tcp_input.c@ 13984

Last change on this file since 13984 was 13984, checked in by vboxsync, 16 years ago
simplefied syncronization work on Unixes only
Property svn:eol-style set to `native`
File size: 48.4 KB

Line
1	/*
2	* Copyright (c) 1982, 1986, 1988, 1990, 1993, 1994
3	* The Regents of the University of California. All rights reserved.
4	*
5	* Redistribution and use in source and binary forms, with or without
6	* modification, are permitted provided that the following conditions
7	* are met:
8	* 1. Redistributions of source code must retain the above copyright
9	* notice, this list of conditions and the following disclaimer.
10	* 2. Redistributions in binary form must reproduce the above copyright
11	* notice, this list of conditions and the following disclaimer in the
12	* documentation and/or other materials provided with the distribution.
13	* 3. All advertising materials mentioning features or use of this software
14	* must display the following acknowledgement:
15	* This product includes software developed by the University of
16	* California, Berkeley and its contributors.
17	* 4. Neither the name of the University nor the names of its contributors
18	* may be used to endorse or promote products derived from this software
19	* without specific prior written permission.
20	*
21	* THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22	* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23	* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24	* ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25	* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26	* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27	* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28	* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29	* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30	* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31	* SUCH DAMAGE.
32	*
33	* @(#)tcp_input.c 8.5 (Berkeley) 4/10/94
34	* tcp_input.c,v 1.10 1994/10/13 18:36:32 wollman Exp
35	*/
36
37	/*
38	* Changes and additions relating to SLiRP
39	* Copyright (c) 1995 Danny Gasparovski.
40	*
41	* Please read the file COPYRIGHT for the
42	* terms and conditions of the copyright.
43	*/
44
45	#include <slirp.h>
46	#include "ip_icmp.h"
47
48
49	#define TCP_PAWS_IDLE (24 * 24 * 60 * 60 * PR_SLOWHZ)
50
51	/* for modulo comparisons of timestamps */
52	#define TSTMP_LT(a,b) ((int)((a)-(b)) < 0)
53	#define TSTMP_GEQ(a,b) ((int)((a)-(b)) >= 0)
54
55	/*
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
394	sbreserve(&so->so_snd, tcp_sndspace);
395	sbreserve(&so->so_rcv, tcp_rcvspace);
396
397	/* tcp_last_so = so; / / XXX ? */
398	/* tp = sototcpcb(so); */
399
400	so->so_laddr = ti->ti_src;
401	so->so_lport = ti->ti_sport;
402	so->so_faddr = ti->ti_dst;
403	so->so_fport = ti->ti_dport;
404
405	if ((so->so_iptos = tcp_tos(so)) == 0)
406	so->so_iptos = ((struct ip *)ti)->ip_tos;
407
408	tp = sototcpcb(so);
409	tp->t_state = TCPS_LISTEN;
410	}
411
412	/*
413	* If this is a still-connecting socket, this probably
414	* a retransmit of the SYN. Whether it's a retransmit SYN
415	* or something else, we nuke it.
416	*/
417	if (so->so_state & SS_ISFCONNECTING)
418	goto drop;
419
420	tp = sototcpcb(so);
421
422	/* XXX Should never fail */
423	if (tp == 0)
424	goto dropwithreset;
425	if (tp->t_state == TCPS_CLOSED)
426	goto drop;
427
428	/* Unscale the window into a 32-bit value. */
429	/* if ((tiflags & TH_SYN) == 0)
430	* tiwin = ti->ti_win << tp->snd_scale;
431	* else
432	*/
433	tiwin = ti->ti_win;
434
435	/*
436	* Segment received on connection.
437	* Reset idle time and keep-alive timer.
438	*/
439	tp->t_idle = 0;
440	if (so_options)
441	tp->t_timer[TCPT_KEEP] = tcp_keepintvl;
442	else
443	tp->t_timer[TCPT_KEEP] = tcp_keepidle;
444
445	/*
446	* Process options if not in LISTEN state,
447	* else do it below (after getting remote address).
448	*/
449	if (optp && tp->t_state != TCPS_LISTEN)
450	tcp_dooptions(pData, tp, (u_char *)optp, optlen, ti);
451	/* , */
452	/* &ts_present, &ts_val, &ts_ecr); */
453
454	/*
455	* Header prediction: check for the two common cases
456	* of a uni-directional data xfer. If the packet has
457	* no control flags, is in-sequence, the window didn't
458	* change and we're not retransmitting, it's a
459	* candidate. If the length is zero and the ack moved
460	* forward, we're the sender side of the xfer. Just
461	* free the data acked & wake any higher level process
462	* that was blocked waiting for space. If the length
463	* is non-zero and the ack didn't move, we're the
464	* receiver side. If we're getting packets in-order
465	* (the reassembly queue is empty), add the data to
466	* the socket buffer and note that we need a delayed ack.
467	*
468	* XXX Some of these tests are not needed
469	* eg: the tiwin == tp->snd_wnd prevents many more
470	* predictions.. with no real advantage..
471	*/
472	if (tp->t_state == TCPS_ESTABLISHED &&
473	(tiflags & (TH_SYN\|TH_FIN\|TH_RST\|TH_URG\|TH_ACK)) == TH_ACK &&
474	/* (!ts_present \|\| TSTMP_GEQ(ts_val, tp->ts_recent)) && */
475	ti->ti_seq == tp->rcv_nxt &&
476	tiwin && tiwin == tp->snd_wnd &&
477	tp->snd_nxt == tp->snd_max) {
478	/*
479	* If last ACK falls within this segment's sequence numbers,
480	* record the timestamp.
481	*/
482	/* if (ts_present && SEQ_LEQ(ti->ti_seq, tp->last_ack_sent) &&
483	* SEQ_LT(tp->last_ack_sent, ti->ti_seq + ti->ti_len)) {
484	* tp->ts_recent_age = tcp_now;
485	* tp->ts_recent = ts_val;
486	* }
487	*/
488	if (ti->ti_len == 0) {
489	if (SEQ_GT(ti->ti_ack, tp->snd_una) &&
490	SEQ_LEQ(ti->ti_ack, tp->snd_max) &&
491	tp->snd_cwnd >= tp->snd_wnd) {
492	/*
493	* this is a pure ack for outstanding data.
494	*/
495	++tcpstat.tcps_predack;
496	/* if (ts_present)
497	* tcp_xmit_timer(tp, tcp_now-ts_ecr+1);
498	* else
499	*/ if (tp->t_rtt &&
500	SEQ_GT(ti->ti_ack, tp->t_rtseq))
501	tcp_xmit_timer(pData, tp, tp->t_rtt);
502	acked = ti->ti_ack - tp->snd_una;
503	tcpstat.tcps_rcvackpack++;
504	tcpstat.tcps_rcvackbyte += acked;
505	sbdrop(&so->so_snd, acked);
506	tp->snd_una = ti->ti_ack;
507	m_freem(pData, m);
508
509	/*
510	* If all outstanding data are acked, stop
511	* retransmit timer, otherwise restart timer
512	* using current (possibly backed-off) value.
513	* If process is waiting for space,
514	* wakeup/selwakeup/signal. If data
515	* are ready to send, let tcp_output
516	* decide between more output or persist.
517	*/
518	if (tp->snd_una == tp->snd_max)
519	tp->t_timer[TCPT_REXMT] = 0;
520	else if (tp->t_timer[TCPT_PERSIST] == 0)
521	tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
522
523	/*
524	* There's room in so_snd, sowwakup will read()
525	* from the socket if we can
526	*/
527	/* if (so->so_snd.sb_flags & SB_NOTIFY)
528	* sowwakeup(so);
529	*/
530	/*
531	* This is called because sowwakeup might have
532	* put data into so_snd. Since we don't so sowwakeup,
533	* we don't need this.. XXX???
534	*/
535	if (so->so_snd.sb_cc)
536	(void) tcp_output(pData, tp);
537
538	return;
539	}
540	} else if (ti->ti_ack == tp->snd_una &&
541	u32_to_ptr(pData, tp->seg_next, struct tcpcb *) == tp &&
542	ti->ti_len <= sbspace(&so->so_rcv)) {
543	/*
544	* this is a pure, in-sequence data packet
545	* with nothing on the reassembly queue and
546	* we have enough buffer space to take it.
547	*/
548	++tcpstat.tcps_preddat;
549	tp->rcv_nxt += ti->ti_len;
550	tcpstat.tcps_rcvpack++;
551	tcpstat.tcps_rcvbyte += ti->ti_len;
552	/*
553	* Add data to socket buffer.
554	*/
555	if (so->so_emu) {
556	if (tcp_emu(pData, so,m)) sbappend(pData, so, m);
557	} else
558	sbappend(pData, so, m);
559
560	/*
561	* XXX This is called when data arrives. Later, check
562	* if we can actually write() to the socket
563	* XXX Need to check? It's be NON_BLOCKING
564	*/
565	/* sorwakeup(so); */
566
567	/*
568	* If this is a short packet, then ACK now - with Nagel
569	* congestion avoidance sender won't send more until
570	* he gets an ACK.
571	*
572	* It is better to not delay acks at all to maximize
573	* TCP throughput. See RFC 2581.
574	*/
575	tp->t_flags \|= TF_ACKNOW;
576	tcp_output(pData, tp);
577	return;
578	}
579	} /* header prediction */
580	/*
581	* Calculate amount of space in receive window,
582	* and then do TCP input processing.
583	* Receive window is amount of space in rcv queue,
584	* but not less than advertised window.
585	*/
586	{ int win;
587	win = sbspace(&so->so_rcv);
588	if (win < 0)
589	win = 0;
590	tp->rcv_wnd = max(win, (int)(tp->rcv_adv - tp->rcv_nxt));
591	}
592
593	switch (tp->t_state) {
594
595	/*
596	* If the state is LISTEN then ignore segment if it contains an RST.
597	* If the segment contains an ACK then it is bad and send a RST.
598	* If it does not contain a SYN then it is not interesting; drop it.
599	* Don't bother responding if the destination was a broadcast.
600	* Otherwise initialize tp->rcv_nxt, and tp->irs, select an initial
601	* tp->iss, and send a segment:
602	* <SEQ=ISS><ACK=RCV_NXT><CTL=SYN,ACK>
603	* Also initialize tp->snd_nxt to tp->iss+1 and tp->snd_una to tp->iss.
604	* Fill in remote peer address fields if not previously specified.
605	* Enter SYN_RECEIVED state, and process any other fields of this
606	* segment in this state.
607	*/
608	case TCPS_LISTEN: {
609
610	if (tiflags & TH_RST)
611	goto drop;
612	if (tiflags & TH_ACK)
613	goto dropwithreset;
614	if ((tiflags & TH_SYN) == 0)
615	goto drop;
616
617	/*
618	* This has way too many gotos...
619	* But a bit of spaghetti code never hurt anybody :)
620	*/
621
622	/*
623	* If this is destined for the control address, then flag to
624	* tcp_ctl once connected, otherwise connect
625	*/
626	if ((so->so_faddr.s_addr&htonl(pData->netmask)) == special_addr.s_addr) {
627	int lastbyte=ntohl(so->so_faddr.s_addr) & ~pData->netmask;
628	if (lastbyte!=CTL_ALIAS && lastbyte!=CTL_DNS) {
629	#if 0
630	if(lastbyte==CTL_CMD \|\| lastbyte==CTL_EXEC) {
631	/* Command or exec adress */
632	so->so_state \|= SS_CTL;
633	} else
634	#endif
635	{
636	/* May be an add exec */
637	struct ex_list *ex_ptr;
638	for(ex_ptr = exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next) {
639	if(ex_ptr->ex_fport == so->so_fport &&
640	lastbyte == ex_ptr->ex_addr) {
641	so->so_state \|= SS_CTL;
642	break;
643	}
644	}
645	}
646	if(so->so_state & SS_CTL) goto cont_input;
647	}
648	/* CTL_ALIAS: Do nothing, tcp_fconnect will be called on it */
649	}
650
651	if (so->so_emu & EMU_NOCONNECT) {
652	so->so_emu &= ~EMU_NOCONNECT;
653	goto cont_input;
654	}
655
656	if((tcp_fconnect(pData, so) == -1) && (errno != EINPROGRESS) && (errno != EWOULDBLOCK)) {
657	u_char code=ICMP_UNREACH_NET;
658	DEBUG_MISC((dfd," tcp fconnect errno = %d-%s\n",
659	errno,strerror(errno)));
660	if(errno == ECONNREFUSED) {
661	/* ACK the SYN, send RST to refuse the connection */
662	tcp_respond(pData, tp, ti, m, ti->ti_seq+1, (tcp_seq)0,
663	TH_RST\|TH_ACK);
664	} else {
665	if(errno == EHOSTUNREACH) code=ICMP_UNREACH_HOST;
666	HTONL(ti->ti_seq); /* restore tcp header */
667	HTONL(ti->ti_ack);
668	HTONS(ti->ti_win);
669	HTONS(ti->ti_urp);
670	m->m_data -= sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
671	m->m_len += sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
672	*ip=save_ip;
673	icmp_error(pData, m, ICMP_UNREACH,code, 0,strerror(errno));
674	}
675	tp = tcp_close(pData, tp);
676	m_free(pData, m);
677	} else {
678	/*
679	* Haven't connected yet, save the current mbuf
680	* and ti, and return
681	* XXX Some OS's don't tell us whether the connect()
682	* succeeded or not. So we must time it out.
683	*/
684	so->so_m = m;
685	so->so_ti = ti;
686	tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT;
687	tp->t_state = TCPS_SYN_RECEIVED;
688	}
689	return;
690
691	cont_conn:
692	/* m==NULL
693	* Check if the connect succeeded
694	*/
695	if (so->so_state & SS_NOFDREF) {
696	tp = tcp_close(pData, tp);
697	goto dropwithreset;
698	}
699	cont_input:
700	tcp_template(tp);
701
702	if (optp)
703	tcp_dooptions(pData, tp, (u_char *)optp, optlen, ti);
704	/* , */
705	/* &ts_present, &ts_val, &ts_ecr); */
706
707	if (iss)
708	tp->iss = iss;
709	else
710	tp->iss = tcp_iss;
711	tcp_iss += TCP_ISSINCR/2;
712	tp->irs = ti->ti_seq;
713	tcp_sendseqinit(tp);
714	tcp_rcvseqinit(tp);
715	tp->t_flags \|= TF_ACKNOW;
716	tp->t_state = TCPS_SYN_RECEIVED;
717	tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT;
718	tcpstat.tcps_accepts++;
719	goto trimthenstep6;
720	} /* case TCPS_LISTEN */
721
722	/*
723	* If the state is SYN_SENT:
724	* if seg contains an ACK, but not for our SYN, drop the input.
725	* if seg contains a RST, then drop the connection.
726	* if seg does not contain SYN, then drop it.
727	* Otherwise this is an acceptable SYN segment
728	* initialize tp->rcv_nxt and tp->irs
729	* if seg contains ack then advance tp->snd_una
730	* if SYN has been acked change to ESTABLISHED else SYN_RCVD state
731	* arrange for segment to be acked (eventually)
732	* continue processing rest of data/controls, beginning with URG
733	*/
734	case TCPS_SYN_SENT:
735	if ((tiflags & TH_ACK) &&
736	(SEQ_LEQ(ti->ti_ack, tp->iss) \|\|
737	SEQ_GT(ti->ti_ack, tp->snd_max)))
738	goto dropwithreset;
739
740	if (tiflags & TH_RST) {
741	if (tiflags & TH_ACK)
742	tp = tcp_drop(pData, tp,0); /* XXX Check t_softerror! */
743	goto drop;
744	}
745
746	if ((tiflags & TH_SYN) == 0)
747	goto drop;
748	if (tiflags & TH_ACK) {
749	tp->snd_una = ti->ti_ack;
750	if (SEQ_LT(tp->snd_nxt, tp->snd_una))
751	tp->snd_nxt = tp->snd_una;
752	}
753
754	tp->t_timer[TCPT_REXMT] = 0;
755	tp->irs = ti->ti_seq;
756	tcp_rcvseqinit(tp);
757	tp->t_flags \|= TF_ACKNOW;
758	if (tiflags & TH_ACK && SEQ_GT(tp->snd_una, tp->iss)) {
759	tcpstat.tcps_connects++;
760	soisfconnected(so);
761	tp->t_state = TCPS_ESTABLISHED;
762
763	/* Do window scaling on this connection? */
764	/* if ((tp->t_flags & (TF_RCVD_SCALE\|TF_REQ_SCALE)) ==
765	* (TF_RCVD_SCALE\|TF_REQ_SCALE)) {
766	* tp->snd_scale = tp->requested_s_scale;
767	* tp->rcv_scale = tp->request_r_scale;
768	* }
769	*/
770	(void) tcp_reass(pData, tp, (struct tcpiphdr *)0,
771	(struct mbuf *)0);
772	/*
773	* if we didn't have to retransmit the SYN,
774	* use its rtt as our initial srtt & rtt var.
775	*/
776	if (tp->t_rtt)
777	tcp_xmit_timer(pData, tp, tp->t_rtt);
778	} else
779	tp->t_state = TCPS_SYN_RECEIVED;
780
781	trimthenstep6:
782	/*
783	* Advance ti->ti_seq to correspond to first data byte.
784	* If data, trim to stay within window,
785	* dropping FIN if necessary.
786	*/
787	ti->ti_seq++;
788	if (ti->ti_len > tp->rcv_wnd) {
789	todrop = ti->ti_len - tp->rcv_wnd;
790	m_adj(m, -todrop);
791	ti->ti_len = tp->rcv_wnd;
792	tiflags &= ~TH_FIN;
793	tcpstat.tcps_rcvpackafterwin++;
794	tcpstat.tcps_rcvbyteafterwin += todrop;
795	}
796	tp->snd_wl1 = ti->ti_seq - 1;
797	tp->rcv_up = ti->ti_seq;
798	goto step6;
799	} /* switch tp->t_state */
800	/*
801	* States other than LISTEN or SYN_SENT.
802	* First check timestamp, if present.
803	* Then check that at least some bytes of segment are within
804	* receive window. If segment begins before rcv_nxt,
805	* drop leading data (and SYN); if nothing left, just ack.
806	*
807	* RFC 1323 PAWS: If we have a timestamp reply on this segment
808	* and it's less than ts_recent, drop it.
809	*/
810	/* if (ts_present && (tiflags & TH_RST) == 0 && tp->ts_recent &&
811	* TSTMP_LT(ts_val, tp->ts_recent)) {
812	*
813	/ / Check to see if ts_recent is over 24 days old. */
814	/* if ((int)(tcp_now - tp->ts_recent_age) > TCP_PAWS_IDLE) {
815	/ /
816	* * Invalidate ts_recent. If this segment updates
817	* * ts_recent, the age will be reset later and ts_recent
818	* * will get a valid value. If it does not, setting
819	* * ts_recent to zero will at least satisfy the
820	* * requirement that zero be placed in the timestamp
821	* * echo reply when ts_recent isn't valid. The
822	* * age isn't reset until we get a valid ts_recent
823	* * because we don't want out-of-order segments to be
824	* * dropped when ts_recent is old.
825	* */
826	/* tp->ts_recent = 0;
827	* } else {
828	* tcpstat.tcps_rcvduppack++;
829	* tcpstat.tcps_rcvdupbyte += ti->ti_len;
830	* tcpstat.tcps_pawsdrop++;
831	* goto dropafterack;
832	* }
833	* }
834	*/
835
836	todrop = tp->rcv_nxt - ti->ti_seq;
837	if (todrop > 0) {
838	if (tiflags & TH_SYN) {
839	tiflags &= ~TH_SYN;
840	ti->ti_seq++;
841	if (ti->ti_urp > 1)
842	ti->ti_urp--;
843	else
844	tiflags &= ~TH_URG;
845	todrop--;
846	}
847	/*
848	* Following if statement from Stevens, vol. 2, p. 960.
849	*/
850	if (todrop > ti->ti_len
851	\|\| (todrop == ti->ti_len && (tiflags & TH_FIN) == 0)) {
852	/*
853	* Any valid FIN must be to the left of the window.
854	* At this point the FIN must be a duplicate or out
855	* of sequence; drop it.
856	*/
857	tiflags &= ~TH_FIN;
858
859	/*
860	* Send an ACK to resynchronize and drop any data.
861	* But keep on processing for RST or ACK.
862	*/
863	tp->t_flags \|= TF_ACKNOW;
864	todrop = ti->ti_len;
865	tcpstat.tcps_rcvduppack++;
866	tcpstat.tcps_rcvdupbyte += todrop;
867	} else {
868	tcpstat.tcps_rcvpartduppack++;
869	tcpstat.tcps_rcvpartdupbyte += todrop;
870	}
871	m_adj(m, todrop);
872	ti->ti_seq += todrop;
873	ti->ti_len -= todrop;
874	if (ti->ti_urp > todrop)
875	ti->ti_urp -= todrop;
876	else {
877	tiflags &= ~TH_URG;
878	ti->ti_urp = 0;
879	}
880	}
881	/*
882	* If new data are received on a connection after the
883	* user processes are gone, then RST the other end.
884	*/
885	if ((so->so_state & SS_NOFDREF) &&
886	tp->t_state > TCPS_CLOSE_WAIT && ti->ti_len) {
887	tp = tcp_close(pData, tp);
888	tcpstat.tcps_rcvafterclose++;
889	goto dropwithreset;
890	}
891
892	/*
893	* If segment ends after window, drop trailing data
894	* (and PUSH and FIN); if nothing left, just ACK.
895	*/
896	todrop = (ti->ti_seq+ti->ti_len) - (tp->rcv_nxt+tp->rcv_wnd);
897	if (todrop > 0) {
898	tcpstat.tcps_rcvpackafterwin++;
899	if (todrop >= ti->ti_len) {
900	tcpstat.tcps_rcvbyteafterwin += ti->ti_len;
901	/*
902	* If a new connection request is received
903	* while in TIME_WAIT, drop the old connection
904	* and start over if the sequence numbers
905	* are above the previous ones.
906	*/
907	if (tiflags & TH_SYN &&
908	tp->t_state == TCPS_TIME_WAIT &&
909	SEQ_GT(ti->ti_seq, tp->rcv_nxt)) {
910	iss = tp->rcv_nxt + TCP_ISSINCR;
911	tp = tcp_close(pData, tp);
912	goto findso;
913	}
914	/*
915	* If window is closed can only take segments at
916	* window edge, and have to drop data and PUSH from
917	* incoming segments. Continue processing, but
918	* remember to ack. Otherwise, drop segment
919	* and ack.
920	*/
921	if (tp->rcv_wnd == 0 && ti->ti_seq == tp->rcv_nxt) {
922	tp->t_flags \|= TF_ACKNOW;
923	tcpstat.tcps_rcvwinprobe++;
924	} else
925	goto dropafterack;
926	} else
927	tcpstat.tcps_rcvbyteafterwin += todrop;
928	m_adj(m, -todrop);
929	ti->ti_len -= todrop;
930	tiflags &= ~(TH_PUSH\|TH_FIN);
931	}
932
933	/*
934	* If last ACK falls within this segment's sequence numbers,
935	* record its timestamp.
936	*/
937	/* if (ts_present && SEQ_LEQ(ti->ti_seq, tp->last_ack_sent) &&
938	* SEQ_LT(tp->last_ack_sent, ti->ti_seq + ti->ti_len +
939	* ((tiflags & (TH_SYN\|TH_FIN)) != 0))) {
940	* tp->ts_recent_age = tcp_now;
941	* tp->ts_recent = ts_val;
942	* }
943	*/
944
945	/*
946	* If the RST bit is set examine the state:
947	* SYN_RECEIVED STATE:
948	* If passive open, return to LISTEN state.
949	* If active open, inform user that connection was refused.
950	* ESTABLISHED, FIN_WAIT_1, FIN_WAIT2, CLOSE_WAIT STATES:
951	* Inform user that connection was reset, and close tcb.
952	* CLOSING, LAST_ACK, TIME_WAIT STATES
953	* Close the tcb.
954	*/
955	if (tiflags&TH_RST) switch (tp->t_state) {
956
957	case TCPS_SYN_RECEIVED:
958	/* so->so_error = ECONNREFUSED; */
959	goto close;
960
961	case TCPS_ESTABLISHED:
962	case TCPS_FIN_WAIT_1:
963	case TCPS_FIN_WAIT_2:
964	case TCPS_CLOSE_WAIT:
965	/* so->so_error = ECONNRESET; */
966	close:
967	tp->t_state = TCPS_CLOSED;
968	tcpstat.tcps_drops++;
969	tp = tcp_close(pData, tp);
970	goto drop;
971
972	case TCPS_CLOSING:
973	case TCPS_LAST_ACK:
974	case TCPS_TIME_WAIT:
975	tp = tcp_close(pData, tp);
976	goto drop;
977	}
978
979	/*
980	* If a SYN is in the window, then this is an
981	* error and we send an RST and drop the connection.
982	*/
983	if (tiflags & TH_SYN) {
984	tp = tcp_drop(pData, tp,0);
985	goto dropwithreset;
986	}
987
988	/*
989	* If the ACK bit is off we drop the segment and return.
990	*/
991	if ((tiflags & TH_ACK) == 0) goto drop;
992
993	/*
994	* Ack processing.
995	*/
996	switch (tp->t_state) {
997	/*
998	* In SYN_RECEIVED state if the ack ACKs our SYN then enter
999	* ESTABLISHED state and continue processing, otherwise
1000	* send an RST. una<=ack<=max
1001	*/
1002	case TCPS_SYN_RECEIVED:
1003
1004	if (SEQ_GT(tp->snd_una, ti->ti_ack) \|\|
1005	SEQ_GT(ti->ti_ack, tp->snd_max))
1006	goto dropwithreset;
1007	tcpstat.tcps_connects++;
1008	tp->t_state = TCPS_ESTABLISHED;
1009	/*
1010	* The sent SYN is ack'ed with our sequence number +1
1011	* The first data byte already in the buffer will get
1012	* lost if no correction is made. This is only needed for
1013	* SS_CTL since the buffer is empty otherwise.
1014	* tp->snd_una++; or:
1015	*/
1016	tp->snd_una=ti->ti_ack;
1017	if (so->so_state & SS_CTL) {
1018	/* So tcp_ctl reports the right state */
1019	ret = tcp_ctl(pData, so);
1020	if (ret == 1) {
1021	soisfconnected(so);
1022	so->so_state &= ~SS_CTL; /* success XXX */
1023	} else if (ret == 2) {
1024	so->so_state = SS_NOFDREF; /* CTL_CMD */
1025	} else {
1026	needoutput = 1;
1027	tp->t_state = TCPS_FIN_WAIT_1;
1028	}
1029	} else {
1030	soisfconnected(so);
1031	}
1032
1033	/* Do window scaling? */
1034	/* if ((tp->t_flags & (TF_RCVD_SCALE\|TF_REQ_SCALE)) ==
1035	* (TF_RCVD_SCALE\|TF_REQ_SCALE)) {
1036	* tp->snd_scale = tp->requested_s_scale;
1037	* tp->rcv_scale = tp->request_r_scale;
1038	* }
1039	*/
1040	(void) tcp_reass(pData, tp, (struct tcpiphdr )0, (struct mbuf )0);
1041	tp->snd_wl1 = ti->ti_seq - 1;
1042	/* Avoid ack processing; snd_una==ti_ack => dup ack */
1043	goto synrx_to_est;
1044	/* fall into ... */
1045
1046	/*
1047	* In ESTABLISHED state: drop duplicate ACKs; ACK out of range
1048	* ACKs. If the ack is in the range
1049	* tp->snd_una < ti->ti_ack <= tp->snd_max
1050	* then advance tp->snd_una to ti->ti_ack and drop
1051	* data from the retransmission queue. If this ACK reflects
1052	* more up to date window information we update our window information.
1053	*/
1054	case TCPS_ESTABLISHED:
1055	case TCPS_FIN_WAIT_1:
1056	case TCPS_FIN_WAIT_2:
1057	case TCPS_CLOSE_WAIT:
1058	case TCPS_CLOSING:
1059	case TCPS_LAST_ACK:
1060	case TCPS_TIME_WAIT:
1061
1062	if (SEQ_LEQ(ti->ti_ack, tp->snd_una)) {
1063	if (ti->ti_len == 0 && tiwin == tp->snd_wnd) {
1064	tcpstat.tcps_rcvdupack++;
1065	DEBUG_MISC((dfd," dup ack m = %lx so = %lx \n",
1066	(long )m, (long )so));
1067	/*
1068	* If we have outstanding data (other than
1069	* a window probe), this is a completely
1070	* duplicate ack (ie, window info didn't
1071	* change), the ack is the biggest we've
1072	* seen and we've seen exactly our rexmt
1073	* threshold of them, assume a packet
1074	* has been dropped and retransmit it.
1075	* Kludge snd_nxt & the congestion
1076	* window so we send only this one
1077	* packet.
1078	*
1079	* We know we're losing at the current
1080	* window size so do congestion avoidance
1081	* (set ssthresh to half the current window
1082	* and pull our congestion window back to
1083	* the new ssthresh).
1084	*
1085	* Dup acks mean that packets have left the
1086	* network (they're now cached at the receiver)
1087	* so bump cwnd by the amount in the receiver
1088	* to keep a constant cwnd packets in the
1089	* network.
1090	*/
1091	if (tp->t_timer[TCPT_REXMT] == 0 \|\|
1092	ti->ti_ack != tp->snd_una)
1093	tp->t_dupacks = 0;
1094	else if (++tp->t_dupacks == tcprexmtthresh) {
1095	tcp_seq onxt = tp->snd_nxt;
1096	u_int win =
1097	min(tp->snd_wnd, tp->snd_cwnd) / 2 /
1098	tp->t_maxseg;
1099
1100	if (win < 2)
1101	win = 2;
1102	tp->snd_ssthresh = win * tp->t_maxseg;
1103	tp->t_timer[TCPT_REXMT] = 0;
1104	tp->t_rtt = 0;
1105	tp->snd_nxt = ti->ti_ack;
1106	tp->snd_cwnd = tp->t_maxseg;
1107	(void) tcp_output(pData, tp);
1108	tp->snd_cwnd = tp->snd_ssthresh +
1109	tp->t_maxseg * tp->t_dupacks;
1110	if (SEQ_GT(onxt, tp->snd_nxt))
1111	tp->snd_nxt = onxt;
1112	goto drop;
1113	} else if (tp->t_dupacks > tcprexmtthresh) {
1114	tp->snd_cwnd += tp->t_maxseg;
1115	(void) tcp_output(pData, tp);
1116	goto drop;
1117	}
1118	} else
1119	tp->t_dupacks = 0;
1120	break;
1121	}
1122	synrx_to_est:
1123	/*
1124	* If the congestion window was inflated to account
1125	* for the other side's cached packets, retract it.
1126	*/
1127	if (tp->t_dupacks > tcprexmtthresh &&
1128	tp->snd_cwnd > tp->snd_ssthresh)
1129	tp->snd_cwnd = tp->snd_ssthresh;
1130	tp->t_dupacks = 0;
1131	if (SEQ_GT(ti->ti_ack, tp->snd_max)) {
1132	tcpstat.tcps_rcvacktoomuch++;
1133	goto dropafterack;
1134	}
1135	acked = ti->ti_ack - tp->snd_una;
1136	tcpstat.tcps_rcvackpack++;
1137	tcpstat.tcps_rcvackbyte += acked;
1138
1139	/*
1140	* If we have a timestamp reply, update smoothed
1141	* round trip time. If no timestamp is present but
1142	* transmit timer is running and timed sequence
1143	* number was acked, update smoothed round trip time.
1144	* Since we now have an rtt measurement, cancel the
1145	* timer backoff (cf., Phil Karn's retransmit alg.).
1146	* Recompute the initial retransmit timer.
1147	*/
1148	/* if (ts_present)
1149	* tcp_xmit_timer(tp, tcp_now-ts_ecr+1);
1150	* else
1151	*/
1152	if (tp->t_rtt && SEQ_GT(ti->ti_ack, tp->t_rtseq))
1153	tcp_xmit_timer(pData, tp,tp->t_rtt);
1154
1155	/*
1156	* If all outstanding data is acked, stop retransmit
1157	* timer and remember to restart (more output or persist).
1158	* If there is more data to be acked, restart retransmit
1159	* timer, using current (possibly backed-off) value.
1160	*/
1161	if (ti->ti_ack == tp->snd_max) {
1162	tp->t_timer[TCPT_REXMT] = 0;
1163	needoutput = 1;
1164	} else if (tp->t_timer[TCPT_PERSIST] == 0)
1165	tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
1166	/*
1167	* When new data is acked, open the congestion window.
1168	* If the window gives us less than ssthresh packets
1169	* in flight, open exponentially (maxseg per packet).
1170	* Otherwise open linearly: maxseg per window
1171	* (maxseg^2 / cwnd per packet).
1172	*/
1173	{
1174	register u_int cw = tp->snd_cwnd;
1175	register u_int incr = tp->t_maxseg;
1176
1177	if (cw > tp->snd_ssthresh)
1178	incr = incr * incr / cw;
1179	tp->snd_cwnd = min(cw + incr, TCP_MAXWIN<<tp->snd_scale);
1180	}
1181	if (acked > so->so_snd.sb_cc) {
1182	tp->snd_wnd -= so->so_snd.sb_cc;
1183	sbdrop(&so->so_snd, (int )so->so_snd.sb_cc);
1184	ourfinisacked = 1;
1185	} else {
1186	sbdrop(&so->so_snd, acked);
1187	tp->snd_wnd -= acked;
1188	ourfinisacked = 0;
1189	}
1190	/*
1191	* XXX sowwakup is called when data is acked and there's room for
1192	* for more data... it should read() the socket
1193	*/
1194	/* if (so->so_snd.sb_flags & SB_NOTIFY)
1195	* sowwakeup(so);
1196	*/
1197	tp->snd_una = ti->ti_ack;
1198	if (SEQ_LT(tp->snd_nxt, tp->snd_una))
1199	tp->snd_nxt = tp->snd_una;
1200
1201	switch (tp->t_state) {
1202
1203	/*
1204	* In FIN_WAIT_1 STATE in addition to the processing
1205	* for the ESTABLISHED state if our FIN is now acknowledged
1206	* then enter FIN_WAIT_2.
1207	*/
1208	case TCPS_FIN_WAIT_1:
1209	if (ourfinisacked) {
1210	/*
1211	* If we can't receive any more
1212	* data, then closing user can proceed.
1213	* Starting the timer is contrary to the
1214	* specification, but if we don't get a FIN
1215	* we'll hang forever.
1216	*/
1217	if (so->so_state & SS_FCANTRCVMORE) {
1218	soisfdisconnected(so);
1219	tp->t_timer[TCPT_2MSL] = tcp_maxidle;
1220	}
1221	tp->t_state = TCPS_FIN_WAIT_2;
1222	}
1223	break;
1224
1225	/*
1226	* In CLOSING STATE in addition to the processing for
1227	* the ESTABLISHED state if the ACK acknowledges our FIN
1228	* then enter the TIME-WAIT state, otherwise ignore
1229	* the segment.
1230	*/
1231	case TCPS_CLOSING:
1232	if (ourfinisacked) {
1233	tp->t_state = TCPS_TIME_WAIT;
1234	tcp_canceltimers(tp);
1235	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1236	soisfdisconnected(so);
1237	}
1238	break;
1239
1240	/*
1241	* In LAST_ACK, we may still be waiting for data to drain
1242	* and/or to be acked, as well as for the ack of our FIN.
1243	* If our FIN is now acknowledged, delete the TCB,
1244	* enter the closed state and return.
1245	*/
1246	case TCPS_LAST_ACK:
1247	if (ourfinisacked) {
1248	tp = tcp_close(pData, tp);
1249	goto drop;
1250	}
1251	break;
1252
1253	/*
1254	* In TIME_WAIT state the only thing that should arrive
1255	* is a retransmission of the remote FIN. Acknowledge
1256	* it and restart the finack timer.
1257	*/
1258	case TCPS_TIME_WAIT:
1259	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1260	goto dropafterack;
1261	}
1262	} /* switch(tp->t_state) */
1263
1264	step6:
1265	/*
1266	* Update window information.
1267	* Don't look at window if no ACK: TAC's send garbage on first SYN.
1268	*/
1269	if ((tiflags & TH_ACK) &&
1270	(SEQ_LT(tp->snd_wl1, ti->ti_seq) \|\|
1271	(tp->snd_wl1 == ti->ti_seq && (SEQ_LT(tp->snd_wl2, ti->ti_ack) \|\|
1272	(tp->snd_wl2 == ti->ti_ack && tiwin > tp->snd_wnd))))) {
1273	/* keep track of pure window updates */
1274	if (ti->ti_len == 0 &&
1275	tp->snd_wl2 == ti->ti_ack && tiwin > tp->snd_wnd)
1276	tcpstat.tcps_rcvwinupd++;
1277	tp->snd_wnd = tiwin;
1278	tp->snd_wl1 = ti->ti_seq;
1279	tp->snd_wl2 = ti->ti_ack;
1280	if (tp->snd_wnd > tp->max_sndwnd)
1281	tp->max_sndwnd = tp->snd_wnd;
1282	needoutput = 1;
1283	}
1284
1285	/*
1286	* Process segments with URG.
1287	*/
1288	if ((tiflags & TH_URG) && ti->ti_urp &&
1289	TCPS_HAVERCVDFIN(tp->t_state) == 0) {
1290	/*
1291	* This is a kludge, but if we receive and accept
1292	* random urgent pointers, we'll crash in
1293	* soreceive. It's hard to imagine someone
1294	* actually wanting to send this much urgent data.
1295	*/
1296	if (ti->ti_urp + so->so_rcv.sb_cc > so->so_rcv.sb_datalen) {
1297	ti->ti_urp = 0;
1298	tiflags &= ~TH_URG;
1299	goto dodata;
1300	}
1301	/*
1302	* If this segment advances the known urgent pointer,
1303	* then mark the data stream. This should not happen
1304	* in CLOSE_WAIT, CLOSING, LAST_ACK or TIME_WAIT STATES since
1305	* a FIN has been received from the remote side.
1306	* In these states we ignore the URG.
1307	*
1308	* According to RFC961 (Assigned Protocols),
1309	* the urgent pointer points to the last octet
1310	* of urgent data. We continue, however,
1311	* to consider it to indicate the first octet
1312	* of data past the urgent section as the original
1313	* spec states (in one of two places).
1314	*/
1315	if (SEQ_GT(ti->ti_seq+ti->ti_urp, tp->rcv_up)) {
1316	tp->rcv_up = ti->ti_seq + ti->ti_urp;
1317	so->so_urgc = so->so_rcv.sb_cc +
1318	(tp->rcv_up - tp->rcv_nxt); /* -1; */
1319	tp->rcv_up = ti->ti_seq + ti->ti_urp;
1320
1321	}
1322	} else
1323	/*
1324	* If no out of band data is expected,
1325	* pull receive urgent pointer along
1326	* with the receive window.
1327	*/
1328	if (SEQ_GT(tp->rcv_nxt, tp->rcv_up))
1329	tp->rcv_up = tp->rcv_nxt;
1330	dodata:
1331
1332	/*
1333	* Process the segment text, merging it into the TCP sequencing queue,
1334	* and arranging for acknowledgment of receipt if necessary.
1335	* This process logically involves adjusting tp->rcv_wnd as data
1336	* is presented to the user (this happens in tcp_usrreq.c,
1337	* case PRU_RCVD). If a FIN has already been received on this
1338	* connection then we just ignore the text.
1339	*/
1340	if ((ti->ti_len \|\| (tiflags&TH_FIN)) &&
1341	TCPS_HAVERCVDFIN(tp->t_state) == 0) {
1342	TCP_REASS(pData, tp, ti, m, so, tiflags);
1343	/*
1344	* Note the amount of data that peer has sent into
1345	* our window, in order to estimate the sender's
1346	* buffer size.
1347	*/
1348	len = so->so_rcv.sb_datalen - (tp->rcv_adv - tp->rcv_nxt);
1349	} else {
1350	mbuf_freed = 1; /* The mbuf must be freed, but only when its content is not needed anymore. */
1351	tiflags &= ~TH_FIN;
1352	}
1353
1354	/*
1355	* If FIN is received ACK the FIN and let the user know
1356	* that the connection is closing.
1357	*/
1358	if (tiflags & TH_FIN) {
1359	if (TCPS_HAVERCVDFIN(tp->t_state) == 0) {
1360	/*
1361	* If we receive a FIN we can't send more data,
1362	* set it SS_FDRAIN
1363	* Shutdown the socket if there is no rx data in the
1364	* buffer.
1365	* soread() is called on completion of shutdown() and
1366	* will got to TCPS_LAST_ACK, and use tcp_output()
1367	* to send the FIN.
1368	*/
1369	/* sofcantrcvmore(so); */
1370	sofwdrain(so);
1371
1372	tp->t_flags \|= TF_ACKNOW;
1373	tp->rcv_nxt++;
1374	}
1375	switch (tp->t_state) {
1376
1377	/*
1378	* In SYN_RECEIVED and ESTABLISHED STATES
1379	* enter the CLOSE_WAIT state.
1380	*/
1381	case TCPS_SYN_RECEIVED:
1382	case TCPS_ESTABLISHED:
1383	if(so->so_emu == EMU_CTL) /* no shutdown on socket */
1384	tp->t_state = TCPS_LAST_ACK;
1385	else
1386	tp->t_state = TCPS_CLOSE_WAIT;
1387	break;
1388
1389	/*
1390	* If still in FIN_WAIT_1 STATE FIN has not been acked so
1391	* enter the CLOSING state.
1392	*/
1393	case TCPS_FIN_WAIT_1:
1394	tp->t_state = TCPS_CLOSING;
1395	break;
1396
1397	/*
1398	* In FIN_WAIT_2 state enter the TIME_WAIT state,
1399	* starting the time-wait timer, turning off the other
1400	* standard timers.
1401	*/
1402	case TCPS_FIN_WAIT_2:
1403	tp->t_state = TCPS_TIME_WAIT;
1404	tcp_canceltimers(tp);
1405	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1406	soisfdisconnected(so);
1407	break;
1408
1409	/*
1410	* In TIME_WAIT state restart the 2 MSL time_wait timer.
1411	*/
1412	case TCPS_TIME_WAIT:
1413	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1414	break;
1415	}
1416	}
1417
1418	/*
1419	* If this is a small packet, then ACK now - with Nagel
1420	* congestion avoidance sender won't send more until
1421	* he gets an ACK.
1422	*
1423	* See above.
1424	*/
1425	/* if (ti->ti_len && (unsigned)ti->ti_len < tp->t_maxseg) {
1426	*/
1427	/* if ((ti->ti_len && (unsigned)ti->ti_len < tp->t_maxseg &&
1428	* (so->so_iptos & IPTOS_LOWDELAY) == 0) \|\|
1429	* ((so->so_iptos & IPTOS_LOWDELAY) &&
1430	* ((struct tcpiphdr_2 *)ti)->first_char == (char)27)) {
1431	*/
1432	if (ti->ti_len && (unsigned)ti->ti_len <= 5 &&
1433	((struct tcpiphdr_2 *)ti)->first_char == (char)27) {
1434	tp->t_flags \|= TF_ACKNOW;
1435	}
1436
1437	if (mbuf_freed) {
1438	m_free(pData, m);
1439	}
1440	/*
1441	* Return any desired output.
1442	*/
1443	if (needoutput \|\| (tp->t_flags & TF_ACKNOW)) {
1444	(void) tcp_output(pData, tp);
1445	}
1446	return;
1447
1448	dropafterack:
1449	/*
1450	* Generate an ACK dropping incoming segment if it occupies
1451	* sequence space, where the ACK reflects our state.
1452	*/
1453	if (tiflags & TH_RST)
1454	goto drop;
1455	m_freem(pData, m);
1456	tp->t_flags \|= TF_ACKNOW;
1457	(void) tcp_output(pData, tp);
1458	return;
1459
1460	dropwithreset:
1461	/* reuses m if m!=NULL, m_free() unnecessary */
1462	if (tiflags & TH_ACK)
1463	tcp_respond(pData, tp, ti, m, (tcp_seq)0, ti->ti_ack, TH_RST);
1464	else {
1465	if (tiflags & TH_SYN) ti->ti_len++;
1466	tcp_respond(pData, tp, ti, m, ti->ti_seq+ti->ti_len, (tcp_seq)0,
1467	TH_RST\|TH_ACK);
1468	}
1469
1470	return;
1471
1472	drop:
1473	/*
1474	* Drop space held by incoming segment and return.
1475	*/
1476	m_free(pData, m);
1477
1478	return;
1479	}
1480
1481	/* , ts_present, ts_val, ts_ecr) */
1482	/* int *ts_present;
1483	* u_int32_t ts_val, ts_ecr;
1484	*/
1485	void
1486	tcp_dooptions(PNATState pData, struct tcpcb tp, u_char cp, int cnt, struct tcpiphdr *ti)
1487	{
1488	u_int16_t mss;
1489	int opt, optlen;
1490
1491	DEBUG_CALL("tcp_dooptions");
1492	DEBUG_ARGS((dfd," tp = %lx cnt=%i \n", (long )tp, cnt));
1493
1494	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1495	opt = cp[0];
1496	if (opt == TCPOPT_EOL)
1497	break;
1498	if (opt == TCPOPT_NOP)
1499	optlen = 1;
1500	else {
1501	optlen = cp[1];
1502	if (optlen <= 0)
1503	break;
1504	}
1505	switch (opt) {
1506
1507	default:
1508	continue;
1509
1510	case TCPOPT_MAXSEG:
1511	if (optlen != TCPOLEN_MAXSEG)
1512	continue;
1513	if (!(ti->ti_flags & TH_SYN))
1514	continue;
1515	memcpy((char ) &mss, (char ) cp + 2, sizeof(mss));
1516	NTOHS(mss);
1517	(void) tcp_mss(pData, tp, mss); /* sets t_maxseg */
1518	break;
1519
1520	/* case TCPOPT_WINDOW:
1521	* if (optlen != TCPOLEN_WINDOW)
1522	* continue;
1523	* if (!(ti->ti_flags & TH_SYN))
1524	* continue;
1525	* tp->t_flags \|= TF_RCVD_SCALE;
1526	* tp->requested_s_scale = min(cp[2], TCP_MAX_WINSHIFT);
1527	* break;
1528	*/
1529	/* case TCPOPT_TIMESTAMP:
1530	* if (optlen != TCPOLEN_TIMESTAMP)
1531	* continue;
1532	* *ts_present = 1;
1533	* memcpy((char ) ts_val, (char )cp + 2, sizeof(*ts_val));
1534	* NTOHL(*ts_val);
1535	* memcpy((char ) ts_ecr, (char )cp + 6, sizeof(*ts_ecr));
1536	* NTOHL(*ts_ecr);
1537	*
1538	/ /
1539	* * A timestamp received in a SYN makes
1540	* * it ok to send timestamp requests and replies.
1541	* */
1542	/* if (ti->ti_flags & TH_SYN) {
1543	* tp->t_flags \|= TF_RCVD_TSTMP;
1544	* tp->ts_recent = *ts_val;
1545	* tp->ts_recent_age = tcp_now;
1546	* }
1547	*/ break;
1548	}
1549	}
1550	}
1551
1552
1553	/*
1554	* Pull out of band byte out of a segment so
1555	* it doesn't appear in the user's data queue.
1556	* It is still reflected in the segment length for
1557	* sequencing purposes.
1558	*/
1559
1560	#ifdef notdef
1561
1562	void
1563	tcp_pulloutofband(so, ti, m)
1564	struct socket *so;
1565	struct tcpiphdr *ti;
1566	register struct mbuf *m;
1567	{
1568	int cnt = ti->ti_urp - 1;
1569
1570	while (cnt >= 0) {
1571	if (m->m_len > cnt) {
1572	char *cp = mtod(m, caddr_t) + cnt;
1573	struct tcpcb *tp = sototcpcb(so);
1574
1575	tp->t_iobc = *cp;
1576	tp->t_oobflags \|= TCPOOB_HAVEDATA;
1577	memcpy(sp, cp+1, (unsigned)(m->m_len - cnt - 1));
1578	m->m_len--;
1579	return;
1580	}
1581	cnt -= m->m_len;
1582	m = m->m_next; /* XXX WRONG! Fix it! */
1583	if (m == 0)
1584	break;
1585	}
1586	panic("tcp_pulloutofband");
1587	}
1588
1589	#endif /* notdef */
1590
1591	/*
1592	* Collect new round-trip time estimate
1593	* and update averages and current timeout.
1594	*/
1595
1596	void
1597	tcp_xmit_timer(PNATState pData, register struct tcpcb *tp, int rtt)
1598	{
1599	register short delta;
1600
1601	DEBUG_CALL("tcp_xmit_timer");
1602	DEBUG_ARG("tp = %lx", (long)tp);
1603	DEBUG_ARG("rtt = %d", rtt);
1604
1605	tcpstat.tcps_rttupdated++;
1606	if (tp->t_srtt != 0) {
1607	/*
1608	* srtt is stored as fixed point with 3 bits after the
1609	* binary point (i.e., scaled by 8). The following magic
1610	* is equivalent to the smoothing algorithm in rfc793 with
1611	* an alpha of .875 (srtt = rtt/8 + srtt*7/8 in fixed
1612	* point). Adjust rtt to origin 0.
1613	*/
1614	delta = rtt - 1 - (tp->t_srtt >> TCP_RTT_SHIFT);
1615	if ((tp->t_srtt += delta) <= 0)
1616	tp->t_srtt = 1;
1617	/*
1618	* We accumulate a smoothed rtt variance (actually, a
1619	* smoothed mean difference), then set the retransmit
1620	* timer to smoothed rtt + 4 times the smoothed variance.
1621	* rttvar is stored as fixed point with 2 bits after the
1622	* binary point (scaled by 4). The following is
1623	* equivalent to rfc793 smoothing with an alpha of .75
1624	* (rttvar = rttvar*3/4 + \|delta\| / 4). This replaces
1625	* rfc793's wired-in beta.
1626	*/
1627	if (delta < 0)
1628	delta = -delta;
1629	delta -= (tp->t_rttvar >> TCP_RTTVAR_SHIFT);
1630	if ((tp->t_rttvar += delta) <= 0)
1631	tp->t_rttvar = 1;
1632	} else {
1633	/*
1634	* No rtt measurement yet - use the unsmoothed rtt.
1635	* Set the variance to half the rtt (so our first
1636	* retransmit happens at 3*rtt).
1637	*/
1638	tp->t_srtt = rtt << TCP_RTT_SHIFT;
1639	tp->t_rttvar = rtt << (TCP_RTTVAR_SHIFT - 1);
1640	}
1641	tp->t_rtt = 0;
1642	tp->t_rxtshift = 0;
1643
1644	/*
1645	* the retransmit should happen at rtt + 4 * rttvar.
1646	* Because of the way we do the smoothing, srtt and rttvar
1647	* will each average +1/2 tick of bias. When we compute
1648	* the retransmit timer, we want 1/2 tick of rounding and
1649	* 1 extra tick because of +-1/2 tick uncertainty in the
1650	* firing of the timer. The bias will give us exactly the
1651	* 1.5 tick we need. But, because the bias is
1652	* statistical, we have to test that we don't drop below
1653	* the minimum feasible timer (which is 2 ticks).
1654	*/
1655	TCPT_RANGESET(tp->t_rxtcur, TCP_REXMTVAL(tp),
1656	(short)tp->t_rttmin, TCPTV_REXMTMAX); /* XXX */
1657
1658	/*
1659	* We received an ack for a packet that wasn't retransmitted;
1660	* it is probably safe to discard any error indications we've
1661	* received recently. This isn't quite right, but close enough
1662	* for now (a route might have failed after we sent a segment,
1663	* and the return path might not be symmetrical).
1664	*/
1665	tp->t_softerror = 0;
1666	}
1667
1668	/*
1669	* Determine a reasonable value for maxseg size.
1670	* If the route is known, check route for mtu.
1671	* If none, use an mss that can be handled on the outgoing
1672	* interface without forcing IP to fragment; if bigger than
1673	* an mbuf cluster (MCLBYTES), round down to nearest multiple of MCLBYTES
1674	* to utilize large mbufs. If no route is found, route has no mtu,
1675	* or the destination isn't local, use a default, hopefully conservative
1676	* size (usually 512 or the default IP max size, but no more than the mtu
1677	* of the interface), as we can't discover anything about intervening
1678	* gateways or networks. We also initialize the congestion/slow start
1679	* window to be a single segment if the destination isn't local.
1680	* While looking at the routing entry, we also initialize other path-dependent
1681	* parameters from pre-set or cached values in the routing entry.
1682	*/
1683
1684	int
1685	tcp_mss(PNATState pData, register struct tcpcb *tp, u_int offer)
1686	{
1687	struct socket *so = tp->t_socket;
1688	int mss;
1689
1690	DEBUG_CALL("tcp_mss");
1691	DEBUG_ARG("tp = %lx", (long)tp);
1692	DEBUG_ARG("offer = %d", offer);
1693
1694	mss = min(if_mtu, if_mru) - sizeof(struct tcpiphdr);
1695	if (offer)
1696	mss = min(mss, offer);
1697	mss = max(mss, 32);
1698	if (mss < tp->t_maxseg \|\| offer != 0)
1699	tp->t_maxseg = mss;
1700
1701	tp->snd_cwnd = mss;
1702
1703	sbreserve(&so->so_snd, tcp_sndspace+((tcp_sndspace%mss)?(mss-(tcp_sndspace%mss)):0));
1704	sbreserve(&so->so_rcv, tcp_rcvspace+((tcp_rcvspace%mss)?(mss-(tcp_rcvspace%mss)):0));
1705
1706	DEBUG_MISC((dfd, " returning mss = %d\n", mss));
1707
1708	return mss;
1709	}

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

source: vbox/trunk/src/VBox/Devices/Network/slirp/tcp_input.c@ 13984

Download in other formats: