVD_SCALE\|TF_REQ_SCALE))
1238	{
1239	tp->snd_scale = tp->requested_s_scale;
1240	tp->rcv_scale = tp->request_r_scale;
1241	}
1242	#endif
1243	(void) tcp_reass(pData, tp, (struct tcphdr )0, (int )0, (struct mbuf *)0);
1244	tp->snd_wl1 = ti->ti_seq - 1;
1245	/* Avoid ack processing; snd_una==ti_ack => dup ack */
1246	LogFlowFunc(("%d -> synrx_to_est\n", __LINE__));
1247	goto synrx_to_est;
1248	/* fall into ... */
1249
1250	/*
1251	* In ESTABLISHED state: drop duplicate ACKs; ACK out of range
1252	* ACKs. If the ack is in the range
1253	* tp->snd_una < ti->ti_ack <= tp->snd_max
1254	* then advance tp->snd_una to ti->ti_ack and drop
1255	* data from the retransmission queue. If this ACK reflects
1256	* more up to date window information we update our window information.
1257	*/
1258	case TCPS_ESTABLISHED:
1259	case TCPS_FIN_WAIT_1:
1260	case TCPS_FIN_WAIT_2:
1261	case TCPS_CLOSE_WAIT:
1262	case TCPS_CLOSING:
1263	case TCPS_LAST_ACK:
1264	case TCPS_TIME_WAIT:
1265	if (SEQ_LEQ(ti->ti_ack, tp->snd_una))
1266	{
1267	if (ti->ti_len == 0 && tiwin == tp->snd_wnd)
1268	{
1269	tcpstat.tcps_rcvdupack++;
1270	Log2((" dup ack m = %lx, so = %lx\n", (long)m, (long)so));
1271	/*
1272	* If we have outstanding data (other than
1273	* a window probe), this is a completely
1274	* duplicate ack (ie, window info didn't
1275	* change), the ack is the biggest we've
1276	* seen and we've seen exactly our rexmt
1277	* threshold of them, assume a packet
1278	* has been dropped and retransmit it.
1279	* Kludge snd_nxt & the congestion
1280	* window so we send only this one
1281	* packet.
1282	*
1283	* We know we're losing at the current
1284	* window size so do congestion avoidance
1285	* (set ssthresh to half the current window
1286	* and pull our congestion window back to
1287	* the new ssthresh).
1288	*
1289	* Dup acks mean that packets have left the
1290	* network (they're now cached at the receiver)
1291	* so bump cwnd by the amount in the receiver
1292	* to keep a constant cwnd packets in the
1293	* network.
1294	*/
1295	if ( tp->t_timer[TCPT_REXMT] == 0
1296	\|\| ti->ti_ack != tp->snd_una)
1297	tp->t_dupacks = 0;
1298	else if (++tp->t_dupacks == tcprexmtthresh)
1299	{
1300	tcp_seq onxt = tp->snd_nxt;
1301	u_int win = min(tp->snd_wnd, tp->snd_cwnd) / 2 / tp->t_maxseg;
1302	if (win < 2)
1303	win = 2;
1304	tp->snd_ssthresh = win * tp->t_maxseg;
1305	tp->t_timer[TCPT_REXMT] = 0;
1306	tp->t_rtt = 0;
1307	tp->snd_nxt = ti->ti_ack;
1308	tp->snd_cwnd = tp->t_maxseg;
1309	(void) tcp_output(pData, tp);
1310	tp->snd_cwnd = tp->snd_ssthresh +
1311	tp->t_maxseg * tp->t_dupacks;
1312	if (SEQ_GT(onxt, tp->snd_nxt))
1313	tp->snd_nxt = onxt;
1314	LogFlowFunc(("%d -> drop\n", __LINE__));
1315	goto drop;
1316	}
1317	else if (tp->t_dupacks > tcprexmtthresh)
1318	{
1319	tp->snd_cwnd += tp->t_maxseg;
1320	(void) tcp_output(pData, tp);
1321	LogFlowFunc(("%d -> drop\n", __LINE__));
1322	goto drop;
1323	}
1324	}
1325	else
1326	tp->t_dupacks = 0;
1327	break;
1328	}
1329	synrx_to_est:
1330	LogFlowFunc(("synrx_to_est:\n"));
1331	/*
1332	* If the congestion window was inflated to account
1333	* for the other side's cached packets, retract it.
1334	*/
1335	if ( tp->t_dupacks > tcprexmtthresh
1336	&& tp->snd_cwnd > tp->snd_ssthresh)
1337	tp->snd_cwnd = tp->snd_ssthresh;
1338	tp->t_dupacks = 0;
1339	if (SEQ_GT(ti->ti_ack, tp->snd_max))
1340	{
1341	tcpstat.tcps_rcvacktoomuch++;
1342	LogFlowFunc(("%d -> dropafterack\n", __LINE__));
1343	goto dropafterack;
1344	}
1345	acked = ti->ti_ack - tp->snd_una;
1346	tcpstat.tcps_rcvackpack++;
1347	tcpstat.tcps_rcvackbyte += acked;
1348
1349	/*
1350	* If we have a timestamp reply, update smoothed
1351	* round trip time. If no timestamp is present but
1352	* transmit timer is running and timed sequence
1353	* number was acked, update smoothed round trip time.
1354	* Since we now have an rtt measurement, cancel the
1355	* timer backoff (cf., Phil Karn's retransmit alg.).
1356	* Recompute the initial retransmit timer.
1357	*/
1358	#if 0
1359	if (ts_present)
1360	tcp_xmit_timer(tp, tcp_now-ts_ecr+1);
1361	else
1362	#endif
1363	if (tp->t_rtt && SEQ_GT(ti->ti_ack, tp->t_rtseq))
1364	tcp_xmit_timer(pData, tp, tp->t_rtt);
1365
1366	/*
1367	* If all outstanding data is acked, stop retransmit
1368	* timer and remember to restart (more output or persist).
1369	* If there is more data to be acked, restart retransmit
1370	* timer, using current (possibly backed-off) value.
1371	*/
1372	if (ti->ti_ack == tp->snd_max)
1373	{
1374	tp->t_timer[TCPT_REXMT] = 0;
1375	needoutput = 1;
1376	}
1377	else if (tp->t_timer[TCPT_PERSIST] == 0)
1378	tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
1379	/*
1380	* When new data is acked, open the congestion window.
1381	* If the window gives us less than ssthresh packets
1382	* in flight, open exponentially (maxseg per packet).
1383	* Otherwise open linearly: maxseg per window
1384	* (maxseg^2 / cwnd per packet).
1385	*/
1386	{
1387	register u_int cw = tp->snd_cwnd;
1388	register u_int incr = tp->t_maxseg;
1389
1390	if (cw > tp->snd_ssthresh)
1391	incr = incr * incr / cw;
1392	tp->snd_cwnd = min(cw + incr, TCP_MAXWIN<<tp->snd_scale);
1393	}
1394	if (acked > SBUF_LEN(&so->so_snd))
1395	{
1396	tp->snd_wnd -= SBUF_LEN(&so->so_snd);
1397	#ifndef VBOX_WITH_SLIRP_BSD_SBUF
1398	sbdrop(&so->so_snd, (int)so->so_snd.sb_cc);
1399	#else
1400	sbuf_clear(&so->so_snd);
1401	#endif
1402	ourfinisacked = 1;
1403	}
1404	else
1405	{
1406	#ifndef VBOX_WITH_SLIRP_BSD_SBUF
1407	sbdrop(&so->so_snd, acked);
1408	#else
1409	sbuf_setpos(&so->so_snd, sbuf_len(&so->so_snd) - acked);
1410	#endif
1411	tp->snd_wnd -= acked;
1412	ourfinisacked = 0;
1413	}
1414	/*
1415	* XXX sowwakup is called when data is acked and there's room for
1416	* for more data... it should read() the socket
1417	*/
1418	#if 0
1419	if (so->so_snd.sb_flags & SB_NOTIFY)
1420	sowwakeup(so);
1421	#endif
1422	tp->snd_una = ti->ti_ack;
1423	if (SEQ_LT(tp->snd_nxt, tp->snd_una))
1424	tp->snd_nxt = tp->snd_una;
1425
1426	switch (tp->t_state)
1427	{
1428	/*
1429	* In FIN_WAIT_1 STATE in addition to the processing
1430	* for the ESTABLISHED state if our FIN is now acknowledged
1431	* then enter FIN_WAIT_2.
1432	*/
1433	case TCPS_FIN_WAIT_1:
1434	if (ourfinisacked)
1435	{
1436	/*
1437	* If we can't receive any more
1438	* data, then closing user can proceed.
1439	* Starting the timer is contrary to the
1440	* specification, but if we don't get a FIN
1441	* we'll hang forever.
1442	*/
1443	if (so->so_state & SS_FCANTRCVMORE)
1444	{
1445	soisfdisconnected(so);
1446	tp->t_timer[TCPT_2MSL] = tcp_maxidle;
1447	}
1448	TCP_STATE_SWITCH_TO(tp, TCPS_FIN_WAIT_2);
1449	}
1450	break;
1451
1452	/*
1453	* In CLOSING STATE in addition to the processing for
1454	* the ESTABLISHED state if the ACK acknowledges our FIN
1455	* then enter the TIME-WAIT state, otherwise ignore
1456	* the segment.
1457	*/
1458	case TCPS_CLOSING:
1459	if (ourfinisacked)
1460	{
1461	TCP_STATE_SWITCH_TO(tp, TCPS_TIME_WAIT);
1462	tcp_canceltimers(tp);
1463	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1464	soisfdisconnected(so);
1465	}
1466	break;
1467
1468	/*
1469	* In LAST_ACK, we may still be waiting for data to drain
1470	* and/or to be acked, as well as for the ack of our FIN.
1471	* If our FIN is now acknowledged, delete the TCB,
1472	* enter the closed state and return.
1473	*/
1474	case TCPS_LAST_ACK:
1475	if (ourfinisacked)
1476	{
1477	tp = tcp_close(pData, tp);
1478	LogFlowFunc(("%d -> drop\n", __LINE__));
1479	goto drop;
1480	}
1481	break;
1482
1483	/*
1484	* In TIME_WAIT state the only thing that should arrive
1485	* is a retransmission of the remote FIN. Acknowledge
1486	* it and restart the finack timer.
1487	*/
1488	case TCPS_TIME_WAIT:
1489	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1490	LogFlowFunc(("%d -> dropafterack\n", __LINE__));
1491	goto dropafterack;
1492	}
1493	} /* switch(tp->t_state) */
1494
1495	step6:
1496	LogFlowFunc(("step6:\n"));
1497	/*
1498	* Update window information.
1499	* Don't look at window if no ACK: TAC's send garbage on first SYN.
1500	*/
1501	if ( (tiflags & TH_ACK)
1502	&& ( SEQ_LT(tp->snd_wl1, ti->ti_seq)
1503	\|\| ( tp->snd_wl1 == ti->ti_seq
1504	&& ( SEQ_LT(tp->snd_wl2, ti->ti_ack)
1505	\|\| ( tp->snd_wl2 == ti->ti_ack
1506	&& tiwin > tp->snd_wnd)))))
1507	{
1508	/* keep track of pure window updates */
1509	if ( ti->ti_len == 0
1510	&& tp->snd_wl2 == ti->ti_ack
1511	&& tiwin > tp->snd_wnd)
1512	tcpstat.tcps_rcvwinupd++;
1513	tp->snd_wnd = tiwin;
1514	tp->snd_wl1 = ti->ti_seq;
1515	tp->snd_wl2 = ti->ti_ack;
1516	if (tp->snd_wnd > tp->max_sndwnd)
1517	tp->max_sndwnd = tp->snd_wnd;
1518	needoutput = 1;
1519	}
1520
1521	/*
1522	* Process segments with URG.
1523	*/
1524	if ((tiflags & TH_URG) && ti->ti_urp &&
1525	TCPS_HAVERCVDFIN(tp->t_state) == 0)
1526	{
1527	/* BSD's sbufs are auto extent so we shouldn't worry here */
1528	#ifndef VBOX_WITH_SLIRP_BSD_SBUF
1529	/*
1530	* This is a kludge, but if we receive and accept
1531	* random urgent pointers, we'll crash in
1532	* soreceive. It's hard to imagine someone
1533	* actually wanting to send this much urgent data.
1534	*/
1535	if (ti->ti_urp + so->so_rcv.sb_cc > so->so_rcv.sb_datalen)
1536	{
1537	ti->ti_urp = 0;
1538	tiflags &= ~TH_URG;
1539	LogFlowFunc(("%d -> dodata\n", __LINE__));
1540	goto dodata;
1541	}
1542	#endif
1543	/*
1544	* If this segment advances the known urgent pointer,
1545	* then mark the data stream. This should not happen
1546	* in CLOSE_WAIT, CLOSING, LAST_ACK or TIME_WAIT STATES since
1547	* a FIN has been received from the remote side.
1548	* In these states we ignore the URG.
1549	*
1550	* According to RFC961 (Assigned Protocols),
1551	* the urgent pointer points to the last octet
1552	* of urgent data. We continue, however,
1553	* to consider it to indicate the first octet
1554	* of data past the urgent section as the original
1555	* spec states (in one of two places).
1556	*/
1557	if (SEQ_GT(ti->ti_seq+ti->ti_urp, tp->rcv_up))
1558	{
1559	tp->rcv_up = ti->ti_seq + ti->ti_urp;
1560	so->so_urgc = SBUF_LEN(&so->so_rcv) +
1561	(tp->rcv_up - tp->rcv_nxt); /* -1; */
1562	tp->rcv_up = ti->ti_seq + ti->ti_urp;
1563	}
1564	}
1565	else
1566	/*
1567	* If no out of band data is expected,
1568	* pull receive urgent pointer along
1569	* with the receive window.
1570	*/
1571	if (SEQ_GT(tp->rcv_nxt, tp->rcv_up))
1572	tp->rcv_up = tp->rcv_nxt;
1573	dodata:
1574	LogFlowFunc(("dodata:\n"));
1575
1576	/*
1577	* If this is a small packet, then ACK now - with Nagel
1578	* congestion avoidance sender won't send more until
1579	* he gets an ACK.
1580	*
1581	* See above.
1582	*/
1583	if ( ti->ti_len
1584	&& (unsigned)ti->ti_len <= 5
1585	&& ((struct tcpiphdr_2 *)ti)->first_char == (char)27)
1586	{
1587	tp->t_flags \|= TF_ACKNOW;
1588	}
1589
1590	/*
1591	* Process the segment text, merging it into the TCP sequencing queue,
1592	* and arranging for acknowledgment of receipt if necessary.
1593	* This process logically involves adjusting tp->rcv_wnd as data
1594	* is presented to the user (this happens in tcp_usrreq.c,
1595	* case PRU_RCVD). If a FIN has already been received on this
1596	* connection then we just ignore the text.
1597	*/
1598	if ( (ti->ti_len \|\| (tiflags&TH_FIN))
1599	&& TCPS_HAVERCVDFIN(tp->t_state) == 0)
1600	{
1601	if ( ti->ti_seq == tp->rcv_nxt
1602	&& LIST_EMPTY(&tp->t_segq)
1603	&& tp->t_state == TCPS_ESTABLISHED)
1604	{
1605	DELAY_ACK(tp, ti); /* little bit different from BSD declaration see netinet/tcp_input.c */
1606	tp->rcv_nxt += tlen;
1607	tiflags = ti->ti_t.th_flags & TH_FIN;
1608	tcpstat.tcps_rcvpack++;
1609	tcpstat.tcps_rcvbyte += tlen;
1610	if (so->so_state & SS_FCANTRCVMORE)
1611	m_freem(pData, m);
1612	else
1613	sbappend(pData, so, m);
1614	}
1615	else
1616	{
1617	tiflags = tcp_reass(pData, tp, &ti->ti_t, &tlen, m);
1618	tiflags \|= TF_ACKNOW;
1619	}
1620	/*
1621	* Note the amount of data that peer has sent into
1622	* our window, in order to estimate the sender's
1623	* buffer size.
1624	*/
1625	len = SBUF_SIZE(&so->so_rcv) - (tp->rcv_adv - tp->rcv_nxt);
1626	}
1627	else
1628	{
1629	m_freem(pData, m);
1630	tiflags &= ~TH_FIN;
1631	}
1632
1633	/*
1634	* If FIN is received ACK the FIN and let the user know
1635	* that the connection is closing.
1636	*/
1637	if (tiflags & TH_FIN)
1638	{
1639	if (TCPS_HAVERCVDFIN(tp->t_state) == 0)
1640	{
1641	/*
1642	* If we receive a FIN we can't send more data,
1643	* set it SS_FDRAIN
1644	* Shutdown the socket if there is no rx data in the
1645	* buffer.
1646	* soread() is called on completion of shutdown() and
1647	* will got to TCPS_LAST_ACK, and use tcp_output()
1648	* to send the FIN.
1649	*/
1650	/* sofcantrcvmore(so); */
1651	sofwdrain(so);
1652
1653	tp->t_flags \|= TF_ACKNOW;
1654	tp->rcv_nxt++;
1655	}
1656	switch (tp->t_state)
1657	{
1658	/*
1659	* In SYN_RECEIVED and ESTABLISHED STATES
1660	* enter the CLOSE_WAIT state.
1661	*/
1662	case TCPS_SYN_RECEIVED:
1663	case TCPS_ESTABLISHED:
1664	TCP_STATE_SWITCH_TO(tp, TCPS_CLOSE_WAIT);
1665	break;
1666
1667	/*
1668	* If still in FIN_WAIT_1 STATE FIN has not been acked so
1669	* enter the CLOSING state.
1670	*/
1671	case TCPS_FIN_WAIT_1:
1672	TCP_STATE_SWITCH_TO(tp, TCPS_CLOSING);
1673	break;
1674
1675	/*
1676	* In FIN_WAIT_2 state enter the TIME_WAIT state,
1677	* starting the time-wait timer, turning off the other
1678	* standard timers.
1679	*/
1680	case TCPS_FIN_WAIT_2:
1681	TCP_STATE_SWITCH_TO(tp, TCPS_TIME_WAIT);
1682	tcp_canceltimers(tp);
1683	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1684	soisfdisconnected(so);
1685	break;
1686
1687	/*
1688	* In TIME_WAIT state restart the 2 MSL time_wait timer.
1689	*/
1690	case TCPS_TIME_WAIT:
1691	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1692	break;
1693	}
1694	}
1695
1696	/*
1697	* Return any desired output.
1698	*/
1699	if (needoutput \|\| (tp->t_flags & TF_ACKNOW))
1700	tcp_output(pData, tp);
1701
1702	SOCKET_UNLOCK(so);
1703	STAM_PROFILE_STOP(&pData->StatTCP_input, counter_input);
1704	LogFlowFuncLeave();
1705	return;
1706
1707	dropafterack:
1708	LogFlowFunc(("dropafterack:\n"));
1709	/*
1710	* Generate an ACK dropping incoming segment if it occupies
1711	* sequence space, where the ACK reflects our state.
1712	*/
1713	if (tiflags & TH_RST)
1714	{
1715	LogFlowFunc(("%d -> drop\n", __LINE__));
1716	goto drop;
1717	}
1718	m_freem(pData, m);
1719	tp->t_flags \|= TF_ACKNOW;
1720	(void) tcp_output(pData, tp);
1721	SOCKET_UNLOCK(so);
1722	STAM_PROFILE_STOP(&pData->StatTCP_input, counter_input);
1723	LogFlowFuncLeave();
1724	return;
1725
1726	dropwithreset:
1727	LogFlowFunc(("dropwithreset:\n"));
1728	/* reuses m if m!=NULL, m_free() unnecessary */
1729	if (tiflags & TH_ACK)
1730	tcp_respond(pData, tp, ti, m, (tcp_seq)0, ti->ti_ack, TH_RST);
1731	else
1732	{
1733	if (tiflags & TH_SYN)
1734	ti->ti_len++;
1735	tcp_respond(pData, tp, ti, m, ti->ti_seq+ti->ti_len, (tcp_seq)0,
1736	TH_RST\|TH_ACK);
1737	}
1738
1739	if (so != &tcb)
1740	SOCKET_UNLOCK(so);
1741	STAM_PROFILE_STOP(&pData->StatTCP_input, counter_input);
1742	LogFlowFuncLeave();
1743	return;
1744
1745	drop:
1746	LogFlowFunc(("drop:\n"));
1747	/*
1748	* Drop space held by incoming segment and return.
1749	*/
1750	m_freem(pData, m);
1751
1752	#ifdef VBOX_WITH_SLIRP_MT
1753	if (RTCritSectIsOwned(&so->so_mutex))
1754	{
1755	SOCKET_UNLOCK(so);
1756	}
1757	#endif
1758
1759	STAM_PROFILE_STOP(&pData->StatTCP_input, counter_input);
1760	LogFlowFuncLeave();
1761	return;
1762	}
1763
1764	void
1765	tcp_dooptions(PNATState pData, struct tcpcb tp, u_char cp, int cnt, struct tcpiphdr *ti)
1766	{
1767	u_int16_t mss;
1768	int opt, optlen;
1769
1770	LogFlow(("tcp_dooptions: tp = %lx, cnt=%i\n", (long)tp, cnt));
1771
1772	for (; cnt > 0; cnt -= optlen, cp += optlen)
1773	{
1774	opt = cp[0];
1775	if (opt == TCPOPT_EOL)
1776	break;
1777	if (opt == TCPOPT_NOP)
1778	optlen = 1;
1779	else
1780	{
1781	optlen = cp[1];
1782	if (optlen <= 0)
1783	break;
1784	}
1785	switch (opt)
1786	{
1787	default:
1788	continue;
1789
1790	case TCPOPT_MAXSEG:
1791	if (optlen != TCPOLEN_MAXSEG)
1792	continue;
1793	if (!(ti->ti_flags & TH_SYN))
1794	continue;
1795	memcpy((char ) &mss, (char ) cp + 2, sizeof(mss));
1796	NTOHS(mss);
1797	(void) tcp_mss(pData, tp, mss); /* sets t_maxseg */
1798	break;
1799
1800	#if 0
1801	case TCPOPT_WINDOW:
1802	if (optlen != TCPOLEN_WINDOW)
1803	continue;
1804	if (!(ti->ti_flags & TH_SYN))
1805	continue;
1806	tp->t_flags \|= TF_RCVD_SCALE;
1807	tp->requested_s_scale = min(cp[2], TCP_MAX_WINSHIFT);
1808	break;
1809
1810	case TCPOPT_TIMESTAMP:
1811	if (optlen != TCPOLEN_TIMESTAMP)
1812	continue;
1813	*ts_present = 1;
1814	memcpy((char ) ts_val, (char )cp + 2, sizeof(*ts_val));
1815	NTOHL(*ts_val);
1816	memcpy((char ) ts_ecr, (char )cp + 6, sizeof(*ts_ecr));
1817	NTOHL(*ts_ecr);
1818
1819	/*
1820	* A timestamp received in a SYN makes
1821	* it ok to send timestamp requests and replies.
1822	*/
1823	if (ti->ti_flags & TH_SYN)
1824	{
1825	tp->t_flags \|= TF_RCVD_TSTMP;
1826	tp->ts_recent = *ts_val;
1827	tp->ts_recent_age = tcp_now;
1828	}
1829	break;
1830	#endif
1831	}
1832	}
1833	}
1834
1835
1836	/*
1837	* Pull out of band byte out of a segment so
1838	* it doesn't appear in the user's data queue.
1839	* It is still reflected in the segment length for
1840	* sequencing purposes.
1841	*/
1842
1843	#if 0
1844	void
1845	tcp_pulloutofband(struct socket so, struct tcpiphdr ti, struct mbuf *m)
1846	{
1847	int cnt = ti->ti_urp - 1;
1848
1849	while (cnt >= 0)
1850	{
1851	if (m->m_len > cnt)
1852	{
1853	char *cp = mtod(m, caddr_t) + cnt;
1854	struct tcpcb *tp = sototcpcb(so);
1855
1856	tp->t_iobc = *cp;
1857	tp->t_oobflags \|= TCPOOB_HAVEDATA;
1858	memcpy(sp, cp+1, (unsigned)(m->m_len - cnt - 1));
1859	m->m_len--;
1860	return;
1861	}
1862	cnt -= m->m_len;
1863	m = m->m_next; /* XXX WRONG! Fix it! */
1864	if (m == 0)
1865	break;
1866	}
1867	panic("tcp_pulloutofband");
1868	}
1869	#endif
1870
1871	/*
1872	* Collect new round-trip time estimate
1873	* and update averages and current timeout.
1874	*/
1875
1876	void
1877	tcp_xmit_timer(PNATState pData, register struct tcpcb *tp, int rtt)
1878	{
1879	register short delta;
1880
1881	LogFlow(("tcp_xmit_timer: tp = %lx rtt = %d\n", (long)tp, rtt));
1882
1883	tcpstat.tcps_rttupdated++;
1884	if (tp->t_srtt != 0)
1885	{
1886	/*
1887	* srtt is stored as fixed point with 3 bits after the
1888	* binary point (i.e., scaled by 8). The following magic
1889	* is equivalent to the smoothing algorithm in rfc793 with
1890	* an alpha of .875 (srtt = rtt/8 + srtt*7/8 in fixed
1891	* point). Adjust rtt to origin 0.
1892	*/
1893	delta = rtt - 1 - (tp->t_srtt >> TCP_RTT_SHIFT);
1894	if ((tp->t_srtt += delta) <= 0)
1895	tp->t_srtt = 1;
1896	/*
1897	* We accumulate a smoothed rtt variance (actually, a
1898	* smoothed mean difference), then set the retransmit
1899	* timer to smoothed rtt + 4 times the smoothed variance.
1900	* rttvar is stored as fixed point with 2 bits after the
1901	* binary point (scaled by 4). The following is
1902	* equivalent to rfc793 smoothing with an alpha of .75
1903	* (rttvar = rttvar*3/4 + \|delta\| / 4). This replaces
1904	* rfc793's wired-in beta.
1905	*/
1906	if (delta < 0)
1907	delta = -delta;
1908	delta -= (tp->t_rttvar >> TCP_RTTVAR_SHIFT);
1909	if ((tp->t_rttvar += delta) <= 0)
1910	tp->t_rttvar = 1;
1911	}
1912	else
1913	{
1914	/*
1915	* No rtt measurement yet - use the unsmoothed rtt.
1916	* Set the variance to half the rtt (so our first
1917	* retransmit happens at 3*rtt).
1918	*/
1919	tp->t_srtt = rtt << TCP_RTT_SHIFT;
1920	tp->t_rttvar = rtt << (TCP_RTTVAR_SHIFT - 1);
1921	}
1922	tp->t_rtt = 0;
1923	tp->t_rxtshift = 0;
1924
1925	/*
1926	* the retransmit should happen at rtt + 4 * rttvar.
1927	* Because of the way we do the smoothing, srtt and rttvar
1928	* will each average +1/2 tick of bias. When we compute
1929	* the retransmit timer, we want 1/2 tick of rounding and
1930	* 1 extra tick because of +-1/2 tick uncertainty in the
1931	* firing of the timer. The bias will give us exactly the
1932	* 1.5 tick we need. But, because the bias is
1933	* statistical, we have to test that we don't drop below
1934	* the minimum feasible timer (which is 2 ticks).
1935	*/
1936	TCPT_RANGESET(tp->t_rxtcur, TCP_REXMTVAL(tp),
1937	(short)tp->t_rttmin, TCPTV_REXMTMAX); /* XXX */
1938
1939	/*
1940	* We received an ack for a packet that wasn't retransmitted;
1941	* it is probably safe to discard any error indications we've
1942	* received recently. This isn't quite right, but close enough
1943	* for now (a route might have failed after we sent a segment,
1944	* and the return path might not be symmetrical).
1945	*/
1946	tp->t_softerror = 0;
1947	}
1948
1949	/*
1950	* Determine a reasonable value for maxseg size.
1951	* If the route is known, check route for mtu.
1952	* If none, use an mss that can be handled on the outgoing
1953	* interface without forcing IP to fragment; if bigger than
1954	* an mbuf cluster (MCLBYTES), round down to nearest multiple of MCLBYTES
1955	* to utilize large mbufs. If no route is found, route has no mtu,
1956	* or the destination isn't local, use a default, hopefully conservative
1957	* size (usually 512 or the default IP max size, but no more than the mtu
1958	* of the interface), as we can't discover anything about intervening
1959	* gateways or networks. We also initialize the congestion/slow start
1960	* window to be a single segment if the destination isn't local.
1961	* While looking at the routing entry, we also initialize other path-dependent
1962	* parameters from pre-set or cached values in the routing entry.
1963	*/
1964
1965	int
1966	tcp_mss(PNATState pData, register struct tcpcb *tp, u_int offer)
1967	{
1968	struct socket *so = tp->t_socket;
1969	int mss;
1970
1971	LogFlow(("tcp_mss: tp = %lx, offet = %d\n", (long)tp, offer));
1972
1973	mss = min(if_mtu, if_mru) - sizeof(struct tcpiphdr);
1974	if (offer)
1975	mss = min(mss, offer);
1976	mss = max(mss, 32);
1977	if (mss < tp->t_maxseg \|\| offer != 0)
1978	tp->t_maxseg = mss;
1979
1980	tp->snd_cwnd = mss;
1981
1982	#ifndef VBOX_WITH_SLIRP_BSD_SBUF
1983	sbreserve(pData, &so->so_snd, tcp_sndspace+((tcp_sndspace%mss)?(mss-(tcp_sndspace%mss)):0));
1984	sbreserve(pData, &so->so_rcv, tcp_rcvspace+((tcp_rcvspace%mss)?(mss-(tcp_rcvspace%mss)):0));
1985	#else
1986	sbuf_new(&so->so_snd, NULL, tcp_sndspace+((tcp_sndspace%mss)?(mss-(tcp_sndspace%mss)):0), SBUF_AUTOEXTEND);
1987	sbuf_new(&so->so_rcv, NULL, tcp_rcvspace+((tcp_rcvspace%mss)?(mss-(tcp_rcvspace%mss)):0), SBUF_AUTOEXTEND);
1988	#endif
1989
1990	Log2((" returning mss = %d\n", mss));
1991
1992	return mss;
1993	}

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

source: vbox/trunk/src/VBox/Devices/Network/slirp/tcp_input.c@ 37910

Download in other formats: