xref: /netbsd-src/sys/netinet/tcp_input.c (revision 4724848cf0da353df257f730694b7882798e5daf)
1 /*	$NetBSD: tcp_input.c,v 1.438 2022/11/04 09:01:53 ozaki-r Exp $	*/
2 
3 /*
4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the project nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  */
31 
32 /*
33  *      @(#)COPYRIGHT   1.1 (NRL) 17 January 1995
34  *
35  * NRL grants permission for redistribution and use in source and binary
36  * forms, with or without modification, of the software and documentation
37  * created at NRL provided that the following conditions are met:
38  *
39  * 1. Redistributions of source code must retain the above copyright
40  *    notice, this list of conditions and the following disclaimer.
41  * 2. Redistributions in binary form must reproduce the above copyright
42  *    notice, this list of conditions and the following disclaimer in the
43  *    documentation and/or other materials provided with the distribution.
44  * 3. All advertising materials mentioning features or use of this software
45  *    must display the following acknowledgements:
46  *      This product includes software developed by the University of
47  *      California, Berkeley and its contributors.
48  *      This product includes software developed at the Information
49  *      Technology Division, US Naval Research Laboratory.
50  * 4. Neither the name of the NRL nor the names of its contributors
51  *    may be used to endorse or promote products derived from this software
52  *    without specific prior written permission.
53  *
54  * THE SOFTWARE PROVIDED BY NRL IS PROVIDED BY NRL AND CONTRIBUTORS ``AS
55  * IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
56  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A
57  * PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL NRL OR
58  * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
59  * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
60  * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
61  * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
62  * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
63  * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
64  * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
65  *
66  * The views and conclusions contained in the software and documentation
67  * are those of the authors and should not be interpreted as representing
68  * official policies, either expressed or implied, of the US Naval
69  * Research Laboratory (NRL).
70  */
71 
72 /*-
73  * Copyright (c) 1997, 1998, 1999, 2001, 2005, 2006,
74  * 2011 The NetBSD Foundation, Inc.
75  * All rights reserved.
76  *
77  * This code is derived from software contributed to The NetBSD Foundation
78  * by Coyote Point Systems, Inc.
79  * This code is derived from software contributed to The NetBSD Foundation
80  * by Jason R. Thorpe and Kevin M. Lahey of the Numerical Aerospace Simulation
81  * Facility, NASA Ames Research Center.
82  * This code is derived from software contributed to The NetBSD Foundation
83  * by Charles M. Hannum.
84  * This code is derived from software contributed to The NetBSD Foundation
85  * by Rui Paulo.
86  *
87  * Redistribution and use in source and binary forms, with or without
88  * modification, are permitted provided that the following conditions
89  * are met:
90  * 1. Redistributions of source code must retain the above copyright
91  *    notice, this list of conditions and the following disclaimer.
92  * 2. Redistributions in binary form must reproduce the above copyright
93  *    notice, this list of conditions and the following disclaimer in the
94  *    documentation and/or other materials provided with the distribution.
95  *
96  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
97  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
98  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
99  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
100  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
101  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
102  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
103  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
104  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
105  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
106  * POSSIBILITY OF SUCH DAMAGE.
107  */
108 
109 /*
110  * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1994, 1995
111  *	The Regents of the University of California.  All rights reserved.
112  *
113  * Redistribution and use in source and binary forms, with or without
114  * modification, are permitted provided that the following conditions
115  * are met:
116  * 1. Redistributions of source code must retain the above copyright
117  *    notice, this list of conditions and the following disclaimer.
118  * 2. Redistributions in binary form must reproduce the above copyright
119  *    notice, this list of conditions and the following disclaimer in the
120  *    documentation and/or other materials provided with the distribution.
121  * 3. Neither the name of the University nor the names of its contributors
122  *    may be used to endorse or promote products derived from this software
123  *    without specific prior written permission.
124  *
125  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
126  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
127  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
128  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
129  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
130  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
131  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
132  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
133  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
134  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
135  * SUCH DAMAGE.
136  *
137  *	@(#)tcp_input.c	8.12 (Berkeley) 5/24/95
138  */
139 
140 #include <sys/cdefs.h>
141 __KERNEL_RCSID(0, "$NetBSD: tcp_input.c,v 1.438 2022/11/04 09:01:53 ozaki-r Exp $");
142 
143 #ifdef _KERNEL_OPT
144 #include "opt_inet.h"
145 #include "opt_ipsec.h"
146 #include "opt_inet_csum.h"
147 #include "opt_tcp_debug.h"
148 #endif
149 
150 #include <sys/param.h>
151 #include <sys/systm.h>
152 #include <sys/malloc.h>
153 #include <sys/mbuf.h>
154 #include <sys/protosw.h>
155 #include <sys/socket.h>
156 #include <sys/socketvar.h>
157 #include <sys/errno.h>
158 #include <sys/syslog.h>
159 #include <sys/pool.h>
160 #include <sys/domain.h>
161 #include <sys/kernel.h>
162 #ifdef TCP_SIGNATURE
163 #include <sys/md5.h>
164 #endif
165 #include <sys/lwp.h> /* for lwp0 */
166 #include <sys/cprng.h>
167 
168 #include <net/if.h>
169 #include <net/if_types.h>
170 
171 #include <netinet/in.h>
172 #include <netinet/in_systm.h>
173 #include <netinet/ip.h>
174 #include <netinet/in_pcb.h>
175 #include <netinet/in_var.h>
176 #include <netinet/ip_var.h>
177 #include <netinet/in_offload.h>
178 
179 #if NARP > 0
180 #include <netinet/if_inarp.h>
181 #endif
182 #ifdef INET6
183 #include <netinet/ip6.h>
184 #include <netinet6/ip6_var.h>
185 #include <netinet6/in6_pcb.h>
186 #include <netinet6/ip6_var.h>
187 #include <netinet6/in6_var.h>
188 #include <netinet/icmp6.h>
189 #include <netinet6/nd6.h>
190 #ifdef TCP_SIGNATURE
191 #include <netinet6/scope6_var.h>
192 #endif
193 #endif
194 
195 #ifndef INET6
196 #include <netinet/ip6.h>
197 #endif
198 
199 #include <netinet/tcp.h>
200 #include <netinet/tcp_fsm.h>
201 #include <netinet/tcp_seq.h>
202 #include <netinet/tcp_timer.h>
203 #include <netinet/tcp_var.h>
204 #include <netinet/tcp_private.h>
205 #include <netinet/tcp_congctl.h>
206 #include <netinet/tcp_debug.h>
207 #include <netinet/tcp_syncache.h>
208 
209 #ifdef INET6
210 #include "faith.h"
211 #if defined(NFAITH) && NFAITH > 0
212 #include <net/if_faith.h>
213 #endif
214 #endif
215 
216 #ifdef IPSEC
217 #include <netipsec/ipsec.h>
218 #include <netipsec/key.h>
219 #ifdef INET6
220 #include <netipsec/ipsec6.h>
221 #endif
222 #endif	/* IPSEC*/
223 
224 #include <netinet/tcp_vtw.h>
225 
226 int	tcprexmtthresh = 3;
227 int	tcp_log_refused;
228 
229 int	tcp_do_autorcvbuf = 1;
230 int	tcp_autorcvbuf_inc = 16 * 1024;
231 int	tcp_autorcvbuf_max = 256 * 1024;
232 int	tcp_msl = (TCPTV_MSL / PR_SLOWHZ);
233 
234 static int tcp_rst_ppslim_count = 0;
235 static struct timeval tcp_rst_ppslim_last;
236 static int tcp_ackdrop_ppslim_count = 0;
237 static struct timeval tcp_ackdrop_ppslim_last;
238 
239 #define TCP_PAWS_IDLE	(24U * 24 * 60 * 60 * PR_SLOWHZ)
240 
241 /* for modulo comparisons of timestamps */
242 #define TSTMP_LT(a,b)	((int)((a)-(b)) < 0)
243 #define TSTMP_GEQ(a,b)	((int)((a)-(b)) >= 0)
244 
245 /*
246  * Neighbor Discovery, Neighbor Unreachability Detection Upper layer hint.
247  */
248 static void
249 nd_hint(struct tcpcb *tp)
250 {
251 	struct route *ro = NULL;
252 	struct rtentry *rt;
253 
254 	if (tp == NULL)
255 		return;
256 
257 	ro = &tp->t_inpcb->inp_route;
258 	if (ro == NULL)
259 		return;
260 
261 	rt = rtcache_validate(ro);
262 	if (rt == NULL)
263 		return;
264 
265 	switch (tp->t_family) {
266 #if NARP > 0
267 	case AF_INET:
268 		arp_nud_hint(rt);
269 		break;
270 #endif
271 #ifdef INET6
272 	case AF_INET6:
273 		nd6_nud_hint(rt);
274 		break;
275 #endif
276 	}
277 
278 	rtcache_unref(rt, ro);
279 }
280 
281 /*
282  * Compute ACK transmission behavior.  Delay the ACK unless
283  * we have already delayed an ACK (must send an ACK every two segments).
284  * We also ACK immediately if we received a PUSH and the ACK-on-PUSH
285  * option is enabled.
286  */
287 static void
288 tcp_setup_ack(struct tcpcb *tp, const struct tcphdr *th)
289 {
290 
291 	if (tp->t_flags & TF_DELACK ||
292 	    (tcp_ack_on_push && th->th_flags & TH_PUSH))
293 		tp->t_flags |= TF_ACKNOW;
294 	else
295 		TCP_SET_DELACK(tp);
296 }
297 
298 static void
299 icmp_check(struct tcpcb *tp, const struct tcphdr *th, int acked)
300 {
301 
302 	/*
303 	 * If we had a pending ICMP message that refers to data that have
304 	 * just been acknowledged, disregard the recorded ICMP message.
305 	 */
306 	if ((tp->t_flags & TF_PMTUD_PEND) &&
307 	    SEQ_GT(th->th_ack, tp->t_pmtud_th_seq))
308 		tp->t_flags &= ~TF_PMTUD_PEND;
309 
310 	/*
311 	 * Keep track of the largest chunk of data
312 	 * acknowledged since last PMTU update
313 	 */
314 	if (tp->t_pmtud_mss_acked < acked)
315 		tp->t_pmtud_mss_acked = acked;
316 }
317 
318 /*
319  * Convert TCP protocol fields to host order for easier processing.
320  */
321 static void
322 tcp_fields_to_host(struct tcphdr *th)
323 {
324 
325 	NTOHL(th->th_seq);
326 	NTOHL(th->th_ack);
327 	NTOHS(th->th_win);
328 	NTOHS(th->th_urp);
329 }
330 
331 /*
332  * ... and reverse the above.
333  */
334 static void
335 tcp_fields_to_net(struct tcphdr *th)
336 {
337 
338 	HTONL(th->th_seq);
339 	HTONL(th->th_ack);
340 	HTONS(th->th_win);
341 	HTONS(th->th_urp);
342 }
343 
344 static void
345 tcp_urp_drop(struct tcphdr *th, int todrop, int *tiflags)
346 {
347 	if (th->th_urp > todrop) {
348 		th->th_urp -= todrop;
349 	} else {
350 		*tiflags &= ~TH_URG;
351 		th->th_urp = 0;
352 	}
353 }
354 
355 #ifdef TCP_CSUM_COUNTERS
356 #include <sys/device.h>
357 
358 extern struct evcnt tcp_hwcsum_ok;
359 extern struct evcnt tcp_hwcsum_bad;
360 extern struct evcnt tcp_hwcsum_data;
361 extern struct evcnt tcp_swcsum;
362 #if defined(INET6)
363 extern struct evcnt tcp6_hwcsum_ok;
364 extern struct evcnt tcp6_hwcsum_bad;
365 extern struct evcnt tcp6_hwcsum_data;
366 extern struct evcnt tcp6_swcsum;
367 #endif /* defined(INET6) */
368 
369 #define	TCP_CSUM_COUNTER_INCR(ev)	(ev)->ev_count++
370 
371 #else
372 
373 #define	TCP_CSUM_COUNTER_INCR(ev)	/* nothing */
374 
375 #endif /* TCP_CSUM_COUNTERS */
376 
377 #ifdef TCP_REASS_COUNTERS
378 #include <sys/device.h>
379 
380 extern struct evcnt tcp_reass_;
381 extern struct evcnt tcp_reass_empty;
382 extern struct evcnt tcp_reass_iteration[8];
383 extern struct evcnt tcp_reass_prependfirst;
384 extern struct evcnt tcp_reass_prepend;
385 extern struct evcnt tcp_reass_insert;
386 extern struct evcnt tcp_reass_inserttail;
387 extern struct evcnt tcp_reass_append;
388 extern struct evcnt tcp_reass_appendtail;
389 extern struct evcnt tcp_reass_overlaptail;
390 extern struct evcnt tcp_reass_overlapfront;
391 extern struct evcnt tcp_reass_segdup;
392 extern struct evcnt tcp_reass_fragdup;
393 
394 #define	TCP_REASS_COUNTER_INCR(ev)	(ev)->ev_count++
395 
396 #else
397 
398 #define	TCP_REASS_COUNTER_INCR(ev)	/* nothing */
399 
400 #endif /* TCP_REASS_COUNTERS */
401 
402 static int tcp_reass(struct tcpcb *, const struct tcphdr *, struct mbuf *,
403     int);
404 
405 static void tcp4_log_refused(const struct ip *, const struct tcphdr *);
406 #ifdef INET6
407 static void tcp6_log_refused(const struct ip6_hdr *, const struct tcphdr *);
408 #endif
409 
410 #if defined(MBUFTRACE)
411 struct mowner tcp_reass_mowner = MOWNER_INIT("tcp", "reass");
412 #endif /* defined(MBUFTRACE) */
413 
414 static struct pool tcpipqent_pool;
415 
416 void
417 tcpipqent_init(void)
418 {
419 
420 	pool_init(&tcpipqent_pool, sizeof(struct ipqent), 0, 0, 0, "tcpipqepl",
421 	    NULL, IPL_VM);
422 }
423 
424 struct ipqent *
425 tcpipqent_alloc(void)
426 {
427 	struct ipqent *ipqe;
428 	int s;
429 
430 	s = splvm();
431 	ipqe = pool_get(&tcpipqent_pool, PR_NOWAIT);
432 	splx(s);
433 
434 	return ipqe;
435 }
436 
437 void
438 tcpipqent_free(struct ipqent *ipqe)
439 {
440 	int s;
441 
442 	s = splvm();
443 	pool_put(&tcpipqent_pool, ipqe);
444 	splx(s);
445 }
446 
447 /*
448  * Insert segment ti into reassembly queue of tcp with
449  * control block tp.  Return TH_FIN if reassembly now includes
450  * a segment with FIN.
451  */
452 static int
453 tcp_reass(struct tcpcb *tp, const struct tcphdr *th, struct mbuf *m, int tlen)
454 {
455 	struct ipqent *p, *q, *nq, *tiqe = NULL;
456 	struct socket *so = NULL;
457 	int pkt_flags;
458 	tcp_seq pkt_seq;
459 	unsigned pkt_len;
460 	u_long rcvpartdupbyte = 0;
461 	u_long rcvoobyte;
462 #ifdef TCP_REASS_COUNTERS
463 	u_int count = 0;
464 #endif
465 	uint64_t *tcps;
466 
467 	so = tp->t_inpcb->inp_socket;
468 
469 	TCP_REASS_LOCK_CHECK(tp);
470 
471 	/*
472 	 * Call with th==NULL after become established to
473 	 * force pre-ESTABLISHED data up to user socket.
474 	 */
475 	if (th == NULL)
476 		goto present;
477 
478 	m_claimm(m, &tcp_reass_mowner);
479 
480 	rcvoobyte = tlen;
481 	/*
482 	 * Copy these to local variables because the TCP header gets munged
483 	 * while we are collapsing mbufs.
484 	 */
485 	pkt_seq = th->th_seq;
486 	pkt_len = tlen;
487 	pkt_flags = th->th_flags;
488 
489 	TCP_REASS_COUNTER_INCR(&tcp_reass_);
490 
491 	if ((p = TAILQ_LAST(&tp->segq, ipqehead)) != NULL) {
492 		/*
493 		 * When we miss a packet, the vast majority of time we get
494 		 * packets that follow it in order.  So optimize for that.
495 		 */
496 		if (pkt_seq == p->ipqe_seq + p->ipqe_len) {
497 			p->ipqe_len += pkt_len;
498 			p->ipqe_flags |= pkt_flags;
499 			m_cat(p->ipqe_m, m);
500 			m = NULL;
501 			tiqe = p;
502 			TAILQ_REMOVE(&tp->timeq, p, ipqe_timeq);
503 			TCP_REASS_COUNTER_INCR(&tcp_reass_appendtail);
504 			goto skip_replacement;
505 		}
506 		/*
507 		 * While we're here, if the pkt is completely beyond
508 		 * anything we have, just insert it at the tail.
509 		 */
510 		if (SEQ_GT(pkt_seq, p->ipqe_seq + p->ipqe_len)) {
511 			TCP_REASS_COUNTER_INCR(&tcp_reass_inserttail);
512 			goto insert_it;
513 		}
514 	}
515 
516 	q = TAILQ_FIRST(&tp->segq);
517 
518 	if (q != NULL) {
519 		/*
520 		 * If this segment immediately precedes the first out-of-order
521 		 * block, simply slap the segment in front of it and (mostly)
522 		 * skip the complicated logic.
523 		 */
524 		if (pkt_seq + pkt_len == q->ipqe_seq) {
525 			q->ipqe_seq = pkt_seq;
526 			q->ipqe_len += pkt_len;
527 			q->ipqe_flags |= pkt_flags;
528 			m_cat(m, q->ipqe_m);
529 			q->ipqe_m = m;
530 			tiqe = q;
531 			TAILQ_REMOVE(&tp->timeq, q, ipqe_timeq);
532 			TCP_REASS_COUNTER_INCR(&tcp_reass_prependfirst);
533 			goto skip_replacement;
534 		}
535 	} else {
536 		TCP_REASS_COUNTER_INCR(&tcp_reass_empty);
537 	}
538 
539 	/*
540 	 * Find a segment which begins after this one does.
541 	 */
542 	for (p = NULL; q != NULL; q = nq) {
543 		nq = TAILQ_NEXT(q, ipqe_q);
544 #ifdef TCP_REASS_COUNTERS
545 		count++;
546 #endif
547 
548 		/*
549 		 * If the received segment is just right after this
550 		 * fragment, merge the two together and then check
551 		 * for further overlaps.
552 		 */
553 		if (q->ipqe_seq + q->ipqe_len == pkt_seq) {
554 			pkt_len += q->ipqe_len;
555 			pkt_flags |= q->ipqe_flags;
556 			pkt_seq = q->ipqe_seq;
557 			m_cat(q->ipqe_m, m);
558 			m = q->ipqe_m;
559 			TCP_REASS_COUNTER_INCR(&tcp_reass_append);
560 			goto free_ipqe;
561 		}
562 
563 		/*
564 		 * If the received segment is completely past this
565 		 * fragment, we need to go to the next fragment.
566 		 */
567 		if (SEQ_LT(q->ipqe_seq + q->ipqe_len, pkt_seq)) {
568 			p = q;
569 			continue;
570 		}
571 
572 		/*
573 		 * If the fragment is past the received segment,
574 		 * it (or any following) can't be concatenated.
575 		 */
576 		if (SEQ_GT(q->ipqe_seq, pkt_seq + pkt_len)) {
577 			TCP_REASS_COUNTER_INCR(&tcp_reass_insert);
578 			break;
579 		}
580 
581 		/*
582 		 * We've received all the data in this segment before.
583 		 * Mark it as a duplicate and return.
584 		 */
585 		if (SEQ_LEQ(q->ipqe_seq, pkt_seq) &&
586 		    SEQ_GEQ(q->ipqe_seq + q->ipqe_len, pkt_seq + pkt_len)) {
587 			tcps = TCP_STAT_GETREF();
588 			tcps[TCP_STAT_RCVDUPPACK]++;
589 			tcps[TCP_STAT_RCVDUPBYTE] += pkt_len;
590 			TCP_STAT_PUTREF();
591 			tcp_new_dsack(tp, pkt_seq, pkt_len);
592 			m_freem(m);
593 			if (tiqe != NULL) {
594 				tcpipqent_free(tiqe);
595 			}
596 			TCP_REASS_COUNTER_INCR(&tcp_reass_segdup);
597 			goto out;
598 		}
599 
600 		/*
601 		 * Received segment completely overlaps this fragment
602 		 * so we drop the fragment (this keeps the temporal
603 		 * ordering of segments correct).
604 		 */
605 		if (SEQ_GEQ(q->ipqe_seq, pkt_seq) &&
606 		    SEQ_LEQ(q->ipqe_seq + q->ipqe_len, pkt_seq + pkt_len)) {
607 			rcvpartdupbyte += q->ipqe_len;
608 			m_freem(q->ipqe_m);
609 			TCP_REASS_COUNTER_INCR(&tcp_reass_fragdup);
610 			goto free_ipqe;
611 		}
612 
613 		/*
614 		 * Received segment extends past the end of the fragment.
615 		 * Drop the overlapping bytes, merge the fragment and
616 		 * segment, and treat as a longer received packet.
617 		 */
618 		if (SEQ_LT(q->ipqe_seq, pkt_seq) &&
619 		    SEQ_GT(q->ipqe_seq + q->ipqe_len, pkt_seq))  {
620 			int overlap = q->ipqe_seq + q->ipqe_len - pkt_seq;
621 			m_adj(m, overlap);
622 			rcvpartdupbyte += overlap;
623 			m_cat(q->ipqe_m, m);
624 			m = q->ipqe_m;
625 			pkt_seq = q->ipqe_seq;
626 			pkt_len += q->ipqe_len - overlap;
627 			rcvoobyte -= overlap;
628 			TCP_REASS_COUNTER_INCR(&tcp_reass_overlaptail);
629 			goto free_ipqe;
630 		}
631 
632 		/*
633 		 * Received segment extends past the front of the fragment.
634 		 * Drop the overlapping bytes on the received packet. The
635 		 * packet will then be concatenated with this fragment a
636 		 * bit later.
637 		 */
638 		if (SEQ_GT(q->ipqe_seq, pkt_seq) &&
639 		    SEQ_LT(q->ipqe_seq, pkt_seq + pkt_len))  {
640 			int overlap = pkt_seq + pkt_len - q->ipqe_seq;
641 			m_adj(m, -overlap);
642 			pkt_len -= overlap;
643 			rcvpartdupbyte += overlap;
644 			TCP_REASS_COUNTER_INCR(&tcp_reass_overlapfront);
645 			rcvoobyte -= overlap;
646 		}
647 
648 		/*
649 		 * If the received segment immediately precedes this
650 		 * fragment then tack the fragment onto this segment
651 		 * and reinsert the data.
652 		 */
653 		if (q->ipqe_seq == pkt_seq + pkt_len) {
654 			pkt_len += q->ipqe_len;
655 			pkt_flags |= q->ipqe_flags;
656 			m_cat(m, q->ipqe_m);
657 			TAILQ_REMOVE(&tp->segq, q, ipqe_q);
658 			TAILQ_REMOVE(&tp->timeq, q, ipqe_timeq);
659 			tp->t_segqlen--;
660 			KASSERT(tp->t_segqlen >= 0);
661 			KASSERT(tp->t_segqlen != 0 ||
662 			    (TAILQ_EMPTY(&tp->segq) &&
663 			    TAILQ_EMPTY(&tp->timeq)));
664 			if (tiqe == NULL) {
665 				tiqe = q;
666 			} else {
667 				tcpipqent_free(q);
668 			}
669 			TCP_REASS_COUNTER_INCR(&tcp_reass_prepend);
670 			break;
671 		}
672 
673 		/*
674 		 * If the fragment is before the segment, remember it.
675 		 * When this loop is terminated, p will contain the
676 		 * pointer to the fragment that is right before the
677 		 * received segment.
678 		 */
679 		if (SEQ_LEQ(q->ipqe_seq, pkt_seq))
680 			p = q;
681 
682 		continue;
683 
684 		/*
685 		 * This is a common operation.  It also will allow
686 		 * to save doing a malloc/free in most instances.
687 		 */
688 	  free_ipqe:
689 		TAILQ_REMOVE(&tp->segq, q, ipqe_q);
690 		TAILQ_REMOVE(&tp->timeq, q, ipqe_timeq);
691 		tp->t_segqlen--;
692 		KASSERT(tp->t_segqlen >= 0);
693 		KASSERT(tp->t_segqlen != 0 ||
694 		    (TAILQ_EMPTY(&tp->segq) && TAILQ_EMPTY(&tp->timeq)));
695 		if (tiqe == NULL) {
696 			tiqe = q;
697 		} else {
698 			tcpipqent_free(q);
699 		}
700 	}
701 
702 #ifdef TCP_REASS_COUNTERS
703 	if (count > 7)
704 		TCP_REASS_COUNTER_INCR(&tcp_reass_iteration[0]);
705 	else if (count > 0)
706 		TCP_REASS_COUNTER_INCR(&tcp_reass_iteration[count]);
707 #endif
708 
709 insert_it:
710 	/*
711 	 * Allocate a new queue entry (block) since the received segment
712 	 * did not collapse onto any other out-of-order block. If it had
713 	 * collapsed, tiqe would not be NULL and we would be reusing it.
714 	 *
715 	 * If the allocation fails, drop the packet.
716 	 */
717 	if (tiqe == NULL) {
718 		tiqe = tcpipqent_alloc();
719 		if (tiqe == NULL) {
720 			TCP_STATINC(TCP_STAT_RCVMEMDROP);
721 			m_freem(m);
722 			goto out;
723 		}
724 	}
725 
726 	/*
727 	 * Update the counters.
728 	 */
729 	tp->t_rcvoopack++;
730 	tcps = TCP_STAT_GETREF();
731 	tcps[TCP_STAT_RCVOOPACK]++;
732 	tcps[TCP_STAT_RCVOOBYTE] += rcvoobyte;
733 	if (rcvpartdupbyte) {
734 	    tcps[TCP_STAT_RCVPARTDUPPACK]++;
735 	    tcps[TCP_STAT_RCVPARTDUPBYTE] += rcvpartdupbyte;
736 	}
737 	TCP_STAT_PUTREF();
738 
739 	/*
740 	 * Insert the new fragment queue entry into both queues.
741 	 */
742 	tiqe->ipqe_m = m;
743 	tiqe->ipqe_seq = pkt_seq;
744 	tiqe->ipqe_len = pkt_len;
745 	tiqe->ipqe_flags = pkt_flags;
746 	if (p == NULL) {
747 		TAILQ_INSERT_HEAD(&tp->segq, tiqe, ipqe_q);
748 	} else {
749 		TAILQ_INSERT_AFTER(&tp->segq, p, tiqe, ipqe_q);
750 	}
751 	tp->t_segqlen++;
752 
753 skip_replacement:
754 	TAILQ_INSERT_HEAD(&tp->timeq, tiqe, ipqe_timeq);
755 
756 present:
757 	/*
758 	 * Present data to user, advancing rcv_nxt through
759 	 * completed sequence space.
760 	 */
761 	if (TCPS_HAVEESTABLISHED(tp->t_state) == 0)
762 		goto out;
763 	q = TAILQ_FIRST(&tp->segq);
764 	if (q == NULL || q->ipqe_seq != tp->rcv_nxt)
765 		goto out;
766 	if (tp->t_state == TCPS_SYN_RECEIVED && q->ipqe_len)
767 		goto out;
768 
769 	tp->rcv_nxt += q->ipqe_len;
770 	pkt_flags = q->ipqe_flags & TH_FIN;
771 	nd_hint(tp);
772 
773 	TAILQ_REMOVE(&tp->segq, q, ipqe_q);
774 	TAILQ_REMOVE(&tp->timeq, q, ipqe_timeq);
775 	tp->t_segqlen--;
776 	KASSERT(tp->t_segqlen >= 0);
777 	KASSERT(tp->t_segqlen != 0 ||
778 	    (TAILQ_EMPTY(&tp->segq) && TAILQ_EMPTY(&tp->timeq)));
779 	if (so->so_state & SS_CANTRCVMORE)
780 		m_freem(q->ipqe_m);
781 	else
782 		sbappendstream(&so->so_rcv, q->ipqe_m);
783 	tcpipqent_free(q);
784 	TCP_REASS_UNLOCK(tp);
785 	sorwakeup(so);
786 	return pkt_flags;
787 
788 out:
789 	TCP_REASS_UNLOCK(tp);
790 	return 0;
791 }
792 
793 #ifdef INET6
794 int
795 tcp6_input(struct mbuf **mp, int *offp, int proto)
796 {
797 	struct mbuf *m = *mp;
798 
799 	/*
800 	 * draft-itojun-ipv6-tcp-to-anycast
801 	 * better place to put this in?
802 	 */
803 	if (m->m_flags & M_ANYCAST6) {
804 		struct ip6_hdr *ip6;
805 		if (m->m_len < sizeof(struct ip6_hdr)) {
806 			if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
807 				TCP_STATINC(TCP_STAT_RCVSHORT);
808 				return IPPROTO_DONE;
809 			}
810 		}
811 		ip6 = mtod(m, struct ip6_hdr *);
812 		icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_ADDR,
813 		    (char *)&ip6->ip6_dst - (char *)ip6);
814 		return IPPROTO_DONE;
815 	}
816 
817 	tcp_input(m, *offp, proto);
818 	return IPPROTO_DONE;
819 }
820 #endif
821 
822 static void
823 tcp4_log_refused(const struct ip *ip, const struct tcphdr *th)
824 {
825 	char src[INET_ADDRSTRLEN];
826 	char dst[INET_ADDRSTRLEN];
827 
828 	if (ip) {
829 		in_print(src, sizeof(src), &ip->ip_src);
830 		in_print(dst, sizeof(dst), &ip->ip_dst);
831 	} else {
832 		strlcpy(src, "(unknown)", sizeof(src));
833 		strlcpy(dst, "(unknown)", sizeof(dst));
834 	}
835 	log(LOG_INFO,
836 	    "Connection attempt to TCP %s:%d from %s:%d\n",
837 	    dst, ntohs(th->th_dport),
838 	    src, ntohs(th->th_sport));
839 }
840 
841 #ifdef INET6
842 static void
843 tcp6_log_refused(const struct ip6_hdr *ip6, const struct tcphdr *th)
844 {
845 	char src[INET6_ADDRSTRLEN];
846 	char dst[INET6_ADDRSTRLEN];
847 
848 	if (ip6) {
849 		in6_print(src, sizeof(src), &ip6->ip6_src);
850 		in6_print(dst, sizeof(dst), &ip6->ip6_dst);
851 	} else {
852 		strlcpy(src, "(unknown v6)", sizeof(src));
853 		strlcpy(dst, "(unknown v6)", sizeof(dst));
854 	}
855 	log(LOG_INFO,
856 	    "Connection attempt to TCP [%s]:%d from [%s]:%d\n",
857 	    dst, ntohs(th->th_dport),
858 	    src, ntohs(th->th_sport));
859 }
860 #endif
861 
862 /*
863  * Checksum extended TCP header and data.
864  */
865 int
866 tcp_input_checksum(int af, struct mbuf *m, const struct tcphdr *th,
867     int toff, int off, int tlen)
868 {
869 	struct ifnet *rcvif;
870 	int s;
871 
872 	/*
873 	 * XXX it's better to record and check if this mbuf is
874 	 * already checked.
875 	 */
876 
877 	rcvif = m_get_rcvif(m, &s);
878 	if (__predict_false(rcvif == NULL))
879 		goto badcsum; /* XXX */
880 
881 	switch (af) {
882 	case AF_INET:
883 		switch (m->m_pkthdr.csum_flags &
884 			((rcvif->if_csum_flags_rx & M_CSUM_TCPv4) |
885 			 M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) {
886 		case M_CSUM_TCPv4|M_CSUM_TCP_UDP_BAD:
887 			TCP_CSUM_COUNTER_INCR(&tcp_hwcsum_bad);
888 			goto badcsum;
889 
890 		case M_CSUM_TCPv4|M_CSUM_DATA: {
891 			u_int32_t hw_csum = m->m_pkthdr.csum_data;
892 
893 			TCP_CSUM_COUNTER_INCR(&tcp_hwcsum_data);
894 			if (m->m_pkthdr.csum_flags & M_CSUM_NO_PSEUDOHDR) {
895 				const struct ip *ip =
896 				    mtod(m, const struct ip *);
897 
898 				hw_csum = in_cksum_phdr(ip->ip_src.s_addr,
899 				    ip->ip_dst.s_addr,
900 				    htons(hw_csum + tlen + off + IPPROTO_TCP));
901 			}
902 			if ((hw_csum ^ 0xffff) != 0)
903 				goto badcsum;
904 			break;
905 		}
906 
907 		case M_CSUM_TCPv4:
908 			/* Checksum was okay. */
909 			TCP_CSUM_COUNTER_INCR(&tcp_hwcsum_ok);
910 			break;
911 
912 		default:
913 			/*
914 			 * Must compute it ourselves.  Maybe skip checksum
915 			 * on loopback interfaces.
916 			 */
917 			if (__predict_true(!(rcvif->if_flags & IFF_LOOPBACK) ||
918 					   tcp_do_loopback_cksum)) {
919 				TCP_CSUM_COUNTER_INCR(&tcp_swcsum);
920 				if (in4_cksum(m, IPPROTO_TCP, toff,
921 					      tlen + off) != 0)
922 					goto badcsum;
923 			}
924 			break;
925 		}
926 		break;
927 
928 #ifdef INET6
929 	case AF_INET6:
930 		switch (m->m_pkthdr.csum_flags &
931 			((rcvif->if_csum_flags_rx & M_CSUM_TCPv6) |
932 			 M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) {
933 		case M_CSUM_TCPv6|M_CSUM_TCP_UDP_BAD:
934 			TCP_CSUM_COUNTER_INCR(&tcp6_hwcsum_bad);
935 			goto badcsum;
936 
937 #if 0 /* notyet */
938 		case M_CSUM_TCPv6|M_CSUM_DATA:
939 #endif
940 
941 		case M_CSUM_TCPv6:
942 			/* Checksum was okay. */
943 			TCP_CSUM_COUNTER_INCR(&tcp6_hwcsum_ok);
944 			break;
945 
946 		default:
947 			/*
948 			 * Must compute it ourselves.  Maybe skip checksum
949 			 * on loopback interfaces.
950 			 */
951 			if (__predict_true((m->m_flags & M_LOOP) == 0 ||
952 			    tcp_do_loopback_cksum)) {
953 				TCP_CSUM_COUNTER_INCR(&tcp6_swcsum);
954 				if (in6_cksum(m, IPPROTO_TCP, toff,
955 				    tlen + off) != 0)
956 					goto badcsum;
957 			}
958 		}
959 		break;
960 #endif /* INET6 */
961 	}
962 	m_put_rcvif(rcvif, &s);
963 
964 	return 0;
965 
966 badcsum:
967 	m_put_rcvif(rcvif, &s);
968 	TCP_STATINC(TCP_STAT_RCVBADSUM);
969 	return -1;
970 }
971 
972 /*
973  * When a packet arrives addressed to a vestigial tcpbp, we
974  * nevertheless have to respond to it per the spec.
975  *
976  * This code is duplicated from the one in tcp_input().
977  */
978 static void tcp_vtw_input(struct tcphdr *th, vestigial_inpcb_t *vp,
979     struct mbuf *m, int tlen)
980 {
981 	int tiflags;
982 	int todrop;
983 	uint32_t t_flags = 0;
984 	uint64_t *tcps;
985 
986 	tiflags = th->th_flags;
987 	todrop  = vp->rcv_nxt - th->th_seq;
988 
989 	if (todrop > 0) {
990 		if (tiflags & TH_SYN) {
991 			tiflags &= ~TH_SYN;
992 			th->th_seq++;
993 			tcp_urp_drop(th, 1, &tiflags);
994 			todrop--;
995 		}
996 		if (todrop > tlen ||
997 		    (todrop == tlen && (tiflags & TH_FIN) == 0)) {
998 			/*
999 			 * Any valid FIN or RST must be to the left of the
1000 			 * window.  At this point the FIN or RST must be a
1001 			 * duplicate or out of sequence; drop it.
1002 			 */
1003 			if (tiflags & TH_RST)
1004 				goto drop;
1005 			tiflags &= ~(TH_FIN|TH_RST);
1006 
1007 			/*
1008 			 * Send an ACK to resynchronize and drop any data.
1009 			 * But keep on processing for RST or ACK.
1010 			 */
1011 			t_flags |= TF_ACKNOW;
1012 			todrop = tlen;
1013 			tcps = TCP_STAT_GETREF();
1014 			tcps[TCP_STAT_RCVDUPPACK] += 1;
1015 			tcps[TCP_STAT_RCVDUPBYTE] += todrop;
1016 			TCP_STAT_PUTREF();
1017 		} else if ((tiflags & TH_RST) &&
1018 		    th->th_seq != vp->rcv_nxt) {
1019 			/*
1020 			 * Test for reset before adjusting the sequence
1021 			 * number for overlapping data.
1022 			 */
1023 			goto dropafterack_ratelim;
1024 		} else {
1025 			tcps = TCP_STAT_GETREF();
1026 			tcps[TCP_STAT_RCVPARTDUPPACK] += 1;
1027 			tcps[TCP_STAT_RCVPARTDUPBYTE] += todrop;
1028 			TCP_STAT_PUTREF();
1029 		}
1030 
1031 //		tcp_new_dsack(tp, th->th_seq, todrop);
1032 //		hdroptlen += todrop;	/*drop from head afterwards*/
1033 
1034 		th->th_seq += todrop;
1035 		tlen -= todrop;
1036 		tcp_urp_drop(th, todrop, &tiflags);
1037 	}
1038 
1039 	/*
1040 	 * If new data are received on a connection after the
1041 	 * user processes are gone, then RST the other end.
1042 	 */
1043 	if (tlen) {
1044 		TCP_STATINC(TCP_STAT_RCVAFTERCLOSE);
1045 		goto dropwithreset;
1046 	}
1047 
1048 	/*
1049 	 * If segment ends after window, drop trailing data
1050 	 * (and PUSH and FIN); if nothing left, just ACK.
1051 	 */
1052 	todrop = (th->th_seq + tlen) - (vp->rcv_nxt + vp->rcv_wnd);
1053 
1054 	if (todrop > 0) {
1055 		TCP_STATINC(TCP_STAT_RCVPACKAFTERWIN);
1056 		if (todrop >= tlen) {
1057 			/*
1058 			 * The segment actually starts after the window.
1059 			 * th->th_seq + tlen - vp->rcv_nxt - vp->rcv_wnd >= tlen
1060 			 * th->th_seq - vp->rcv_nxt - vp->rcv_wnd >= 0
1061 			 * th->th_seq >= vp->rcv_nxt + vp->rcv_wnd
1062 			 */
1063 			TCP_STATADD(TCP_STAT_RCVBYTEAFTERWIN, tlen);
1064 
1065 			/*
1066 			 * If a new connection request is received
1067 			 * while in TIME_WAIT, drop the old connection
1068 			 * and start over if the sequence numbers
1069 			 * are above the previous ones.
1070 			 */
1071 			if ((tiflags & TH_SYN) &&
1072 			    SEQ_GT(th->th_seq, vp->rcv_nxt)) {
1073 				/*
1074 				 * We only support this in the !NOFDREF case, which
1075 				 * is to say: not here.
1076 				 */
1077 				goto dropwithreset;
1078 			}
1079 
1080 			/*
1081 			 * If window is closed can only take segments at
1082 			 * window edge, and have to drop data and PUSH from
1083 			 * incoming segments.  Continue processing, but
1084 			 * remember to ack.  Otherwise, drop segment
1085 			 * and (if not RST) ack.
1086 			 */
1087 			if (vp->rcv_wnd == 0 && th->th_seq == vp->rcv_nxt) {
1088 				t_flags |= TF_ACKNOW;
1089 				TCP_STATINC(TCP_STAT_RCVWINPROBE);
1090 			} else {
1091 				goto dropafterack;
1092 			}
1093 		} else {
1094 			TCP_STATADD(TCP_STAT_RCVBYTEAFTERWIN, todrop);
1095 		}
1096 		m_adj(m, -todrop);
1097 		tlen -= todrop;
1098 		tiflags &= ~(TH_PUSH|TH_FIN);
1099 	}
1100 
1101 	if (tiflags & TH_RST) {
1102 		if (th->th_seq != vp->rcv_nxt)
1103 			goto dropafterack_ratelim;
1104 
1105 		vtw_del(vp->ctl, vp->vtw);
1106 		goto drop;
1107 	}
1108 
1109 	/*
1110 	 * If the ACK bit is off we drop the segment and return.
1111 	 */
1112 	if ((tiflags & TH_ACK) == 0) {
1113 		if (t_flags & TF_ACKNOW)
1114 			goto dropafterack;
1115 		goto drop;
1116 	}
1117 
1118 	/*
1119 	 * In TIME_WAIT state the only thing that should arrive
1120 	 * is a retransmission of the remote FIN.  Acknowledge
1121 	 * it and restart the finack timer.
1122 	 */
1123 	vtw_restart(vp);
1124 	goto dropafterack;
1125 
1126 dropafterack:
1127 	/*
1128 	 * Generate an ACK dropping incoming segment if it occupies
1129 	 * sequence space, where the ACK reflects our state.
1130 	 */
1131 	if (tiflags & TH_RST)
1132 		goto drop;
1133 	goto dropafterack2;
1134 
1135 dropafterack_ratelim:
1136 	/*
1137 	 * We may want to rate-limit ACKs against SYN/RST attack.
1138 	 */
1139 	if (ppsratecheck(&tcp_ackdrop_ppslim_last, &tcp_ackdrop_ppslim_count,
1140 	    tcp_ackdrop_ppslim) == 0) {
1141 		/* XXX stat */
1142 		goto drop;
1143 	}
1144 	/* ...fall into dropafterack2... */
1145 
1146 dropafterack2:
1147 	(void)tcp_respond(0, m, m, th, th->th_seq + tlen, th->th_ack, TH_ACK);
1148 	return;
1149 
1150 dropwithreset:
1151 	/*
1152 	 * Generate a RST, dropping incoming segment.
1153 	 * Make ACK acceptable to originator of segment.
1154 	 */
1155 	if (tiflags & TH_RST)
1156 		goto drop;
1157 
1158 	if (tiflags & TH_ACK) {
1159 		tcp_respond(0, m, m, th, (tcp_seq)0, th->th_ack, TH_RST);
1160 	} else {
1161 		if (tiflags & TH_SYN)
1162 			++tlen;
1163 		(void)tcp_respond(0, m, m, th, th->th_seq + tlen, (tcp_seq)0,
1164 		    TH_RST|TH_ACK);
1165 	}
1166 	return;
1167 drop:
1168 	m_freem(m);
1169 }
1170 
1171 /*
1172  * TCP input routine, follows pages 65-76 of RFC 793 very closely.
1173  */
1174 void
1175 tcp_input(struct mbuf *m, int off, int proto)
1176 {
1177 	struct tcphdr *th;
1178 	struct ip *ip;
1179 	struct inpcb *inp;
1180 #ifdef INET6
1181 	struct ip6_hdr *ip6;
1182 #endif
1183 	u_int8_t *optp = NULL;
1184 	int optlen = 0;
1185 	int len, tlen, hdroptlen = 0;
1186 	struct tcpcb *tp = NULL;
1187 	int tiflags;
1188 	struct socket *so = NULL;
1189 	int todrop, acked, ourfinisacked, needoutput = 0;
1190 	bool dupseg;
1191 #ifdef TCP_DEBUG
1192 	short ostate = 0;
1193 #endif
1194 	u_long tiwin;
1195 	struct tcp_opt_info opti;
1196 	int thlen, iphlen;
1197 	int af;		/* af on the wire */
1198 	struct mbuf *tcp_saveti = NULL;
1199 	uint32_t ts_rtt;
1200 	uint8_t iptos;
1201 	uint64_t *tcps;
1202 	vestigial_inpcb_t vestige;
1203 
1204 	vestige.valid = 0;
1205 
1206 	MCLAIM(m, &tcp_rx_mowner);
1207 
1208 	TCP_STATINC(TCP_STAT_RCVTOTAL);
1209 
1210 	memset(&opti, 0, sizeof(opti));
1211 	opti.ts_present = 0;
1212 	opti.maxseg = 0;
1213 
1214 	/*
1215 	 * RFC1122 4.2.3.10, p. 104: discard bcast/mcast SYN.
1216 	 *
1217 	 * TCP is, by definition, unicast, so we reject all
1218 	 * multicast outright.
1219 	 *
1220 	 * Note, there are additional src/dst address checks in
1221 	 * the AF-specific code below.
1222 	 */
1223 	if (m->m_flags & (M_BCAST|M_MCAST)) {
1224 		/* XXX stat */
1225 		goto drop;
1226 	}
1227 #ifdef INET6
1228 	if (m->m_flags & M_ANYCAST6) {
1229 		/* XXX stat */
1230 		goto drop;
1231 	}
1232 #endif
1233 
1234 	M_REGION_GET(th, struct tcphdr *, m, off, sizeof(struct tcphdr));
1235 	if (th == NULL) {
1236 		TCP_STATINC(TCP_STAT_RCVSHORT);
1237 		return;
1238 	}
1239 
1240 	/*
1241 	 * Enforce alignment requirements that are violated in
1242 	 * some cases, see kern/50766 for details.
1243 	 */
1244 	if (ACCESSIBLE_POINTER(th, struct tcphdr) == 0) {
1245 		m = m_copyup(m, off + sizeof(struct tcphdr), 0);
1246 		if (m == NULL) {
1247 			TCP_STATINC(TCP_STAT_RCVSHORT);
1248 			return;
1249 		}
1250 		th = (struct tcphdr *)(mtod(m, char *) + off);
1251 	}
1252 	KASSERT(ACCESSIBLE_POINTER(th, struct tcphdr));
1253 
1254 	/*
1255 	 * Get IP and TCP header.
1256 	 * Note: IP leaves IP header in first mbuf.
1257 	 */
1258 	ip = mtod(m, struct ip *);
1259 #ifdef INET6
1260 	ip6 = mtod(m, struct ip6_hdr *);
1261 #endif
1262 	switch (ip->ip_v) {
1263 	case 4:
1264 		af = AF_INET;
1265 		iphlen = sizeof(struct ip);
1266 
1267 		if (IN_MULTICAST(ip->ip_dst.s_addr) ||
1268 		    in_broadcast(ip->ip_dst, m_get_rcvif_NOMPSAFE(m)))
1269 			goto drop;
1270 
1271 		/* We do the checksum after PCB lookup... */
1272 		len = ntohs(ip->ip_len);
1273 		tlen = len - off;
1274 		iptos = ip->ip_tos;
1275 		break;
1276 #ifdef INET6
1277 	case 6:
1278 		iphlen = sizeof(struct ip6_hdr);
1279 		af = AF_INET6;
1280 
1281 		/*
1282 		 * Be proactive about unspecified IPv6 address in source.
1283 		 * As we use all-zero to indicate unbounded/unconnected pcb,
1284 		 * unspecified IPv6 address can be used to confuse us.
1285 		 *
1286 		 * Note that packets with unspecified IPv6 destination is
1287 		 * already dropped in ip6_input.
1288 		 */
1289 		if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) {
1290 			/* XXX stat */
1291 			goto drop;
1292 		}
1293 
1294 		/*
1295 		 * Make sure destination address is not multicast.
1296 		 * Source address checked in ip6_input().
1297 		 */
1298 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1299 			/* XXX stat */
1300 			goto drop;
1301 		}
1302 
1303 		/* We do the checksum after PCB lookup... */
1304 		len = m->m_pkthdr.len;
1305 		tlen = len - off;
1306 		iptos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
1307 		break;
1308 #endif
1309 	default:
1310 		m_freem(m);
1311 		return;
1312 	}
1313 
1314 
1315 	/*
1316 	 * Check that TCP offset makes sense, pull out TCP options and
1317 	 * adjust length.
1318 	 */
1319 	thlen = th->th_off << 2;
1320 	if (thlen < sizeof(struct tcphdr) || thlen > tlen) {
1321 		TCP_STATINC(TCP_STAT_RCVBADOFF);
1322 		goto drop;
1323 	}
1324 	tlen -= thlen;
1325 
1326 	if (thlen > sizeof(struct tcphdr)) {
1327 		M_REGION_GET(th, struct tcphdr *, m, off, thlen);
1328 		if (th == NULL) {
1329 			TCP_STATINC(TCP_STAT_RCVSHORT);
1330 			return;
1331 		}
1332 		KASSERT(ACCESSIBLE_POINTER(th, struct tcphdr));
1333 		optlen = thlen - sizeof(struct tcphdr);
1334 		optp = ((u_int8_t *)th) + sizeof(struct tcphdr);
1335 
1336 		/*
1337 		 * Do quick retrieval of timestamp options.
1338 		 *
1339 		 * If timestamp is the only option and it's formatted as
1340 		 * recommended in RFC 1323 appendix A, we quickly get the
1341 		 * values now and don't bother calling tcp_dooptions(),
1342 		 * etc.
1343 		 */
1344 		if ((optlen == TCPOLEN_TSTAMP_APPA ||
1345 		     (optlen > TCPOLEN_TSTAMP_APPA &&
1346 		      optp[TCPOLEN_TSTAMP_APPA] == TCPOPT_EOL)) &&
1347 		    be32dec(optp) == TCPOPT_TSTAMP_HDR &&
1348 		    (th->th_flags & TH_SYN) == 0) {
1349 			opti.ts_present = 1;
1350 			opti.ts_val = be32dec(optp + 4);
1351 			opti.ts_ecr = be32dec(optp + 8);
1352 			optp = NULL;	/* we've parsed the options */
1353 		}
1354 	}
1355 	tiflags = th->th_flags;
1356 
1357 	/*
1358 	 * Checksum extended TCP header and data
1359 	 */
1360 	if (tcp_input_checksum(af, m, th, off, thlen, tlen))
1361 		goto badcsum;
1362 
1363 	/*
1364 	 * Locate pcb for segment.
1365 	 */
1366 findpcb:
1367 	inp = NULL;
1368 	switch (af) {
1369 	case AF_INET:
1370 		inp = inpcb_lookup(&tcbtable, ip->ip_src, th->th_sport,
1371 		    ip->ip_dst, th->th_dport, &vestige);
1372 		if (inp == NULL && !vestige.valid) {
1373 			TCP_STATINC(TCP_STAT_PCBHASHMISS);
1374 			inp = inpcb_lookup_bound(&tcbtable, ip->ip_dst,
1375 			    th->th_dport);
1376 		}
1377 #ifdef INET6
1378 		if (inp == NULL && !vestige.valid) {
1379 			struct in6_addr s, d;
1380 
1381 			/* mapped addr case */
1382 			in6_in_2_v4mapin6(&ip->ip_src, &s);
1383 			in6_in_2_v4mapin6(&ip->ip_dst, &d);
1384 			inp = in6pcb_lookup(&tcbtable, &s,
1385 			    th->th_sport, &d, th->th_dport, 0, &vestige);
1386 			if (inp == NULL && !vestige.valid) {
1387 				TCP_STATINC(TCP_STAT_PCBHASHMISS);
1388 				inp = in6pcb_lookup_bound(&tcbtable, &d,
1389 				    th->th_dport, 0);
1390 			}
1391 		}
1392 #endif
1393 		if (inp == NULL && !vestige.valid) {
1394 			TCP_STATINC(TCP_STAT_NOPORT);
1395 			if (tcp_log_refused &&
1396 			    (tiflags & (TH_RST|TH_ACK|TH_SYN)) == TH_SYN) {
1397 				tcp4_log_refused(ip, th);
1398 			}
1399 			tcp_fields_to_host(th);
1400 			goto dropwithreset_ratelim;
1401 		}
1402 #if defined(IPSEC)
1403 		if (ipsec_used) {
1404 			if (inp && ipsec_in_reject(m, inp))
1405 				goto drop;
1406 		}
1407 #endif /*IPSEC*/
1408 		break;
1409 #ifdef INET6
1410 	case AF_INET6:
1411 	    {
1412 		int faith;
1413 
1414 #if defined(NFAITH) && NFAITH > 0
1415 		faith = faithprefix(&ip6->ip6_dst);
1416 #else
1417 		faith = 0;
1418 #endif
1419 		inp = in6pcb_lookup(&tcbtable, &ip6->ip6_src,
1420 		    th->th_sport, &ip6->ip6_dst, th->th_dport, faith, &vestige);
1421 		if (inp == NULL && !vestige.valid) {
1422 			TCP_STATINC(TCP_STAT_PCBHASHMISS);
1423 			inp = in6pcb_lookup_bound(&tcbtable, &ip6->ip6_dst,
1424 			    th->th_dport, faith);
1425 		}
1426 		if (inp == NULL && !vestige.valid) {
1427 			TCP_STATINC(TCP_STAT_NOPORT);
1428 			if (tcp_log_refused &&
1429 			    (tiflags & (TH_RST|TH_ACK|TH_SYN)) == TH_SYN) {
1430 				tcp6_log_refused(ip6, th);
1431 			}
1432 			tcp_fields_to_host(th);
1433 			goto dropwithreset_ratelim;
1434 		}
1435 #if defined(IPSEC)
1436 		if (ipsec_used && inp && ipsec_in_reject(m, inp))
1437 			goto drop;
1438 #endif
1439 		break;
1440 	    }
1441 #endif
1442 	}
1443 
1444 	tcp_fields_to_host(th);
1445 
1446 	/*
1447 	 * If the state is CLOSED (i.e., TCB does not exist) then
1448 	 * all data in the incoming segment is discarded.
1449 	 * If the TCB exists but is in CLOSED state, it is embryonic,
1450 	 * but should either do a listen or a connect soon.
1451 	 */
1452 	tp = NULL;
1453 	so = NULL;
1454 	if (inp) {
1455 		/* Check the minimum TTL for socket. */
1456 		if (inp->inp_af == AF_INET && ip->ip_ttl < in4p_ip_minttl(inp))
1457 			goto drop;
1458 
1459 		tp = intotcpcb(inp);
1460 		so = inp->inp_socket;
1461 	} else if (vestige.valid) {
1462 		/* We do not support the resurrection of vtw tcpcps. */
1463 		tcp_vtw_input(th, &vestige, m, tlen);
1464 		m = NULL;
1465 		goto drop;
1466 	}
1467 
1468 	if (tp == NULL)
1469 		goto dropwithreset_ratelim;
1470 	if (tp->t_state == TCPS_CLOSED)
1471 		goto drop;
1472 
1473 	KASSERT(so->so_lock == softnet_lock);
1474 	KASSERT(solocked(so));
1475 
1476 	/* Unscale the window into a 32-bit value. */
1477 	if ((tiflags & TH_SYN) == 0)
1478 		tiwin = th->th_win << tp->snd_scale;
1479 	else
1480 		tiwin = th->th_win;
1481 
1482 #ifdef INET6
1483 	/* save packet options if user wanted */
1484 	if (inp->inp_af == AF_INET6 && (inp->inp_flags & IN6P_CONTROLOPTS)) {
1485 		if (inp->inp_options) {
1486 			m_freem(inp->inp_options);
1487 			inp->inp_options = NULL;
1488 		}
1489 		ip6_savecontrol(inp, &inp->inp_options, ip6, m);
1490 	}
1491 #endif
1492 
1493 	if (so->so_options & SO_DEBUG) {
1494 #ifdef TCP_DEBUG
1495 		ostate = tp->t_state;
1496 #endif
1497 
1498 		tcp_saveti = NULL;
1499 		if (iphlen + sizeof(struct tcphdr) > MHLEN)
1500 			goto nosave;
1501 
1502 		if (m->m_len > iphlen && (m->m_flags & M_EXT) == 0) {
1503 			tcp_saveti = m_copym(m, 0, iphlen, M_DONTWAIT);
1504 			if (tcp_saveti == NULL)
1505 				goto nosave;
1506 		} else {
1507 			MGETHDR(tcp_saveti, M_DONTWAIT, MT_HEADER);
1508 			if (tcp_saveti == NULL)
1509 				goto nosave;
1510 			MCLAIM(m, &tcp_mowner);
1511 			tcp_saveti->m_len = iphlen;
1512 			m_copydata(m, 0, iphlen,
1513 			    mtod(tcp_saveti, void *));
1514 		}
1515 
1516 		if (M_TRAILINGSPACE(tcp_saveti) < sizeof(struct tcphdr)) {
1517 			m_freem(tcp_saveti);
1518 			tcp_saveti = NULL;
1519 		} else {
1520 			tcp_saveti->m_len += sizeof(struct tcphdr);
1521 			memcpy(mtod(tcp_saveti, char *) + iphlen, th,
1522 			    sizeof(struct tcphdr));
1523 		}
1524 nosave:;
1525 	}
1526 
1527 	if (so->so_options & SO_ACCEPTCONN) {
1528 		union syn_cache_sa src;
1529 		union syn_cache_sa dst;
1530 
1531 		KASSERT(tp->t_state == TCPS_LISTEN);
1532 
1533 		memset(&src, 0, sizeof(src));
1534 		memset(&dst, 0, sizeof(dst));
1535 		switch (af) {
1536 		case AF_INET:
1537 			src.sin.sin_len = sizeof(struct sockaddr_in);
1538 			src.sin.sin_family = AF_INET;
1539 			src.sin.sin_addr = ip->ip_src;
1540 			src.sin.sin_port = th->th_sport;
1541 
1542 			dst.sin.sin_len = sizeof(struct sockaddr_in);
1543 			dst.sin.sin_family = AF_INET;
1544 			dst.sin.sin_addr = ip->ip_dst;
1545 			dst.sin.sin_port = th->th_dport;
1546 			break;
1547 #ifdef INET6
1548 		case AF_INET6:
1549 			src.sin6.sin6_len = sizeof(struct sockaddr_in6);
1550 			src.sin6.sin6_family = AF_INET6;
1551 			src.sin6.sin6_addr = ip6->ip6_src;
1552 			src.sin6.sin6_port = th->th_sport;
1553 
1554 			dst.sin6.sin6_len = sizeof(struct sockaddr_in6);
1555 			dst.sin6.sin6_family = AF_INET6;
1556 			dst.sin6.sin6_addr = ip6->ip6_dst;
1557 			dst.sin6.sin6_port = th->th_dport;
1558 			break;
1559 #endif
1560 		}
1561 
1562 		if ((tiflags & (TH_RST|TH_ACK|TH_SYN)) != TH_SYN) {
1563 			if (tiflags & TH_RST) {
1564 				syn_cache_reset(&src.sa, &dst.sa, th);
1565 			} else if ((tiflags & (TH_ACK|TH_SYN)) ==
1566 			    (TH_ACK|TH_SYN)) {
1567 				/*
1568 				 * Received a SYN,ACK. This should never
1569 				 * happen while we are in LISTEN. Send an RST.
1570 				 */
1571 				goto badsyn;
1572 			} else if (tiflags & TH_ACK) {
1573 				so = syn_cache_get(&src.sa, &dst.sa, th, so, m);
1574 				if (so == NULL) {
1575 					/*
1576 					 * We don't have a SYN for this ACK;
1577 					 * send an RST.
1578 					 */
1579 					goto badsyn;
1580 				} else if (so == (struct socket *)(-1)) {
1581 					/*
1582 					 * We were unable to create the
1583 					 * connection. If the 3-way handshake
1584 					 * was completed, and RST has been
1585 					 * sent to the peer. Since the mbuf
1586 					 * might be in use for the reply, do
1587 					 * not free it.
1588 					 */
1589 					m = NULL;
1590 				} else {
1591 					/*
1592 					 * We have created a full-blown
1593 					 * connection.
1594 					 */
1595 					inp = sotoinpcb(so);
1596 					tp = intotcpcb(inp);
1597 					if (tp == NULL)
1598 						goto badsyn;	/*XXX*/
1599 					tiwin <<= tp->snd_scale;
1600 					goto after_listen;
1601 				}
1602 			} else {
1603 				/*
1604 				 * None of RST, SYN or ACK was set.
1605 				 * This is an invalid packet for a
1606 				 * TCB in LISTEN state.  Send a RST.
1607 				 */
1608 				goto badsyn;
1609 			}
1610 		} else {
1611 			/*
1612 			 * Received a SYN.
1613 			 */
1614 
1615 #ifdef INET6
1616 			/*
1617 			 * If deprecated address is forbidden, we do
1618 			 * not accept SYN to deprecated interface
1619 			 * address to prevent any new inbound
1620 			 * connection from getting established.
1621 			 * When we do not accept SYN, we send a TCP
1622 			 * RST, with deprecated source address (instead
1623 			 * of dropping it).  We compromise it as it is
1624 			 * much better for peer to send a RST, and
1625 			 * RST will be the final packet for the
1626 			 * exchange.
1627 			 *
1628 			 * If we do not forbid deprecated addresses, we
1629 			 * accept the SYN packet.  RFC2462 does not
1630 			 * suggest dropping SYN in this case.
1631 			 * If we decipher RFC2462 5.5.4, it says like
1632 			 * this:
1633 			 * 1. use of deprecated addr with existing
1634 			 *    communication is okay - "SHOULD continue
1635 			 *    to be used"
1636 			 * 2. use of it with new communication:
1637 			 *   (2a) "SHOULD NOT be used if alternate
1638 			 *        address with sufficient scope is
1639 			 *        available"
1640 			 *   (2b) nothing mentioned otherwise.
1641 			 * Here we fall into (2b) case as we have no
1642 			 * choice in our source address selection - we
1643 			 * must obey the peer.
1644 			 *
1645 			 * The wording in RFC2462 is confusing, and
1646 			 * there are multiple description text for
1647 			 * deprecated address handling - worse, they
1648 			 * are not exactly the same.  I believe 5.5.4
1649 			 * is the best one, so we follow 5.5.4.
1650 			 */
1651 			if (af == AF_INET6 && !ip6_use_deprecated) {
1652 				struct in6_ifaddr *ia6;
1653 				int s;
1654 				struct ifnet *rcvif = m_get_rcvif(m, &s);
1655 				if (rcvif == NULL)
1656 					goto dropwithreset; /* XXX */
1657 				if ((ia6 = in6ifa_ifpwithaddr(rcvif,
1658 				    &ip6->ip6_dst)) &&
1659 				    (ia6->ia6_flags & IN6_IFF_DEPRECATED)) {
1660 					tp = NULL;
1661 					m_put_rcvif(rcvif, &s);
1662 					goto dropwithreset;
1663 				}
1664 				m_put_rcvif(rcvif, &s);
1665 			}
1666 #endif
1667 
1668 			/*
1669 			 * LISTEN socket received a SYN from itself? This
1670 			 * can't possibly be valid; drop the packet.
1671 			 */
1672 			if (th->th_sport == th->th_dport) {
1673 				int eq = 0;
1674 
1675 				switch (af) {
1676 				case AF_INET:
1677 					eq = in_hosteq(ip->ip_src, ip->ip_dst);
1678 					break;
1679 #ifdef INET6
1680 				case AF_INET6:
1681 					eq = IN6_ARE_ADDR_EQUAL(&ip6->ip6_src,
1682 					    &ip6->ip6_dst);
1683 					break;
1684 #endif
1685 				}
1686 				if (eq) {
1687 					TCP_STATINC(TCP_STAT_BADSYN);
1688 					goto drop;
1689 				}
1690 			}
1691 
1692 			/*
1693 			 * SYN looks ok; create compressed TCP
1694 			 * state for it.
1695 			 */
1696 			if (so->so_qlen <= so->so_qlimit &&
1697 			    syn_cache_add(&src.sa, &dst.sa, th, off,
1698 			    so, m, optp, optlen, &opti))
1699 				m = NULL;
1700 		}
1701 
1702 		goto drop;
1703 	}
1704 
1705 after_listen:
1706 	/*
1707 	 * From here on, we're dealing with !LISTEN.
1708 	 */
1709 	KASSERT(tp->t_state != TCPS_LISTEN);
1710 
1711 	/*
1712 	 * Segment received on connection.
1713 	 * Reset idle time and keep-alive timer.
1714 	 */
1715 	tp->t_rcvtime = tcp_now;
1716 	if (TCPS_HAVEESTABLISHED(tp->t_state))
1717 		TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepidle);
1718 
1719 	/*
1720 	 * Process options.
1721 	 */
1722 #ifdef TCP_SIGNATURE
1723 	if (optp || (tp->t_flags & TF_SIGNATURE))
1724 #else
1725 	if (optp)
1726 #endif
1727 		if (tcp_dooptions(tp, optp, optlen, th, m, off, &opti) < 0)
1728 			goto drop;
1729 
1730 	if (TCP_SACK_ENABLED(tp)) {
1731 		tcp_del_sackholes(tp, th);
1732 	}
1733 
1734 	if (TCP_ECN_ALLOWED(tp)) {
1735 		if (tiflags & TH_CWR) {
1736 			tp->t_flags &= ~TF_ECN_SND_ECE;
1737 		}
1738 		switch (iptos & IPTOS_ECN_MASK) {
1739 		case IPTOS_ECN_CE:
1740 			tp->t_flags |= TF_ECN_SND_ECE;
1741 			TCP_STATINC(TCP_STAT_ECN_CE);
1742 			break;
1743 		case IPTOS_ECN_ECT0:
1744 			TCP_STATINC(TCP_STAT_ECN_ECT);
1745 			break;
1746 		case IPTOS_ECN_ECT1:
1747 			/* XXX: ignore for now -- rpaulo */
1748 			break;
1749 		}
1750 		/*
1751 		 * Congestion experienced.
1752 		 * Ignore if we are already trying to recover.
1753 		 */
1754 		if ((tiflags & TH_ECE) && SEQ_GEQ(tp->snd_una, tp->snd_recover))
1755 			tp->t_congctl->cong_exp(tp);
1756 	}
1757 
1758 	if (opti.ts_present && opti.ts_ecr) {
1759 		/*
1760 		 * Calculate the RTT from the returned time stamp and the
1761 		 * connection's time base.  If the time stamp is later than
1762 		 * the current time, or is extremely old, fall back to non-1323
1763 		 * RTT calculation.  Since ts_rtt is unsigned, we can test both
1764 		 * at the same time.
1765 		 *
1766 		 * Note that ts_rtt is in units of slow ticks (500
1767 		 * ms).  Since most earthbound RTTs are < 500 ms,
1768 		 * observed values will have large quantization noise.
1769 		 * Our smoothed RTT is then the fraction of observed
1770 		 * samples that are 1 tick instead of 0 (times 500
1771 		 * ms).
1772 		 *
1773 		 * ts_rtt is increased by 1 to denote a valid sample,
1774 		 * with 0 indicating an invalid measurement.  This
1775 		 * extra 1 must be removed when ts_rtt is used, or
1776 		 * else an erroneous extra 500 ms will result.
1777 		 */
1778 		ts_rtt = TCP_TIMESTAMP(tp) - opti.ts_ecr + 1;
1779 		if (ts_rtt > TCP_PAWS_IDLE)
1780 			ts_rtt = 0;
1781 	} else {
1782 		ts_rtt = 0;
1783 	}
1784 
1785 	/*
1786 	 * Fast path: check for the two common cases of a uni-directional
1787 	 * data transfer. If:
1788 	 *    o We are in the ESTABLISHED state, and
1789 	 *    o The packet has no control flags, and
1790 	 *    o The packet is in-sequence, and
1791 	 *    o The window didn't change, and
1792 	 *    o We are not retransmitting
1793 	 * It's a candidate.
1794 	 *
1795 	 * If the length (tlen) is zero and the ack moved forward, we're
1796 	 * the sender side of the transfer. Just free the data acked and
1797 	 * wake any higher level process that was blocked waiting for
1798 	 * space.
1799 	 *
1800 	 * If the length is non-zero and the ack didn't move, we're the
1801 	 * receiver side. If we're getting packets in-order (the reassembly
1802 	 * queue is empty), add the data to the socket buffer and note
1803 	 * that we need a delayed ack.
1804 	 */
1805 	if (tp->t_state == TCPS_ESTABLISHED &&
1806 	    (tiflags & (TH_SYN|TH_FIN|TH_RST|TH_URG|TH_ECE|TH_CWR|TH_ACK))
1807 	        == TH_ACK &&
1808 	    (!opti.ts_present || TSTMP_GEQ(opti.ts_val, tp->ts_recent)) &&
1809 	    th->th_seq == tp->rcv_nxt &&
1810 	    tiwin && tiwin == tp->snd_wnd &&
1811 	    tp->snd_nxt == tp->snd_max) {
1812 
1813 		/*
1814 		 * If last ACK falls within this segment's sequence numbers,
1815 		 * record the timestamp.
1816 		 * NOTE that the test is modified according to the latest
1817 		 * proposal of the tcplw@cray.com list (Braden 1993/04/26).
1818 		 *
1819 		 * note that we already know
1820 		 *	TSTMP_GEQ(opti.ts_val, tp->ts_recent)
1821 		 */
1822 		if (opti.ts_present && SEQ_LEQ(th->th_seq, tp->last_ack_sent)) {
1823 			tp->ts_recent_age = tcp_now;
1824 			tp->ts_recent = opti.ts_val;
1825 		}
1826 
1827 		if (tlen == 0) {
1828 			/* Ack prediction. */
1829 			if (SEQ_GT(th->th_ack, tp->snd_una) &&
1830 			    SEQ_LEQ(th->th_ack, tp->snd_max) &&
1831 			    tp->snd_cwnd >= tp->snd_wnd &&
1832 			    tp->t_partialacks < 0) {
1833 				/*
1834 				 * this is a pure ack for outstanding data.
1835 				 */
1836 				if (ts_rtt)
1837 					tcp_xmit_timer(tp, ts_rtt - 1);
1838 				else if (tp->t_rtttime &&
1839 				    SEQ_GT(th->th_ack, tp->t_rtseq))
1840 					tcp_xmit_timer(tp,
1841 					  tcp_now - tp->t_rtttime);
1842 				acked = th->th_ack - tp->snd_una;
1843 				tcps = TCP_STAT_GETREF();
1844 				tcps[TCP_STAT_PREDACK]++;
1845 				tcps[TCP_STAT_RCVACKPACK]++;
1846 				tcps[TCP_STAT_RCVACKBYTE] += acked;
1847 				TCP_STAT_PUTREF();
1848 				nd_hint(tp);
1849 
1850 				if (acked > (tp->t_lastoff - tp->t_inoff))
1851 					tp->t_lastm = NULL;
1852 				sbdrop(&so->so_snd, acked);
1853 				tp->t_lastoff -= acked;
1854 
1855 				icmp_check(tp, th, acked);
1856 
1857 				tp->snd_una = th->th_ack;
1858 				tp->snd_fack = tp->snd_una;
1859 				if (SEQ_LT(tp->snd_high, tp->snd_una))
1860 					tp->snd_high = tp->snd_una;
1861 				/*
1862 				 * drag snd_wl2 along so only newer
1863 				 * ACKs can update the window size.
1864 				 * also avoids the state where snd_wl2
1865 				 * is eventually larger than th_ack and thus
1866 				 * blocking the window update mechanism and
1867 				 * the connection gets stuck for a loooong
1868 				 * time in the zero sized send window state.
1869 				 *
1870 				 * see PR/kern 55567
1871 				 */
1872 				tp->snd_wl2 = tp->snd_una;
1873 
1874 				m_freem(m);
1875 
1876 				/*
1877 				 * If all outstanding data are acked, stop
1878 				 * retransmit timer, otherwise restart timer
1879 				 * using current (possibly backed-off) value.
1880 				 * If process is waiting for space,
1881 				 * wakeup/selnotify/signal.  If data
1882 				 * are ready to send, let tcp_output
1883 				 * decide between more output or persist.
1884 				 */
1885 				if (tp->snd_una == tp->snd_max)
1886 					TCP_TIMER_DISARM(tp, TCPT_REXMT);
1887 				else if (TCP_TIMER_ISARMED(tp,
1888 				    TCPT_PERSIST) == 0)
1889 					TCP_TIMER_ARM(tp, TCPT_REXMT,
1890 					    tp->t_rxtcur);
1891 
1892 				sowwakeup(so);
1893 				if (so->so_snd.sb_cc) {
1894 					KERNEL_LOCK(1, NULL);
1895 					(void)tcp_output(tp);
1896 					KERNEL_UNLOCK_ONE(NULL);
1897 				}
1898 				if (tcp_saveti)
1899 					m_freem(tcp_saveti);
1900 				return;
1901 			}
1902 		} else if (th->th_ack == tp->snd_una &&
1903 		    TAILQ_FIRST(&tp->segq) == NULL &&
1904 		    tlen <= sbspace(&so->so_rcv)) {
1905 			int newsize = 0;
1906 
1907 			/*
1908 			 * this is a pure, in-sequence data packet
1909 			 * with nothing on the reassembly queue and
1910 			 * we have enough buffer space to take it.
1911 			 */
1912 			tp->rcv_nxt += tlen;
1913 
1914 			/*
1915 			 * Pull rcv_up up to prevent seq wrap relative to
1916 			 * rcv_nxt.
1917 			 */
1918 			tp->rcv_up = tp->rcv_nxt;
1919 
1920 			/*
1921 			 * Pull snd_wl1 up to prevent seq wrap relative to
1922 			 * th_seq.
1923 			 */
1924 			tp->snd_wl1 = th->th_seq;
1925 
1926 			tcps = TCP_STAT_GETREF();
1927 			tcps[TCP_STAT_PREDDAT]++;
1928 			tcps[TCP_STAT_RCVPACK]++;
1929 			tcps[TCP_STAT_RCVBYTE] += tlen;
1930 			TCP_STAT_PUTREF();
1931 			nd_hint(tp);
1932 		/*
1933 		 * Automatic sizing enables the performance of large buffers
1934 		 * and most of the efficiency of small ones by only allocating
1935 		 * space when it is needed.
1936 		 *
1937 		 * On the receive side the socket buffer memory is only rarely
1938 		 * used to any significant extent.  This allows us to be much
1939 		 * more aggressive in scaling the receive socket buffer.  For
1940 		 * the case that the buffer space is actually used to a large
1941 		 * extent and we run out of kernel memory we can simply drop
1942 		 * the new segments; TCP on the sender will just retransmit it
1943 		 * later.  Setting the buffer size too big may only consume too
1944 		 * much kernel memory if the application doesn't read() from
1945 		 * the socket or packet loss or reordering makes use of the
1946 		 * reassembly queue.
1947 		 *
1948 		 * The criteria to step up the receive buffer one notch are:
1949 		 *  1. the number of bytes received during the time it takes
1950 		 *     one timestamp to be reflected back to us (the RTT);
1951 		 *  2. received bytes per RTT is within seven eighth of the
1952 		 *     current socket buffer size;
1953 		 *  3. receive buffer size has not hit maximal automatic size;
1954 		 *
1955 		 * This algorithm does one step per RTT at most and only if
1956 		 * we receive a bulk stream w/o packet losses or reorderings.
1957 		 * Shrinking the buffer during idle times is not necessary as
1958 		 * it doesn't consume any memory when idle.
1959 		 *
1960 		 * TODO: Only step up if the application is actually serving
1961 		 * the buffer to better manage the socket buffer resources.
1962 		 */
1963 			if (tcp_do_autorcvbuf &&
1964 			    opti.ts_ecr &&
1965 			    (so->so_rcv.sb_flags & SB_AUTOSIZE)) {
1966 				if (opti.ts_ecr > tp->rfbuf_ts &&
1967 				    opti.ts_ecr - tp->rfbuf_ts < PR_SLOWHZ) {
1968 					if (tp->rfbuf_cnt >
1969 					    (so->so_rcv.sb_hiwat / 8 * 7) &&
1970 					    so->so_rcv.sb_hiwat <
1971 					    tcp_autorcvbuf_max) {
1972 						newsize =
1973 						    uimin(so->so_rcv.sb_hiwat +
1974 						    tcp_autorcvbuf_inc,
1975 						    tcp_autorcvbuf_max);
1976 					}
1977 					/* Start over with next RTT. */
1978 					tp->rfbuf_ts = 0;
1979 					tp->rfbuf_cnt = 0;
1980 				} else
1981 					tp->rfbuf_cnt += tlen;	/* add up */
1982 			}
1983 
1984 			/*
1985 			 * Drop TCP, IP headers and TCP options then add data
1986 			 * to socket buffer.
1987 			 */
1988 			if (so->so_state & SS_CANTRCVMORE) {
1989 				m_freem(m);
1990 			} else {
1991 				/*
1992 				 * Set new socket buffer size.
1993 				 * Give up when limit is reached.
1994 				 */
1995 				if (newsize)
1996 					if (!sbreserve(&so->so_rcv,
1997 					    newsize, so))
1998 						so->so_rcv.sb_flags &= ~SB_AUTOSIZE;
1999 				m_adj(m, off + thlen);
2000 				sbappendstream(&so->so_rcv, m);
2001 			}
2002 			sorwakeup(so);
2003 			tcp_setup_ack(tp, th);
2004 			if (tp->t_flags & TF_ACKNOW) {
2005 				KERNEL_LOCK(1, NULL);
2006 				(void)tcp_output(tp);
2007 				KERNEL_UNLOCK_ONE(NULL);
2008 			}
2009 			if (tcp_saveti)
2010 				m_freem(tcp_saveti);
2011 			return;
2012 		}
2013 	}
2014 
2015 	/*
2016 	 * Compute mbuf offset to TCP data segment.
2017 	 */
2018 	hdroptlen = off + thlen;
2019 
2020 	/*
2021 	 * Calculate amount of space in receive window. Receive window is
2022 	 * amount of space in rcv queue, but not less than advertised
2023 	 * window.
2024 	 */
2025 	{
2026 		int win;
2027 		win = sbspace(&so->so_rcv);
2028 		if (win < 0)
2029 			win = 0;
2030 		tp->rcv_wnd = imax(win, (int)(tp->rcv_adv - tp->rcv_nxt));
2031 	}
2032 
2033 	/* Reset receive buffer auto scaling when not in bulk receive mode. */
2034 	tp->rfbuf_ts = 0;
2035 	tp->rfbuf_cnt = 0;
2036 
2037 	switch (tp->t_state) {
2038 	/*
2039 	 * If the state is SYN_SENT:
2040 	 *	if seg contains an ACK, but not for our SYN, drop the input.
2041 	 *	if seg contains a RST, then drop the connection.
2042 	 *	if seg does not contain SYN, then drop it.
2043 	 * Otherwise this is an acceptable SYN segment
2044 	 *	initialize tp->rcv_nxt and tp->irs
2045 	 *	if seg contains ack then advance tp->snd_una
2046 	 *	if seg contains a ECE and ECN support is enabled, the stream
2047 	 *	    is ECN capable.
2048 	 *	if SYN has been acked change to ESTABLISHED else SYN_RCVD state
2049 	 *	arrange for segment to be acked (eventually)
2050 	 *	continue processing rest of data/controls, beginning with URG
2051 	 */
2052 	case TCPS_SYN_SENT:
2053 		if ((tiflags & TH_ACK) &&
2054 		    (SEQ_LEQ(th->th_ack, tp->iss) ||
2055 		     SEQ_GT(th->th_ack, tp->snd_max)))
2056 			goto dropwithreset;
2057 		if (tiflags & TH_RST) {
2058 			if (tiflags & TH_ACK)
2059 				tp = tcp_drop(tp, ECONNREFUSED);
2060 			goto drop;
2061 		}
2062 		if ((tiflags & TH_SYN) == 0)
2063 			goto drop;
2064 		if (tiflags & TH_ACK) {
2065 			tp->snd_una = th->th_ack;
2066 			if (SEQ_LT(tp->snd_nxt, tp->snd_una))
2067 				tp->snd_nxt = tp->snd_una;
2068 			if (SEQ_LT(tp->snd_high, tp->snd_una))
2069 				tp->snd_high = tp->snd_una;
2070 			TCP_TIMER_DISARM(tp, TCPT_REXMT);
2071 
2072 			if ((tiflags & TH_ECE) && tcp_do_ecn) {
2073 				tp->t_flags |= TF_ECN_PERMIT;
2074 				TCP_STATINC(TCP_STAT_ECN_SHS);
2075 			}
2076 		}
2077 		tp->irs = th->th_seq;
2078 		tcp_rcvseqinit(tp);
2079 		tp->t_flags |= TF_ACKNOW;
2080 		tcp_mss_from_peer(tp, opti.maxseg);
2081 
2082 		/*
2083 		 * Initialize the initial congestion window.  If we
2084 		 * had to retransmit the SYN, we must initialize cwnd
2085 		 * to 1 segment (i.e. the Loss Window).
2086 		 */
2087 		if (tp->t_flags & TF_SYN_REXMT)
2088 			tp->snd_cwnd = tp->t_peermss;
2089 		else {
2090 			int ss = tcp_init_win;
2091 			if (inp->inp_af == AF_INET && in_localaddr(in4p_faddr(inp)))
2092 				ss = tcp_init_win_local;
2093 #ifdef INET6
2094 			else if (inp->inp_af == AF_INET6 && in6_localaddr(&in6p_faddr(inp)))
2095 				ss = tcp_init_win_local;
2096 #endif
2097 			tp->snd_cwnd = TCP_INITIAL_WINDOW(ss, tp->t_peermss);
2098 		}
2099 
2100 		tcp_rmx_rtt(tp);
2101 		if (tiflags & TH_ACK) {
2102 			TCP_STATINC(TCP_STAT_CONNECTS);
2103 			/*
2104 			 * move tcp_established before soisconnected
2105 			 * because upcall handler can drive tcp_output
2106 			 * functionality.
2107 			 * XXX we might call soisconnected at the end of
2108 			 * all processing
2109 			 */
2110 			tcp_established(tp);
2111 			soisconnected(so);
2112 			/* Do window scaling on this connection? */
2113 			if ((tp->t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) ==
2114 			    (TF_RCVD_SCALE|TF_REQ_SCALE)) {
2115 				tp->snd_scale = tp->requested_s_scale;
2116 				tp->rcv_scale = tp->request_r_scale;
2117 			}
2118 			TCP_REASS_LOCK(tp);
2119 			(void)tcp_reass(tp, NULL, NULL, tlen);
2120 			/*
2121 			 * if we didn't have to retransmit the SYN,
2122 			 * use its rtt as our initial srtt & rtt var.
2123 			 */
2124 			if (tp->t_rtttime)
2125 				tcp_xmit_timer(tp, tcp_now - tp->t_rtttime);
2126 		} else {
2127 			tp->t_state = TCPS_SYN_RECEIVED;
2128 		}
2129 
2130 		/*
2131 		 * Advance th->th_seq to correspond to first data byte.
2132 		 * If data, trim to stay within window,
2133 		 * dropping FIN if necessary.
2134 		 */
2135 		th->th_seq++;
2136 		if (tlen > tp->rcv_wnd) {
2137 			todrop = tlen - tp->rcv_wnd;
2138 			m_adj(m, -todrop);
2139 			tlen = tp->rcv_wnd;
2140 			tiflags &= ~TH_FIN;
2141 			tcps = TCP_STAT_GETREF();
2142 			tcps[TCP_STAT_RCVPACKAFTERWIN]++;
2143 			tcps[TCP_STAT_RCVBYTEAFTERWIN] += todrop;
2144 			TCP_STAT_PUTREF();
2145 		}
2146 		tp->snd_wl1 = th->th_seq - 1;
2147 		tp->rcv_up = th->th_seq;
2148 		goto step6;
2149 
2150 	/*
2151 	 * If the state is SYN_RECEIVED:
2152 	 *	If seg contains an ACK, but not for our SYN, drop the input
2153 	 *	and generate an RST.  See page 36, rfc793
2154 	 */
2155 	case TCPS_SYN_RECEIVED:
2156 		if ((tiflags & TH_ACK) &&
2157 		    (SEQ_LEQ(th->th_ack, tp->iss) ||
2158 		     SEQ_GT(th->th_ack, tp->snd_max)))
2159 			goto dropwithreset;
2160 		break;
2161 	}
2162 
2163 	/*
2164 	 * From here on, we're dealing with !LISTEN and !SYN_SENT.
2165 	 */
2166 	KASSERT(tp->t_state != TCPS_LISTEN &&
2167 	    tp->t_state != TCPS_SYN_SENT);
2168 
2169 	/*
2170 	 * RFC1323 PAWS: if we have a timestamp reply on this segment and
2171 	 * it's less than ts_recent, drop it.
2172 	 */
2173 	if (opti.ts_present && (tiflags & TH_RST) == 0 && tp->ts_recent &&
2174 	    TSTMP_LT(opti.ts_val, tp->ts_recent)) {
2175 		/* Check to see if ts_recent is over 24 days old.  */
2176 		if (tcp_now - tp->ts_recent_age > TCP_PAWS_IDLE) {
2177 			/*
2178 			 * Invalidate ts_recent.  If this segment updates
2179 			 * ts_recent, the age will be reset later and ts_recent
2180 			 * will get a valid value.  If it does not, setting
2181 			 * ts_recent to zero will at least satisfy the
2182 			 * requirement that zero be placed in the timestamp
2183 			 * echo reply when ts_recent isn't valid.  The
2184 			 * age isn't reset until we get a valid ts_recent
2185 			 * because we don't want out-of-order segments to be
2186 			 * dropped when ts_recent is old.
2187 			 */
2188 			tp->ts_recent = 0;
2189 		} else {
2190 			tcps = TCP_STAT_GETREF();
2191 			tcps[TCP_STAT_RCVDUPPACK]++;
2192 			tcps[TCP_STAT_RCVDUPBYTE] += tlen;
2193 			tcps[TCP_STAT_PAWSDROP]++;
2194 			TCP_STAT_PUTREF();
2195 			tcp_new_dsack(tp, th->th_seq, tlen);
2196 			goto dropafterack;
2197 		}
2198 	}
2199 
2200 	/*
2201 	 * Check that at least some bytes of the segment are within the
2202 	 * receive window. If segment begins before rcv_nxt, drop leading
2203 	 * data (and SYN); if nothing left, just ack.
2204 	 */
2205 	todrop = tp->rcv_nxt - th->th_seq;
2206 	dupseg = false;
2207 	if (todrop > 0) {
2208 		if (tiflags & TH_SYN) {
2209 			tiflags &= ~TH_SYN;
2210 			th->th_seq++;
2211 			tcp_urp_drop(th, 1, &tiflags);
2212 			todrop--;
2213 		}
2214 		if (todrop > tlen ||
2215 		    (todrop == tlen && (tiflags & TH_FIN) == 0)) {
2216 			/*
2217 			 * Any valid FIN or RST must be to the left of the
2218 			 * window.  At this point the FIN or RST must be a
2219 			 * duplicate or out of sequence; drop it.
2220 			 */
2221 			if (tiflags & TH_RST)
2222 				goto drop;
2223 			tiflags &= ~(TH_FIN|TH_RST);
2224 
2225 			/*
2226 			 * Send an ACK to resynchronize and drop any data.
2227 			 * But keep on processing for RST or ACK.
2228 			 */
2229 			tp->t_flags |= TF_ACKNOW;
2230 			todrop = tlen;
2231 			dupseg = true;
2232 			tcps = TCP_STAT_GETREF();
2233 			tcps[TCP_STAT_RCVDUPPACK]++;
2234 			tcps[TCP_STAT_RCVDUPBYTE] += todrop;
2235 			TCP_STAT_PUTREF();
2236 		} else if ((tiflags & TH_RST) && th->th_seq != tp->rcv_nxt) {
2237 			/*
2238 			 * Test for reset before adjusting the sequence
2239 			 * number for overlapping data.
2240 			 */
2241 			goto dropafterack_ratelim;
2242 		} else {
2243 			tcps = TCP_STAT_GETREF();
2244 			tcps[TCP_STAT_RCVPARTDUPPACK]++;
2245 			tcps[TCP_STAT_RCVPARTDUPBYTE] += todrop;
2246 			TCP_STAT_PUTREF();
2247 		}
2248 		tcp_new_dsack(tp, th->th_seq, todrop);
2249 		hdroptlen += todrop;	/* drop from head afterwards (m_adj) */
2250 		th->th_seq += todrop;
2251 		tlen -= todrop;
2252 		tcp_urp_drop(th, todrop, &tiflags);
2253 	}
2254 
2255 	/*
2256 	 * If new data is received on a connection after the user processes
2257 	 * are gone, then RST the other end.
2258 	 */
2259 	if ((so->so_state & SS_NOFDREF) &&
2260 	    tp->t_state > TCPS_CLOSE_WAIT && tlen) {
2261 		tp = tcp_close(tp);
2262 		TCP_STATINC(TCP_STAT_RCVAFTERCLOSE);
2263 		goto dropwithreset;
2264 	}
2265 
2266 	/*
2267 	 * If the segment ends after the window, drop trailing data (and
2268 	 * PUSH and FIN); if nothing left, just ACK.
2269 	 */
2270 	todrop = (th->th_seq + tlen) - (tp->rcv_nxt + tp->rcv_wnd);
2271 	if (todrop > 0) {
2272 		TCP_STATINC(TCP_STAT_RCVPACKAFTERWIN);
2273 		if (todrop >= tlen) {
2274 			/*
2275 			 * The segment actually starts after the window.
2276 			 * th->th_seq + tlen - tp->rcv_nxt - tp->rcv_wnd >= tlen
2277 			 * th->th_seq - tp->rcv_nxt - tp->rcv_wnd >= 0
2278 			 * th->th_seq >= tp->rcv_nxt + tp->rcv_wnd
2279 			 */
2280 			TCP_STATADD(TCP_STAT_RCVBYTEAFTERWIN, tlen);
2281 
2282 			/*
2283 			 * If a new connection request is received while in
2284 			 * TIME_WAIT, drop the old connection and start over
2285 			 * if the sequence numbers are above the previous
2286 			 * ones.
2287 			 *
2288 			 * NOTE: We need to put the header fields back into
2289 			 * network order.
2290 			 */
2291 			if ((tiflags & TH_SYN) &&
2292 			    tp->t_state == TCPS_TIME_WAIT &&
2293 			    SEQ_GT(th->th_seq, tp->rcv_nxt)) {
2294 				tp = tcp_close(tp);
2295 				tcp_fields_to_net(th);
2296 				m_freem(tcp_saveti);
2297 				tcp_saveti = NULL;
2298 				goto findpcb;
2299 			}
2300 
2301 			/*
2302 			 * If window is closed can only take segments at
2303 			 * window edge, and have to drop data and PUSH from
2304 			 * incoming segments.  Continue processing, but
2305 			 * remember to ack.  Otherwise, drop segment
2306 			 * and (if not RST) ack.
2307 			 */
2308 			if (tp->rcv_wnd == 0 && th->th_seq == tp->rcv_nxt) {
2309 				KASSERT(todrop == tlen);
2310 				tp->t_flags |= TF_ACKNOW;
2311 				TCP_STATINC(TCP_STAT_RCVWINPROBE);
2312 			} else {
2313 				goto dropafterack;
2314 			}
2315 		} else {
2316 			TCP_STATADD(TCP_STAT_RCVBYTEAFTERWIN, todrop);
2317 		}
2318 		m_adj(m, -todrop);
2319 		tlen -= todrop;
2320 		tiflags &= ~(TH_PUSH|TH_FIN);
2321 	}
2322 
2323 	/*
2324 	 * If last ACK falls within this segment's sequence numbers,
2325 	 *  record the timestamp.
2326 	 * NOTE:
2327 	 * 1) That the test incorporates suggestions from the latest
2328 	 *    proposal of the tcplw@cray.com list (Braden 1993/04/26).
2329 	 * 2) That updating only on newer timestamps interferes with
2330 	 *    our earlier PAWS tests, so this check should be solely
2331 	 *    predicated on the sequence space of this segment.
2332 	 * 3) That we modify the segment boundary check to be
2333 	 *        Last.ACK.Sent <= SEG.SEQ + SEG.Len
2334 	 *    instead of RFC1323's
2335 	 *        Last.ACK.Sent < SEG.SEQ + SEG.Len,
2336 	 *    This modified check allows us to overcome RFC1323's
2337 	 *    limitations as described in Stevens TCP/IP Illustrated
2338 	 *    Vol. 2 p.869. In such cases, we can still calculate the
2339 	 *    RTT correctly when RCV.NXT == Last.ACK.Sent.
2340 	 */
2341 	if (opti.ts_present &&
2342 	    SEQ_LEQ(th->th_seq, tp->last_ack_sent) &&
2343 	    SEQ_LEQ(tp->last_ack_sent, th->th_seq + tlen +
2344 	         ((tiflags & (TH_SYN|TH_FIN)) != 0))) {
2345 		tp->ts_recent_age = tcp_now;
2346 		tp->ts_recent = opti.ts_val;
2347 	}
2348 
2349 	/*
2350 	 * If the RST bit is set examine the state:
2351 	 *    RECEIVED state:
2352 	 *        If passive open, return to LISTEN state.
2353 	 *        If active open, inform user that connection was refused.
2354 	 *    ESTABLISHED, FIN_WAIT_1, FIN_WAIT2, CLOSE_WAIT states:
2355 	 *        Inform user that connection was reset, and close tcb.
2356 	 *    CLOSING, LAST_ACK, TIME_WAIT states:
2357 	 *        Close the tcb.
2358 	 */
2359 	if (tiflags & TH_RST) {
2360 		if (th->th_seq != tp->rcv_nxt)
2361 			goto dropafterack_ratelim;
2362 
2363 		switch (tp->t_state) {
2364 		case TCPS_SYN_RECEIVED:
2365 			so->so_error = ECONNREFUSED;
2366 			goto close;
2367 
2368 		case TCPS_ESTABLISHED:
2369 		case TCPS_FIN_WAIT_1:
2370 		case TCPS_FIN_WAIT_2:
2371 		case TCPS_CLOSE_WAIT:
2372 			so->so_error = ECONNRESET;
2373 		close:
2374 			tp->t_state = TCPS_CLOSED;
2375 			TCP_STATINC(TCP_STAT_DROPS);
2376 			tp = tcp_close(tp);
2377 			goto drop;
2378 
2379 		case TCPS_CLOSING:
2380 		case TCPS_LAST_ACK:
2381 		case TCPS_TIME_WAIT:
2382 			tp = tcp_close(tp);
2383 			goto drop;
2384 		}
2385 	}
2386 
2387 	/*
2388 	 * Since we've covered the SYN-SENT and SYN-RECEIVED states above
2389 	 * we must be in a synchronized state.  RFC793 states (under Reset
2390 	 * Generation) that any unacceptable segment (an out-of-order SYN
2391 	 * qualifies) received in a synchronized state must elicit only an
2392 	 * empty acknowledgment segment ... and the connection remains in
2393 	 * the same state.
2394 	 */
2395 	if (tiflags & TH_SYN) {
2396 		if (tp->rcv_nxt == th->th_seq) {
2397 			tcp_respond(tp, m, m, th, (tcp_seq)0, th->th_ack - 1,
2398 			    TH_ACK);
2399 			if (tcp_saveti)
2400 				m_freem(tcp_saveti);
2401 			return;
2402 		}
2403 
2404 		goto dropafterack_ratelim;
2405 	}
2406 
2407 	/*
2408 	 * If the ACK bit is off we drop the segment and return.
2409 	 */
2410 	if ((tiflags & TH_ACK) == 0) {
2411 		if (tp->t_flags & TF_ACKNOW)
2412 			goto dropafterack;
2413 		goto drop;
2414 	}
2415 
2416 	/*
2417 	 * From here on, we're doing ACK processing.
2418 	 */
2419 
2420 	switch (tp->t_state) {
2421 	/*
2422 	 * In SYN_RECEIVED state if the ack ACKs our SYN then enter
2423 	 * ESTABLISHED state and continue processing, otherwise
2424 	 * send an RST.
2425 	 */
2426 	case TCPS_SYN_RECEIVED:
2427 		if (SEQ_GT(tp->snd_una, th->th_ack) ||
2428 		    SEQ_GT(th->th_ack, tp->snd_max))
2429 			goto dropwithreset;
2430 		TCP_STATINC(TCP_STAT_CONNECTS);
2431 		soisconnected(so);
2432 		tcp_established(tp);
2433 		/* Do window scaling? */
2434 		if ((tp->t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) ==
2435 		    (TF_RCVD_SCALE|TF_REQ_SCALE)) {
2436 			tp->snd_scale = tp->requested_s_scale;
2437 			tp->rcv_scale = tp->request_r_scale;
2438 		}
2439 		TCP_REASS_LOCK(tp);
2440 		(void)tcp_reass(tp, NULL, NULL, tlen);
2441 		tp->snd_wl1 = th->th_seq - 1;
2442 		/* FALLTHROUGH */
2443 
2444 	/*
2445 	 * In ESTABLISHED state: drop duplicate ACKs; ACK out of range
2446 	 * ACKs.  If the ack is in the range
2447 	 *	tp->snd_una < th->th_ack <= tp->snd_max
2448 	 * then advance tp->snd_una to th->th_ack and drop
2449 	 * data from the retransmission queue.  If this ACK reflects
2450 	 * more up to date window information we update our window information.
2451 	 */
2452 	case TCPS_ESTABLISHED:
2453 	case TCPS_FIN_WAIT_1:
2454 	case TCPS_FIN_WAIT_2:
2455 	case TCPS_CLOSE_WAIT:
2456 	case TCPS_CLOSING:
2457 	case TCPS_LAST_ACK:
2458 	case TCPS_TIME_WAIT:
2459 		if (SEQ_LEQ(th->th_ack, tp->snd_una)) {
2460 			if (tlen == 0 && !dupseg && tiwin == tp->snd_wnd) {
2461 				TCP_STATINC(TCP_STAT_RCVDUPACK);
2462 				/*
2463 				 * If we have outstanding data (other than
2464 				 * a window probe), this is a completely
2465 				 * duplicate ack (ie, window info didn't
2466 				 * change), the ack is the biggest we've
2467 				 * seen and we've seen exactly our rexmt
2468 				 * threshold of them, assume a packet
2469 				 * has been dropped and retransmit it.
2470 				 * Kludge snd_nxt & the congestion
2471 				 * window so we send only this one
2472 				 * packet.
2473 				 */
2474 				if (TCP_TIMER_ISARMED(tp, TCPT_REXMT) == 0 ||
2475 				    th->th_ack != tp->snd_una)
2476 					tp->t_dupacks = 0;
2477 				else if (tp->t_partialacks < 0 &&
2478 				    (++tp->t_dupacks == tcprexmtthresh ||
2479 				     TCP_FACK_FASTRECOV(tp))) {
2480 					/*
2481 					 * Do the fast retransmit, and adjust
2482 					 * congestion control parameters.
2483 					 */
2484 					if (tp->t_congctl->fast_retransmit(tp, th)) {
2485 						/* False fast retransmit */
2486 						break;
2487 					}
2488 					goto drop;
2489 				} else if (tp->t_dupacks > tcprexmtthresh) {
2490 					tp->snd_cwnd += tp->t_segsz;
2491 					KERNEL_LOCK(1, NULL);
2492 					(void)tcp_output(tp);
2493 					KERNEL_UNLOCK_ONE(NULL);
2494 					goto drop;
2495 				}
2496 			} else {
2497 				/*
2498 				 * If the ack appears to be very old, only
2499 				 * allow data that is in-sequence.  This
2500 				 * makes it somewhat more difficult to insert
2501 				 * forged data by guessing sequence numbers.
2502 				 * Sent an ack to try to update the send
2503 				 * sequence number on the other side.
2504 				 */
2505 				if (tlen && th->th_seq != tp->rcv_nxt &&
2506 				    SEQ_LT(th->th_ack,
2507 				    tp->snd_una - tp->max_sndwnd))
2508 					goto dropafterack;
2509 			}
2510 			break;
2511 		}
2512 		/*
2513 		 * If the congestion window was inflated to account
2514 		 * for the other side's cached packets, retract it.
2515 		 */
2516 		tp->t_congctl->fast_retransmit_newack(tp, th);
2517 
2518 		if (SEQ_GT(th->th_ack, tp->snd_max)) {
2519 			TCP_STATINC(TCP_STAT_RCVACKTOOMUCH);
2520 			goto dropafterack;
2521 		}
2522 		acked = th->th_ack - tp->snd_una;
2523 		tcps = TCP_STAT_GETREF();
2524 		tcps[TCP_STAT_RCVACKPACK]++;
2525 		tcps[TCP_STAT_RCVACKBYTE] += acked;
2526 		TCP_STAT_PUTREF();
2527 
2528 		/*
2529 		 * If we have a timestamp reply, update smoothed
2530 		 * round trip time.  If no timestamp is present but
2531 		 * transmit timer is running and timed sequence
2532 		 * number was acked, update smoothed round trip time.
2533 		 * Since we now have an rtt measurement, cancel the
2534 		 * timer backoff (cf., Phil Karn's retransmit alg.).
2535 		 * Recompute the initial retransmit timer.
2536 		 */
2537 		if (ts_rtt)
2538 			tcp_xmit_timer(tp, ts_rtt - 1);
2539 		else if (tp->t_rtttime && SEQ_GT(th->th_ack, tp->t_rtseq))
2540 			tcp_xmit_timer(tp, tcp_now - tp->t_rtttime);
2541 
2542 		/*
2543 		 * If all outstanding data is acked, stop retransmit
2544 		 * timer and remember to restart (more output or persist).
2545 		 * If there is more data to be acked, restart retransmit
2546 		 * timer, using current (possibly backed-off) value.
2547 		 */
2548 		if (th->th_ack == tp->snd_max) {
2549 			TCP_TIMER_DISARM(tp, TCPT_REXMT);
2550 			needoutput = 1;
2551 		} else if (TCP_TIMER_ISARMED(tp, TCPT_PERSIST) == 0)
2552 			TCP_TIMER_ARM(tp, TCPT_REXMT, tp->t_rxtcur);
2553 
2554 		/*
2555 		 * New data has been acked, adjust the congestion window.
2556 		 */
2557 		tp->t_congctl->newack(tp, th);
2558 
2559 		nd_hint(tp);
2560 		if (acked > so->so_snd.sb_cc) {
2561 			tp->snd_wnd -= so->so_snd.sb_cc;
2562 			sbdrop(&so->so_snd, (int)so->so_snd.sb_cc);
2563 			ourfinisacked = 1;
2564 		} else {
2565 			if (acked > (tp->t_lastoff - tp->t_inoff))
2566 				tp->t_lastm = NULL;
2567 			sbdrop(&so->so_snd, acked);
2568 			tp->t_lastoff -= acked;
2569 			if (tp->snd_wnd > acked)
2570 				tp->snd_wnd -= acked;
2571 			else
2572 				tp->snd_wnd = 0;
2573 			ourfinisacked = 0;
2574 		}
2575 		sowwakeup(so);
2576 
2577 		icmp_check(tp, th, acked);
2578 
2579 		tp->snd_una = th->th_ack;
2580 		if (SEQ_GT(tp->snd_una, tp->snd_fack))
2581 			tp->snd_fack = tp->snd_una;
2582 		if (SEQ_LT(tp->snd_nxt, tp->snd_una))
2583 			tp->snd_nxt = tp->snd_una;
2584 		if (SEQ_LT(tp->snd_high, tp->snd_una))
2585 			tp->snd_high = tp->snd_una;
2586 
2587 		switch (tp->t_state) {
2588 
2589 		/*
2590 		 * In FIN_WAIT_1 STATE in addition to the processing
2591 		 * for the ESTABLISHED state if our FIN is now acknowledged
2592 		 * then enter FIN_WAIT_2.
2593 		 */
2594 		case TCPS_FIN_WAIT_1:
2595 			if (ourfinisacked) {
2596 				/*
2597 				 * If we can't receive any more
2598 				 * data, then closing user can proceed.
2599 				 * Starting the timer is contrary to the
2600 				 * specification, but if we don't get a FIN
2601 				 * we'll hang forever.
2602 				 */
2603 				if (so->so_state & SS_CANTRCVMORE) {
2604 					soisdisconnected(so);
2605 					if (tp->t_maxidle > 0)
2606 						TCP_TIMER_ARM(tp, TCPT_2MSL,
2607 						    tp->t_maxidle);
2608 				}
2609 				tp->t_state = TCPS_FIN_WAIT_2;
2610 			}
2611 			break;
2612 
2613 	 	/*
2614 		 * In CLOSING STATE in addition to the processing for
2615 		 * the ESTABLISHED state if the ACK acknowledges our FIN
2616 		 * then enter the TIME-WAIT state, otherwise ignore
2617 		 * the segment.
2618 		 */
2619 		case TCPS_CLOSING:
2620 			if (ourfinisacked) {
2621 				tp->t_state = TCPS_TIME_WAIT;
2622 				tcp_canceltimers(tp);
2623 				TCP_TIMER_ARM(tp, TCPT_2MSL, 2 * tp->t_msl);
2624 				soisdisconnected(so);
2625 			}
2626 			break;
2627 
2628 		/*
2629 		 * In LAST_ACK, we may still be waiting for data to drain
2630 		 * and/or to be acked, as well as for the ack of our FIN.
2631 		 * If our FIN is now acknowledged, delete the TCB,
2632 		 * enter the closed state and return.
2633 		 */
2634 		case TCPS_LAST_ACK:
2635 			if (ourfinisacked) {
2636 				tp = tcp_close(tp);
2637 				goto drop;
2638 			}
2639 			break;
2640 
2641 		/*
2642 		 * In TIME_WAIT state the only thing that should arrive
2643 		 * is a retransmission of the remote FIN.  Acknowledge
2644 		 * it and restart the finack timer.
2645 		 */
2646 		case TCPS_TIME_WAIT:
2647 			TCP_TIMER_ARM(tp, TCPT_2MSL, 2 * tp->t_msl);
2648 			goto dropafterack;
2649 		}
2650 	}
2651 
2652 step6:
2653 	/*
2654 	 * Update window information.
2655 	 * Don't look at window if no ACK: TAC's send garbage on first SYN.
2656 	 */
2657 	if ((tiflags & TH_ACK) && (SEQ_LT(tp->snd_wl1, th->th_seq) ||
2658 	    (tp->snd_wl1 == th->th_seq && (SEQ_LT(tp->snd_wl2, th->th_ack) ||
2659 	    (tp->snd_wl2 == th->th_ack && tiwin > tp->snd_wnd))))) {
2660 		/* keep track of pure window updates */
2661 		if (tlen == 0 &&
2662 		    tp->snd_wl2 == th->th_ack && tiwin > tp->snd_wnd)
2663 			TCP_STATINC(TCP_STAT_RCVWINUPD);
2664 		tp->snd_wnd = tiwin;
2665 		tp->snd_wl1 = th->th_seq;
2666 		tp->snd_wl2 = th->th_ack;
2667 		if (tp->snd_wnd > tp->max_sndwnd)
2668 			tp->max_sndwnd = tp->snd_wnd;
2669 		needoutput = 1;
2670 	}
2671 
2672 	/*
2673 	 * Process segments with URG.
2674 	 */
2675 	if ((tiflags & TH_URG) && th->th_urp &&
2676 	    TCPS_HAVERCVDFIN(tp->t_state) == 0) {
2677 		/*
2678 		 * This is a kludge, but if we receive and accept
2679 		 * random urgent pointers, we'll crash in
2680 		 * soreceive.  It's hard to imagine someone
2681 		 * actually wanting to send this much urgent data.
2682 		 */
2683 		if (th->th_urp + so->so_rcv.sb_cc > sb_max) {
2684 			th->th_urp = 0;			/* XXX */
2685 			tiflags &= ~TH_URG;		/* XXX */
2686 			goto dodata;			/* XXX */
2687 		}
2688 
2689 		/*
2690 		 * If this segment advances the known urgent pointer,
2691 		 * then mark the data stream.  This should not happen
2692 		 * in CLOSE_WAIT, CLOSING, LAST_ACK or TIME_WAIT STATES since
2693 		 * a FIN has been received from the remote side.
2694 		 * In these states we ignore the URG.
2695 		 *
2696 		 * According to RFC961 (Assigned Protocols),
2697 		 * the urgent pointer points to the last octet
2698 		 * of urgent data.  We continue, however,
2699 		 * to consider it to indicate the first octet
2700 		 * of data past the urgent section as the original
2701 		 * spec states (in one of two places).
2702 		 */
2703 		if (SEQ_GT(th->th_seq+th->th_urp, tp->rcv_up)) {
2704 			tp->rcv_up = th->th_seq + th->th_urp;
2705 			so->so_oobmark = so->so_rcv.sb_cc +
2706 			    (tp->rcv_up - tp->rcv_nxt) - 1;
2707 			if (so->so_oobmark == 0)
2708 				so->so_state |= SS_RCVATMARK;
2709 			sohasoutofband(so);
2710 			tp->t_oobflags &= ~(TCPOOB_HAVEDATA | TCPOOB_HADDATA);
2711 		}
2712 
2713 		/*
2714 		 * Remove out of band data so doesn't get presented to user.
2715 		 * This can happen independent of advancing the URG pointer,
2716 		 * but if two URG's are pending at once, some out-of-band
2717 		 * data may creep in... ick.
2718 		 */
2719 		if (th->th_urp <= (u_int16_t)tlen &&
2720 		    (so->so_options & SO_OOBINLINE) == 0)
2721 			tcp_pulloutofband(so, th, m, hdroptlen);
2722 	} else {
2723 		/*
2724 		 * If no out of band data is expected,
2725 		 * pull receive urgent pointer along
2726 		 * with the receive window.
2727 		 */
2728 		if (SEQ_GT(tp->rcv_nxt, tp->rcv_up))
2729 			tp->rcv_up = tp->rcv_nxt;
2730 	}
2731 dodata:
2732 
2733 	/*
2734 	 * Process the segment text, merging it into the TCP sequencing queue,
2735 	 * and arranging for acknowledgement of receipt if necessary.
2736 	 * This process logically involves adjusting tp->rcv_wnd as data
2737 	 * is presented to the user (this happens in tcp_usrreq.c,
2738 	 * tcp_rcvd()).  If a FIN has already been received on this
2739 	 * connection then we just ignore the text.
2740 	 */
2741 	if ((tlen || (tiflags & TH_FIN)) &&
2742 	    TCPS_HAVERCVDFIN(tp->t_state) == 0) {
2743 		/*
2744 		 * Handle the common case:
2745 		 *  o Segment is the next to be received, and
2746 		 *  o The queue is empty, and
2747 		 *  o The connection is established
2748 		 * In this case, we avoid calling tcp_reass.
2749 		 *
2750 		 * tcp_setup_ack: set DELACK for segments received in order,
2751 		 * but ack immediately when segments are out of order (so that
2752 		 * fast retransmit can work).
2753 		 */
2754 		TCP_REASS_LOCK(tp);
2755 		if (th->th_seq == tp->rcv_nxt &&
2756 		    TAILQ_FIRST(&tp->segq) == NULL &&
2757 		    tp->t_state == TCPS_ESTABLISHED) {
2758 			tcp_setup_ack(tp, th);
2759 			tp->rcv_nxt += tlen;
2760 			tiflags = th->th_flags & TH_FIN;
2761 			tcps = TCP_STAT_GETREF();
2762 			tcps[TCP_STAT_RCVPACK]++;
2763 			tcps[TCP_STAT_RCVBYTE] += tlen;
2764 			TCP_STAT_PUTREF();
2765 			nd_hint(tp);
2766 			if (so->so_state & SS_CANTRCVMORE) {
2767 				m_freem(m);
2768 			} else {
2769 				m_adj(m, hdroptlen);
2770 				sbappendstream(&(so)->so_rcv, m);
2771 			}
2772 			TCP_REASS_UNLOCK(tp);
2773 			sorwakeup(so);
2774 		} else {
2775 			m_adj(m, hdroptlen);
2776 			tiflags = tcp_reass(tp, th, m, tlen);
2777 			tp->t_flags |= TF_ACKNOW;
2778 		}
2779 
2780 		/*
2781 		 * Note the amount of data that peer has sent into
2782 		 * our window, in order to estimate the sender's
2783 		 * buffer size.
2784 		 */
2785 		len = so->so_rcv.sb_hiwat - (tp->rcv_adv - tp->rcv_nxt);
2786 	} else {
2787 		m_freem(m);
2788 		m = NULL;
2789 		tiflags &= ~TH_FIN;
2790 	}
2791 
2792 	/*
2793 	 * If FIN is received ACK the FIN and let the user know
2794 	 * that the connection is closing.  Ignore a FIN received before
2795 	 * the connection is fully established.
2796 	 */
2797 	if ((tiflags & TH_FIN) && TCPS_HAVEESTABLISHED(tp->t_state)) {
2798 		if (TCPS_HAVERCVDFIN(tp->t_state) == 0) {
2799 			socantrcvmore(so);
2800 			tp->t_flags |= TF_ACKNOW;
2801 			tp->rcv_nxt++;
2802 		}
2803 		switch (tp->t_state) {
2804 
2805 	 	/*
2806 		 * In ESTABLISHED STATE enter the CLOSE_WAIT state.
2807 		 */
2808 		case TCPS_ESTABLISHED:
2809 			tp->t_state = TCPS_CLOSE_WAIT;
2810 			break;
2811 
2812 	 	/*
2813 		 * If still in FIN_WAIT_1 STATE FIN has not been acked so
2814 		 * enter the CLOSING state.
2815 		 */
2816 		case TCPS_FIN_WAIT_1:
2817 			tp->t_state = TCPS_CLOSING;
2818 			break;
2819 
2820 	 	/*
2821 		 * In FIN_WAIT_2 state enter the TIME_WAIT state,
2822 		 * starting the time-wait timer, turning off the other
2823 		 * standard timers.
2824 		 */
2825 		case TCPS_FIN_WAIT_2:
2826 			tp->t_state = TCPS_TIME_WAIT;
2827 			tcp_canceltimers(tp);
2828 			TCP_TIMER_ARM(tp, TCPT_2MSL, 2 * tp->t_msl);
2829 			soisdisconnected(so);
2830 			break;
2831 
2832 		/*
2833 		 * In TIME_WAIT state restart the 2 MSL time_wait timer.
2834 		 */
2835 		case TCPS_TIME_WAIT:
2836 			TCP_TIMER_ARM(tp, TCPT_2MSL, 2 * tp->t_msl);
2837 			break;
2838 		}
2839 	}
2840 #ifdef TCP_DEBUG
2841 	if (so->so_options & SO_DEBUG)
2842 		tcp_trace(TA_INPUT, ostate, tp, tcp_saveti, 0);
2843 #endif
2844 
2845 	/*
2846 	 * Return any desired output.
2847 	 */
2848 	if (needoutput || (tp->t_flags & TF_ACKNOW)) {
2849 		KERNEL_LOCK(1, NULL);
2850 		(void)tcp_output(tp);
2851 		KERNEL_UNLOCK_ONE(NULL);
2852 	}
2853 	if (tcp_saveti)
2854 		m_freem(tcp_saveti);
2855 
2856 	if (tp->t_state == TCPS_TIME_WAIT
2857 	    && (so->so_state & SS_NOFDREF)
2858 	    && (tp->t_inpcb || af != AF_INET || af != AF_INET6)
2859 	    && ((af == AF_INET ? tcp4_vtw_enable : tcp6_vtw_enable) & 1) != 0
2860 	    && TAILQ_EMPTY(&tp->segq)
2861 	    && vtw_add(af, tp)) {
2862 		;
2863 	}
2864 	return;
2865 
2866 badsyn:
2867 	/*
2868 	 * Received a bad SYN.  Increment counters and dropwithreset.
2869 	 */
2870 	TCP_STATINC(TCP_STAT_BADSYN);
2871 	tp = NULL;
2872 	goto dropwithreset;
2873 
2874 dropafterack:
2875 	/*
2876 	 * Generate an ACK dropping incoming segment if it occupies
2877 	 * sequence space, where the ACK reflects our state.
2878 	 */
2879 	if (tiflags & TH_RST)
2880 		goto drop;
2881 	goto dropafterack2;
2882 
2883 dropafterack_ratelim:
2884 	/*
2885 	 * We may want to rate-limit ACKs against SYN/RST attack.
2886 	 */
2887 	if (ppsratecheck(&tcp_ackdrop_ppslim_last, &tcp_ackdrop_ppslim_count,
2888 	    tcp_ackdrop_ppslim) == 0) {
2889 		/* XXX stat */
2890 		goto drop;
2891 	}
2892 
2893 dropafterack2:
2894 	m_freem(m);
2895 	tp->t_flags |= TF_ACKNOW;
2896 	KERNEL_LOCK(1, NULL);
2897 	(void)tcp_output(tp);
2898 	KERNEL_UNLOCK_ONE(NULL);
2899 	if (tcp_saveti)
2900 		m_freem(tcp_saveti);
2901 	return;
2902 
2903 dropwithreset_ratelim:
2904 	/*
2905 	 * We may want to rate-limit RSTs in certain situations,
2906 	 * particularly if we are sending an RST in response to
2907 	 * an attempt to connect to or otherwise communicate with
2908 	 * a port for which we have no socket.
2909 	 */
2910 	if (ppsratecheck(&tcp_rst_ppslim_last, &tcp_rst_ppslim_count,
2911 	    tcp_rst_ppslim) == 0) {
2912 		/* XXX stat */
2913 		goto drop;
2914 	}
2915 
2916 dropwithreset:
2917 	/*
2918 	 * Generate a RST, dropping incoming segment.
2919 	 * Make ACK acceptable to originator of segment.
2920 	 */
2921 	if (tiflags & TH_RST)
2922 		goto drop;
2923 	if (tiflags & TH_ACK) {
2924 		(void)tcp_respond(tp, m, m, th, (tcp_seq)0, th->th_ack, TH_RST);
2925 	} else {
2926 		if (tiflags & TH_SYN)
2927 			tlen++;
2928 		(void)tcp_respond(tp, m, m, th, th->th_seq + tlen, (tcp_seq)0,
2929 		    TH_RST|TH_ACK);
2930 	}
2931 	if (tcp_saveti)
2932 		m_freem(tcp_saveti);
2933 	return;
2934 
2935 badcsum:
2936 drop:
2937 	/*
2938 	 * Drop space held by incoming segment and return.
2939 	 */
2940 	if (tp) {
2941 		so = tp->t_inpcb->inp_socket;
2942 #ifdef TCP_DEBUG
2943 		if (so && (so->so_options & SO_DEBUG) != 0)
2944 			tcp_trace(TA_DROP, ostate, tp, tcp_saveti, 0);
2945 #endif
2946 	}
2947 	if (tcp_saveti)
2948 		m_freem(tcp_saveti);
2949 	m_freem(m);
2950 	return;
2951 }
2952 
2953 #ifdef TCP_SIGNATURE
2954 int
2955 tcp_signature_apply(void *fstate, void *data, u_int len)
2956 {
2957 
2958 	MD5Update(fstate, (u_char *)data, len);
2959 	return (0);
2960 }
2961 
2962 struct secasvar *
2963 tcp_signature_getsav(struct mbuf *m)
2964 {
2965 	struct ip *ip;
2966 	struct ip6_hdr *ip6;
2967 
2968 	ip = mtod(m, struct ip *);
2969 	switch (ip->ip_v) {
2970 	case 4:
2971 		ip = mtod(m, struct ip *);
2972 		ip6 = NULL;
2973 		break;
2974 	case 6:
2975 		ip = NULL;
2976 		ip6 = mtod(m, struct ip6_hdr *);
2977 		break;
2978 	default:
2979 		return (NULL);
2980 	}
2981 
2982 #ifdef IPSEC
2983 	union sockaddr_union dst;
2984 
2985 	/* Extract the destination from the IP header in the mbuf. */
2986 	memset(&dst, 0, sizeof(union sockaddr_union));
2987 	if (ip != NULL) {
2988 		dst.sa.sa_len = sizeof(struct sockaddr_in);
2989 		dst.sa.sa_family = AF_INET;
2990 		dst.sin.sin_addr = ip->ip_dst;
2991 	} else {
2992 		dst.sa.sa_len = sizeof(struct sockaddr_in6);
2993 		dst.sa.sa_family = AF_INET6;
2994 		dst.sin6.sin6_addr = ip6->ip6_dst;
2995 	}
2996 
2997 	/*
2998 	 * Look up an SADB entry which matches the address of the peer.
2999 	 */
3000 	return KEY_LOOKUP_SA(&dst, IPPROTO_TCP, htonl(TCP_SIG_SPI), 0, 0);
3001 #else
3002 	return NULL;
3003 #endif
3004 }
3005 
3006 int
3007 tcp_signature(struct mbuf *m, struct tcphdr *th, int thoff,
3008     struct secasvar *sav, char *sig)
3009 {
3010 	MD5_CTX ctx;
3011 	struct ip *ip;
3012 	struct ipovly *ipovly;
3013 #ifdef INET6
3014 	struct ip6_hdr *ip6;
3015 	struct ip6_hdr_pseudo ip6pseudo;
3016 #endif
3017 	struct ippseudo ippseudo;
3018 	struct tcphdr th0;
3019 	int l, tcphdrlen;
3020 
3021 	if (sav == NULL)
3022 		return (-1);
3023 
3024 	tcphdrlen = th->th_off * 4;
3025 
3026 	switch (mtod(m, struct ip *)->ip_v) {
3027 	case 4:
3028 		MD5Init(&ctx);
3029 		ip = mtod(m, struct ip *);
3030 		memset(&ippseudo, 0, sizeof(ippseudo));
3031 		ipovly = (struct ipovly *)ip;
3032 		ippseudo.ippseudo_src = ipovly->ih_src;
3033 		ippseudo.ippseudo_dst = ipovly->ih_dst;
3034 		ippseudo.ippseudo_pad = 0;
3035 		ippseudo.ippseudo_p = IPPROTO_TCP;
3036 		ippseudo.ippseudo_len = htons(m->m_pkthdr.len - thoff);
3037 		MD5Update(&ctx, (char *)&ippseudo, sizeof(ippseudo));
3038 		break;
3039 #if INET6
3040 	case 6:
3041 		MD5Init(&ctx);
3042 		ip6 = mtod(m, struct ip6_hdr *);
3043 		memset(&ip6pseudo, 0, sizeof(ip6pseudo));
3044 		ip6pseudo.ip6ph_src = ip6->ip6_src;
3045 		in6_clearscope(&ip6pseudo.ip6ph_src);
3046 		ip6pseudo.ip6ph_dst = ip6->ip6_dst;
3047 		in6_clearscope(&ip6pseudo.ip6ph_dst);
3048 		ip6pseudo.ip6ph_len = htons(m->m_pkthdr.len - thoff);
3049 		ip6pseudo.ip6ph_nxt = IPPROTO_TCP;
3050 		MD5Update(&ctx, (char *)&ip6pseudo, sizeof(ip6pseudo));
3051 		break;
3052 #endif
3053 	default:
3054 		return (-1);
3055 	}
3056 
3057 	th0 = *th;
3058 	th0.th_sum = 0;
3059 	MD5Update(&ctx, (char *)&th0, sizeof(th0));
3060 
3061 	l = m->m_pkthdr.len - thoff - tcphdrlen;
3062 	if (l > 0)
3063 		m_apply(m, thoff + tcphdrlen,
3064 		    m->m_pkthdr.len - thoff - tcphdrlen,
3065 		    tcp_signature_apply, &ctx);
3066 
3067 	MD5Update(&ctx, _KEYBUF(sav->key_auth), _KEYLEN(sav->key_auth));
3068 	MD5Final(sig, &ctx);
3069 
3070 	return (0);
3071 }
3072 #endif
3073 
3074 /*
3075  * Parse and process tcp options.
3076  *
3077  * Returns -1 if this segment should be dropped.  (eg. wrong signature)
3078  * Otherwise returns 0.
3079  */
3080 int
3081 tcp_dooptions(struct tcpcb *tp, const u_char *cp, int cnt, struct tcphdr *th,
3082     struct mbuf *m, int toff, struct tcp_opt_info *oi)
3083 {
3084 	u_int16_t mss;
3085 	int opt, optlen = 0;
3086 #ifdef TCP_SIGNATURE
3087 	void *sigp = NULL;
3088 	char sigbuf[TCP_SIGLEN];
3089 	struct secasvar *sav = NULL;
3090 #endif
3091 
3092 	for (; cp && cnt > 0; cnt -= optlen, cp += optlen) {
3093 		opt = cp[0];
3094 		if (opt == TCPOPT_EOL)
3095 			break;
3096 		if (opt == TCPOPT_NOP)
3097 			optlen = 1;
3098 		else {
3099 			if (cnt < 2)
3100 				break;
3101 			optlen = cp[1];
3102 			if (optlen < 2 || optlen > cnt)
3103 				break;
3104 		}
3105 		switch (opt) {
3106 
3107 		default:
3108 			continue;
3109 
3110 		case TCPOPT_MAXSEG:
3111 			if (optlen != TCPOLEN_MAXSEG)
3112 				continue;
3113 			if (!(th->th_flags & TH_SYN))
3114 				continue;
3115 			if (TCPS_HAVERCVDSYN(tp->t_state))
3116 				continue;
3117 			memcpy(&mss, cp + 2, sizeof(mss));
3118 			oi->maxseg = ntohs(mss);
3119 			break;
3120 
3121 		case TCPOPT_WINDOW:
3122 			if (optlen != TCPOLEN_WINDOW)
3123 				continue;
3124 			if (!(th->th_flags & TH_SYN))
3125 				continue;
3126 			if (TCPS_HAVERCVDSYN(tp->t_state))
3127 				continue;
3128 			tp->t_flags |= TF_RCVD_SCALE;
3129 			tp->requested_s_scale = cp[2];
3130 			if (tp->requested_s_scale > TCP_MAX_WINSHIFT) {
3131 				char buf[INET6_ADDRSTRLEN];
3132 				struct ip *ip = mtod(m, struct ip *);
3133 #ifdef INET6
3134 				struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
3135 #endif
3136 
3137 				switch (ip->ip_v) {
3138 				case 4:
3139 					in_print(buf, sizeof(buf),
3140 					    &ip->ip_src);
3141 					break;
3142 #ifdef INET6
3143 				case 6:
3144 					in6_print(buf, sizeof(buf),
3145 					    &ip6->ip6_src);
3146 					break;
3147 #endif
3148 				default:
3149 					strlcpy(buf, "(unknown)", sizeof(buf));
3150 					break;
3151 				}
3152 
3153 				log(LOG_ERR, "TCP: invalid wscale %d from %s, "
3154 				    "assuming %d\n",
3155 				    tp->requested_s_scale, buf,
3156 				    TCP_MAX_WINSHIFT);
3157 				tp->requested_s_scale = TCP_MAX_WINSHIFT;
3158 			}
3159 			break;
3160 
3161 		case TCPOPT_TIMESTAMP:
3162 			if (optlen != TCPOLEN_TIMESTAMP)
3163 				continue;
3164 			oi->ts_present = 1;
3165 			memcpy(&oi->ts_val, cp + 2, sizeof(oi->ts_val));
3166 			NTOHL(oi->ts_val);
3167 			memcpy(&oi->ts_ecr, cp + 6, sizeof(oi->ts_ecr));
3168 			NTOHL(oi->ts_ecr);
3169 
3170 			if (!(th->th_flags & TH_SYN))
3171 				continue;
3172 			if (TCPS_HAVERCVDSYN(tp->t_state))
3173 				continue;
3174 			/*
3175 			 * A timestamp received in a SYN makes
3176 			 * it ok to send timestamp requests and replies.
3177 			 */
3178 			tp->t_flags |= TF_RCVD_TSTMP;
3179 			tp->ts_recent = oi->ts_val;
3180 			tp->ts_recent_age = tcp_now;
3181                         break;
3182 
3183 		case TCPOPT_SACK_PERMITTED:
3184 			if (optlen != TCPOLEN_SACK_PERMITTED)
3185 				continue;
3186 			if (!(th->th_flags & TH_SYN))
3187 				continue;
3188 			if (TCPS_HAVERCVDSYN(tp->t_state))
3189 				continue;
3190 			if (tcp_do_sack) {
3191 				tp->t_flags |= TF_SACK_PERMIT;
3192 				tp->t_flags |= TF_WILL_SACK;
3193 			}
3194 			break;
3195 
3196 		case TCPOPT_SACK:
3197 			tcp_sack_option(tp, th, cp, optlen);
3198 			break;
3199 #ifdef TCP_SIGNATURE
3200 		case TCPOPT_SIGNATURE:
3201 			if (optlen != TCPOLEN_SIGNATURE)
3202 				continue;
3203 			if (sigp &&
3204 			    !consttime_memequal(sigp, cp + 2, TCP_SIGLEN))
3205 				return (-1);
3206 
3207 			sigp = sigbuf;
3208 			memcpy(sigbuf, cp + 2, TCP_SIGLEN);
3209 			tp->t_flags |= TF_SIGNATURE;
3210 			break;
3211 #endif
3212 		}
3213 	}
3214 
3215 #ifndef TCP_SIGNATURE
3216 	return 0;
3217 #else
3218 	if (tp->t_flags & TF_SIGNATURE) {
3219 		sav = tcp_signature_getsav(m);
3220 		if (sav == NULL && tp->t_state == TCPS_LISTEN)
3221 			return (-1);
3222 	}
3223 
3224 	if ((sigp ? TF_SIGNATURE : 0) ^ (tp->t_flags & TF_SIGNATURE))
3225 		goto out;
3226 
3227 	if (sigp) {
3228 		char sig[TCP_SIGLEN];
3229 
3230 		tcp_fields_to_net(th);
3231 		if (tcp_signature(m, th, toff, sav, sig) < 0) {
3232 			tcp_fields_to_host(th);
3233 			goto out;
3234 		}
3235 		tcp_fields_to_host(th);
3236 
3237 		if (!consttime_memequal(sig, sigp, TCP_SIGLEN)) {
3238 			TCP_STATINC(TCP_STAT_BADSIG);
3239 			goto out;
3240 		} else
3241 			TCP_STATINC(TCP_STAT_GOODSIG);
3242 
3243 		key_sa_recordxfer(sav, m);
3244 		KEY_SA_UNREF(&sav);
3245 	}
3246 	return 0;
3247 out:
3248 	if (sav != NULL)
3249 		KEY_SA_UNREF(&sav);
3250 	return -1;
3251 #endif
3252 }
3253 
3254 /*
3255  * Pull out of band byte out of a segment so
3256  * it doesn't appear in the user's data queue.
3257  * It is still reflected in the segment length for
3258  * sequencing purposes.
3259  */
3260 void
3261 tcp_pulloutofband(struct socket *so, struct tcphdr *th,
3262     struct mbuf *m, int off)
3263 {
3264 	int cnt = off + th->th_urp - 1;
3265 
3266 	while (cnt >= 0) {
3267 		if (m->m_len > cnt) {
3268 			char *cp = mtod(m, char *) + cnt;
3269 			struct tcpcb *tp = sototcpcb(so);
3270 
3271 			tp->t_iobc = *cp;
3272 			tp->t_oobflags |= TCPOOB_HAVEDATA;
3273 			memmove(cp, cp + 1, (unsigned)(m->m_len - cnt - 1));
3274 			m->m_len--;
3275 			return;
3276 		}
3277 		cnt -= m->m_len;
3278 		m = m->m_next;
3279 		if (m == NULL)
3280 			break;
3281 	}
3282 	panic("tcp_pulloutofband");
3283 }
3284 
3285 /*
3286  * Collect new round-trip time estimate
3287  * and update averages and current timeout.
3288  *
3289  * rtt is in units of slow ticks (typically 500 ms) -- essentially the
3290  * difference of two timestamps.
3291  */
3292 void
3293 tcp_xmit_timer(struct tcpcb *tp, uint32_t rtt)
3294 {
3295 	int32_t delta;
3296 
3297 	TCP_STATINC(TCP_STAT_RTTUPDATED);
3298 	if (tp->t_srtt != 0) {
3299 		/*
3300 		 * Compute the amount to add to srtt for smoothing,
3301 		 * *alpha, or 2^(-TCP_RTT_SHIFT).  Because
3302 		 * srtt is stored in 1/32 slow ticks, we conceptually
3303 		 * shift left 5 bits, subtract srtt to get the
3304 		 * difference, and then shift right by TCP_RTT_SHIFT
3305 		 * (3) to obtain 1/8 of the difference.
3306 		 */
3307 		delta = (rtt << 2) - (tp->t_srtt >> TCP_RTT_SHIFT);
3308 		/*
3309 		 * This can never happen, because delta's lowest
3310 		 * possible value is 1/8 of t_srtt.  But if it does,
3311 		 * set srtt to some reasonable value, here chosen
3312 		 * as 1/8 tick.
3313 		 */
3314 		if ((tp->t_srtt += delta) <= 0)
3315 			tp->t_srtt = 1 << 2;
3316 		/*
3317 		 * RFC2988 requires that rttvar be updated first.
3318 		 * This code is compliant because "delta" is the old
3319 		 * srtt minus the new observation (scaled).
3320 		 *
3321 		 * RFC2988 says:
3322 		 *   rttvar = (1-beta) * rttvar + beta * |srtt-observed|
3323 		 *
3324 		 * delta is in units of 1/32 ticks, and has then been
3325 		 * divided by 8.  This is equivalent to being in 1/16s
3326 		 * units and divided by 4.  Subtract from it 1/4 of
3327 		 * the existing rttvar to form the (signed) amount to
3328 		 * adjust.
3329 		 */
3330 		if (delta < 0)
3331 			delta = -delta;
3332 		delta -= (tp->t_rttvar >> TCP_RTTVAR_SHIFT);
3333 		/*
3334 		 * As with srtt, this should never happen.  There is
3335 		 * no support in RFC2988 for this operation.  But 1/4s
3336 		 * as rttvar when faced with something arguably wrong
3337 		 * is ok.
3338 		 */
3339 		if ((tp->t_rttvar += delta) <= 0)
3340 			tp->t_rttvar = 1 << 2;
3341 
3342 		/*
3343 		 * If srtt exceeds .01 second, ensure we use the 'remote' MSL
3344 		 * Problem is: it doesn't work.  Disabled by defaulting
3345 		 * tcp_rttlocal to 0; see corresponding code in
3346 		 * tcp_subr that selects local vs remote in a different way.
3347 		 *
3348 		 * The static branch prediction hint here should be removed
3349 		 * when the rtt estimator is fixed and the rtt_enable code
3350 		 * is turned back on.
3351 		 */
3352 		if (__predict_false(tcp_rttlocal) && tcp_msl_enable
3353 		    && tp->t_srtt > tcp_msl_remote_threshold
3354 		    && tp->t_msl  < tcp_msl_remote) {
3355 			tp->t_msl = MIN(tcp_msl_remote, TCP_MAXMSL);
3356 		}
3357 	} else {
3358 		/*
3359 		 * This is the first measurement.  Per RFC2988, 2.2,
3360 		 * set rtt=R and srtt=R/2.
3361 		 * For srtt, storage representation is 1/32 ticks,
3362 		 * so shift left by 5.
3363 		 * For rttvar, storage representation is 1/16 ticks,
3364 		 * So shift left by 4, but then right by 1 to halve.
3365 		 */
3366 		tp->t_srtt = rtt << (TCP_RTT_SHIFT + 2);
3367 		tp->t_rttvar = rtt << (TCP_RTTVAR_SHIFT + 2 - 1);
3368 	}
3369 	tp->t_rtttime = 0;
3370 	tp->t_rxtshift = 0;
3371 
3372 	/*
3373 	 * the retransmit should happen at rtt + 4 * rttvar.
3374 	 * Because of the way we do the smoothing, srtt and rttvar
3375 	 * will each average +1/2 tick of bias.  When we compute
3376 	 * the retransmit timer, we want 1/2 tick of rounding and
3377 	 * 1 extra tick because of +-1/2 tick uncertainty in the
3378 	 * firing of the timer.  The bias will give us exactly the
3379 	 * 1.5 tick we need.  But, because the bias is
3380 	 * statistical, we have to test that we don't drop below
3381 	 * the minimum feasible timer (which is 2 ticks).
3382 	 */
3383 	TCPT_RANGESET(tp->t_rxtcur, TCP_REXMTVAL(tp),
3384 	    uimax(tp->t_rttmin, rtt + 2), TCPTV_REXMTMAX);
3385 
3386 	/*
3387 	 * We received an ack for a packet that wasn't retransmitted;
3388 	 * it is probably safe to discard any error indications we've
3389 	 * received recently.  This isn't quite right, but close enough
3390 	 * for now (a route might have failed after we sent a segment,
3391 	 * and the return path might not be symmetrical).
3392 	 */
3393 	tp->t_softerror = 0;
3394 }
3395