tcp_input.c@ 14160

Last change on this file since 14160 was 14159, checked in by vboxsync, 16 years ago
registration several sockets per event, to avoid artificial barrier with WSA_MAXIMUM_WAIT_EVENTS (64). Currently registered 64 sockets per event. should work fine ;)
Property svn:eol-style set to `native`
File size: 48.5 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
249	/*
250	* If called with m == 0, then we're continuing the connect
251	*/
252	if (m == NULL) {
253	so = inso;
254
255	/* Re-set a few variables */
256	tp = sototcpcb(so);
257	m = so->so_m;
258	so->so_m = 0;
259	ti = so->so_ti;
260	tiwin = ti->ti_win;
261	tiflags = ti->ti_flags;
262
263	goto cont_conn;
264	}
265
266
267	tcpstat.tcps_rcvtotal++;
268	/*
269	* Get IP and TCP header together in first mbuf.
270	* Note: IP leaves IP header in first mbuf.
271	*/
272	ti = mtod(m, struct tcpiphdr *);
273	if (iphlen > sizeof(struct ip )) {
274	ip_stripoptions(m, (struct mbuf *)0);
275	iphlen=sizeof(struct ip );
276	}
277	/* XXX Check if too short */
278
279
280	/*
281	* Save a copy of the IP header in case we want restore it
282	* for sending an ICMP error message in response.
283	*/
284	ip=mtod(m, struct ip *);
285	save_ip = *ip;
286	save_ip.ip_len+= iphlen;
287
288	/*
289	* Checksum extended TCP header and data.
290	*/
291	tlen = ((struct ip *)ti)->ip_len;
292	ti->ti_next = ti->ti_prev = 0;
293	ti->ti_x1 = 0;
294	ti->ti_len = htons((u_int16_t)tlen);
295	len = sizeof(struct ip ) + tlen;
296	/* keep checksum for ICMP reply
297	* ti->ti_sum = cksum(m, len);
298	* if (ti->ti_sum) { */
299	if(cksum(m, len)) {
300	tcpstat.tcps_rcvbadsum++;
301	goto drop;
302	}
303
304	/*
305	* Check that TCP offset makes sense,
306	* pull out TCP options and adjust length. XXX
307	*/
308	off = ti->ti_off << 2;
309	if (off < sizeof (struct tcphdr) \|\| off > tlen) {
310	tcpstat.tcps_rcvbadoff++;
311	goto drop;
312	}
313	tlen -= off;
314	ti->ti_len = tlen;
315	if (off > sizeof (struct tcphdr)) {
316	optlen = off - sizeof (struct tcphdr);
317	optp = mtod(m, caddr_t) + sizeof (struct tcpiphdr);
318
319	/*
320	* Do quick retrieval of timestamp options ("options
321	* prediction?"). If timestamp is the only option and it's
322	* formatted as recommended in RFC 1323 appendix A, we
323	* quickly get the values now and not bother calling
324	* tcp_dooptions(), etc.
325	*/
326	/* if ((optlen == TCPOLEN_TSTAMP_APPA \|\|
327	* (optlen > TCPOLEN_TSTAMP_APPA &&
328	* optp[TCPOLEN_TSTAMP_APPA] == TCPOPT_EOL)) &&
329	* (u_int32_t )optp == htonl(TCPOPT_TSTAMP_HDR) &&
330	* (ti->ti_flags & TH_SYN) == 0) {
331	* ts_present = 1;
332	* ts_val = ntohl((u_int32_t )(optp + 4));
333	* ts_ecr = ntohl((u_int32_t )(optp + 8));
334	* optp = NULL; / * we've parsed the options * /
335	* }
336	*/
337	}
338	tiflags = ti->ti_flags;
339
340	/*
341	* Convert TCP protocol specific fields to host format.
342	*/
343	NTOHL(ti->ti_seq);
344	NTOHL(ti->ti_ack);
345	NTOHS(ti->ti_win);
346	NTOHS(ti->ti_urp);
347
348	/*
349	* Drop TCP, IP headers and TCP options.
350	*/
351	m->m_data += sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
352	m->m_len -= sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
353
354	/*
355	* Locate pcb for segment.
356	*/
357	findso:
358	so = tcp_last_so;
359	if (so->so_fport != ti->ti_dport \|\|
360	so->so_lport != ti->ti_sport \|\|
361	so->so_laddr.s_addr != ti->ti_src.s_addr \|\|
362	so->so_faddr.s_addr != ti->ti_dst.s_addr) {
363	so = solookup(&tcb, ti->ti_src, ti->ti_sport,
364	ti->ti_dst, ti->ti_dport);
365	if (so)
366	tcp_last_so = so;
367	++tcpstat.tcps_socachemiss;
368	}
369
370	/*
371	* If the state is CLOSED (i.e., TCB does not exist) then
372	* all data in the incoming segment is discarded.
373	* If the TCB exists but is in CLOSED state, it is embryonic,
374	* but should either do a listen or a connect soon.
375	*
376	* state == CLOSED means we've done socreate() but haven't
377	* attached it to a protocol yet...
378	*
379	* XXX If a TCB does not exist, and the TH_SYN flag is
380	* the only flag set, then create a session, mark it
381	* as if it was LISTENING, and continue...
382	*/
383	if (so == 0) {
384	if ((tiflags & (TH_SYN\|TH_FIN\|TH_RST\|TH_URG\|TH_ACK)) != TH_SYN)
385	goto dropwithreset;
386
387	if ((so = socreate()) == NULL)
388	goto dropwithreset;
389	if (tcp_attach(pData, so) < 0) {
390	free(so); /* Not sofree (if it failed, it's not insqued) */
391	goto dropwithreset;
392	}
393	#if defined(VBOX_WITH_SIMPLEFIED_SLIRP_SYNC) && defined(RT_OS_WINDOWS)
394	soregister_event(pData, so);
395	#endif
396
397	sbreserve(&so->so_snd, tcp_sndspace);
398	sbreserve(&so->so_rcv, tcp_rcvspace);
399
400	/* tcp_last_so = so; / / XXX ? */
401	/* tp = sototcpcb(so); */
402
403	so->so_laddr = ti->ti_src;
404	so->so_lport = ti->ti_sport;
405	so->so_faddr = ti->ti_dst;
406	so->so_fport = ti->ti_dport;
407
408	if ((so->so_iptos = tcp_tos(so)) == 0)
409	so->so_iptos = ((struct ip *)ti)->ip_tos;
410
411	tp = sototcpcb(so);
412	tp->t_state = TCPS_LISTEN;
413	}
414
415	/*
416	* If this is a still-connecting socket, this probably
417	* a retransmit of the SYN. Whether it's a retransmit SYN
418	* or something else, we nuke it.
419	*/
420	if (so->so_state & SS_ISFCONNECTING)
421	goto drop;
422
423	tp = sototcpcb(so);
424
425	/* XXX Should never fail */
426	if (tp == 0)
427	goto dropwithreset;
428	if (tp->t_state == TCPS_CLOSED)
429	goto drop;
430
431	/* Unscale the window into a 32-bit value. */
432	/* if ((tiflags & TH_SYN) == 0)
433	* tiwin = ti->ti_win << tp->snd_scale;
434	* else
435	*/
436	tiwin = ti->ti_win;
437
438	/*
439	* Segment received on connection.
440	* Reset idle time and keep-alive timer.
441	*/
442	tp->t_idle = 0;
443	if (so_options)
444	tp->t_timer[TCPT_KEEP] = tcp_keepintvl;
445	else
446	tp->t_timer[TCPT_KEEP] = tcp_keepidle;
447
448	/*
449	* Process options if not in LISTEN state,
450	* else do it below (after getting remote address).
451	*/
452	if (optp && tp->t_state != TCPS_LISTEN)
453	tcp_dooptions(pData, tp, (u_char *)optp, optlen, ti);
454	/* , */
455	/* &ts_present, &ts_val, &ts_ecr); */
456
457	/*
458	* Header prediction: check for the two common cases
459	* of a uni-directional data xfer. If the packet has
460	* no control flags, is in-sequence, the window didn't
461	* change and we're not retransmitting, it's a
462	* candidate. If the length is zero and the ack moved
463	* forward, we're the sender side of the xfer. Just
464	* free the data acked & wake any higher level process
465	* that was blocked waiting for space. If the length
466	* is non-zero and the ack didn't move, we're the
467	* receiver side. If we're getting packets in-order
468	* (the reassembly queue is empty), add the data to
469	* the socket buffer and note that we need a delayed ack.
470	*
471	* XXX Some of these tests are not needed
472	* eg: the tiwin == tp->snd_wnd prevents many more
473	* predictions.. with no real advantage..
474	*/
475	if (tp->t_state == TCPS_ESTABLISHED &&
476	(tiflags & (TH_SYN\|TH_FIN\|TH_RST\|TH_URG\|TH_ACK)) == TH_ACK &&
477	/* (!ts_present \|\| TSTMP_GEQ(ts_val, tp->ts_recent)) && */
478	ti->ti_seq == tp->rcv_nxt &&
479	tiwin && tiwin == tp->snd_wnd &&
480	tp->snd_nxt == tp->snd_max) {
481	/*
482	* If last ACK falls within this segment's sequence numbers,
483	* record the timestamp.
484	*/
485	/* if (ts_present && SEQ_LEQ(ti->ti_seq, tp->last_ack_sent) &&
486	* SEQ_LT(tp->last_ack_sent, ti->ti_seq + ti->ti_len)) {
487	* tp->ts_recent_age = tcp_now;
488	* tp->ts_recent = ts_val;
489	* }
490	*/
491	if (ti->ti_len == 0) {
492	if (SEQ_GT(ti->ti_ack, tp->snd_una) &&
493	SEQ_LEQ(ti->ti_ack, tp->snd_max) &&
494	tp->snd_cwnd >= tp->snd_wnd) {
495	/*
496	* this is a pure ack for outstanding data.
497	*/
498	++tcpstat.tcps_predack;
499	/* if (ts_present)
500	* tcp_xmit_timer(tp, tcp_now-ts_ecr+1);
501	* else
502	*/ if (tp->t_rtt &&
503	SEQ_GT(ti->ti_ack, tp->t_rtseq))
504	tcp_xmit_timer(pData, tp, tp->t_rtt);
505	acked = ti->ti_ack - tp->snd_una;
506	tcpstat.tcps_rcvackpack++;
507	tcpstat.tcps_rcvackbyte += acked;
508	sbdrop(&so->so_snd, acked);
509	tp->snd_una = ti->ti_ack;
510	m_freem(pData, m);
511
512	/*
513	* If all outstanding data are acked, stop
514	* retransmit timer, otherwise restart timer
515	* using current (possibly backed-off) value.
516	* If process is waiting for space,
517	* wakeup/selwakeup/signal. If data
518	* are ready to send, let tcp_output
519	* decide between more output or persist.
520	*/
521	if (tp->snd_una == tp->snd_max)
522	tp->t_timer[TCPT_REXMT] = 0;
523	else if (tp->t_timer[TCPT_PERSIST] == 0)
524	tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
525
526	/*
527	* There's room in so_snd, sowwakup will read()
528	* from the socket if we can
529	*/
530	/* if (so->so_snd.sb_flags & SB_NOTIFY)
531	* sowwakeup(so);
532	*/
533	/*
534	* This is called because sowwakeup might have
535	* put data into so_snd. Since we don't so sowwakeup,
536	* we don't need this.. XXX???
537	*/
538	if (so->so_snd.sb_cc)
539	(void) tcp_output(pData, tp);
540
541	return;
542	}
543	} else if (ti->ti_ack == tp->snd_una &&
544	u32_to_ptr(pData, tp->seg_next, struct tcpcb *) == tp &&
545	ti->ti_len <= sbspace(&so->so_rcv)) {
546	/*
547	* this is a pure, in-sequence data packet
548	* with nothing on the reassembly queue and
549	* we have enough buffer space to take it.
550	*/
551	++tcpstat.tcps_preddat;
552	tp->rcv_nxt += ti->ti_len;
553	tcpstat.tcps_rcvpack++;
554	tcpstat.tcps_rcvbyte += ti->ti_len;
555	/*
556	* Add data to socket buffer.
557	*/
558	if (so->so_emu) {
559	if (tcp_emu(pData, so,m)) sbappend(pData, so, m);
560	} else
561	sbappend(pData, so, m);
562
563	/*
564	* XXX This is called when data arrives. Later, check
565	* if we can actually write() to the socket
566	* XXX Need to check? It's be NON_BLOCKING
567	*/
568	/* sorwakeup(so); */
569
570	/*
571	* If this is a short packet, then ACK now - with Nagel
572	* congestion avoidance sender won't send more until
573	* he gets an ACK.
574	*
575	* It is better to not delay acks at all to maximize
576	* TCP throughput. See RFC 2581.
577	*/
578	tp->t_flags \|= TF_ACKNOW;
579	tcp_output(pData, tp);
580	return;
581	}
582	} /* header prediction */
583	/*
584	* Calculate amount of space in receive window,
585	* and then do TCP input processing.
586	* Receive window is amount of space in rcv queue,
587	* but not less than advertised window.
588	*/
589	{ int win;
590	win = sbspace(&so->so_rcv);
591	if (win < 0)
592	win = 0;
593	tp->rcv_wnd = max(win, (int)(tp->rcv_adv - tp->rcv_nxt));
594	}
595
596	switch (tp->t_state) {
597
598	/*
599	* If the state is LISTEN then ignore segment if it contains an RST.
600	* If the segment contains an ACK then it is bad and send a RST.
601	* If it does not contain a SYN then it is not interesting; drop it.
602	* Don't bother responding if the destination was a broadcast.
603	* Otherwise initialize tp->rcv_nxt, and tp->irs, select an initial
604	* tp->iss, and send a segment:
605	* <SEQ=ISS><ACK=RCV_NXT><CTL=SYN,ACK>
606	* Also initialize tp->snd_nxt to tp->iss+1 and tp->snd_una to tp->iss.
607	* Fill in remote peer address fields if not previously specified.
608	* Enter SYN_RECEIVED state, and process any other fields of this
609	* segment in this state.
610	*/
611	case TCPS_LISTEN: {
612
613	if (tiflags & TH_RST)
614	goto drop;
615	if (tiflags & TH_ACK)
616	goto dropwithreset;
617	if ((tiflags & TH_SYN) == 0)
618	goto drop;
619
620	/*
621	* This has way too many gotos...
622	* But a bit of spaghetti code never hurt anybody :)
623	*/
624
625	/*
626	* If this is destined for the control address, then flag to
627	* tcp_ctl once connected, otherwise connect
628	*/
629	if ((so->so_faddr.s_addr&htonl(pData->netmask)) == special_addr.s_addr) {
630	int lastbyte=ntohl(so->so_faddr.s_addr) & ~pData->netmask;
631	if (lastbyte!=CTL_ALIAS && lastbyte!=CTL_DNS) {
632	#if 0
633	if(lastbyte==CTL_CMD \|\| lastbyte==CTL_EXEC) {
634	/* Command or exec adress */
635	so->so_state \|= SS_CTL;
636	} else
637	#endif
638	{
639	/* May be an add exec */
640	struct ex_list *ex_ptr;
641	for(ex_ptr = exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next) {
642	if(ex_ptr->ex_fport == so->so_fport &&
643	lastbyte == ex_ptr->ex_addr) {
644	so->so_state \|= SS_CTL;
645	break;
646	}
647	}
648	}
649	if(so->so_state & SS_CTL) goto cont_input;
650	}
651	/* CTL_ALIAS: Do nothing, tcp_fconnect will be called on it */
652	}
653
654	if (so->so_emu & EMU_NOCONNECT) {
655	so->so_emu &= ~EMU_NOCONNECT;
656	goto cont_input;
657	}
658
659	if((tcp_fconnect(pData, so) == -1) && (errno != EINPROGRESS) && (errno != EWOULDBLOCK)) {
660	u_char code=ICMP_UNREACH_NET;
661	DEBUG_MISC((dfd," tcp fconnect errno = %d-%s\n",
662	errno,strerror(errno)));
663	if(errno == ECONNREFUSED) {
664	/* ACK the SYN, send RST to refuse the connection */
665	tcp_respond(pData, tp, ti, m, ti->ti_seq+1, (tcp_seq)0,
666	TH_RST\|TH_ACK);
667	} else {
668	if(errno == EHOSTUNREACH) code=ICMP_UNREACH_HOST;
669	HTONL(ti->ti_seq); /* restore tcp header */
670	HTONL(ti->ti_ack);
671	HTONS(ti->ti_win);
672	HTONS(ti->ti_urp);
673	m->m_data -= sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
674	m->m_len += sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
675	*ip=save_ip;
676	icmp_error(pData, m, ICMP_UNREACH,code, 0,strerror(errno));
677	}
678	tp = tcp_close(pData, tp);
679	m_free(pData, m);
680	} else {
681	/*
682	* Haven't connected yet, save the current mbuf
683	* and ti, and return
684	* XXX Some OS's don't tell us whether the connect()
685	* succeeded or not. So we must time it out.
686	*/
687	so->so_m = m;
688	so->so_ti = ti;
689	tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT;
690	tp->t_state = TCPS_SYN_RECEIVED;
691	}
692	return;
693
694	cont_conn:
695	/* m==NULL
696	* Check if the connect succeeded
697	*/
698	if (so->so_state & SS_NOFDREF) {
699	tp = tcp_close(pData, tp);
700	goto dropwithreset;
701	}
702	cont_input:
703	tcp_template(tp);
704
705	if (optp)
706	tcp_dooptions(pData, tp, (u_char *)optp, optlen, ti);
707	/* , */
708	/* &ts_present, &ts_val, &ts_ecr); */
709
710	if (iss)
711	tp->iss = iss;
712	else
713	tp->iss = tcp_iss;
714	tcp_iss += TCP_ISSINCR/2;
715	tp->irs = ti->ti_seq;
716	tcp_sendseqinit(tp);
717	tcp_rcvseqinit(tp);
718	tp->t_flags \|= TF_ACKNOW;
719	tp->t_state = TCPS_SYN_RECEIVED;
720	tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT;
721	tcpstat.tcps_accepts++;
722	goto trimthenstep6;
723	} /* case TCPS_LISTEN */
724
725	/*
726	* If the state is SYN_SENT:
727	* if seg contains an ACK, but not for our SYN, drop the input.
728	* if seg contains a RST, then drop the connection.
729	* if seg does not contain SYN, then drop it.
730	* Otherwise this is an acceptable SYN segment
731	* initialize tp->rcv_nxt and tp->irs
732	* if seg contains ack then advance tp->snd_una
733	* if SYN has been acked change to ESTABLISHED else SYN_RCVD state
734	* arrange for segment to be acked (eventually)
735	* continue processing rest of data/controls, beginning with URG
736	*/
737	case TCPS_SYN_SENT:
738	if ((tiflags & TH_ACK) &&
739	(SEQ_LEQ(ti->ti_ack, tp->iss) \|\|
740	SEQ_GT(ti->ti_ack, tp->snd_max)))
741	goto dropwithreset;
742
743	if (tiflags & TH_RST) {
744	if (tiflags & TH_ACK)
745	tp = tcp_drop(pData, tp,0); /* XXX Check t_softerror! */
746	goto drop;
747	}
748
749	if ((tiflags & TH_SYN) == 0)
750	goto drop;
751	if (tiflags & TH_ACK) {
752	tp->snd_una = ti->ti_ack;
753	if (SEQ_LT(tp->snd_nxt, tp->snd_una))
754	tp->snd_nxt = tp->snd_una;
755	}
756
757	tp->t_timer[TCPT_REXMT] = 0;
758	tp->irs = ti->ti_seq;
759	tcp_rcvseqinit(tp);
760	tp->t_flags \|= TF_ACKNOW;
761	if (tiflags & TH_ACK && SEQ_GT(tp->snd_una, tp->iss)) {
762	tcpstat.tcps_connects++;
763	soisfconnected(so);
764	tp->t_state = TCPS_ESTABLISHED;
765
766	/* Do window scaling on this connection? */
767	/* if ((tp->t_flags & (TF_RCVD_SCALE\|TF_REQ_SCALE)) ==
768	* (TF_RCVD_SCALE\|TF_REQ_SCALE)) {
769	* tp->snd_scale = tp->requested_s_scale;
770	* tp->rcv_scale = tp->request_r_scale;
771	* }
772	*/
773	(void) tcp_reass(pData, tp, (struct tcpiphdr *)0,
774	(struct mbuf *)0);
775	/*
776	* if we didn't have to retransmit the SYN,
777	* use its rtt as our initial srtt & rtt var.
778	*/
779	if (tp->t_rtt)
780	tcp_xmit_timer(pData, tp, tp->t_rtt);
781	} else
782	tp->t_state = TCPS_SYN_RECEIVED;
783
784	trimthenstep6:
785	/*
786	* Advance ti->ti_seq to correspond to first data byte.
787	* If data, trim to stay within window,
788	* dropping FIN if necessary.
789	*/
790	ti->ti_seq++;
791	if (ti->ti_len > tp->rcv_wnd) {
792	todrop = ti->ti_len - tp->rcv_wnd;
793	m_adj(m, -todrop);
794	ti->ti_len = tp->rcv_wnd;
795	tiflags &= ~TH_FIN;
796	tcpstat.tcps_rcvpackafterwin++;
797	tcpstat.tcps_rcvbyteafterwin += todrop;
798	}
799	tp->snd_wl1 = ti->ti_seq - 1;
800	tp->rcv_up = ti->ti_seq;
801	goto step6;
802	} /* switch tp->t_state */
803	/*
804	* States other than LISTEN or SYN_SENT.
805	* First check timestamp, if present.
806	* Then check that at least some bytes of segment are within
807	* receive window. If segment begins before rcv_nxt,
808	* drop leading data (and SYN); if nothing left, just ack.
809	*
810	* RFC 1323 PAWS: If we have a timestamp reply on this segment
811	* and it's less than ts_recent, drop it.
812	*/
813	/* if (ts_present && (tiflags & TH_RST) == 0 && tp->ts_recent &&
814	* TSTMP_LT(ts_val, tp->ts_recent)) {
815	*
816	/ / Check to see if ts_recent is over 24 days old. */
817	/* if ((int)(tcp_now - tp->ts_recent_age) > TCP_PAWS_IDLE) {
818	/ /
819	* * Invalidate ts_recent. If this segment updates
820	* * ts_recent, the age will be reset later and ts_recent
821	* * will get a valid value. If it does not, setting
822	* * ts_recent to zero will at least satisfy the
823	* * requirement that zero be placed in the timestamp
824	* * echo reply when ts_recent isn't valid. The
825	* * age isn't reset until we get a valid ts_recent
826	* * because we don't want out-of-order segments to be
827	* * dropped when ts_recent is old.
828	* */
829	/* tp->ts_recent = 0;
830	* } else {
831	* tcpstat.tcps_rcvduppack++;
832	* tcpstat.tcps_rcvdupbyte += ti->ti_len;
833	* tcpstat.tcps_pawsdrop++;
834	* goto dropafterack;
835	* }
836	* }
837	*/
838
839	todrop = tp->rcv_nxt - ti->ti_seq;
840	if (todrop > 0) {
841	if (tiflags & TH_SYN) {
842	tiflags &= ~TH_SYN;
843	ti->ti_seq++;
844	if (ti->ti_urp > 1)
845	ti->ti_urp--;
846	else
847	tiflags &= ~TH_URG;
848	todrop--;
849	}
850	/*
851	* Following if statement from Stevens, vol. 2, p. 960.
852	*/
853	if (todrop > ti->ti_len
854	\|\| (todrop == ti->ti_len && (tiflags & TH_FIN) == 0)) {
855	/*
856	* Any valid FIN must be to the left of the window.
857	* At this point the FIN must be a duplicate or out
858	* of sequence; drop it.
859	*/
860	tiflags &= ~TH_FIN;
861
862	/*
863	* Send an ACK to resynchronize and drop any data.
864	* But keep on processing for RST or ACK.
865	*/
866	tp->t_flags \|= TF_ACKNOW;
867	todrop = ti->ti_len;
868	tcpstat.tcps_rcvduppack++;
869	tcpstat.tcps_rcvdupbyte += todrop;
870	} else {
871	tcpstat.tcps_rcvpartduppack++;
872	tcpstat.tcps_rcvpartdupbyte += todrop;
873	}
874	m_adj(m, todrop);
875	ti->ti_seq += todrop;
876	ti->ti_len -= todrop;
877	if (ti->ti_urp > todrop)
878	ti->ti_urp -= todrop;
879	else {
880	tiflags &= ~TH_URG;
881	ti->ti_urp = 0;
882	}
883	}
884	/*
885	* If new data are received on a connection after the
886	* user processes are gone, then RST the other end.
887	*/
888	if ((so->so_state & SS_NOFDREF) &&
889	tp->t_state > TCPS_CLOSE_WAIT && ti->ti_len) {
890	tp = tcp_close(pData, tp);
891	tcpstat.tcps_rcvafterclose++;
892	goto dropwithreset;
893	}
894
895	/*
896	* If segment ends after window, drop trailing data
897	* (and PUSH and FIN); if nothing left, just ACK.
898	*/
899	todrop = (ti->ti_seq+ti->ti_len) - (tp->rcv_nxt+tp->rcv_wnd);
900	if (todrop > 0) {
901	tcpstat.tcps_rcvpackafterwin++;
902	if (todrop >= ti->ti_len) {
903	tcpstat.tcps_rcvbyteafterwin += ti->ti_len;
904	/*
905	* If a new connection request is received
906	* while in TIME_WAIT, drop the old connection
907	* and start over if the sequence numbers
908	* are above the previous ones.
909	*/
910	if (tiflags & TH_SYN &&
911	tp->t_state == TCPS_TIME_WAIT &&
912	SEQ_GT(ti->ti_seq, tp->rcv_nxt)) {
913	iss = tp->rcv_nxt + TCP_ISSINCR;
914	tp = tcp_close(pData, tp);
915	goto findso;
916	}
917	/*
918	* If window is closed can only take segments at
919	* window edge, and have to drop data and PUSH from
920	* incoming segments. Continue processing, but
921	* remember to ack. Otherwise, drop segment
922	* and ack.
923	*/
924	if (tp->rcv_wnd == 0 && ti->ti_seq == tp->rcv_nxt) {
925	tp->t_flags \|= TF_ACKNOW;
926	tcpstat.tcps_rcvwinprobe++;
927	} else
928	goto dropafterack;
929	} else
930	tcpstat.tcps_rcvbyteafterwin += todrop;
931	m_adj(m, -todrop);
932	ti->ti_len -= todrop;
933	tiflags &= ~(TH_PUSH\|TH_FIN);
934	}
935
936	/*
937	* If last ACK falls within this segment's sequence numbers,
938	* record its timestamp.
939	*/
940	/* if (ts_present && SEQ_LEQ(ti->ti_seq, tp->last_ack_sent) &&
941	* SEQ_LT(tp->last_ack_sent, ti->ti_seq + ti->ti_len +
942	* ((tiflags & (TH_SYN\|TH_FIN)) != 0))) {
943	* tp->ts_recent_age = tcp_now;
944	* tp->ts_recent = ts_val;
945	* }
946	*/
947
948	/*
949	* If the RST bit is set examine the state:
950	* SYN_RECEIVED STATE:
951	* If passive open, return to LISTEN state.
952	* If active open, inform user that connection was refused.
953	* ESTABLISHED, FIN_WAIT_1, FIN_WAIT2, CLOSE_WAIT STATES:
954	* Inform user that connection was reset, and close tcb.
955	* CLOSING, LAST_ACK, TIME_WAIT STATES
956	* Close the tcb.
957	*/
958	if (tiflags&TH_RST) switch (tp->t_state) {
959
960	case TCPS_SYN_RECEIVED:
961	/* so->so_error = ECONNREFUSED; */
962	goto close;
963
964	case TCPS_ESTABLISHED:
965	case TCPS_FIN_WAIT_1:
966	case TCPS_FIN_WAIT_2:
967	case TCPS_CLOSE_WAIT:
968	/* so->so_error = ECONNRESET; */
969	close:
970	tp->t_state = TCPS_CLOSED;
971	tcpstat.tcps_drops++;
972	tp = tcp_close(pData, tp);
973	goto drop;
974
975	case TCPS_CLOSING:
976	case TCPS_LAST_ACK:
977	case TCPS_TIME_WAIT:
978	tp = tcp_close(pData, tp);
979	goto drop;
980	}
981
982	/*
983	* If a SYN is in the window, then this is an
984	* error and we send an RST and drop the connection.
985	*/
986	if (tiflags & TH_SYN) {
987	tp = tcp_drop(pData, tp,0);
988	goto dropwithreset;
989	}
990
991	/*
992	* If the ACK bit is off we drop the segment and return.
993	*/
994	if ((tiflags & TH_ACK) == 0) goto drop;
995
996	/*
997	* Ack processing.
998	*/
999	switch (tp->t_state) {
1000	/*
1001	* In SYN_RECEIVED state if the ack ACKs our SYN then enter
1002	* ESTABLISHED state and continue processing, otherwise
1003	* send an RST. una<=ack<=max
1004	*/
1005	case TCPS_SYN_RECEIVED:
1006
1007	if (SEQ_GT(tp->snd_una, ti->ti_ack) \|\|
1008	SEQ_GT(ti->ti_ack, tp->snd_max))
1009	goto dropwithreset;
1010	tcpstat.tcps_connects++;
1011	tp->t_state = TCPS_ESTABLISHED;
1012	/*
1013	* The sent SYN is ack'ed with our sequence number +1
1014	* The first data byte already in the buffer will get
1015	* lost if no correction is made. This is only needed for
1016	* SS_CTL since the buffer is empty otherwise.
1017	* tp->snd_una++; or:
1018	*/
1019	tp->snd_una=ti->ti_ack;
1020	if (so->so_state & SS_CTL) {
1021	/* So tcp_ctl reports the right state */
1022	ret = tcp_ctl(pData, so);
1023	if (ret == 1) {
1024	soisfconnected(so);
1025	so->so_state &= ~SS_CTL; /* success XXX */
1026	} else if (ret == 2) {
1027	so->so_state = SS_NOFDREF; /* CTL_CMD */
1028	} else {
1029	needoutput = 1;
1030	tp->t_state = TCPS_FIN_WAIT_1;
1031	}
1032	} else {
1033	soisfconnected(so);
1034	}
1035
1036	/* Do window scaling? */
1037	/* if ((tp->t_flags & (TF_RCVD_SCALE\|TF_REQ_SCALE)) ==
1038	* (TF_RCVD_SCALE\|TF_REQ_SCALE)) {
1039	* tp->snd_scale = tp->requested_s_scale;
1040	* tp->rcv_scale = tp->request_r_scale;
1041	* }
1042	*/
1043	(void) tcp_reass(pData, tp, (struct tcpiphdr )0, (struct mbuf )0);
1044	tp->snd_wl1 = ti->ti_seq - 1;
1045	/* Avoid ack processing; snd_una==ti_ack => dup ack */
1046	goto synrx_to_est;
1047	/* fall into ... */
1048
1049	/*
1050	* In ESTABLISHED state: drop duplicate ACKs; ACK out of range
1051	* ACKs. If the ack is in the range
1052	* tp->snd_una < ti->ti_ack <= tp->snd_max
1053	* then advance tp->snd_una to ti->ti_ack and drop
1054	* data from the retransmission queue. If this ACK reflects
1055	* more up to date window information we update our window information.
1056	*/
1057	case TCPS_ESTABLISHED:
1058	case TCPS_FIN_WAIT_1:
1059	case TCPS_FIN_WAIT_2:
1060	case TCPS_CLOSE_WAIT:
1061	case TCPS_CLOSING:
1062	case TCPS_LAST_ACK:
1063	case TCPS_TIME_WAIT:
1064
1065	if (SEQ_LEQ(ti->ti_ack, tp->snd_una)) {
1066	if (ti->ti_len == 0 && tiwin == tp->snd_wnd) {
1067	tcpstat.tcps_rcvdupack++;
1068	DEBUG_MISC((dfd," dup ack m = %lx so = %lx \n",
1069	(long )m, (long )so));
1070	/*
1071	* If we have outstanding data (other than
1072	* a window probe), this is a completely
1073	* duplicate ack (ie, window info didn't
1074	* change), the ack is the biggest we've
1075	* seen and we've seen exactly our rexmt
1076	* threshold of them, assume a packet
1077	* has been dropped and retransmit it.
1078	* Kludge snd_nxt & the congestion
1079	* window so we send only this one
1080	* packet.
1081	*
1082	* We know we're losing at the current
1083	* window size so do congestion avoidance
1084	* (set ssthresh to half the current window
1085	* and pull our congestion window back to
1086	* the new ssthresh).
1087	*
1088	* Dup acks mean that packets have left the
1089	* network (they're now cached at the receiver)
1090	* so bump cwnd by the amount in the receiver
1091	* to keep a constant cwnd packets in the
1092	* network.
1093	*/
1094	if (tp->t_timer[TCPT_REXMT] == 0 \|\|
1095	ti->ti_ack != tp->snd_una)
1096	tp->t_dupacks = 0;
1097	else if (++tp->t_dupacks == tcprexmtthresh) {
1098	tcp_seq onxt = tp->snd_nxt;
1099	u_int win =
1100	min(tp->snd_wnd, tp->snd_cwnd) / 2 /
1101	tp->t_maxseg;
1102
1103	if (win < 2)
1104	win = 2;
1105	tp->snd_ssthresh = win * tp->t_maxseg;
1106	tp->t_timer[TCPT_REXMT] = 0;
1107	tp->t_rtt = 0;
1108	tp->snd_nxt = ti->ti_ack;
1109	tp->snd_cwnd = tp->t_maxseg;
1110	(void) tcp_output(pData, tp);
1111	tp->snd_cwnd = tp->snd_ssthresh +
1112	tp->t_maxseg * tp->t_dupacks;
1113	if (SEQ_GT(onxt, tp->snd_nxt))
1114	tp->snd_nxt = onxt;
1115	goto drop;
1116	} else if (tp->t_dupacks > tcprexmtthresh) {
1117	tp->snd_cwnd += tp->t_maxseg;
1118	(void) tcp_output(pData, tp);
1119	goto drop;
1120	}
1121	} else
1122	tp->t_dupacks = 0;
1123	break;
1124	}
1125	synrx_to_est:
1126	/*
1127	* If the congestion window was inflated to account
1128	* for the other side's cached packets, retract it.
1129	*/
1130	if (tp->t_dupacks > tcprexmtthresh &&
1131	tp->snd_cwnd > tp->snd_ssthresh)
1132	tp->snd_cwnd = tp->snd_ssthresh;
1133	tp->t_dupacks = 0;
1134	if (SEQ_GT(ti->ti_ack, tp->snd_max)) {
1135	tcpstat.tcps_rcvacktoomuch++;
1136	goto dropafterack;
1137	}
1138	acked = ti->ti_ack - tp->snd_una;
1139	tcpstat.tcps_rcvackpack++;
1140	tcpstat.tcps_rcvackbyte += acked;
1141
1142	/*
1143	* If we have a timestamp reply, update smoothed
1144	* round trip time. If no timestamp is present but
1145	* transmit timer is running and timed sequence
1146	* number was acked, update smoothed round trip time.
1147	* Since we now have an rtt measurement, cancel the
1148	* timer backoff (cf., Phil Karn's retransmit alg.).
1149	* Recompute the initial retransmit timer.
1150	*/
1151	/* if (ts_present)
1152	* tcp_xmit_timer(tp, tcp_now-ts_ecr+1);
1153	* else
1154	*/
1155	if (tp->t_rtt && SEQ_GT(ti->ti_ack, tp->t_rtseq))
1156	tcp_xmit_timer(pData, tp,tp->t_rtt);
1157
1158	/*
1159	* If all outstanding data is acked, stop retransmit
1160	* timer and remember to restart (more output or persist).
1161	* If there is more data to be acked, restart retransmit
1162	* timer, using current (possibly backed-off) value.
1163	*/
1164	if (ti->ti_ack == tp->snd_max) {
1165	tp->t_timer[TCPT_REXMT] = 0;
1166	needoutput = 1;
1167	} else if (tp->t_timer[TCPT_PERSIST] == 0)
1168	tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
1169	/*
1170	* When new data is acked, open the congestion window.
1171	* If the window gives us less than ssthresh packets
1172	* in flight, open exponentially (maxseg per packet).
1173	* Otherwise open linearly: maxseg per window
1174	* (maxseg^2 / cwnd per packet).
1175	*/
1176	{
1177	register u_int cw = tp->snd_cwnd;
1178	register u_int incr = tp->t_maxseg;
1179
1180	if (cw > tp->snd_ssthresh)
1181	incr = incr * incr / cw;
1182	tp->snd_cwnd = min(cw + incr, TCP_MAXWIN<<tp->snd_scale);
1183	}
1184	if (acked > so->so_snd.sb_cc) {
1185	tp->snd_wnd -= so->so_snd.sb_cc;
1186	sbdrop(&so->so_snd, (int )so->so_snd.sb_cc);
1187	ourfinisacked = 1;
1188	} else {
1189	sbdrop(&so->so_snd, acked);
1190	tp->snd_wnd -= acked;
1191	ourfinisacked = 0;
1192	}
1193	/*
1194	* XXX sowwakup is called when data is acked and there's room for
1195	* for more data... it should read() the socket
1196	*/
1197	/* if (so->so_snd.sb_flags & SB_NOTIFY)
1198	* sowwakeup(so);
1199	*/
1200	tp->snd_una = ti->ti_ack;
1201	if (SEQ_LT(tp->snd_nxt, tp->snd_una))
1202	tp->snd_nxt = tp->snd_una;
1203
1204	switch (tp->t_state) {
1205
1206	/*
1207	* In FIN_WAIT_1 STATE in addition to the processing
1208	* for the ESTABLISHED state if our FIN is now acknowledged
1209	* then enter FIN_WAIT_2.
1210	*/
1211	case TCPS_FIN_WAIT_1:
1212	if (ourfinisacked) {
1213	/*
1214	* If we can't receive any more
1215	* data, then closing user can proceed.
1216	* Starting the timer is contrary to the
1217	* specification, but if we don't get a FIN
1218	* we'll hang forever.
1219	*/
1220	if (so->so_state & SS_FCANTRCVMORE) {
1221	soisfdisconnected(so);
1222	tp->t_timer[TCPT_2MSL] = tcp_maxidle;
1223	}
1224	tp->t_state = TCPS_FIN_WAIT_2;
1225	}
1226	break;
1227
1228	/*
1229	* In CLOSING STATE in addition to the processing for
1230	* the ESTABLISHED state if the ACK acknowledges our FIN
1231	* then enter the TIME-WAIT state, otherwise ignore
1232	* the segment.
1233	*/
1234	case TCPS_CLOSING:
1235	if (ourfinisacked) {
1236	tp->t_state = TCPS_TIME_WAIT;
1237	tcp_canceltimers(tp);
1238	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1239	soisfdisconnected(so);
1240	}
1241	break;
1242
1243	/*
1244	* In LAST_ACK, we may still be waiting for data to drain
1245	* and/or to be acked, as well as for the ack of our FIN.
1246	* If our FIN is now acknowledged, delete the TCB,
1247	* enter the closed state and return.
1248	*/
1249	case TCPS_LAST_ACK:
1250	if (ourfinisacked) {
1251	tp = tcp_close(pData, tp);
1252	goto drop;
1253	}
1254	break;
1255
1256	/*
1257	* In TIME_WAIT state the only thing that should arrive
1258	* is a retransmission of the remote FIN. Acknowledge
1259	* it and restart the finack timer.
1260	*/
1261	case TCPS_TIME_WAIT:
1262	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1263	goto dropafterack;
1264	}
1265	} /* switch(tp->t_state) */
1266
1267	step6:
1268	/*
1269	* Update window information.
1270	* Don't look at window if no ACK: TAC's send garbage on first SYN.
1271	*/
1272	if ((tiflags & TH_ACK) &&
1273	(SEQ_LT(tp->snd_wl1, ti->ti_seq) \|\|
1274	(tp->snd_wl1 == ti->ti_seq && (SEQ_LT(tp->snd_wl2, ti->ti_ack) \|\|
1275	(tp->snd_wl2 == ti->ti_ack && tiwin > tp->snd_wnd))))) {
1276	/* keep track of pure window updates */
1277	if (ti->ti_len == 0 &&
1278	tp->snd_wl2 == ti->ti_ack && tiwin > tp->snd_wnd)
1279	tcpstat.tcps_rcvwinupd++;
1280	tp->snd_wnd = tiwin;
1281	tp->snd_wl1 = ti->ti_seq;
1282	tp->snd_wl2 = ti->ti_ack;
1283	if (tp->snd_wnd > tp->max_sndwnd)
1284	tp->max_sndwnd = tp->snd_wnd;
1285	needoutput = 1;
1286	}
1287
1288	/*
1289	* Process segments with URG.
1290	*/
1291	if ((tiflags & TH_URG) && ti->ti_urp &&
1292	TCPS_HAVERCVDFIN(tp->t_state) == 0) {
1293	/*
1294	* This is a kludge, but if we receive and accept
1295	* random urgent pointers, we'll crash in
1296	* soreceive. It's hard to imagine someone
1297	* actually wanting to send this much urgent data.
1298	*/
1299	if (ti->ti_urp + so->so_rcv.sb_cc > so->so_rcv.sb_datalen) {
1300	ti->ti_urp = 0;
1301	tiflags &= ~TH_URG;
1302	goto dodata;
1303	}
1304	/*
1305	* If this segment advances the known urgent pointer,
1306	* then mark the data stream. This should not happen
1307	* in CLOSE_WAIT, CLOSING, LAST_ACK or TIME_WAIT STATES since
1308	* a FIN has been received from the remote side.
1309	* In these states we ignore the URG.
1310	*
1311	* According to RFC961 (Assigned Protocols),
1312	* the urgent pointer points to the last octet
1313	* of urgent data. We continue, however,
1314	* to consider it to indicate the first octet
1315	* of data past the urgent section as the original
1316	* spec states (in one of two places).
1317	*/
1318	if (SEQ_GT(ti->ti_seq+ti->ti_urp, tp->rcv_up)) {
1319	tp->rcv_up = ti->ti_seq + ti->ti_urp;
1320	so->so_urgc = so->so_rcv.sb_cc +
1321	(tp->rcv_up - tp->rcv_nxt); /* -1; */
1322	tp->rcv_up = ti->ti_seq + ti->ti_urp;
1323
1324	}
1325	} else
1326	/*
1327	* If no out of band data is expected,
1328	* pull receive urgent pointer along
1329	* with the receive window.
1330	*/
1331	if (SEQ_GT(tp->rcv_nxt, tp->rcv_up))
1332	tp->rcv_up = tp->rcv_nxt;
1333	dodata:
1334
1335	/*
1336	* Process the segment text, merging it into the TCP sequencing queue,
1337	* and arranging for acknowledgment of receipt if necessary.
1338	* This process logically involves adjusting tp->rcv_wnd as data
1339	* is presented to the user (this happens in tcp_usrreq.c,
1340	* case PRU_RCVD). If a FIN has already been received on this
1341	* connection then we just ignore the text.
1342	*/
1343	if ((ti->ti_len \|\| (tiflags&TH_FIN)) &&
1344	TCPS_HAVERCVDFIN(tp->t_state) == 0) {
1345	TCP_REASS(pData, tp, ti, m, so, tiflags);
1346	/*
1347	* Note the amount of data that peer has sent into
1348	* our window, in order to estimate the sender's
1349	* buffer size.
1350	*/
1351	len = so->so_rcv.sb_datalen - (tp->rcv_adv - tp->rcv_nxt);
1352	} else {
1353	mbuf_freed = 1; /* The mbuf must be freed, but only when its content is not needed anymore. */
1354	tiflags &= ~TH_FIN;
1355	}
1356
1357	/*
1358	* If FIN is received ACK the FIN and let the user know
1359	* that the connection is closing.
1360	*/
1361	if (tiflags & TH_FIN) {
1362	if (TCPS_HAVERCVDFIN(tp->t_state) == 0) {
1363	/*
1364	* If we receive a FIN we can't send more data,
1365	* set it SS_FDRAIN
1366	* Shutdown the socket if there is no rx data in the
1367	* buffer.
1368	* soread() is called on completion of shutdown() and
1369	* will got to TCPS_LAST_ACK, and use tcp_output()
1370	* to send the FIN.
1371	*/
1372	/* sofcantrcvmore(so); */
1373	sofwdrain(so);
1374
1375	tp->t_flags \|= TF_ACKNOW;
1376	tp->rcv_nxt++;
1377	}
1378	switch (tp->t_state) {
1379
1380	/*
1381	* In SYN_RECEIVED and ESTABLISHED STATES
1382	* enter the CLOSE_WAIT state.
1383	*/
1384	case TCPS_SYN_RECEIVED:
1385	case TCPS_ESTABLISHED:
1386	if(so->so_emu == EMU_CTL) /* no shutdown on socket */
1387	tp->t_state = TCPS_LAST_ACK;
1388	else
1389	tp->t_state = TCPS_CLOSE_WAIT;
1390	break;
1391
1392	/*
1393	* If still in FIN_WAIT_1 STATE FIN has not been acked so
1394	* enter the CLOSING state.
1395	*/
1396	case TCPS_FIN_WAIT_1:
1397	tp->t_state = TCPS_CLOSING;
1398	break;
1399
1400	/*
1401	* In FIN_WAIT_2 state enter the TIME_WAIT state,
1402	* starting the time-wait timer, turning off the other
1403	* standard timers.
1404	*/
1405	case TCPS_FIN_WAIT_2:
1406	tp->t_state = TCPS_TIME_WAIT;
1407	tcp_canceltimers(tp);
1408	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1409	soisfdisconnected(so);
1410	break;
1411
1412	/*
1413	* In TIME_WAIT state restart the 2 MSL time_wait timer.
1414	*/
1415	case TCPS_TIME_WAIT:
1416	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1417	break;
1418	}
1419	}
1420
1421	/*
1422	* If this is a small packet, then ACK now - with Nagel
1423	* congestion avoidance sender won't send more until
1424	* he gets an ACK.
1425	*
1426	* See above.
1427	*/
1428	/* if (ti->ti_len && (unsigned)ti->ti_len < tp->t_maxseg) {
1429	*/
1430	/* if ((ti->ti_len && (unsigned)ti->ti_len < tp->t_maxseg &&
1431	* (so->so_iptos & IPTOS_LOWDELAY) == 0) \|\|
1432	* ((so->so_iptos & IPTOS_LOWDELAY) &&
1433	* ((struct tcpiphdr_2 *)ti)->first_char == (char)27)) {
1434	*/
1435	if (ti->ti_len && (unsigned)ti->ti_len <= 5 &&
1436	((struct tcpiphdr_2 *)ti)->first_char == (char)27) {
1437	tp->t_flags \|= TF_ACKNOW;
1438	}
1439
1440	if (mbuf_freed) {
1441	m_free(pData, m);
1442	}
1443	/*
1444	* Return any desired output.
1445	*/
1446	if (needoutput \|\| (tp->t_flags & TF_ACKNOW)) {
1447	(void) tcp_output(pData, tp);
1448	}
1449	return;
1450
1451	dropafterack:
1452	/*
1453	* Generate an ACK dropping incoming segment if it occupies
1454	* sequence space, where the ACK reflects our state.
1455	*/
1456	if (tiflags & TH_RST)
1457	goto drop;
1458	m_freem(pData, m);
1459	tp->t_flags \|= TF_ACKNOW;
1460	(void) tcp_output(pData, tp);
1461	return;
1462
1463	dropwithreset:
1464	/* reuses m if m!=NULL, m_free() unnecessary */
1465	if (tiflags & TH_ACK)
1466	tcp_respond(pData, tp, ti, m, (tcp_seq)0, ti->ti_ack, TH_RST);
1467	else {
1468	if (tiflags & TH_SYN) ti->ti_len++;
1469	tcp_respond(pData, tp, ti, m, ti->ti_seq+ti->ti_len, (tcp_seq)0,
1470	TH_RST\|TH_ACK);
1471	}
1472
1473	return;
1474
1475	drop:
1476	/*
1477	* Drop space held by incoming segment and return.
1478	*/
1479	m_free(pData, m);
1480
1481	return;
1482	}
1483
1484	/* , ts_present, ts_val, ts_ecr) */
1485	/* int *ts_present;
1486	* u_int32_t ts_val, ts_ecr;
1487	*/
1488	void
1489	tcp_dooptions(PNATState pData, struct tcpcb tp, u_char cp, int cnt, struct tcpiphdr *ti)
1490	{
1491	u_int16_t mss;
1492	int opt, optlen;
1493
1494	DEBUG_CALL("tcp_dooptions");
1495	DEBUG_ARGS((dfd," tp = %lx cnt=%i \n", (long )tp, cnt));
1496
1497	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1498	opt = cp[0];
1499	if (opt == TCPOPT_EOL)
1500	break;
1501	if (opt == TCPOPT_NOP)
1502	optlen = 1;
1503	else {
1504	optlen = cp[1];
1505	if (optlen <= 0)
1506	break;
1507	}
1508	switch (opt) {
1509
1510	default:
1511	continue;
1512
1513	case TCPOPT_MAXSEG:
1514	if (optlen != TCPOLEN_MAXSEG)
1515	continue;
1516	if (!(ti->ti_flags & TH_SYN))
1517	continue;
1518	memcpy((char ) &mss, (char ) cp + 2, sizeof(mss));
1519	NTOHS(mss);
1520	(void) tcp_mss(pData, tp, mss); /* sets t_maxseg */
1521	break;
1522
1523	/* case TCPOPT_WINDOW:
1524	* if (optlen != TCPOLEN_WINDOW)
1525	* continue;
1526	* if (!(ti->ti_flags & TH_SYN))
1527	* continue;
1528	* tp->t_flags \|= TF_RCVD_SCALE;
1529	* tp->requested_s_scale = min(cp[2], TCP_MAX_WINSHIFT);
1530	* break;
1531	*/
1532	/* case TCPOPT_TIMESTAMP:
1533	* if (optlen != TCPOLEN_TIMESTAMP)
1534	* continue;
1535	* *ts_present = 1;
1536	* memcpy((char ) ts_val, (char )cp + 2, sizeof(*ts_val));
1537	* NTOHL(*ts_val);
1538	* memcpy((char ) ts_ecr, (char )cp + 6, sizeof(*ts_ecr));
1539	* NTOHL(*ts_ecr);
1540	*
1541	/ /
1542	* * A timestamp received in a SYN makes
1543	* * it ok to send timestamp requests and replies.
1544	* */
1545	/* if (ti->ti_flags & TH_SYN) {
1546	* tp->t_flags \|= TF_RCVD_TSTMP;
1547	* tp->ts_recent = *ts_val;
1548	* tp->ts_recent_age = tcp_now;
1549	* }
1550	*/ break;
1551	}
1552	}
1553	}
1554
1555
1556	/*
1557	* Pull out of band byte out of a segment so
1558	* it doesn't appear in the user's data queue.
1559	* It is still reflected in the segment length for
1560	* sequencing purposes.
1561	*/
1562
1563	#ifdef notdef
1564
1565	void
1566	tcp_pulloutofband(so, ti, m)
1567	struct socket *so;
1568	struct tcpiphdr *ti;
1569	register struct mbuf *m;
1570	{
1571	int cnt = ti->ti_urp - 1;
1572
1573	while (cnt >= 0) {
1574	if (m->m_len > cnt) {
1575	char *cp = mtod(m, caddr_t) + cnt;
1576	struct tcpcb *tp = sototcpcb(so);
1577
1578	tp->t_iobc = *cp;
1579	tp->t_oobflags \|= TCPOOB_HAVEDATA;
1580	memcpy(sp, cp+1, (unsigned)(m->m_len - cnt - 1));
1581	m->m_len--;
1582	return;
1583	}
1584	cnt -= m->m_len;
1585	m = m->m_next; /* XXX WRONG! Fix it! */
1586	if (m == 0)
1587	break;
1588	}
1589	panic("tcp_pulloutofband");
1590	}
1591
1592	#endif /* notdef */
1593
1594	/*
1595	* Collect new round-trip time estimate
1596	* and update averages and current timeout.
1597	*/
1598
1599	void
1600	tcp_xmit_timer(PNATState pData, register struct tcpcb *tp, int rtt)
1601	{
1602	register short delta;
1603
1604	DEBUG_CALL("tcp_xmit_timer");
1605	DEBUG_ARG("tp = %lx", (long)tp);
1606	DEBUG_ARG("rtt = %d", rtt);
1607
1608	tcpstat.tcps_rttupdated++;
1609	if (tp->t_srtt != 0) {
1610	/*
1611	* srtt is stored as fixed point with 3 bits after the
1612	* binary point (i.e., scaled by 8). The following magic
1613	* is equivalent to the smoothing algorithm in rfc793 with
1614	* an alpha of .875 (srtt = rtt/8 + srtt*7/8 in fixed
1615	* point). Adjust rtt to origin 0.
1616	*/
1617	delta = rtt - 1 - (tp->t_srtt >> TCP_RTT_SHIFT);
1618	if ((tp->t_srtt += delta) <= 0)
1619	tp->t_srtt = 1;
1620	/*
1621	* We accumulate a smoothed rtt variance (actually, a
1622	* smoothed mean difference), then set the retransmit
1623	* timer to smoothed rtt + 4 times the smoothed variance.
1624	* rttvar is stored as fixed point with 2 bits after the
1625	* binary point (scaled by 4). The following is
1626	* equivalent to rfc793 smoothing with an alpha of .75
1627	* (rttvar = rttvar*3/4 + \|delta\| / 4). This replaces
1628	* rfc793's wired-in beta.
1629	*/
1630	if (delta < 0)
1631	delta = -delta;
1632	delta -= (tp->t_rttvar >> TCP_RTTVAR_SHIFT);
1633	if ((tp->t_rttvar += delta) <= 0)
1634	tp->t_rttvar = 1;
1635	} else {
1636	/*
1637	* No rtt measurement yet - use the unsmoothed rtt.
1638	* Set the variance to half the rtt (so our first
1639	* retransmit happens at 3*rtt).
1640	*/
1641	tp->t_srtt = rtt << TCP_RTT_SHIFT;
1642	tp->t_rttvar = rtt << (TCP_RTTVAR_SHIFT - 1);
1643	}
1644	tp->t_rtt = 0;
1645	tp->t_rxtshift = 0;
1646
1647	/*
1648	* the retransmit should happen at rtt + 4 * rttvar.
1649	* Because of the way we do the smoothing, srtt and rttvar
1650	* will each average +1/2 tick of bias. When we compute
1651	* the retransmit timer, we want 1/2 tick of rounding and
1652	* 1 extra tick because of +-1/2 tick uncertainty in the
1653	* firing of the timer. The bias will give us exactly the
1654	* 1.5 tick we need. But, because the bias is
1655	* statistical, we have to test that we don't drop below
1656	* the minimum feasible timer (which is 2 ticks).
1657	*/
1658	TCPT_RANGESET(tp->t_rxtcur, TCP_REXMTVAL(tp),
1659	(short)tp->t_rttmin, TCPTV_REXMTMAX); /* XXX */
1660
1661	/*
1662	* We received an ack for a packet that wasn't retransmitted;
1663	* it is probably safe to discard any error indications we've
1664	* received recently. This isn't quite right, but close enough
1665	* for now (a route might have failed after we sent a segment,
1666	* and the return path might not be symmetrical).
1667	*/
1668	tp->t_softerror = 0;
1669	}
1670
1671	/*
1672	* Determine a reasonable value for maxseg size.
1673	* If the route is known, check route for mtu.
1674	* If none, use an mss that can be handled on the outgoing
1675	* interface without forcing IP to fragment; if bigger than
1676	* an mbuf cluster (MCLBYTES), round down to nearest multiple of MCLBYTES
1677	* to utilize large mbufs. If no route is found, route has no mtu,
1678	* or the destination isn't local, use a default, hopefully conservative
1679	* size (usually 512 or the default IP max size, but no more than the mtu
1680	* of the interface), as we can't discover anything about intervening
1681	* gateways or networks. We also initialize the congestion/slow start
1682	* window to be a single segment if the destination isn't local.
1683	* While looking at the routing entry, we also initialize other path-dependent
1684	* parameters from pre-set or cached values in the routing entry.
1685	*/
1686
1687	int
1688	tcp_mss(PNATState pData, register struct tcpcb *tp, u_int offer)
1689	{
1690	struct socket *so = tp->t_socket;
1691	int mss;
1692
1693	DEBUG_CALL("tcp_mss");
1694	DEBUG_ARG("tp = %lx", (long)tp);
1695	DEBUG_ARG("offer = %d", offer);
1696
1697	mss = min(if_mtu, if_mru) - sizeof(struct tcpiphdr);
1698	if (offer)
1699	mss = min(mss, offer);
1700	mss = max(mss, 32);
1701	if (mss < tp->t_maxseg \|\| offer != 0)
1702	tp->t_maxseg = mss;
1703
1704	tp->snd_cwnd = mss;
1705
1706	sbreserve(&so->so_snd, tcp_sndspace+((tcp_sndspace%mss)?(mss-(tcp_sndspace%mss)):0));
1707	sbreserve(&so->so_rcv, tcp_rcvspace+((tcp_rcvspace%mss)?(mss-(tcp_rcvspace%mss)):0));
1708
1709	DEBUG_MISC((dfd, " returning mss = %d\n", mss));
1710
1711	return mss;
1712	}

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

source: vbox/trunk/src/VBox/Devices/Network/slirp/tcp_input.c@ 14160

Download in other formats: