tcp_input.c@ 93115

Last change on this file since 93115 was 93115, checked in by vboxsync, 3 years ago
scm --update-copyright-year
Property svn:eol-style set to `native` Property svn:keywords set to `Author Date Id Revision`
File size: 67.1 KB

Line
1	/* $Id: tcp_input.c 93115 2022-01-01 11:31:46Z vboxsync $ */
2	/** @file
3	* NAT - TCP input.
4	*/
5
6	/*
7	* Copyright (C) 2006-2022 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, off;
297	int tlen = 0; /* Shut up MSC (didn't check whether MSC was right). */
298	register struct tcpcb *tp = 0;
299	register int tiflags;
300	struct socket *so = 0;
301	int todrop, acked, ourfinisacked, needoutput = 0;
302	/* int dropsocket = 0; */
303	int iss = 0;
304	u_long tiwin;
305	/* int ts_present = 0; */
306	unsigned ohdrlen;
307	uint8_t ohdr[60 + 8]; /* max IP header plus 8 bytes of payload for icmp */
308
309	STAM_PROFILE_START(&pData->StatTCP_input, counter_input);
310
311	LogFlow(("tcp_input: m = %p, iphlen = %2d, inso = %R[natsock]\n", 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
327	/* Re-set a few variables */
328	tp = sototcpcb(so);
329
330	m = so->so_m;
331	optp = so->so_optp; /* points into m if set */
332	optlen = so->so_optlen;
333	so->so_m = NULL;
334	so->so_optp = 0;
335	so->so_optlen = 0;
336
337	if (RT_LIKELY(so->so_ohdr != NULL))
338	{
339	RTMemFree(so->so_ohdr);
340	so->so_ohdr = NULL;
341	}
342
343	ti = so->so_ti;
344
345	/** @todo (vvl) clarify why it might happens */
346	if (ti == NULL)
347	{
348	LogRel(("NAT: ti is null. can't do any reseting connection actions\n"));
349	/* mbuf should be cleared in sofree called from tcp_close */
350	tcp_close(pData, tp);
351	STAM_PROFILE_STOP(&pData->StatTCP_input, counter_input);
352	LogFlowFuncLeave();
353	return;
354	}
355
356	tiwin = ti->ti_win;
357	tiflags = ti->ti_flags;
358
359	LogFlowFunc(("%d -> cont_conn\n", __LINE__));
360	goto cont_conn;
361	}
362
363	tcpstat.tcps_rcvtotal++;
364
365	ip = mtod(m, struct ip *);
366
367	/* ip_input() subtracts iphlen from ip::ip_len */
368	AssertStmt(ip->ip_len + iphlen == (ssize_t)m_length(m, NULL), goto drop);
369	if (RT_UNLIKELY(ip->ip_len < sizeof(struct tcphdr)))
370	{
371	/* tcps_rcvshort++; */
372	goto drop;
373	}
374
375	/*
376	* Save a copy of the IP header in case we want to restore it for
377	* sending an ICMP error message in response.
378	*
379	* XXX: This function should really be fixed to not strip IP
380	* options, to not overwrite IP header and to use "tlen" local
381	* variable (instead of ti->ti_len), then "m" could be passed to
382	* icmp_error() directly.
383	*/
384	ohdrlen = iphlen + 8;
385	m_copydata(m, 0, ohdrlen, (caddr_t)ohdr);
386	save_ip = (struct ip *)ohdr;
387	save_ip->ip_len += iphlen; /* undo change by ip_input() */
388
389
390	/*
391	* Get IP and TCP header together in first mbuf.
392	* Note: IP leaves IP header in first mbuf.
393	*/
394	ti = mtod(m, struct tcpiphdr *);
395	if (iphlen > sizeof(struct ip))
396	{
397	ip_stripoptions(m, (struct mbuf *)0);
398	iphlen = sizeof(struct ip);
399	}
400
401	/*
402	* Checksum extended TCP header and data.
403	*/
404	tlen = ((struct ip *)ti)->ip_len;
405	memset(ti->ti_x1, 0, 9);
406	ti->ti_len = RT_H2N_U16((u_int16_t)tlen);
407	len = sizeof(struct ip) + tlen;
408	/* keep checksum for ICMP reply
409	* ti->ti_sum = cksum(m, len);
410	* if (ti->ti_sum) { */
411	if (cksum(m, len))
412	{
413	tcpstat.tcps_rcvbadsum++;
414	LogFlowFunc(("%d -> drop\n", __LINE__));
415	goto drop;
416	}
417
418	/*
419	* Check that TCP offset makes sense,
420	* pull out TCP options and adjust length. XXX
421	*/
422	off = ti->ti_off << 2;
423	if ( off < sizeof (struct tcphdr)
424	\|\| off > tlen)
425	{
426	tcpstat.tcps_rcvbadoff++;
427	LogFlowFunc(("%d -> drop\n", __LINE__));
428	goto drop;
429	}
430	tlen -= off;
431	ti->ti_len = tlen;
432	if (off > sizeof (struct tcphdr))
433	{
434	optlen = off - sizeof (struct tcphdr);
435	optp = mtod(m, caddr_t) + sizeof (struct tcpiphdr);
436
437	/*
438	* Do quick retrieval of timestamp options ("options
439	* prediction?"). If timestamp is the only option and it's
440	* formatted as recommended in RFC 1323 appendix A, we
441	* quickly get the values now and not bother calling
442	* tcp_dooptions(), etc.
443	*/
444	#if 0
445	if (( optlen == TCPOLEN_TSTAMP_APPA
446	\|\| ( optlen > TCPOLEN_TSTAMP_APPA
447	&& optp[TCPOLEN_TSTAMP_APPA] == TCPOPT_EOL)) &&
448	(u_int32_t )optp == RT_H2N_U32_C(TCPOPT_TSTAMP_HDR) &&
449	(ti->ti_flags & TH_SYN) == 0)
450	{
451	ts_present = 1;
452	ts_val = RT_N2H_U32((u_int32_t )(optp + 4));
453	ts_ecr = RT_N2H_U32((u_int32_t )(optp + 8));
454	optp = NULL; / * we have parsed the options * /
455	}
456	#endif
457	}
458	tiflags = ti->ti_flags;
459
460	/*
461	* Convert TCP protocol specific fields to host format.
462	*/
463	NTOHL(ti->ti_seq);
464	NTOHL(ti->ti_ack);
465	NTOHS(ti->ti_win);
466	NTOHS(ti->ti_urp);
467
468	/*
469	* Drop TCP, IP headers and TCP options.
470	*/
471	m->m_data += sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
472	m->m_len -= sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
473
474	/*
475	* Locate pcb for segment.
476	*/
477	findso:
478	LogFlowFunc(("(enter) findso: %R[natsock]\n", so));
479	if (so != NULL && so != &tcb)
480	SOCKET_UNLOCK(so);
481	QSOCKET_LOCK(tcb);
482	so = tcp_last_so;
483	if ( so->so_fport != ti->ti_dport
484	\|\| so->so_lport != ti->ti_sport
485	\|\| so->so_laddr.s_addr != ti->ti_src.s_addr
486	\|\| so->so_faddr.s_addr != ti->ti_dst.s_addr)
487	{
488	QSOCKET_UNLOCK(tcb);
489	/** @todo fix SOLOOKUP macrodefinition to be usable here */
490	so = solookup(&tcb, ti->ti_src, ti->ti_sport,
491	ti->ti_dst, ti->ti_dport);
492	if (so)
493	{
494	tcp_last_so = so;
495	}
496	++tcpstat.tcps_socachemiss;
497	}
498	else
499	{
500	SOCKET_LOCK(so);
501	QSOCKET_UNLOCK(tcb);
502	}
503	LogFlowFunc(("(leave) findso: %R[natsock]\n", so));
504
505	/*
506	* Check whether the packet is targeting CTL_ALIAS and drop it if the connection wasn't
507	* initiated by localhost (so == NULL), see @bugref{9896}.
508	*/
509	if ( (RT_N2H_U32(ti->ti_dst.s_addr) & ~pData->netmask) == CTL_ALIAS
510	&& !pData->fLocalhostReachable
511	&& !so)
512	{
513	LogFlowFunc(("Packet for CTL_ALIAS and fLocalhostReachable=false so=NULL -> drop\n"));
514	goto drop;
515	}
516
517	/*
518	* If the state is CLOSED (i.e., TCB does not exist) then
519	* all data in the incoming segment is discarded.
520	* If the TCB exists but is in CLOSED state, it is embryonic,
521	* but should either do a listen or a connect soon.
522	*
523	* state == CLOSED means we've done socreate() but haven't
524	* attached it to a protocol yet...
525	*
526	* XXX If a TCB does not exist, and the TH_SYN flag is
527	* the only flag set, then create a session, mark it
528	* as if it was LISTENING, and continue...
529	*/
530	if (so == 0)
531	{
532	if ((tiflags & (TH_SYN\|TH_FIN\|TH_RST\|TH_URG\|TH_ACK)) != TH_SYN)
533	{
534	LogFlowFunc(("%d -> dropwithreset\n", __LINE__));
535	goto dropwithreset;
536	}
537
538	if ((so = socreate()) == NULL)
539	{
540	LogFlowFunc(("%d -> dropwithreset\n", __LINE__));
541	goto dropwithreset;
542	}
543	if (tcp_attach(pData, so) < 0)
544	{
545	RTMemFree(so); /* Not sofree (if it failed, it's not insqued) */
546	LogFlowFunc(("%d -> dropwithreset\n", __LINE__));
547	goto dropwithreset;
548	}
549	SOCKET_LOCK(so);
550	sbreserve(pData, &so->so_snd, tcp_sndspace);
551	sbreserve(pData, &so->so_rcv, tcp_rcvspace);
552
553	/* tcp_last_so = so; / / XXX ? */
554	/* tp = sototcpcb(so); */
555
556	so->so_laddr = ti->ti_src;
557	so->so_lport = ti->ti_sport;
558	so->so_faddr = ti->ti_dst;
559	so->so_fport = ti->ti_dport;
560
561	so->so_iptos = ((struct ip *)ti)->ip_tos;
562
563	tp = sototcpcb(so);
564	TCP_STATE_SWITCH_TO(tp, TCPS_LISTEN);
565	}
566
567	/*
568	* If this is a still-connecting socket, this probably
569	* a retransmit of the SYN. Whether it's a retransmit SYN
570	* or something else, we nuke it.
571	*/
572	if (so->so_state & SS_ISFCONNECTING)
573	{
574	LogFlowFunc(("%d -> drop\n", __LINE__));
575	goto drop;
576	}
577
578	tp = sototcpcb(so);
579
580	/* XXX Should never fail */
581	if (tp == 0)
582	{
583	LogFlowFunc(("%d -> dropwithreset\n", __LINE__));
584	goto dropwithreset;
585	}
586	if (tp->t_state == TCPS_CLOSED)
587	{
588	LogFlowFunc(("%d -> drop\n", __LINE__));
589	goto drop;
590	}
591
592	/* Unscale the window into a 32-bit value. */
593	/* if ((tiflags & TH_SYN) == 0)
594	* tiwin = ti->ti_win << tp->snd_scale;
595	* else
596	*/
597	tiwin = ti->ti_win;
598
599	/*
600	* Segment received on connection.
601	* Reset idle time and keep-alive timer.
602	*/
603	tp->t_idle = 0;
604	if (so_options)
605	tp->t_timer[TCPT_KEEP] = tcp_keepintvl;
606	else
607	tp->t_timer[TCPT_KEEP] = tcp_keepidle;
608
609	/*
610	* Process options if not in LISTEN state,
611	* else do it below (after getting remote address).
612	*/
613	if (optp && tp->t_state != TCPS_LISTEN)
614	tcp_dooptions(pData, tp, (u_char *)optp, optlen, ti);
615	/* , */
616	/* &ts_present, &ts_val, &ts_ecr); */
617
618	/*
619	* Header prediction: check for the two common cases
620	* of a uni-directional data xfer. If the packet has
621	* no control flags, is in-sequence, the window didn't
622	* change and we're not retransmitting, it's a
623	* candidate. If the length is zero and the ack moved
624	* forward, we're the sender side of the xfer. Just
625	* free the data acked & wake any higher level process
626	* that was blocked waiting for space. If the length
627	* is non-zero and the ack didn't move, we're the
628	* receiver side. If we're getting packets in-order
629	* (the reassembly queue is empty), add the data to
630	* the socket buffer and note that we need a delayed ack.
631	*
632	* XXX Some of these tests are not needed
633	* eg: the tiwin == tp->snd_wnd prevents many more
634	* predictions.. with no real advantage..
635	*/
636	if ( tp->t_state == TCPS_ESTABLISHED
637	&& (tiflags & (TH_SYN\|TH_FIN\|TH_RST\|TH_URG\|TH_ACK)) == TH_ACK
638	/* && (!ts_present \|\| TSTMP_GEQ(ts_val, tp->ts_recent)) */
639	&& ti->ti_seq == tp->rcv_nxt
640	&& tiwin && tiwin == tp->snd_wnd
641	&& tp->snd_nxt == tp->snd_max)
642	{
643	/*
644	* If last ACK falls within this segment's sequence numbers,
645	* record the timestamp.
646	*/
647	#if 0
648	if (ts_present && SEQ_LEQ(ti->ti_seq, tp->last_ack_sent) &&
649	SEQ_LT(tp->last_ack_sent, ti->ti_seq + ti->ti_len))
650	{
651	tp->ts_recent_age = tcp_now;
652	tp->ts_recent = ts_val;
653	}
654	#endif
655
656	if (ti->ti_len == 0)
657	{
658	if ( SEQ_GT(ti->ti_ack, tp->snd_una)
659	&& SEQ_LEQ(ti->ti_ack, tp->snd_max)
660	&& tp->snd_cwnd >= tp->snd_wnd)
661	{
662	/*
663	* this is a pure ack for outstanding data.
664	*/
665	++tcpstat.tcps_predack;
666	#if 0
667	if (ts_present)
668	tcp_xmit_timer(tp, tcp_now-ts_ecr+1);
669	else
670	#endif
671	if ( tp->t_rtt
672	&& SEQ_GT(ti->ti_ack, tp->t_rtseq))
673	tcp_xmit_timer(pData, tp, tp->t_rtt);
674	acked = ti->ti_ack - tp->snd_una;
675	tcpstat.tcps_rcvackpack++;
676	tcpstat.tcps_rcvackbyte += acked;
677	sbdrop(&so->so_snd, acked);
678	tp->snd_una = ti->ti_ack;
679	m_freem(pData, m);
680
681	/*
682	* If all outstanding data are acked, stop
683	* retransmit timer, otherwise restart timer
684	* using current (possibly backed-off) value.
685	* If process is waiting for space,
686	* wakeup/selwakeup/signal. If data
687	* are ready to send, let tcp_output
688	* decide between more output or persist.
689	*/
690	if (tp->snd_una == tp->snd_max)
691	tp->t_timer[TCPT_REXMT] = 0;
692	else if (tp->t_timer[TCPT_PERSIST] == 0)
693	tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
694
695	/*
696	* There's room in so_snd, sowwakup will read()
697	* from the socket if we can
698	*/
699	#if 0
700	if (so->so_snd.sb_flags & SB_NOTIFY)
701	sowwakeup(so);
702	#endif
703	/*
704	* This is called because sowwakeup might have
705	* put data into so_snd. Since we don't so sowwakeup,
706	* we don't need this.. XXX???
707	*/
708	if (SBUF_LEN(&so->so_snd))
709	(void) tcp_output(pData, tp);
710
711	SOCKET_UNLOCK(so);
712	STAM_PROFILE_STOP(&pData->StatTCP_input, counter_input);
713	return;
714	}
715	}
716	else if ( ti->ti_ack == tp->snd_una
717	&& LIST_EMPTY(&tp->t_segq)
718	&& ti->ti_len <= sbspace(&so->so_rcv))
719	{
720	/*
721	* this is a pure, in-sequence data packet
722	* with nothing on the reassembly queue and
723	* we have enough buffer space to take it.
724	*/
725	++tcpstat.tcps_preddat;
726	tp->rcv_nxt += ti->ti_len;
727	tcpstat.tcps_rcvpack++;
728	tcpstat.tcps_rcvbyte += ti->ti_len;
729	/*
730	* Add data to socket buffer.
731	*/
732	sbappend(pData, so, m);
733
734	/*
735	* XXX This is called when data arrives. Later, check
736	* if we can actually write() to the socket
737	* XXX Need to check? It's be NON_BLOCKING
738	*/
739	/* sorwakeup(so); */
740
741	/*
742	* If this is a short packet, then ACK now - with Nagle
743	* congestion avoidance sender won't send more until
744	* he gets an ACK.
745	*
746	* It is better to not delay acks at all to maximize
747	* TCP throughput. See RFC 2581.
748	*/
749	tp->t_flags \|= TF_ACKNOW;
750	tcp_output(pData, tp);
751	SOCKET_UNLOCK(so);
752	STAM_PROFILE_STOP(&pData->StatTCP_input, counter_input);
753	return;
754	}
755	} /* header prediction */
756	/*
757	* Calculate amount of space in receive window,
758	* and then do TCP input processing.
759	* Receive window is amount of space in rcv queue,
760	* but not less than advertised window.
761	*/
762	{
763	int win;
764	win = sbspace(&so->so_rcv);
765	if (win < 0)
766	win = 0;
767	tp->rcv_wnd = max(win, (int)(tp->rcv_adv - tp->rcv_nxt));
768	}
769
770	switch (tp->t_state)
771	{
772	/*
773	* If the state is LISTEN then ignore segment if it contains an RST.
774	* If the segment contains an ACK then it is bad and send a RST.
775	* If it does not contain a SYN then it is not interesting; drop it.
776	* Don't bother responding if the destination was a broadcast.
777	* Otherwise initialize tp->rcv_nxt, and tp->irs, select an initial
778	* tp->iss, and send a segment:
779	* <SEQ=ISS><ACK=RCV_NXT><CTL=SYN,ACK>
780	* Also initialize tp->snd_nxt to tp->iss+1 and tp->snd_una to tp->iss.
781	* Fill in remote peer address fields if not previously specified.
782	* Enter SYN_RECEIVED state, and process any other fields of this
783	* segment in this state.
784	*/
785	case TCPS_LISTEN:
786	{
787	if (tiflags & TH_RST)
788	{
789	LogFlowFunc(("%d -> drop\n", __LINE__));
790	goto drop;
791	}
792	if (tiflags & TH_ACK)
793	{
794	LogFlowFunc(("%d -> dropwithreset\n", __LINE__));
795	goto dropwithreset;
796	}
797	if ((tiflags & TH_SYN) == 0)
798	{
799	LogFlowFunc(("%d -> drop\n", __LINE__));
800	goto drop;
801	}
802
803	/*
804	* This has way too many gotos...
805	* But a bit of spaghetti code never hurt anybody :)
806	*/
807	if ( (tcp_fconnect(pData, so) == -1)
808	&& errno != EINPROGRESS
809	&& errno != EWOULDBLOCK)
810	{
811	u_char code = ICMP_UNREACH_NET;
812	Log2((" tcp fconnect errno = %d (%s)\n", errno, strerror(errno)));
813	if (errno == ECONNREFUSED)
814	{
815	/* ACK the SYN, send RST to refuse the connection */
816	tcp_respond(pData, tp, ti, m, ti->ti_seq+1, (tcp_seq)0,
817	TH_RST\|TH_ACK);
818	}
819	else
820	{
821	if (errno == EHOSTUNREACH)
822	code = ICMP_UNREACH_HOST;
823	HTONL(ti->ti_seq); /* restore tcp header */
824	HTONL(ti->ti_ack);
825	HTONS(ti->ti_win);
826	HTONS(ti->ti_urp);
827	m->m_data -= sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
828	m->m_len += sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
829	ip = save_ip;
830	icmp_error(pData, m, ICMP_UNREACH, code, 0, strerror(errno));
831	tp->t_socket->so_m = NULL;
832	}
833	tp = tcp_close(pData, tp);
834	}
835	else
836	{
837	/*
838	* Haven't connected yet, save the current mbuf
839	* and ti, and return
840	* XXX Some OS's don't tell us whether the connect()
841	* succeeded or not. So we must time it out.
842	*/
843	so->so_m = m;
844	so->so_ti = ti;
845	so->so_ohdr = RTMemDup(ohdr, ohdrlen);
846	so->so_optp = optp;
847	so->so_optlen = optlen;
848	tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT;
849	TCP_STATE_SWITCH_TO(tp, TCPS_SYN_RECEIVED);
850	}
851	SOCKET_UNLOCK(so);
852	STAM_PROFILE_STOP(&pData->StatTCP_input, counter_input);
853	LogFlowFuncLeave();
854	return;
855
856	cont_conn:
857	/* m==NULL
858	* Check if the connect succeeded
859	*/
860	LogFlowFunc(("cont_conn:\n"));
861	if (so->so_state & SS_NOFDREF)
862	{
863	tp = tcp_close(pData, tp);
864	LogFlowFunc(("%d -> dropwithreset\n", __LINE__));
865	goto dropwithreset;
866	}
867
868	tcp_template(tp);
869
870	if (optp)
871	tcp_dooptions(pData, tp, (u_char *)optp, optlen, ti);
872
873	if (iss)
874	tp->iss = iss;
875	else
876	tp->iss = tcp_iss;
877	tcp_iss += TCP_ISSINCR/2;
878	tp->irs = ti->ti_seq;
879	tcp_sendseqinit(tp);
880	tcp_rcvseqinit(tp);
881	tp->t_flags \|= TF_ACKNOW;
882	TCP_STATE_SWITCH_TO(tp, TCPS_SYN_RECEIVED);
883	tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT;
884	tcpstat.tcps_accepts++;
885	LogFlowFunc(("%d -> trimthenstep6\n", __LINE__));
886	goto trimthenstep6;
887	} /* case TCPS_LISTEN */
888
889	/*
890	* If the state is SYN_SENT:
891	* if seg contains an ACK, but not for our SYN, drop the input.
892	* if seg contains a RST, then drop the connection.
893	* if seg does not contain SYN, then drop it.
894	* Otherwise this is an acceptable SYN segment
895	* initialize tp->rcv_nxt and tp->irs
896	* if seg contains ack then advance tp->snd_una
897	* if SYN has been acked change to ESTABLISHED else SYN_RCVD state
898	* arrange for segment to be acked (eventually)
899	* continue processing rest of data/controls, beginning with URG
900	*/
901	case TCPS_SYN_SENT:
902	if ( (tiflags & TH_ACK)
903	&& ( SEQ_LEQ(ti->ti_ack, tp->iss)
904	\|\| SEQ_GT(ti->ti_ack, tp->snd_max)))
905	{
906	LogFlowFunc(("%d -> dropwithreset\n", __LINE__));
907	goto dropwithreset;
908	}
909
910	if (tiflags & TH_RST)
911	{
912	if (tiflags & TH_ACK)
913	tp = tcp_drop(pData, tp, 0); /* XXX Check t_softerror! */
914	LogFlowFunc(("%d -> drop\n", __LINE__));
915	goto drop;
916	}
917
918	if ((tiflags & TH_SYN) == 0)
919	{
920	LogFlowFunc(("%d -> drop\n", __LINE__));
921	goto drop;
922	}
923	if (tiflags & TH_ACK)
924	{
925	tp->snd_una = ti->ti_ack;
926	if (SEQ_LT(tp->snd_nxt, tp->snd_una))
927	tp->snd_nxt = tp->snd_una;
928	}
929
930	tp->t_timer[TCPT_REXMT] = 0;
931	tp->irs = ti->ti_seq;
932	tcp_rcvseqinit(tp);
933	tp->t_flags \|= TF_ACKNOW;
934	if (tiflags & TH_ACK && SEQ_GT(tp->snd_una, tp->iss))
935	{
936	tcpstat.tcps_connects++;
937	soisfconnected(so);
938	TCP_STATE_SWITCH_TO(tp, TCPS_ESTABLISHED);
939
940	/* Do window scaling on this connection? */
941	#if 0
942	if (( tp->t_flags & (TF_RCVD_SCALE\|TF_REQ_SCALE))
943	== (TF_RCVD_SCALE\|TF_REQ_SCALE))
944	{
945	tp->snd_scale = tp->requested_s_scale;
946	tp->rcv_scale = tp->request_r_scale;
947	}
948	#endif
949	(void) tcp_reass(pData, tp, (struct tcphdr )0, NULL, (struct mbuf )0);
950	/*
951	* if we didn't have to retransmit the SYN,
952	* use its rtt as our initial srtt & rtt var.
953	*/
954	if (tp->t_rtt)
955	tcp_xmit_timer(pData, tp, tp->t_rtt);
956	}
957	else
958	TCP_STATE_SWITCH_TO(tp, TCPS_SYN_RECEIVED);
959
960	trimthenstep6:
961	LogFlowFunc(("trimthenstep6:\n"));
962	/*
963	* Advance ti->ti_seq to correspond to first data byte.
964	* If data, trim to stay within window,
965	* dropping FIN if necessary.
966	*/
967	ti->ti_seq++;
968	if (ti->ti_len > tp->rcv_wnd)
969	{
970	todrop = ti->ti_len - tp->rcv_wnd;
971	m_adj(m, -todrop);
972	ti->ti_len = tp->rcv_wnd;
973	tiflags &= ~TH_FIN;
974	tcpstat.tcps_rcvpackafterwin++;
975	tcpstat.tcps_rcvbyteafterwin += todrop;
976	}
977	tp->snd_wl1 = ti->ti_seq - 1;
978	tp->rcv_up = ti->ti_seq;
979	LogFlowFunc(("%d -> step6\n", __LINE__));
980	goto step6;
981	} /* switch tp->t_state */
982	/*
983	* States other than LISTEN or SYN_SENT.
984	* First check timestamp, if present.
985	* Then check that at least some bytes of segment are within
986	* receive window. If segment begins before rcv_nxt,
987	* drop leading data (and SYN); if nothing left, just ack.
988	*
989	* RFC 1323 PAWS: If we have a timestamp reply on this segment
990	* and it's less than ts_recent, drop it.
991	*/
992	#if 0
993	if ( ts_present
994	&& (tiflags & TH_RST) == 0
995	&& tp->ts_recent
996	&& TSTMP_LT(ts_val, tp->ts_recent))
997	{
998	/* Check to see if ts_recent is over 24 days old. */
999	if ((int)(tcp_now - tp->ts_recent_age) > TCP_PAWS_IDLE)
1000	{
1001	/*
1002	* Invalidate ts_recent. If this segment updates
1003	* ts_recent, the age will be reset later and ts_recent
1004	* will get a valid value. If it does not, setting
1005	* ts_recent to zero will at least satisfy the
1006	* requirement that zero be placed in the timestamp
1007	* echo reply when ts_recent isn't valid. The
1008	* age isn't reset until we get a valid ts_recent
1009	* because we don't want out-of-order segments to be
1010	* dropped when ts_recent is old.
1011	*/
1012	tp->ts_recent = 0;
1013	}
1014	else
1015	{
1016	tcpstat.tcps_rcvduppack++;
1017	tcpstat.tcps_rcvdupbyte += ti->ti_len;
1018	tcpstat.tcps_pawsdrop++;
1019	goto dropafterack;
1020	}
1021	}
1022	#endif
1023
1024	todrop = tp->rcv_nxt - ti->ti_seq;
1025	if (todrop > 0)
1026	{
1027	if (tiflags & TH_SYN)
1028	{
1029	tiflags &= ~TH_SYN;
1030	ti->ti_seq++;
1031	if (ti->ti_urp > 1)
1032	ti->ti_urp--;
1033	else
1034	tiflags &= ~TH_URG;
1035	todrop--;
1036	}
1037	/*
1038	* Following if statement from Stevens, vol. 2, p. 960.
1039	*/
1040	if ( todrop > ti->ti_len
1041	\|\| ( todrop == ti->ti_len
1042	&& (tiflags & TH_FIN) == 0))
1043	{
1044	/*
1045	* Any valid FIN must be to the left of the window.
1046	* At this point the FIN must be a duplicate or out
1047	* of sequence; drop it.
1048	*/
1049	tiflags &= ~TH_FIN;
1050
1051	/*
1052	* Send an ACK to resynchronize and drop any data.
1053	* But keep on processing for RST or ACK.
1054	*/
1055	tp->t_flags \|= TF_ACKNOW;
1056	todrop = ti->ti_len;
1057	tcpstat.tcps_rcvduppack++;
1058	tcpstat.tcps_rcvdupbyte += todrop;
1059	}
1060	else
1061	{
1062	tcpstat.tcps_rcvpartduppack++;
1063	tcpstat.tcps_rcvpartdupbyte += todrop;
1064	}
1065	m_adj(m, todrop);
1066	ti->ti_seq += todrop;
1067	ti->ti_len -= todrop;
1068	if (ti->ti_urp > todrop)
1069	ti->ti_urp -= todrop;
1070	else
1071	{
1072	tiflags &= ~TH_URG;
1073	ti->ti_urp = 0;
1074	}
1075	}
1076	/*
1077	* If new data are received on a connection after the
1078	* user processes are gone, then RST the other end.
1079	*/
1080	if ( (so->so_state & SS_NOFDREF)
1081	&& tp->t_state > TCPS_CLOSE_WAIT && ti->ti_len)
1082	{
1083	tp = tcp_close(pData, tp);
1084	tcpstat.tcps_rcvafterclose++;
1085	LogFlowFunc(("%d -> dropwithreset\n", __LINE__));
1086	goto dropwithreset;
1087	}
1088
1089	/*
1090	* If segment ends after window, drop trailing data
1091	* (and PUSH and FIN); if nothing left, just ACK.
1092	*/
1093	todrop = (ti->ti_seq+ti->ti_len) - (tp->rcv_nxt+tp->rcv_wnd);
1094	if (todrop > 0)
1095	{
1096	tcpstat.tcps_rcvpackafterwin++;
1097	if (todrop >= ti->ti_len)
1098	{
1099	tcpstat.tcps_rcvbyteafterwin += ti->ti_len;
1100	/*
1101	* If a new connection request is received
1102	* while in TIME_WAIT, drop the old connection
1103	* and start over if the sequence numbers
1104	* are above the previous ones.
1105	*/
1106	if ( tiflags & TH_SYN
1107	&& tp->t_state == TCPS_TIME_WAIT
1108	&& SEQ_GT(ti->ti_seq, tp->rcv_nxt))
1109	{
1110	iss = tp->rcv_nxt + TCP_ISSINCR;
1111	tp = tcp_close(pData, tp);
1112	SOCKET_UNLOCK(tp->t_socket);
1113	LogFlowFunc(("%d -> findso\n", __LINE__));
1114	goto findso;
1115	}
1116	/*
1117	* If window is closed can only take segments at
1118	* window edge, and have to drop data and PUSH from
1119	* incoming segments. Continue processing, but
1120	* remember to ack. Otherwise, drop segment
1121	* and ack.
1122	*/
1123	if (tp->rcv_wnd == 0 && ti->ti_seq == tp->rcv_nxt)
1124	{
1125	tp->t_flags \|= TF_ACKNOW;
1126	tcpstat.tcps_rcvwinprobe++;
1127	}
1128	else
1129	{
1130	LogFlowFunc(("%d -> dropafterack\n", __LINE__));
1131	goto dropafterack;
1132	}
1133	}
1134	else
1135	tcpstat.tcps_rcvbyteafterwin += todrop;
1136	m_adj(m, -todrop);
1137	ti->ti_len -= todrop;
1138	tiflags &= ~(TH_PUSH\|TH_FIN);
1139	}
1140
1141	/*
1142	* If last ACK falls within this segment's sequence numbers,
1143	* record its timestamp.
1144	*/
1145	#if 0
1146	if ( ts_present
1147	&& SEQ_LEQ(ti->ti_seq, tp->last_ack_sent)
1148	&& SEQ_LT(tp->last_ack_sent, ti->ti_seq + ti->ti_len + ((tiflags & (TH_SYN\|TH_FIN)) != 0)))
1149	{
1150	tp->ts_recent_age = tcp_now;
1151	tp->ts_recent = ts_val;
1152	}
1153	#endif
1154
1155	/*
1156	* If the RST bit is set examine the state:
1157	* SYN_RECEIVED STATE:
1158	* If passive open, return to LISTEN state.
1159	* If active open, inform user that connection was refused.
1160	* ESTABLISHED, FIN_WAIT_1, FIN_WAIT2, CLOSE_WAIT STATES:
1161	* Inform user that connection was reset, and close tcb.
1162	* CLOSING, LAST_ACK, TIME_WAIT STATES
1163	* Close the tcb.
1164	*/
1165	if (tiflags&TH_RST)
1166	switch (tp->t_state)
1167	{
1168	case TCPS_SYN_RECEIVED:
1169	/* so->so_error = ECONNREFUSED; */
1170	LogFlowFunc(("%d -> close\n", __LINE__));
1171	goto close;
1172
1173	case TCPS_ESTABLISHED:
1174	case TCPS_FIN_WAIT_1:
1175	case TCPS_FIN_WAIT_2:
1176	case TCPS_CLOSE_WAIT:
1177	/* so->so_error = ECONNRESET; */
1178	close:
1179	LogFlowFunc(("close:\n"));
1180	TCP_STATE_SWITCH_TO(tp, TCPS_CLOSED);
1181	tcpstat.tcps_drops++;
1182	tp = tcp_close(pData, tp);
1183	LogFlowFunc(("%d -> drop\n", __LINE__));
1184	goto drop;
1185
1186	case TCPS_CLOSING:
1187	case TCPS_LAST_ACK:
1188	case TCPS_TIME_WAIT:
1189	tp = tcp_close(pData, tp);
1190	LogFlowFunc(("%d -> drop\n", __LINE__));
1191	goto drop;
1192	}
1193
1194	/*
1195	* If a SYN is in the window, then this is an
1196	* error and we send an RST and drop the connection.
1197	*/
1198	if (tiflags & TH_SYN)
1199	{
1200	tp = tcp_drop(pData, tp, 0);
1201	LogFlowFunc(("%d -> dropwithreset\n", __LINE__));
1202	goto dropwithreset;
1203	}
1204
1205	/*
1206	* If the ACK bit is off we drop the segment and return.
1207	*/
1208	if ((tiflags & TH_ACK) == 0)
1209	{
1210	LogFlowFunc(("%d -> drop\n", __LINE__));
1211	goto drop;
1212	}
1213
1214	/*
1215	* Ack processing.
1216	*/
1217	switch (tp->t_state)
1218	{
1219	/*
1220	* In SYN_RECEIVED state if the ack ACKs our SYN then enter
1221	* ESTABLISHED state and continue processing, otherwise
1222	* send an RST. una<=ack<=max
1223	*/
1224	case TCPS_SYN_RECEIVED:
1225	LogFlowFunc(("%d -> TCPS_SYN_RECEIVED\n", __LINE__));
1226	if ( SEQ_GT(tp->snd_una, ti->ti_ack)
1227	\|\| SEQ_GT(ti->ti_ack, tp->snd_max))
1228	goto dropwithreset;
1229	tcpstat.tcps_connects++;
1230	TCP_STATE_SWITCH_TO(tp, TCPS_ESTABLISHED);
1231	/*
1232	* The sent SYN is ack'ed with our sequence number +1
1233	* The first data byte already in the buffer will get
1234	* lost if no correction is made. This is only needed for
1235	* SS_CTL since the buffer is empty otherwise.
1236	* tp->snd_una++; or:
1237	*/
1238	tp->snd_una = ti->ti_ack;
1239	soisfconnected(so);
1240
1241	/* Do window scaling? */
1242	#if 0
1243	if ( (tp->t_flags & (TF_RCVD_SCALE\|TF_REQ_SCALE))
1244	== (TF_RCVD_SCALE\|TF_REQ_SCALE))
1245	{
1246	tp->snd_scale = tp->requested_s_scale;
1247	tp->rcv_scale = tp->request_r_scale;
1248	}
1249	#endif
1250	(void) tcp_reass(pData, tp, (struct tcphdr )0, (int )0, (struct mbuf *)0);
1251	tp->snd_wl1 = ti->ti_seq - 1;
1252	/* Avoid ack processing; snd_una==ti_ack => dup ack */
1253	LogFlowFunc(("%d -> synrx_to_est\n", __LINE__));
1254	goto synrx_to_est;
1255	/* fall into ... */
1256
1257	/*
1258	* In ESTABLISHED state: drop duplicate ACKs; ACK out of range
1259	* ACKs. If the ack is in the range
1260	* tp->snd_una < ti->ti_ack <= tp->snd_max
1261	* then advance tp->snd_una to ti->ti_ack and drop
1262	* data from the retransmission queue. If this ACK reflects
1263	* more up to date window information we update our window information.
1264	*/
1265	case TCPS_ESTABLISHED:
1266	case TCPS_FIN_WAIT_1:
1267	case TCPS_FIN_WAIT_2:
1268	case TCPS_CLOSE_WAIT:
1269	case TCPS_CLOSING:
1270	case TCPS_LAST_ACK:
1271	case TCPS_TIME_WAIT:
1272	LogFlowFunc(("%d -> TCPS_ESTABLISHED\|TCPS_FIN_WAIT_1\|TCPS_FIN_WAIT_2\|TCPS_CLOSE_WAIT\|"
1273	"TCPS_CLOSING\|TCPS_LAST_ACK\|TCPS_TIME_WAIT\n", __LINE__));
1274	if (SEQ_LEQ(ti->ti_ack, tp->snd_una))
1275	{
1276	if (ti->ti_len == 0 && tiwin == tp->snd_wnd)
1277	{
1278	tcpstat.tcps_rcvdupack++;
1279	Log2((" dup ack m = %p, so = %p\n", m, so));
1280	/*
1281	* If we have outstanding data (other than
1282	* a window probe), this is a completely
1283	* duplicate ack (ie, window info didn't
1284	* change), the ack is the biggest we've
1285	* seen and we've seen exactly our rexmt
1286	* threshold of them, assume a packet
1287	* has been dropped and retransmit it.
1288	* Kludge snd_nxt & the congestion
1289	* window so we send only this one
1290	* packet.
1291	*
1292	* We know we're losing at the current
1293	* window size so do congestion avoidance
1294	* (set ssthresh to half the current window
1295	* and pull our congestion window back to
1296	* the new ssthresh).
1297	*
1298	* Dup acks mean that packets have left the
1299	* network (they're now cached at the receiver)
1300	* so bump cwnd by the amount in the receiver
1301	* to keep a constant cwnd packets in the
1302	* network.
1303	*/
1304	if ( tp->t_timer[TCPT_REXMT] == 0
1305	\|\| ti->ti_ack != tp->snd_una)
1306	tp->t_dupacks = 0;
1307	else if (++tp->t_dupacks == tcprexmtthresh)
1308	{
1309	tcp_seq onxt = tp->snd_nxt;
1310	u_int win = min(tp->snd_wnd, tp->snd_cwnd) / 2 / tp->t_maxseg;
1311	if (win < 2)
1312	win = 2;
1313	tp->snd_ssthresh = win * tp->t_maxseg;
1314	tp->t_timer[TCPT_REXMT] = 0;
1315	tp->t_rtt = 0;
1316	tp->snd_nxt = ti->ti_ack;
1317	tp->snd_cwnd = tp->t_maxseg;
1318	(void) tcp_output(pData, tp);
1319	tp->snd_cwnd = tp->snd_ssthresh +
1320	tp->t_maxseg * tp->t_dupacks;
1321	if (SEQ_GT(onxt, tp->snd_nxt))
1322	tp->snd_nxt = onxt;
1323	LogFlowFunc(("%d -> drop\n", __LINE__));
1324	goto drop;
1325	}
1326	else if (tp->t_dupacks > tcprexmtthresh)
1327	{
1328	tp->snd_cwnd += tp->t_maxseg;
1329	(void) tcp_output(pData, tp);
1330	LogFlowFunc(("%d -> drop\n", __LINE__));
1331	goto drop;
1332	}
1333	}
1334	else
1335	tp->t_dupacks = 0;
1336	break;
1337	}
1338	synrx_to_est:
1339	LogFlowFunc(("synrx_to_est:\n"));
1340	/*
1341	* If the congestion window was inflated to account
1342	* for the other side's cached packets, retract it.
1343	*/
1344	if ( tp->t_dupacks > tcprexmtthresh
1345	&& tp->snd_cwnd > tp->snd_ssthresh)
1346	tp->snd_cwnd = tp->snd_ssthresh;
1347	tp->t_dupacks = 0;
1348	if (SEQ_GT(ti->ti_ack, tp->snd_max))
1349	{
1350	tcpstat.tcps_rcvacktoomuch++;
1351	LogFlowFunc(("%d -> dropafterack\n", __LINE__));
1352	goto dropafterack;
1353	}
1354	acked = ti->ti_ack - tp->snd_una;
1355	tcpstat.tcps_rcvackpack++;
1356	tcpstat.tcps_rcvackbyte += acked;
1357
1358	/*
1359	* If we have a timestamp reply, update smoothed
1360	* round trip time. If no timestamp is present but
1361	* transmit timer is running and timed sequence
1362	* number was acked, update smoothed round trip time.
1363	* Since we now have an rtt measurement, cancel the
1364	* timer backoff (cf., Phil Karn's retransmit alg.).
1365	* Recompute the initial retransmit timer.
1366	*/
1367	#if 0
1368	if (ts_present)
1369	tcp_xmit_timer(tp, tcp_now-ts_ecr+1);
1370	else
1371	#endif
1372	if (tp->t_rtt && SEQ_GT(ti->ti_ack, tp->t_rtseq))
1373	tcp_xmit_timer(pData, tp, tp->t_rtt);
1374
1375	/*
1376	* If all outstanding data is acked, stop retransmit
1377	* timer and remember to restart (more output or persist).
1378	* If there is more data to be acked, restart retransmit
1379	* timer, using current (possibly backed-off) value.
1380	*/
1381	if (ti->ti_ack == tp->snd_max)
1382	{
1383	tp->t_timer[TCPT_REXMT] = 0;
1384	needoutput = 1;
1385	}
1386	else if (tp->t_timer[TCPT_PERSIST] == 0)
1387	tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
1388	/*
1389	* When new data is acked, open the congestion window.
1390	* If the window gives us less than ssthresh packets
1391	* in flight, open exponentially (maxseg per packet).
1392	* Otherwise open linearly: maxseg per window
1393	* (maxseg^2 / cwnd per packet).
1394	*/
1395	{
1396	register u_int cw = tp->snd_cwnd;
1397	register u_int incr = tp->t_maxseg;
1398
1399	if (cw > tp->snd_ssthresh)
1400	incr = incr * incr / cw;
1401	tp->snd_cwnd = min(cw + incr, TCP_MAXWIN<<tp->snd_scale);
1402	}
1403	if (acked > SBUF_LEN(&so->so_snd))
1404	{
1405	tp->snd_wnd -= SBUF_LEN(&so->so_snd);
1406	sbdrop(&so->so_snd, (int)so->so_snd.sb_cc);
1407	ourfinisacked = 1;
1408	}
1409	else
1410	{
1411	sbdrop(&so->so_snd, acked);
1412	tp->snd_wnd -= acked;
1413	ourfinisacked = 0;
1414	}
1415	/*
1416	* XXX sowwakup is called when data is acked and there's room for
1417	* for more data... it should read() the socket
1418	*/
1419	#if 0
1420	if (so->so_snd.sb_flags & SB_NOTIFY)
1421	sowwakeup(so);
1422	#endif
1423	tp->snd_una = ti->ti_ack;
1424	if (SEQ_LT(tp->snd_nxt, tp->snd_una))
1425	tp->snd_nxt = tp->snd_una;
1426
1427	switch (tp->t_state)
1428	{
1429	/*
1430	* In FIN_WAIT_1 STATE in addition to the processing
1431	* for the ESTABLISHED state if our FIN is now acknowledged
1432	* then enter FIN_WAIT_2.
1433	*/
1434	case TCPS_FIN_WAIT_1:
1435	if (ourfinisacked)
1436	{
1437	/*
1438	* If we can't receive any more
1439	* data, then closing user can proceed.
1440	* Starting the timer is contrary to the
1441	* specification, but if we don't get a FIN
1442	* we'll hang forever.
1443	*/
1444	if (so->so_state & SS_FCANTRCVMORE)
1445	{
1446	soisfdisconnected(so);
1447	tp->t_timer[TCPT_2MSL] = tcp_maxidle;
1448	}
1449	TCP_STATE_SWITCH_TO(tp, TCPS_FIN_WAIT_2);
1450	}
1451	break;
1452
1453	/*
1454	* In CLOSING STATE in addition to the processing for
1455	* the ESTABLISHED state if the ACK acknowledges our FIN
1456	* then enter the TIME-WAIT state, otherwise ignore
1457	* the segment.
1458	*/
1459	case TCPS_CLOSING:
1460	if (ourfinisacked)
1461	{
1462	TCP_STATE_SWITCH_TO(tp, TCPS_TIME_WAIT);
1463	tcp_canceltimers(tp);
1464	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1465	soisfdisconnected(so);
1466	}
1467	break;
1468
1469	/*
1470	* In LAST_ACK, we may still be waiting for data to drain
1471	* and/or to be acked, as well as for the ack of our FIN.
1472	* If our FIN is now acknowledged, delete the TCB,
1473	* enter the closed state and return.
1474	*/
1475	case TCPS_LAST_ACK:
1476	if (ourfinisacked)
1477	{
1478	tp = tcp_close(pData, tp);
1479	LogFlowFunc(("%d -> drop\n", __LINE__));
1480	goto drop;
1481	}
1482	break;
1483
1484	/*
1485	* In TIME_WAIT state the only thing that should arrive
1486	* is a retransmission of the remote FIN. Acknowledge
1487	* it and restart the finack timer.
1488	*/
1489	case TCPS_TIME_WAIT:
1490	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1491	LogFlowFunc(("%d -> dropafterack\n", __LINE__));
1492	goto dropafterack;
1493	}
1494	} /* switch(tp->t_state) */
1495
1496	step6:
1497	LogFlowFunc(("step6:\n"));
1498	/*
1499	* Update window information.
1500	* Don't look at window if no ACK: TAC's send garbage on first SYN.
1501	*/
1502	if ( (tiflags & TH_ACK)
1503	&& ( SEQ_LT(tp->snd_wl1, ti->ti_seq)
1504	\|\| ( tp->snd_wl1 == ti->ti_seq
1505	&& ( SEQ_LT(tp->snd_wl2, ti->ti_ack)
1506	\|\| ( tp->snd_wl2 == ti->ti_ack
1507	&& tiwin > tp->snd_wnd)))))
1508	{
1509	/* keep track of pure window updates */
1510	if ( ti->ti_len == 0
1511	&& tp->snd_wl2 == ti->ti_ack
1512	&& tiwin > tp->snd_wnd)
1513	tcpstat.tcps_rcvwinupd++;
1514	tp->snd_wnd = tiwin;
1515	tp->snd_wl1 = ti->ti_seq;
1516	tp->snd_wl2 = ti->ti_ack;
1517	if (tp->snd_wnd > tp->max_sndwnd)
1518	tp->max_sndwnd = tp->snd_wnd;
1519	needoutput = 1;
1520	}
1521
1522	/*
1523	* Process segments with URG.
1524	*/
1525	if ((tiflags & TH_URG) && ti->ti_urp &&
1526	TCPS_HAVERCVDFIN(tp->t_state) == 0)
1527	{
1528	/*
1529	* This is a kludge, but if we receive and accept
1530	* random urgent pointers, we'll crash in
1531	* soreceive. It's hard to imagine someone
1532	* actually wanting to send this much urgent data.
1533	*/
1534	if (ti->ti_urp + so->so_rcv.sb_cc > so->so_rcv.sb_datalen)
1535	{
1536	ti->ti_urp = 0;
1537	tiflags &= ~TH_URG;
1538	LogFlowFunc(("%d -> dodata\n", __LINE__));
1539	goto dodata;
1540	}
1541
1542	/*
1543	* If this segment advances the known urgent pointer,
1544	* then mark the data stream. This should not happen
1545	* in CLOSE_WAIT, CLOSING, LAST_ACK or TIME_WAIT STATES since
1546	* a FIN has been received from the remote side.
1547	* In these states we ignore the URG.
1548	*
1549	* According to RFC961 (Assigned Protocols),
1550	* the urgent pointer points to the last octet
1551	* of urgent data. We continue, however,
1552	* to consider it to indicate the first octet
1553	* of data past the urgent section as the original
1554	* spec states (in one of two places).
1555	*/
1556	if (SEQ_GT(ti->ti_seq+ti->ti_urp, tp->rcv_up))
1557	{
1558	tp->rcv_up = ti->ti_seq + ti->ti_urp;
1559	so->so_urgc = SBUF_LEN(&so->so_rcv) +
1560	(tp->rcv_up - tp->rcv_nxt); /* -1; */
1561	tp->rcv_up = ti->ti_seq + ti->ti_urp;
1562	}
1563	}
1564	else
1565	/*
1566	* If no out of band data is expected,
1567	* pull receive urgent pointer along
1568	* with the receive window.
1569	*/
1570	if (SEQ_GT(tp->rcv_nxt, tp->rcv_up))
1571	tp->rcv_up = tp->rcv_nxt;
1572	dodata:
1573	LogFlowFunc(("dodata:\n"));
1574
1575	/*
1576	* If this is a small packet, then ACK now - with Nagel
1577	* congestion avoidance sender won't send more until
1578	* he gets an ACK.
1579	*
1580	* XXX: In case you wonder... The magic "27" below is ESC that
1581	* presumably starts a terminal escape-sequence and that we want
1582	* to ACK ASAP. [Original slirp code had three different
1583	* heuristics to chose from here and in the header prediction case
1584	* above, but the commented out alternatives were lost and the
1585	* header prediction case that had an expanded comment about this
1586	* has been modified to always send an ACK].
1587	*/
1588	if ( ti->ti_len
1589	&& (unsigned)ti->ti_len <= 5
1590	&& ((struct tcpiphdr_2 *)ti)->first_char == (char)27)
1591	{
1592	tp->t_flags \|= TF_ACKNOW;
1593	}
1594
1595	/*
1596	* Process the segment text, merging it into the TCP sequencing queue,
1597	* and arranging for acknowledgment of receipt if necessary.
1598	* This process logically involves adjusting tp->rcv_wnd as data
1599	* is presented to the user (this happens in tcp_usrreq.c,
1600	* case PRU_RCVD). If a FIN has already been received on this
1601	* connection then we just ignore the text.
1602	*/
1603	if ( (ti->ti_len \|\| (tiflags&TH_FIN))
1604	&& TCPS_HAVERCVDFIN(tp->t_state) == 0)
1605	{
1606	if ( ti->ti_seq == tp->rcv_nxt
1607	&& LIST_EMPTY(&tp->t_segq)
1608	&& tp->t_state == TCPS_ESTABLISHED)
1609	{
1610	DELAY_ACK(tp, ti); /* little bit different from BSD declaration see netinet/tcp_input.c */
1611	tp->rcv_nxt += tlen;
1612	tiflags = ti->ti_t.th_flags & TH_FIN;
1613	tcpstat.tcps_rcvpack++;
1614	tcpstat.tcps_rcvbyte += tlen;
1615	if (so->so_state & SS_FCANTRCVMORE)
1616	m_freem(pData, m);
1617	else
1618	sbappend(pData, so, m);
1619	}
1620	else
1621	{
1622	tiflags = tcp_reass(pData, tp, &ti->ti_t, &tlen, m);
1623	tp->t_flags \|= TF_ACKNOW;
1624	}
1625	/*
1626	* Note the amount of data that peer has sent into
1627	* our window, in order to estimate the sender's
1628	* buffer size.
1629	*/
1630	len = SBUF_SIZE(&so->so_rcv) - (tp->rcv_adv - tp->rcv_nxt);
1631	}
1632	else
1633	{
1634	m_freem(pData, m);
1635	tiflags &= ~TH_FIN;
1636	}
1637
1638	/*
1639	* If FIN is received ACK the FIN and let the user know
1640	* that the connection is closing.
1641	*/
1642	if (tiflags & TH_FIN)
1643	{
1644	if (TCPS_HAVERCVDFIN(tp->t_state) == 0)
1645	{
1646	/*
1647	* If we receive a FIN we can't send more data,
1648	* set it SS_FDRAIN
1649	* Shutdown the socket if there is no rx data in the
1650	* buffer.
1651	* soread() is called on completion of shutdown() and
1652	* will got to TCPS_LAST_ACK, and use tcp_output()
1653	* to send the FIN.
1654	*/
1655	/* sofcantrcvmore(so); */
1656	sofwdrain(so);
1657
1658	tp->t_flags \|= TF_ACKNOW;
1659	tp->rcv_nxt++;
1660	}
1661	switch (tp->t_state)
1662	{
1663	/*
1664	* In SYN_RECEIVED and ESTABLISHED STATES
1665	* enter the CLOSE_WAIT state.
1666	*/
1667	case TCPS_SYN_RECEIVED:
1668	case TCPS_ESTABLISHED:
1669	TCP_STATE_SWITCH_TO(tp, TCPS_CLOSE_WAIT);
1670	break;
1671
1672	/*
1673	* If still in FIN_WAIT_1 STATE FIN has not been acked so
1674	* enter the CLOSING state.
1675	*/
1676	case TCPS_FIN_WAIT_1:
1677	TCP_STATE_SWITCH_TO(tp, TCPS_CLOSING);
1678	break;
1679
1680	/*
1681	* In FIN_WAIT_2 state enter the TIME_WAIT state,
1682	* starting the time-wait timer, turning off the other
1683	* standard timers.
1684	*/
1685	case TCPS_FIN_WAIT_2:
1686	TCP_STATE_SWITCH_TO(tp, TCPS_TIME_WAIT);
1687	tcp_canceltimers(tp);
1688	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1689	soisfdisconnected(so);
1690	break;
1691
1692	/*
1693	* In TIME_WAIT state restart the 2 MSL time_wait timer.
1694	*/
1695	case TCPS_TIME_WAIT:
1696	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1697	break;
1698	}
1699	}
1700
1701	/*
1702	* Return any desired output.
1703	*/
1704	if (needoutput \|\| (tp->t_flags & TF_ACKNOW))
1705	tcp_output(pData, tp);
1706
1707	SOCKET_UNLOCK(so);
1708	STAM_PROFILE_STOP(&pData->StatTCP_input, counter_input);
1709	LogFlowFuncLeave();
1710	return;
1711
1712	dropafterack:
1713	LogFlowFunc(("dropafterack:\n"));
1714	/*
1715	* Generate an ACK dropping incoming segment if it occupies
1716	* sequence space, where the ACK reflects our state.
1717	*/
1718	if (tiflags & TH_RST)
1719	{
1720	LogFlowFunc(("%d -> drop\n", __LINE__));
1721	goto drop;
1722	}
1723	m_freem(pData, m);
1724	tp->t_flags \|= TF_ACKNOW;
1725	(void) tcp_output(pData, tp);
1726	SOCKET_UNLOCK(so);
1727	STAM_PROFILE_STOP(&pData->StatTCP_input, counter_input);
1728	LogFlowFuncLeave();
1729	return;
1730
1731	dropwithreset:
1732	LogFlowFunc(("dropwithreset:\n"));
1733	/* reuses m if m!=NULL, m_free() unnecessary */
1734	if (tiflags & TH_ACK)
1735	tcp_respond(pData, tp, ti, m, (tcp_seq)0, ti->ti_ack, TH_RST);
1736	else
1737	{
1738	if (tiflags & TH_SYN)
1739	ti->ti_len++;
1740	tcp_respond(pData, tp, ti, m, ti->ti_seq+ti->ti_len, (tcp_seq)0,
1741	TH_RST\|TH_ACK);
1742	}
1743
1744	if (so != &tcb)
1745	SOCKET_UNLOCK(so);
1746	STAM_PROFILE_STOP(&pData->StatTCP_input, counter_input);
1747	LogFlowFuncLeave();
1748	return;
1749
1750	drop:
1751	LogFlowFunc(("drop:\n"));
1752	/*
1753	* Drop space held by incoming segment and return.
1754	*/
1755	m_freem(pData, m);
1756
1757	#ifdef VBOX_WITH_SLIRP_MT
1758	if (RTCritSectIsOwned(&so->so_mutex))
1759	{
1760	SOCKET_UNLOCK(so);
1761	}
1762	#endif
1763
1764	STAM_PROFILE_STOP(&pData->StatTCP_input, counter_input);
1765	LogFlowFuncLeave();
1766	return;
1767	}
1768
1769
1770	void
1771	tcp_fconnect_failed(PNATState pData, struct socket *so, int sockerr)
1772	{
1773	struct tcpcb *tp;
1774	int code;
1775
1776	Log2(("NAT: connect error %d %R[natsock]\n", sockerr, so));
1777
1778	Assert(so->so_state & SS_ISFCONNECTING);
1779	so->so_state = SS_NOFDREF;
1780
1781	if (sockerr == ECONNREFUSED \|\| sockerr == ECONNRESET)
1782	{
1783	/* hand off to tcp_input():cont_conn to send RST */
1784	TCP_INPUT(pData, NULL, 0, so);
1785	return;
1786	}
1787
1788	tp = sototcpcb(so);
1789	if (RT_UNLIKELY(tp == NULL)) /* should never happen */
1790	{
1791	LogRel(("NAT: tp == NULL %R[natsock]\n", so));
1792	sofree(pData, so);
1793	return;
1794	}
1795
1796	if (sockerr == ENETUNREACH \|\| sockerr == ENETDOWN)
1797	code = ICMP_UNREACH_NET;
1798	else if (sockerr == EHOSTUNREACH \|\| sockerr == EHOSTDOWN)
1799	code = ICMP_UNREACH_HOST;
1800	else
1801	code = -1;
1802
1803	if (code >= 0)
1804	{
1805	struct ip *oip;
1806	unsigned ohdrlen;
1807	struct mbuf *m;
1808
1809	if (RT_UNLIKELY(so->so_ohdr == NULL))
1810	goto out;
1811
1812	oip = (struct ip *)so->so_ohdr;
1813	ohdrlen = oip->ip_hl * 4 + 8;
1814
1815	m = m_gethdr(pData, M_NOWAIT, MT_HEADER);
1816	if (RT_UNLIKELY(m == NULL))
1817	goto out;
1818
1819	m_copyback(pData, m, 0, ohdrlen, (caddr_t)so->so_ohdr);
1820	m->m_pkthdr.header = mtod(m, void *);
1821
1822	icmp_error(pData, m, ICMP_UNREACH, code, 0, NULL);
1823	}
1824
1825	out:
1826	tcp_close(pData, tp);
1827	}
1828
1829
1830	void
1831	tcp_dooptions(PNATState pData, struct tcpcb tp, u_char cp, int cnt, struct tcpiphdr *ti)
1832	{
1833	u_int16_t mss;
1834	int opt, optlen;
1835
1836	LogFlowFunc(("tcp_dooptions: tp = %R[tcpcb793], cnt=%i\n", tp, cnt));
1837
1838	for (; cnt > 0; cnt -= optlen, cp += optlen)
1839	{
1840	opt = cp[0];
1841	if (opt == TCPOPT_EOL)
1842	break;
1843	if (opt == TCPOPT_NOP)
1844	optlen = 1;
1845	else
1846	{
1847	optlen = cp[1];
1848	if (optlen <= 0)
1849	break;
1850	}
1851	switch (opt)
1852	{
1853	default:
1854	continue;
1855
1856	case TCPOPT_MAXSEG:
1857	if (optlen != TCPOLEN_MAXSEG)
1858	continue;
1859	if (!(ti->ti_flags & TH_SYN))
1860	continue;
1861	memcpy((char ) &mss, (char ) cp + 2, sizeof(mss));
1862	NTOHS(mss);
1863	(void) tcp_mss(pData, tp, mss); /* sets t_maxseg */
1864	break;
1865
1866	#if 0
1867	case TCPOPT_WINDOW:
1868	if (optlen != TCPOLEN_WINDOW)
1869	continue;
1870	if (!(ti->ti_flags & TH_SYN))
1871	continue;
1872	tp->t_flags \|= TF_RCVD_SCALE;
1873	tp->requested_s_scale = min(cp[2], TCP_MAX_WINSHIFT);
1874	break;
1875
1876	case TCPOPT_TIMESTAMP:
1877	if (optlen != TCPOLEN_TIMESTAMP)
1878	continue;
1879	*ts_present = 1;
1880	memcpy((char ) ts_val, (char )cp + 2, sizeof(*ts_val));
1881	NTOHL(*ts_val);
1882	memcpy((char ) ts_ecr, (char )cp + 6, sizeof(*ts_ecr));
1883	NTOHL(*ts_ecr);
1884
1885	/*
1886	* A timestamp received in a SYN makes
1887	* it ok to send timestamp requests and replies.
1888	*/
1889	if (ti->ti_flags & TH_SYN)
1890	{
1891	tp->t_flags \|= TF_RCVD_TSTMP;
1892	tp->ts_recent = *ts_val;
1893	tp->ts_recent_age = tcp_now;
1894	}
1895	break;
1896	#endif
1897	}
1898	}
1899	}
1900
1901
1902	/*
1903	* Pull out of band byte out of a segment so
1904	* it doesn't appear in the user's data queue.
1905	* It is still reflected in the segment length for
1906	* sequencing purposes.
1907	*/
1908
1909	#if 0
1910	void
1911	tcp_pulloutofband(struct socket so, struct tcpiphdr ti, struct mbuf *m)
1912	{
1913	int cnt = ti->ti_urp - 1;
1914
1915	while (cnt >= 0)
1916	{
1917	if (m->m_len > cnt)
1918	{
1919	char *cp = mtod(m, caddr_t) + cnt;
1920	struct tcpcb *tp = sototcpcb(so);
1921
1922	tp->t_iobc = *cp;
1923	tp->t_oobflags \|= TCPOOB_HAVEDATA;
1924	memcpy(sp, cp+1, (unsigned)(m->m_len - cnt - 1));
1925	m->m_len--;
1926	return;
1927	}
1928	cnt -= m->m_len;
1929	m = m->m_next; /* XXX WRONG! Fix it! */
1930	if (m == 0)
1931	break;
1932	}
1933	panic("tcp_pulloutofband");
1934	}
1935	#endif
1936
1937	/*
1938	* Collect new round-trip time estimate
1939	* and update averages and current timeout.
1940	*/
1941
1942	void
1943	tcp_xmit_timer(PNATState pData, register struct tcpcb *tp, int rtt)
1944	{
1945	register short delta;
1946
1947	LogFlowFunc(("ENTER: tcp_xmit_timer: tp = %R[tcpcb793] rtt = %d\n", tp, rtt));
1948
1949	tcpstat.tcps_rttupdated++;
1950	if (tp->t_srtt != 0)
1951	{
1952	/*
1953	* srtt is stored as fixed point with 3 bits after the
1954	* binary point (i.e., scaled by 8). The following magic
1955	* is equivalent to the smoothing algorithm in rfc793 with
1956	* an alpha of .875 (srtt = rtt/8 + srtt*7/8 in fixed
1957	* point). Adjust rtt to origin 0.
1958	*/
1959	delta = rtt - 1 - (tp->t_srtt >> TCP_RTT_SHIFT);
1960	if ((tp->t_srtt += delta) <= 0)
1961	tp->t_srtt = 1;
1962	/*
1963	* We accumulate a smoothed rtt variance (actually, a
1964	* smoothed mean difference), then set the retransmit
1965	* timer to smoothed rtt + 4 times the smoothed variance.
1966	* rttvar is stored as fixed point with 2 bits after the
1967	* binary point (scaled by 4). The following is
1968	* equivalent to rfc793 smoothing with an alpha of .75
1969	* (rttvar = rttvar*3/4 + \|delta\| / 4). This replaces
1970	* rfc793's wired-in beta.
1971	*/
1972	if (delta < 0)
1973	delta = -delta;
1974	delta -= (tp->t_rttvar >> TCP_RTTVAR_SHIFT);
1975	if ((tp->t_rttvar += delta) <= 0)
1976	tp->t_rttvar = 1;
1977	}
1978	else
1979	{
1980	/*
1981	* No rtt measurement yet - use the unsmoothed rtt.
1982	* Set the variance to half the rtt (so our first
1983	* retransmit happens at 3*rtt).
1984	*/
1985	tp->t_srtt = rtt << TCP_RTT_SHIFT;
1986	tp->t_rttvar = rtt << (TCP_RTTVAR_SHIFT - 1);
1987	}
1988	tp->t_rtt = 0;
1989	tp->t_rxtshift = 0;
1990
1991	/*
1992	* the retransmit should happen at rtt + 4 * rttvar.
1993	* Because of the way we do the smoothing, srtt and rttvar
1994	* will each average +1/2 tick of bias. When we compute
1995	* the retransmit timer, we want 1/2 tick of rounding and
1996	* 1 extra tick because of +-1/2 tick uncertainty in the
1997	* firing of the timer. The bias will give us exactly the
1998	* 1.5 tick we need. But, because the bias is
1999	* statistical, we have to test that we don't drop below
2000	* the minimum feasible timer (which is 2 ticks).
2001	*/
2002	TCPT_RANGESET(tp->t_rxtcur, TCP_REXMTVAL(tp),
2003	(short)tp->t_rttmin, TCPTV_REXMTMAX); /* XXX */
2004
2005	/*
2006	* We received an ack for a packet that wasn't retransmitted;
2007	* it is probably safe to discard any error indications we've
2008	* received recently. This isn't quite right, but close enough
2009	* for now (a route might have failed after we sent a segment,
2010	* and the return path might not be symmetrical).
2011	*/
2012	tp->t_softerror = 0;
2013	}
2014
2015	/*
2016	* Determine a reasonable value for maxseg size.
2017	* If the route is known, check route for mtu.
2018	* If none, use an mss that can be handled on the outgoing
2019	* interface without forcing IP to fragment; if bigger than
2020	* an mbuf cluster (MCLBYTES), round down to nearest multiple of MCLBYTES
2021	* to utilize large mbufs. If no route is found, route has no mtu,
2022	* or the destination isn't local, use a default, hopefully conservative
2023	* size (usually 512 or the default IP max size, but no more than the mtu
2024	* of the interface), as we can't discover anything about intervening
2025	* gateways or networks. We also initialize the congestion/slow start
2026	* window to be a single segment if the destination isn't local.
2027	* While looking at the routing entry, we also initialize other path-dependent
2028	* parameters from pre-set or cached values in the routing entry.
2029	*/
2030
2031	int
2032	tcp_mss(PNATState pData, register struct tcpcb *tp, u_int offer)
2033	{
2034	struct socket *so = tp->t_socket;
2035	int mss;
2036
2037	LogFlowFunc(("ENTER: tcp_mss: offer=%u, t_maxseg=%u; tp=%R[natsock]\n",
2038	offer, (unsigned int)tp->t_maxseg, so));
2039
2040	mss = min(if_mtu, if_mru) - sizeof(struct tcpiphdr);
2041	if (offer)
2042	mss = min(mss, offer);
2043	mss = max(mss, 32);
2044	if (mss < tp->t_maxseg \|\| offer != 0)
2045	tp->t_maxseg = mss;
2046
2047	tp->snd_cwnd = mss;
2048
2049	sbreserve(pData, &so->so_snd, tcp_sndspace+((tcp_sndspace%mss)?(mss-(tcp_sndspace%mss)):0));
2050	sbreserve(pData, &so->so_rcv, tcp_rcvspace+((tcp_rcvspace%mss)?(mss-(tcp_rcvspace%mss)):0));
2051
2052	LogFlowFunc(("LEAVE: mss=%d\n", mss));
2053	return mss;
2054	}

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

source: vbox/trunk/src/VBox/Devices/Network/slirp/tcp_input.c@ 93115

Download in other formats: