tcp_input.c@ 14309

Last change on this file since 14309 was 14309, checked in by vboxsync, 16 years ago
#else /* what comes now */
Property svn:eol-style set to `native`
File size: 54.4 KB

Line
1	/*
2	* Copyright (c) 1982, 1986, 1988, 1990, 1993, 1994
3	* The Regents of the University of California. All rights reserved.
4	*
5	* Redistribution and use in source and binary forms, with or without
6	* modification, are permitted provided that the following conditions
7	* are met:
8	* 1. Redistributions of source code must retain the above copyright
9	* notice, this list of conditions and the following disclaimer.
10	* 2. Redistributions in binary form must reproduce the above copyright
11	* notice, this list of conditions and the following disclaimer in the
12	* documentation and/or other materials provided with the distribution.
13	* 3. All advertising materials mentioning features or use of this software
14	* must display the following acknowledgement:
15	* This product includes software developed by the University of
16	* California, Berkeley and its contributors.
17	* 4. Neither the name of the University nor the names of its contributors
18	* may be used to endorse or promote products derived from this software
19	* without specific prior written permission.
20	*
21	* THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22	* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23	* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24	* ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25	* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26	* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27	* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28	* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29	* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30	* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31	* SUCH DAMAGE.
32	*
33	* @(#)tcp_input.c 8.5 (Berkeley) 4/10/94
34	* tcp_input.c,v 1.10 1994/10/13 18:36:32 wollman Exp
35	*/
36
37	/*
38	* Changes and additions relating to SLiRP
39	* Copyright (c) 1995 Danny Gasparovski.
40	*
41	* Please read the file COPYRIGHT for the
42	* terms and conditions of the copyright.
43	*/
44
45	#include <slirp.h>
46	#include "ip_icmp.h"
47
48
49	#define TCP_PAWS_IDLE (24 * 24 * 60 * 60 * PR_SLOWHZ)
50
51	/* for modulo comparisons of timestamps */
52	#define TSTMP_LT(a,b) ((int)((a)-(b)) < 0)
53	#define TSTMP_GEQ(a,b) ((int)((a)-(b)) >= 0)
54
55	#ifndef VBOX_WITH_BSD_TCP_REASS
56	/*
57	* Insert segment ti into reassembly queue of tcp with
58	* control block tp. Return TH_FIN if reassembly now includes
59	* a segment with FIN. The macro form does the common case inline
60	* (segment is the next to be received on an established connection,
61	* and the queue is empty), avoiding linkage into and removal
62	* from the queue and repetition of various conversions.
63	* Set DELACK for segments received in order, but ack immediately
64	* when segments are out of order (so fast retransmit can work).
65	*/
66	#ifdef TCP_ACK_HACK
67	#define TCP_REASS(pData, tp, ti, m, so, flags) {\
68	if ((ti)->ti_seq == (tp)->rcv_nxt && \
69	u32_to_ptr((pData), (tp)->seg_next, struct tcpcb *) == (tp) && \
70	(tp)->t_state == TCPS_ESTABLISHED) {\
71	if (ti->ti_flags & TH_PUSH) \
72	tp->t_flags \|= TF_ACKNOW; \
73	else \
74	tp->t_flags \|= TF_DELACK; \
75	(tp)->rcv_nxt += (ti)->ti_len; \
76	flags = (ti)->ti_flags & TH_FIN; \
77	tcpstat.tcps_rcvpack++;\
78	tcpstat.tcps_rcvbyte += (ti)->ti_len;\
79	if (so->so_emu) { \
80	if (tcp_emu((pData), (so),(m))) sbappend((pData), (so), (m)); \
81	} else \
82	sbappend((pData), (so), (m)); \
83	/* sorwakeup(so); */ \
84	} else {\
85	(flags) = tcp_reass((pData), (tp), (ti), (m)); \
86	tp->t_flags \|= TF_ACKNOW; \
87	} \
88	}
89	#else
90	#define TCP_REASS(pData, tp, ti, m, so, flags) { \
91	if ((ti)->ti_seq == (tp)->rcv_nxt && \
92	u32_to_ptr((pData), (tp)->seg_next, struct tcpcb *) == (tp) && \
93	(tp)->t_state == TCPS_ESTABLISHED) { \
94	tp->t_flags \|= TF_DELACK; \
95	(tp)->rcv_nxt += (ti)->ti_len; \
96	flags = (ti)->ti_flags & TH_FIN; \
97	tcpstat.tcps_rcvpack++;\
98	tcpstat.tcps_rcvbyte += (ti)->ti_len;\
99	if (so->so_emu) { \
100	if (tcp_emu((pData), (so),(m))) sbappend((pData), (so), (m)); \
101	} else \
102	sbappend((pData), (so), (m)); \
103	/* sorwakeup(so); */ \
104	} else { \
105	(flags) = tcp_reass((pData), (tp), (ti), (m)); \
106	tp->t_flags \|= TF_ACKNOW; \
107	} \
108	}
109	#endif
110
111	int
112	tcp_reass(PNATState pData, register struct tcpcb tp, register struct tcpiphdr ti, struct mbuf *m)
113	{
114	register struct tcpiphdr *q;
115	struct socket *so = tp->t_socket;
116	int flags;
117
118	/*
119	* Call with ti==0 after become established to
120	* force pre-ESTABLISHED data up to user socket.
121	*/
122	if (ti == 0)
123	goto present;
124
125	/*
126	* Find a segment which begins after this one does.
127	*/
128	for (q = u32_to_ptr(pData, tp->seg_next, struct tcpiphdr ); q != (struct tcpiphdr )tp;
129	q = u32_to_ptr(pData, q->ti_next, struct tcpiphdr *))
130	if (SEQ_GT(q->ti_seq, ti->ti_seq))
131	break;
132
133	/*
134	* If there is a preceding segment, it may provide some of
135	* our data already. If so, drop the data from the incoming
136	* segment. If it provides all of our data, drop us.
137	*/
138	if (u32_to_ptr(pData, q->ti_prev, struct tcpiphdr ) != (struct tcpiphdr )tp) {
139	register int i;
140	q = u32_to_ptr(pData, q->ti_prev, struct tcpiphdr *);
141	/* conversion to int (in i) handles seq wraparound */
142	i = q->ti_seq + q->ti_len - ti->ti_seq;
143	if (i > 0) {
144	if (i >= ti->ti_len) {
145	tcpstat.tcps_rcvduppack++;
146	tcpstat.tcps_rcvdupbyte += ti->ti_len;
147	m_freem(pData, m);
148	/*
149	* Try to present any queued data
150	* at the left window edge to the user.
151	* This is needed after the 3-WHS
152	* completes.
153	*/
154	goto present; /* ??? */
155	}
156	m_adj(m, i);
157	ti->ti_len -= i;
158	ti->ti_seq += i;
159	}
160	q = u32_to_ptr(pData, q->ti_next, struct tcpiphdr *);
161	}
162	tcpstat.tcps_rcvoopack++;
163	tcpstat.tcps_rcvoobyte += ti->ti_len;
164	REASS_MBUF_SET(ti, m); /* XXX */
165
166	/*
167	* While we overlap succeeding segments trim them or,
168	* if they are completely covered, dequeue them.
169	*/
170	while (q != (struct tcpiphdr *)tp) {
171	register int i = (ti->ti_seq + ti->ti_len) - q->ti_seq;
172	if (i <= 0)
173	break;
174	if (i < q->ti_len) {
175	q->ti_seq += i;
176	q->ti_len -= i;
177	m_adj(REASS_MBUF_GET(q), i);
178	break;
179	}
180	q = u32_to_ptr(pData, q->ti_next, struct tcpiphdr *);
181	m = REASS_MBUF_GET(u32_to_ptr(pData, q->ti_prev, struct tcpiphdr *));
182	remque_32(pData, u32_to_ptr(pData, q->ti_prev, struct tcpiphdr *));
183	m_freem(pData, m);
184	}
185
186	/*
187	* Stick new segment in its place.
188	*/
189	insque_32(pData, ti, u32_to_ptr(pData, q->ti_prev, struct tcpiphdr *));
190
191	present:
192	/*
193	* Present data to user, advancing rcv_nxt through
194	* completed sequence space.
195	*/
196	if (!TCPS_HAVEESTABLISHED(tp->t_state))
197	return (0);
198	ti = u32_to_ptr(pData, tp->seg_next, struct tcpiphdr *);
199	if (ti == (struct tcpiphdr *)tp \|\| ti->ti_seq != tp->rcv_nxt)
200	return (0);
201	if (tp->t_state == TCPS_SYN_RECEIVED && ti->ti_len)
202	return (0);
203	do {
204	tp->rcv_nxt += ti->ti_len;
205	flags = ti->ti_flags & TH_FIN;
206	remque_32(pData, ti);
207	m = REASS_MBUF_GET(ti); /* XXX */
208	ti = u32_to_ptr(pData, ti->ti_next, struct tcpiphdr *);
209	/* if (so->so_state & SS_FCANTRCVMORE) */
210	if (so->so_state & SS_FCANTSENDMORE)
211	m_freem(pData, m);
212	else {
213	if (so->so_emu) {
214	if (tcp_emu(pData, so,m)) sbappend(pData, so, m);
215	} else
216	sbappend(pData, so, m);
217	}
218	} while (ti != (struct tcpiphdr *)tp && ti->ti_seq == tp->rcv_nxt);
219	/* sorwakeup(so); */
220	return (flags);
221	}
222
223	#else /* VBOX_WITH_BSD_TCP_REASS */
224
225	#ifndef TCP_ACK_HACK
226	#define DELAY_ACK(tp, ti) \
227	if (ti->ti_flags & TH_PUSH) \
228	tp->t_flags \|= TF_ACKNOW; \
229	else \
230	tp->t_flags \|= TF_DELACK;
231	#else /* !TCP_ACK_HACK */
232	#define DELAY_ACK(tp, ign) \
233	tp->t_flags \|= TF_DELACK;
234	#endif /* TCP_ACK_HACK */
235
236
237	/*
238	* deps: netinet/tcp_reass.c
239	* tcp_reass_maxqlen = 48 (deafault)
240	* tcp_reass_maxseg = nmbclusters/16 (nmbclusters = 1024 + maxusers * 64 from kern/kern_mbuf.c let's say 256)
241	*/
242	int
243	tcp_reass(PNATState pData, struct tcpcb tp, struct tcphdr th, int tlenp, struct mbuf m)
244	{
245	struct tseg_qent *q;
246	struct tseg_qent *p = NULL;
247	struct tseg_qent *nq;
248	struct tseg_qent *te = NULL;
249	struct socket *so = tp->t_socket;
250	int flags;
251
252	/*
253	* XXX: tcp_reass() is rather inefficient with its data structures
254	* and should be rewritten (see NetBSD for optimizations). While
255	* doing that it should move to its own file tcp_reass.c.
256	*/
257
258	/*
259	* Call with th==NULL after become established to
260	* force pre-ESTABLISHED data up to user socket.
261	*/
262	if (th == NULL)
263	goto present;
264
265	/*
266	* Limit the number of segments in the reassembly queue to prevent
267	* holding on to too many segments (and thus running out of mbufs).
268	* Make sure to let the missing segment through which caused this
269	* queue. Always keep one global queue entry spare to be able to
270	* process the missing segment.
271	*/
272	if (th->th_seq != tp->rcv_nxt &&
273	(tcp_reass_qsize + 1 >= tcp_reass_maxseg \|\|
274	tp->t_segqlen >= tcp_reass_maxqlen)) {
275	tcp_reass_overflows++;
276	tcpstat.tcps_rcvmemdrop++;
277	m_freem(pData, m);
278	*tlenp = 0;
279	return (0);
280	}
281
282	/*
283	* Allocate a new queue entry. If we can't, or hit the zone limit
284	* just drop the pkt.
285	*/
286	te = malloc(sizeof(struct tseg_qent));
287	if (te == NULL) {
288	tcpstat.tcps_rcvmemdrop++;
289	m_freem(pData, m);
290	*tlenp = 0;
291	return (0);
292	}
293	tp->t_segqlen++;
294	tcp_reass_qsize++;
295
296	/*
297	* Find a segment which begins after this one does.
298	*/
299	LIST_FOREACH(q, &tp->t_segq, tqe_q) {
300	if (SEQ_GT(q->tqe_th->th_seq, th->th_seq))
301	break;
302	p = q;
303	}
304
305	/*
306	* If there is a preceding segment, it may provide some of
307	* our data already. If so, drop the data from the incoming
308	* segment. If it provides all of our data, drop us.
309	*/
310	if (p != NULL) {
311	int i;
312	/* conversion to int (in i) handles seq wraparound */
313	i = p->tqe_th->th_seq + p->tqe_len - th->th_seq;
314	if (i > 0) {
315	if (i >= *tlenp) {
316	tcpstat.tcps_rcvduppack++;
317	tcpstat.tcps_rcvdupbyte += *tlenp;
318	m_freem(pData, m);
319	free(te);
320	tp->t_segqlen--;
321	tcp_reass_qsize--;
322	/*
323	* Try to present any queued data
324	* at the left window edge to the user.
325	* This is needed after the 3-WHS
326	* completes.
327	*/
328	goto present; /* ??? */
329	}
330	m_adj(m, i);
331	*tlenp -= i;
332	th->th_seq += i;
333	}
334	}
335	tcpstat.tcps_rcvoopack++;
336	tcpstat.tcps_rcvoobyte += *tlenp;
337
338	/*
339	* While we overlap succeeding segments trim them or,
340	* if they are completely covered, dequeue them.
341	*/
342	while (q) {
343	int i = (th->th_seq + *tlenp) - q->tqe_th->th_seq;
344	if (i <= 0)
345	break;
346	if (i < q->tqe_len) {
347	q->tqe_th->th_seq += i;
348	q->tqe_len -= i;
349	m_adj(q->tqe_m, i);
350	break;
351	}
352
353	nq = LIST_NEXT(q, tqe_q);
354	LIST_REMOVE(q, tqe_q);
355	m_freem(pData, q->tqe_m);
356	free(q);
357	tp->t_segqlen--;
358	tcp_reass_qsize--;
359	q = nq;
360	}
361
362	/* Insert the new segment queue entry into place. */
363	te->tqe_m = m;
364	te->tqe_th = th;
365	te->tqe_len = *tlenp;
366
367	if (p == NULL) {
368	LIST_INSERT_HEAD(&tp->t_segq, te, tqe_q);
369	} else {
370	LIST_INSERT_AFTER(p, te, tqe_q);
371	}
372
373	present:
374	/*
375	* Present data to user, advancing rcv_nxt through
376	* completed sequence space.
377	*/
378	if (!TCPS_HAVEESTABLISHED(tp->t_state))
379	return (0);
380	q = LIST_FIRST(&tp->t_segq);
381	if (!q \|\| q->tqe_th->th_seq != tp->rcv_nxt)
382	return (0);
383	do {
384	tp->rcv_nxt += q->tqe_len;
385	flags = q->tqe_th->th_flags & TH_FIN;
386	nq = LIST_NEXT(q, tqe_q);
387	LIST_REMOVE(q, tqe_q);
388	/* XXX: This place should be checked for the same code in
389	* original BSD code for Slirp and current BSD used SS_FCANTRCVMORE
390	*/
391	if (so->so_state & SS_FCANTSENDMORE)
392	m_freem(pData, q->tqe_m);
393	else
394	sbappend(pData, so, q->tqe_m);
395	free(q);
396	tp->t_segqlen--;
397	tcp_reass_qsize--;
398	q = nq;
399	} while (q && q->tqe_th->th_seq == tp->rcv_nxt);
400	return (flags);
401	}
402	#endif /* VBOX_WITH_BSD_TCP_REASS */
403
404	/*
405	* TCP input routine, follows pages 65-76 of the
406	* protocol specification dated September, 1981 very closely.
407	*/
408	void
409	tcp_input(PNATState pData, register struct mbuf m, int iphlen, struct socket inso)
410	{
411	struct ip save_ip, *ip;
412	register struct tcpiphdr *ti;
413	caddr_t optp = NULL;
414	int optlen = 0;
415	int len, tlen, off;
416	register struct tcpcb *tp = 0;
417	register int tiflags;
418	struct socket *so = 0;
419	int todrop, acked, ourfinisacked, needoutput = 0;
420	/* int dropsocket = 0; */
421	int iss = 0;
422	u_long tiwin;
423	int ret;
424	/* int ts_present = 0; */
425	int mbuf_freed = 0;
426
427	DEBUG_CALL("tcp_input");
428	DEBUG_ARGS((dfd," m = %8lx iphlen = %2d inso = %lx\n",
429	(long )m, iphlen, (long )inso ));
430
431	/*
432	* If called with m == 0, then we're continuing the connect
433	*/
434	if (m == NULL) {
435	so = inso;
436
437	/* Re-set a few variables */
438	tp = sototcpcb(so);
439	m = so->so_m;
440	so->so_m = 0;
441	ti = so->so_ti;
442	tiwin = ti->ti_win;
443	tiflags = ti->ti_flags;
444
445	goto cont_conn;
446	}
447
448
449	tcpstat.tcps_rcvtotal++;
450	/*
451	* Get IP and TCP header together in first mbuf.
452	* Note: IP leaves IP header in first mbuf.
453	*/
454	ti = mtod(m, struct tcpiphdr *);
455	if (iphlen > sizeof(struct ip )) {
456	ip_stripoptions(m, (struct mbuf *)0);
457	iphlen=sizeof(struct ip );
458	}
459	/* XXX Check if too short */
460
461
462	/*
463	* Save a copy of the IP header in case we want restore it
464	* for sending an ICMP error message in response.
465	*/
466	ip=mtod(m, struct ip *);
467	save_ip = *ip;
468	save_ip.ip_len+= iphlen;
469
470	/*
471	* Checksum extended TCP header and data.
472	*/
473	tlen = ((struct ip *)ti)->ip_len;
474	ti->ti_next = ti->ti_prev = 0;
475	ti->ti_x1 = 0;
476	ti->ti_len = htons((u_int16_t)tlen);
477	len = sizeof(struct ip ) + tlen;
478	/* keep checksum for ICMP reply
479	* ti->ti_sum = cksum(m, len);
480	* if (ti->ti_sum) { */
481	#ifdef VBOX_WITH_BSD_REASS_CKSUM_HACK
482	if (m->m_sum_recalculate == 0 && ip->ip_sum != 0) {
483	#endif /* VBOX_WITH_BSD_REASS_CKSUM_HACK */
484	if(cksum(m, len)) {
485	tcpstat.tcps_rcvbadsum++;
486	goto drop;
487	}
488	#ifdef VBOX_WITH_BSD_REASS_CKSUM_HACK
489	}
490	#endif /* VBOX_WITH_BSD_REASS_CKSUM_HACK */
491
492	/*
493	* Check that TCP offset makes sense,
494	* pull out TCP options and adjust length. XXX
495	*/
496	off = ti->ti_off << 2;
497	if (off < sizeof (struct tcphdr) \|\| off > tlen) {
498	tcpstat.tcps_rcvbadoff++;
499	goto drop;
500	}
501	tlen -= off;
502	ti->ti_len = tlen;
503	if (off > sizeof (struct tcphdr)) {
504	optlen = off - sizeof (struct tcphdr);
505	optp = mtod(m, caddr_t) + sizeof (struct tcpiphdr);
506
507	/*
508	* Do quick retrieval of timestamp options ("options
509	* prediction?"). If timestamp is the only option and it's
510	* formatted as recommended in RFC 1323 appendix A, we
511	* quickly get the values now and not bother calling
512	* tcp_dooptions(), etc.
513	*/
514	/* if ((optlen == TCPOLEN_TSTAMP_APPA \|\|
515	* (optlen > TCPOLEN_TSTAMP_APPA &&
516	* optp[TCPOLEN_TSTAMP_APPA] == TCPOPT_EOL)) &&
517	* (u_int32_t )optp == htonl(TCPOPT_TSTAMP_HDR) &&
518	* (ti->ti_flags & TH_SYN) == 0) {
519	* ts_present = 1;
520	* ts_val = ntohl((u_int32_t )(optp + 4));
521	* ts_ecr = ntohl((u_int32_t )(optp + 8));
522	* optp = NULL; / * we've parsed the options * /
523	* }
524	*/
525	}
526	tiflags = ti->ti_flags;
527
528	/*
529	* Convert TCP protocol specific fields to host format.
530	*/
531	NTOHL(ti->ti_seq);
532	NTOHL(ti->ti_ack);
533	NTOHS(ti->ti_win);
534	NTOHS(ti->ti_urp);
535
536	/*
537	* Drop TCP, IP headers and TCP options.
538	*/
539	m->m_data += sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
540	m->m_len -= sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
541
542	/*
543	* Locate pcb for segment.
544	*/
545	findso:
546	so = tcp_last_so;
547	if (so->so_fport != ti->ti_dport \|\|
548	so->so_lport != ti->ti_sport \|\|
549	so->so_laddr.s_addr != ti->ti_src.s_addr \|\|
550	so->so_faddr.s_addr != ti->ti_dst.s_addr) {
551	so = solookup(&tcb, ti->ti_src, ti->ti_sport,
552	ti->ti_dst, ti->ti_dport);
553	if (so)
554	tcp_last_so = so;
555	++tcpstat.tcps_socachemiss;
556	}
557
558	/*
559	* If the state is CLOSED (i.e., TCB does not exist) then
560	* all data in the incoming segment is discarded.
561	* If the TCB exists but is in CLOSED state, it is embryonic,
562	* but should either do a listen or a connect soon.
563	*
564	* state == CLOSED means we've done socreate() but haven't
565	* attached it to a protocol yet...
566	*
567	* XXX If a TCB does not exist, and the TH_SYN flag is
568	* the only flag set, then create a session, mark it
569	* as if it was LISTENING, and continue...
570	*/
571	if (so == 0) {
572	if ((tiflags & (TH_SYN\|TH_FIN\|TH_RST\|TH_URG\|TH_ACK)) != TH_SYN)
573	goto dropwithreset;
574
575	if ((so = socreate()) == NULL)
576	goto dropwithreset;
577	if (tcp_attach(pData, so) < 0) {
578	free(so); /* Not sofree (if it failed, it's not insqued) */
579	goto dropwithreset;
580	}
581
582	sbreserve(&so->so_snd, tcp_sndspace);
583	sbreserve(&so->so_rcv, tcp_rcvspace);
584
585	/* tcp_last_so = so; / / XXX ? */
586	/* tp = sototcpcb(so); */
587
588	so->so_laddr = ti->ti_src;
589	so->so_lport = ti->ti_sport;
590	so->so_faddr = ti->ti_dst;
591	so->so_fport = ti->ti_dport;
592
593	if ((so->so_iptos = tcp_tos(so)) == 0)
594	so->so_iptos = ((struct ip *)ti)->ip_tos;
595
596	tp = sototcpcb(so);
597	tp->t_state = TCPS_LISTEN;
598	}
599
600	/*
601	* If this is a still-connecting socket, this probably
602	* a retransmit of the SYN. Whether it's a retransmit SYN
603	* or something else, we nuke it.
604	*/
605	if (so->so_state & SS_ISFCONNECTING)
606	goto drop;
607
608	tp = sototcpcb(so);
609
610	/* XXX Should never fail */
611	if (tp == 0)
612	goto dropwithreset;
613	if (tp->t_state == TCPS_CLOSED)
614	goto drop;
615
616	/* Unscale the window into a 32-bit value. */
617	/* if ((tiflags & TH_SYN) == 0)
618	* tiwin = ti->ti_win << tp->snd_scale;
619	* else
620	*/
621	tiwin = ti->ti_win;
622
623	/*
624	* Segment received on connection.
625	* Reset idle time and keep-alive timer.
626	*/
627	tp->t_idle = 0;
628	if (so_options)
629	tp->t_timer[TCPT_KEEP] = tcp_keepintvl;
630	else
631	tp->t_timer[TCPT_KEEP] = tcp_keepidle;
632
633	/*
634	* Process options if not in LISTEN state,
635	* else do it below (after getting remote address).
636	*/
637	if (optp && tp->t_state != TCPS_LISTEN)
638	tcp_dooptions(pData, tp, (u_char *)optp, optlen, ti);
639	/* , */
640	/* &ts_present, &ts_val, &ts_ecr); */
641
642	/*
643	* Header prediction: check for the two common cases
644	* of a uni-directional data xfer. If the packet has
645	* no control flags, is in-sequence, the window didn't
646	* change and we're not retransmitting, it's a
647	* candidate. If the length is zero and the ack moved
648	* forward, we're the sender side of the xfer. Just
649	* free the data acked & wake any higher level process
650	* that was blocked waiting for space. If the length
651	* is non-zero and the ack didn't move, we're the
652	* receiver side. If we're getting packets in-order
653	* (the reassembly queue is empty), add the data to
654	* the socket buffer and note that we need a delayed ack.
655	*
656	* XXX Some of these tests are not needed
657	* eg: the tiwin == tp->snd_wnd prevents many more
658	* predictions.. with no real advantage..
659	*/
660	if (tp->t_state == TCPS_ESTABLISHED &&
661	(tiflags & (TH_SYN\|TH_FIN\|TH_RST\|TH_URG\|TH_ACK)) == TH_ACK &&
662	/* (!ts_present \|\| TSTMP_GEQ(ts_val, tp->ts_recent)) && */
663	ti->ti_seq == tp->rcv_nxt &&
664	tiwin && tiwin == tp->snd_wnd &&
665	tp->snd_nxt == tp->snd_max) {
666	/*
667	* If last ACK falls within this segment's sequence numbers,
668	* record the timestamp.
669	*/
670	/* if (ts_present && SEQ_LEQ(ti->ti_seq, tp->last_ack_sent) &&
671	* SEQ_LT(tp->last_ack_sent, ti->ti_seq + ti->ti_len)) {
672	* tp->ts_recent_age = tcp_now;
673	* tp->ts_recent = ts_val;
674	* }
675	*/
676	if (ti->ti_len == 0) {
677	if (SEQ_GT(ti->ti_ack, tp->snd_una) &&
678	SEQ_LEQ(ti->ti_ack, tp->snd_max) &&
679	tp->snd_cwnd >= tp->snd_wnd) {
680	/*
681	* this is a pure ack for outstanding data.
682	*/
683	++tcpstat.tcps_predack;
684	/* if (ts_present)
685	* tcp_xmit_timer(tp, tcp_now-ts_ecr+1);
686	* else
687	*/ if (tp->t_rtt &&
688	SEQ_GT(ti->ti_ack, tp->t_rtseq))
689	tcp_xmit_timer(pData, tp, tp->t_rtt);
690	acked = ti->ti_ack - tp->snd_una;
691	tcpstat.tcps_rcvackpack++;
692	tcpstat.tcps_rcvackbyte += acked;
693	sbdrop(&so->so_snd, acked);
694	tp->snd_una = ti->ti_ack;
695	m_freem(pData, m);
696
697	/*
698	* If all outstanding data are acked, stop
699	* retransmit timer, otherwise restart timer
700	* using current (possibly backed-off) value.
701	* If process is waiting for space,
702	* wakeup/selwakeup/signal. If data
703	* are ready to send, let tcp_output
704	* decide between more output or persist.
705	*/
706	if (tp->snd_una == tp->snd_max)
707	tp->t_timer[TCPT_REXMT] = 0;
708	else if (tp->t_timer[TCPT_PERSIST] == 0)
709	tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
710
711	/*
712	* There's room in so_snd, sowwakup will read()
713	* from the socket if we can
714	*/
715	/* if (so->so_snd.sb_flags & SB_NOTIFY)
716	* sowwakeup(so);
717	*/
718	/*
719	* This is called because sowwakeup might have
720	* put data into so_snd. Since we don't so sowwakeup,
721	* we don't need this.. XXX???
722	*/
723	if (so->so_snd.sb_cc)
724	(void) tcp_output(pData, tp);
725
726	return;
727	}
728	} else if (ti->ti_ack == tp->snd_una &&
729	#ifndef VBOX_WITH_BSD_TCP_REASS
730	u32_to_ptr(pData, tp->seg_next, struct tcpcb *) == tp &&
731	#else /* VBOX_WITH_BSD_TCP_REASS */
732	LIST_FIRST(&tp->t_segq) &&
733	#endif /* VBOX_WITH_BSD_TCP_REASS */
734	ti->ti_len <= sbspace(&so->so_rcv)) {
735	/*
736	* this is a pure, in-sequence data packet
737	* with nothing on the reassembly queue and
738	* we have enough buffer space to take it.
739	*/
740	++tcpstat.tcps_preddat;
741	tp->rcv_nxt += ti->ti_len;
742	tcpstat.tcps_rcvpack++;
743	tcpstat.tcps_rcvbyte += ti->ti_len;
744	/*
745	* Add data to socket buffer.
746	*/
747	if (so->so_emu) {
748	if (tcp_emu(pData, so,m)) sbappend(pData, so, m);
749	} else
750	sbappend(pData, so, m);
751
752	/*
753	* XXX This is called when data arrives. Later, check
754	* if we can actually write() to the socket
755	* XXX Need to check? It's be NON_BLOCKING
756	*/
757	/* sorwakeup(so); */
758
759	/*
760	* If this is a short packet, then ACK now - with Nagel
761	* congestion avoidance sender won't send more until
762	* he gets an ACK.
763	*
764	* It is better to not delay acks at all to maximize
765	* TCP throughput. See RFC 2581.
766	*/
767	tp->t_flags \|= TF_ACKNOW;
768	tcp_output(pData, tp);
769	return;
770	}
771	} /* header prediction */
772	/*
773	* Calculate amount of space in receive window,
774	* and then do TCP input processing.
775	* Receive window is amount of space in rcv queue,
776	* but not less than advertised window.
777	*/
778	{ int win;
779	win = sbspace(&so->so_rcv);
780	if (win < 0)
781	win = 0;
782	tp->rcv_wnd = max(win, (int)(tp->rcv_adv - tp->rcv_nxt));
783	}
784
785	switch (tp->t_state) {
786
787	/*
788	* If the state is LISTEN then ignore segment if it contains an RST.
789	* If the segment contains an ACK then it is bad and send a RST.
790	* If it does not contain a SYN then it is not interesting; drop it.
791	* Don't bother responding if the destination was a broadcast.
792	* Otherwise initialize tp->rcv_nxt, and tp->irs, select an initial
793	* tp->iss, and send a segment:
794	* <SEQ=ISS><ACK=RCV_NXT><CTL=SYN,ACK>
795	* Also initialize tp->snd_nxt to tp->iss+1 and tp->snd_una to tp->iss.
796	* Fill in remote peer address fields if not previously specified.
797	* Enter SYN_RECEIVED state, and process any other fields of this
798	* segment in this state.
799	*/
800	case TCPS_LISTEN: {
801
802	if (tiflags & TH_RST)
803	goto drop;
804	if (tiflags & TH_ACK)
805	goto dropwithreset;
806	if ((tiflags & TH_SYN) == 0)
807	goto drop;
808
809	/*
810	* This has way too many gotos...
811	* But a bit of spaghetti code never hurt anybody :)
812	*/
813
814	/*
815	* If this is destined for the control address, then flag to
816	* tcp_ctl once connected, otherwise connect
817	*/
818	if ((so->so_faddr.s_addr&htonl(pData->netmask)) == special_addr.s_addr) {
819	int lastbyte=ntohl(so->so_faddr.s_addr) & ~pData->netmask;
820	if (lastbyte!=CTL_ALIAS && lastbyte!=CTL_DNS) {
821	#if 0
822	if(lastbyte==CTL_CMD \|\| lastbyte==CTL_EXEC) {
823	/* Command or exec adress */
824	so->so_state \|= SS_CTL;
825	} else
826	#endif
827	{
828	/* May be an add exec */
829	struct ex_list *ex_ptr;
830	for(ex_ptr = exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next) {
831	if(ex_ptr->ex_fport == so->so_fport &&
832	lastbyte == ex_ptr->ex_addr) {
833	so->so_state \|= SS_CTL;
834	break;
835	}
836	}
837	}
838	if(so->so_state & SS_CTL) goto cont_input;
839	}
840	/* CTL_ALIAS: Do nothing, tcp_fconnect will be called on it */
841	}
842
843	if (so->so_emu & EMU_NOCONNECT) {
844	so->so_emu &= ~EMU_NOCONNECT;
845	goto cont_input;
846	}
847
848	if((tcp_fconnect(pData, so) == -1) && (errno != EINPROGRESS) && (errno != EWOULDBLOCK)) {
849	u_char code=ICMP_UNREACH_NET;
850	DEBUG_MISC((dfd," tcp fconnect errno = %d-%s\n",
851	errno,strerror(errno)));
852	if(errno == ECONNREFUSED) {
853	/* ACK the SYN, send RST to refuse the connection */
854	tcp_respond(pData, tp, ti, m, ti->ti_seq+1, (tcp_seq)0,
855	TH_RST\|TH_ACK);
856	} else {
857	if(errno == EHOSTUNREACH) code=ICMP_UNREACH_HOST;
858	HTONL(ti->ti_seq); /* restore tcp header */
859	HTONL(ti->ti_ack);
860	HTONS(ti->ti_win);
861	HTONS(ti->ti_urp);
862	m->m_data -= sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
863	m->m_len += sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
864	*ip=save_ip;
865	icmp_error(pData, m, ICMP_UNREACH,code, 0,strerror(errno));
866	}
867	tp = tcp_close(pData, tp);
868	m_free(pData, m);
869	} else {
870	/*
871	* Haven't connected yet, save the current mbuf
872	* and ti, and return
873	* XXX Some OS's don't tell us whether the connect()
874	* succeeded or not. So we must time it out.
875	*/
876	so->so_m = m;
877	so->so_ti = ti;
878	tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT;
879	tp->t_state = TCPS_SYN_RECEIVED;
880	}
881	return;
882
883	cont_conn:
884	/* m==NULL
885	* Check if the connect succeeded
886	*/
887	if (so->so_state & SS_NOFDREF) {
888	tp = tcp_close(pData, tp);
889	goto dropwithreset;
890	}
891	cont_input:
892	tcp_template(tp);
893
894	if (optp)
895	tcp_dooptions(pData, tp, (u_char *)optp, optlen, ti);
896	/* , */
897	/* &ts_present, &ts_val, &ts_ecr); */
898
899	if (iss)
900	tp->iss = iss;
901	else
902	tp->iss = tcp_iss;
903	tcp_iss += TCP_ISSINCR/2;
904	tp->irs = ti->ti_seq;
905	tcp_sendseqinit(tp);
906	tcp_rcvseqinit(tp);
907	tp->t_flags \|= TF_ACKNOW;
908	tp->t_state = TCPS_SYN_RECEIVED;
909	tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT;
910	tcpstat.tcps_accepts++;
911	goto trimthenstep6;
912	} /* case TCPS_LISTEN */
913
914	/*
915	* If the state is SYN_SENT:
916	* if seg contains an ACK, but not for our SYN, drop the input.
917	* if seg contains a RST, then drop the connection.
918	* if seg does not contain SYN, then drop it.
919	* Otherwise this is an acceptable SYN segment
920	* initialize tp->rcv_nxt and tp->irs
921	* if seg contains ack then advance tp->snd_una
922	* if SYN has been acked change to ESTABLISHED else SYN_RCVD state
923	* arrange for segment to be acked (eventually)
924	* continue processing rest of data/controls, beginning with URG
925	*/
926	case TCPS_SYN_SENT:
927	if ((tiflags & TH_ACK) &&
928	(SEQ_LEQ(ti->ti_ack, tp->iss) \|\|
929	SEQ_GT(ti->ti_ack, tp->snd_max)))
930	goto dropwithreset;
931
932	if (tiflags & TH_RST) {
933	if (tiflags & TH_ACK)
934	tp = tcp_drop(pData, tp,0); /* XXX Check t_softerror! */
935	goto drop;
936	}
937
938	if ((tiflags & TH_SYN) == 0)
939	goto drop;
940	if (tiflags & TH_ACK) {
941	tp->snd_una = ti->ti_ack;
942	if (SEQ_LT(tp->snd_nxt, tp->snd_una))
943	tp->snd_nxt = tp->snd_una;
944	}
945
946	tp->t_timer[TCPT_REXMT] = 0;
947	tp->irs = ti->ti_seq;
948	tcp_rcvseqinit(tp);
949	tp->t_flags \|= TF_ACKNOW;
950	if (tiflags & TH_ACK && SEQ_GT(tp->snd_una, tp->iss)) {
951	tcpstat.tcps_connects++;
952	soisfconnected(so);
953	tp->t_state = TCPS_ESTABLISHED;
954
955	/* Do window scaling on this connection? */
956	/* if ((tp->t_flags & (TF_RCVD_SCALE\|TF_REQ_SCALE)) ==
957	* (TF_RCVD_SCALE\|TF_REQ_SCALE)) {
958	* tp->snd_scale = tp->requested_s_scale;
959	* tp->rcv_scale = tp->request_r_scale;
960	* }
961	*/
962	#ifndef VBOX_WITH_BSD_TCP_REASS
963	(void) tcp_reass(pData, tp, (struct tcpiphdr *)0,
964	(struct mbuf *)0);
965	#else /* VBOX_WITH_BSD_TCP_REASS */
966	(void) tcp_reass(pData, tp, (struct tcphdr )0, NULL, (struct mbuf )0);
967	#endif /* VBOX_WITH_BSD_TCP_REASS */
968	/*
969	* if we didn't have to retransmit the SYN,
970	* use its rtt as our initial srtt & rtt var.
971	*/
972	if (tp->t_rtt)
973	tcp_xmit_timer(pData, tp, tp->t_rtt);
974	} else
975	tp->t_state = TCPS_SYN_RECEIVED;
976
977	trimthenstep6:
978	/*
979	* Advance ti->ti_seq to correspond to first data byte.
980	* If data, trim to stay within window,
981	* dropping FIN if necessary.
982	*/
983	ti->ti_seq++;
984	if (ti->ti_len > tp->rcv_wnd) {
985	todrop = ti->ti_len - tp->rcv_wnd;
986	m_adj(m, -todrop);
987	ti->ti_len = tp->rcv_wnd;
988	tiflags &= ~TH_FIN;
989	tcpstat.tcps_rcvpackafterwin++;
990	tcpstat.tcps_rcvbyteafterwin += todrop;
991	}
992	tp->snd_wl1 = ti->ti_seq - 1;
993	tp->rcv_up = ti->ti_seq;
994	goto step6;
995	} /* switch tp->t_state */
996	/*
997	* States other than LISTEN or SYN_SENT.
998	* First check timestamp, if present.
999	* Then check that at least some bytes of segment are within
1000	* receive window. If segment begins before rcv_nxt,
1001	* drop leading data (and SYN); if nothing left, just ack.
1002	*
1003	* RFC 1323 PAWS: If we have a timestamp reply on this segment
1004	* and it's less than ts_recent, drop it.
1005	*/
1006	/* if (ts_present && (tiflags & TH_RST) == 0 && tp->ts_recent &&
1007	* TSTMP_LT(ts_val, tp->ts_recent)) {
1008	*
1009	/ / Check to see if ts_recent is over 24 days old. */
1010	/* if ((int)(tcp_now - tp->ts_recent_age) > TCP_PAWS_IDLE) {
1011	/ /
1012	* * Invalidate ts_recent. If this segment updates
1013	* * ts_recent, the age will be reset later and ts_recent
1014	* * will get a valid value. If it does not, setting
1015	* * ts_recent to zero will at least satisfy the
1016	* * requirement that zero be placed in the timestamp
1017	* * echo reply when ts_recent isn't valid. The
1018	* * age isn't reset until we get a valid ts_recent
1019	* * because we don't want out-of-order segments to be
1020	* * dropped when ts_recent is old.
1021	* */
1022	/* tp->ts_recent = 0;
1023	* } else {
1024	* tcpstat.tcps_rcvduppack++;
1025	* tcpstat.tcps_rcvdupbyte += ti->ti_len;
1026	* tcpstat.tcps_pawsdrop++;
1027	* goto dropafterack;
1028	* }
1029	* }
1030	*/
1031
1032	todrop = tp->rcv_nxt - ti->ti_seq;
1033	if (todrop > 0) {
1034	if (tiflags & TH_SYN) {
1035	tiflags &= ~TH_SYN;
1036	ti->ti_seq++;
1037	if (ti->ti_urp > 1)
1038	ti->ti_urp--;
1039	else
1040	tiflags &= ~TH_URG;
1041	todrop--;
1042	}
1043	/*
1044	* Following if statement from Stevens, vol. 2, p. 960.
1045	*/
1046	if (todrop > ti->ti_len
1047	\|\| (todrop == ti->ti_len && (tiflags & TH_FIN) == 0)) {
1048	/*
1049	* Any valid FIN must be to the left of the window.
1050	* At this point the FIN must be a duplicate or out
1051	* of sequence; drop it.
1052	*/
1053	tiflags &= ~TH_FIN;
1054
1055	/*
1056	* Send an ACK to resynchronize and drop any data.
1057	* But keep on processing for RST or ACK.
1058	*/
1059	tp->t_flags \|= TF_ACKNOW;
1060	todrop = ti->ti_len;
1061	tcpstat.tcps_rcvduppack++;
1062	tcpstat.tcps_rcvdupbyte += todrop;
1063	} else {
1064	tcpstat.tcps_rcvpartduppack++;
1065	tcpstat.tcps_rcvpartdupbyte += todrop;
1066	}
1067	m_adj(m, todrop);
1068	ti->ti_seq += todrop;
1069	ti->ti_len -= todrop;
1070	if (ti->ti_urp > todrop)
1071	ti->ti_urp -= todrop;
1072	else {
1073	tiflags &= ~TH_URG;
1074	ti->ti_urp = 0;
1075	}
1076	}
1077	/*
1078	* If new data are received on a connection after the
1079	* user processes are gone, then RST the other end.
1080	*/
1081	if ((so->so_state & SS_NOFDREF) &&
1082	tp->t_state > TCPS_CLOSE_WAIT && ti->ti_len) {
1083	tp = tcp_close(pData, tp);
1084	tcpstat.tcps_rcvafterclose++;
1085	goto dropwithreset;
1086	}
1087
1088	/*
1089	* If segment ends after window, drop trailing data
1090	* (and PUSH and FIN); if nothing left, just ACK.
1091	*/
1092	todrop = (ti->ti_seq+ti->ti_len) - (tp->rcv_nxt+tp->rcv_wnd);
1093	if (todrop > 0) {
1094	tcpstat.tcps_rcvpackafterwin++;
1095	if (todrop >= ti->ti_len) {
1096	tcpstat.tcps_rcvbyteafterwin += ti->ti_len;
1097	/*
1098	* If a new connection request is received
1099	* while in TIME_WAIT, drop the old connection
1100	* and start over if the sequence numbers
1101	* are above the previous ones.
1102	*/
1103	if (tiflags & TH_SYN &&
1104	tp->t_state == TCPS_TIME_WAIT &&
1105	SEQ_GT(ti->ti_seq, tp->rcv_nxt)) {
1106	iss = tp->rcv_nxt + TCP_ISSINCR;
1107	tp = tcp_close(pData, tp);
1108	goto findso;
1109	}
1110	/*
1111	* If window is closed can only take segments at
1112	* window edge, and have to drop data and PUSH from
1113	* incoming segments. Continue processing, but
1114	* remember to ack. Otherwise, drop segment
1115	* and ack.
1116	*/
1117	if (tp->rcv_wnd == 0 && ti->ti_seq == tp->rcv_nxt) {
1118	tp->t_flags \|= TF_ACKNOW;
1119	tcpstat.tcps_rcvwinprobe++;
1120	} else
1121	goto dropafterack;
1122	} else
1123	tcpstat.tcps_rcvbyteafterwin += todrop;
1124	m_adj(m, -todrop);
1125	ti->ti_len -= todrop;
1126	tiflags &= ~(TH_PUSH\|TH_FIN);
1127	}
1128
1129	/*
1130	* If last ACK falls within this segment's sequence numbers,
1131	* record its timestamp.
1132	*/
1133	/* if (ts_present && SEQ_LEQ(ti->ti_seq, tp->last_ack_sent) &&
1134	* SEQ_LT(tp->last_ack_sent, ti->ti_seq + ti->ti_len +
1135	* ((tiflags & (TH_SYN\|TH_FIN)) != 0))) {
1136	* tp->ts_recent_age = tcp_now;
1137	* tp->ts_recent = ts_val;
1138	* }
1139	*/
1140
1141	/*
1142	* If the RST bit is set examine the state:
1143	* SYN_RECEIVED STATE:
1144	* If passive open, return to LISTEN state.
1145	* If active open, inform user that connection was refused.
1146	* ESTABLISHED, FIN_WAIT_1, FIN_WAIT2, CLOSE_WAIT STATES:
1147	* Inform user that connection was reset, and close tcb.
1148	* CLOSING, LAST_ACK, TIME_WAIT STATES
1149	* Close the tcb.
1150	*/
1151	if (tiflags&TH_RST) switch (tp->t_state) {
1152
1153	case TCPS_SYN_RECEIVED:
1154	/* so->so_error = ECONNREFUSED; */
1155	goto close;
1156
1157	case TCPS_ESTABLISHED:
1158	case TCPS_FIN_WAIT_1:
1159	case TCPS_FIN_WAIT_2:
1160	case TCPS_CLOSE_WAIT:
1161	/* so->so_error = ECONNRESET; */
1162	close:
1163	tp->t_state = TCPS_CLOSED;
1164	tcpstat.tcps_drops++;
1165	tp = tcp_close(pData, tp);
1166	goto drop;
1167
1168	case TCPS_CLOSING:
1169	case TCPS_LAST_ACK:
1170	case TCPS_TIME_WAIT:
1171	tp = tcp_close(pData, tp);
1172	goto drop;
1173	}
1174
1175	/*
1176	* If a SYN is in the window, then this is an
1177	* error and we send an RST and drop the connection.
1178	*/
1179	if (tiflags & TH_SYN) {
1180	tp = tcp_drop(pData, tp,0);
1181	goto dropwithreset;
1182	}
1183
1184	/*
1185	* If the ACK bit is off we drop the segment and return.
1186	*/
1187	if ((tiflags & TH_ACK) == 0) goto drop;
1188
1189	/*
1190	* Ack processing.
1191	*/
1192	switch (tp->t_state) {
1193	/*
1194	* In SYN_RECEIVED state if the ack ACKs our SYN then enter
1195	* ESTABLISHED state and continue processing, otherwise
1196	* send an RST. una<=ack<=max
1197	*/
1198	case TCPS_SYN_RECEIVED:
1199
1200	if (SEQ_GT(tp->snd_una, ti->ti_ack) \|\|
1201	SEQ_GT(ti->ti_ack, tp->snd_max))
1202	goto dropwithreset;
1203	tcpstat.tcps_connects++;
1204	tp->t_state = TCPS_ESTABLISHED;
1205	/*
1206	* The sent SYN is ack'ed with our sequence number +1
1207	* The first data byte already in the buffer will get
1208	* lost if no correction is made. This is only needed for
1209	* SS_CTL since the buffer is empty otherwise.
1210	* tp->snd_una++; or:
1211	*/
1212	tp->snd_una=ti->ti_ack;
1213	if (so->so_state & SS_CTL) {
1214	/* So tcp_ctl reports the right state */
1215	ret = tcp_ctl(pData, so);
1216	if (ret == 1) {
1217	soisfconnected(so);
1218	so->so_state &= ~SS_CTL; /* success XXX */
1219	} else if (ret == 2) {
1220	so->so_state = SS_NOFDREF; /* CTL_CMD */
1221	} else {
1222	needoutput = 1;
1223	tp->t_state = TCPS_FIN_WAIT_1;
1224	}
1225	} else {
1226	soisfconnected(so);
1227	}
1228
1229	/* Do window scaling? */
1230	/* if ((tp->t_flags & (TF_RCVD_SCALE\|TF_REQ_SCALE)) ==
1231	* (TF_RCVD_SCALE\|TF_REQ_SCALE)) {
1232	* tp->snd_scale = tp->requested_s_scale;
1233	* tp->rcv_scale = tp->request_r_scale;
1234	* }
1235	*/
1236	#ifndef VBOX_WITH_BSD_TCP_REASS
1237	(void) tcp_reass(pData, tp, (struct tcpiphdr )0, (struct mbuf )0);
1238	#else /* VBOX_WITH_BSD_TCP_REASS */
1239	(void) tcp_reass(pData, tp, (struct tcphdr )0, (int )0, (struct mbuf *)0);
1240	#endif /VBOX_WITH_BSD_TCP_REASS/
1241	tp->snd_wl1 = ti->ti_seq - 1;
1242	/* Avoid ack processing; snd_una==ti_ack => dup ack */
1243	goto synrx_to_est;
1244	/* fall into ... */
1245
1246	/*
1247	* In ESTABLISHED state: drop duplicate ACKs; ACK out of range
1248	* ACKs. If the ack is in the range
1249	* tp->snd_una < ti->ti_ack <= tp->snd_max
1250	* then advance tp->snd_una to ti->ti_ack and drop
1251	* data from the retransmission queue. If this ACK reflects
1252	* more up to date window information we update our window information.
1253	*/
1254	case TCPS_ESTABLISHED:
1255	case TCPS_FIN_WAIT_1:
1256	case TCPS_FIN_WAIT_2:
1257	case TCPS_CLOSE_WAIT:
1258	case TCPS_CLOSING:
1259	case TCPS_LAST_ACK:
1260	case TCPS_TIME_WAIT:
1261
1262	if (SEQ_LEQ(ti->ti_ack, tp->snd_una)) {
1263	if (ti->ti_len == 0 && tiwin == tp->snd_wnd) {
1264	tcpstat.tcps_rcvdupack++;
1265	DEBUG_MISC((dfd," dup ack m = %lx so = %lx \n",
1266	(long )m, (long )so));
1267	/*
1268	* If we have outstanding data (other than
1269	* a window probe), this is a completely
1270	* duplicate ack (ie, window info didn't
1271	* change), the ack is the biggest we've
1272	* seen and we've seen exactly our rexmt
1273	* threshold of them, assume a packet
1274	* has been dropped and retransmit it.
1275	* Kludge snd_nxt & the congestion
1276	* window so we send only this one
1277	* packet.
1278	*
1279	* We know we're losing at the current
1280	* window size so do congestion avoidance
1281	* (set ssthresh to half the current window
1282	* and pull our congestion window back to
1283	* the new ssthresh).
1284	*
1285	* Dup acks mean that packets have left the
1286	* network (they're now cached at the receiver)
1287	* so bump cwnd by the amount in the receiver
1288	* to keep a constant cwnd packets in the
1289	* network.
1290	*/
1291	if (tp->t_timer[TCPT_REXMT] == 0 \|\|
1292	ti->ti_ack != tp->snd_una)
1293	tp->t_dupacks = 0;
1294	else if (++tp->t_dupacks == tcprexmtthresh) {
1295	tcp_seq onxt = tp->snd_nxt;
1296	u_int win =
1297	min(tp->snd_wnd, tp->snd_cwnd) / 2 /
1298	tp->t_maxseg;
1299
1300	if (win < 2)
1301	win = 2;
1302	tp->snd_ssthresh = win * tp->t_maxseg;
1303	tp->t_timer[TCPT_REXMT] = 0;
1304	tp->t_rtt = 0;
1305	tp->snd_nxt = ti->ti_ack;
1306	tp->snd_cwnd = tp->t_maxseg;
1307	(void) tcp_output(pData, tp);
1308	tp->snd_cwnd = tp->snd_ssthresh +
1309	tp->t_maxseg * tp->t_dupacks;
1310	if (SEQ_GT(onxt, tp->snd_nxt))
1311	tp->snd_nxt = onxt;
1312	goto drop;
1313	} else if (tp->t_dupacks > tcprexmtthresh) {
1314	tp->snd_cwnd += tp->t_maxseg;
1315	(void) tcp_output(pData, tp);
1316	goto drop;
1317	}
1318	} else
1319	tp->t_dupacks = 0;
1320	break;
1321	}
1322	synrx_to_est:
1323	/*
1324	* If the congestion window was inflated to account
1325	* for the other side's cached packets, retract it.
1326	*/
1327	if (tp->t_dupacks > tcprexmtthresh &&
1328	tp->snd_cwnd > tp->snd_ssthresh)
1329	tp->snd_cwnd = tp->snd_ssthresh;
1330	tp->t_dupacks = 0;
1331	if (SEQ_GT(ti->ti_ack, tp->snd_max)) {
1332	tcpstat.tcps_rcvacktoomuch++;
1333	goto dropafterack;
1334	}
1335	acked = ti->ti_ack - tp->snd_una;
1336	tcpstat.tcps_rcvackpack++;
1337	tcpstat.tcps_rcvackbyte += acked;
1338
1339	/*
1340	* If we have a timestamp reply, update smoothed
1341	* round trip time. If no timestamp is present but
1342	* transmit timer is running and timed sequence
1343	* number was acked, update smoothed round trip time.
1344	* Since we now have an rtt measurement, cancel the
1345	* timer backoff (cf., Phil Karn's retransmit alg.).
1346	* Recompute the initial retransmit timer.
1347	*/
1348	/* if (ts_present)
1349	* tcp_xmit_timer(tp, tcp_now-ts_ecr+1);
1350	* else
1351	*/
1352	if (tp->t_rtt && SEQ_GT(ti->ti_ack, tp->t_rtseq))
1353	tcp_xmit_timer(pData, tp,tp->t_rtt);
1354
1355	/*
1356	* If all outstanding data is acked, stop retransmit
1357	* timer and remember to restart (more output or persist).
1358	* If there is more data to be acked, restart retransmit
1359	* timer, using current (possibly backed-off) value.
1360	*/
1361	if (ti->ti_ack == tp->snd_max) {
1362	tp->t_timer[TCPT_REXMT] = 0;
1363	needoutput = 1;
1364	} else if (tp->t_timer[TCPT_PERSIST] == 0)
1365	tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
1366	/*
1367	* When new data is acked, open the congestion window.
1368	* If the window gives us less than ssthresh packets
1369	* in flight, open exponentially (maxseg per packet).
1370	* Otherwise open linearly: maxseg per window
1371	* (maxseg^2 / cwnd per packet).
1372	*/
1373	{
1374	register u_int cw = tp->snd_cwnd;
1375	register u_int incr = tp->t_maxseg;
1376
1377	if (cw > tp->snd_ssthresh)
1378	incr = incr * incr / cw;
1379	tp->snd_cwnd = min(cw + incr, TCP_MAXWIN<<tp->snd_scale);
1380	}
1381	if (acked > so->so_snd.sb_cc) {
1382	tp->snd_wnd -= so->so_snd.sb_cc;
1383	sbdrop(&so->so_snd, (int )so->so_snd.sb_cc);
1384	ourfinisacked = 1;
1385	} else {
1386	sbdrop(&so->so_snd, acked);
1387	tp->snd_wnd -= acked;
1388	ourfinisacked = 0;
1389	}
1390	/*
1391	* XXX sowwakup is called when data is acked and there's room for
1392	* for more data... it should read() the socket
1393	*/
1394	/* if (so->so_snd.sb_flags & SB_NOTIFY)
1395	* sowwakeup(so);
1396	*/
1397	tp->snd_una = ti->ti_ack;
1398	if (SEQ_LT(tp->snd_nxt, tp->snd_una))
1399	tp->snd_nxt = tp->snd_una;
1400
1401	switch (tp->t_state) {
1402
1403	/*
1404	* In FIN_WAIT_1 STATE in addition to the processing
1405	* for the ESTABLISHED state if our FIN is now acknowledged
1406	* then enter FIN_WAIT_2.
1407	*/
1408	case TCPS_FIN_WAIT_1:
1409	if (ourfinisacked) {
1410	/*
1411	* If we can't receive any more
1412	* data, then closing user can proceed.
1413	* Starting the timer is contrary to the
1414	* specification, but if we don't get a FIN
1415	* we'll hang forever.
1416	*/
1417	if (so->so_state & SS_FCANTRCVMORE) {
1418	soisfdisconnected(so);
1419	tp->t_timer[TCPT_2MSL] = tcp_maxidle;
1420	}
1421	tp->t_state = TCPS_FIN_WAIT_2;
1422	}
1423	break;
1424
1425	/*
1426	* In CLOSING STATE in addition to the processing for
1427	* the ESTABLISHED state if the ACK acknowledges our FIN
1428	* then enter the TIME-WAIT state, otherwise ignore
1429	* the segment.
1430	*/
1431	case TCPS_CLOSING:
1432	if (ourfinisacked) {
1433	tp->t_state = TCPS_TIME_WAIT;
1434	tcp_canceltimers(tp);
1435	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1436	soisfdisconnected(so);
1437	}
1438	break;
1439
1440	/*
1441	* In LAST_ACK, we may still be waiting for data to drain
1442	* and/or to be acked, as well as for the ack of our FIN.
1443	* If our FIN is now acknowledged, delete the TCB,
1444	* enter the closed state and return.
1445	*/
1446	case TCPS_LAST_ACK:
1447	if (ourfinisacked) {
1448	tp = tcp_close(pData, tp);
1449	goto drop;
1450	}
1451	break;
1452
1453	/*
1454	* In TIME_WAIT state the only thing that should arrive
1455	* is a retransmission of the remote FIN. Acknowledge
1456	* it and restart the finack timer.
1457	*/
1458	case TCPS_TIME_WAIT:
1459	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1460	goto dropafterack;
1461	}
1462	} /* switch(tp->t_state) */
1463
1464	step6:
1465	/*
1466	* Update window information.
1467	* Don't look at window if no ACK: TAC's send garbage on first SYN.
1468	*/
1469	if ((tiflags & TH_ACK) &&
1470	(SEQ_LT(tp->snd_wl1, ti->ti_seq) \|\|
1471	(tp->snd_wl1 == ti->ti_seq && (SEQ_LT(tp->snd_wl2, ti->ti_ack) \|\|
1472	(tp->snd_wl2 == ti->ti_ack && tiwin > tp->snd_wnd))))) {
1473	/* keep track of pure window updates */
1474	if (ti->ti_len == 0 &&
1475	tp->snd_wl2 == ti->ti_ack && tiwin > tp->snd_wnd)
1476	tcpstat.tcps_rcvwinupd++;
1477	tp->snd_wnd = tiwin;
1478	tp->snd_wl1 = ti->ti_seq;
1479	tp->snd_wl2 = ti->ti_ack;
1480	if (tp->snd_wnd > tp->max_sndwnd)
1481	tp->max_sndwnd = tp->snd_wnd;
1482	needoutput = 1;
1483	}
1484
1485	/*
1486	* Process segments with URG.
1487	*/
1488	if ((tiflags & TH_URG) && ti->ti_urp &&
1489	TCPS_HAVERCVDFIN(tp->t_state) == 0) {
1490	/*
1491	* This is a kludge, but if we receive and accept
1492	* random urgent pointers, we'll crash in
1493	* soreceive. It's hard to imagine someone
1494	* actually wanting to send this much urgent data.
1495	*/
1496	if (ti->ti_urp + so->so_rcv.sb_cc > so->so_rcv.sb_datalen) {
1497	ti->ti_urp = 0;
1498	tiflags &= ~TH_URG;
1499	goto dodata;
1500	}
1501	/*
1502	* If this segment advances the known urgent pointer,
1503	* then mark the data stream. This should not happen
1504	* in CLOSE_WAIT, CLOSING, LAST_ACK or TIME_WAIT STATES since
1505	* a FIN has been received from the remote side.
1506	* In these states we ignore the URG.
1507	*
1508	* According to RFC961 (Assigned Protocols),
1509	* the urgent pointer points to the last octet
1510	* of urgent data. We continue, however,
1511	* to consider it to indicate the first octet
1512	* of data past the urgent section as the original
1513	* spec states (in one of two places).
1514	*/
1515	if (SEQ_GT(ti->ti_seq+ti->ti_urp, tp->rcv_up)) {
1516	tp->rcv_up = ti->ti_seq + ti->ti_urp;
1517	so->so_urgc = so->so_rcv.sb_cc +
1518	(tp->rcv_up - tp->rcv_nxt); /* -1; */
1519	tp->rcv_up = ti->ti_seq + ti->ti_urp;
1520
1521	}
1522	} else
1523	/*
1524	* If no out of band data is expected,
1525	* pull receive urgent pointer along
1526	* with the receive window.
1527	*/
1528	if (SEQ_GT(tp->rcv_nxt, tp->rcv_up))
1529	tp->rcv_up = tp->rcv_nxt;
1530	dodata:
1531
1532	/*
1533	* Process the segment text, merging it into the TCP sequencing queue,
1534	* and arranging for acknowledgment of receipt if necessary.
1535	* This process logically involves adjusting tp->rcv_wnd as data
1536	* is presented to the user (this happens in tcp_usrreq.c,
1537	* case PRU_RCVD). If a FIN has already been received on this
1538	* connection then we just ignore the text.
1539	*/
1540	if ((ti->ti_len \|\| (tiflags&TH_FIN)) &&
1541	TCPS_HAVERCVDFIN(tp->t_state) == 0) {
1542	#ifndef VBOX_WITH_BSD_TCP_REASS
1543	TCP_REASS(pData, tp, ti, m, so, tiflags);
1544	#else /* VBOX_WITH_BSD_TCP_REASS */
1545	if (ti->ti_seq == tp->rcv_nxt
1546	&& LIST_EMPTY(&tp->t_segq)
1547	&& tp->t_state == TCPS_ESTABLISHED) {
1548	DELAY_ACK(tp, ti); /* little bit different from BSD declaration see netinet/tcp_input.c */
1549	tp->rcv_nxt += tlen;
1550	tiflags = ti->ti_t.th_flags & TH_FIN;
1551	tcpstat.tcps_rcvpack++;
1552	tcpstat.tcps_rcvbyte += tlen;
1553	if (so->so_state & SS_FCANTRCVMORE)
1554	m_freem(pData, m);
1555	else
1556	sbappend(pData, so, m);
1557	}
1558	else {
1559	tiflags = tcp_reass(pData, tp, &ti->ti_t, &tlen, m);
1560	tiflags \|= TF_ACKNOW;
1561	}
1562	#endif /* VBOX_WITH_BSD_TCP_REASS */
1563	/*
1564	* Note the amount of data that peer has sent into
1565	* our window, in order to estimate the sender's
1566	* buffer size.
1567	*/
1568	len = so->so_rcv.sb_datalen - (tp->rcv_adv - tp->rcv_nxt);
1569	} else {
1570	mbuf_freed = 1; /* The mbuf must be freed, but only when its content is not needed anymore. */
1571	tiflags &= ~TH_FIN;
1572	}
1573
1574	/*
1575	* If FIN is received ACK the FIN and let the user know
1576	* that the connection is closing.
1577	*/
1578	if (tiflags & TH_FIN) {
1579	if (TCPS_HAVERCVDFIN(tp->t_state) == 0) {
1580	/*
1581	* If we receive a FIN we can't send more data,
1582	* set it SS_FDRAIN
1583	* Shutdown the socket if there is no rx data in the
1584	* buffer.
1585	* soread() is called on completion of shutdown() and
1586	* will got to TCPS_LAST_ACK, and use tcp_output()
1587	* to send the FIN.
1588	*/
1589	/* sofcantrcvmore(so); */
1590	sofwdrain(so);
1591
1592	tp->t_flags \|= TF_ACKNOW;
1593	tp->rcv_nxt++;
1594	}
1595	switch (tp->t_state) {
1596
1597	/*
1598	* In SYN_RECEIVED and ESTABLISHED STATES
1599	* enter the CLOSE_WAIT state.
1600	*/
1601	case TCPS_SYN_RECEIVED:
1602	case TCPS_ESTABLISHED:
1603	if(so->so_emu == EMU_CTL) /* no shutdown on socket */
1604	tp->t_state = TCPS_LAST_ACK;
1605	else
1606	tp->t_state = TCPS_CLOSE_WAIT;
1607	break;
1608
1609	/*
1610	* If still in FIN_WAIT_1 STATE FIN has not been acked so
1611	* enter the CLOSING state.
1612	*/
1613	case TCPS_FIN_WAIT_1:
1614	tp->t_state = TCPS_CLOSING;
1615	break;
1616
1617	/*
1618	* In FIN_WAIT_2 state enter the TIME_WAIT state,
1619	* starting the time-wait timer, turning off the other
1620	* standard timers.
1621	*/
1622	case TCPS_FIN_WAIT_2:
1623	tp->t_state = TCPS_TIME_WAIT;
1624	tcp_canceltimers(tp);
1625	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1626	soisfdisconnected(so);
1627	break;
1628
1629	/*
1630	* In TIME_WAIT state restart the 2 MSL time_wait timer.
1631	*/
1632	case TCPS_TIME_WAIT:
1633	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1634	break;
1635	}
1636	}
1637
1638	/*
1639	* If this is a small packet, then ACK now - with Nagel
1640	* congestion avoidance sender won't send more until
1641	* he gets an ACK.
1642	*
1643	* See above.
1644	*/
1645	/* if (ti->ti_len && (unsigned)ti->ti_len < tp->t_maxseg) {
1646	*/
1647	/* if ((ti->ti_len && (unsigned)ti->ti_len < tp->t_maxseg &&
1648	* (so->so_iptos & IPTOS_LOWDELAY) == 0) \|\|
1649	* ((so->so_iptos & IPTOS_LOWDELAY) &&
1650	* ((struct tcpiphdr_2 *)ti)->first_char == (char)27)) {
1651	*/
1652	if (ti->ti_len && (unsigned)ti->ti_len <= 5 &&
1653	((struct tcpiphdr_2 *)ti)->first_char == (char)27) {
1654	tp->t_flags \|= TF_ACKNOW;
1655	}
1656
1657	if (mbuf_freed) {
1658	m_free(pData, m);
1659	}
1660	/*
1661	* Return any desired output.
1662	*/
1663	if (needoutput \|\| (tp->t_flags & TF_ACKNOW)) {
1664	(void) tcp_output(pData, tp);
1665	}
1666	return;
1667
1668	dropafterack:
1669	/*
1670	* Generate an ACK dropping incoming segment if it occupies
1671	* sequence space, where the ACK reflects our state.
1672	*/
1673	if (tiflags & TH_RST)
1674	goto drop;
1675	m_freem(pData, m);
1676	tp->t_flags \|= TF_ACKNOW;
1677	(void) tcp_output(pData, tp);
1678	return;
1679
1680	dropwithreset:
1681	/* reuses m if m!=NULL, m_free() unnecessary */
1682	if (tiflags & TH_ACK)
1683	tcp_respond(pData, tp, ti, m, (tcp_seq)0, ti->ti_ack, TH_RST);
1684	else {
1685	if (tiflags & TH_SYN) ti->ti_len++;
1686	tcp_respond(pData, tp, ti, m, ti->ti_seq+ti->ti_len, (tcp_seq)0,
1687	TH_RST\|TH_ACK);
1688	}
1689
1690	return;
1691
1692	drop:
1693	/*
1694	* Drop space held by incoming segment and return.
1695	*/
1696	m_free(pData, m);
1697
1698	return;
1699	}
1700
1701	/* , ts_present, ts_val, ts_ecr) */
1702	/* int *ts_present;
1703	* u_int32_t ts_val, ts_ecr;
1704	*/
1705	void
1706	tcp_dooptions(PNATState pData, struct tcpcb tp, u_char cp, int cnt, struct tcpiphdr *ti)
1707	{
1708	u_int16_t mss;
1709	int opt, optlen;
1710
1711	DEBUG_CALL("tcp_dooptions");
1712	DEBUG_ARGS((dfd," tp = %lx cnt=%i \n", (long )tp, cnt));
1713
1714	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1715	opt = cp[0];
1716	if (opt == TCPOPT_EOL)
1717	break;
1718	if (opt == TCPOPT_NOP)
1719	optlen = 1;
1720	else {
1721	optlen = cp[1];
1722	if (optlen <= 0)
1723	break;
1724	}
1725	switch (opt) {
1726
1727	default:
1728	continue;
1729
1730	case TCPOPT_MAXSEG:
1731	if (optlen != TCPOLEN_MAXSEG)
1732	continue;
1733	if (!(ti->ti_flags & TH_SYN))
1734	continue;
1735	memcpy((char ) &mss, (char ) cp + 2, sizeof(mss));
1736	NTOHS(mss);
1737	(void) tcp_mss(pData, tp, mss); /* sets t_maxseg */
1738	break;
1739
1740	/* case TCPOPT_WINDOW:
1741	* if (optlen != TCPOLEN_WINDOW)
1742	* continue;
1743	* if (!(ti->ti_flags & TH_SYN))
1744	* continue;
1745	* tp->t_flags \|= TF_RCVD_SCALE;
1746	* tp->requested_s_scale = min(cp[2], TCP_MAX_WINSHIFT);
1747	* break;
1748	*/
1749	/* case TCPOPT_TIMESTAMP:
1750	* if (optlen != TCPOLEN_TIMESTAMP)
1751	* continue;
1752	* *ts_present = 1;
1753	* memcpy((char ) ts_val, (char )cp + 2, sizeof(*ts_val));
1754	* NTOHL(*ts_val);
1755	* memcpy((char ) ts_ecr, (char )cp + 6, sizeof(*ts_ecr));
1756	* NTOHL(*ts_ecr);
1757	*
1758	/ /
1759	* * A timestamp received in a SYN makes
1760	* * it ok to send timestamp requests and replies.
1761	* */
1762	/* if (ti->ti_flags & TH_SYN) {
1763	* tp->t_flags \|= TF_RCVD_TSTMP;
1764	* tp->ts_recent = *ts_val;
1765	* tp->ts_recent_age = tcp_now;
1766	* }
1767	*/ break;
1768	}
1769	}
1770	}
1771
1772
1773	/*
1774	* Pull out of band byte out of a segment so
1775	* it doesn't appear in the user's data queue.
1776	* It is still reflected in the segment length for
1777	* sequencing purposes.
1778	*/
1779
1780	#ifdef notdef
1781
1782	void
1783	tcp_pulloutofband(so, ti, m)
1784	struct socket *so;
1785	struct tcpiphdr *ti;
1786	register struct mbuf *m;
1787	{
1788	int cnt = ti->ti_urp - 1;
1789
1790	while (cnt >= 0) {
1791	if (m->m_len > cnt) {
1792	char *cp = mtod(m, caddr_t) + cnt;
1793	struct tcpcb *tp = sototcpcb(so);
1794
1795	tp->t_iobc = *cp;
1796	tp->t_oobflags \|= TCPOOB_HAVEDATA;
1797	memcpy(sp, cp+1, (unsigned)(m->m_len - cnt - 1));
1798	m->m_len--;
1799	return;
1800	}
1801	cnt -= m->m_len;
1802	m = m->m_next; /* XXX WRONG! Fix it! */
1803	if (m == 0)
1804	break;
1805	}
1806	panic("tcp_pulloutofband");
1807	}
1808
1809	#endif /* notdef */
1810
1811	/*
1812	* Collect new round-trip time estimate
1813	* and update averages and current timeout.
1814	*/
1815
1816	void
1817	tcp_xmit_timer(PNATState pData, register struct tcpcb *tp, int rtt)
1818	{
1819	register short delta;
1820
1821	DEBUG_CALL("tcp_xmit_timer");
1822	DEBUG_ARG("tp = %lx", (long)tp);
1823	DEBUG_ARG("rtt = %d", rtt);
1824
1825	tcpstat.tcps_rttupdated++;
1826	if (tp->t_srtt != 0) {
1827	/*
1828	* srtt is stored as fixed point with 3 bits after the
1829	* binary point (i.e., scaled by 8). The following magic
1830	* is equivalent to the smoothing algorithm in rfc793 with
1831	* an alpha of .875 (srtt = rtt/8 + srtt*7/8 in fixed
1832	* point). Adjust rtt to origin 0.
1833	*/
1834	delta = rtt - 1 - (tp->t_srtt >> TCP_RTT_SHIFT);
1835	if ((tp->t_srtt += delta) <= 0)
1836	tp->t_srtt = 1;
1837	/*
1838	* We accumulate a smoothed rtt variance (actually, a
1839	* smoothed mean difference), then set the retransmit
1840	* timer to smoothed rtt + 4 times the smoothed variance.
1841	* rttvar is stored as fixed point with 2 bits after the
1842	* binary point (scaled by 4). The following is
1843	* equivalent to rfc793 smoothing with an alpha of .75
1844	* (rttvar = rttvar*3/4 + \|delta\| / 4). This replaces
1845	* rfc793's wired-in beta.
1846	*/
1847	if (delta < 0)
1848	delta = -delta;
1849	delta -= (tp->t_rttvar >> TCP_RTTVAR_SHIFT);
1850	if ((tp->t_rttvar += delta) <= 0)
1851	tp->t_rttvar = 1;
1852	} else {
1853	/*
1854	* No rtt measurement yet - use the unsmoothed rtt.
1855	* Set the variance to half the rtt (so our first
1856	* retransmit happens at 3*rtt).
1857	*/
1858	tp->t_srtt = rtt << TCP_RTT_SHIFT;
1859	tp->t_rttvar = rtt << (TCP_RTTVAR_SHIFT - 1);
1860	}
1861	tp->t_rtt = 0;
1862	tp->t_rxtshift = 0;
1863
1864	/*
1865	* the retransmit should happen at rtt + 4 * rttvar.
1866	* Because of the way we do the smoothing, srtt and rttvar
1867	* will each average +1/2 tick of bias. When we compute
1868	* the retransmit timer, we want 1/2 tick of rounding and
1869	* 1 extra tick because of +-1/2 tick uncertainty in the
1870	* firing of the timer. The bias will give us exactly the
1871	* 1.5 tick we need. But, because the bias is
1872	* statistical, we have to test that we don't drop below
1873	* the minimum feasible timer (which is 2 ticks).
1874	*/
1875	TCPT_RANGESET(tp->t_rxtcur, TCP_REXMTVAL(tp),
1876	(short)tp->t_rttmin, TCPTV_REXMTMAX); /* XXX */
1877
1878	/*
1879	* We received an ack for a packet that wasn't retransmitted;
1880	* it is probably safe to discard any error indications we've
1881	* received recently. This isn't quite right, but close enough
1882	* for now (a route might have failed after we sent a segment,
1883	* and the return path might not be symmetrical).
1884	*/
1885	tp->t_softerror = 0;
1886	}
1887
1888	/*
1889	* Determine a reasonable value for maxseg size.
1890	* If the route is known, check route for mtu.
1891	* If none, use an mss that can be handled on the outgoing
1892	* interface without forcing IP to fragment; if bigger than
1893	* an mbuf cluster (MCLBYTES), round down to nearest multiple of MCLBYTES
1894	* to utilize large mbufs. If no route is found, route has no mtu,
1895	* or the destination isn't local, use a default, hopefully conservative
1896	* size (usually 512 or the default IP max size, but no more than the mtu
1897	* of the interface), as we can't discover anything about intervening
1898	* gateways or networks. We also initialize the congestion/slow start
1899	* window to be a single segment if the destination isn't local.
1900	* While looking at the routing entry, we also initialize other path-dependent
1901	* parameters from pre-set or cached values in the routing entry.
1902	*/
1903
1904	int
1905	tcp_mss(PNATState pData, register struct tcpcb *tp, u_int offer)
1906	{
1907	struct socket *so = tp->t_socket;
1908	int mss;
1909
1910	DEBUG_CALL("tcp_mss");
1911	DEBUG_ARG("tp = %lx", (long)tp);
1912	DEBUG_ARG("offer = %d", offer);
1913
1914	mss = min(if_mtu, if_mru) - sizeof(struct tcpiphdr);
1915	if (offer)
1916	mss = min(mss, offer);
1917	mss = max(mss, 32);
1918	if (mss < tp->t_maxseg \|\| offer != 0)
1919	tp->t_maxseg = mss;
1920
1921	tp->snd_cwnd = mss;
1922
1923	sbreserve(&so->so_snd, tcp_sndspace+((tcp_sndspace%mss)?(mss-(tcp_sndspace%mss)):0));
1924	sbreserve(&so->so_rcv, tcp_rcvspace+((tcp_rcvspace%mss)?(mss-(tcp_rcvspace%mss)):0));
1925
1926	DEBUG_MISC((dfd, " returning mss = %d\n", mss));
1927
1928	return mss;
1929	}

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

source: vbox/trunk/src/VBox/Devices/Network/slirp/tcp_input.c@ 14309

Download in other formats: