tcp_input.c@ 53814

Last change on this file since 53814 was 52798, checked in by vboxsync, 11 years ago
NAT: when outgoing connect(2) fails decide what to do based on the errno. The code to call icmp_error() is a bit icky because tcp_input() vivisects incoming datagram and original mbuf can't be used.
Property svn:eol-style set to `native` Property svn:keywords set to `Author Date Id Revision`
File size: 67.0 KB

Line
1	/* $Id: tcp_input.c 52798 2014-09-21 21:19:38Z vboxsync $ */
2	/** @file
3	* NAT - TCP input.
4	*/
5
6	/*
7	* Copyright (C) 2006-2012 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 ip, save_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	size_t ohdrlen;
306	uint8_t ohdr[60 + 8]; /* max IP header plus 8 bytes of payload for icmp */
307
308	STAM_PROFILE_START(&pData->StatTCP_input, counter_input);
309
310	LogFlow(("tcp_input: m = %8lx, iphlen = %2d, inso = %R[natsock]\n",
311	(long)m, iphlen, inso));
312
313	if (inso != NULL)
314	{
315	QSOCKET_LOCK(tcb);
316	SOCKET_LOCK(inso);
317	QSOCKET_UNLOCK(tcb);
318	}
319	/*
320	* If called with m == 0, then we're continuing the connect
321	*/
322	if (m == NULL)
323	{
324	so = inso;
325	Log4(("NAT: tcp_input: %R[natsock]\n", so));
326	/* Re-set a few variables */
327	tp = sototcpcb(so);
328	m = so->so_m;
329	so->so_m = 0;
330
331	if (RT_LIKELY(so->so_ohdr != NULL))
332	{
333	RTMemFree(so->so_ohdr);
334	so->so_ohdr = NULL;
335	}
336
337	ti = so->so_ti;
338
339	/** @todo (vvl) clarify why it might happens */
340	if (ti == NULL)
341	{
342	LogRel(("NAT: ti is null. can't do any reseting connection actions\n"));
343	/* mbuf should be cleared in sofree called from tcp_close */
344	tcp_close(pData, tp);
345	STAM_PROFILE_STOP(&pData->StatTCP_input, counter_input);
346	LogFlowFuncLeave();
347	return;
348	}
349
350	tiwin = ti->ti_win;
351	tiflags = ti->ti_flags;
352
353	LogFlowFunc(("%d -> cont_conn\n", __LINE__));
354	goto cont_conn;
355	}
356
357	tcpstat.tcps_rcvtotal++;
358
359	ip = mtod(m, struct ip *);
360
361	/* ip_input() subtracts iphlen from ip::ip_len */
362	AssertStmt((ip->ip_len + iphlen == m_length(m, NULL)), goto drop);
363	if (RT_UNLIKELY(ip->ip_len < sizeof(struct tcphdr)))
364	{
365	/* tcps_rcvshort++; */
366	goto drop;
367	}
368
369	/*
370	* Save a copy of the IP header in case we want to restore it for
371	* sending an ICMP error message in response.
372	*
373	* XXX: This function should really be fixed to not strip IP
374	* options, to not overwrite IP header and to use "tlen" local
375	* variable (instead of ti->ti_len), then "m" could be passed to
376	* icmp_error() directly.
377	*/
378	ohdrlen = iphlen + 8;
379	m_copydata(m, 0, ohdrlen, (caddr_t)ohdr);
380	save_ip = (struct ip *)ohdr;
381	save_ip->ip_len += iphlen; /* undo change by ip_input() */
382
383
384	/*
385	* Get IP and TCP header together in first mbuf.
386	* Note: IP leaves IP header in first mbuf.
387	*/
388	ti = mtod(m, struct tcpiphdr *);
389	if (iphlen > sizeof(struct ip))
390	{
391	ip_stripoptions(m, (struct mbuf *)0);
392	iphlen = sizeof(struct ip);
393	}
394
395	/*
396	* Checksum extended TCP header and data.
397	*/
398	tlen = ((struct ip *)ti)->ip_len;
399	memset(ti->ti_x1, 0, 9);
400	ti->ti_len = RT_H2N_U16((u_int16_t)tlen);
401	len = sizeof(struct ip) + tlen;
402	/* keep checksum for ICMP reply
403	* ti->ti_sum = cksum(m, len);
404	* if (ti->ti_sum) { */
405	if (cksum(m, len))
406	{
407	tcpstat.tcps_rcvbadsum++;
408	LogFlowFunc(("%d -> drop\n", __LINE__));
409	goto drop;
410	}
411
412	/*
413	* Check that TCP offset makes sense,
414	* pull out TCP options and adjust length. XXX
415	*/
416	off = ti->ti_off << 2;
417	if ( off < sizeof (struct tcphdr)
418	\|\| off > tlen)
419	{
420	tcpstat.tcps_rcvbadoff++;
421	LogFlowFunc(("%d -> drop\n", __LINE__));
422	goto drop;
423	}
424	tlen -= off;
425	ti->ti_len = tlen;
426	if (off > sizeof (struct tcphdr))
427	{
428	optlen = off - sizeof (struct tcphdr);
429	optp = mtod(m, caddr_t) + sizeof (struct tcpiphdr);
430
431	/*
432	* Do quick retrieval of timestamp options ("options
433	* prediction?"). If timestamp is the only option and it's
434	* formatted as recommended in RFC 1323 appendix A, we
435	* quickly get the values now and not bother calling
436	* tcp_dooptions(), etc.
437	*/
438	#if 0
439	if (( optlen == TCPOLEN_TSTAMP_APPA
440	\|\| ( optlen > TCPOLEN_TSTAMP_APPA
441	&& optp[TCPOLEN_TSTAMP_APPA] == TCPOPT_EOL)) &&
442	(u_int32_t )optp == RT_H2N_U32_C(TCPOPT_TSTAMP_HDR) &&
443	(ti->ti_flags & TH_SYN) == 0)
444	{
445	ts_present = 1;
446	ts_val = RT_N2H_U32((u_int32_t )(optp + 4));
447	ts_ecr = RT_N2H_U32((u_int32_t )(optp + 8));
448	optp = NULL; / * we have parsed the options * /
449	}
450	#endif
451	}
452	tiflags = ti->ti_flags;
453
454	/*
455	* Convert TCP protocol specific fields to host format.
456	*/
457	NTOHL(ti->ti_seq);
458	NTOHL(ti->ti_ack);
459	NTOHS(ti->ti_win);
460	NTOHS(ti->ti_urp);
461
462	/*
463	* Drop TCP, IP headers and TCP options.
464	*/
465	m->m_data += sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
466	m->m_len -= sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
467
468	/*
469	* Locate pcb for segment.
470	*/
471	findso:
472	LogFlowFunc(("(enter) findso: %R[natsock]\n", so));
473	if (so != NULL && so != &tcb)
474	SOCKET_UNLOCK(so);
475	QSOCKET_LOCK(tcb);
476	so = tcp_last_so;
477	if ( so->so_fport != ti->ti_dport
478	\|\| so->so_lport != ti->ti_sport
479	\|\| so->so_laddr.s_addr != ti->ti_src.s_addr
480	\|\| so->so_faddr.s_addr != ti->ti_dst.s_addr)
481	{
482	QSOCKET_UNLOCK(tcb);
483	/* @todo fix SOLOOKUP macrodefinition to be usable here */
484	so = solookup(&tcb, ti->ti_src, ti->ti_sport,
485	ti->ti_dst, ti->ti_dport);
486	if (so)
487	{
488	tcp_last_so = so;
489	}
490	++tcpstat.tcps_socachemiss;
491	}
492	else
493	{
494	SOCKET_LOCK(so);
495	QSOCKET_UNLOCK(tcb);
496	}
497	LogFlowFunc(("(leave) findso: %R[natsock]\n", so));
498
499	/*
500	* If the state is CLOSED (i.e., TCB does not exist) then
501	* all data in the incoming segment is discarded.
502	* If the TCB exists but is in CLOSED state, it is embryonic,
503	* but should either do a listen or a connect soon.
504	*
505	* state == CLOSED means we've done socreate() but haven't
506	* attached it to a protocol yet...
507	*
508	* XXX If a TCB does not exist, and the TH_SYN flag is
509	* the only flag set, then create a session, mark it
510	* as if it was LISTENING, and continue...
511	*/
512	if (so == 0)
513	{
514	if ((tiflags & (TH_SYN\|TH_FIN\|TH_RST\|TH_URG\|TH_ACK)) != TH_SYN)
515	{
516	LogFlowFunc(("%d -> dropwithreset\n", __LINE__));
517	goto dropwithreset;
518	}
519
520	if ((so = socreate()) == NULL)
521	{
522	LogFlowFunc(("%d -> dropwithreset\n", __LINE__));
523	goto dropwithreset;
524	}
525	if (tcp_attach(pData, so) < 0)
526	{
527	RTMemFree(so); /* Not sofree (if it failed, it's not insqued) */
528	LogFlowFunc(("%d -> dropwithreset\n", __LINE__));
529	goto dropwithreset;
530	}
531	SOCKET_LOCK(so);
532	#ifndef VBOX_WITH_SLIRP_BSD_SBUF
533	sbreserve(pData, &so->so_snd, tcp_sndspace);
534	sbreserve(pData, &so->so_rcv, tcp_rcvspace);
535	#else
536	sbuf_new(&so->so_snd, NULL, tcp_sndspace, SBUF_AUTOEXTEND);
537	sbuf_new(&so->so_rcv, NULL, tcp_rcvspace, SBUF_AUTOEXTEND);
538	#endif
539
540	/* tcp_last_so = so; / / XXX ? */
541	/* tp = sototcpcb(so); */
542
543	so->so_laddr = ti->ti_src;
544	so->so_lport = ti->ti_sport;
545	so->so_faddr = ti->ti_dst;
546	so->so_fport = ti->ti_dport;
547
548	so->so_iptos = ((struct ip *)ti)->ip_tos;
549
550	tp = sototcpcb(so);
551	TCP_STATE_SWITCH_TO(tp, TCPS_LISTEN);
552	}
553
554	/*
555	* If this is a still-connecting socket, this probably
556	* a retransmit of the SYN. Whether it's a retransmit SYN
557	* or something else, we nuke it.
558	*/
559	if (so->so_state & SS_ISFCONNECTING)
560	{
561	LogFlowFunc(("%d -> drop\n", __LINE__));
562	goto drop;
563	}
564
565	tp = sototcpcb(so);
566
567	/* XXX Should never fail */
568	if (tp == 0)
569	{
570	LogFlowFunc(("%d -> dropwithreset\n", __LINE__));
571	goto dropwithreset;
572	}
573	if (tp->t_state == TCPS_CLOSED)
574	{
575	LogFlowFunc(("%d -> drop\n", __LINE__));
576	goto drop;
577	}
578
579	/* Unscale the window into a 32-bit value. */
580	/* if ((tiflags & TH_SYN) == 0)
581	* tiwin = ti->ti_win << tp->snd_scale;
582	* else
583	*/
584	tiwin = ti->ti_win;
585
586	/*
587	* Segment received on connection.
588	* Reset idle time and keep-alive timer.
589	*/
590	tp->t_idle = 0;
591	if (so_options)
592	tp->t_timer[TCPT_KEEP] = tcp_keepintvl;
593	else
594	tp->t_timer[TCPT_KEEP] = tcp_keepidle;
595
596	/*
597	* Process options if not in LISTEN state,
598	* else do it below (after getting remote address).
599	*/
600	if (optp && tp->t_state != TCPS_LISTEN)
601	tcp_dooptions(pData, tp, (u_char *)optp, optlen, ti);
602	/* , */
603	/* &ts_present, &ts_val, &ts_ecr); */
604
605	/*
606	* Header prediction: check for the two common cases
607	* of a uni-directional data xfer. If the packet has
608	* no control flags, is in-sequence, the window didn't
609	* change and we're not retransmitting, it's a
610	* candidate. If the length is zero and the ack moved
611	* forward, we're the sender side of the xfer. Just
612	* free the data acked & wake any higher level process
613	* that was blocked waiting for space. If the length
614	* is non-zero and the ack didn't move, we're the
615	* receiver side. If we're getting packets in-order
616	* (the reassembly queue is empty), add the data to
617	* the socket buffer and note that we need a delayed ack.
618	*
619	* XXX Some of these tests are not needed
620	* eg: the tiwin == tp->snd_wnd prevents many more
621	* predictions.. with no real advantage..
622	*/
623	if ( tp->t_state == TCPS_ESTABLISHED
624	&& (tiflags & (TH_SYN\|TH_FIN\|TH_RST\|TH_URG\|TH_ACK)) == TH_ACK
625	/* && (!ts_present \|\| TSTMP_GEQ(ts_val, tp->ts_recent)) */
626	&& ti->ti_seq == tp->rcv_nxt
627	&& tiwin && tiwin == tp->snd_wnd
628	&& tp->snd_nxt == tp->snd_max)
629	{
630	/*
631	* If last ACK falls within this segment's sequence numbers,
632	* record the timestamp.
633	*/
634	#if 0
635	if (ts_present && SEQ_LEQ(ti->ti_seq, tp->last_ack_sent) &&
636	SEQ_LT(tp->last_ack_sent, ti->ti_seq + ti->ti_len))
637	{
638	tp->ts_recent_age = tcp_now;
639	tp->ts_recent = ts_val;
640	}
641	#endif
642
643	if (ti->ti_len == 0)
644	{
645	if ( SEQ_GT(ti->ti_ack, tp->snd_una)
646	&& SEQ_LEQ(ti->ti_ack, tp->snd_max)
647	&& tp->snd_cwnd >= tp->snd_wnd)
648	{
649	/*
650	* this is a pure ack for outstanding data.
651	*/
652	++tcpstat.tcps_predack;
653	#if 0
654	if (ts_present)
655	tcp_xmit_timer(tp, tcp_now-ts_ecr+1);
656	else
657	#endif
658	if ( tp->t_rtt
659	&& SEQ_GT(ti->ti_ack, tp->t_rtseq))
660	tcp_xmit_timer(pData, tp, tp->t_rtt);
661	acked = ti->ti_ack - tp->snd_una;
662	tcpstat.tcps_rcvackpack++;
663	tcpstat.tcps_rcvackbyte += acked;
664	#ifndef VBOX_WITH_SLIRP_BSD_SBUF
665	sbdrop(&so->so_snd, acked);
666	#else
667	if (sbuf_len(&so->so_snd) < acked)
668	/* drop all what sbuf have */
669	sbuf_setpos(&so->so_snd, 0);
670	else
671	sbuf_setpos(&so->so_snd, sbuf_len(&so->so_snd) - acked);
672	#endif
673	tp->snd_una = ti->ti_ack;
674	m_freem(pData, m);
675
676	/*
677	* If all outstanding data are acked, stop
678	* retransmit timer, otherwise restart timer
679	* using current (possibly backed-off) value.
680	* If process is waiting for space,
681	* wakeup/selwakeup/signal. If data
682	* are ready to send, let tcp_output
683	* decide between more output or persist.
684	*/
685	if (tp->snd_una == tp->snd_max)
686	tp->t_timer[TCPT_REXMT] = 0;
687	else if (tp->t_timer[TCPT_PERSIST] == 0)
688	tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
689
690	/*
691	* There's room in so_snd, sowwakup will read()
692	* from the socket if we can
693	*/
694	#if 0
695	if (so->so_snd.sb_flags & SB_NOTIFY)
696	sowwakeup(so);
697	#endif
698	/*
699	* This is called because sowwakeup might have
700	* put data into so_snd. Since we don't so sowwakeup,
701	* we don't need this.. XXX???
702	*/
703	if (SBUF_LEN(&so->so_snd))
704	(void) tcp_output(pData, tp);
705
706	SOCKET_UNLOCK(so);
707	STAM_PROFILE_STOP(&pData->StatTCP_input, counter_input);
708	return;
709	}
710	}
711	else if ( ti->ti_ack == tp->snd_una
712	&& LIST_FIRST(&tp->t_segq)
713	&& ti->ti_len <= sbspace(&so->so_rcv))
714	{
715	/*
716	* this is a pure, in-sequence data packet
717	* with nothing on the reassembly queue and
718	* we have enough buffer space to take it.
719	*/
720	++tcpstat.tcps_preddat;
721	tp->rcv_nxt += ti->ti_len;
722	tcpstat.tcps_rcvpack++;
723	tcpstat.tcps_rcvbyte += ti->ti_len;
724	/*
725	* Add data to socket buffer.
726	*/
727	sbappend(pData, so, m);
728
729	/*
730	* XXX This is called when data arrives. Later, check
731	* if we can actually write() to the socket
732	* XXX Need to check? It's be NON_BLOCKING
733	*/
734	/* sorwakeup(so); */
735
736	/*
737	* If this is a short packet, then ACK now - with Nagel
738	* congestion avoidance sender won't send more until
739	* he gets an ACK.
740	*
741	* It is better to not delay acks at all to maximize
742	* TCP throughput. See RFC 2581.
743	*/
744	tp->t_flags \|= TF_ACKNOW;
745	tcp_output(pData, tp);
746	SOCKET_UNLOCK(so);
747	STAM_PROFILE_STOP(&pData->StatTCP_input, counter_input);
748	return;
749	}
750	} /* header prediction */
751	/*
752	* Calculate amount of space in receive window,
753	* and then do TCP input processing.
754	* Receive window is amount of space in rcv queue,
755	* but not less than advertised window.
756	*/
757	{
758	int win;
759	win = sbspace(&so->so_rcv);
760	if (win < 0)
761	win = 0;
762	tp->rcv_wnd = max(win, (int)(tp->rcv_adv - tp->rcv_nxt));
763	}
764
765	switch (tp->t_state)
766	{
767	/*
768	* If the state is LISTEN then ignore segment if it contains an RST.
769	* If the segment contains an ACK then it is bad and send a RST.
770	* If it does not contain a SYN then it is not interesting; drop it.
771	* Don't bother responding if the destination was a broadcast.
772	* Otherwise initialize tp->rcv_nxt, and tp->irs, select an initial
773	* tp->iss, and send a segment:
774	* <SEQ=ISS><ACK=RCV_NXT><CTL=SYN,ACK>
775	* Also initialize tp->snd_nxt to tp->iss+1 and tp->snd_una to tp->iss.
776	* Fill in remote peer address fields if not previously specified.
777	* Enter SYN_RECEIVED state, and process any other fields of this
778	* segment in this state.
779	*/
780	case TCPS_LISTEN:
781	{
782	if (tiflags & TH_RST)
783	{
784	LogFlowFunc(("%d -> drop\n", __LINE__));
785	goto drop;
786	}
787	if (tiflags & TH_ACK)
788	{
789	LogFlowFunc(("%d -> dropwithreset\n", __LINE__));
790	goto dropwithreset;
791	}
792	if ((tiflags & TH_SYN) == 0)
793	{
794	LogFlowFunc(("%d -> drop\n", __LINE__));
795	goto drop;
796	}
797
798	/*
799	* This has way too many gotos...
800	* But a bit of spaghetti code never hurt anybody :)
801	*/
802	if ( (tcp_fconnect(pData, so) == -1)
803	&& errno != EINPROGRESS
804	&& errno != EWOULDBLOCK)
805	{
806	u_char code = ICMP_UNREACH_NET;
807	Log2((" tcp fconnect errno = %d (%s)\n", errno, strerror(errno)));
808	if (errno == ECONNREFUSED)
809	{
810	/* ACK the SYN, send RST to refuse the connection */
811	tcp_respond(pData, tp, ti, m, ti->ti_seq+1, (tcp_seq)0,
812	TH_RST\|TH_ACK);
813	}
814	else
815	{
816	if (errno == EHOSTUNREACH)
817	code = ICMP_UNREACH_HOST;
818	HTONL(ti->ti_seq); /* restore tcp header */
819	HTONL(ti->ti_ack);
820	HTONS(ti->ti_win);
821	HTONS(ti->ti_urp);
822	m->m_data -= sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
823	m->m_len += sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
824	ip = save_ip;
825	icmp_error(pData, m, ICMP_UNREACH, code, 0, strerror(errno));
826	tp->t_socket->so_m = NULL;
827	}
828	tp = tcp_close(pData, tp);
829	}
830	else
831	{
832	/*
833	* Haven't connected yet, save the current mbuf
834	* and ti, and return
835	* XXX Some OS's don't tell us whether the connect()
836	* succeeded or not. So we must time it out.
837	*/
838	so->so_m = m;
839	so->so_ti = ti;
840	so->so_ohdr = RTMemDup(ohdr, ohdrlen);
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
861	tcp_template(tp);
862
863	if (optp)
864	tcp_dooptions(pData, tp, (u_char *)optp, optlen, ti);
865
866	if (iss)
867	tp->iss = iss;
868	else
869	tp->iss = tcp_iss;
870	tcp_iss += TCP_ISSINCR/2;
871	tp->irs = ti->ti_seq;
872	tcp_sendseqinit(tp);
873	tcp_rcvseqinit(tp);
874	tp->t_flags \|= TF_ACKNOW;
875	TCP_STATE_SWITCH_TO(tp, TCPS_SYN_RECEIVED);
876	tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT;
877	tcpstat.tcps_accepts++;
878	LogFlowFunc(("%d -> trimthenstep6\n", __LINE__));
879	goto trimthenstep6;
880	} /* case TCPS_LISTEN */
881
882	/*
883	* If the state is SYN_SENT:
884	* if seg contains an ACK, but not for our SYN, drop the input.
885	* if seg contains a RST, then drop the connection.
886	* if seg does not contain SYN, then drop it.
887	* Otherwise this is an acceptable SYN segment
888	* initialize tp->rcv_nxt and tp->irs
889	* if seg contains ack then advance tp->snd_una
890	* if SYN has been acked change to ESTABLISHED else SYN_RCVD state
891	* arrange for segment to be acked (eventually)
892	* continue processing rest of data/controls, beginning with URG
893	*/
894	case TCPS_SYN_SENT:
895	if ( (tiflags & TH_ACK)
896	&& ( SEQ_LEQ(ti->ti_ack, tp->iss)
897	\|\| SEQ_GT(ti->ti_ack, tp->snd_max)))
898	{
899	LogFlowFunc(("%d -> dropwithreset\n", __LINE__));
900	goto dropwithreset;
901	}
902
903	if (tiflags & TH_RST)
904	{
905	if (tiflags & TH_ACK)
906	tp = tcp_drop(pData, tp, 0); /* XXX Check t_softerror! */
907	LogFlowFunc(("%d -> drop\n", __LINE__));
908	goto drop;
909	}
910
911	if ((tiflags & TH_SYN) == 0)
912	{
913	LogFlowFunc(("%d -> drop\n", __LINE__));
914	goto drop;
915	}
916	if (tiflags & TH_ACK)
917	{
918	tp->snd_una = ti->ti_ack;
919	if (SEQ_LT(tp->snd_nxt, tp->snd_una))
920	tp->snd_nxt = tp->snd_una;
921	}
922
923	tp->t_timer[TCPT_REXMT] = 0;
924	tp->irs = ti->ti_seq;
925	tcp_rcvseqinit(tp);
926	tp->t_flags \|= TF_ACKNOW;
927	if (tiflags & TH_ACK && SEQ_GT(tp->snd_una, tp->iss))
928	{
929	tcpstat.tcps_connects++;
930	soisfconnected(so);
931	TCP_STATE_SWITCH_TO(tp, TCPS_ESTABLISHED);
932
933	/* Do window scaling on this connection? */
934	#if 0
935	if (( tp->t_flags & (TF_RCVD_SCALE\|TF_REQ_SCALE))
936	== (TF_RCVD_SCALE\|TF_REQ_SCALE))
937	{
938	tp->snd_scale = tp->requested_s_scale;
939	tp->rcv_scale = tp->request_r_scale;
940	}
941	#endif
942	(void) tcp_reass(pData, tp, (struct tcphdr )0, NULL, (struct mbuf )0);
943	/*
944	* if we didn't have to retransmit the SYN,
945	* use its rtt as our initial srtt & rtt var.
946	*/
947	if (tp->t_rtt)
948	tcp_xmit_timer(pData, tp, tp->t_rtt);
949	}
950	else
951	TCP_STATE_SWITCH_TO(tp, TCPS_SYN_RECEIVED);
952
953	trimthenstep6:
954	LogFlowFunc(("trimthenstep6:\n"));
955	/*
956	* Advance ti->ti_seq to correspond to first data byte.
957	* If data, trim to stay within window,
958	* dropping FIN if necessary.
959	*/
960	ti->ti_seq++;
961	if (ti->ti_len > tp->rcv_wnd)
962	{
963	todrop = ti->ti_len - tp->rcv_wnd;
964	m_adj(m, -todrop);
965	ti->ti_len = tp->rcv_wnd;
966	tiflags &= ~TH_FIN;
967	tcpstat.tcps_rcvpackafterwin++;
968	tcpstat.tcps_rcvbyteafterwin += todrop;
969	}
970	tp->snd_wl1 = ti->ti_seq - 1;
971	tp->rcv_up = ti->ti_seq;
972	LogFlowFunc(("%d -> step6\n", __LINE__));
973	goto step6;
974	} /* switch tp->t_state */
975	/*
976	* States other than LISTEN or SYN_SENT.
977	* First check timestamp, if present.
978	* Then check that at least some bytes of segment are within
979	* receive window. If segment begins before rcv_nxt,
980	* drop leading data (and SYN); if nothing left, just ack.
981	*
982	* RFC 1323 PAWS: If we have a timestamp reply on this segment
983	* and it's less than ts_recent, drop it.
984	*/
985	#if 0
986	if ( ts_present
987	&& (tiflags & TH_RST) == 0
988	&& tp->ts_recent
989	&& TSTMP_LT(ts_val, tp->ts_recent))
990	{
991	/* Check to see if ts_recent is over 24 days old. */
992	if ((int)(tcp_now - tp->ts_recent_age) > TCP_PAWS_IDLE)
993	{
994	/*
995	* Invalidate ts_recent. If this segment updates
996	* ts_recent, the age will be reset later and ts_recent
997	* will get a valid value. If it does not, setting
998	* ts_recent to zero will at least satisfy the
999	* requirement that zero be placed in the timestamp
1000	* echo reply when ts_recent isn't valid. The
1001	* age isn't reset until we get a valid ts_recent
1002	* because we don't want out-of-order segments to be
1003	* dropped when ts_recent is old.
1004	*/
1005	tp->ts_recent = 0;
1006	}
1007	else
1008	{
1009	tcpstat.tcps_rcvduppack++;
1010	tcpstat.tcps_rcvdupbyte += ti->ti_len;
1011	tcpstat.tcps_pawsdrop++;
1012	goto dropafterack;
1013	}
1014	}
1015	#endif
1016
1017	todrop = tp->rcv_nxt - ti->ti_seq;
1018	if (todrop > 0)
1019	{
1020	if (tiflags & TH_SYN)
1021	{
1022	tiflags &= ~TH_SYN;
1023	ti->ti_seq++;
1024	if (ti->ti_urp > 1)
1025	ti->ti_urp--;
1026	else
1027	tiflags &= ~TH_URG;
1028	todrop--;
1029	}
1030	/*
1031	* Following if statement from Stevens, vol. 2, p. 960.
1032	*/
1033	if ( todrop > ti->ti_len
1034	\|\| ( todrop == ti->ti_len
1035	&& (tiflags & TH_FIN) == 0))
1036	{
1037	/*
1038	* Any valid FIN must be to the left of the window.
1039	* At this point the FIN must be a duplicate or out
1040	* of sequence; drop it.
1041	*/
1042	tiflags &= ~TH_FIN;
1043
1044	/*
1045	* Send an ACK to resynchronize and drop any data.
1046	* But keep on processing for RST or ACK.
1047	*/
1048	tp->t_flags \|= TF_ACKNOW;
1049	todrop = ti->ti_len;
1050	tcpstat.tcps_rcvduppack++;
1051	tcpstat.tcps_rcvdupbyte += todrop;
1052	}
1053	else
1054	{
1055	tcpstat.tcps_rcvpartduppack++;
1056	tcpstat.tcps_rcvpartdupbyte += todrop;
1057	}
1058	m_adj(m, todrop);
1059	ti->ti_seq += todrop;
1060	ti->ti_len -= todrop;
1061	if (ti->ti_urp > todrop)
1062	ti->ti_urp -= todrop;
1063	else
1064	{
1065	tiflags &= ~TH_URG;
1066	ti->ti_urp = 0;
1067	}
1068	}
1069	/*
1070	* If new data are received on a connection after the
1071	* user processes are gone, then RST the other end.
1072	*/
1073	if ( (so->so_state & SS_NOFDREF)
1074	&& tp->t_state > TCPS_CLOSE_WAIT && ti->ti_len)
1075	{
1076	tp = tcp_close(pData, tp);
1077	tcpstat.tcps_rcvafterclose++;
1078	LogFlowFunc(("%d -> dropwithreset\n", __LINE__));
1079	goto dropwithreset;
1080	}
1081
1082	/*
1083	* If segment ends after window, drop trailing data
1084	* (and PUSH and FIN); if nothing left, just ACK.
1085	*/
1086	todrop = (ti->ti_seq+ti->ti_len) - (tp->rcv_nxt+tp->rcv_wnd);
1087	if (todrop > 0)
1088	{
1089	tcpstat.tcps_rcvpackafterwin++;
1090	if (todrop >= ti->ti_len)
1091	{
1092	tcpstat.tcps_rcvbyteafterwin += ti->ti_len;
1093	/*
1094	* If a new connection request is received
1095	* while in TIME_WAIT, drop the old connection
1096	* and start over if the sequence numbers
1097	* are above the previous ones.
1098	*/
1099	if ( tiflags & TH_SYN
1100	&& tp->t_state == TCPS_TIME_WAIT
1101	&& SEQ_GT(ti->ti_seq, tp->rcv_nxt))
1102	{
1103	iss = tp->rcv_nxt + TCP_ISSINCR;
1104	tp = tcp_close(pData, tp);
1105	SOCKET_UNLOCK(tp->t_socket);
1106	LogFlowFunc(("%d -> findso\n", __LINE__));
1107	goto findso;
1108	}
1109	/*
1110	* If window is closed can only take segments at
1111	* window edge, and have to drop data and PUSH from
1112	* incoming segments. Continue processing, but
1113	* remember to ack. Otherwise, drop segment
1114	* and ack.
1115	*/
1116	if (tp->rcv_wnd == 0 && ti->ti_seq == tp->rcv_nxt)
1117	{
1118	tp->t_flags \|= TF_ACKNOW;
1119	tcpstat.tcps_rcvwinprobe++;
1120	}
1121	else
1122	{
1123	LogFlowFunc(("%d -> dropafterack\n", __LINE__));
1124	goto dropafterack;
1125	}
1126	}
1127	else
1128	tcpstat.tcps_rcvbyteafterwin += todrop;
1129	m_adj(m, -todrop);
1130	ti->ti_len -= todrop;
1131	tiflags &= ~(TH_PUSH\|TH_FIN);
1132	}
1133
1134	/*
1135	* If last ACK falls within this segment's sequence numbers,
1136	* record its timestamp.
1137	*/
1138	#if 0
1139	if ( ts_present
1140	&& SEQ_LEQ(ti->ti_seq, tp->last_ack_sent)
1141	&& SEQ_LT(tp->last_ack_sent, ti->ti_seq + ti->ti_len + ((tiflags & (TH_SYN\|TH_FIN)) != 0)))
1142	{
1143	tp->ts_recent_age = tcp_now;
1144	tp->ts_recent = ts_val;
1145	}
1146	#endif
1147
1148	/*
1149	* If the RST bit is set examine the state:
1150	* SYN_RECEIVED STATE:
1151	* If passive open, return to LISTEN state.
1152	* If active open, inform user that connection was refused.
1153	* ESTABLISHED, FIN_WAIT_1, FIN_WAIT2, CLOSE_WAIT STATES:
1154	* Inform user that connection was reset, and close tcb.
1155	* CLOSING, LAST_ACK, TIME_WAIT STATES
1156	* Close the tcb.
1157	*/
1158	if (tiflags&TH_RST)
1159	switch (tp->t_state)
1160	{
1161	case TCPS_SYN_RECEIVED:
1162	/* so->so_error = ECONNREFUSED; */
1163	LogFlowFunc(("%d -> close\n", __LINE__));
1164	goto close;
1165
1166	case TCPS_ESTABLISHED:
1167	case TCPS_FIN_WAIT_1:
1168	case TCPS_FIN_WAIT_2:
1169	case TCPS_CLOSE_WAIT:
1170	/* so->so_error = ECONNRESET; */
1171	close:
1172	LogFlowFunc(("close:\n"));
1173	TCP_STATE_SWITCH_TO(tp, TCPS_CLOSED);
1174	tcpstat.tcps_drops++;
1175	tp = tcp_close(pData, tp);
1176	LogFlowFunc(("%d -> drop\n", __LINE__));
1177	goto drop;
1178
1179	case TCPS_CLOSING:
1180	case TCPS_LAST_ACK:
1181	case TCPS_TIME_WAIT:
1182	tp = tcp_close(pData, tp);
1183	LogFlowFunc(("%d -> drop\n", __LINE__));
1184	goto drop;
1185	}
1186
1187	/*
1188	* If a SYN is in the window, then this is an
1189	* error and we send an RST and drop the connection.
1190	*/
1191	if (tiflags & TH_SYN)
1192	{
1193	tp = tcp_drop(pData, tp, 0);
1194	LogFlowFunc(("%d -> dropwithreset\n", __LINE__));
1195	goto dropwithreset;
1196	}
1197
1198	/*
1199	* If the ACK bit is off we drop the segment and return.
1200	*/
1201	if ((tiflags & TH_ACK) == 0)
1202	{
1203	LogFlowFunc(("%d -> drop\n", __LINE__));
1204	goto drop;
1205	}
1206
1207	/*
1208	* Ack processing.
1209	*/
1210	switch (tp->t_state)
1211	{
1212	/*
1213	* In SYN_RECEIVED state if the ack ACKs our SYN then enter
1214	* ESTABLISHED state and continue processing, otherwise
1215	* send an RST. una<=ack<=max
1216	*/
1217	case TCPS_SYN_RECEIVED:
1218	LogFlowFunc(("%d -> TCPS_SYN_RECEIVED\n", __LINE__));
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	LogFlowFunc(("%d -> TCPS_ESTABLISHED\|TCPS_FIN_WAIT_1\|TCPS_FIN_WAIT_2\|TCPS_CLOSE_WAIT\|"
1266	"TCPS_CLOSING\|TCPS_LAST_ACK\|TCPS_TIME_WAIT\n", __LINE__));
1267	if (SEQ_LEQ(ti->ti_ack, tp->snd_una))
1268	{
1269	if (ti->ti_len == 0 && tiwin == tp->snd_wnd)
1270	{
1271	tcpstat.tcps_rcvdupack++;
1272	Log2((" dup ack m = %lx, so = %lx\n", (long)m, (long)so));
1273	/*
1274	* If we have outstanding data (other than
1275	* a window probe), this is a completely
1276	* duplicate ack (ie, window info didn't
1277	* change), the ack is the biggest we've
1278	* seen and we've seen exactly our rexmt
1279	* threshold of them, assume a packet
1280	* has been dropped and retransmit it.
1281	* Kludge snd_nxt & the congestion
1282	* window so we send only this one
1283	* packet.
1284	*
1285	* We know we're losing at the current
1286	* window size so do congestion avoidance
1287	* (set ssthresh to half the current window
1288	* and pull our congestion window back to
1289	* the new ssthresh).
1290	*
1291	* Dup acks mean that packets have left the
1292	* network (they're now cached at the receiver)
1293	* so bump cwnd by the amount in the receiver
1294	* to keep a constant cwnd packets in the
1295	* network.
1296	*/
1297	if ( tp->t_timer[TCPT_REXMT] == 0
1298	\|\| ti->ti_ack != tp->snd_una)
1299	tp->t_dupacks = 0;
1300	else if (++tp->t_dupacks == tcprexmtthresh)
1301	{
1302	tcp_seq onxt = tp->snd_nxt;
1303	u_int win = min(tp->snd_wnd, tp->snd_cwnd) / 2 / tp->t_maxseg;
1304	if (win < 2)
1305	win = 2;
1306	tp->snd_ssthresh = win * tp->t_maxseg;
1307	tp->t_timer[TCPT_REXMT] = 0;
1308	tp->t_rtt = 0;
1309	tp->snd_nxt = ti->ti_ack;
1310	tp->snd_cwnd = tp->t_maxseg;
1311	(void) tcp_output(pData, tp);
1312	tp->snd_cwnd = tp->snd_ssthresh +
1313	tp->t_maxseg * tp->t_dupacks;
1314	if (SEQ_GT(onxt, tp->snd_nxt))
1315	tp->snd_nxt = onxt;
1316	LogFlowFunc(("%d -> drop\n", __LINE__));
1317	goto drop;
1318	}
1319	else if (tp->t_dupacks > tcprexmtthresh)
1320	{
1321	tp->snd_cwnd += tp->t_maxseg;
1322	(void) tcp_output(pData, tp);
1323	LogFlowFunc(("%d -> drop\n", __LINE__));
1324	goto drop;
1325	}
1326	}
1327	else
1328	tp->t_dupacks = 0;
1329	break;
1330	}
1331	synrx_to_est:
1332	LogFlowFunc(("synrx_to_est:\n"));
1333	/*
1334	* If the congestion window was inflated to account
1335	* for the other side's cached packets, retract it.
1336	*/
1337	if ( tp->t_dupacks > tcprexmtthresh
1338	&& tp->snd_cwnd > tp->snd_ssthresh)
1339	tp->snd_cwnd = tp->snd_ssthresh;
1340	tp->t_dupacks = 0;
1341	if (SEQ_GT(ti->ti_ack, tp->snd_max))
1342	{
1343	tcpstat.tcps_rcvacktoomuch++;
1344	LogFlowFunc(("%d -> dropafterack\n", __LINE__));
1345	goto dropafterack;
1346	}
1347	acked = ti->ti_ack - tp->snd_una;
1348	tcpstat.tcps_rcvackpack++;
1349	tcpstat.tcps_rcvackbyte += acked;
1350
1351	/*
1352	* If we have a timestamp reply, update smoothed
1353	* round trip time. If no timestamp is present but
1354	* transmit timer is running and timed sequence
1355	* number was acked, update smoothed round trip time.
1356	* Since we now have an rtt measurement, cancel the
1357	* timer backoff (cf., Phil Karn's retransmit alg.).
1358	* Recompute the initial retransmit timer.
1359	*/
1360	#if 0
1361	if (ts_present)
1362	tcp_xmit_timer(tp, tcp_now-ts_ecr+1);
1363	else
1364	#endif
1365	if (tp->t_rtt && SEQ_GT(ti->ti_ack, tp->t_rtseq))
1366	tcp_xmit_timer(pData, tp, tp->t_rtt);
1367
1368	/*
1369	* If all outstanding data is acked, stop retransmit
1370	* timer and remember to restart (more output or persist).
1371	* If there is more data to be acked, restart retransmit
1372	* timer, using current (possibly backed-off) value.
1373	*/
1374	if (ti->ti_ack == tp->snd_max)
1375	{
1376	tp->t_timer[TCPT_REXMT] = 0;
1377	needoutput = 1;
1378	}
1379	else if (tp->t_timer[TCPT_PERSIST] == 0)
1380	tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
1381	/*
1382	* When new data is acked, open the congestion window.
1383	* If the window gives us less than ssthresh packets
1384	* in flight, open exponentially (maxseg per packet).
1385	* Otherwise open linearly: maxseg per window
1386	* (maxseg^2 / cwnd per packet).
1387	*/
1388	{
1389	register u_int cw = tp->snd_cwnd;
1390	register u_int incr = tp->t_maxseg;
1391
1392	if (cw > tp->snd_ssthresh)
1393	incr = incr * incr / cw;
1394	tp->snd_cwnd = min(cw + incr, TCP_MAXWIN<<tp->snd_scale);
1395	}
1396	if (acked > SBUF_LEN(&so->so_snd))
1397	{
1398	tp->snd_wnd -= SBUF_LEN(&so->so_snd);
1399	#ifndef VBOX_WITH_SLIRP_BSD_SBUF
1400	sbdrop(&so->so_snd, (int)so->so_snd.sb_cc);
1401	#else
1402	sbuf_clear(&so->so_snd);
1403	#endif
1404	ourfinisacked = 1;
1405	}
1406	else
1407	{
1408	#ifndef VBOX_WITH_SLIRP_BSD_SBUF
1409	sbdrop(&so->so_snd, acked);
1410	#else
1411	sbuf_setpos(&so->so_snd, sbuf_len(&so->so_snd) - acked);
1412	#endif
1413	tp->snd_wnd -= acked;
1414	ourfinisacked = 0;
1415	}
1416	/*
1417	* XXX sowwakup is called when data is acked and there's room for
1418	* for more data... it should read() the socket
1419	*/
1420	#if 0
1421	if (so->so_snd.sb_flags & SB_NOTIFY)
1422	sowwakeup(so);
1423	#endif
1424	tp->snd_una = ti->ti_ack;
1425	if (SEQ_LT(tp->snd_nxt, tp->snd_una))
1426	tp->snd_nxt = tp->snd_una;
1427
1428	switch (tp->t_state)
1429	{
1430	/*
1431	* In FIN_WAIT_1 STATE in addition to the processing
1432	* for the ESTABLISHED state if our FIN is now acknowledged
1433	* then enter FIN_WAIT_2.
1434	*/
1435	case TCPS_FIN_WAIT_1:
1436	if (ourfinisacked)
1437	{
1438	/*
1439	* If we can't receive any more
1440	* data, then closing user can proceed.
1441	* Starting the timer is contrary to the
1442	* specification, but if we don't get a FIN
1443	* we'll hang forever.
1444	*/
1445	if (so->so_state & SS_FCANTRCVMORE)
1446	{
1447	soisfdisconnected(so);
1448	tp->t_timer[TCPT_2MSL] = tcp_maxidle;
1449	}
1450	TCP_STATE_SWITCH_TO(tp, TCPS_FIN_WAIT_2);
1451	}
1452	break;
1453
1454	/*
1455	* In CLOSING STATE in addition to the processing for
1456	* the ESTABLISHED state if the ACK acknowledges our FIN
1457	* then enter the TIME-WAIT state, otherwise ignore
1458	* the segment.
1459	*/
1460	case TCPS_CLOSING:
1461	if (ourfinisacked)
1462	{
1463	TCP_STATE_SWITCH_TO(tp, TCPS_TIME_WAIT);
1464	tcp_canceltimers(tp);
1465	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1466	soisfdisconnected(so);
1467	}
1468	break;
1469
1470	/*
1471	* In LAST_ACK, we may still be waiting for data to drain
1472	* and/or to be acked, as well as for the ack of our FIN.
1473	* If our FIN is now acknowledged, delete the TCB,
1474	* enter the closed state and return.
1475	*/
1476	case TCPS_LAST_ACK:
1477	if (ourfinisacked)
1478	{
1479	tp = tcp_close(pData, tp);
1480	LogFlowFunc(("%d -> drop\n", __LINE__));
1481	goto drop;
1482	}
1483	break;
1484
1485	/*
1486	* In TIME_WAIT state the only thing that should arrive
1487	* is a retransmission of the remote FIN. Acknowledge
1488	* it and restart the finack timer.
1489	*/
1490	case TCPS_TIME_WAIT:
1491	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1492	LogFlowFunc(("%d -> dropafterack\n", __LINE__));
1493	goto dropafterack;
1494	}
1495	} /* switch(tp->t_state) */
1496
1497	step6:
1498	LogFlowFunc(("step6:\n"));
1499	/*
1500	* Update window information.
1501	* Don't look at window if no ACK: TAC's send garbage on first SYN.
1502	*/
1503	if ( (tiflags & TH_ACK)
1504	&& ( SEQ_LT(tp->snd_wl1, ti->ti_seq)
1505	\|\| ( tp->snd_wl1 == ti->ti_seq
1506	&& ( SEQ_LT(tp->snd_wl2, ti->ti_ack)
1507	\|\| ( tp->snd_wl2 == ti->ti_ack
1508	&& tiwin > tp->snd_wnd)))))
1509	{
1510	/* keep track of pure window updates */
1511	if ( ti->ti_len == 0
1512	&& tp->snd_wl2 == ti->ti_ack
1513	&& tiwin > tp->snd_wnd)
1514	tcpstat.tcps_rcvwinupd++;
1515	tp->snd_wnd = tiwin;
1516	tp->snd_wl1 = ti->ti_seq;
1517	tp->snd_wl2 = ti->ti_ack;
1518	if (tp->snd_wnd > tp->max_sndwnd)
1519	tp->max_sndwnd = tp->snd_wnd;
1520	needoutput = 1;
1521	}
1522
1523	/*
1524	* Process segments with URG.
1525	*/
1526	if ((tiflags & TH_URG) && ti->ti_urp &&
1527	TCPS_HAVERCVDFIN(tp->t_state) == 0)
1528	{
1529	/* BSD's sbufs are auto extent so we shouldn't worry here */
1530	#ifndef VBOX_WITH_SLIRP_BSD_SBUF
1531	/*
1532	* This is a kludge, but if we receive and accept
1533	* random urgent pointers, we'll crash in
1534	* soreceive. It's hard to imagine someone
1535	* actually wanting to send this much urgent data.
1536	*/
1537	if (ti->ti_urp + so->so_rcv.sb_cc > so->so_rcv.sb_datalen)
1538	{
1539	ti->ti_urp = 0;
1540	tiflags &= ~TH_URG;
1541	LogFlowFunc(("%d -> dodata\n", __LINE__));
1542	goto dodata;
1543	}
1544	#endif
1545	/*
1546	* If this segment advances the known urgent pointer,
1547	* then mark the data stream. This should not happen
1548	* in CLOSE_WAIT, CLOSING, LAST_ACK or TIME_WAIT STATES since
1549	* a FIN has been received from the remote side.
1550	* In these states we ignore the URG.
1551	*
1552	* According to RFC961 (Assigned Protocols),
1553	* the urgent pointer points to the last octet
1554	* of urgent data. We continue, however,
1555	* to consider it to indicate the first octet
1556	* of data past the urgent section as the original
1557	* spec states (in one of two places).
1558	*/
1559	if (SEQ_GT(ti->ti_seq+ti->ti_urp, tp->rcv_up))
1560	{
1561	tp->rcv_up = ti->ti_seq + ti->ti_urp;
1562	so->so_urgc = SBUF_LEN(&so->so_rcv) +
1563	(tp->rcv_up - tp->rcv_nxt); /* -1; */
1564	tp->rcv_up = ti->ti_seq + ti->ti_urp;
1565	}
1566	}
1567	else
1568	/*
1569	* If no out of band data is expected,
1570	* pull receive urgent pointer along
1571	* with the receive window.
1572	*/
1573	if (SEQ_GT(tp->rcv_nxt, tp->rcv_up))
1574	tp->rcv_up = tp->rcv_nxt;
1575	dodata:
1576	LogFlowFunc(("dodata:\n"));
1577
1578	/*
1579	* If this is a small packet, then ACK now - with Nagel
1580	* congestion avoidance sender won't send more until
1581	* he gets an ACK.
1582	*
1583	* See above.
1584	*/
1585	if ( ti->ti_len
1586	&& (unsigned)ti->ti_len <= 5
1587	&& ((struct tcpiphdr_2 *)ti)->first_char == (char)27)
1588	{
1589	tp->t_flags \|= TF_ACKNOW;
1590	}
1591
1592	/*
1593	* Process the segment text, merging it into the TCP sequencing queue,
1594	* and arranging for acknowledgment of receipt if necessary.
1595	* This process logically involves adjusting tp->rcv_wnd as data
1596	* is presented to the user (this happens in tcp_usrreq.c,
1597	* case PRU_RCVD). If a FIN has already been received on this
1598	* connection then we just ignore the text.
1599	*/
1600	if ( (ti->ti_len \|\| (tiflags&TH_FIN))
1601	&& TCPS_HAVERCVDFIN(tp->t_state) == 0)
1602	{
1603	if ( ti->ti_seq == tp->rcv_nxt
1604	&& LIST_EMPTY(&tp->t_segq)
1605	&& tp->t_state == TCPS_ESTABLISHED)
1606	{
1607	DELAY_ACK(tp, ti); /* little bit different from BSD declaration see netinet/tcp_input.c */
1608	tp->rcv_nxt += tlen;
1609	tiflags = ti->ti_t.th_flags & TH_FIN;
1610	tcpstat.tcps_rcvpack++;
1611	tcpstat.tcps_rcvbyte += tlen;
1612	if (so->so_state & SS_FCANTRCVMORE)
1613	m_freem(pData, m);
1614	else
1615	sbappend(pData, so, m);
1616	}
1617	else
1618	{
1619	tiflags = tcp_reass(pData, tp, &ti->ti_t, &tlen, m);
1620	tp->t_flags \|= TF_ACKNOW;
1621	}
1622	/*
1623	* Note the amount of data that peer has sent into
1624	* our window, in order to estimate the sender's
1625	* buffer size.
1626	*/
1627	len = SBUF_SIZE(&so->so_rcv) - (tp->rcv_adv - tp->rcv_nxt);
1628	}
1629	else
1630	{
1631	m_freem(pData, m);
1632	tiflags &= ~TH_FIN;
1633	}
1634
1635	/*
1636	* If FIN is received ACK the FIN and let the user know
1637	* that the connection is closing.
1638	*/
1639	if (tiflags & TH_FIN)
1640	{
1641	if (TCPS_HAVERCVDFIN(tp->t_state) == 0)
1642	{
1643	/*
1644	* If we receive a FIN we can't send more data,
1645	* set it SS_FDRAIN
1646	* Shutdown the socket if there is no rx data in the
1647	* buffer.
1648	* soread() is called on completion of shutdown() and
1649	* will got to TCPS_LAST_ACK, and use tcp_output()
1650	* to send the FIN.
1651	*/
1652	/* sofcantrcvmore(so); */
1653	sofwdrain(so);
1654
1655	tp->t_flags \|= TF_ACKNOW;
1656	tp->rcv_nxt++;
1657	}
1658	switch (tp->t_state)
1659	{
1660	/*
1661	* In SYN_RECEIVED and ESTABLISHED STATES
1662	* enter the CLOSE_WAIT state.
1663	*/
1664	case TCPS_SYN_RECEIVED:
1665	case TCPS_ESTABLISHED:
1666	TCP_STATE_SWITCH_TO(tp, TCPS_CLOSE_WAIT);
1667	break;
1668
1669	/*
1670	* If still in FIN_WAIT_1 STATE FIN has not been acked so
1671	* enter the CLOSING state.
1672	*/
1673	case TCPS_FIN_WAIT_1:
1674	TCP_STATE_SWITCH_TO(tp, TCPS_CLOSING);
1675	break;
1676
1677	/*
1678	* In FIN_WAIT_2 state enter the TIME_WAIT state,
1679	* starting the time-wait timer, turning off the other
1680	* standard timers.
1681	*/
1682	case TCPS_FIN_WAIT_2:
1683	TCP_STATE_SWITCH_TO(tp, TCPS_TIME_WAIT);
1684	tcp_canceltimers(tp);
1685	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1686	soisfdisconnected(so);
1687	break;
1688
1689	/*
1690	* In TIME_WAIT state restart the 2 MSL time_wait timer.
1691	*/
1692	case TCPS_TIME_WAIT:
1693	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1694	break;
1695	}
1696	}
1697
1698	/*
1699	* Return any desired output.
1700	*/
1701	if (needoutput \|\| (tp->t_flags & TF_ACKNOW))
1702	tcp_output(pData, tp);
1703
1704	SOCKET_UNLOCK(so);
1705	STAM_PROFILE_STOP(&pData->StatTCP_input, counter_input);
1706	LogFlowFuncLeave();
1707	return;
1708
1709	dropafterack:
1710	LogFlowFunc(("dropafterack:\n"));
1711	/*
1712	* Generate an ACK dropping incoming segment if it occupies
1713	* sequence space, where the ACK reflects our state.
1714	*/
1715	if (tiflags & TH_RST)
1716	{
1717	LogFlowFunc(("%d -> drop\n", __LINE__));
1718	goto drop;
1719	}
1720	m_freem(pData, m);
1721	tp->t_flags \|= TF_ACKNOW;
1722	(void) tcp_output(pData, tp);
1723	SOCKET_UNLOCK(so);
1724	STAM_PROFILE_STOP(&pData->StatTCP_input, counter_input);
1725	LogFlowFuncLeave();
1726	return;
1727
1728	dropwithreset:
1729	LogFlowFunc(("dropwithreset:\n"));
1730	/* reuses m if m!=NULL, m_free() unnecessary */
1731	if (tiflags & TH_ACK)
1732	tcp_respond(pData, tp, ti, m, (tcp_seq)0, ti->ti_ack, TH_RST);
1733	else
1734	{
1735	if (tiflags & TH_SYN)
1736	ti->ti_len++;
1737	tcp_respond(pData, tp, ti, m, ti->ti_seq+ti->ti_len, (tcp_seq)0,
1738	TH_RST\|TH_ACK);
1739	}
1740
1741	if (so != &tcb)
1742	SOCKET_UNLOCK(so);
1743	STAM_PROFILE_STOP(&pData->StatTCP_input, counter_input);
1744	LogFlowFuncLeave();
1745	return;
1746
1747	drop:
1748	LogFlowFunc(("drop:\n"));
1749	/*
1750	* Drop space held by incoming segment and return.
1751	*/
1752	m_freem(pData, m);
1753
1754	#ifdef VBOX_WITH_SLIRP_MT
1755	if (RTCritSectIsOwned(&so->so_mutex))
1756	{
1757	SOCKET_UNLOCK(so);
1758	}
1759	#endif
1760
1761	STAM_PROFILE_STOP(&pData->StatTCP_input, counter_input);
1762	LogFlowFuncLeave();
1763	return;
1764	}
1765
1766
1767	void
1768	tcp_fconnect_failed(PNATState pData, struct socket *so, int sockerr)
1769	{
1770	struct tcpcb *tp;
1771	int code;
1772
1773	Log2(("NAT: connect error %d %R[natsock]\n", sockerr, so));
1774
1775	Assert(so->so_state & SS_ISFCONNECTING);
1776	so->so_state = SS_NOFDREF;
1777
1778	if (sockerr == ECONNREFUSED \|\| sockerr == ECONNRESET)
1779	{
1780	/* hand off to tcp_input():cont_conn to send RST */
1781	TCP_INPUT(pData, NULL, 0, so);
1782	return;
1783	}
1784
1785	tp = sototcpcb(so);
1786	if (RT_UNLIKELY(tp == NULL)) /* should never happen */
1787	{
1788	LogRel(("NAT: tp == NULL %R[natsock]\n", so));
1789	sofree(pData, so);
1790	return;
1791	}
1792
1793	if (sockerr == ENETUNREACH \|\| sockerr == ENETDOWN)
1794	code = ICMP_UNREACH_NET;
1795	else if (sockerr == EHOSTUNREACH \|\| sockerr == EHOSTDOWN)
1796	code = ICMP_UNREACH_HOST;
1797	else
1798	code = -1;
1799
1800	if (code >= 0)
1801	{
1802	struct ip *oip;
1803	size_t ohdrlen;
1804	struct mbuf *m;
1805
1806	if (RT_UNLIKELY(so->so_ohdr == NULL))
1807	goto out;
1808
1809	oip = (struct ip *)so->so_ohdr;
1810	ohdrlen = oip->ip_hl * 4 + 8;
1811
1812	m = m_gethdr(pData, M_NOWAIT, MT_HEADER);
1813	if (RT_UNLIKELY(m == NULL))
1814	goto out;
1815
1816	m_copyback(pData, m, 0, ohdrlen, (caddr_t)so->so_ohdr);
1817	m->m_pkthdr.header = mtod(m, void *);
1818
1819	icmp_error(pData, m, ICMP_UNREACH, code, 0, NULL);
1820	}
1821
1822	out:
1823	tcp_close(pData, tp);
1824	}
1825
1826
1827	void
1828	tcp_dooptions(PNATState pData, struct tcpcb tp, u_char cp, int cnt, struct tcpiphdr *ti)
1829	{
1830	u_int16_t mss;
1831	int opt, optlen;
1832
1833	LogFlowFunc(("tcp_dooptions: tp = %R[tcpcb793], cnt=%i\n", tp, cnt));
1834
1835	for (; cnt > 0; cnt -= optlen, cp += optlen)
1836	{
1837	opt = cp[0];
1838	if (opt == TCPOPT_EOL)
1839	break;
1840	if (opt == TCPOPT_NOP)
1841	optlen = 1;
1842	else
1843	{
1844	optlen = cp[1];
1845	if (optlen <= 0)
1846	break;
1847	}
1848	switch (opt)
1849	{
1850	default:
1851	continue;
1852
1853	case TCPOPT_MAXSEG:
1854	if (optlen != TCPOLEN_MAXSEG)
1855	continue;
1856	if (!(ti->ti_flags & TH_SYN))
1857	continue;
1858	memcpy((char ) &mss, (char ) cp + 2, sizeof(mss));
1859	NTOHS(mss);
1860	(void) tcp_mss(pData, tp, mss); /* sets t_maxseg */
1861	break;
1862
1863	#if 0
1864	case TCPOPT_WINDOW:
1865	if (optlen != TCPOLEN_WINDOW)
1866	continue;
1867	if (!(ti->ti_flags & TH_SYN))
1868	continue;
1869	tp->t_flags \|= TF_RCVD_SCALE;
1870	tp->requested_s_scale = min(cp[2], TCP_MAX_WINSHIFT);
1871	break;
1872
1873	case TCPOPT_TIMESTAMP:
1874	if (optlen != TCPOLEN_TIMESTAMP)
1875	continue;
1876	*ts_present = 1;
1877	memcpy((char ) ts_val, (char )cp + 2, sizeof(*ts_val));
1878	NTOHL(*ts_val);
1879	memcpy((char ) ts_ecr, (char )cp + 6, sizeof(*ts_ecr));
1880	NTOHL(*ts_ecr);
1881
1882	/*
1883	* A timestamp received in a SYN makes
1884	* it ok to send timestamp requests and replies.
1885	*/
1886	if (ti->ti_flags & TH_SYN)
1887	{
1888	tp->t_flags \|= TF_RCVD_TSTMP;
1889	tp->ts_recent = *ts_val;
1890	tp->ts_recent_age = tcp_now;
1891	}
1892	break;
1893	#endif
1894	}
1895	}
1896	}
1897
1898
1899	/*
1900	* Pull out of band byte out of a segment so
1901	* it doesn't appear in the user's data queue.
1902	* It is still reflected in the segment length for
1903	* sequencing purposes.
1904	*/
1905
1906	#if 0
1907	void
1908	tcp_pulloutofband(struct socket so, struct tcpiphdr ti, struct mbuf *m)
1909	{
1910	int cnt = ti->ti_urp - 1;
1911
1912	while (cnt >= 0)
1913	{
1914	if (m->m_len > cnt)
1915	{
1916	char *cp = mtod(m, caddr_t) + cnt;
1917	struct tcpcb *tp = sototcpcb(so);
1918
1919	tp->t_iobc = *cp;
1920	tp->t_oobflags \|= TCPOOB_HAVEDATA;
1921	memcpy(sp, cp+1, (unsigned)(m->m_len - cnt - 1));
1922	m->m_len--;
1923	return;
1924	}
1925	cnt -= m->m_len;
1926	m = m->m_next; /* XXX WRONG! Fix it! */
1927	if (m == 0)
1928	break;
1929	}
1930	panic("tcp_pulloutofband");
1931	}
1932	#endif
1933
1934	/*
1935	* Collect new round-trip time estimate
1936	* and update averages and current timeout.
1937	*/
1938
1939	void
1940	tcp_xmit_timer(PNATState pData, register struct tcpcb *tp, int rtt)
1941	{
1942	register short delta;
1943
1944	LogFlowFunc(("ENTER: tcp_xmit_timer: tp = %R[tcpcb793] rtt = %d\n", tp, rtt));
1945
1946	tcpstat.tcps_rttupdated++;
1947	if (tp->t_srtt != 0)
1948	{
1949	/*
1950	* srtt is stored as fixed point with 3 bits after the
1951	* binary point (i.e., scaled by 8). The following magic
1952	* is equivalent to the smoothing algorithm in rfc793 with
1953	* an alpha of .875 (srtt = rtt/8 + srtt*7/8 in fixed
1954	* point). Adjust rtt to origin 0.
1955	*/
1956	delta = rtt - 1 - (tp->t_srtt >> TCP_RTT_SHIFT);
1957	if ((tp->t_srtt += delta) <= 0)
1958	tp->t_srtt = 1;
1959	/*
1960	* We accumulate a smoothed rtt variance (actually, a
1961	* smoothed mean difference), then set the retransmit
1962	* timer to smoothed rtt + 4 times the smoothed variance.
1963	* rttvar is stored as fixed point with 2 bits after the
1964	* binary point (scaled by 4). The following is
1965	* equivalent to rfc793 smoothing with an alpha of .75
1966	* (rttvar = rttvar*3/4 + \|delta\| / 4). This replaces
1967	* rfc793's wired-in beta.
1968	*/
1969	if (delta < 0)
1970	delta = -delta;
1971	delta -= (tp->t_rttvar >> TCP_RTTVAR_SHIFT);
1972	if ((tp->t_rttvar += delta) <= 0)
1973	tp->t_rttvar = 1;
1974	}
1975	else
1976	{
1977	/*
1978	* No rtt measurement yet - use the unsmoothed rtt.
1979	* Set the variance to half the rtt (so our first
1980	* retransmit happens at 3*rtt).
1981	*/
1982	tp->t_srtt = rtt << TCP_RTT_SHIFT;
1983	tp->t_rttvar = rtt << (TCP_RTTVAR_SHIFT - 1);
1984	}
1985	tp->t_rtt = 0;
1986	tp->t_rxtshift = 0;
1987
1988	/*
1989	* the retransmit should happen at rtt + 4 * rttvar.
1990	* Because of the way we do the smoothing, srtt and rttvar
1991	* will each average +1/2 tick of bias. When we compute
1992	* the retransmit timer, we want 1/2 tick of rounding and
1993	* 1 extra tick because of +-1/2 tick uncertainty in the
1994	* firing of the timer. The bias will give us exactly the
1995	* 1.5 tick we need. But, because the bias is
1996	* statistical, we have to test that we don't drop below
1997	* the minimum feasible timer (which is 2 ticks).
1998	*/
1999	TCPT_RANGESET(tp->t_rxtcur, TCP_REXMTVAL(tp),
2000	(short)tp->t_rttmin, TCPTV_REXMTMAX); /* XXX */
2001
2002	/*
2003	* We received an ack for a packet that wasn't retransmitted;
2004	* it is probably safe to discard any error indications we've
2005	* received recently. This isn't quite right, but close enough
2006	* for now (a route might have failed after we sent a segment,
2007	* and the return path might not be symmetrical).
2008	*/
2009	tp->t_softerror = 0;
2010	}
2011
2012	/*
2013	* Determine a reasonable value for maxseg size.
2014	* If the route is known, check route for mtu.
2015	* If none, use an mss that can be handled on the outgoing
2016	* interface without forcing IP to fragment; if bigger than
2017	* an mbuf cluster (MCLBYTES), round down to nearest multiple of MCLBYTES
2018	* to utilize large mbufs. If no route is found, route has no mtu,
2019	* or the destination isn't local, use a default, hopefully conservative
2020	* size (usually 512 or the default IP max size, but no more than the mtu
2021	* of the interface), as we can't discover anything about intervening
2022	* gateways or networks. We also initialize the congestion/slow start
2023	* window to be a single segment if the destination isn't local.
2024	* While looking at the routing entry, we also initialize other path-dependent
2025	* parameters from pre-set or cached values in the routing entry.
2026	*/
2027
2028	int
2029	tcp_mss(PNATState pData, register struct tcpcb *tp, u_int offer)
2030	{
2031	struct socket *so = tp->t_socket;
2032	int mss;
2033
2034	LogFlowFunc(("ENTER: tcp_mss: tp = %R[tcpcb793], offer = %d\n", tp, offer));
2035
2036	mss = min(if_mtu, if_mru) - sizeof(struct tcpiphdr);
2037	if (offer)
2038	mss = min(mss, offer);
2039	mss = max(mss, 32);
2040	if (mss < tp->t_maxseg \|\| offer != 0)
2041	tp->t_maxseg = mss;
2042
2043	tp->snd_cwnd = mss;
2044
2045	#ifndef VBOX_WITH_SLIRP_BSD_SBUF
2046	sbreserve(pData, &so->so_snd, tcp_sndspace+((tcp_sndspace%mss)?(mss-(tcp_sndspace%mss)):0));
2047	sbreserve(pData, &so->so_rcv, tcp_rcvspace+((tcp_rcvspace%mss)?(mss-(tcp_rcvspace%mss)):0));
2048	#else
2049	sbuf_new(&so->so_snd, NULL, tcp_sndspace+((tcp_sndspace%mss)?(mss-(tcp_sndspace%mss)):0), SBUF_AUTOEXTEND);
2050	sbuf_new(&so->so_rcv, NULL, tcp_rcvspace+((tcp_rcvspace%mss)?(mss-(tcp_rcvspace%mss)):0), SBUF_AUTOEXTEND);
2051	#endif
2052
2053	Log2((" returning mss = %d\n", mss));
2054
2055	return mss;
2056	}

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

source: vbox/trunk/src/VBox/Devices/Network/slirp/tcp_input.c@ 53814

Download in other formats: