tcp_input.c@ 3021

Last change on this file since 3021 was 1076, checked in by vboxsync, 18 years ago
Removed tons of ifdef VBOX conditionals to make slirp readable again
Property svn:eol-style set to `native`
File size: 48.2 KB

Line
1	/*
2	* Copyright (c) 1982, 1986, 1988, 1990, 1993, 1994
3	* The Regents of the University of California. All rights reserved.
4	*
5	* Redistribution and use in source and binary forms, with or without
6	* modification, are permitted provided that the following conditions
7	* are met:
8	* 1. Redistributions of source code must retain the above copyright
9	* notice, this list of conditions and the following disclaimer.
10	* 2. Redistributions in binary form must reproduce the above copyright
11	* notice, this list of conditions and the following disclaimer in the
12	* documentation and/or other materials provided with the distribution.
13	* 3. All advertising materials mentioning features or use of this software
14	* must display the following acknowledgement:
15	* This product includes software developed by the University of
16	* California, Berkeley and its contributors.
17	* 4. Neither the name of the University nor the names of its contributors
18	* may be used to endorse or promote products derived from this software
19	* without specific prior written permission.
20	*
21	* THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22	* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23	* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24	* ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25	* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26	* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27	* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28	* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29	* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30	* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31	* SUCH DAMAGE.
32	*
33	* @(#)tcp_input.c 8.5 (Berkeley) 4/10/94
34	* tcp_input.c,v 1.10 1994/10/13 18:36:32 wollman Exp
35	*/
36
37	/*
38	* Changes and additions relating to SLiRP
39	* Copyright (c) 1995 Danny Gasparovski.
40	*
41	* Please read the file COPYRIGHT for the
42	* terms and conditions of the copyright.
43	*/
44
45	#include <slirp.h>
46	#include "ip_icmp.h"
47
48
49	#define TCP_PAWS_IDLE (24 * 24 * 60 * 60 * PR_SLOWHZ)
50
51	/* for modulo comparisons of timestamps */
52	#define TSTMP_LT(a,b) ((int)((a)-(b)) < 0)
53	#define TSTMP_GEQ(a,b) ((int)((a)-(b)) >= 0)
54
55	/*
56	* Insert segment ti into reassembly queue of tcp with
57	* control block tp. Return TH_FIN if reassembly now includes
58	* a segment with FIN. The macro form does the common case inline
59	* (segment is the next to be received on an established connection,
60	* and the queue is empty), avoiding linkage into and removal
61	* from the queue and repetition of various conversions.
62	* Set DELACK for segments received in order, but ack immediately
63	* when segments are out of order (so fast retransmit can work).
64	*/
65	#ifdef TCP_ACK_HACK
66	#define TCP_REASS(pData, tp, ti, m, so, flags) {\
67	if ((ti)->ti_seq == (tp)->rcv_nxt && \
68	u32_to_ptr((pData), (tp)->seg_next, struct tcpcb *) == (tp) && \
69	(tp)->t_state == TCPS_ESTABLISHED) {\
70	if (ti->ti_flags & TH_PUSH) \
71	tp->t_flags \|= TF_ACKNOW; \
72	else \
73	tp->t_flags \|= TF_DELACK; \
74	(tp)->rcv_nxt += (ti)->ti_len; \
75	flags = (ti)->ti_flags & TH_FIN; \
76	tcpstat.tcps_rcvpack++;\
77	tcpstat.tcps_rcvbyte += (ti)->ti_len;\
78	if (so->so_emu) { \
79	if (tcp_emu((pData), (so),(m))) sbappend((pData), (so), (m)); \
80	} else \
81	sbappend((pData), (so), (m)); \
82	/* sorwakeup(so); */ \
83	} else {\
84	(flags) = tcp_reass((pData), (tp), (ti), (m)); \
85	tp->t_flags \|= TF_ACKNOW; \
86	} \
87	}
88	#else
89	#define TCP_REASS(pData, tp, ti, m, so, flags) { \
90	if ((ti)->ti_seq == (tp)->rcv_nxt && \
91	u32_to_ptr((pData), (tp)->seg_next, struct tcpcb *) == (tp) && \
92	(tp)->t_state == TCPS_ESTABLISHED) { \
93	tp->t_flags \|= TF_DELACK; \
94	(tp)->rcv_nxt += (ti)->ti_len; \
95	flags = (ti)->ti_flags & TH_FIN; \
96	tcpstat.tcps_rcvpack++;\
97	tcpstat.tcps_rcvbyte += (ti)->ti_len;\
98	if (so->so_emu) { \
99	if (tcp_emu((pData), (so),(m))) sbappend((pData), (so), (m)); \
100	} else \
101	sbappend((pData), (so), (m)); \
102	/* sorwakeup(so); */ \
103	} else { \
104	(flags) = tcp_reass((pData), (tp), (ti), (m)); \
105	tp->t_flags \|= TF_ACKNOW; \
106	} \
107	}
108	#endif
109
110	int
111	tcp_reass(PNATState pData, register struct tcpcb tp, register struct tcpiphdr ti, struct mbuf *m)
112	{
113	register struct tcpiphdr *q;
114	struct socket *so = tp->t_socket;
115	int flags;
116
117	/*
118	* Call with ti==0 after become established to
119	* force pre-ESTABLISHED data up to user socket.
120	*/
121	if (ti == 0)
122	goto present;
123
124	/*
125	* Find a segment which begins after this one does.
126	*/
127	for (q = u32_to_ptr(pData, tp->seg_next, struct tcpiphdr ); q != (struct tcpiphdr )tp;
128	q = u32_to_ptr(pData, q->ti_next, struct tcpiphdr *))
129	if (SEQ_GT(q->ti_seq, ti->ti_seq))
130	break;
131
132	/*
133	* If there is a preceding segment, it may provide some of
134	* our data already. If so, drop the data from the incoming
135	* segment. If it provides all of our data, drop us.
136	*/
137	if (u32_to_ptr(pData, q->ti_prev, struct tcpiphdr ) != (struct tcpiphdr )tp) {
138	register int i;
139	q = u32_to_ptr(pData, q->ti_prev, struct tcpiphdr *);
140	/* conversion to int (in i) handles seq wraparound */
141	i = q->ti_seq + q->ti_len - ti->ti_seq;
142	if (i > 0) {
143	if (i >= ti->ti_len) {
144	tcpstat.tcps_rcvduppack++;
145	tcpstat.tcps_rcvdupbyte += ti->ti_len;
146	m_freem(pData, m);
147	/*
148	* Try to present any queued data
149	* at the left window edge to the user.
150	* This is needed after the 3-WHS
151	* completes.
152	*/
153	goto present; /* ??? */
154	}
155	m_adj(m, i);
156	ti->ti_len -= i;
157	ti->ti_seq += i;
158	}
159	q = u32_to_ptr(pData, q->ti_next, struct tcpiphdr *);
160	}
161	tcpstat.tcps_rcvoopack++;
162	tcpstat.tcps_rcvoobyte += ti->ti_len;
163	REASS_MBUF_SET(ti, m); /* XXX */
164
165	/*
166	* While we overlap succeeding segments trim them or,
167	* if they are completely covered, dequeue them.
168	*/
169	while (q != (struct tcpiphdr *)tp) {
170	register int i = (ti->ti_seq + ti->ti_len) - q->ti_seq;
171	if (i <= 0)
172	break;
173	if (i < q->ti_len) {
174	q->ti_seq += i;
175	q->ti_len -= i;
176	m_adj(REASS_MBUF_GET(q), i);
177	break;
178	}
179	q = u32_to_ptr(pData, q->ti_next, struct tcpiphdr *);
180	m = REASS_MBUF_GET(u32_to_ptr(pData, q->ti_prev, struct tcpiphdr *));
181	remque_32(pData, u32_to_ptr(pData, q->ti_prev, struct tcpiphdr *));
182	m_freem(pData, m);
183	}
184
185	/*
186	* Stick new segment in its place.
187	*/
188	insque_32(pData, ti, u32_to_ptr(pData, q->ti_prev, struct tcpiphdr *));
189
190	present:
191	/*
192	* Present data to user, advancing rcv_nxt through
193	* completed sequence space.
194	*/
195	if (!TCPS_HAVEESTABLISHED(tp->t_state))
196	return (0);
197	ti = u32_to_ptr(pData, tp->seg_next, struct tcpiphdr *);
198	if (ti == (struct tcpiphdr *)tp \|\| ti->ti_seq != tp->rcv_nxt)
199	return (0);
200	if (tp->t_state == TCPS_SYN_RECEIVED && ti->ti_len)
201	return (0);
202	do {
203	tp->rcv_nxt += ti->ti_len;
204	flags = ti->ti_flags & TH_FIN;
205	remque_32(pData, ti);
206	m = REASS_MBUF_GET(ti); /* XXX */
207	ti = u32_to_ptr(pData, ti->ti_next, struct tcpiphdr *);
208	/* if (so->so_state & SS_FCANTRCVMORE) */
209	if (so->so_state & SS_FCANTSENDMORE)
210	m_freem(pData, m);
211	else {
212	if (so->so_emu) {
213	if (tcp_emu(pData, so,m)) sbappend(pData, so, m);
214	} else
215	sbappend(pData, so, m);
216	}
217	} while (ti != (struct tcpiphdr *)tp && ti->ti_seq == tp->rcv_nxt);
218	/* sorwakeup(so); */
219	return (flags);
220	}
221
222	/*
223	* TCP input routine, follows pages 65-76 of the
224	* protocol specification dated September, 1981 very closely.
225	*/
226	void
227	tcp_input(PNATState pData, register struct mbuf m, int iphlen, struct socket inso)
228	{
229	struct ip save_ip, *ip;
230	register struct tcpiphdr *ti;
231	caddr_t optp = NULL;
232	int optlen = 0;
233	int len, tlen, off;
234	register struct tcpcb *tp = 0;
235	register int tiflags;
236	struct socket *so = 0;
237	int todrop, acked, ourfinisacked, needoutput = 0;
238	/* int dropsocket = 0; */
239	int iss = 0;
240	u_long tiwin;
241	int ret;
242	/* int ts_present = 0; */
243
244	DEBUG_CALL("tcp_input");
245	DEBUG_ARGS((dfd," m = %8lx iphlen = %2d inso = %lx\n",
246	(long )m, iphlen, (long )inso ));
247
248	/*
249	* If called with m == 0, then we're continuing the connect
250	*/
251	if (m == NULL) {
252	so = inso;
253
254	/* Re-set a few variables */
255	tp = sototcpcb(so);
256	m = so->so_m;
257	so->so_m = 0;
258	ti = so->so_ti;
259	tiwin = ti->ti_win;
260	tiflags = ti->ti_flags;
261
262	goto cont_conn;
263	}
264
265
266	tcpstat.tcps_rcvtotal++;
267	/*
268	* Get IP and TCP header together in first mbuf.
269	* Note: IP leaves IP header in first mbuf.
270	*/
271	ti = mtod(m, struct tcpiphdr *);
272	if (iphlen > sizeof(struct ip )) {
273	ip_stripoptions(m, (struct mbuf *)0);
274	iphlen=sizeof(struct ip );
275	}
276	/* XXX Check if too short */
277
278
279	/*
280	* Save a copy of the IP header in case we want restore it
281	* for sending an ICMP error message in response.
282	*/
283	ip=mtod(m, struct ip *);
284	save_ip = *ip;
285	save_ip.ip_len+= iphlen;
286
287	/*
288	* Checksum extended TCP header and data.
289	*/
290	tlen = ((struct ip *)ti)->ip_len;
291	ti->ti_next = ti->ti_prev = 0;
292	ti->ti_x1 = 0;
293	ti->ti_len = htons((u_int16_t)tlen);
294	len = sizeof(struct ip ) + tlen;
295	/* keep checksum for ICMP reply
296	* ti->ti_sum = cksum(m, len);
297	* if (ti->ti_sum) { */
298	if(cksum(m, len)) {
299	tcpstat.tcps_rcvbadsum++;
300	goto drop;
301	}
302
303	/*
304	* Check that TCP offset makes sense,
305	* pull out TCP options and adjust length. XXX
306	*/
307	off = ti->ti_off << 2;
308	if (off < sizeof (struct tcphdr) \|\| off > tlen) {
309	tcpstat.tcps_rcvbadoff++;
310	goto drop;
311	}
312	tlen -= off;
313	ti->ti_len = tlen;
314	if (off > sizeof (struct tcphdr)) {
315	optlen = off - sizeof (struct tcphdr);
316	optp = mtod(m, caddr_t) + sizeof (struct tcpiphdr);
317
318	/*
319	* Do quick retrieval of timestamp options ("options
320	* prediction?"). If timestamp is the only option and it's
321	* formatted as recommended in RFC 1323 appendix A, we
322	* quickly get the values now and not bother calling
323	* tcp_dooptions(), etc.
324	*/
325	/* if ((optlen == TCPOLEN_TSTAMP_APPA \|\|
326	* (optlen > TCPOLEN_TSTAMP_APPA &&
327	* optp[TCPOLEN_TSTAMP_APPA] == TCPOPT_EOL)) &&
328	* (u_int32_t )optp == htonl(TCPOPT_TSTAMP_HDR) &&
329	* (ti->ti_flags & TH_SYN) == 0) {
330	* ts_present = 1;
331	* ts_val = ntohl((u_int32_t )(optp + 4));
332	* ts_ecr = ntohl((u_int32_t )(optp + 8));
333	* optp = NULL; / * we've parsed the options * /
334	* }
335	*/
336	}
337	tiflags = ti->ti_flags;
338
339	/*
340	* Convert TCP protocol specific fields to host format.
341	*/
342	NTOHL(ti->ti_seq);
343	NTOHL(ti->ti_ack);
344	NTOHS(ti->ti_win);
345	NTOHS(ti->ti_urp);
346
347	/*
348	* Drop TCP, IP headers and TCP options.
349	*/
350	m->m_data += sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
351	m->m_len -= sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
352
353	/*
354	* Locate pcb for segment.
355	*/
356	findso:
357	so = tcp_last_so;
358	if (so->so_fport != ti->ti_dport \|\|
359	so->so_lport != ti->ti_sport \|\|
360	so->so_laddr.s_addr != ti->ti_src.s_addr \|\|
361	so->so_faddr.s_addr != ti->ti_dst.s_addr) {
362	so = solookup(&tcb, ti->ti_src, ti->ti_sport,
363	ti->ti_dst, ti->ti_dport);
364	if (so)
365	tcp_last_so = so;
366	++tcpstat.tcps_socachemiss;
367	}
368
369	/*
370	* If the state is CLOSED (i.e., TCB does not exist) then
371	* all data in the incoming segment is discarded.
372	* If the TCB exists but is in CLOSED state, it is embryonic,
373	* but should either do a listen or a connect soon.
374	*
375	* state == CLOSED means we've done socreate() but haven't
376	* attached it to a protocol yet...
377	*
378	* XXX If a TCB does not exist, and the TH_SYN flag is
379	* the only flag set, then create a session, mark it
380	* as if it was LISTENING, and continue...
381	*/
382	if (so == 0) {
383	if ((tiflags & (TH_SYN\|TH_FIN\|TH_RST\|TH_URG\|TH_ACK)) != TH_SYN)
384	goto dropwithreset;
385
386	if ((so = socreate()) == NULL)
387	goto dropwithreset;
388	if (tcp_attach(pData, so) < 0) {
389	free(so); /* Not sofree (if it failed, it's not insqued) */
390	goto dropwithreset;
391	}
392
393	sbreserve(&so->so_snd, tcp_sndspace);
394	sbreserve(&so->so_rcv, tcp_rcvspace);
395
396	/* tcp_last_so = so; / / XXX ? */
397	/* tp = sototcpcb(so); */
398
399	so->so_laddr = ti->ti_src;
400	so->so_lport = ti->ti_sport;
401	so->so_faddr = ti->ti_dst;
402	so->so_fport = ti->ti_dport;
403
404	if ((so->so_iptos = tcp_tos(so)) == 0)
405	so->so_iptos = ((struct ip *)ti)->ip_tos;
406
407	tp = sototcpcb(so);
408	tp->t_state = TCPS_LISTEN;
409	}
410
411	/*
412	* If this is a still-connecting socket, this probably
413	* a retransmit of the SYN. Whether it's a retransmit SYN
414	* or something else, we nuke it.
415	*/
416	if (so->so_state & SS_ISFCONNECTING)
417	goto drop;
418
419	tp = sototcpcb(so);
420
421	/* XXX Should never fail */
422	if (tp == 0)
423	goto dropwithreset;
424	if (tp->t_state == TCPS_CLOSED)
425	goto drop;
426
427	/* Unscale the window into a 32-bit value. */
428	/* if ((tiflags & TH_SYN) == 0)
429	* tiwin = ti->ti_win << tp->snd_scale;
430	* else
431	*/
432	tiwin = ti->ti_win;
433
434	/*
435	* Segment received on connection.
436	* Reset idle time and keep-alive timer.
437	*/
438	tp->t_idle = 0;
439	if (so_options)
440	tp->t_timer[TCPT_KEEP] = tcp_keepintvl;
441	else
442	tp->t_timer[TCPT_KEEP] = tcp_keepidle;
443
444	/*
445	* Process options if not in LISTEN state,
446	* else do it below (after getting remote address).
447	*/
448	if (optp && tp->t_state != TCPS_LISTEN)
449	tcp_dooptions(pData, tp, (u_char *)optp, optlen, ti);
450	/* , */
451	/* &ts_present, &ts_val, &ts_ecr); */
452
453	/*
454	* Header prediction: check for the two common cases
455	* of a uni-directional data xfer. If the packet has
456	* no control flags, is in-sequence, the window didn't
457	* change and we're not retransmitting, it's a
458	* candidate. If the length is zero and the ack moved
459	* forward, we're the sender side of the xfer. Just
460	* free the data acked & wake any higher level process
461	* that was blocked waiting for space. If the length
462	* is non-zero and the ack didn't move, we're the
463	* receiver side. If we're getting packets in-order
464	* (the reassembly queue is empty), add the data to
465	* the socket buffer and note that we need a delayed ack.
466	*
467	* XXX Some of these tests are not needed
468	* eg: the tiwin == tp->snd_wnd prevents many more
469	* predictions.. with no real advantage..
470	*/
471	if (tp->t_state == TCPS_ESTABLISHED &&
472	(tiflags & (TH_SYN\|TH_FIN\|TH_RST\|TH_URG\|TH_ACK)) == TH_ACK &&
473	/* (!ts_present \|\| TSTMP_GEQ(ts_val, tp->ts_recent)) && */
474	ti->ti_seq == tp->rcv_nxt &&
475	tiwin && tiwin == tp->snd_wnd &&
476	tp->snd_nxt == tp->snd_max) {
477	/*
478	* If last ACK falls within this segment's sequence numbers,
479	* record the timestamp.
480	*/
481	/* if (ts_present && SEQ_LEQ(ti->ti_seq, tp->last_ack_sent) &&
482	* SEQ_LT(tp->last_ack_sent, ti->ti_seq + ti->ti_len)) {
483	* tp->ts_recent_age = tcp_now;
484	* tp->ts_recent = ts_val;
485	* }
486	*/
487	if (ti->ti_len == 0) {
488	if (SEQ_GT(ti->ti_ack, tp->snd_una) &&
489	SEQ_LEQ(ti->ti_ack, tp->snd_max) &&
490	tp->snd_cwnd >= tp->snd_wnd) {
491	/*
492	* this is a pure ack for outstanding data.
493	*/
494	++tcpstat.tcps_predack;
495	/* if (ts_present)
496	* tcp_xmit_timer(tp, tcp_now-ts_ecr+1);
497	* else
498	*/ if (tp->t_rtt &&
499	SEQ_GT(ti->ti_ack, tp->t_rtseq))
500	tcp_xmit_timer(pData, tp, tp->t_rtt);
501	acked = ti->ti_ack - tp->snd_una;
502	tcpstat.tcps_rcvackpack++;
503	tcpstat.tcps_rcvackbyte += acked;
504	sbdrop(&so->so_snd, acked);
505	tp->snd_una = ti->ti_ack;
506	m_freem(pData, m);
507
508	/*
509	* If all outstanding data are acked, stop
510	* retransmit timer, otherwise restart timer
511	* using current (possibly backed-off) value.
512	* If process is waiting for space,
513	* wakeup/selwakeup/signal. If data
514	* are ready to send, let tcp_output
515	* decide between more output or persist.
516	*/
517	if (tp->snd_una == tp->snd_max)
518	tp->t_timer[TCPT_REXMT] = 0;
519	else if (tp->t_timer[TCPT_PERSIST] == 0)
520	tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
521
522	/*
523	* There's room in so_snd, sowwakup will read()
524	* from the socket if we can
525	*/
526	/* if (so->so_snd.sb_flags & SB_NOTIFY)
527	* sowwakeup(so);
528	*/
529	/*
530	* This is called because sowwakeup might have
531	* put data into so_snd. Since we don't so sowwakeup,
532	* we don't need this.. XXX???
533	*/
534	if (so->so_snd.sb_cc)
535	(void) tcp_output(pData, tp);
536
537	return;
538	}
539	} else if (ti->ti_ack == tp->snd_una &&
540	u32_to_ptr(pData, tp->seg_next, struct tcpcb *) == tp &&
541	ti->ti_len <= sbspace(&so->so_rcv)) {
542	/*
543	* this is a pure, in-sequence data packet
544	* with nothing on the reassembly queue and
545	* we have enough buffer space to take it.
546	*/
547	++tcpstat.tcps_preddat;
548	tp->rcv_nxt += ti->ti_len;
549	tcpstat.tcps_rcvpack++;
550	tcpstat.tcps_rcvbyte += ti->ti_len;
551	/*
552	* Add data to socket buffer.
553	*/
554	if (so->so_emu) {
555	if (tcp_emu(pData, so,m)) sbappend(pData, so, m);
556	} else
557	sbappend(pData, so, m);
558
559	/*
560	* XXX This is called when data arrives. Later, check
561	* if we can actually write() to the socket
562	* XXX Need to check? It's be NON_BLOCKING
563	*/
564	/* sorwakeup(so); */
565
566	/*
567	* If this is a short packet, then ACK now - with Nagel
568	* congestion avoidance sender won't send more until
569	* he gets an ACK.
570	*
571	* It is better to not delay acks at all to maximize
572	* TCP throughput. See RFC 2581.
573	*/
574	tp->t_flags \|= TF_ACKNOW;
575	tcp_output(pData, tp);
576	return;
577	}
578	} /* header prediction */
579	/*
580	* Calculate amount of space in receive window,
581	* and then do TCP input processing.
582	* Receive window is amount of space in rcv queue,
583	* but not less than advertised window.
584	*/
585	{ int win;
586	win = sbspace(&so->so_rcv);
587	if (win < 0)
588	win = 0;
589	tp->rcv_wnd = max(win, (int)(tp->rcv_adv - tp->rcv_nxt));
590	}
591
592	switch (tp->t_state) {
593
594	/*
595	* If the state is LISTEN then ignore segment if it contains an RST.
596	* If the segment contains an ACK then it is bad and send a RST.
597	* If it does not contain a SYN then it is not interesting; drop it.
598	* Don't bother responding if the destination was a broadcast.
599	* Otherwise initialize tp->rcv_nxt, and tp->irs, select an initial
600	* tp->iss, and send a segment:
601	* <SEQ=ISS><ACK=RCV_NXT><CTL=SYN,ACK>
602	* Also initialize tp->snd_nxt to tp->iss+1 and tp->snd_una to tp->iss.
603	* Fill in remote peer address fields if not previously specified.
604	* Enter SYN_RECEIVED state, and process any other fields of this
605	* segment in this state.
606	*/
607	case TCPS_LISTEN: {
608
609	if (tiflags & TH_RST)
610	goto drop;
611	if (tiflags & TH_ACK)
612	goto dropwithreset;
613	if ((tiflags & TH_SYN) == 0)
614	goto drop;
615
616	/*
617	* This has way too many gotos...
618	* But a bit of spaghetti code never hurt anybody :)
619	*/
620
621	/*
622	* If this is destined for the control address, then flag to
623	* tcp_ctl once connected, otherwise connect
624	*/
625	if ((so->so_faddr.s_addr&htonl(0xffffff00)) == special_addr.s_addr) {
626	int lastbyte=ntohl(so->so_faddr.s_addr) & 0xff;
627	if (lastbyte!=CTL_ALIAS && lastbyte!=CTL_DNS) {
628	#if 0
629	if(lastbyte==CTL_CMD \|\| lastbyte==CTL_EXEC) {
630	/* Command or exec adress */
631	so->so_state \|= SS_CTL;
632	} else
633	#endif
634	{
635	/* May be an add exec */
636	struct ex_list *ex_ptr;
637	for(ex_ptr = exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next) {
638	if(ex_ptr->ex_fport == so->so_fport &&
639	lastbyte == ex_ptr->ex_addr) {
640	so->so_state \|= SS_CTL;
641	break;
642	}
643	}
644	}
645	if(so->so_state & SS_CTL) goto cont_input;
646	}
647	/* CTL_ALIAS: Do nothing, tcp_fconnect will be called on it */
648	}
649
650	if (so->so_emu & EMU_NOCONNECT) {
651	so->so_emu &= ~EMU_NOCONNECT;
652	goto cont_input;
653	}
654
655	if((tcp_fconnect(pData, so) == -1) && (errno != EINPROGRESS) && (errno != EWOULDBLOCK)) {
656	u_char code=ICMP_UNREACH_NET;
657	DEBUG_MISC((dfd," tcp fconnect errno = %d-%s\n",
658	errno,strerror(errno)));
659	if(errno == ECONNREFUSED) {
660	/* ACK the SYN, send RST to refuse the connection */
661	tcp_respond(pData, tp, ti, m, ti->ti_seq+1, (tcp_seq)0,
662	TH_RST\|TH_ACK);
663	} else {
664	if(errno == EHOSTUNREACH) code=ICMP_UNREACH_HOST;
665	HTONL(ti->ti_seq); /* restore tcp header */
666	HTONL(ti->ti_ack);
667	HTONS(ti->ti_win);
668	HTONS(ti->ti_urp);
669	m->m_data -= sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
670	m->m_len += sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
671	*ip=save_ip;
672	icmp_error(pData, m, ICMP_UNREACH,code, 0,strerror(errno));
673	}
674	tp = tcp_close(pData, tp);
675	m_free(pData, m);
676	} else {
677	/*
678	* Haven't connected yet, save the current mbuf
679	* and ti, and return
680	* XXX Some OS's don't tell us whether the connect()
681	* succeeded or not. So we must time it out.
682	*/
683	so->so_m = m;
684	so->so_ti = ti;
685	tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT;
686	tp->t_state = TCPS_SYN_RECEIVED;
687	}
688	return;
689
690	cont_conn:
691	/* m==NULL
692	* Check if the connect succeeded
693	*/
694	if (so->so_state & SS_NOFDREF) {
695	tp = tcp_close(pData, tp);
696	goto dropwithreset;
697	}
698	cont_input:
699	tcp_template(tp);
700
701	if (optp)
702	tcp_dooptions(pData, tp, (u_char *)optp, optlen, ti);
703	/* , */
704	/* &ts_present, &ts_val, &ts_ecr); */
705
706	if (iss)
707	tp->iss = iss;
708	else
709	tp->iss = tcp_iss;
710	tcp_iss += TCP_ISSINCR/2;
711	tp->irs = ti->ti_seq;
712	tcp_sendseqinit(tp);
713	tcp_rcvseqinit(tp);
714	tp->t_flags \|= TF_ACKNOW;
715	tp->t_state = TCPS_SYN_RECEIVED;
716	tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT;
717	tcpstat.tcps_accepts++;
718	goto trimthenstep6;
719	} /* case TCPS_LISTEN */
720
721	/*
722	* If the state is SYN_SENT:
723	* if seg contains an ACK, but not for our SYN, drop the input.
724	* if seg contains a RST, then drop the connection.
725	* if seg does not contain SYN, then drop it.
726	* Otherwise this is an acceptable SYN segment
727	* initialize tp->rcv_nxt and tp->irs
728	* if seg contains ack then advance tp->snd_una
729	* if SYN has been acked change to ESTABLISHED else SYN_RCVD state
730	* arrange for segment to be acked (eventually)
731	* continue processing rest of data/controls, beginning with URG
732	*/
733	case TCPS_SYN_SENT:
734	if ((tiflags & TH_ACK) &&
735	(SEQ_LEQ(ti->ti_ack, tp->iss) \|\|
736	SEQ_GT(ti->ti_ack, tp->snd_max)))
737	goto dropwithreset;
738
739	if (tiflags & TH_RST) {
740	if (tiflags & TH_ACK)
741	tp = tcp_drop(pData, tp,0); /* XXX Check t_softerror! */
742	goto drop;
743	}
744
745	if ((tiflags & TH_SYN) == 0)
746	goto drop;
747	if (tiflags & TH_ACK) {
748	tp->snd_una = ti->ti_ack;
749	if (SEQ_LT(tp->snd_nxt, tp->snd_una))
750	tp->snd_nxt = tp->snd_una;
751	}
752
753	tp->t_timer[TCPT_REXMT] = 0;
754	tp->irs = ti->ti_seq;
755	tcp_rcvseqinit(tp);
756	tp->t_flags \|= TF_ACKNOW;
757	if (tiflags & TH_ACK && SEQ_GT(tp->snd_una, tp->iss)) {
758	tcpstat.tcps_connects++;
759	soisfconnected(so);
760	tp->t_state = TCPS_ESTABLISHED;
761
762	/* Do window scaling on this connection? */
763	/* if ((tp->t_flags & (TF_RCVD_SCALE\|TF_REQ_SCALE)) ==
764	* (TF_RCVD_SCALE\|TF_REQ_SCALE)) {
765	* tp->snd_scale = tp->requested_s_scale;
766	* tp->rcv_scale = tp->request_r_scale;
767	* }
768	*/
769	(void) tcp_reass(pData, tp, (struct tcpiphdr *)0,
770	(struct mbuf *)0);
771	/*
772	* if we didn't have to retransmit the SYN,
773	* use its rtt as our initial srtt & rtt var.
774	*/
775	if (tp->t_rtt)
776	tcp_xmit_timer(pData, tp, tp->t_rtt);
777	} else
778	tp->t_state = TCPS_SYN_RECEIVED;
779
780	trimthenstep6:
781	/*
782	* Advance ti->ti_seq to correspond to first data byte.
783	* If data, trim to stay within window,
784	* dropping FIN if necessary.
785	*/
786	ti->ti_seq++;
787	if (ti->ti_len > tp->rcv_wnd) {
788	todrop = ti->ti_len - tp->rcv_wnd;
789	m_adj(m, -todrop);
790	ti->ti_len = tp->rcv_wnd;
791	tiflags &= ~TH_FIN;
792	tcpstat.tcps_rcvpackafterwin++;
793	tcpstat.tcps_rcvbyteafterwin += todrop;
794	}
795	tp->snd_wl1 = ti->ti_seq - 1;
796	tp->rcv_up = ti->ti_seq;
797	goto step6;
798	} /* switch tp->t_state */
799	/*
800	* States other than LISTEN or SYN_SENT.
801	* First check timestamp, if present.
802	* Then check that at least some bytes of segment are within
803	* receive window. If segment begins before rcv_nxt,
804	* drop leading data (and SYN); if nothing left, just ack.
805	*
806	* RFC 1323 PAWS: If we have a timestamp reply on this segment
807	* and it's less than ts_recent, drop it.
808	*/
809	/* if (ts_present && (tiflags & TH_RST) == 0 && tp->ts_recent &&
810	* TSTMP_LT(ts_val, tp->ts_recent)) {
811	*
812	/ / Check to see if ts_recent is over 24 days old. */
813	/* if ((int)(tcp_now - tp->ts_recent_age) > TCP_PAWS_IDLE) {
814	/ /
815	* * Invalidate ts_recent. If this segment updates
816	* * ts_recent, the age will be reset later and ts_recent
817	* * will get a valid value. If it does not, setting
818	* * ts_recent to zero will at least satisfy the
819	* * requirement that zero be placed in the timestamp
820	* * echo reply when ts_recent isn't valid. The
821	* * age isn't reset until we get a valid ts_recent
822	* * because we don't want out-of-order segments to be
823	* * dropped when ts_recent is old.
824	* */
825	/* tp->ts_recent = 0;
826	* } else {
827	* tcpstat.tcps_rcvduppack++;
828	* tcpstat.tcps_rcvdupbyte += ti->ti_len;
829	* tcpstat.tcps_pawsdrop++;
830	* goto dropafterack;
831	* }
832	* }
833	*/
834
835	todrop = tp->rcv_nxt - ti->ti_seq;
836	if (todrop > 0) {
837	if (tiflags & TH_SYN) {
838	tiflags &= ~TH_SYN;
839	ti->ti_seq++;
840	if (ti->ti_urp > 1)
841	ti->ti_urp--;
842	else
843	tiflags &= ~TH_URG;
844	todrop--;
845	}
846	/*
847	* Following if statement from Stevens, vol. 2, p. 960.
848	*/
849	if (todrop > ti->ti_len
850	\|\| (todrop == ti->ti_len && (tiflags & TH_FIN) == 0)) {
851	/*
852	* Any valid FIN must be to the left of the window.
853	* At this point the FIN must be a duplicate or out
854	* of sequence; drop it.
855	*/
856	tiflags &= ~TH_FIN;
857
858	/*
859	* Send an ACK to resynchronize and drop any data.
860	* But keep on processing for RST or ACK.
861	*/
862	tp->t_flags \|= TF_ACKNOW;
863	todrop = ti->ti_len;
864	tcpstat.tcps_rcvduppack++;
865	tcpstat.tcps_rcvdupbyte += todrop;
866	} else {
867	tcpstat.tcps_rcvpartduppack++;
868	tcpstat.tcps_rcvpartdupbyte += todrop;
869	}
870	m_adj(m, todrop);
871	ti->ti_seq += todrop;
872	ti->ti_len -= todrop;
873	if (ti->ti_urp > todrop)
874	ti->ti_urp -= todrop;
875	else {
876	tiflags &= ~TH_URG;
877	ti->ti_urp = 0;
878	}
879	}
880	/*
881	* If new data are received on a connection after the
882	* user processes are gone, then RST the other end.
883	*/
884	if ((so->so_state & SS_NOFDREF) &&
885	tp->t_state > TCPS_CLOSE_WAIT && ti->ti_len) {
886	tp = tcp_close(pData, tp);
887	tcpstat.tcps_rcvafterclose++;
888	goto dropwithreset;
889	}
890
891	/*
892	* If segment ends after window, drop trailing data
893	* (and PUSH and FIN); if nothing left, just ACK.
894	*/
895	todrop = (ti->ti_seq+ti->ti_len) - (tp->rcv_nxt+tp->rcv_wnd);
896	if (todrop > 0) {
897	tcpstat.tcps_rcvpackafterwin++;
898	if (todrop >= ti->ti_len) {
899	tcpstat.tcps_rcvbyteafterwin += ti->ti_len;
900	/*
901	* If a new connection request is received
902	* while in TIME_WAIT, drop the old connection
903	* and start over if the sequence numbers
904	* are above the previous ones.
905	*/
906	if (tiflags & TH_SYN &&
907	tp->t_state == TCPS_TIME_WAIT &&
908	SEQ_GT(ti->ti_seq, tp->rcv_nxt)) {
909	iss = tp->rcv_nxt + TCP_ISSINCR;
910	tp = tcp_close(pData, tp);
911	goto findso;
912	}
913	/*
914	* If window is closed can only take segments at
915	* window edge, and have to drop data and PUSH from
916	* incoming segments. Continue processing, but
917	* remember to ack. Otherwise, drop segment
918	* and ack.
919	*/
920	if (tp->rcv_wnd == 0 && ti->ti_seq == tp->rcv_nxt) {
921	tp->t_flags \|= TF_ACKNOW;
922	tcpstat.tcps_rcvwinprobe++;
923	} else
924	goto dropafterack;
925	} else
926	tcpstat.tcps_rcvbyteafterwin += todrop;
927	m_adj(m, -todrop);
928	ti->ti_len -= todrop;
929	tiflags &= ~(TH_PUSH\|TH_FIN);
930	}
931
932	/*
933	* If last ACK falls within this segment's sequence numbers,
934	* record its timestamp.
935	*/
936	/* if (ts_present && SEQ_LEQ(ti->ti_seq, tp->last_ack_sent) &&
937	* SEQ_LT(tp->last_ack_sent, ti->ti_seq + ti->ti_len +
938	* ((tiflags & (TH_SYN\|TH_FIN)) != 0))) {
939	* tp->ts_recent_age = tcp_now;
940	* tp->ts_recent = ts_val;
941	* }
942	*/
943
944	/*
945	* If the RST bit is set examine the state:
946	* SYN_RECEIVED STATE:
947	* If passive open, return to LISTEN state.
948	* If active open, inform user that connection was refused.
949	* ESTABLISHED, FIN_WAIT_1, FIN_WAIT2, CLOSE_WAIT STATES:
950	* Inform user that connection was reset, and close tcb.
951	* CLOSING, LAST_ACK, TIME_WAIT STATES
952	* Close the tcb.
953	*/
954	if (tiflags&TH_RST) switch (tp->t_state) {
955
956	case TCPS_SYN_RECEIVED:
957	/* so->so_error = ECONNREFUSED; */
958	goto close;
959
960	case TCPS_ESTABLISHED:
961	case TCPS_FIN_WAIT_1:
962	case TCPS_FIN_WAIT_2:
963	case TCPS_CLOSE_WAIT:
964	/* so->so_error = ECONNRESET; */
965	close:
966	tp->t_state = TCPS_CLOSED;
967	tcpstat.tcps_drops++;
968	tp = tcp_close(pData, tp);
969	goto drop;
970
971	case TCPS_CLOSING:
972	case TCPS_LAST_ACK:
973	case TCPS_TIME_WAIT:
974	tp = tcp_close(pData, tp);
975	goto drop;
976	}
977
978	/*
979	* If a SYN is in the window, then this is an
980	* error and we send an RST and drop the connection.
981	*/
982	if (tiflags & TH_SYN) {
983	tp = tcp_drop(pData, tp,0);
984	goto dropwithreset;
985	}
986
987	/*
988	* If the ACK bit is off we drop the segment and return.
989	*/
990	if ((tiflags & TH_ACK) == 0) goto drop;
991
992	/*
993	* Ack processing.
994	*/
995	switch (tp->t_state) {
996	/*
997	* In SYN_RECEIVED state if the ack ACKs our SYN then enter
998	* ESTABLISHED state and continue processing, otherwise
999	* send an RST. una<=ack<=max
1000	*/
1001	case TCPS_SYN_RECEIVED:
1002
1003	if (SEQ_GT(tp->snd_una, ti->ti_ack) \|\|
1004	SEQ_GT(ti->ti_ack, tp->snd_max))
1005	goto dropwithreset;
1006	tcpstat.tcps_connects++;
1007	tp->t_state = TCPS_ESTABLISHED;
1008	/*
1009	* The sent SYN is ack'ed with our sequence number +1
1010	* The first data byte already in the buffer will get
1011	* lost if no correction is made. This is only needed for
1012	* SS_CTL since the buffer is empty otherwise.
1013	* tp->snd_una++; or:
1014	*/
1015	tp->snd_una=ti->ti_ack;
1016	if (so->so_state & SS_CTL) {
1017	/* So tcp_ctl reports the right state */
1018	ret = tcp_ctl(pData, so);
1019	if (ret == 1) {
1020	soisfconnected(so);
1021	so->so_state &= ~SS_CTL; /* success XXX */
1022	} else if (ret == 2) {
1023	so->so_state = SS_NOFDREF; /* CTL_CMD */
1024	} else {
1025	needoutput = 1;
1026	tp->t_state = TCPS_FIN_WAIT_1;
1027	}
1028	} else {
1029	soisfconnected(so);
1030	}
1031
1032	/* Do window scaling? */
1033	/* if ((tp->t_flags & (TF_RCVD_SCALE\|TF_REQ_SCALE)) ==
1034	* (TF_RCVD_SCALE\|TF_REQ_SCALE)) {
1035	* tp->snd_scale = tp->requested_s_scale;
1036	* tp->rcv_scale = tp->request_r_scale;
1037	* }
1038	*/
1039	(void) tcp_reass(pData, tp, (struct tcpiphdr )0, (struct mbuf )0);
1040	tp->snd_wl1 = ti->ti_seq - 1;
1041	/* Avoid ack processing; snd_una==ti_ack => dup ack */
1042	goto synrx_to_est;
1043	/* fall into ... */
1044
1045	/*
1046	* In ESTABLISHED state: drop duplicate ACKs; ACK out of range
1047	* ACKs. If the ack is in the range
1048	* tp->snd_una < ti->ti_ack <= tp->snd_max
1049	* then advance tp->snd_una to ti->ti_ack and drop
1050	* data from the retransmission queue. If this ACK reflects
1051	* more up to date window information we update our window information.
1052	*/
1053	case TCPS_ESTABLISHED:
1054	case TCPS_FIN_WAIT_1:
1055	case TCPS_FIN_WAIT_2:
1056	case TCPS_CLOSE_WAIT:
1057	case TCPS_CLOSING:
1058	case TCPS_LAST_ACK:
1059	case TCPS_TIME_WAIT:
1060
1061	if (SEQ_LEQ(ti->ti_ack, tp->snd_una)) {
1062	if (ti->ti_len == 0 && tiwin == tp->snd_wnd) {
1063	tcpstat.tcps_rcvdupack++;
1064	DEBUG_MISC((dfd," dup ack m = %lx so = %lx \n",
1065	(long )m, (long )so));
1066	/*
1067	* If we have outstanding data (other than
1068	* a window probe), this is a completely
1069	* duplicate ack (ie, window info didn't
1070	* change), the ack is the biggest we've
1071	* seen and we've seen exactly our rexmt
1072	* threshold of them, assume a packet
1073	* has been dropped and retransmit it.
1074	* Kludge snd_nxt & the congestion
1075	* window so we send only this one
1076	* packet.
1077	*
1078	* We know we're losing at the current
1079	* window size so do congestion avoidance
1080	* (set ssthresh to half the current window
1081	* and pull our congestion window back to
1082	* the new ssthresh).
1083	*
1084	* Dup acks mean that packets have left the
1085	* network (they're now cached at the receiver)
1086	* so bump cwnd by the amount in the receiver
1087	* to keep a constant cwnd packets in the
1088	* network.
1089	*/
1090	if (tp->t_timer[TCPT_REXMT] == 0 \|\|
1091	ti->ti_ack != tp->snd_una)
1092	tp->t_dupacks = 0;
1093	else if (++tp->t_dupacks == tcprexmtthresh) {
1094	tcp_seq onxt = tp->snd_nxt;
1095	u_int win =
1096	min(tp->snd_wnd, tp->snd_cwnd) / 2 /
1097	tp->t_maxseg;
1098
1099	if (win < 2)
1100	win = 2;
1101	tp->snd_ssthresh = win * tp->t_maxseg;
1102	tp->t_timer[TCPT_REXMT] = 0;
1103	tp->t_rtt = 0;
1104	tp->snd_nxt = ti->ti_ack;
1105	tp->snd_cwnd = tp->t_maxseg;
1106	(void) tcp_output(pData, tp);
1107	tp->snd_cwnd = tp->snd_ssthresh +
1108	tp->t_maxseg * tp->t_dupacks;
1109	if (SEQ_GT(onxt, tp->snd_nxt))
1110	tp->snd_nxt = onxt;
1111	goto drop;
1112	} else if (tp->t_dupacks > tcprexmtthresh) {
1113	tp->snd_cwnd += tp->t_maxseg;
1114	(void) tcp_output(pData, tp);
1115	goto drop;
1116	}
1117	} else
1118	tp->t_dupacks = 0;
1119	break;
1120	}
1121	synrx_to_est:
1122	/*
1123	* If the congestion window was inflated to account
1124	* for the other side's cached packets, retract it.
1125	*/
1126	if (tp->t_dupacks > tcprexmtthresh &&
1127	tp->snd_cwnd > tp->snd_ssthresh)
1128	tp->snd_cwnd = tp->snd_ssthresh;
1129	tp->t_dupacks = 0;
1130	if (SEQ_GT(ti->ti_ack, tp->snd_max)) {
1131	tcpstat.tcps_rcvacktoomuch++;
1132	goto dropafterack;
1133	}
1134	acked = ti->ti_ack - tp->snd_una;
1135	tcpstat.tcps_rcvackpack++;
1136	tcpstat.tcps_rcvackbyte += acked;
1137
1138	/*
1139	* If we have a timestamp reply, update smoothed
1140	* round trip time. If no timestamp is present but
1141	* transmit timer is running and timed sequence
1142	* number was acked, update smoothed round trip time.
1143	* Since we now have an rtt measurement, cancel the
1144	* timer backoff (cf., Phil Karn's retransmit alg.).
1145	* Recompute the initial retransmit timer.
1146	*/
1147	/* if (ts_present)
1148	* tcp_xmit_timer(tp, tcp_now-ts_ecr+1);
1149	* else
1150	*/
1151	if (tp->t_rtt && SEQ_GT(ti->ti_ack, tp->t_rtseq))
1152	tcp_xmit_timer(pData, tp,tp->t_rtt);
1153
1154	/*
1155	* If all outstanding data is acked, stop retransmit
1156	* timer and remember to restart (more output or persist).
1157	* If there is more data to be acked, restart retransmit
1158	* timer, using current (possibly backed-off) value.
1159	*/
1160	if (ti->ti_ack == tp->snd_max) {
1161	tp->t_timer[TCPT_REXMT] = 0;
1162	needoutput = 1;
1163	} else if (tp->t_timer[TCPT_PERSIST] == 0)
1164	tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
1165	/*
1166	* When new data is acked, open the congestion window.
1167	* If the window gives us less than ssthresh packets
1168	* in flight, open exponentially (maxseg per packet).
1169	* Otherwise open linearly: maxseg per window
1170	* (maxseg^2 / cwnd per packet).
1171	*/
1172	{
1173	register u_int cw = tp->snd_cwnd;
1174	register u_int incr = tp->t_maxseg;
1175
1176	if (cw > tp->snd_ssthresh)
1177	incr = incr * incr / cw;
1178	tp->snd_cwnd = min(cw + incr, TCP_MAXWIN<<tp->snd_scale);
1179	}
1180	if (acked > so->so_snd.sb_cc) {
1181	tp->snd_wnd -= so->so_snd.sb_cc;
1182	sbdrop(&so->so_snd, (int )so->so_snd.sb_cc);
1183	ourfinisacked = 1;
1184	} else {
1185	sbdrop(&so->so_snd, acked);
1186	tp->snd_wnd -= acked;
1187	ourfinisacked = 0;
1188	}
1189	/*
1190	* XXX sowwakup is called when data is acked and there's room for
1191	* for more data... it should read() the socket
1192	*/
1193	/* if (so->so_snd.sb_flags & SB_NOTIFY)
1194	* sowwakeup(so);
1195	*/
1196	tp->snd_una = ti->ti_ack;
1197	if (SEQ_LT(tp->snd_nxt, tp->snd_una))
1198	tp->snd_nxt = tp->snd_una;
1199
1200	switch (tp->t_state) {
1201
1202	/*
1203	* In FIN_WAIT_1 STATE in addition to the processing
1204	* for the ESTABLISHED state if our FIN is now acknowledged
1205	* then enter FIN_WAIT_2.
1206	*/
1207	case TCPS_FIN_WAIT_1:
1208	if (ourfinisacked) {
1209	/*
1210	* If we can't receive any more
1211	* data, then closing user can proceed.
1212	* Starting the timer is contrary to the
1213	* specification, but if we don't get a FIN
1214	* we'll hang forever.
1215	*/
1216	if (so->so_state & SS_FCANTRCVMORE) {
1217	soisfdisconnected(so);
1218	tp->t_timer[TCPT_2MSL] = tcp_maxidle;
1219	}
1220	tp->t_state = TCPS_FIN_WAIT_2;
1221	}
1222	break;
1223
1224	/*
1225	* In CLOSING STATE in addition to the processing for
1226	* the ESTABLISHED state if the ACK acknowledges our FIN
1227	* then enter the TIME-WAIT state, otherwise ignore
1228	* the segment.
1229	*/
1230	case TCPS_CLOSING:
1231	if (ourfinisacked) {
1232	tp->t_state = TCPS_TIME_WAIT;
1233	tcp_canceltimers(tp);
1234	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1235	soisfdisconnected(so);
1236	}
1237	break;
1238
1239	/*
1240	* In LAST_ACK, we may still be waiting for data to drain
1241	* and/or to be acked, as well as for the ack of our FIN.
1242	* If our FIN is now acknowledged, delete the TCB,
1243	* enter the closed state and return.
1244	*/
1245	case TCPS_LAST_ACK:
1246	if (ourfinisacked) {
1247	tp = tcp_close(pData, tp);
1248	goto drop;
1249	}
1250	break;
1251
1252	/*
1253	* In TIME_WAIT state the only thing that should arrive
1254	* is a retransmission of the remote FIN. Acknowledge
1255	* it and restart the finack timer.
1256	*/
1257	case TCPS_TIME_WAIT:
1258	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1259	goto dropafterack;
1260	}
1261	} /* switch(tp->t_state) */
1262
1263	step6:
1264	/*
1265	* Update window information.
1266	* Don't look at window if no ACK: TAC's send garbage on first SYN.
1267	*/
1268	if ((tiflags & TH_ACK) &&
1269	(SEQ_LT(tp->snd_wl1, ti->ti_seq) \|\|
1270	(tp->snd_wl1 == ti->ti_seq && (SEQ_LT(tp->snd_wl2, ti->ti_ack) \|\|
1271	(tp->snd_wl2 == ti->ti_ack && tiwin > tp->snd_wnd))))) {
1272	/* keep track of pure window updates */
1273	if (ti->ti_len == 0 &&
1274	tp->snd_wl2 == ti->ti_ack && tiwin > tp->snd_wnd)
1275	tcpstat.tcps_rcvwinupd++;
1276	tp->snd_wnd = tiwin;
1277	tp->snd_wl1 = ti->ti_seq;
1278	tp->snd_wl2 = ti->ti_ack;
1279	if (tp->snd_wnd > tp->max_sndwnd)
1280	tp->max_sndwnd = tp->snd_wnd;
1281	needoutput = 1;
1282	}
1283
1284	/*
1285	* Process segments with URG.
1286	*/
1287	if ((tiflags & TH_URG) && ti->ti_urp &&
1288	TCPS_HAVERCVDFIN(tp->t_state) == 0) {
1289	/*
1290	* This is a kludge, but if we receive and accept
1291	* random urgent pointers, we'll crash in
1292	* soreceive. It's hard to imagine someone
1293	* actually wanting to send this much urgent data.
1294	*/
1295	if (ti->ti_urp + so->so_rcv.sb_cc > so->so_rcv.sb_datalen) {
1296	ti->ti_urp = 0;
1297	tiflags &= ~TH_URG;
1298	goto dodata;
1299	}
1300	/*
1301	* If this segment advances the known urgent pointer,
1302	* then mark the data stream. This should not happen
1303	* in CLOSE_WAIT, CLOSING, LAST_ACK or TIME_WAIT STATES since
1304	* a FIN has been received from the remote side.
1305	* In these states we ignore the URG.
1306	*
1307	* According to RFC961 (Assigned Protocols),
1308	* the urgent pointer points to the last octet
1309	* of urgent data. We continue, however,
1310	* to consider it to indicate the first octet
1311	* of data past the urgent section as the original
1312	* spec states (in one of two places).
1313	*/
1314	if (SEQ_GT(ti->ti_seq+ti->ti_urp, tp->rcv_up)) {
1315	tp->rcv_up = ti->ti_seq + ti->ti_urp;
1316	so->so_urgc = so->so_rcv.sb_cc +
1317	(tp->rcv_up - tp->rcv_nxt); /* -1; */
1318	tp->rcv_up = ti->ti_seq + ti->ti_urp;
1319
1320	}
1321	} else
1322	/*
1323	* If no out of band data is expected,
1324	* pull receive urgent pointer along
1325	* with the receive window.
1326	*/
1327	if (SEQ_GT(tp->rcv_nxt, tp->rcv_up))
1328	tp->rcv_up = tp->rcv_nxt;
1329	dodata:
1330
1331	/*
1332	* Process the segment text, merging it into the TCP sequencing queue,
1333	* and arranging for acknowledgment of receipt if necessary.
1334	* This process logically involves adjusting tp->rcv_wnd as data
1335	* is presented to the user (this happens in tcp_usrreq.c,
1336	* case PRU_RCVD). If a FIN has already been received on this
1337	* connection then we just ignore the text.
1338	*/
1339	if ((ti->ti_len \|\| (tiflags&TH_FIN)) &&
1340	TCPS_HAVERCVDFIN(tp->t_state) == 0) {
1341	TCP_REASS(pData, tp, ti, m, so, tiflags);
1342	/*
1343	* Note the amount of data that peer has sent into
1344	* our window, in order to estimate the sender's
1345	* buffer size.
1346	*/
1347	len = so->so_rcv.sb_datalen - (tp->rcv_adv - tp->rcv_nxt);
1348	} else {
1349	m_free(pData, m);
1350	tiflags &= ~TH_FIN;
1351	}
1352
1353	/*
1354	* If FIN is received ACK the FIN and let the user know
1355	* that the connection is closing.
1356	*/
1357	if (tiflags & TH_FIN) {
1358	if (TCPS_HAVERCVDFIN(tp->t_state) == 0) {
1359	/*
1360	* If we receive a FIN we can't send more data,
1361	* set it SS_FDRAIN
1362	* Shutdown the socket if there is no rx data in the
1363	* buffer.
1364	* soread() is called on completion of shutdown() and
1365	* will got to TCPS_LAST_ACK, and use tcp_output()
1366	* to send the FIN.
1367	*/
1368	/* sofcantrcvmore(so); */
1369	sofwdrain(so);
1370
1371	tp->t_flags \|= TF_ACKNOW;
1372	tp->rcv_nxt++;
1373	}
1374	switch (tp->t_state) {
1375
1376	/*
1377	* In SYN_RECEIVED and ESTABLISHED STATES
1378	* enter the CLOSE_WAIT state.
1379	*/
1380	case TCPS_SYN_RECEIVED:
1381	case TCPS_ESTABLISHED:
1382	if(so->so_emu == EMU_CTL) /* no shutdown on socket */
1383	tp->t_state = TCPS_LAST_ACK;
1384	else
1385	tp->t_state = TCPS_CLOSE_WAIT;
1386	break;
1387
1388	/*
1389	* If still in FIN_WAIT_1 STATE FIN has not been acked so
1390	* enter the CLOSING state.
1391	*/
1392	case TCPS_FIN_WAIT_1:
1393	tp->t_state = TCPS_CLOSING;
1394	break;
1395
1396	/*
1397	* In FIN_WAIT_2 state enter the TIME_WAIT state,
1398	* starting the time-wait timer, turning off the other
1399	* standard timers.
1400	*/
1401	case TCPS_FIN_WAIT_2:
1402	tp->t_state = TCPS_TIME_WAIT;
1403	tcp_canceltimers(tp);
1404	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1405	soisfdisconnected(so);
1406	break;
1407
1408	/*
1409	* In TIME_WAIT state restart the 2 MSL time_wait timer.
1410	*/
1411	case TCPS_TIME_WAIT:
1412	tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1413	break;
1414	}
1415	}
1416
1417	/*
1418	* If this is a small packet, then ACK now - with Nagel
1419	* congestion avoidance sender won't send more until
1420	* he gets an ACK.
1421	*
1422	* See above.
1423	*/
1424	/* if (ti->ti_len && (unsigned)ti->ti_len < tp->t_maxseg) {
1425	*/
1426	/* if ((ti->ti_len && (unsigned)ti->ti_len < tp->t_maxseg &&
1427	* (so->so_iptos & IPTOS_LOWDELAY) == 0) \|\|
1428	* ((so->so_iptos & IPTOS_LOWDELAY) &&
1429	* ((struct tcpiphdr_2 *)ti)->first_char == (char)27)) {
1430	*/
1431	if (ti->ti_len && (unsigned)ti->ti_len <= 5 &&
1432	((struct tcpiphdr_2 *)ti)->first_char == (char)27) {
1433	tp->t_flags \|= TF_ACKNOW;
1434	}
1435
1436	/*
1437	* Return any desired output.
1438	*/
1439	if (needoutput \|\| (tp->t_flags & TF_ACKNOW)) {
1440	(void) tcp_output(pData, tp);
1441	}
1442	return;
1443
1444	dropafterack:
1445	/*
1446	* Generate an ACK dropping incoming segment if it occupies
1447	* sequence space, where the ACK reflects our state.
1448	*/
1449	if (tiflags & TH_RST)
1450	goto drop;
1451	m_freem(pData, m);
1452	tp->t_flags \|= TF_ACKNOW;
1453	(void) tcp_output(pData, tp);
1454	return;
1455
1456	dropwithreset:
1457	/* reuses m if m!=NULL, m_free() unnecessary */
1458	if (tiflags & TH_ACK)
1459	tcp_respond(pData, tp, ti, m, (tcp_seq)0, ti->ti_ack, TH_RST);
1460	else {
1461	if (tiflags & TH_SYN) ti->ti_len++;
1462	tcp_respond(pData, tp, ti, m, ti->ti_seq+ti->ti_len, (tcp_seq)0,
1463	TH_RST\|TH_ACK);
1464	}
1465
1466	return;
1467
1468	drop:
1469	/*
1470	* Drop space held by incoming segment and return.
1471	*/
1472	m_free(pData, m);
1473
1474	return;
1475	}
1476
1477	/* , ts_present, ts_val, ts_ecr) */
1478	/* int *ts_present;
1479	* u_int32_t ts_val, ts_ecr;
1480	*/
1481	void
1482	tcp_dooptions(PNATState pData, struct tcpcb tp, u_char cp, int cnt, struct tcpiphdr *ti)
1483	{
1484	u_int16_t mss;
1485	int opt, optlen;
1486
1487	DEBUG_CALL("tcp_dooptions");
1488	DEBUG_ARGS((dfd," tp = %lx cnt=%i \n", (long )tp, cnt));
1489
1490	for (; cnt > 0; cnt -= optlen, cp += optlen) {
1491	opt = cp[0];
1492	if (opt == TCPOPT_EOL)
1493	break;
1494	if (opt == TCPOPT_NOP)
1495	optlen = 1;
1496	else {
1497	optlen = cp[1];
1498	if (optlen <= 0)
1499	break;
1500	}
1501	switch (opt) {
1502
1503	default:
1504	continue;
1505
1506	case TCPOPT_MAXSEG:
1507	if (optlen != TCPOLEN_MAXSEG)
1508	continue;
1509	if (!(ti->ti_flags & TH_SYN))
1510	continue;
1511	memcpy((char ) &mss, (char ) cp + 2, sizeof(mss));
1512	NTOHS(mss);
1513	(void) tcp_mss(pData, tp, mss); /* sets t_maxseg */
1514	break;
1515
1516	/* case TCPOPT_WINDOW:
1517	* if (optlen != TCPOLEN_WINDOW)
1518	* continue;
1519	* if (!(ti->ti_flags & TH_SYN))
1520	* continue;
1521	* tp->t_flags \|= TF_RCVD_SCALE;
1522	* tp->requested_s_scale = min(cp[2], TCP_MAX_WINSHIFT);
1523	* break;
1524	*/
1525	/* case TCPOPT_TIMESTAMP:
1526	* if (optlen != TCPOLEN_TIMESTAMP)
1527	* continue;
1528	* *ts_present = 1;
1529	* memcpy((char ) ts_val, (char )cp + 2, sizeof(*ts_val));
1530	* NTOHL(*ts_val);
1531	* memcpy((char ) ts_ecr, (char )cp + 6, sizeof(*ts_ecr));
1532	* NTOHL(*ts_ecr);
1533	*
1534	/ /
1535	* * A timestamp received in a SYN makes
1536	* * it ok to send timestamp requests and replies.
1537	* */
1538	/* if (ti->ti_flags & TH_SYN) {
1539	* tp->t_flags \|= TF_RCVD_TSTMP;
1540	* tp->ts_recent = *ts_val;
1541	* tp->ts_recent_age = tcp_now;
1542	* }
1543	*/ break;
1544	}
1545	}
1546	}
1547
1548
1549	/*
1550	* Pull out of band byte out of a segment so
1551	* it doesn't appear in the user's data queue.
1552	* It is still reflected in the segment length for
1553	* sequencing purposes.
1554	*/
1555
1556	#ifdef notdef
1557
1558	void
1559	tcp_pulloutofband(so, ti, m)
1560	struct socket *so;
1561	struct tcpiphdr *ti;
1562	register struct mbuf *m;
1563	{
1564	int cnt = ti->ti_urp - 1;
1565
1566	while (cnt >= 0) {
1567	if (m->m_len > cnt) {
1568	char *cp = mtod(m, caddr_t) + cnt;
1569	struct tcpcb *tp = sototcpcb(so);
1570
1571	tp->t_iobc = *cp;
1572	tp->t_oobflags \|= TCPOOB_HAVEDATA;
1573	memcpy(sp, cp+1, (unsigned)(m->m_len - cnt - 1));
1574	m->m_len--;
1575	return;
1576	}
1577	cnt -= m->m_len;
1578	m = m->m_next; /* XXX WRONG! Fix it! */
1579	if (m == 0)
1580	break;
1581	}
1582	panic("tcp_pulloutofband");
1583	}
1584
1585	#endif /* notdef */
1586
1587	/*
1588	* Collect new round-trip time estimate
1589	* and update averages and current timeout.
1590	*/
1591
1592	void
1593	tcp_xmit_timer(PNATState pData, register struct tcpcb *tp, int rtt)
1594	{
1595	register short delta;
1596
1597	DEBUG_CALL("tcp_xmit_timer");
1598	DEBUG_ARG("tp = %lx", (long)tp);
1599	DEBUG_ARG("rtt = %d", rtt);
1600
1601	tcpstat.tcps_rttupdated++;
1602	if (tp->t_srtt != 0) {
1603	/*
1604	* srtt is stored as fixed point with 3 bits after the
1605	* binary point (i.e., scaled by 8). The following magic
1606	* is equivalent to the smoothing algorithm in rfc793 with
1607	* an alpha of .875 (srtt = rtt/8 + srtt*7/8 in fixed
1608	* point). Adjust rtt to origin 0.
1609	*/
1610	delta = rtt - 1 - (tp->t_srtt >> TCP_RTT_SHIFT);
1611	if ((tp->t_srtt += delta) <= 0)
1612	tp->t_srtt = 1;
1613	/*
1614	* We accumulate a smoothed rtt variance (actually, a
1615	* smoothed mean difference), then set the retransmit
1616	* timer to smoothed rtt + 4 times the smoothed variance.
1617	* rttvar is stored as fixed point with 2 bits after the
1618	* binary point (scaled by 4). The following is
1619	* equivalent to rfc793 smoothing with an alpha of .75
1620	* (rttvar = rttvar*3/4 + \|delta\| / 4). This replaces
1621	* rfc793's wired-in beta.
1622	*/
1623	if (delta < 0)
1624	delta = -delta;
1625	delta -= (tp->t_rttvar >> TCP_RTTVAR_SHIFT);
1626	if ((tp->t_rttvar += delta) <= 0)
1627	tp->t_rttvar = 1;
1628	} else {
1629	/*
1630	* No rtt measurement yet - use the unsmoothed rtt.
1631	* Set the variance to half the rtt (so our first
1632	* retransmit happens at 3*rtt).
1633	*/
1634	tp->t_srtt = rtt << TCP_RTT_SHIFT;
1635	tp->t_rttvar = rtt << (TCP_RTTVAR_SHIFT - 1);
1636	}
1637	tp->t_rtt = 0;
1638	tp->t_rxtshift = 0;
1639
1640	/*
1641	* the retransmit should happen at rtt + 4 * rttvar.
1642	* Because of the way we do the smoothing, srtt and rttvar
1643	* will each average +1/2 tick of bias. When we compute
1644	* the retransmit timer, we want 1/2 tick of rounding and
1645	* 1 extra tick because of +-1/2 tick uncertainty in the
1646	* firing of the timer. The bias will give us exactly the
1647	* 1.5 tick we need. But, because the bias is
1648	* statistical, we have to test that we don't drop below
1649	* the minimum feasible timer (which is 2 ticks).
1650	*/
1651	TCPT_RANGESET(tp->t_rxtcur, TCP_REXMTVAL(tp),
1652	(short)tp->t_rttmin, TCPTV_REXMTMAX); /* XXX */
1653
1654	/*
1655	* We received an ack for a packet that wasn't retransmitted;
1656	* it is probably safe to discard any error indications we've
1657	* received recently. This isn't quite right, but close enough
1658	* for now (a route might have failed after we sent a segment,
1659	* and the return path might not be symmetrical).
1660	*/
1661	tp->t_softerror = 0;
1662	}
1663
1664	/*
1665	* Determine a reasonable value for maxseg size.
1666	* If the route is known, check route for mtu.
1667	* If none, use an mss that can be handled on the outgoing
1668	* interface without forcing IP to fragment; if bigger than
1669	* an mbuf cluster (MCLBYTES), round down to nearest multiple of MCLBYTES
1670	* to utilize large mbufs. If no route is found, route has no mtu,
1671	* or the destination isn't local, use a default, hopefully conservative
1672	* size (usually 512 or the default IP max size, but no more than the mtu
1673	* of the interface), as we can't discover anything about intervening
1674	* gateways or networks. We also initialize the congestion/slow start
1675	* window to be a single segment if the destination isn't local.
1676	* While looking at the routing entry, we also initialize other path-dependent
1677	* parameters from pre-set or cached values in the routing entry.
1678	*/
1679
1680	int
1681	tcp_mss(PNATState pData, register struct tcpcb *tp, u_int offer)
1682	{
1683	struct socket *so = tp->t_socket;
1684	int mss;
1685
1686	DEBUG_CALL("tcp_mss");
1687	DEBUG_ARG("tp = %lx", (long)tp);
1688	DEBUG_ARG("offer = %d", offer);
1689
1690	mss = min(if_mtu, if_mru) - sizeof(struct tcpiphdr);
1691	if (offer)
1692	mss = min(mss, offer);
1693	mss = max(mss, 32);
1694	if (mss < tp->t_maxseg \|\| offer != 0)
1695	tp->t_maxseg = mss;
1696
1697	tp->snd_cwnd = mss;
1698
1699	sbreserve(&so->so_snd, tcp_sndspace+((tcp_sndspace%mss)?(mss-(tcp_sndspace%mss)):0));
1700	sbreserve(&so->so_rcv, tcp_rcvspace+((tcp_rcvspace%mss)?(mss-(tcp_rcvspace%mss)):0));
1701
1702	DEBUG_MISC((dfd, " returning mss = %d\n", mss));
1703
1704	return mss;
1705	}

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

source: vbox/trunk/src/VBox/Devices/Network/slirp/tcp_input.c@ 3021

Download in other formats: