tcp_input.c@ 13958

Last change on this file since 13958 was 13783, checked in by vboxsync, 16 years ago
per-socket and per-mbuf mutexes are removed only global locks are stayin on their places
Property svn:eol-style set to `native`
File size: 49.2 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	/*
56	* Insert segment ti into reassembly queue of tcp with
57	* control block tp. Return TH_FIN if reassembly now includes
58	* a segment with FIN. The macro form does the common case inline
59	* (segment is the next to be received on an established connection,
60	* and the queue is empty), avoiding linkage into and removal
61	* from the queue and repetition of various conversions.
62	* Set DELACK for segments received in order, but ack immediately
63	* when segments are out of order (so fast retransmit can work).
64	*/
65	#ifdef TCP_ACK_HACK
66	#define TCP_REASS(pData, tp, ti, m, so, flags) {\
67	if ((ti)->ti_seq == (tp)->rcv_nxt && \
68	u32_to_ptr((pData), (tp)->seg_next, struct tcpcb *) == (tp) && \
69	(tp)->t_state == TCPS_ESTABLISHED) {\
70	if (ti->ti_flags & TH_PUSH) \
71	tp->t_flags \|= TF_ACKNOW; \
72	else \
73	tp->t_flags \|= TF_DELACK; \
74	(tp)->rcv_nxt += (ti)->ti_len; \
75	flags = (ti)->ti_flags & TH_FIN; \
76	tcpstat.tcps_rcvpack++;\
77	tcpstat.tcps_rcvbyte += (ti)->ti_len;\
78	if (so->so_emu) { \
79	if (tcp_emu((pData), (so),(m))) sbappend((pData), (so), (m)); \
80	} else \
81	sbappend((pData), (so), (m)); \
82	/* sorwakeup(so); */ \
83	} else {\
84	(flags) = tcp_reass((pData), (tp), (ti), (m)); \
85	tp->t_flags \|= TF_ACKNOW; \
86	} \
87	}
88	#else
89	#define TCP_REASS(pData, tp, ti, m, so, flags) { \
90	if ((ti)->ti_seq == (tp)->rcv_nxt && \
91	u32_to_ptr((pData), (tp)->seg_next, struct tcpcb *) == (tp) && \
92	(tp)->t_state == TCPS_ESTABLISHED) { \
93	tp->t_flags \|= TF_DELACK; \
94	(tp)->rcv_nxt += (ti)->ti_len; \
95	flags = (ti)->ti_flags & TH_FIN; \
96	tcpstat.tcps_rcvpack++;\
97	tcpstat.tcps_rcvbyte += (ti)->ti_len;\
98	if (so->so_emu) { \
99	if (tcp_emu((pData), (so),(m))) sbappend((pData), (so), (m)); \
100	} else \
101	sbappend((pData), (so), (m)); \
102	/* sorwakeup(so); */ \
103	} else { \
104	(flags) = tcp_reass((pData), (tp), (ti), (m)); \
105	tp->t_flags \|= TF_ACKNOW; \
106	} \
107	}
108	#endif
109
110	int
111	tcp_reass(PNATState pData, register struct tcpcb tp, register struct tcpiphdr ti, struct mbuf *m)
112	{
113	register struct tcpiphdr *q;
114	struct socket *so = tp->t_socket;
115	int flags;
116
117	/*
118	* Call with ti==0 after become established to
119	* force pre-ESTABLISHED data up to user socket.
120	*/
121	if (ti == 0)
122	goto present;
123
124	/*
125	* Find a segment which begins after this one does.
126	*/
127	for (q = u32_to_ptr(pData, tp->seg_next, struct tcpiphdr ); q != (struct tcpiphdr )tp;
128	q = u32_to_ptr(pData, q->ti_next, struct tcpiphdr *))
129	if (SEQ_GT(q->ti_seq, ti->ti_seq))
130	break;
131
132	/*
133	* If there is a preceding segment, it may provide some of
134	* our data already. If so, drop the data from the incoming
135	* segment. If it provides all of our data, drop us.
136	*/
137	if (u32_to_ptr(pData, q->ti_prev, struct tcpiphdr ) != (struct tcpiphdr )tp) {
138	register int i;
139	q = u32_to_ptr(pData, q->ti_prev, struct tcpiphdr *);
140	/* conversion to int (in i) handles seq wraparound */
141	i = q->ti_seq + q->ti_len - ti->ti_seq;
142	if (i > 0) {
143	if (i >= ti->ti_len) {
144	tcpstat.tcps_rcvduppack++;
145	tcpstat.tcps_rcvdupbyte += ti->ti_len;
146	m_freem(pData, m);
147	/*
148	* Try to present any queued data
149	* at the left window edge to the user.
150	* This is needed after the 3-WHS
151	* completes.
152	*/
153	goto present; /* ??? */
154	}
155	m_adj(m, i);
156	ti->ti_len -= i;
157	ti->ti_seq += i;
158	}
159	q = u32_to_ptr(pData, q->ti_next, struct tcpiphdr *);
160	}
161	tcpstat.tcps_rcvoopack++;
162	tcpstat.tcps_rcvoobyte += ti->ti_len;
163	REASS_MBUF_SET(ti, m); /* XXX */
164
165	/*
166	* While we overlap succeeding segments trim them or,
167	* if they are completely covered, dequeue them.
168	*/
169	while (q != (struct tcpiphdr *)tp) {
170	register int i = (ti->ti_seq + ti->ti_len) - q->ti_seq;
171	if (i <= 0)
172	break;
173	if (i < q->ti_len) {
174	q->ti_seq += i;
175	q->ti_len -= i;
176	m_adj(REASS_MBUF_GET(q), i);
177	break;
178	}
179	q = u32_to_ptr(pData, q->ti_next, struct tcpiphdr *);
180	m = REASS_MBUF_GET(u32_to_ptr(pData, q->ti_prev, struct tcpiphdr *));
181	remque_32(pData, u32_to_ptr(pData, q->ti_prev, struct tcpiphdr *));
182	m_freem(pData, m);
183	}
184
185	/*
186	* Stick new segment in its place.
187	*/
188	insque_32(pData, ti, u32_to_ptr(pData, q->ti_prev, struct tcpiphdr *));
189
190	present:
191	/*
192	* Present data to user, advancing rcv_nxt through
193	* completed sequence space.
194	*/
195	if (!TCPS_HAVEESTABLISHED(tp->t_state))
196	return (0);
197	ti = u32_to_ptr(pData, tp->seg_next, struct tcpiphdr *);
198	if (ti == (struct tcpiphdr *)tp \|\| ti->ti_seq != tp->rcv_nxt)
199	return (0);
200	if (tp->t_state == TCPS_SYN_RECEIVED && ti->ti_len)
201	return (0);
202	do {
203	tp->rcv_nxt += ti->ti_len;
204	flags = ti->ti_flags & TH_FIN;
205	remque_32(pData, ti);
206	m = REASS_MBUF_GET(ti); /* XXX */
207	ti = u32_to_ptr(pData, ti->ti_next, struct tcpiphdr *);
208	/* if (so->so_state & SS_FCANTRCVMORE) */
209	if (so->so_state & SS_FCANTSENDMORE)
210	m_freem(pData, m);
211	else {
212	if (so->so_emu) {
213	if (tcp_emu(pData, so,m)) sbappend(pData, so, m);
214	} else
215	sbappend(pData, so, m);
216	}
217	} while (ti != (struct tcpiphdr *)tp && ti->ti_seq == tp->rcv_nxt);
218	/* sorwakeup(so); */
219	return (flags);
220	}
221
222	/*
223	* TCP input routine, follows pages 65-76 of the
224	* protocol specification dated September, 1981 very closely.
225	*/
226	void
227	tcp_input(PNATState pData, register struct mbuf m, int iphlen, struct socket inso)
228	{
229	struct ip save_ip, *ip;
230	register struct tcpiphdr *ti;
231	caddr_t optp = NULL;
232	int optlen = 0;
233	int len, tlen, off;
234	register struct tcpcb *tp = 0;
235	register int tiflags;
236	struct socket *so = 0;
237	int todrop, acked, ourfinisacked, needoutput = 0;
238	/* int dropsocket = 0; */
239	int iss = 0;
240	u_long tiwin;
241	int ret;
242	/* int ts_present = 0; */
243	int mbuf_freed = 0;
244
245	DEBUG_CALL("tcp_input");
246	DEBUG_ARGS((dfd," m = %8lx iphlen = %2d inso = %lx\n",
247	(long )m, iphlen, (long )inso ));
248	#ifdef VBOX_WITH_SYNC_SLIRP
249	#if 0
250	#define return \
251	do { \
252	fprintf(stderr, "%s:%d\n", __FILE__, __LINE__); \
253	return; \
254	}while(0)
255	#endif
256	#endif
257
258	/*
259	* If called with m == 0, then we're continuing the connect
260	*/
261	if (m == NULL) {
262	so = inso;
263
264	/* Re-set a few variables */
265	tp = sototcpcb(so);
266	m = so->so_m;
267	so->so_m = 0;
268	ti = so->so_ti;
269	tiwin = ti->ti_win;
270	tiflags = ti->ti_flags;
271
272	goto cont_conn;
273	}
274
275
276	tcpstat.tcps_rcvtotal++;
277	/*
278	* Get IP and TCP header together in first mbuf.
279	* Note: IP leaves IP header in first mbuf.
280	*/
281	ti = mtod(m, struct tcpiphdr *);
282	if (iphlen > sizeof(struct ip )) {
283	ip_stripoptions(m, (struct mbuf *)0);
284	iphlen=sizeof(struct ip );
285	}
286	/* XXX Check if too short */
287
288
289	/*
290	* Save a copy of the IP header in case we want restore it
291	* for sending an ICMP error message in response.
292	*/
293	ip=mtod(m, struct ip *);
294	save_ip = *ip;
295	save_ip.ip_len+= iphlen;
296
297	/*
298	* Checksum extended TCP header and data.
299	*/
300	tlen = ((struct ip *)ti)->ip_len;
301	ti->ti_next = ti->ti_prev = 0;
302	ti->ti_x1 = 0;
303	ti->ti_len = htons((u_int16_t)tlen);
304	len = sizeof(struct ip ) + tlen;
305	/* keep checksum for ICMP reply
306	* ti->ti_sum = cksum(m, len);
307	* if (ti->ti_sum) { */
308	if(cksum(m, len)) {
309	tcpstat.tcps_rcvbadsum++;
310	goto drop;
311	}
312
313	/*
314	* Check that TCP offset makes sense,
315	* pull out TCP options and adjust length. XXX
316	*/
317	off = ti->ti_off << 2;
318	if (off < sizeof (struct tcphdr) \|\| off > tlen) {
319	tcpstat.tcps_rcvbadoff++;
320	goto drop;
321	}
322	tlen -= off;
323	ti->ti_len = tlen;
324	if (off > sizeof (struct tcphdr)) {
325	optlen = off - sizeof (struct tcphdr);
326	optp = mtod(m, caddr_t) + sizeof (struct tcpiphdr);
327
328	/*
329	* Do quick retrieval of timestamp options ("options
330	* prediction?"). If timestamp is the only option and it's
331	* formatted as recommended in RFC 1323 appendix A, we
332	* quickly get the values now and not bother calling
333	* tcp_dooptions(), etc.
334	*/
335	/* if ((optlen == TCPOLEN_TSTAMP_APPA \|\|
336	* (optlen > TCPOLEN_TSTAMP_APPA &&
337	* optp[TCPOLEN_TSTAMP_APPA] == TCPOPT_EOL)) &&
338	* (u_int32_t )optp == htonl(TCPOPT_TSTAMP_HDR) &&
339	* (ti->ti_flags & TH_SYN) == 0) {
340	* ts_present = 1;
341	* ts_val = ntohl((u_int32_t )(optp + 4));
342	* ts_ecr = ntohl((u_int32_t )(optp + 8));
343	* optp = NULL; / * we've parsed the options * /
344	* }
345	*/
346	}
347	tiflags = ti->ti_flags;
348
349	/*
350	* Convert TCP protocol specific fields to host format.
351	*/
352	NTOHL(ti->ti_seq);
353	NTOHL(ti->ti_ack);
354	NTOHS(ti->ti_win);
355	NTOHS(ti->ti_urp);
356
357	/*
358	* Drop TCP, IP headers and TCP options.
359	*/
360	m->m_data += sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
361	m->m_len -= sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
362
363	/*
364	* Locate pcb for segment.
365	*/
366	findso:
367	VBOX_SLIRP_LOCK(pData->tcp_last_so_mutex);
368	so = tcp_last_so;
369	/* this checking for making sure that we're not trying to hold mutex on head list*/
370	VBOX_SLIRP_UNLOCK(pData->tcp_last_so_mutex);
371
372	if (so->so_fport != ti->ti_dport \|\|
373	so->so_lport != ti->ti_sport \|\|
374	so->so_laddr.s_addr != ti->ti_src.s_addr \|\|
375	so->so_faddr.s_addr != ti->ti_dst.s_addr) {
376	/To make sure that we don't try to release mutex on head of the socket queue/
377	so = solookup(&tcb, ti->ti_src, ti->ti_sport,
378	ti->ti_dst, ti->ti_dport);
379	if (so) {
380	VBOX_SLIRP_LOCK(pData->tcp_last_so_mutex);
381	tcp_last_so = so;
382	VBOX_SLIRP_UNLOCK(pData->tcp_last_so_mutex);
383	}
384	++tcpstat.tcps_socachemiss;
385	}
386
387	/*
388	* If the state is CLOSED (i.e., TCB does not exist) then
389	* all data in the incoming segment is discarded.
390	* If the TCB exists but is in CLOSED state, it is embryonic,
391	* but should either do a listen or a connect soon.
392	*
393	* state == CLOSED means we've done socreate() but haven't
394	* attached it to a protocol yet...
395	*
396	* XXX If a TCB does not exist, and the TH_SYN flag is
397	* the only flag set, then create a session, mark it
398	* as if it was LISTENING, and continue...
399	*/
400	if (so == 0) {
401	if ((tiflags & (TH_SYN\|TH_FIN\|TH_RST\|TH_URG\|TH_ACK)) != TH_SYN)
402	goto dropwithreset;
403
404	if ((so = socreate()) == NULL)
405	goto dropwithreset;
406
407	if (tcp_attach(pData, so) < 0) {
408	free(so); /* Not sofree (if it failed, it's not insqued) */
409	#ifdef VBOX_WITH_SYNC_SLIRP
410	so = NULL;
411	#endif
412	goto dropwithreset;
413	}
414
415	sbreserve(&so->so_snd, tcp_sndspace);
416	sbreserve(&so->so_rcv, tcp_rcvspace);
417
418	/* tcp_last_so = so; / / XXX ? */
419	/* tp = sototcpcb(so); */
420
421	so->so_laddr = ti->ti_src;
422	so->so_lport = ti->ti_sport;
423	so->so_faddr = ti->ti_dst;
424	so->so_fport = ti->ti_dport;
425
426	if ((so->so_iptos = tcp_tos(so)) == 0)
427	so->so_iptos = ((struct ip *)ti)->ip_tos;
428
429	tp = sototcpcb(so);
430	tp->t_state = TCPS_LISTEN;
431	}
432
433	/*
434	* If this is a still-connecting socket, this probably
435	* a retransmit of the SYN. Whether it's a retransmit SYN
436	* or something else, we nuke it.
437	*/
438	if (so->so_state & SS_ISFCONNECTING)
439	goto drop;
440
441	tp = sototcpcb(so);
442
443	/* XXX Should never fail */
444	if (tp == 0)
445	goto dropwithreset;
446	if (tp->t_state == TCPS_CLOSED)
447	goto drop;
448
449	/* Unscale the window into a 32-bit value. */
450	/* if ((tiflags & TH_SYN) == 0)
451	* tiwin = ti->ti_win << tp->snd_scale;
452	* else
453	*/
454	tiwin = ti->ti_win;
455
456	/*
457	* Segment received on connection.
458	* Reset idle time and keep-alive timer.
459	*/
460	tp->t_idle = 0;
461	if (so_options)
462	tp->t_timer[TCPT_KEEP] = tcp_keepintvl;
463	else
464	tp->t_timer[TCPT_KEEP] = tcp_keepidle;
465
466	/*
467	* Process options if not in LISTEN state,
468	* else do it below (after getting remote address).
469	*/
470	if (optp && tp->t_state != TCPS_LISTEN)
471	tcp_dooptions(pData, tp, (u_char *)optp, optlen, ti);
472	/* , */
473	/* &ts_present, &ts_val, &ts_ecr); */
474
475	/*
476	* Header prediction: check for the two common cases
477	* of a uni-directional data xfer. If the packet has
478	* no control flags, is in-sequence, the window didn't
479	* change and we're not retransmitting, it's a
480	* candidate. If the length is zero and the ack moved
481	* forward, we're the sender side of the xfer. Just
482	* free the data acked & wake any higher level process
483	* that was blocked waiting for space. If the length
484	* is non-zero and the ack didn't move, we're the
485	* receiver side. If we're getting packets in-order
486	* (the reassembly queue is empty), add the data to
487	* the socket buffer and note that we need a delayed ack.
488	*
489	* XXX Some of these tests are not needed
490	* eg: the tiwin == tp->snd_wnd prevents many more
491	* predictions.. with no real advantage..
492	*/
493	if (tp->t_state == TCPS_ESTABLISHED &&
494	(tiflags & (TH_SYN\|TH_FIN\|TH_RST\|TH_URG\|TH_ACK)) == TH_ACK &&
495	/* (!ts_present \|\| TSTMP_GEQ(ts_val, tp->ts_recent)) && */
496	ti->ti_seq == tp->rcv_nxt &&
497	tiwin && tiwin == tp->snd_wnd &&
498	tp->snd_nxt == tp->snd_max) {
499	/*
500	* If last ACK falls within this segment's sequence numbers,
501	* record the timestamp.
502	*/
503	/* if (ts_present && SEQ_LEQ(ti->ti_seq, tp->last_ack_sent) &&
504	* SEQ_LT(tp->last_ack_sent, ti->ti_seq + ti->ti_len)) {
505	* tp->ts_recent_age = tcp_now;
506	* tp->ts_recent = ts_val;
507	* }
508	*/
509	if (ti->ti_len == 0) {
510	if (SEQ_GT(ti->ti_ack, tp->snd_una) &&
511	SEQ_LEQ(ti->ti_ack, tp->snd_max) &&
512	tp->snd_cwnd >= tp->snd_wnd) {
513	/*
514	* this is a pure ack for outstanding data.
515	*/
516	++tcpstat.tcps_predack;
517	/* if (ts_present)
518	* tcp_xmit_timer(tp, tcp_now-ts_ecr+1);
519	* else
520	*/ if (tp->t_rtt &&
521	SEQ_GT(ti->ti_ack, tp->t_rtseq))
522	tcp_xmit_timer(pData, tp, tp->t_rtt);
523	acked = ti->ti_ack - tp->snd_una;
524	tcpstat.tcps_rcvackpack++;
525	tcpstat.tcps_rcvackbyte += acked;
526	sbdrop(&so->so_snd, acked);
527	tp->snd_una = ti->ti_ack;
528	m_freem(pData, m);
529
530	/*
531	* If all outstanding data are acked, stop
532	* retransmit timer, otherwise restart timer
533	* using current (possibly backed-off) value.
534	* If process is waiting for space,
535	* wakeup/selwakeup/signal. If data
536	* are ready to send, let tcp_output
537	* decide between more output or persist.
538	*/
539	if (tp->snd_una == tp->snd_max)
540	tp->t_timer[TCPT_REXMT] = 0;
541	else if (tp->t_timer[TCPT_PERSIST] == 0)
542	tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
543
544	/*
545	* There's room in so_snd, sowwakup will read()
546	* from the socket if we can
547	*/
548	/* if (so->so_snd.sb_flags & SB_NOTIFY)
549	* sowwakeup(so);
550	*/
551	/*
552	* This is called because sowwakeup might have
553	* put data into so_snd. Since we don't so sowwakeup,
554	* we don't need this.. XXX???
555	*/
556	if (so->so_snd.sb_cc)
557	(void) tcp_output(pData, tp);
558	return;
559	}
560	} else if (ti->ti_ack == tp->snd_una &&
561	u32_to_ptr(pData, tp->seg_next, struct tcpcb *) == tp &&
562	ti->ti_len <= sbspace(&so->so_rcv)) {
563	/*
564	* this is a pure, in-sequence data packet
565	* with nothing on the reassembly queue and
566	* we have enough buffer space to take it.
567	*/
568	++tcpstat.tcps_preddat;
569	tp->rcv_nxt += ti->ti_len;
570	tcpstat.tcps_rcvpack++;
571	tcpstat.tcps_rcvbyte += ti->ti_len;
572	/*
573	* Add data to socket buffer.
574	*/
575	if (so->so_emu) {
576	if (tcp_emu(pData, so,m)) sbappend(pData, so, m);
577	} else
578	sbappend(pData, so, m);
579
580	/*
581	* XXX This is called when data arrives. Later, check
582	* if we can actually write() to the socket
583	* XXX Need to check? It's be NON_BLOCKING
584	*/
585	/* sorwakeup(so); */
586
587	/*
588	* If this is a short packet, then ACK now - with Nagel
589	* congestion avoidance sender won't send more until
590	* he gets an ACK.
591	*
592	* It is better to not delay acks at all to maximize
593	* TCP throughput. See RFC 2581.
594	*/
595	tp->t_flags \|= TF_ACKNOW;
596	tcp_output(pData, tp);
597	return;
598	}
599	} /* header prediction */
600	/*
601	* Calculate amount of space in receive window,
602	* and then do TCP input processing.
603	* Receive window is amount of space in rcv queue,
604	* but not less than advertised window.
605	*/
606	{ int win;
607	win = sbspace(&so->so_rcv);
608	if (win < 0)
609	win = 0;
610	tp->rcv_wnd = max(win, (int)(tp->rcv_adv - tp->rcv_nxt));
611	}
612
613	switch (tp->t_state) {
614
615	/*
616	* If the state is LISTEN then ignore segment if it contains an RST.
617	* If the segment contains an ACK then it is bad and send a RST.
618	* If it does not contain a SYN then it is not interesting; drop it.
619	* Don't bother responding if the destination was a broadcast.
620	* Otherwise initialize tp->rcv_nxt, and tp->irs, select an initial
621	* tp->iss, and send a segment:
622	* <SEQ=ISS><ACK=RCV_NXT><CTL=SYN,ACK>
623	* Also initialize tp->snd_nxt to tp->iss+1 and tp->snd_una to tp->iss.
624	* Fill in remote peer address fields if not previously specified.
625	* Enter SYN_RECEIVED state, and process any other fields of this
626	* segment in this state.
627	*/
628	case TCPS_LISTEN: {
629
630	if (tiflags & TH_RST)
631	goto drop;
632	if (tiflags & TH_ACK)
633	goto dropwithreset;
634	if ((tiflags & TH_SYN) == 0)
635	goto drop;
636
637	/*
638	* This has way too many gotos...
639	* But a bit of spaghetti code never hurt anybody :)
640	*/
641
642	/*
643	* If this is destined for the control address, then flag to
644	* tcp_ctl once connected, otherwise connect
645	*/
646	if ((so->so_faddr.s_addr&htonl(pData->netmask)) == special_addr.s_addr) {
647	int lastbyte=ntohl(so->so_faddr.s_addr) & ~pData->netmask;
648	if (lastbyte!=CTL_ALIAS && lastbyte!=CTL_DNS) {
649	#if 0
650	if(lastbyte==CTL_CMD \|\| lastbyte==CTL_EXEC) {
651	/* Command or exec adress */
652	so->so_state \|= SS_CTL;
653	} else
654	#endif
655	{
656	/* May be an add exec */
657	struct ex_list *ex_ptr;
658	for(ex_ptr = exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next) {
659	if(ex_ptr->ex_fport == so->so_fport &&
660	lastbyte == ex_ptr->ex_addr) {
661	so->so_state \|= SS_CTL;
662	break;
663	}
664	}
665	}
666	if(so->so_state & SS_CTL) goto cont_input;
667	}
668	/* CTL_ALIAS: Do nothing, tcp_fconnect will be called on it */
669	}
670
671	if (so->so_emu & EMU_NOCONNECT) {
672	so->so_emu &= ~EMU_NOCONNECT;
673	goto cont_input;
674	}
675
676	if((tcp_fconnect(pData, so) == -1) && (errno != EINPROGRESS) && (errno != EWOULDBLOCK)) {
677	u_char code=ICMP_UNREACH_NET;
678	DEBUG_MISC((dfd," tcp fconnect errno = %d-%s\n",
679	errno,strerror(errno)));
680	if(errno == ECONNREFUSED) {
681	/* ACK the SYN, send RST to refuse the connection */
682	tcp_respond(pData, tp, ti, m, ti->ti_seq+1, (tcp_seq)0,
683	TH_RST\|TH_ACK);
684	} else {
685	if(errno == EHOSTUNREACH) code=ICMP_UNREACH_HOST;
686	HTONL(ti->ti_seq); /* restore tcp header */
687	HTONL(ti->ti_ack);
688	HTONS(ti->ti_win);
689	HTONS(ti->ti_urp);
690	m->m_data -= sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
691	m->m_len += sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
692	*ip=save_ip;
693	icmp_error(pData, m, ICMP_UNREACH,code, 0,strerror(errno));
694	}
695	tp = tcp_close(pData, tp);
696	m_free(pData, m);
697	} else {
698	/*
699	* Haven't connected yet, save the current mbuf
700	* and ti, and return
701	* XXX Some OS's don't tell us whether the connect()
702	* succeeded or not. So we must time it out.
703	*/
704	so->so_m = m;
705	so->so_ti = ti;
706	tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT;
707	tp->t_state = TCPS_SYN_RECEIVED;
708	}
709	return;
710
711	cont_conn:
712	/* m==NULL
713	* Check if the connect succeeded
714	*/
715	if (so->so_state & SS_NOFDREF) {
716	tp = tcp_close(pData, tp);
717	goto dropwithreset;
718	}
719	cont_input:
720	tcp_template(tp);
721
722	if (optp)
723	tcp_dooptions(pData, tp, (u_char *)optp, optlen, ti);
724	/* , */
725	/* &ts_present, &ts_val, &ts_ecr); */
726
727	if (iss)
728	tp->iss = iss;
729	else
730	tp->iss = tcp_iss;
731	tcp_iss += TCP_ISSINCR/2;
732	tp->irs = ti->ti_seq;
733	tcp_sendseqinit(tp);
734	tcp_rcvseqinit(tp);
735	tp->t_flags \|= TF_ACKNOW;
736	tp->t_state = TCPS_SYN_RECEIVED;
737	tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT;
738	tcpstat.tcps_accepts++;
739	goto trimthenstep6;
740	} /* case TCPS_LISTEN */
741
742	/*
743	* If the state is SYN_SENT:
744	* if seg contains an ACK, but not for our SYN, drop the input.
745	* if seg contains a RST, then drop the connection.
746	* if seg does not contain SYN, then drop it.
747	* Otherwise this is an acceptable SYN segment
748	* initialize tp->rcv_nxt and tp->irs
749	* if seg contains ack then advance tp->snd_una
750	* if SYN has been acked change to ESTABLISHED else SYN_RCVD state
751	* arrange for segment to be acked (eventually)
752	* continue processing rest of data/controls, beginning with URG
753	*/
754	case TCPS_SYN_SENT:
755	if ((tiflags & TH_ACK) &&
756	(SEQ_LEQ(ti->ti_ack, tp->iss) \|\|
757	SEQ_GT(ti->ti_ack, tp->snd_max)))
758	goto dropwithreset;
759
760	if (tiflags & TH_RST) {
761	if (tiflags & TH_ACK)
762	tp = tcp_drop(pData, tp,0); /* XXX Check t_softerror! */
763	goto drop;
764	}
765
766	if ((tiflags & TH_SYN) == 0)
767	goto drop;
768	if (tiflags & TH_ACK) {
769	tp->snd_una = ti->ti_ack;
770	if (SEQ_LT(tp->snd_nxt, tp->snd_una))
771	tp->snd_nxt = tp->snd_una;
772	}
773
774	tp->t_timer[TCPT_REXMT] = 0;
775	tp->irs = ti->ti_seq;
776	tcp_rcvseqinit(tp);
777	tp->t_flags \|= TF_ACKNOW;
778	if (tiflags & TH_ACK && SEQ_GT(tp->snd_una, tp->iss)) {
779	tcpstat.tcps_connects++;
780	soisfconnected(so);
781	tp->t_state = TCPS_ESTABLISHED;
782
783	/* Do window scaling on this connection? */
784	/* if ((tp->t_flags & (TF_RCVD_SCALE\|TF_REQ_SCALE)) ==
785	* (TF_RCVD_SCALE\|TF_REQ_SCALE)) {
786	* tp->snd_scale = tp->requested_s_scale;
787	* tp->rcv_scale = tp->request_r_scale;
788	* }
789	*/
790	(void) tcp_reass(pData, tp, (struct tcpiphdr *)0,
791	(struct mbuf *)0);
792	/*
793	* if we didn't have to retransmit the SYN,
794	* use its rtt as our initial srtt & rtt var.
795	*/
796	if (tp->t_rtt)
797	tcp_xmit_timer(pData, tp, tp->t_rtt);
798	} else
799	tp->t_state = TCPS_SYN_RECEIVED;
800
801	trimthenstep6:
802	/*
803	* Advance ti->ti_seq to correspond to first data byte.
804	* If data, trim to stay within window,
805	* dropping FIN if necessary.
806	*/
807	ti->ti_seq++;
808	if (ti->ti_len > tp->rcv_wnd) {
809	todrop = ti->ti_len - tp->rcv_wnd;
810	m_adj(m, -todrop);
811	ti->ti_len = tp->rcv_wnd;
812	tiflags &= ~TH_FIN;
813	tcpstat.tcps_rcvpackafterwin++;
814	tcpstat.tcps_rcvbyteafterwin += todrop;
815	}
816	tp->snd_wl1 = ti->ti_seq - 1;
817	tp->rcv_up = ti->ti_seq;
818	goto step6;
819	} /* switch tp->t_state */
820	/*
821	* States other than LISTEN or SYN_SENT.
822	* First check timestamp, if present.
823	* Then check that at least some bytes of segment are within
824	* receive window. If segment begins before rcv_nxt,
825	* drop leading data (and SYN); if nothing left, just ack.
826	*
827	* RFC 1323 PAWS: If we have a timestamp reply on this segment
828	* and it's less than ts_recent, drop it.
829	*/
830	/* if (ts_present && (tiflags & TH_RST) == 0 && tp->ts_recent &&
831	* TSTMP_LT(ts_val, tp->ts_recent)) {
832	*
833	/ / Check to see if ts_recent is over 24 days old. */
834	/* if ((int)(tcp_now - tp->ts_recent_age) > TCP_PAWS_IDLE) {
835	/ /
836	* * Invalidate ts_recent. If this segment updates
837	* * ts_recent, the age will be reset later and ts_recent
838	* * will get a valid value. If it does not, setting
839	* * ts_recent to zero will at least satisfy the
840	* * requirement that zero be placed in the timestamp
841	* * echo reply when ts_recent isn't valid. The
842	* * age isn't reset until we get a valid ts_recent
843	* * because we don't want out-of-order segments to be
844	* * dropped when ts_recent is old.
845	* */
846	/* tp->ts_recent = 0;
847	* } else {
848	* tcpstat.tcps_rcvduppack++;
849	* tcpstat.tcps_rcvdupbyte += ti->ti_len;
850	* tcpstat.tcps_pawsdrop++;
851	* goto dropafterack;
852	* }
853	* }
854	*/
855
856	todrop = tp->rcv_nxt - ti->ti_seq;
857	if (todrop > 0) {
858	if (tiflags & TH_SYN) {
859	tiflags &= ~TH_SYN;
860	ti->ti_seq++;
861	if (ti->ti_urp > 1)
862	ti->ti_urp--;
863	else
864	tiflags &= ~TH_URG;
865	todrop--;
866	}
867	/*
868	* Following if statement from Stevens, vol. 2, p. 960.
869	*/
870	if (todrop > ti->ti_len
871	\|\| (todrop == ti->ti_len && (tiflags & TH_FIN) == 0)) {
872	/*
873	* Any valid FIN must be to the left of the window.
874	* At this point the FIN must be a duplicate or out
875	* of sequence; drop it.
876	*/
877	tiflags &= ~TH_FIN;
878
879	/*
880	* Send an ACK to resynchronize and drop any data.
881	* But keep on processing for RST or ACK.
882	*/
883	tp->t_flags \|= TF_ACKNOW;
884	todrop = ti->ti_len;
885	tcpstat.tcps_rcvduppack++;
886	tcpstat.tcps_rcvdupbyte += todrop;
887	} else {
888	tcpstat.tcps_rcvpartduppack++;
889	tcpstat.tcps_rcvpartdupbyte += todrop;
890	}
891	m_adj(m, todrop);
892	ti->ti_seq += todrop;
893	ti->ti_len -= todrop;
894	if (ti->ti_urp > todrop)
895	ti->ti_urp -= todrop;
896	else {
897	tiflags &= ~TH_URG;
898	ti->ti_urp = 0;
899	}
900	}
901	/*
902	* If new data are received on a connection after the
903	* user processes are gone, then RST the other end.
904	*/
905	if ((so->so_state & SS_NOFDREF) &&
906	tp->t_state > TCPS_CLOSE_WAIT && ti->ti_len) {
907	tp = tcp_close(pData, tp);
908	tcpstat.tcps_rcvafterclose++;
909	goto dropwithreset;
910	}
911
912	/*
913	* If segment ends after window, drop trailing data
914	* (and PUSH and FIN); if nothing left, just ACK.
915	*/
916	todrop = (ti->ti_seq+ti->ti_len) - (tp->rcv_nxt+tp->rcv_wnd);
917	if (todrop > 0) {
918	tcpstat.tcps_rcvpackafterwin++;
919	if (todrop >= ti->ti_len) {
920	tcpstat.tcps_rcvbyteafterwin += ti->ti_len;
921	/*
922	* If a new connection request is received
923	* while in TIME_WAIT, drop the old connection
924	* and start over if the sequence numbers
925	* are above the previous ones.
926	*/
927	if (tiflags & TH_SYN &&
928	tp->t_state == TCPS_TIME_WAIT &&
929	SEQ_GT(ti->ti_seq, tp->rcv_nxt)) {
930	iss = tp->rcv_nxt + TCP_ISSINCR;
931	tp = tcp_close(pData, tp);
932	goto findso;
933	}
934	/*
935	* If window is closed can only take segments at
936	* window edge, and have to drop data and PUSH from
937	* incoming segments. Continue processing, but
938	* remember to ack. Otherwise, drop segment
939	* and ack.
940	*/
941	if (tp->rcv_wnd == 0 && ti->ti_seq == tp->rcv_nxt) {
942	tp->t_flags \|= TF_ACKNOW;
943	tcpstat.tcps_rcvwinprobe++;
944	} else
945	goto dropafterack;
946	} else
947	tcpstat.tcps_rcvbyteafterwin += todrop;
948	m_adj(m, -todrop);
949	ti->ti_len -= todrop;
950	tiflags &= ~(TH_PUSH\|TH_FIN);
951	}
952
953	/*
954	* If last ACK falls within this segment's sequence numbers,
955	* record its timestamp.
956	*/
957	/* if (ts_present && SEQ_LEQ(ti->ti_seq, tp->last_ack_sent) &&
958	* SEQ_LT(tp->last_ack_sent, ti->ti_seq + ti->ti_len +
959	* ((tiflags & (TH_SYN\|TH_FIN)) != 0))) {
960	* tp->ts_recent_age = tcp_now;
961	* tp->ts_recent = ts_val;
962	* }
963	*/
964
965	/*
966	* If the RST bit is set examine the state:
967	* SYN_RECEIVED STATE:
968	* If passive open, return to LISTEN state.
969	* If active open, inform user that connection was refused.
970	* ESTABLISHED, FIN_WAIT_1, FIN_WAIT2, CLOSE_WAIT STATES:
971	* Inform user that connection was reset, and close tcb.
972	* CLOSING, LAST_ACK, TIME_WAIT STATES
973	* Close the tcb.
974	*/
975	if (tiflags&TH_RST) switch (tp->t_state) {
976
977	case TCPS_SYN_RECEIVED:
978	/* so->so_error = ECONNREFUSED; */
979	goto close;
980
981	case TCPS_ESTABLISHED:
982	case TCPS_FIN_WAIT_1:
983	case TCPS_FIN_WAIT_2:
984	case TCPS_CLOSE_WAIT:
985	/* so->so_error = ECONNRESET; */
986	close:
987	tp->t_state = TCPS_CLOSED;
988	tcpstat.tcps_drops++;
989	tp = tcp_close(pData, tp);
990	goto drop;
991
992	case TCPS_CLOSING:
993	case TCPS_LAST_ACK:
994	case TCPS_TIME_WAIT:
995	tp = tcp_close(pData, tp);
996	goto drop;
997	}
998
999	/*
1000	* If a SYN is in the window, then this is an
1001	* error and we send an RST and drop the connection.
1002	*/
1003	if (tiflags & TH_SYN) {
1004	tp = tcp_drop(pData, tp,0);
1005	goto dropwithreset;
1006	}
1007
1008	/*
1009	* If the ACK bit is off we drop the segment and return.
1010	*/
1011	if ((tiflags & TH_ACK) == 0) goto drop;
1012
1013	/*
1014	* Ack processing.
1015	*/
1016	switch (tp->t_state) {
1017	/*
1018	* In SYN_RECEIVED state if the ack ACKs our SYN then enter
1019	* ESTABLISHED state and continue processing, otherwise
1020	* send an RST. una<=ack<=max
1021	*/
1022	case TCPS_SYN_RECEIVED:
1023
1024	if (SEQ_GT(tp->snd_una, ti->ti_ack) \|\|
1025	SEQ_GT(ti->ti_ack, tp->snd_max))
1026	goto dropwithreset;
1027	tcpstat.tcps_connects++;
1028	tp->t_state = TCPS_ESTABLISHED;
1029	/*
1030	* The sent SYN is ack'ed with our sequence number +1
1031	* The first data byte already in the buffer will get
1032	* lost if no correction is made. This is only needed for
1033	* SS_CTL since the buffer is empty otherwise.
1034	* tp->snd_una++; or:
1035	*/
1036	tp->snd_una=ti->ti_ack;
1037	if (so->so_state & SS_CTL) {
1038	/* So tcp_ctl reports the right state */
1039	ret = tcp_ctl(pData, so);
1040	if (ret == 1) {
1041	soisfconnected(so);
1042	so->so_state &= ~SS_CTL; /* success XXX */
1043	} else if (ret == 2) {
1044	so->so_state = SS_NOFDREF; /* CTL_CMD */
1045	} else {
1046	needoutput = 1;
1047	tp->t_state = TCPS_FIN_WAIT_1;
1048	}
1049	} else {
1050	soisfconnected(so);
1051	}
1052
1053	/* Do window scaling? */
1054	/* if ((tp->t_flags & (TF_RCVD_SCALE\|TF_REQ_SCALE)) ==
1055	* (TF_RCVD_SCALE\|TF_REQ_SCALE)) {
1056	* tp->snd_scale = tp->requested_s_scale;
1057	* tp->rcv_scale = tp->request_r_scale;
1058	* }
1059	*/
1060	(void) tcp_reass(pData, tp, (struct tcpiphdr )0, (struct mbuf )0);
1061	tp->snd_wl1 = ti->ti_seq - 1;
1062	/* Avoid ack processing; snd_una==ti_ack => dup ack */
1063	goto synrx_to_est;
1064	/* fall into ... */
1065
1066	/*
1067	* In ESTABLISHED state: drop duplicate ACKs; ACK out of range
1068	* ACKs. If the ack is in the range
1069	* tp->snd_una < ti->ti_ack <= tp->snd_max
1070	* then advance tp->snd_una to ti->ti_ack and drop
1071	* data from the retransmission queue. If this ACK reflects
1072	* more up to date window information we update our window information.
1073	*/
1074	case TCPS_ESTABLISHED:
1075	case TCPS_FIN_WAIT_1:
1076	case TCPS_FIN_WAIT_2:
1077	case TCPS_CLOSE_WAIT:
1078	case TCPS_CLOSING:
1079	case TCPS_LAST_ACK:
1080	case TCPS_TIME_WAIT:
1081
1082	if (SEQ_LEQ(ti->ti_ack, tp->snd_una)) {
1083	if (ti->ti_len == 0 && tiwin == tp->snd_wnd) {
1084	tcpstat.tcps_rcvdupack++;
1085	DEBUG_MISC((dfd," dup ack m = %lx so = %lx \n",
1086	(long )m, (long )so));
1087	/*
1088	* If we have outstanding data (other than
1089	* a window probe), this is a completely
1090	* duplicate ack (ie, window info didn't
1091	* change), the ack is the biggest we've
1092	* seen and we've seen exactly our rexmt
1093	* threshold of them, assume a packet
1094	* has been dropped and retransmit it.
1095	* Kludge snd_nxt & the congestion
1096	* window so we send only this one
1097	* packet.
1098	*
1099	* We know we're losing at the current
1100	* window size so do congestion avoidance
1101	* (set ssthresh to half the current window
1102	* and pull our congestion window back to
1103	* the new ssthresh).
1104	*
1105	* Dup acks mean that packets have left the
1106	* network (they're now cached at the receiver)
1107	* so bump cwnd by the amount in the receiver
1108	* to keep a constant cwnd packets in the
1109	* network.
1110	*/
1111	if (tp->t_timer[TCPT_REXMT] == 0 \|\|
1112	ti->ti_ack != tp->snd_una)
1113	tp->t_dupacks = 0;
1114	else if (++tp->t_dupacks == tcprexmtthresh) {
1115	tcp_seq onxt = tp->snd_nxt;
1116	u_int win =
1117	min(tp->snd_wnd, tp->snd_cwnd) / 2 /
1118	tp->t_maxseg;
1119
1120	if (win < 2)
1121	win = 2;
1122	tp->snd_ssthresh = win * tp->t_maxseg;
1123	tp->t_timer[TCPT_REXMT] = 0;
1124	tp->t_rtt = 0;
1125	tp->snd_nxt = ti->ti_ack;
1126	tp->snd_cwnd = tp->t_maxseg;
1127	(void) tcp_output(pData, tp);
1128	tp->snd_cwnd = tp->snd_ssthresh +
1129	tp->t_maxseg * tp->t_dupacks;
1130	if (SEQ_GT(onxt, tp->snd_nxt))
1131	tp->snd_nxt = onxt;
1132	goto drop;
1133	} else if (tp->t_dupacks > tcprexmtthresh) {
1134	tp->snd_cwnd += tp->t_maxseg;
1135	(void) tcp_output(pData, tp);
1136	goto drop;
1137	}
1138	} else
1139	tp->t_dupacks = 0;
1140	break;
1141	}
1142	synrx_to_est:
1143	/*
1144	* If the congestion window was inflated to account
1145	* for the other side's cached packets, retract it.
1146	*/
1147	if (tp->t_dupacks > tcprexmtthresh &&
1148	tp->snd_cwnd > tp->snd_ssthresh)
1149	tp->snd_cwnd = tp->snd_ssthresh;
1150	tp->t_dupacks = 0;
1151	if (SEQ_GT(ti->ti_ack, tp->snd_max)) {
1152	tcpstat.tcps_rcvacktoomuch++;
1153	goto dropafterack;
1154	}
1155	acked = ti->ti_ack - tp->snd_una;
1156	tcpstat.tcps_rcvackpack++;
1157	tcpstat.tcps_rcvackbyte += acked;
1158
1159	/*
1160	* If we have a timestamp reply, update smoothed
1161	* round trip time. If no timestamp is present but
1162	* transmit timer is running and timed sequence
1163	* number was acked, update smoothed round trip time.
1164	* Since we now have an rtt measurement, cancel the
1165	* timer backoff (cf., Phil Karn's retransmit alg.).
1166	* Recompute the initial retransmit timer.
1167	*/
1168	/* if (ts_present)
1169	* tcp_xmit_timer(tp, tcp_now-ts_ecr+1);
1170	* else
1171	*/
1172	if (tp->t_rtt && SEQ_GT(ti->ti_ack, tp->t_rtseq))
1173	tcp_xmit_timer(pData, tp,tp->t_rtt);
1174
1175	/*
1176	* If all outstanding data is acked, stop retransmit
1177	* timer and remember to restart (more output or persist).
1178	* If there is more data to be acked, restart retransmit
1179	* timer, using current (possibly backed-off) value.
1180	*/
1181	if (ti->ti_ack == tp->snd_max) {
1182	tp->t_timer[TCPT_REXMT] = 0;
1183	needoutput = 1;
1184	} else if (tp->t_timer[TCPT_PERSIST] == 0)
1185	tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
1186	/*
1187	* When new data is acked, open the congestion window.
1188	* If the window gives us less than ssthresh packets
1189	* in flight, open exponentially (maxseg per packet).
1190	* Otherwise open linearly: maxseg per window
1191	* (maxseg^2 / cwnd per packet).
1192	*/
1193	{
1194	register u_int cw = tp->snd_cwnd;
1195	register u_int incr = tp->t_maxseg;
1196
1197	if (cw > tp->snd_ssthresh)
1198	incr = incr * incr / cw;
1199	tp->snd_cwnd = min(cw + incr, TCP_MAXWIN<<tp->snd_scale);
1200	}
1201	if (acked > so->so_snd.sb_cc) {
1202	tp->snd_wnd -= so->so_snd.sb_cc;
1203	sbdrop(&so->so_snd, (int )so->so_snd.sb_cc);
1204	ourfinisacked = 1;
1205	} else {
1206	sbdrop(&so->so_snd, acked);
1207	tp->snd_wnd -= acked;
1208	ourfinisacked = 0;
1209	}
1210	/*
1211	* XXX sowwakup is called when data is acked and there's room for
1212	* for more data... it should read() the socket
1213	*/
1214	/* if (so->so_snd.sb_flags & SB_NOTIFY)
1215	* sowwakeup(so);
1216	*/
1217	tp->snd_una = ti->ti_ack;
1218	if (SEQ_LT(tp->snd_nxt, tp->snd_una))
1219	tp->snd_nxt = tp->snd_una;
1220
1221	switch (tp->t_state) {
1222
1223	/*
1224	* In FIN_WAIT_1 STATE in addition to the processing
1225	* for the ESTABLISHED state if our FIN is now acknowledged
1226	* then enter FIN_WAIT_2.
1227	*/
1228	case TCPS_FIN_WAIT_1:
1229	if (ourfinisacked) {
1230	/*
1231	* If we can't receive any more
1232	* data, then closing user can proceed.
1233	* Starting the timer is contrary to the
1234	* specification, but if we don't get a FIN
1235	* we'll hang forever.
1236	*/
1237	if (so->so_state & SS_FCANTRCVMORE) {
1238	soisfdisconnected(so);
1239	tp->t_timer[TCPT_2MSL] = tcp_maxidle;
1240	}
1241	tp->t_state = TCPS_FIN_WAIT_2;
1242	}
1243	break;
1244
1245	/*
1246	* In CLOSING STATE in addition to the processing for
1247	* the ESTABLISHED state if the ACK acknowledges our FIN
1248	* then enter the TIME-WAIT state, otherwise ignore
1249	* the segment.
1250	*/
1251	case TCPS_CLOSING:
1252	if (ourfinisacked) {
1253	tp->t_state = TCPS_TIME_WAIT;
1254	tcp_canceltimers(tp);
1255	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1256	soisfdisconnected(so);
1257	}
1258	break;
1259
1260	/*
1261	* In LAST_ACK, we may still be waiting for data to drain
1262	* and/or to be acked, as well as for the ack of our FIN.
1263	* If our FIN is now acknowledged, delete the TCB,
1264	* enter the closed state and return.
1265	*/
1266	case TCPS_LAST_ACK:
1267	if (ourfinisacked) {
1268	tp = tcp_close(pData, tp);
1269	goto drop;
1270	}
1271	break;
1272
1273	/*
1274	* In TIME_WAIT state the only thing that should arrive
1275	* is a retransmission of the remote FIN. Acknowledge
1276	* it and restart the finack timer.
1277	*/
1278	case TCPS_TIME_WAIT:
1279	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1280	goto dropafterack;
1281	}
1282	} /* switch(tp->t_state) */
1283
1284	step6:
1285	/*
1286	* Update window information.
1287	* Don't look at window if no ACK: TAC's send garbage on first SYN.
1288	*/
1289	if ((tiflags & TH_ACK) &&
1290	(SEQ_LT(tp->snd_wl1, ti->ti_seq) \|\|
1291	(tp->snd_wl1 == ti->ti_seq && (SEQ_LT(tp->snd_wl2, ti->ti_ack) \|\|
1292	(tp->snd_wl2 == ti->ti_ack && tiwin > tp->snd_wnd))))) {
1293	/* keep track of pure window updates */
1294	if (ti->ti_len == 0 &&
1295	tp->snd_wl2 == ti->ti_ack && tiwin > tp->snd_wnd)
1296	tcpstat.tcps_rcvwinupd++;
1297	tp->snd_wnd = tiwin;
1298	tp->snd_wl1 = ti->ti_seq;
1299	tp->snd_wl2 = ti->ti_ack;
1300	if (tp->snd_wnd > tp->max_sndwnd)
1301	tp->max_sndwnd = tp->snd_wnd;
1302	needoutput = 1;
1303	}
1304
1305	/*
1306	* Process segments with URG.
1307	*/
1308	if ((tiflags & TH_URG) && ti->ti_urp &&
1309	TCPS_HAVERCVDFIN(tp->t_state) == 0) {
1310	/*
1311	* This is a kludge, but if we receive and accept
1312	* random urgent pointers, we'll crash in
1313	* soreceive. It's hard to imagine someone
1314	* actually wanting to send this much urgent data.
1315	*/
1316	if (ti->ti_urp + so->so_rcv.sb_cc > so->so_rcv.sb_datalen) {
1317	ti->ti_urp = 0;
1318	tiflags &= ~TH_URG;
1319	goto dodata;
1320	}
1321	/*
1322	* If this segment advances the known urgent pointer,
1323	* then mark the data stream. This should not happen
1324	* in CLOSE_WAIT, CLOSING, LAST_ACK or TIME_WAIT STATES since
1325	* a FIN has been received from the remote side.
1326	* In these states we ignore the URG.
1327	*
1328	* According to RFC961 (Assigned Protocols),
1329	* the urgent pointer points to the last octet
1330	* of urgent data. We continue, however,
1331	* to consider it to indicate the first octet
1332	* of data past the urgent section as the original
1333	* spec states (in one of two places).
1334	*/
1335	if (SEQ_GT(ti->ti_seq+ti->ti_urp, tp->rcv_up)) {
1336	tp->rcv_up = ti->ti_seq + ti->ti_urp;
1337	so->so_urgc = so->so_rcv.sb_cc +
1338	(tp->rcv_up - tp->rcv_nxt); /* -1; */
1339	tp->rcv_up = ti->ti_seq + ti->ti_urp;
1340
1341	}
1342	} else
1343	/*
1344	* If no out of band data is expected,
1345	* pull receive urgent pointer along
1346	* with the receive window.
1347	*/
1348	if (SEQ_GT(tp->rcv_nxt, tp->rcv_up))
1349	tp->rcv_up = tp->rcv_nxt;
1350	dodata:
1351
1352	/*
1353	* Process the segment text, merging it into the TCP sequencing queue,
1354	* and arranging for acknowledgment of receipt if necessary.
1355	* This process logically involves adjusting tp->rcv_wnd as data
1356	* is presented to the user (this happens in tcp_usrreq.c,
1357	* case PRU_RCVD). If a FIN has already been received on this
1358	* connection then we just ignore the text.
1359	*/
1360	if ((ti->ti_len \|\| (tiflags&TH_FIN)) &&
1361	TCPS_HAVERCVDFIN(tp->t_state) == 0) {
1362	TCP_REASS(pData, tp, ti, m, so, tiflags);
1363	/*
1364	* Note the amount of data that peer has sent into
1365	* our window, in order to estimate the sender's
1366	* buffer size.
1367	*/
1368	len = so->so_rcv.sb_datalen - (tp->rcv_adv - tp->rcv_nxt);
1369	} else {
1370	mbuf_freed = 1; /* The mbuf must be freed, but only when its content is not needed anymore. */
1371	tiflags &= ~TH_FIN;
1372	}
1373
1374	/*
1375	* If FIN is received ACK the FIN and let the user know
1376	* that the connection is closing.
1377	*/
1378	if (tiflags & TH_FIN) {
1379	if (TCPS_HAVERCVDFIN(tp->t_state) == 0) {
1380	/*
1381	* If we receive a FIN we can't send more data,
1382	* set it SS_FDRAIN
1383	* Shutdown the socket if there is no rx data in the
1384	* buffer.
1385	* soread() is called on completion of shutdown() and
1386	* will got to TCPS_LAST_ACK, and use tcp_output()
1387	* to send the FIN.
1388	*/
1389	/* sofcantrcvmore(so); */
1390	sofwdrain(so);
1391
1392	tp->t_flags \|= TF_ACKNOW;
1393	tp->rcv_nxt++;
1394	}
1395	switch (tp->t_state) {
1396
1397	/*
1398	* In SYN_RECEIVED and ESTABLISHED STATES
1399	* enter the CLOSE_WAIT state.
1400	*/
1401	case TCPS_SYN_RECEIVED:
1402	case TCPS_ESTABLISHED:
1403	if(so->so_emu == EMU_CTL) /* no shutdown on socket */
1404	tp->t_state = TCPS_LAST_ACK;
1405	else
1406	tp->t_state = TCPS_CLOSE_WAIT;
1407	break;
1408
1409	/*
1410	* If still in FIN_WAIT_1 STATE FIN has not been acked so
1411	* enter the CLOSING state.
1412	*/
1413	case TCPS_FIN_WAIT_1:
1414	tp->t_state = TCPS_CLOSING;
1415	break;
1416
1417	/*
1418	* In FIN_WAIT_2 state enter the TIME_WAIT state,
1419	* starting the time-wait timer, turning off the other
1420	* standard timers.
1421	*/
1422	case TCPS_FIN_WAIT_2:
1423	tp->t_state = TCPS_TIME_WAIT;
1424	tcp_canceltimers(tp);
1425	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1426	soisfdisconnected(so);
1427	break;
1428
1429	/*
1430	* In TIME_WAIT state restart the 2 MSL time_wait timer.
1431	*/
1432	case TCPS_TIME_WAIT:
1433	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1434	break;
1435	}
1436	}
1437
1438	/*
1439	* If this is a small packet, then ACK now - with Nagel
1440	* congestion avoidance sender won't send more until
1441	* he gets an ACK.
1442	*
1443	* See above.
1444	*/
1445	/* if (ti->ti_len && (unsigned)ti->ti_len < tp->t_maxseg) {
1446	*/
1447	/* if ((ti->ti_len && (unsigned)ti->ti_len < tp->t_maxseg &&
1448	* (so->so_iptos & IPTOS_LOWDELAY) == 0) \|\|
1449	* ((so->so_iptos & IPTOS_LOWDELAY) &&
1450	* ((struct tcpiphdr_2 *)ti)->first_char == (char)27)) {
1451	*/
1452	if (ti->ti_len && (unsigned)ti->ti_len <= 5 &&
1453	((struct tcpiphdr_2 *)ti)->first_char == (char)27) {
1454	tp->t_flags \|= TF_ACKNOW;
1455	}
1456
1457	if (mbuf_freed) {
1458	m_free(pData, m);
1459	}
1460	/*
1461	* Return any desired output.
1462	*/
1463	if (needoutput \|\| (tp->t_flags & TF_ACKNOW)) {
1464	(void) tcp_output(pData, tp);
1465	}
1466
1467	return;
1468
1469	dropafterack:
1470	/*
1471	* Generate an ACK dropping incoming segment if it occupies
1472	* sequence space, where the ACK reflects our state.
1473	*/
1474	if (tiflags & TH_RST)
1475	goto drop;
1476	m_freem(pData, m);
1477	tp->t_flags \|= TF_ACKNOW;
1478	(void) tcp_output(pData, tp);
1479	return;
1480
1481	dropwithreset:
1482	/* reuses m if m!=NULL, m_free() unnecessary */
1483	if (tiflags & TH_ACK)
1484	tcp_respond(pData, tp, ti, m, (tcp_seq)0, ti->ti_ack, TH_RST);
1485	else {
1486	if (tiflags & TH_SYN) ti->ti_len++;
1487	tcp_respond(pData, tp, ti, m, ti->ti_seq+ti->ti_len, (tcp_seq)0,
1488	TH_RST\|TH_ACK);
1489	}
1490
1491	return;
1492
1493	drop:
1494	/*
1495	* Drop space held by incoming segment and return.
1496	*/
1497	m_free(pData, m);
1498	return;
1499	#ifdef VBOX_WITH_SYNC_SLIRP
1500	#undef return
1501	#endif
1502	}
1503
1504	/* , ts_present, ts_val, ts_ecr) */
1505	/* int *ts_present;
1506	* u_int32_t ts_val, ts_ecr;
1507	*/
1508	void
1509	tcp_dooptions(PNATState pData, struct tcpcb tp, u_char cp, int cnt, struct tcpiphdr *ti)
1510	{
1511	u_int16_t mss;
1512	int opt, optlen;
1513
1514	DEBUG_CALL("tcp_dooptions");
1515	DEBUG_ARGS((dfd," tp = %lx cnt=%i \n", (long )tp, cnt));
1516
1517	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1518	opt = cp[0];
1519	if (opt == TCPOPT_EOL)
1520	break;
1521	if (opt == TCPOPT_NOP)
1522	optlen = 1;
1523	else {
1524	optlen = cp[1];
1525	if (optlen <= 0)
1526	break;
1527	}
1528	switch (opt) {
1529
1530	default:
1531	continue;
1532
1533	case TCPOPT_MAXSEG:
1534	if (optlen != TCPOLEN_MAXSEG)
1535	continue;
1536	if (!(ti->ti_flags & TH_SYN))
1537	continue;
1538	memcpy((char ) &mss, (char ) cp + 2, sizeof(mss));
1539	NTOHS(mss);
1540	(void) tcp_mss(pData, tp, mss); /* sets t_maxseg */
1541	break;
1542
1543	/* case TCPOPT_WINDOW:
1544	* if (optlen != TCPOLEN_WINDOW)
1545	* continue;
1546	* if (!(ti->ti_flags & TH_SYN))
1547	* continue;
1548	* tp->t_flags \|= TF_RCVD_SCALE;
1549	* tp->requested_s_scale = min(cp[2], TCP_MAX_WINSHIFT);
1550	* break;
1551	*/
1552	/* case TCPOPT_TIMESTAMP:
1553	* if (optlen != TCPOLEN_TIMESTAMP)
1554	* continue;
1555	* *ts_present = 1;
1556	* memcpy((char ) ts_val, (char )cp + 2, sizeof(*ts_val));
1557	* NTOHL(*ts_val);
1558	* memcpy((char ) ts_ecr, (char )cp + 6, sizeof(*ts_ecr));
1559	* NTOHL(*ts_ecr);
1560	*
1561	/ /
1562	* * A timestamp received in a SYN makes
1563	* * it ok to send timestamp requests and replies.
1564	* */
1565	/* if (ti->ti_flags & TH_SYN) {
1566	* tp->t_flags \|= TF_RCVD_TSTMP;
1567	* tp->ts_recent = *ts_val;
1568	* tp->ts_recent_age = tcp_now;
1569	* }
1570	*/ break;
1571	}
1572	}
1573	}
1574
1575
1576	/*
1577	* Pull out of band byte out of a segment so
1578	* it doesn't appear in the user's data queue.
1579	* It is still reflected in the segment length for
1580	* sequencing purposes.
1581	*/
1582
1583	#ifdef notdef
1584
1585	void
1586	tcp_pulloutofband(so, ti, m)
1587	struct socket *so;
1588	struct tcpiphdr *ti;
1589	register struct mbuf *m;
1590	{
1591	int cnt = ti->ti_urp - 1;
1592
1593	while (cnt >= 0) {
1594	if (m->m_len > cnt) {
1595	char *cp = mtod(m, caddr_t) + cnt;
1596	struct tcpcb *tp = sototcpcb(so);
1597
1598	tp->t_iobc = *cp;
1599	tp->t_oobflags \|= TCPOOB_HAVEDATA;
1600	memcpy(sp, cp+1, (unsigned)(m->m_len - cnt - 1));
1601	m->m_len--;
1602	return;
1603	}
1604	cnt -= m->m_len;
1605	m = m->m_next; /* XXX WRONG! Fix it! */
1606	if (m == 0)
1607	break;
1608	}
1609	panic("tcp_pulloutofband");
1610	}
1611
1612	#endif /* notdef */
1613
1614	/*
1615	* Collect new round-trip time estimate
1616	* and update averages and current timeout.
1617	*/
1618
1619	void
1620	tcp_xmit_timer(PNATState pData, register struct tcpcb *tp, int rtt)
1621	{
1622	register short delta;
1623
1624	DEBUG_CALL("tcp_xmit_timer");
1625	DEBUG_ARG("tp = %lx", (long)tp);
1626	DEBUG_ARG("rtt = %d", rtt);
1627
1628	tcpstat.tcps_rttupdated++;
1629	if (tp->t_srtt != 0) {
1630	/*
1631	* srtt is stored as fixed point with 3 bits after the
1632	* binary point (i.e., scaled by 8). The following magic
1633	* is equivalent to the smoothing algorithm in rfc793 with
1634	* an alpha of .875 (srtt = rtt/8 + srtt*7/8 in fixed
1635	* point). Adjust rtt to origin 0.
1636	*/
1637	delta = rtt - 1 - (tp->t_srtt >> TCP_RTT_SHIFT);
1638	if ((tp->t_srtt += delta) <= 0)
1639	tp->t_srtt = 1;
1640	/*
1641	* We accumulate a smoothed rtt variance (actually, a
1642	* smoothed mean difference), then set the retransmit
1643	* timer to smoothed rtt + 4 times the smoothed variance.
1644	* rttvar is stored as fixed point with 2 bits after the
1645	* binary point (scaled by 4). The following is
1646	* equivalent to rfc793 smoothing with an alpha of .75
1647	* (rttvar = rttvar*3/4 + \|delta\| / 4). This replaces
1648	* rfc793's wired-in beta.
1649	*/
1650	if (delta < 0)
1651	delta = -delta;
1652	delta -= (tp->t_rttvar >> TCP_RTTVAR_SHIFT);
1653	if ((tp->t_rttvar += delta) <= 0)
1654	tp->t_rttvar = 1;
1655	} else {
1656	/*
1657	* No rtt measurement yet - use the unsmoothed rtt.
1658	* Set the variance to half the rtt (so our first
1659	* retransmit happens at 3*rtt).
1660	*/
1661	tp->t_srtt = rtt << TCP_RTT_SHIFT;
1662	tp->t_rttvar = rtt << (TCP_RTTVAR_SHIFT - 1);
1663	}
1664	tp->t_rtt = 0;
1665	tp->t_rxtshift = 0;
1666
1667	/*
1668	* the retransmit should happen at rtt + 4 * rttvar.
1669	* Because of the way we do the smoothing, srtt and rttvar
1670	* will each average +1/2 tick of bias. When we compute
1671	* the retransmit timer, we want 1/2 tick of rounding and
1672	* 1 extra tick because of +-1/2 tick uncertainty in the
1673	* firing of the timer. The bias will give us exactly the
1674	* 1.5 tick we need. But, because the bias is
1675	* statistical, we have to test that we don't drop below
1676	* the minimum feasible timer (which is 2 ticks).
1677	*/
1678	TCPT_RANGESET(tp->t_rxtcur, TCP_REXMTVAL(tp),
1679	(short)tp->t_rttmin, TCPTV_REXMTMAX); /* XXX */
1680
1681	/*
1682	* We received an ack for a packet that wasn't retransmitted;
1683	* it is probably safe to discard any error indications we've
1684	* received recently. This isn't quite right, but close enough
1685	* for now (a route might have failed after we sent a segment,
1686	* and the return path might not be symmetrical).
1687	*/
1688	tp->t_softerror = 0;
1689	}
1690
1691	/*
1692	* Determine a reasonable value for maxseg size.
1693	* If the route is known, check route for mtu.
1694	* If none, use an mss that can be handled on the outgoing
1695	* interface without forcing IP to fragment; if bigger than
1696	* an mbuf cluster (MCLBYTES), round down to nearest multiple of MCLBYTES
1697	* to utilize large mbufs. If no route is found, route has no mtu,
1698	* or the destination isn't local, use a default, hopefully conservative
1699	* size (usually 512 or the default IP max size, but no more than the mtu
1700	* of the interface), as we can't discover anything about intervening
1701	* gateways or networks. We also initialize the congestion/slow start
1702	* window to be a single segment if the destination isn't local.
1703	* While looking at the routing entry, we also initialize other path-dependent
1704	* parameters from pre-set or cached values in the routing entry.
1705	*/
1706
1707	int
1708	tcp_mss(PNATState pData, register struct tcpcb *tp, u_int offer)
1709	{
1710	struct socket *so = tp->t_socket;
1711	int mss;
1712
1713	DEBUG_CALL("tcp_mss");
1714	DEBUG_ARG("tp = %lx", (long)tp);
1715	DEBUG_ARG("offer = %d", offer);
1716
1717	mss = min(if_mtu, if_mru) - sizeof(struct tcpiphdr);
1718	if (offer)
1719	mss = min(mss, offer);
1720	mss = max(mss, 32);
1721	if (mss < tp->t_maxseg \|\| offer != 0)
1722	tp->t_maxseg = mss;
1723
1724	tp->snd_cwnd = mss;
1725
1726	sbreserve(&so->so_snd, tcp_sndspace+((tcp_sndspace%mss)?(mss-(tcp_sndspace%mss)):0));
1727	sbreserve(&so->so_rcv, tcp_rcvspace+((tcp_rcvspace%mss)?(mss-(tcp_rcvspace%mss)):0));
1728
1729	DEBUG_MISC((dfd, " returning mss = %d\n", mss));
1730
1731	return mss;
1732	}

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

source: vbox/trunk/src/VBox/Devices/Network/slirp/tcp_input.c@ 13958

Download in other formats: