tcp_input.c@ 13738

Last change on this file since 13738 was 13738, checked in by vboxsync, 16 years ago
removing extra if/ifndefs introduced defered socket removing, to prevent deletion of socket being in lock
Property svn:eol-style set to `native`
File size: 51.0 KB

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

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

source: vbox/trunk/src/VBox/Devices/Network/slirp/tcp_input.c@ 13738

Download in other formats: