xref: /netbsd-src/sys/netinet/tcp_input.c (revision aad9773e38ed2370a628a6416e098f9008fc10a7)
1 /*	$NetBSD: tcp_input.c,v 1.335 2014/12/02 20:25:47 christos 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 /*
141  *	TODO list for SYN cache stuff:
142  *
143  *	Find room for a "state" field, which is needed to keep a
144  *	compressed state for TIME_WAIT TCBs.  It's been noted already
145  *	that this is fairly important for very high-volume web and
146  *	mail servers, which use a large number of short-lived
147  *	connections.
148  */
149 
150 #include <sys/cdefs.h>
151 __KERNEL_RCSID(0, "$NetBSD: tcp_input.c,v 1.335 2014/12/02 20:25:47 christos Exp $");
152 
153 #include "opt_inet.h"
154 #include "opt_ipsec.h"
155 #include "opt_inet_csum.h"
156 #include "opt_tcp_debug.h"
157 
158 #include <sys/param.h>
159 #include <sys/systm.h>
160 #include <sys/malloc.h>
161 #include <sys/mbuf.h>
162 #include <sys/protosw.h>
163 #include <sys/socket.h>
164 #include <sys/socketvar.h>
165 #include <sys/errno.h>
166 #include <sys/syslog.h>
167 #include <sys/pool.h>
168 #include <sys/domain.h>
169 #include <sys/kernel.h>
170 #ifdef TCP_SIGNATURE
171 #include <sys/md5.h>
172 #endif
173 #include <sys/lwp.h> /* for lwp0 */
174 #include <sys/cprng.h>
175 
176 #include <net/if.h>
177 #include <net/route.h>
178 #include <net/if_types.h>
179 
180 #include <netinet/in.h>
181 #include <netinet/in_systm.h>
182 #include <netinet/ip.h>
183 #include <netinet/in_pcb.h>
184 #include <netinet/in_var.h>
185 #include <netinet/ip_var.h>
186 #include <netinet/in_offload.h>
187 
188 #ifdef INET6
189 #ifndef INET
190 #include <netinet/in.h>
191 #endif
192 #include <netinet/ip6.h>
193 #include <netinet6/ip6_var.h>
194 #include <netinet6/in6_pcb.h>
195 #include <netinet6/ip6_var.h>
196 #include <netinet6/in6_var.h>
197 #include <netinet/icmp6.h>
198 #include <netinet6/nd6.h>
199 #ifdef TCP_SIGNATURE
200 #include <netinet6/scope6_var.h>
201 #endif
202 #endif
203 
204 #ifndef INET6
205 /* always need ip6.h for IP6_EXTHDR_GET */
206 #include <netinet/ip6.h>
207 #endif
208 
209 #include <netinet/tcp.h>
210 #include <netinet/tcp_fsm.h>
211 #include <netinet/tcp_seq.h>
212 #include <netinet/tcp_timer.h>
213 #include <netinet/tcp_var.h>
214 #include <netinet/tcp_private.h>
215 #include <netinet/tcpip.h>
216 #include <netinet/tcp_congctl.h>
217 #include <netinet/tcp_debug.h>
218 
219 #ifdef INET6
220 #include "faith.h"
221 #if defined(NFAITH) && NFAITH > 0
222 #include <net/if_faith.h>
223 #endif
224 #endif	/* INET6 */
225 
226 #ifdef IPSEC
227 #include <netipsec/ipsec.h>
228 #include <netipsec/ipsec_var.h>
229 #include <netipsec/ipsec_private.h>
230 #include <netipsec/key.h>
231 #ifdef INET6
232 #include <netipsec/ipsec6.h>
233 #endif
234 #endif	/* IPSEC*/
235 
236 #include <netinet/tcp_vtw.h>
237 
238 int	tcprexmtthresh = 3;
239 int	tcp_log_refused;
240 
241 int	tcp_do_autorcvbuf = 1;
242 int	tcp_autorcvbuf_inc = 16 * 1024;
243 int	tcp_autorcvbuf_max = 256 * 1024;
244 int	tcp_msl = (TCPTV_MSL / PR_SLOWHZ);
245 
246 static int tcp_rst_ppslim_count = 0;
247 static struct timeval tcp_rst_ppslim_last;
248 static int tcp_ackdrop_ppslim_count = 0;
249 static struct timeval tcp_ackdrop_ppslim_last;
250 
251 #define TCP_PAWS_IDLE	(24U * 24 * 60 * 60 * PR_SLOWHZ)
252 
253 /* for modulo comparisons of timestamps */
254 #define TSTMP_LT(a,b)	((int)((a)-(b)) < 0)
255 #define TSTMP_GEQ(a,b)	((int)((a)-(b)) >= 0)
256 
257 /*
258  * Neighbor Discovery, Neighbor Unreachability Detection Upper layer hint.
259  */
260 #ifdef INET6
261 static inline void
262 nd6_hint(struct tcpcb *tp)
263 {
264 	struct rtentry *rt;
265 
266 	if (tp != NULL && tp->t_in6pcb != NULL && tp->t_family == AF_INET6 &&
267 	    (rt = rtcache_validate(&tp->t_in6pcb->in6p_route)) != NULL)
268 		nd6_nud_hint(rt, NULL, 0);
269 }
270 #else
271 static inline void
272 nd6_hint(struct tcpcb *tp)
273 {
274 }
275 #endif
276 
277 /*
278  * Compute ACK transmission behavior.  Delay the ACK unless
279  * we have already delayed an ACK (must send an ACK every two segments).
280  * We also ACK immediately if we received a PUSH and the ACK-on-PUSH
281  * option is enabled.
282  */
283 static void
284 tcp_setup_ack(struct tcpcb *tp, const struct tcphdr *th)
285 {
286 
287 	if (tp->t_flags & TF_DELACK ||
288 	    (tcp_ack_on_push && th->th_flags & TH_PUSH))
289 		tp->t_flags |= TF_ACKNOW;
290 	else
291 		TCP_SET_DELACK(tp);
292 }
293 
294 static void
295 icmp_check(struct tcpcb *tp, const struct tcphdr *th, int acked)
296 {
297 
298 	/*
299 	 * If we had a pending ICMP message that refers to data that have
300 	 * just been acknowledged, disregard the recorded ICMP message.
301 	 */
302 	if ((tp->t_flags & TF_PMTUD_PEND) &&
303 	    SEQ_GT(th->th_ack, tp->t_pmtud_th_seq))
304 		tp->t_flags &= ~TF_PMTUD_PEND;
305 
306 	/*
307 	 * Keep track of the largest chunk of data
308 	 * acknowledged since last PMTU update
309 	 */
310 	if (tp->t_pmtud_mss_acked < acked)
311 		tp->t_pmtud_mss_acked = acked;
312 }
313 
314 /*
315  * Convert TCP protocol fields to host order for easier processing.
316  */
317 static void
318 tcp_fields_to_host(struct tcphdr *th)
319 {
320 
321 	NTOHL(th->th_seq);
322 	NTOHL(th->th_ack);
323 	NTOHS(th->th_win);
324 	NTOHS(th->th_urp);
325 }
326 
327 /*
328  * ... and reverse the above.
329  */
330 static void
331 tcp_fields_to_net(struct tcphdr *th)
332 {
333 
334 	HTONL(th->th_seq);
335 	HTONL(th->th_ack);
336 	HTONS(th->th_win);
337 	HTONS(th->th_urp);
338 }
339 
340 #ifdef TCP_CSUM_COUNTERS
341 #include <sys/device.h>
342 
343 #if defined(INET)
344 extern struct evcnt tcp_hwcsum_ok;
345 extern struct evcnt tcp_hwcsum_bad;
346 extern struct evcnt tcp_hwcsum_data;
347 extern struct evcnt tcp_swcsum;
348 #endif /* defined(INET) */
349 #if defined(INET6)
350 extern struct evcnt tcp6_hwcsum_ok;
351 extern struct evcnt tcp6_hwcsum_bad;
352 extern struct evcnt tcp6_hwcsum_data;
353 extern struct evcnt tcp6_swcsum;
354 #endif /* defined(INET6) */
355 
356 #define	TCP_CSUM_COUNTER_INCR(ev)	(ev)->ev_count++
357 
358 #else
359 
360 #define	TCP_CSUM_COUNTER_INCR(ev)	/* nothing */
361 
362 #endif /* TCP_CSUM_COUNTERS */
363 
364 #ifdef TCP_REASS_COUNTERS
365 #include <sys/device.h>
366 
367 extern struct evcnt tcp_reass_;
368 extern struct evcnt tcp_reass_empty;
369 extern struct evcnt tcp_reass_iteration[8];
370 extern struct evcnt tcp_reass_prependfirst;
371 extern struct evcnt tcp_reass_prepend;
372 extern struct evcnt tcp_reass_insert;
373 extern struct evcnt tcp_reass_inserttail;
374 extern struct evcnt tcp_reass_append;
375 extern struct evcnt tcp_reass_appendtail;
376 extern struct evcnt tcp_reass_overlaptail;
377 extern struct evcnt tcp_reass_overlapfront;
378 extern struct evcnt tcp_reass_segdup;
379 extern struct evcnt tcp_reass_fragdup;
380 
381 #define	TCP_REASS_COUNTER_INCR(ev)	(ev)->ev_count++
382 
383 #else
384 
385 #define	TCP_REASS_COUNTER_INCR(ev)	/* nothing */
386 
387 #endif /* TCP_REASS_COUNTERS */
388 
389 static int tcp_reass(struct tcpcb *, const struct tcphdr *, struct mbuf *,
390     int *);
391 static int tcp_dooptions(struct tcpcb *, const u_char *, int,
392     struct tcphdr *, struct mbuf *, int, struct tcp_opt_info *);
393 
394 #ifdef INET
395 static void tcp4_log_refused(const struct ip *, const struct tcphdr *);
396 #endif
397 #ifdef INET6
398 static void tcp6_log_refused(const struct ip6_hdr *, const struct tcphdr *);
399 #endif
400 
401 #define	TRAVERSE(x) while ((x)->m_next) (x) = (x)->m_next
402 
403 #if defined(MBUFTRACE)
404 struct mowner tcp_reass_mowner = MOWNER_INIT("tcp", "reass");
405 #endif /* defined(MBUFTRACE) */
406 
407 static struct pool tcpipqent_pool;
408 
409 void
410 tcpipqent_init(void)
411 {
412 
413 	pool_init(&tcpipqent_pool, sizeof(struct ipqent), 0, 0, 0, "tcpipqepl",
414 	    NULL, IPL_VM);
415 }
416 
417 struct ipqent *
418 tcpipqent_alloc(void)
419 {
420 	struct ipqent *ipqe;
421 	int s;
422 
423 	s = splvm();
424 	ipqe = pool_get(&tcpipqent_pool, PR_NOWAIT);
425 	splx(s);
426 
427 	return ipqe;
428 }
429 
430 void
431 tcpipqent_free(struct ipqent *ipqe)
432 {
433 	int s;
434 
435 	s = splvm();
436 	pool_put(&tcpipqent_pool, ipqe);
437 	splx(s);
438 }
439 
440 static int
441 tcp_reass(struct tcpcb *tp, const struct tcphdr *th, struct mbuf *m, int *tlen)
442 {
443 	struct ipqent *p, *q, *nq, *tiqe = NULL;
444 	struct socket *so = NULL;
445 	int pkt_flags;
446 	tcp_seq pkt_seq;
447 	unsigned pkt_len;
448 	u_long rcvpartdupbyte = 0;
449 	u_long rcvoobyte;
450 #ifdef TCP_REASS_COUNTERS
451 	u_int count = 0;
452 #endif
453 	uint64_t *tcps;
454 
455 	if (tp->t_inpcb)
456 		so = tp->t_inpcb->inp_socket;
457 #ifdef INET6
458 	else if (tp->t_in6pcb)
459 		so = tp->t_in6pcb->in6p_socket;
460 #endif
461 
462 	TCP_REASS_LOCK_CHECK(tp);
463 
464 	/*
465 	 * Call with th==0 after become established to
466 	 * force pre-ESTABLISHED data up to user socket.
467 	 */
468 	if (th == 0)
469 		goto present;
470 
471 	m_claimm(m, &tcp_reass_mowner);
472 
473 	rcvoobyte = *tlen;
474 	/*
475 	 * Copy these to local variables because the tcpiphdr
476 	 * gets munged while we are collapsing mbufs.
477 	 */
478 	pkt_seq = th->th_seq;
479 	pkt_len = *tlen;
480 	pkt_flags = th->th_flags;
481 
482 	TCP_REASS_COUNTER_INCR(&tcp_reass_);
483 
484 	if ((p = TAILQ_LAST(&tp->segq, ipqehead)) != NULL) {
485 		/*
486 		 * When we miss a packet, the vast majority of time we get
487 		 * packets that follow it in order.  So optimize for that.
488 		 */
489 		if (pkt_seq == p->ipqe_seq + p->ipqe_len) {
490 			p->ipqe_len += pkt_len;
491 			p->ipqe_flags |= pkt_flags;
492 			m_cat(p->ipre_mlast, m);
493 			TRAVERSE(p->ipre_mlast);
494 			m = NULL;
495 			tiqe = p;
496 			TAILQ_REMOVE(&tp->timeq, p, ipqe_timeq);
497 			TCP_REASS_COUNTER_INCR(&tcp_reass_appendtail);
498 			goto skip_replacement;
499 		}
500 		/*
501 		 * While we're here, if the pkt is completely beyond
502 		 * anything we have, just insert it at the tail.
503 		 */
504 		if (SEQ_GT(pkt_seq, p->ipqe_seq + p->ipqe_len)) {
505 			TCP_REASS_COUNTER_INCR(&tcp_reass_inserttail);
506 			goto insert_it;
507 		}
508 	}
509 
510 	q = TAILQ_FIRST(&tp->segq);
511 
512 	if (q != NULL) {
513 		/*
514 		 * If this segment immediately precedes the first out-of-order
515 		 * block, simply slap the segment in front of it and (mostly)
516 		 * skip the complicated logic.
517 		 */
518 		if (pkt_seq + pkt_len == q->ipqe_seq) {
519 			q->ipqe_seq = pkt_seq;
520 			q->ipqe_len += pkt_len;
521 			q->ipqe_flags |= pkt_flags;
522 			m_cat(m, q->ipqe_m);
523 			q->ipqe_m = m;
524 			q->ipre_mlast = m; /* last mbuf may have changed */
525 			TRAVERSE(q->ipre_mlast);
526 			tiqe = q;
527 			TAILQ_REMOVE(&tp->timeq, q, ipqe_timeq);
528 			TCP_REASS_COUNTER_INCR(&tcp_reass_prependfirst);
529 			goto skip_replacement;
530 		}
531 	} else {
532 		TCP_REASS_COUNTER_INCR(&tcp_reass_empty);
533 	}
534 
535 	/*
536 	 * Find a segment which begins after this one does.
537 	 */
538 	for (p = NULL; q != NULL; q = nq) {
539 		nq = TAILQ_NEXT(q, ipqe_q);
540 #ifdef TCP_REASS_COUNTERS
541 		count++;
542 #endif
543 		/*
544 		 * If the received segment is just right after this
545 		 * fragment, merge the two together and then check
546 		 * for further overlaps.
547 		 */
548 		if (q->ipqe_seq + q->ipqe_len == pkt_seq) {
549 #ifdef TCPREASS_DEBUG
550 			printf("tcp_reass[%p]: concat %u:%u(%u) to %u:%u(%u)\n",
551 			       tp, pkt_seq, pkt_seq + pkt_len, pkt_len,
552 			       q->ipqe_seq, q->ipqe_seq + q->ipqe_len, q->ipqe_len);
553 #endif
554 			pkt_len += q->ipqe_len;
555 			pkt_flags |= q->ipqe_flags;
556 			pkt_seq = q->ipqe_seq;
557 			m_cat(q->ipre_mlast, m);
558 			TRAVERSE(q->ipre_mlast);
559 			m = q->ipqe_m;
560 			TCP_REASS_COUNTER_INCR(&tcp_reass_append);
561 			goto free_ipqe;
562 		}
563 		/*
564 		 * If the received segment is completely past this
565 		 * fragment, we need to go the next fragment.
566 		 */
567 		if (SEQ_LT(q->ipqe_seq + q->ipqe_len, pkt_seq)) {
568 			p = q;
569 			continue;
570 		}
571 		/*
572 		 * If the fragment is past the received segment,
573 		 * it (or any following) can't be concatenated.
574 		 */
575 		if (SEQ_GT(q->ipqe_seq, pkt_seq + pkt_len)) {
576 			TCP_REASS_COUNTER_INCR(&tcp_reass_insert);
577 			break;
578 		}
579 
580 		/*
581 		 * We've received all the data in this segment before.
582 		 * mark it as a duplicate and return.
583 		 */
584 		if (SEQ_LEQ(q->ipqe_seq, pkt_seq) &&
585 		    SEQ_GEQ(q->ipqe_seq + q->ipqe_len, pkt_seq + pkt_len)) {
586 			tcps = TCP_STAT_GETREF();
587 			tcps[TCP_STAT_RCVDUPPACK]++;
588 			tcps[TCP_STAT_RCVDUPBYTE] += pkt_len;
589 			TCP_STAT_PUTREF();
590 			tcp_new_dsack(tp, pkt_seq, pkt_len);
591 			m_freem(m);
592 			if (tiqe != NULL) {
593 				tcpipqent_free(tiqe);
594 			}
595 			TCP_REASS_COUNTER_INCR(&tcp_reass_segdup);
596 			goto out;
597 		}
598 		/*
599 		 * Received segment completely overlaps this fragment
600 		 * so we drop the fragment (this keeps the temporal
601 		 * ordering of segments correct).
602 		 */
603 		if (SEQ_GEQ(q->ipqe_seq, pkt_seq) &&
604 		    SEQ_LEQ(q->ipqe_seq + q->ipqe_len, pkt_seq + pkt_len)) {
605 			rcvpartdupbyte += q->ipqe_len;
606 			m_freem(q->ipqe_m);
607 			TCP_REASS_COUNTER_INCR(&tcp_reass_fragdup);
608 			goto free_ipqe;
609 		}
610 		/*
611 		 * RX'ed segment extends past the end of the
612 		 * fragment.  Drop the overlapping bytes.  Then
613 		 * merge the fragment and segment then treat as
614 		 * a longer received packet.
615 		 */
616 		if (SEQ_LT(q->ipqe_seq, pkt_seq) &&
617 		    SEQ_GT(q->ipqe_seq + q->ipqe_len, pkt_seq))  {
618 			int overlap = q->ipqe_seq + q->ipqe_len - pkt_seq;
619 #ifdef TCPREASS_DEBUG
620 			printf("tcp_reass[%p]: trim starting %d bytes of %u:%u(%u)\n",
621 			       tp, overlap,
622 			       pkt_seq, pkt_seq + pkt_len, pkt_len);
623 #endif
624 			m_adj(m, overlap);
625 			rcvpartdupbyte += overlap;
626 			m_cat(q->ipre_mlast, m);
627 			TRAVERSE(q->ipre_mlast);
628 			m = q->ipqe_m;
629 			pkt_seq = q->ipqe_seq;
630 			pkt_len += q->ipqe_len - overlap;
631 			rcvoobyte -= overlap;
632 			TCP_REASS_COUNTER_INCR(&tcp_reass_overlaptail);
633 			goto free_ipqe;
634 		}
635 		/*
636 		 * RX'ed segment extends past the front of the
637 		 * fragment.  Drop the overlapping bytes on the
638 		 * received packet.  The packet will then be
639 		 * contatentated with this fragment a bit later.
640 		 */
641 		if (SEQ_GT(q->ipqe_seq, pkt_seq) &&
642 		    SEQ_LT(q->ipqe_seq, pkt_seq + pkt_len))  {
643 			int overlap = pkt_seq + pkt_len - q->ipqe_seq;
644 #ifdef TCPREASS_DEBUG
645 			printf("tcp_reass[%p]: trim trailing %d bytes of %u:%u(%u)\n",
646 			       tp, overlap,
647 			       pkt_seq, pkt_seq + pkt_len, pkt_len);
648 #endif
649 			m_adj(m, -overlap);
650 			pkt_len -= overlap;
651 			rcvpartdupbyte += overlap;
652 			TCP_REASS_COUNTER_INCR(&tcp_reass_overlapfront);
653 			rcvoobyte -= overlap;
654 		}
655 		/*
656 		 * If the received segment immediates precedes this
657 		 * fragment then tack the fragment onto this segment
658 		 * and reinsert the data.
659 		 */
660 		if (q->ipqe_seq == pkt_seq + pkt_len) {
661 #ifdef TCPREASS_DEBUG
662 			printf("tcp_reass[%p]: append %u:%u(%u) to %u:%u(%u)\n",
663 			       tp, q->ipqe_seq, q->ipqe_seq + q->ipqe_len, q->ipqe_len,
664 			       pkt_seq, pkt_seq + pkt_len, pkt_len);
665 #endif
666 			pkt_len += q->ipqe_len;
667 			pkt_flags |= q->ipqe_flags;
668 			m_cat(m, q->ipqe_m);
669 			TAILQ_REMOVE(&tp->segq, q, ipqe_q);
670 			TAILQ_REMOVE(&tp->timeq, q, ipqe_timeq);
671 			tp->t_segqlen--;
672 			KASSERT(tp->t_segqlen >= 0);
673 			KASSERT(tp->t_segqlen != 0 ||
674 			    (TAILQ_EMPTY(&tp->segq) &&
675 			    TAILQ_EMPTY(&tp->timeq)));
676 			if (tiqe == NULL) {
677 				tiqe = q;
678 			} else {
679 				tcpipqent_free(q);
680 			}
681 			TCP_REASS_COUNTER_INCR(&tcp_reass_prepend);
682 			break;
683 		}
684 		/*
685 		 * If the fragment is before the segment, remember it.
686 		 * When this loop is terminated, p will contain the
687 		 * pointer to fragment that is right before the received
688 		 * segment.
689 		 */
690 		if (SEQ_LEQ(q->ipqe_seq, pkt_seq))
691 			p = q;
692 
693 		continue;
694 
695 		/*
696 		 * This is a common operation.  It also will allow
697 		 * to save doing a malloc/free in most instances.
698 		 */
699 	  free_ipqe:
700 		TAILQ_REMOVE(&tp->segq, q, ipqe_q);
701 		TAILQ_REMOVE(&tp->timeq, q, ipqe_timeq);
702 		tp->t_segqlen--;
703 		KASSERT(tp->t_segqlen >= 0);
704 		KASSERT(tp->t_segqlen != 0 ||
705 		    (TAILQ_EMPTY(&tp->segq) && TAILQ_EMPTY(&tp->timeq)));
706 		if (tiqe == NULL) {
707 			tiqe = q;
708 		} else {
709 			tcpipqent_free(q);
710 		}
711 	}
712 
713 #ifdef TCP_REASS_COUNTERS
714 	if (count > 7)
715 		TCP_REASS_COUNTER_INCR(&tcp_reass_iteration[0]);
716 	else if (count > 0)
717 		TCP_REASS_COUNTER_INCR(&tcp_reass_iteration[count]);
718 #endif
719 
720     insert_it:
721 
722 	/*
723 	 * Allocate a new queue entry since the received segment did not
724 	 * collapse onto any other out-of-order block; thus we are allocating
725 	 * a new block.  If it had collapsed, tiqe would not be NULL and
726 	 * we would be reusing it.
727 	 * XXX If we can't, just drop the packet.  XXX
728 	 */
729 	if (tiqe == NULL) {
730 		tiqe = tcpipqent_alloc();
731 		if (tiqe == NULL) {
732 			TCP_STATINC(TCP_STAT_RCVMEMDROP);
733 			m_freem(m);
734 			goto out;
735 		}
736 	}
737 
738 	/*
739 	 * Update the counters.
740 	 */
741 	tcps = TCP_STAT_GETREF();
742 	tcps[TCP_STAT_RCVOOPACK]++;
743 	tcps[TCP_STAT_RCVOOBYTE] += rcvoobyte;
744 	if (rcvpartdupbyte) {
745 	    tcps[TCP_STAT_RCVPARTDUPPACK]++;
746 	    tcps[TCP_STAT_RCVPARTDUPBYTE] += rcvpartdupbyte;
747 	}
748 	TCP_STAT_PUTREF();
749 
750 	/*
751 	 * Insert the new fragment queue entry into both queues.
752 	 */
753 	tiqe->ipqe_m = m;
754 	tiqe->ipre_mlast = m;
755 	tiqe->ipqe_seq = pkt_seq;
756 	tiqe->ipqe_len = pkt_len;
757 	tiqe->ipqe_flags = pkt_flags;
758 	if (p == NULL) {
759 		TAILQ_INSERT_HEAD(&tp->segq, tiqe, ipqe_q);
760 #ifdef TCPREASS_DEBUG
761 		if (tiqe->ipqe_seq != tp->rcv_nxt)
762 			printf("tcp_reass[%p]: insert %u:%u(%u) at front\n",
763 			       tp, pkt_seq, pkt_seq + pkt_len, pkt_len);
764 #endif
765 	} else {
766 		TAILQ_INSERT_AFTER(&tp->segq, p, tiqe, ipqe_q);
767 #ifdef TCPREASS_DEBUG
768 		printf("tcp_reass[%p]: insert %u:%u(%u) after %u:%u(%u)\n",
769 		       tp, pkt_seq, pkt_seq + pkt_len, pkt_len,
770 		       p->ipqe_seq, p->ipqe_seq + p->ipqe_len, p->ipqe_len);
771 #endif
772 	}
773 	tp->t_segqlen++;
774 
775 skip_replacement:
776 
777 	TAILQ_INSERT_HEAD(&tp->timeq, tiqe, ipqe_timeq);
778 
779 present:
780 	/*
781 	 * Present data to user, advancing rcv_nxt through
782 	 * completed sequence space.
783 	 */
784 	if (TCPS_HAVEESTABLISHED(tp->t_state) == 0)
785 		goto out;
786 	q = TAILQ_FIRST(&tp->segq);
787 	if (q == NULL || q->ipqe_seq != tp->rcv_nxt)
788 		goto out;
789 	if (tp->t_state == TCPS_SYN_RECEIVED && q->ipqe_len)
790 		goto out;
791 
792 	tp->rcv_nxt += q->ipqe_len;
793 	pkt_flags = q->ipqe_flags & TH_FIN;
794 	nd6_hint(tp);
795 
796 	TAILQ_REMOVE(&tp->segq, q, ipqe_q);
797 	TAILQ_REMOVE(&tp->timeq, q, ipqe_timeq);
798 	tp->t_segqlen--;
799 	KASSERT(tp->t_segqlen >= 0);
800 	KASSERT(tp->t_segqlen != 0 ||
801 	    (TAILQ_EMPTY(&tp->segq) && TAILQ_EMPTY(&tp->timeq)));
802 	if (so->so_state & SS_CANTRCVMORE)
803 		m_freem(q->ipqe_m);
804 	else
805 		sbappendstream(&so->so_rcv, q->ipqe_m);
806 	tcpipqent_free(q);
807 	TCP_REASS_UNLOCK(tp);
808 	sorwakeup(so);
809 	return (pkt_flags);
810 out:
811 	TCP_REASS_UNLOCK(tp);
812 	return (0);
813 }
814 
815 #ifdef INET6
816 int
817 tcp6_input(struct mbuf **mp, int *offp, int proto)
818 {
819 	struct mbuf *m = *mp;
820 
821 	/*
822 	 * draft-itojun-ipv6-tcp-to-anycast
823 	 * better place to put this in?
824 	 */
825 	if (m->m_flags & M_ANYCAST6) {
826 		struct ip6_hdr *ip6;
827 		if (m->m_len < sizeof(struct ip6_hdr)) {
828 			if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
829 				TCP_STATINC(TCP_STAT_RCVSHORT);
830 				return IPPROTO_DONE;
831 			}
832 		}
833 		ip6 = mtod(m, struct ip6_hdr *);
834 		icmp6_error(m, ICMP6_DST_UNREACH, ICMP6_DST_UNREACH_ADDR,
835 		    (char *)&ip6->ip6_dst - (char *)ip6);
836 		return IPPROTO_DONE;
837 	}
838 
839 	tcp_input(m, *offp, proto);
840 	return IPPROTO_DONE;
841 }
842 #endif
843 
844 #ifdef INET
845 static void
846 tcp4_log_refused(const struct ip *ip, const struct tcphdr *th)
847 {
848 	char src[INET_ADDRSTRLEN];
849 	char dst[INET_ADDRSTRLEN];
850 
851 	if (ip) {
852 		in_print(src, sizeof(src), &ip->ip_src);
853 		in_print(dst, sizeof(dst), &ip->ip_dst);
854 	}
855 	else {
856 		strlcpy(src, "(unknown)", sizeof(src));
857 		strlcpy(dst, "(unknown)", sizeof(dst));
858 	}
859 	log(LOG_INFO,
860 	    "Connection attempt to TCP %s:%d from %s:%d\n",
861 	    dst, ntohs(th->th_dport),
862 	    src, ntohs(th->th_sport));
863 }
864 #endif
865 
866 #ifdef INET6
867 static void
868 tcp6_log_refused(const struct ip6_hdr *ip6, const struct tcphdr *th)
869 {
870 	char src[INET6_ADDRSTRLEN];
871 	char dst[INET6_ADDRSTRLEN];
872 
873 	if (ip6) {
874 		in6_print(src, sizeof(src), &ip6->ip6_src);
875 		in6_print(dst, sizeof(dst), &ip6->ip6_dst);
876 	}
877 	else {
878 		strlcpy(src, "(unknown v6)", sizeof(src));
879 		strlcpy(dst, "(unknown v6)", sizeof(dst));
880 	}
881 	log(LOG_INFO,
882 	    "Connection attempt to TCP [%s]:%d from [%s]:%d\n",
883 	    dst, ntohs(th->th_dport),
884 	    src, ntohs(th->th_sport));
885 }
886 #endif
887 
888 /*
889  * Checksum extended TCP header and data.
890  */
891 int
892 tcp_input_checksum(int af, struct mbuf *m, const struct tcphdr *th,
893     int toff, int off, int tlen)
894 {
895 
896 	/*
897 	 * XXX it's better to record and check if this mbuf is
898 	 * already checked.
899 	 */
900 
901 	switch (af) {
902 #ifdef INET
903 	case AF_INET:
904 		switch (m->m_pkthdr.csum_flags &
905 			((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_TCPv4) |
906 			 M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) {
907 		case M_CSUM_TCPv4|M_CSUM_TCP_UDP_BAD:
908 			TCP_CSUM_COUNTER_INCR(&tcp_hwcsum_bad);
909 			goto badcsum;
910 
911 		case M_CSUM_TCPv4|M_CSUM_DATA: {
912 			u_int32_t hw_csum = m->m_pkthdr.csum_data;
913 
914 			TCP_CSUM_COUNTER_INCR(&tcp_hwcsum_data);
915 			if (m->m_pkthdr.csum_flags & M_CSUM_NO_PSEUDOHDR) {
916 				const struct ip *ip =
917 				    mtod(m, const struct ip *);
918 
919 				hw_csum = in_cksum_phdr(ip->ip_src.s_addr,
920 				    ip->ip_dst.s_addr,
921 				    htons(hw_csum + tlen + off + IPPROTO_TCP));
922 			}
923 			if ((hw_csum ^ 0xffff) != 0)
924 				goto badcsum;
925 			break;
926 		}
927 
928 		case M_CSUM_TCPv4:
929 			/* Checksum was okay. */
930 			TCP_CSUM_COUNTER_INCR(&tcp_hwcsum_ok);
931 			break;
932 
933 		default:
934 			/*
935 			 * Must compute it ourselves.  Maybe skip checksum
936 			 * on loopback interfaces.
937 			 */
938 			if (__predict_true(!(m->m_pkthdr.rcvif->if_flags &
939 					     IFF_LOOPBACK) ||
940 					   tcp_do_loopback_cksum)) {
941 				TCP_CSUM_COUNTER_INCR(&tcp_swcsum);
942 				if (in4_cksum(m, IPPROTO_TCP, toff,
943 					      tlen + off) != 0)
944 					goto badcsum;
945 			}
946 			break;
947 		}
948 		break;
949 #endif /* INET4 */
950 
951 #ifdef INET6
952 	case AF_INET6:
953 		switch (m->m_pkthdr.csum_flags &
954 			((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_TCPv6) |
955 			 M_CSUM_TCP_UDP_BAD | M_CSUM_DATA)) {
956 		case M_CSUM_TCPv6|M_CSUM_TCP_UDP_BAD:
957 			TCP_CSUM_COUNTER_INCR(&tcp6_hwcsum_bad);
958 			goto badcsum;
959 
960 #if 0 /* notyet */
961 		case M_CSUM_TCPv6|M_CSUM_DATA:
962 #endif
963 
964 		case M_CSUM_TCPv6:
965 			/* Checksum was okay. */
966 			TCP_CSUM_COUNTER_INCR(&tcp6_hwcsum_ok);
967 			break;
968 
969 		default:
970 			/*
971 			 * Must compute it ourselves.  Maybe skip checksum
972 			 * on loopback interfaces.
973 			 */
974 			if (__predict_true((m->m_flags & M_LOOP) == 0 ||
975 			    tcp_do_loopback_cksum)) {
976 				TCP_CSUM_COUNTER_INCR(&tcp6_swcsum);
977 				if (in6_cksum(m, IPPROTO_TCP, toff,
978 				    tlen + off) != 0)
979 					goto badcsum;
980 			}
981 		}
982 		break;
983 #endif /* INET6 */
984 	}
985 
986 	return 0;
987 
988 badcsum:
989 	TCP_STATINC(TCP_STAT_RCVBADSUM);
990 	return -1;
991 }
992 
993 /* When a packet arrives addressed to a vestigial tcpbp, we
994  * nevertheless have to respond to it per the spec.
995  */
996 static void tcp_vtw_input(struct tcphdr *th, vestigial_inpcb_t *vp,
997 			  struct mbuf *m, int tlen, int multicast)
998 {
999 	int		tiflags;
1000 	int		todrop;
1001 	uint32_t	t_flags = 0;
1002 	uint64_t	*tcps;
1003 
1004 	tiflags = th->th_flags;
1005 	todrop  = vp->rcv_nxt - th->th_seq;
1006 
1007 	if (todrop > 0) {
1008 		if (tiflags & TH_SYN) {
1009 			tiflags &= ~TH_SYN;
1010 			++th->th_seq;
1011 			if (th->th_urp > 1)
1012 				--th->th_urp;
1013 			else {
1014 				tiflags &= ~TH_URG;
1015 				th->th_urp = 0;
1016 			}
1017 			--todrop;
1018 		}
1019 		if (todrop > tlen ||
1020 		    (todrop == tlen && (tiflags & TH_FIN) == 0)) {
1021 			/*
1022 			 * Any valid FIN or RST must be to the left of the
1023 			 * window.  At this point the FIN or RST must be a
1024 			 * duplicate or out of sequence; drop it.
1025 			 */
1026 			if (tiflags & TH_RST)
1027 				goto drop;
1028 			tiflags &= ~(TH_FIN|TH_RST);
1029 			/*
1030 			 * Send an ACK to resynchronize and drop any data.
1031 			 * But keep on processing for RST or ACK.
1032 			 */
1033 			t_flags |= TF_ACKNOW;
1034 			todrop = tlen;
1035 			tcps = TCP_STAT_GETREF();
1036 			tcps[TCP_STAT_RCVDUPPACK] += 1;
1037 			tcps[TCP_STAT_RCVDUPBYTE] += todrop;
1038 			TCP_STAT_PUTREF();
1039 		} else if ((tiflags & TH_RST)
1040 			   && th->th_seq != vp->rcv_nxt) {
1041 			/*
1042 			 * Test for reset before adjusting the sequence
1043 			 * number for overlapping data.
1044 			 */
1045 			goto dropafterack_ratelim;
1046 		} else {
1047 			tcps = TCP_STAT_GETREF();
1048 			tcps[TCP_STAT_RCVPARTDUPPACK] += 1;
1049 			tcps[TCP_STAT_RCVPARTDUPBYTE] += todrop;
1050 			TCP_STAT_PUTREF();
1051 		}
1052 
1053 //		tcp_new_dsack(tp, th->th_seq, todrop);
1054 //		hdroptlen += todrop;	/*drop from head afterwards*/
1055 
1056 		th->th_seq += todrop;
1057 		tlen -= todrop;
1058 
1059 		if (th->th_urp > todrop)
1060 			th->th_urp -= todrop;
1061 		else {
1062 			tiflags &= ~TH_URG;
1063 			th->th_urp = 0;
1064 		}
1065 	}
1066 
1067 	/*
1068 	 * If new data are received on a connection after the
1069 	 * user processes are gone, then RST the other end.
1070 	 */
1071 	if (tlen) {
1072 		TCP_STATINC(TCP_STAT_RCVAFTERCLOSE);
1073 		goto dropwithreset;
1074 	}
1075 
1076 	/*
1077 	 * If segment ends after window, drop trailing data
1078 	 * (and PUSH and FIN); if nothing left, just ACK.
1079 	 */
1080 	todrop = (th->th_seq + tlen) - (vp->rcv_nxt+vp->rcv_wnd);
1081 
1082 	if (todrop > 0) {
1083 		TCP_STATINC(TCP_STAT_RCVPACKAFTERWIN);
1084 		if (todrop >= tlen) {
1085 			/*
1086 			 * The segment actually starts after the window.
1087 			 * th->th_seq + tlen - vp->rcv_nxt - vp->rcv_wnd >= tlen
1088 			 * th->th_seq - vp->rcv_nxt - vp->rcv_wnd >= 0
1089 			 * th->th_seq >= vp->rcv_nxt + vp->rcv_wnd
1090 			 */
1091 			TCP_STATADD(TCP_STAT_RCVBYTEAFTERWIN, tlen);
1092 			/*
1093 			 * If a new connection request is received
1094 			 * while in TIME_WAIT, drop the old connection
1095 			 * and start over if the sequence numbers
1096 			 * are above the previous ones.
1097 			 */
1098 			if ((tiflags & TH_SYN)
1099 			    && SEQ_GT(th->th_seq, vp->rcv_nxt)) {
1100 				/* We only support this in the !NOFDREF case, which
1101 				 * is to say: not here.
1102 				 */
1103 				goto dropwithreset;
1104 			}
1105 			/*
1106 			 * If window is closed can only take segments at
1107 			 * window edge, and have to drop data and PUSH from
1108 			 * incoming segments.  Continue processing, but
1109 			 * remember to ack.  Otherwise, drop segment
1110 			 * and (if not RST) ack.
1111 			 */
1112 			if (vp->rcv_wnd == 0 && th->th_seq == vp->rcv_nxt) {
1113 				t_flags |= TF_ACKNOW;
1114 				TCP_STATINC(TCP_STAT_RCVWINPROBE);
1115 			} else
1116 				goto dropafterack;
1117 		} else
1118 			TCP_STATADD(TCP_STAT_RCVBYTEAFTERWIN, todrop);
1119 		m_adj(m, -todrop);
1120 		tlen -= todrop;
1121 		tiflags &= ~(TH_PUSH|TH_FIN);
1122 	}
1123 
1124 	if (tiflags & TH_RST) {
1125 		if (th->th_seq != vp->rcv_nxt)
1126 			goto dropafterack_ratelim;
1127 
1128 		vtw_del(vp->ctl, vp->vtw);
1129 		goto drop;
1130 	}
1131 
1132 	/*
1133 	 * If the ACK bit is off we drop the segment and return.
1134 	 */
1135 	if ((tiflags & TH_ACK) == 0) {
1136 		if (t_flags & TF_ACKNOW)
1137 			goto dropafterack;
1138 		else
1139 			goto drop;
1140 	}
1141 
1142 	/*
1143 	 * In TIME_WAIT state the only thing that should arrive
1144 	 * is a retransmission of the remote FIN.  Acknowledge
1145 	 * it and restart the finack timer.
1146 	 */
1147 	vtw_restart(vp);
1148 	goto dropafterack;
1149 
1150 dropafterack:
1151 	/*
1152 	 * Generate an ACK dropping incoming segment if it occupies
1153 	 * sequence space, where the ACK reflects our state.
1154 	 */
1155 	if (tiflags & TH_RST)
1156 		goto drop;
1157 	goto dropafterack2;
1158 
1159 dropafterack_ratelim:
1160 	/*
1161 	 * We may want to rate-limit ACKs against SYN/RST attack.
1162 	 */
1163 	if (ppsratecheck(&tcp_ackdrop_ppslim_last, &tcp_ackdrop_ppslim_count,
1164 			 tcp_ackdrop_ppslim) == 0) {
1165 		/* XXX stat */
1166 		goto drop;
1167 	}
1168 	/* ...fall into dropafterack2... */
1169 
1170 dropafterack2:
1171 	(void)tcp_respond(0, m, m, th, th->th_seq + tlen, th->th_ack,
1172 			  TH_ACK);
1173 	return;
1174 
1175 dropwithreset:
1176 	/*
1177 	 * Generate a RST, dropping incoming segment.
1178 	 * Make ACK acceptable to originator of segment.
1179 	 */
1180 	if (tiflags & TH_RST)
1181 		goto drop;
1182 
1183 	if (tiflags & TH_ACK)
1184 		tcp_respond(0, m, m, th, (tcp_seq)0, th->th_ack, TH_RST);
1185 	else {
1186 		if (tiflags & TH_SYN)
1187 			++tlen;
1188 		(void)tcp_respond(0, m, m, th, th->th_seq + tlen, (tcp_seq)0,
1189 				  TH_RST|TH_ACK);
1190 	}
1191 	return;
1192 drop:
1193 	m_freem(m);
1194 }
1195 
1196 /*
1197  * TCP input routine, follows pages 65-76 of RFC 793 very closely.
1198  */
1199 void
1200 tcp_input(struct mbuf *m, ...)
1201 {
1202 	struct tcphdr *th;
1203 	struct ip *ip;
1204 	struct inpcb *inp;
1205 #ifdef INET6
1206 	struct ip6_hdr *ip6;
1207 	struct in6pcb *in6p;
1208 #endif
1209 	u_int8_t *optp = NULL;
1210 	int optlen = 0;
1211 	int len, tlen, toff, hdroptlen = 0;
1212 	struct tcpcb *tp = 0;
1213 	int tiflags;
1214 	struct socket *so = NULL;
1215 	int todrop, acked, ourfinisacked, needoutput = 0;
1216 	bool dupseg;
1217 #ifdef TCP_DEBUG
1218 	short ostate = 0;
1219 #endif
1220 	u_long tiwin;
1221 	struct tcp_opt_info opti;
1222 	int off, iphlen;
1223 	va_list ap;
1224 	int af;		/* af on the wire */
1225 	struct mbuf *tcp_saveti = NULL;
1226 	uint32_t ts_rtt;
1227 	uint8_t iptos;
1228 	uint64_t *tcps;
1229 	vestigial_inpcb_t vestige;
1230 
1231 	vestige.valid = 0;
1232 
1233 	MCLAIM(m, &tcp_rx_mowner);
1234 	va_start(ap, m);
1235 	toff = va_arg(ap, int);
1236 	(void)va_arg(ap, int);		/* ignore value, advance ap */
1237 	va_end(ap);
1238 
1239 	TCP_STATINC(TCP_STAT_RCVTOTAL);
1240 
1241 	memset(&opti, 0, sizeof(opti));
1242 	opti.ts_present = 0;
1243 	opti.maxseg = 0;
1244 
1245 	/*
1246 	 * RFC1122 4.2.3.10, p. 104: discard bcast/mcast SYN.
1247 	 *
1248 	 * TCP is, by definition, unicast, so we reject all
1249 	 * multicast outright.
1250 	 *
1251 	 * Note, there are additional src/dst address checks in
1252 	 * the AF-specific code below.
1253 	 */
1254 	if (m->m_flags & (M_BCAST|M_MCAST)) {
1255 		/* XXX stat */
1256 		goto drop;
1257 	}
1258 #ifdef INET6
1259 	if (m->m_flags & M_ANYCAST6) {
1260 		/* XXX stat */
1261 		goto drop;
1262 	}
1263 #endif
1264 
1265 	/*
1266 	 * Get IP and TCP header.
1267 	 * Note: IP leaves IP header in first mbuf.
1268 	 */
1269 	ip = mtod(m, struct ip *);
1270 	switch (ip->ip_v) {
1271 #ifdef INET
1272 	case 4:
1273 #ifdef INET6
1274 		ip6 = NULL;
1275 #endif
1276 		af = AF_INET;
1277 		iphlen = sizeof(struct ip);
1278 		IP6_EXTHDR_GET(th, struct tcphdr *, m, toff,
1279 			sizeof(struct tcphdr));
1280 		if (th == NULL) {
1281 			TCP_STATINC(TCP_STAT_RCVSHORT);
1282 			return;
1283 		}
1284 		/* We do the checksum after PCB lookup... */
1285 		len = ntohs(ip->ip_len);
1286 		tlen = len - toff;
1287 		iptos = ip->ip_tos;
1288 		break;
1289 #endif
1290 #ifdef INET6
1291 	case 6:
1292 		ip = NULL;
1293 		iphlen = sizeof(struct ip6_hdr);
1294 		af = AF_INET6;
1295 		ip6 = mtod(m, struct ip6_hdr *);
1296 		IP6_EXTHDR_GET(th, struct tcphdr *, m, toff,
1297 			sizeof(struct tcphdr));
1298 		if (th == NULL) {
1299 			TCP_STATINC(TCP_STAT_RCVSHORT);
1300 			return;
1301 		}
1302 
1303 		/* Be proactive about malicious use of IPv4 mapped address */
1304 		if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
1305 		    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
1306 			/* XXX stat */
1307 			goto drop;
1308 		}
1309 
1310 		/*
1311 		 * Be proactive about unspecified IPv6 address in source.
1312 		 * As we use all-zero to indicate unbounded/unconnected pcb,
1313 		 * unspecified IPv6 address can be used to confuse us.
1314 		 *
1315 		 * Note that packets with unspecified IPv6 destination is
1316 		 * already dropped in ip6_input.
1317 		 */
1318 		if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) {
1319 			/* XXX stat */
1320 			goto drop;
1321 		}
1322 
1323 		/*
1324 		 * Make sure destination address is not multicast.
1325 		 * Source address checked in ip6_input().
1326 		 */
1327 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1328 			/* XXX stat */
1329 			goto drop;
1330 		}
1331 
1332 		/* We do the checksum after PCB lookup... */
1333 		len = m->m_pkthdr.len;
1334 		tlen = len - toff;
1335 		iptos = (ntohl(ip6->ip6_flow) >> 20) & 0xff;
1336 		break;
1337 #endif
1338 	default:
1339 		m_freem(m);
1340 		return;
1341 	}
1342 
1343 	KASSERT(TCP_HDR_ALIGNED_P(th));
1344 
1345 	/*
1346 	 * Check that TCP offset makes sense,
1347 	 * pull out TCP options and adjust length.		XXX
1348 	 */
1349 	off = th->th_off << 2;
1350 	if (off < sizeof (struct tcphdr) || off > tlen) {
1351 		TCP_STATINC(TCP_STAT_RCVBADOFF);
1352 		goto drop;
1353 	}
1354 	tlen -= off;
1355 
1356 	/*
1357 	 * tcp_input() has been modified to use tlen to mean the TCP data
1358 	 * length throughout the function.  Other functions can use
1359 	 * m->m_pkthdr.len as the basis for calculating the TCP data length.
1360 	 * rja
1361 	 */
1362 
1363 	if (off > sizeof (struct tcphdr)) {
1364 		IP6_EXTHDR_GET(th, struct tcphdr *, m, toff, off);
1365 		if (th == NULL) {
1366 			TCP_STATINC(TCP_STAT_RCVSHORT);
1367 			return;
1368 		}
1369 		/*
1370 		 * NOTE: ip/ip6 will not be affected by m_pulldown()
1371 		 * (as they're before toff) and we don't need to update those.
1372 		 */
1373 		KASSERT(TCP_HDR_ALIGNED_P(th));
1374 		optlen = off - sizeof (struct tcphdr);
1375 		optp = ((u_int8_t *)th) + sizeof(struct tcphdr);
1376 		/*
1377 		 * Do quick retrieval of timestamp options ("options
1378 		 * prediction?").  If timestamp is the only option and it's
1379 		 * formatted as recommended in RFC 1323 appendix A, we
1380 		 * quickly get the values now and not bother calling
1381 		 * tcp_dooptions(), etc.
1382 		 */
1383 		if ((optlen == TCPOLEN_TSTAMP_APPA ||
1384 		     (optlen > TCPOLEN_TSTAMP_APPA &&
1385 			optp[TCPOLEN_TSTAMP_APPA] == TCPOPT_EOL)) &&
1386 		     *(u_int32_t *)optp == htonl(TCPOPT_TSTAMP_HDR) &&
1387 		     (th->th_flags & TH_SYN) == 0) {
1388 			opti.ts_present = 1;
1389 			opti.ts_val = ntohl(*(u_int32_t *)(optp + 4));
1390 			opti.ts_ecr = ntohl(*(u_int32_t *)(optp + 8));
1391 			optp = NULL;	/* we've parsed the options */
1392 		}
1393 	}
1394 	tiflags = th->th_flags;
1395 
1396 	/*
1397 	 * Locate pcb for segment.
1398 	 */
1399 findpcb:
1400 	inp = NULL;
1401 #ifdef INET6
1402 	in6p = NULL;
1403 #endif
1404 	switch (af) {
1405 #ifdef INET
1406 	case AF_INET:
1407 		inp = in_pcblookup_connect(&tcbtable, ip->ip_src, th->th_sport,
1408 					   ip->ip_dst, th->th_dport,
1409 					   &vestige);
1410 		if (inp == 0 && !vestige.valid) {
1411 			TCP_STATINC(TCP_STAT_PCBHASHMISS);
1412 			inp = in_pcblookup_bind(&tcbtable, ip->ip_dst, th->th_dport);
1413 		}
1414 #ifdef INET6
1415 		if (inp == 0 && !vestige.valid) {
1416 			struct in6_addr s, d;
1417 
1418 			/* mapped addr case */
1419 			memset(&s, 0, sizeof(s));
1420 			s.s6_addr16[5] = htons(0xffff);
1421 			bcopy(&ip->ip_src, &s.s6_addr32[3], sizeof(ip->ip_src));
1422 			memset(&d, 0, sizeof(d));
1423 			d.s6_addr16[5] = htons(0xffff);
1424 			bcopy(&ip->ip_dst, &d.s6_addr32[3], sizeof(ip->ip_dst));
1425 			in6p = in6_pcblookup_connect(&tcbtable, &s,
1426 						     th->th_sport, &d, th->th_dport,
1427 						     0, &vestige);
1428 			if (in6p == 0 && !vestige.valid) {
1429 				TCP_STATINC(TCP_STAT_PCBHASHMISS);
1430 				in6p = in6_pcblookup_bind(&tcbtable, &d,
1431 				    th->th_dport, 0);
1432 			}
1433 		}
1434 #endif
1435 #ifndef INET6
1436 		if (inp == 0 && !vestige.valid)
1437 #else
1438 		if (inp == 0 && in6p == 0 && !vestige.valid)
1439 #endif
1440 		{
1441 			TCP_STATINC(TCP_STAT_NOPORT);
1442 			if (tcp_log_refused &&
1443 			    (tiflags & (TH_RST|TH_ACK|TH_SYN)) == TH_SYN) {
1444 				tcp4_log_refused(ip, th);
1445 			}
1446 			tcp_fields_to_host(th);
1447 			goto dropwithreset_ratelim;
1448 		}
1449 #if defined(IPSEC)
1450 		if (ipsec_used) {
1451 			if (inp &&
1452 			    (inp->inp_socket->so_options & SO_ACCEPTCONN) == 0
1453 			    && ipsec4_in_reject(m, inp)) {
1454 				IPSEC_STATINC(IPSEC_STAT_IN_POLVIO);
1455 				goto drop;
1456 			}
1457 #ifdef INET6
1458 			else if (in6p &&
1459 			    (in6p->in6p_socket->so_options & SO_ACCEPTCONN) == 0
1460 			    && ipsec6_in_reject_so(m, in6p->in6p_socket)) {
1461 				IPSEC_STATINC(IPSEC_STAT_IN_POLVIO);
1462 				goto drop;
1463 			}
1464 #endif
1465 		}
1466 #endif /*IPSEC*/
1467 		break;
1468 #endif /*INET*/
1469 #ifdef INET6
1470 	case AF_INET6:
1471 	    {
1472 		int faith;
1473 
1474 #if defined(NFAITH) && NFAITH > 0
1475 		faith = faithprefix(&ip6->ip6_dst);
1476 #else
1477 		faith = 0;
1478 #endif
1479 		in6p = in6_pcblookup_connect(&tcbtable, &ip6->ip6_src,
1480 					     th->th_sport, &ip6->ip6_dst, th->th_dport, faith, &vestige);
1481 		if (!in6p && !vestige.valid) {
1482 			TCP_STATINC(TCP_STAT_PCBHASHMISS);
1483 			in6p = in6_pcblookup_bind(&tcbtable, &ip6->ip6_dst,
1484 				th->th_dport, faith);
1485 		}
1486 		if (!in6p && !vestige.valid) {
1487 			TCP_STATINC(TCP_STAT_NOPORT);
1488 			if (tcp_log_refused &&
1489 			    (tiflags & (TH_RST|TH_ACK|TH_SYN)) == TH_SYN) {
1490 				tcp6_log_refused(ip6, th);
1491 			}
1492 			tcp_fields_to_host(th);
1493 			goto dropwithreset_ratelim;
1494 		}
1495 #if defined(IPSEC)
1496 		if (ipsec_used && in6p
1497 		    && (in6p->in6p_socket->so_options & SO_ACCEPTCONN) == 0
1498 		    && ipsec6_in_reject(m, in6p)) {
1499 			IPSEC6_STATINC(IPSEC_STAT_IN_POLVIO);
1500 			goto drop;
1501 		}
1502 #endif /*IPSEC*/
1503 		break;
1504 	    }
1505 #endif
1506 	}
1507 
1508 	/*
1509 	 * If the state is CLOSED (i.e., TCB does not exist) then
1510 	 * all data in the incoming segment is discarded.
1511 	 * If the TCB exists but is in CLOSED state, it is embryonic,
1512 	 * but should either do a listen or a connect soon.
1513 	 */
1514 	tp = NULL;
1515 	so = NULL;
1516 	if (inp) {
1517 		/* Check the minimum TTL for socket. */
1518 		if (ip->ip_ttl < inp->inp_ip_minttl)
1519 			goto drop;
1520 
1521 		tp = intotcpcb(inp);
1522 		so = inp->inp_socket;
1523 	}
1524 #ifdef INET6
1525 	else if (in6p) {
1526 		tp = in6totcpcb(in6p);
1527 		so = in6p->in6p_socket;
1528 	}
1529 #endif
1530 	else if (vestige.valid) {
1531 		int mc = 0;
1532 
1533 		/* We do not support the resurrection of vtw tcpcps.
1534 		 */
1535 		if (tcp_input_checksum(af, m, th, toff, off, tlen))
1536 			goto badcsum;
1537 
1538 		switch (af) {
1539 #ifdef INET6
1540 		case AF_INET6:
1541 			mc = IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst);
1542 			break;
1543 #endif
1544 
1545 		case AF_INET:
1546 			mc = (IN_MULTICAST(ip->ip_dst.s_addr)
1547 			      || in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif));
1548 			break;
1549 		}
1550 
1551 		tcp_fields_to_host(th);
1552 		tcp_vtw_input(th, &vestige, m, tlen, mc);
1553 		m = 0;
1554 		goto drop;
1555 	}
1556 
1557 	if (tp == 0) {
1558 		tcp_fields_to_host(th);
1559 		goto dropwithreset_ratelim;
1560 	}
1561 	if (tp->t_state == TCPS_CLOSED)
1562 		goto drop;
1563 
1564 	KASSERT(so->so_lock == softnet_lock);
1565 	KASSERT(solocked(so));
1566 
1567 	/*
1568 	 * Checksum extended TCP header and data.
1569 	 */
1570 	if (tcp_input_checksum(af, m, th, toff, off, tlen))
1571 		goto badcsum;
1572 
1573 	tcp_fields_to_host(th);
1574 
1575 	/* Unscale the window into a 32-bit value. */
1576 	if ((tiflags & TH_SYN) == 0)
1577 		tiwin = th->th_win << tp->snd_scale;
1578 	else
1579 		tiwin = th->th_win;
1580 
1581 #ifdef INET6
1582 	/* save packet options if user wanted */
1583 	if (in6p && (in6p->in6p_flags & IN6P_CONTROLOPTS)) {
1584 		if (in6p->in6p_options) {
1585 			m_freem(in6p->in6p_options);
1586 			in6p->in6p_options = 0;
1587 		}
1588 		KASSERT(ip6 != NULL);
1589 		ip6_savecontrol(in6p, &in6p->in6p_options, ip6, m);
1590 	}
1591 #endif
1592 
1593 	if (so->so_options & (SO_DEBUG|SO_ACCEPTCONN)) {
1594 		union syn_cache_sa src;
1595 		union syn_cache_sa dst;
1596 
1597 		memset(&src, 0, sizeof(src));
1598 		memset(&dst, 0, sizeof(dst));
1599 		switch (af) {
1600 #ifdef INET
1601 		case AF_INET:
1602 			src.sin.sin_len = sizeof(struct sockaddr_in);
1603 			src.sin.sin_family = AF_INET;
1604 			src.sin.sin_addr = ip->ip_src;
1605 			src.sin.sin_port = th->th_sport;
1606 
1607 			dst.sin.sin_len = sizeof(struct sockaddr_in);
1608 			dst.sin.sin_family = AF_INET;
1609 			dst.sin.sin_addr = ip->ip_dst;
1610 			dst.sin.sin_port = th->th_dport;
1611 			break;
1612 #endif
1613 #ifdef INET6
1614 		case AF_INET6:
1615 			src.sin6.sin6_len = sizeof(struct sockaddr_in6);
1616 			src.sin6.sin6_family = AF_INET6;
1617 			src.sin6.sin6_addr = ip6->ip6_src;
1618 			src.sin6.sin6_port = th->th_sport;
1619 
1620 			dst.sin6.sin6_len = sizeof(struct sockaddr_in6);
1621 			dst.sin6.sin6_family = AF_INET6;
1622 			dst.sin6.sin6_addr = ip6->ip6_dst;
1623 			dst.sin6.sin6_port = th->th_dport;
1624 			break;
1625 #endif /* INET6 */
1626 		default:
1627 			goto badsyn;	/*sanity*/
1628 		}
1629 
1630 		if (so->so_options & SO_DEBUG) {
1631 #ifdef TCP_DEBUG
1632 			ostate = tp->t_state;
1633 #endif
1634 
1635 			tcp_saveti = NULL;
1636 			if (iphlen + sizeof(struct tcphdr) > MHLEN)
1637 				goto nosave;
1638 
1639 			if (m->m_len > iphlen && (m->m_flags & M_EXT) == 0) {
1640 				tcp_saveti = m_copym(m, 0, iphlen, M_DONTWAIT);
1641 				if (!tcp_saveti)
1642 					goto nosave;
1643 			} else {
1644 				MGETHDR(tcp_saveti, M_DONTWAIT, MT_HEADER);
1645 				if (!tcp_saveti)
1646 					goto nosave;
1647 				MCLAIM(m, &tcp_mowner);
1648 				tcp_saveti->m_len = iphlen;
1649 				m_copydata(m, 0, iphlen,
1650 				    mtod(tcp_saveti, void *));
1651 			}
1652 
1653 			if (M_TRAILINGSPACE(tcp_saveti) < sizeof(struct tcphdr)) {
1654 				m_freem(tcp_saveti);
1655 				tcp_saveti = NULL;
1656 			} else {
1657 				tcp_saveti->m_len += sizeof(struct tcphdr);
1658 				memcpy(mtod(tcp_saveti, char *) + iphlen, th,
1659 				    sizeof(struct tcphdr));
1660 			}
1661 	nosave:;
1662 		}
1663 		if (so->so_options & SO_ACCEPTCONN) {
1664 			if ((tiflags & (TH_RST|TH_ACK|TH_SYN)) != TH_SYN) {
1665 				if (tiflags & TH_RST) {
1666 					syn_cache_reset(&src.sa, &dst.sa, th);
1667 				} else if ((tiflags & (TH_ACK|TH_SYN)) ==
1668 				    (TH_ACK|TH_SYN)) {
1669 					/*
1670 					 * Received a SYN,ACK.  This should
1671 					 * never happen while we are in
1672 					 * LISTEN.  Send an RST.
1673 					 */
1674 					goto badsyn;
1675 				} else if (tiflags & TH_ACK) {
1676 					so = syn_cache_get(&src.sa, &dst.sa,
1677 						th, toff, tlen, so, m);
1678 					if (so == NULL) {
1679 						/*
1680 						 * We don't have a SYN for
1681 						 * this ACK; send an RST.
1682 						 */
1683 						goto badsyn;
1684 					} else if (so ==
1685 					    (struct socket *)(-1)) {
1686 						/*
1687 						 * We were unable to create
1688 						 * the connection.  If the
1689 						 * 3-way handshake was
1690 						 * completed, and RST has
1691 						 * been sent to the peer.
1692 						 * Since the mbuf might be
1693 						 * in use for the reply,
1694 						 * do not free it.
1695 						 */
1696 						m = NULL;
1697 					} else {
1698 						/*
1699 						 * We have created a
1700 						 * full-blown connection.
1701 						 */
1702 						tp = NULL;
1703 						inp = NULL;
1704 #ifdef INET6
1705 						in6p = NULL;
1706 #endif
1707 						switch (so->so_proto->pr_domain->dom_family) {
1708 #ifdef INET
1709 						case AF_INET:
1710 							inp = sotoinpcb(so);
1711 							tp = intotcpcb(inp);
1712 							break;
1713 #endif
1714 #ifdef INET6
1715 						case AF_INET6:
1716 							in6p = sotoin6pcb(so);
1717 							tp = in6totcpcb(in6p);
1718 							break;
1719 #endif
1720 						}
1721 						if (tp == NULL)
1722 							goto badsyn;	/*XXX*/
1723 						tiwin <<= tp->snd_scale;
1724 						goto after_listen;
1725 					}
1726 				} else {
1727 					/*
1728 					 * None of RST, SYN or ACK was set.
1729 					 * This is an invalid packet for a
1730 					 * TCB in LISTEN state.  Send a RST.
1731 					 */
1732 					goto badsyn;
1733 				}
1734 			} else {
1735 				/*
1736 				 * Received a SYN.
1737 				 *
1738 				 * RFC1122 4.2.3.10, p. 104: discard bcast/mcast SYN
1739 				 */
1740 				if (m->m_flags & (M_BCAST|M_MCAST))
1741 					goto drop;
1742 
1743 				switch (af) {
1744 #ifdef INET6
1745 				case AF_INET6:
1746 					if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
1747 						goto drop;
1748 					break;
1749 #endif /* INET6 */
1750 				case AF_INET:
1751 					if (IN_MULTICAST(ip->ip_dst.s_addr) ||
1752 					    in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif))
1753 						goto drop;
1754 				break;
1755 				}
1756 
1757 #ifdef INET6
1758 				/*
1759 				 * If deprecated address is forbidden, we do
1760 				 * not accept SYN to deprecated interface
1761 				 * address to prevent any new inbound
1762 				 * connection from getting established.
1763 				 * When we do not accept SYN, we send a TCP
1764 				 * RST, with deprecated source address (instead
1765 				 * of dropping it).  We compromise it as it is
1766 				 * much better for peer to send a RST, and
1767 				 * RST will be the final packet for the
1768 				 * exchange.
1769 				 *
1770 				 * If we do not forbid deprecated addresses, we
1771 				 * accept the SYN packet.  RFC2462 does not
1772 				 * suggest dropping SYN in this case.
1773 				 * If we decipher RFC2462 5.5.4, it says like
1774 				 * this:
1775 				 * 1. use of deprecated addr with existing
1776 				 *    communication is okay - "SHOULD continue
1777 				 *    to be used"
1778 				 * 2. use of it with new communication:
1779 				 *   (2a) "SHOULD NOT be used if alternate
1780 				 *        address with sufficient scope is
1781 				 *        available"
1782 				 *   (2b) nothing mentioned otherwise.
1783 				 * Here we fall into (2b) case as we have no
1784 				 * choice in our source address selection - we
1785 				 * must obey the peer.
1786 				 *
1787 				 * The wording in RFC2462 is confusing, and
1788 				 * there are multiple description text for
1789 				 * deprecated address handling - worse, they
1790 				 * are not exactly the same.  I believe 5.5.4
1791 				 * is the best one, so we follow 5.5.4.
1792 				 */
1793 				if (af == AF_INET6 && !ip6_use_deprecated) {
1794 					struct in6_ifaddr *ia6;
1795 					if ((ia6 = in6ifa_ifpwithaddr(m->m_pkthdr.rcvif,
1796 					    &ip6->ip6_dst)) &&
1797 					    (ia6->ia6_flags & IN6_IFF_DEPRECATED)) {
1798 						tp = NULL;
1799 						goto dropwithreset;
1800 					}
1801 				}
1802 #endif
1803 
1804 #if defined(IPSEC)
1805 				if (ipsec_used) {
1806 					switch (af) {
1807 #ifdef INET
1808 					case AF_INET:
1809 						if (!ipsec4_in_reject_so(m, so))
1810 							break;
1811 						IPSEC_STATINC(
1812 						    IPSEC_STAT_IN_POLVIO);
1813 						tp = NULL;
1814 						goto dropwithreset;
1815 #endif
1816 #ifdef INET6
1817 					case AF_INET6:
1818 						if (!ipsec6_in_reject_so(m, so))
1819 							break;
1820 						IPSEC6_STATINC(
1821 						    IPSEC_STAT_IN_POLVIO);
1822 						tp = NULL;
1823 						goto dropwithreset;
1824 #endif /*INET6*/
1825 					}
1826 				}
1827 #endif /*IPSEC*/
1828 
1829 				/*
1830 				 * LISTEN socket received a SYN
1831 				 * from itself?  This can't possibly
1832 				 * be valid; drop the packet.
1833 				 */
1834 				if (th->th_sport == th->th_dport) {
1835 					int i;
1836 
1837 					switch (af) {
1838 #ifdef INET
1839 					case AF_INET:
1840 						i = in_hosteq(ip->ip_src, ip->ip_dst);
1841 						break;
1842 #endif
1843 #ifdef INET6
1844 					case AF_INET6:
1845 						i = IN6_ARE_ADDR_EQUAL(&ip6->ip6_src, &ip6->ip6_dst);
1846 						break;
1847 #endif
1848 					default:
1849 						i = 1;
1850 					}
1851 					if (i) {
1852 						TCP_STATINC(TCP_STAT_BADSYN);
1853 						goto drop;
1854 					}
1855 				}
1856 
1857 				/*
1858 				 * SYN looks ok; create compressed TCP
1859 				 * state for it.
1860 				 */
1861 				if (so->so_qlen <= so->so_qlimit &&
1862 				    syn_cache_add(&src.sa, &dst.sa, th, tlen,
1863 						so, m, optp, optlen, &opti))
1864 					m = NULL;
1865 			}
1866 			goto drop;
1867 		}
1868 	}
1869 
1870 after_listen:
1871 #ifdef DIAGNOSTIC
1872 	/*
1873 	 * Should not happen now that all embryonic connections
1874 	 * are handled with compressed state.
1875 	 */
1876 	if (tp->t_state == TCPS_LISTEN)
1877 		panic("tcp_input: TCPS_LISTEN");
1878 #endif
1879 
1880 	/*
1881 	 * Segment received on connection.
1882 	 * Reset idle time and keep-alive timer.
1883 	 */
1884 	tp->t_rcvtime = tcp_now;
1885 	if (TCPS_HAVEESTABLISHED(tp->t_state))
1886 		TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepidle);
1887 
1888 	/*
1889 	 * Process options.
1890 	 */
1891 #ifdef TCP_SIGNATURE
1892 	if (optp || (tp->t_flags & TF_SIGNATURE))
1893 #else
1894 	if (optp)
1895 #endif
1896 		if (tcp_dooptions(tp, optp, optlen, th, m, toff, &opti) < 0)
1897 			goto drop;
1898 
1899 	if (TCP_SACK_ENABLED(tp)) {
1900 		tcp_del_sackholes(tp, th);
1901 	}
1902 
1903 	if (TCP_ECN_ALLOWED(tp)) {
1904 		if (tiflags & TH_CWR) {
1905 			tp->t_flags &= ~TF_ECN_SND_ECE;
1906 		}
1907 		switch (iptos & IPTOS_ECN_MASK) {
1908 		case IPTOS_ECN_CE:
1909 			tp->t_flags |= TF_ECN_SND_ECE;
1910 			TCP_STATINC(TCP_STAT_ECN_CE);
1911 			break;
1912 		case IPTOS_ECN_ECT0:
1913 			TCP_STATINC(TCP_STAT_ECN_ECT);
1914 			break;
1915 		case IPTOS_ECN_ECT1:
1916 			/* XXX: ignore for now -- rpaulo */
1917 			break;
1918 		}
1919 		/*
1920 		 * Congestion experienced.
1921 		 * Ignore if we are already trying to recover.
1922 		 */
1923 		if ((tiflags & TH_ECE) && SEQ_GEQ(tp->snd_una, tp->snd_recover))
1924 			tp->t_congctl->cong_exp(tp);
1925 	}
1926 
1927 	if (opti.ts_present && opti.ts_ecr) {
1928 		/*
1929 		 * Calculate the RTT from the returned time stamp and the
1930 		 * connection's time base.  If the time stamp is later than
1931 		 * the current time, or is extremely old, fall back to non-1323
1932 		 * RTT calculation.  Since ts_rtt is unsigned, we can test both
1933 		 * at the same time.
1934 		 *
1935 		 * Note that ts_rtt is in units of slow ticks (500
1936 		 * ms).  Since most earthbound RTTs are < 500 ms,
1937 		 * observed values will have large quantization noise.
1938 		 * Our smoothed RTT is then the fraction of observed
1939 		 * samples that are 1 tick instead of 0 (times 500
1940 		 * ms).
1941 		 *
1942 		 * ts_rtt is increased by 1 to denote a valid sample,
1943 		 * with 0 indicating an invalid measurement.  This
1944 		 * extra 1 must be removed when ts_rtt is used, or
1945 		 * else an an erroneous extra 500 ms will result.
1946 		 */
1947 		ts_rtt = TCP_TIMESTAMP(tp) - opti.ts_ecr + 1;
1948 		if (ts_rtt > TCP_PAWS_IDLE)
1949 			ts_rtt = 0;
1950 	} else {
1951 		ts_rtt = 0;
1952 	}
1953 
1954 	/*
1955 	 * Header prediction: check for the two common cases
1956 	 * of a uni-directional data xfer.  If the packet has
1957 	 * no control flags, is in-sequence, the window didn't
1958 	 * change and we're not retransmitting, it's a
1959 	 * candidate.  If the length is zero and the ack moved
1960 	 * forward, we're the sender side of the xfer.  Just
1961 	 * free the data acked & wake any higher level process
1962 	 * that was blocked waiting for space.  If the length
1963 	 * is non-zero and the ack didn't move, we're the
1964 	 * receiver side.  If we're getting packets in-order
1965 	 * (the reassembly queue is empty), add the data to
1966 	 * the socket buffer and note that we need a delayed ack.
1967 	 */
1968 	if (tp->t_state == TCPS_ESTABLISHED &&
1969 	    (tiflags & (TH_SYN|TH_FIN|TH_RST|TH_URG|TH_ECE|TH_CWR|TH_ACK))
1970 	        == TH_ACK &&
1971 	    (!opti.ts_present || TSTMP_GEQ(opti.ts_val, tp->ts_recent)) &&
1972 	    th->th_seq == tp->rcv_nxt &&
1973 	    tiwin && tiwin == tp->snd_wnd &&
1974 	    tp->snd_nxt == tp->snd_max) {
1975 
1976 		/*
1977 		 * If last ACK falls within this segment's sequence numbers,
1978 		 * record the timestamp.
1979 		 * NOTE that the test is modified according to the latest
1980 		 * proposal of the tcplw@cray.com list (Braden 1993/04/26).
1981 		 *
1982 		 * note that we already know
1983 		 *	TSTMP_GEQ(opti.ts_val, tp->ts_recent)
1984 		 */
1985 		if (opti.ts_present &&
1986 		    SEQ_LEQ(th->th_seq, tp->last_ack_sent)) {
1987 			tp->ts_recent_age = tcp_now;
1988 			tp->ts_recent = opti.ts_val;
1989 		}
1990 
1991 		if (tlen == 0) {
1992 			/* Ack prediction. */
1993 			if (SEQ_GT(th->th_ack, tp->snd_una) &&
1994 			    SEQ_LEQ(th->th_ack, tp->snd_max) &&
1995 			    tp->snd_cwnd >= tp->snd_wnd &&
1996 			    tp->t_partialacks < 0) {
1997 				/*
1998 				 * this is a pure ack for outstanding data.
1999 				 */
2000 				if (ts_rtt)
2001 					tcp_xmit_timer(tp, ts_rtt - 1);
2002 				else if (tp->t_rtttime &&
2003 				    SEQ_GT(th->th_ack, tp->t_rtseq))
2004 					tcp_xmit_timer(tp,
2005 					  tcp_now - tp->t_rtttime);
2006 				acked = th->th_ack - tp->snd_una;
2007 				tcps = TCP_STAT_GETREF();
2008 				tcps[TCP_STAT_PREDACK]++;
2009 				tcps[TCP_STAT_RCVACKPACK]++;
2010 				tcps[TCP_STAT_RCVACKBYTE] += acked;
2011 				TCP_STAT_PUTREF();
2012 				nd6_hint(tp);
2013 
2014 				if (acked > (tp->t_lastoff - tp->t_inoff))
2015 					tp->t_lastm = NULL;
2016 				sbdrop(&so->so_snd, acked);
2017 				tp->t_lastoff -= acked;
2018 
2019 				icmp_check(tp, th, acked);
2020 
2021 				tp->snd_una = th->th_ack;
2022 				tp->snd_fack = tp->snd_una;
2023 				if (SEQ_LT(tp->snd_high, tp->snd_una))
2024 					tp->snd_high = tp->snd_una;
2025 				m_freem(m);
2026 
2027 				/*
2028 				 * If all outstanding data are acked, stop
2029 				 * retransmit timer, otherwise restart timer
2030 				 * using current (possibly backed-off) value.
2031 				 * If process is waiting for space,
2032 				 * wakeup/selnotify/signal.  If data
2033 				 * are ready to send, let tcp_output
2034 				 * decide between more output or persist.
2035 				 */
2036 				if (tp->snd_una == tp->snd_max)
2037 					TCP_TIMER_DISARM(tp, TCPT_REXMT);
2038 				else if (TCP_TIMER_ISARMED(tp,
2039 				    TCPT_PERSIST) == 0)
2040 					TCP_TIMER_ARM(tp, TCPT_REXMT,
2041 					    tp->t_rxtcur);
2042 
2043 				sowwakeup(so);
2044 				if (so->so_snd.sb_cc) {
2045 					KERNEL_LOCK(1, NULL);
2046 					(void) tcp_output(tp);
2047 					KERNEL_UNLOCK_ONE(NULL);
2048 				}
2049 				if (tcp_saveti)
2050 					m_freem(tcp_saveti);
2051 				return;
2052 			}
2053 		} else if (th->th_ack == tp->snd_una &&
2054 		    TAILQ_FIRST(&tp->segq) == NULL &&
2055 		    tlen <= sbspace(&so->so_rcv)) {
2056 			int newsize = 0;	/* automatic sockbuf scaling */
2057 
2058 			/*
2059 			 * this is a pure, in-sequence data packet
2060 			 * with nothing on the reassembly queue and
2061 			 * we have enough buffer space to take it.
2062 			 */
2063 			tp->rcv_nxt += tlen;
2064 			tcps = TCP_STAT_GETREF();
2065 			tcps[TCP_STAT_PREDDAT]++;
2066 			tcps[TCP_STAT_RCVPACK]++;
2067 			tcps[TCP_STAT_RCVBYTE] += tlen;
2068 			TCP_STAT_PUTREF();
2069 			nd6_hint(tp);
2070 
2071 		/*
2072 		 * Automatic sizing enables the performance of large buffers
2073 		 * and most of the efficiency of small ones by only allocating
2074 		 * space when it is needed.
2075 		 *
2076 		 * On the receive side the socket buffer memory is only rarely
2077 		 * used to any significant extent.  This allows us to be much
2078 		 * more aggressive in scaling the receive socket buffer.  For
2079 		 * the case that the buffer space is actually used to a large
2080 		 * extent and we run out of kernel memory we can simply drop
2081 		 * the new segments; TCP on the sender will just retransmit it
2082 		 * later.  Setting the buffer size too big may only consume too
2083 		 * much kernel memory if the application doesn't read() from
2084 		 * the socket or packet loss or reordering makes use of the
2085 		 * reassembly queue.
2086 		 *
2087 		 * The criteria to step up the receive buffer one notch are:
2088 		 *  1. the number of bytes received during the time it takes
2089 		 *     one timestamp to be reflected back to us (the RTT);
2090 		 *  2. received bytes per RTT is within seven eighth of the
2091 		 *     current socket buffer size;
2092 		 *  3. receive buffer size has not hit maximal automatic size;
2093 		 *
2094 		 * This algorithm does one step per RTT at most and only if
2095 		 * we receive a bulk stream w/o packet losses or reorderings.
2096 		 * Shrinking the buffer during idle times is not necessary as
2097 		 * it doesn't consume any memory when idle.
2098 		 *
2099 		 * TODO: Only step up if the application is actually serving
2100 		 * the buffer to better manage the socket buffer resources.
2101 		 */
2102 			if (tcp_do_autorcvbuf &&
2103 			    opti.ts_ecr &&
2104 			    (so->so_rcv.sb_flags & SB_AUTOSIZE)) {
2105 				if (opti.ts_ecr > tp->rfbuf_ts &&
2106 				    opti.ts_ecr - tp->rfbuf_ts < PR_SLOWHZ) {
2107 					if (tp->rfbuf_cnt >
2108 					    (so->so_rcv.sb_hiwat / 8 * 7) &&
2109 					    so->so_rcv.sb_hiwat <
2110 					    tcp_autorcvbuf_max) {
2111 						newsize =
2112 						    min(so->so_rcv.sb_hiwat +
2113 						    tcp_autorcvbuf_inc,
2114 						    tcp_autorcvbuf_max);
2115 					}
2116 					/* Start over with next RTT. */
2117 					tp->rfbuf_ts = 0;
2118 					tp->rfbuf_cnt = 0;
2119 				} else
2120 					tp->rfbuf_cnt += tlen;	/* add up */
2121 			}
2122 
2123 			/*
2124 			 * Drop TCP, IP headers and TCP options then add data
2125 			 * to socket buffer.
2126 			 */
2127 			if (so->so_state & SS_CANTRCVMORE)
2128 				m_freem(m);
2129 			else {
2130 				/*
2131 				 * Set new socket buffer size.
2132 				 * Give up when limit is reached.
2133 				 */
2134 				if (newsize)
2135 					if (!sbreserve(&so->so_rcv,
2136 					    newsize, so))
2137 						so->so_rcv.sb_flags &= ~SB_AUTOSIZE;
2138 				m_adj(m, toff + off);
2139 				sbappendstream(&so->so_rcv, m);
2140 			}
2141 			sorwakeup(so);
2142 			tcp_setup_ack(tp, th);
2143 			if (tp->t_flags & TF_ACKNOW) {
2144 				KERNEL_LOCK(1, NULL);
2145 				(void) tcp_output(tp);
2146 				KERNEL_UNLOCK_ONE(NULL);
2147 			}
2148 			if (tcp_saveti)
2149 				m_freem(tcp_saveti);
2150 			return;
2151 		}
2152 	}
2153 
2154 	/*
2155 	 * Compute mbuf offset to TCP data segment.
2156 	 */
2157 	hdroptlen = toff + off;
2158 
2159 	/*
2160 	 * Calculate amount of space in receive window,
2161 	 * and then do TCP input processing.
2162 	 * Receive window is amount of space in rcv queue,
2163 	 * but not less than advertised window.
2164 	 */
2165 	{ int win;
2166 
2167 	win = sbspace(&so->so_rcv);
2168 	if (win < 0)
2169 		win = 0;
2170 	tp->rcv_wnd = imax(win, (int)(tp->rcv_adv - tp->rcv_nxt));
2171 	}
2172 
2173 	/* Reset receive buffer auto scaling when not in bulk receive mode. */
2174 	tp->rfbuf_ts = 0;
2175 	tp->rfbuf_cnt = 0;
2176 
2177 	switch (tp->t_state) {
2178 	/*
2179 	 * If the state is SYN_SENT:
2180 	 *	if seg contains an ACK, but not for our SYN, drop the input.
2181 	 *	if seg contains a RST, then drop the connection.
2182 	 *	if seg does not contain SYN, then drop it.
2183 	 * Otherwise this is an acceptable SYN segment
2184 	 *	initialize tp->rcv_nxt and tp->irs
2185 	 *	if seg contains ack then advance tp->snd_una
2186 	 *	if seg contains a ECE and ECN support is enabled, the stream
2187 	 *	    is ECN capable.
2188 	 *	if SYN has been acked change to ESTABLISHED else SYN_RCVD state
2189 	 *	arrange for segment to be acked (eventually)
2190 	 *	continue processing rest of data/controls, beginning with URG
2191 	 */
2192 	case TCPS_SYN_SENT:
2193 		if ((tiflags & TH_ACK) &&
2194 		    (SEQ_LEQ(th->th_ack, tp->iss) ||
2195 		     SEQ_GT(th->th_ack, tp->snd_max)))
2196 			goto dropwithreset;
2197 		if (tiflags & TH_RST) {
2198 			if (tiflags & TH_ACK)
2199 				tp = tcp_drop(tp, ECONNREFUSED);
2200 			goto drop;
2201 		}
2202 		if ((tiflags & TH_SYN) == 0)
2203 			goto drop;
2204 		if (tiflags & TH_ACK) {
2205 			tp->snd_una = th->th_ack;
2206 			if (SEQ_LT(tp->snd_nxt, tp->snd_una))
2207 				tp->snd_nxt = tp->snd_una;
2208 			if (SEQ_LT(tp->snd_high, tp->snd_una))
2209 				tp->snd_high = tp->snd_una;
2210 			TCP_TIMER_DISARM(tp, TCPT_REXMT);
2211 
2212 			if ((tiflags & TH_ECE) && tcp_do_ecn) {
2213 				tp->t_flags |= TF_ECN_PERMIT;
2214 				TCP_STATINC(TCP_STAT_ECN_SHS);
2215 			}
2216 
2217 		}
2218 		tp->irs = th->th_seq;
2219 		tcp_rcvseqinit(tp);
2220 		tp->t_flags |= TF_ACKNOW;
2221 		tcp_mss_from_peer(tp, opti.maxseg);
2222 
2223 		/*
2224 		 * Initialize the initial congestion window.  If we
2225 		 * had to retransmit the SYN, we must initialize cwnd
2226 		 * to 1 segment (i.e. the Loss Window).
2227 		 */
2228 		if (tp->t_flags & TF_SYN_REXMT)
2229 			tp->snd_cwnd = tp->t_peermss;
2230 		else {
2231 			int ss = tcp_init_win;
2232 #ifdef INET
2233 			if (inp != NULL && in_localaddr(inp->inp_faddr))
2234 				ss = tcp_init_win_local;
2235 #endif
2236 #ifdef INET6
2237 			if (in6p != NULL && in6_localaddr(&in6p->in6p_faddr))
2238 				ss = tcp_init_win_local;
2239 #endif
2240 			tp->snd_cwnd = TCP_INITIAL_WINDOW(ss, tp->t_peermss);
2241 		}
2242 
2243 		tcp_rmx_rtt(tp);
2244 		if (tiflags & TH_ACK) {
2245 			TCP_STATINC(TCP_STAT_CONNECTS);
2246 			/*
2247 			 * move tcp_established before soisconnected
2248 			 * because upcall handler can drive tcp_output
2249 			 * functionality.
2250 			 * XXX we might call soisconnected at the end of
2251 			 * all processing
2252 			 */
2253 			tcp_established(tp);
2254 			soisconnected(so);
2255 			/* Do window scaling on this connection? */
2256 			if ((tp->t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) ==
2257 			    (TF_RCVD_SCALE|TF_REQ_SCALE)) {
2258 				tp->snd_scale = tp->requested_s_scale;
2259 				tp->rcv_scale = tp->request_r_scale;
2260 			}
2261 			TCP_REASS_LOCK(tp);
2262 			(void) tcp_reass(tp, NULL, NULL, &tlen);
2263 			/*
2264 			 * if we didn't have to retransmit the SYN,
2265 			 * use its rtt as our initial srtt & rtt var.
2266 			 */
2267 			if (tp->t_rtttime)
2268 				tcp_xmit_timer(tp, tcp_now - tp->t_rtttime);
2269 		} else
2270 			tp->t_state = TCPS_SYN_RECEIVED;
2271 
2272 		/*
2273 		 * Advance th->th_seq to correspond to first data byte.
2274 		 * If data, trim to stay within window,
2275 		 * dropping FIN if necessary.
2276 		 */
2277 		th->th_seq++;
2278 		if (tlen > tp->rcv_wnd) {
2279 			todrop = tlen - tp->rcv_wnd;
2280 			m_adj(m, -todrop);
2281 			tlen = tp->rcv_wnd;
2282 			tiflags &= ~TH_FIN;
2283 			tcps = TCP_STAT_GETREF();
2284 			tcps[TCP_STAT_RCVPACKAFTERWIN]++;
2285 			tcps[TCP_STAT_RCVBYTEAFTERWIN] += todrop;
2286 			TCP_STAT_PUTREF();
2287 		}
2288 		tp->snd_wl1 = th->th_seq - 1;
2289 		tp->rcv_up = th->th_seq;
2290 		goto step6;
2291 
2292 	/*
2293 	 * If the state is SYN_RECEIVED:
2294 	 *	If seg contains an ACK, but not for our SYN, drop the input
2295 	 *	and generate an RST.  See page 36, rfc793
2296 	 */
2297 	case TCPS_SYN_RECEIVED:
2298 		if ((tiflags & TH_ACK) &&
2299 		    (SEQ_LEQ(th->th_ack, tp->iss) ||
2300 		     SEQ_GT(th->th_ack, tp->snd_max)))
2301 			goto dropwithreset;
2302 		break;
2303 	}
2304 
2305 	/*
2306 	 * States other than LISTEN or SYN_SENT.
2307 	 * First check timestamp, if present.
2308 	 * Then check that at least some bytes of segment are within
2309 	 * receive window.  If segment begins before rcv_nxt,
2310 	 * drop leading data (and SYN); if nothing left, just ack.
2311 	 *
2312 	 * RFC 1323 PAWS: If we have a timestamp reply on this segment
2313 	 * and it's less than ts_recent, drop it.
2314 	 */
2315 	if (opti.ts_present && (tiflags & TH_RST) == 0 && tp->ts_recent &&
2316 	    TSTMP_LT(opti.ts_val, tp->ts_recent)) {
2317 
2318 		/* Check to see if ts_recent is over 24 days old.  */
2319 		if (tcp_now - tp->ts_recent_age > TCP_PAWS_IDLE) {
2320 			/*
2321 			 * Invalidate ts_recent.  If this segment updates
2322 			 * ts_recent, the age will be reset later and ts_recent
2323 			 * will get a valid value.  If it does not, setting
2324 			 * ts_recent to zero will at least satisfy the
2325 			 * requirement that zero be placed in the timestamp
2326 			 * echo reply when ts_recent isn't valid.  The
2327 			 * age isn't reset until we get a valid ts_recent
2328 			 * because we don't want out-of-order segments to be
2329 			 * dropped when ts_recent is old.
2330 			 */
2331 			tp->ts_recent = 0;
2332 		} else {
2333 			tcps = TCP_STAT_GETREF();
2334 			tcps[TCP_STAT_RCVDUPPACK]++;
2335 			tcps[TCP_STAT_RCVDUPBYTE] += tlen;
2336 			tcps[TCP_STAT_PAWSDROP]++;
2337 			TCP_STAT_PUTREF();
2338 			tcp_new_dsack(tp, th->th_seq, tlen);
2339 			goto dropafterack;
2340 		}
2341 	}
2342 
2343 	todrop = tp->rcv_nxt - th->th_seq;
2344 	dupseg = false;
2345 	if (todrop > 0) {
2346 		if (tiflags & TH_SYN) {
2347 			tiflags &= ~TH_SYN;
2348 			th->th_seq++;
2349 			if (th->th_urp > 1)
2350 				th->th_urp--;
2351 			else {
2352 				tiflags &= ~TH_URG;
2353 				th->th_urp = 0;
2354 			}
2355 			todrop--;
2356 		}
2357 		if (todrop > tlen ||
2358 		    (todrop == tlen && (tiflags & TH_FIN) == 0)) {
2359 			/*
2360 			 * Any valid FIN or RST must be to the left of the
2361 			 * window.  At this point the FIN or RST must be a
2362 			 * duplicate or out of sequence; drop it.
2363 			 */
2364 			if (tiflags & TH_RST)
2365 				goto drop;
2366 			tiflags &= ~(TH_FIN|TH_RST);
2367 			/*
2368 			 * Send an ACK to resynchronize and drop any data.
2369 			 * But keep on processing for RST or ACK.
2370 			 */
2371 			tp->t_flags |= TF_ACKNOW;
2372 			todrop = tlen;
2373 			dupseg = true;
2374 			tcps = TCP_STAT_GETREF();
2375 			tcps[TCP_STAT_RCVDUPPACK]++;
2376 			tcps[TCP_STAT_RCVDUPBYTE] += todrop;
2377 			TCP_STAT_PUTREF();
2378 		} else if ((tiflags & TH_RST) &&
2379 			   th->th_seq != tp->rcv_nxt) {
2380 			/*
2381 			 * Test for reset before adjusting the sequence
2382 			 * number for overlapping data.
2383 			 */
2384 			goto dropafterack_ratelim;
2385 		} else {
2386 			tcps = TCP_STAT_GETREF();
2387 			tcps[TCP_STAT_RCVPARTDUPPACK]++;
2388 			tcps[TCP_STAT_RCVPARTDUPBYTE] += todrop;
2389 			TCP_STAT_PUTREF();
2390 		}
2391 		tcp_new_dsack(tp, th->th_seq, todrop);
2392 		hdroptlen += todrop;	/*drop from head afterwards*/
2393 		th->th_seq += todrop;
2394 		tlen -= todrop;
2395 		if (th->th_urp > todrop)
2396 			th->th_urp -= todrop;
2397 		else {
2398 			tiflags &= ~TH_URG;
2399 			th->th_urp = 0;
2400 		}
2401 	}
2402 
2403 	/*
2404 	 * If new data are received on a connection after the
2405 	 * user processes are gone, then RST the other end.
2406 	 */
2407 	if ((so->so_state & SS_NOFDREF) &&
2408 	    tp->t_state > TCPS_CLOSE_WAIT && tlen) {
2409 		tp = tcp_close(tp);
2410 		TCP_STATINC(TCP_STAT_RCVAFTERCLOSE);
2411 		goto dropwithreset;
2412 	}
2413 
2414 	/*
2415 	 * If segment ends after window, drop trailing data
2416 	 * (and PUSH and FIN); if nothing left, just ACK.
2417 	 */
2418 	todrop = (th->th_seq + tlen) - (tp->rcv_nxt+tp->rcv_wnd);
2419 	if (todrop > 0) {
2420 		TCP_STATINC(TCP_STAT_RCVPACKAFTERWIN);
2421 		if (todrop >= tlen) {
2422 			/*
2423 			 * The segment actually starts after the window.
2424 			 * th->th_seq + tlen - tp->rcv_nxt - tp->rcv_wnd >= tlen
2425 			 * th->th_seq - tp->rcv_nxt - tp->rcv_wnd >= 0
2426 			 * th->th_seq >= tp->rcv_nxt + tp->rcv_wnd
2427 			 */
2428 			TCP_STATADD(TCP_STAT_RCVBYTEAFTERWIN, tlen);
2429 			/*
2430 			 * If a new connection request is received
2431 			 * while in TIME_WAIT, drop the old connection
2432 			 * and start over if the sequence numbers
2433 			 * are above the previous ones.
2434 			 *
2435 			 * NOTE: We will checksum the packet again, and
2436 			 * so we need to put the header fields back into
2437 			 * network order!
2438 			 * XXX This kind of sucks, but we don't expect
2439 			 * XXX this to happen very often, so maybe it
2440 			 * XXX doesn't matter so much.
2441 			 */
2442 			if (tiflags & TH_SYN &&
2443 			    tp->t_state == TCPS_TIME_WAIT &&
2444 			    SEQ_GT(th->th_seq, tp->rcv_nxt)) {
2445 				tp = tcp_close(tp);
2446 				tcp_fields_to_net(th);
2447 				goto findpcb;
2448 			}
2449 			/*
2450 			 * If window is closed can only take segments at
2451 			 * window edge, and have to drop data and PUSH from
2452 			 * incoming segments.  Continue processing, but
2453 			 * remember to ack.  Otherwise, drop segment
2454 			 * and (if not RST) ack.
2455 			 */
2456 			if (tp->rcv_wnd == 0 && th->th_seq == tp->rcv_nxt) {
2457 				tp->t_flags |= TF_ACKNOW;
2458 				TCP_STATINC(TCP_STAT_RCVWINPROBE);
2459 			} else
2460 				goto dropafterack;
2461 		} else
2462 			TCP_STATADD(TCP_STAT_RCVBYTEAFTERWIN, todrop);
2463 		m_adj(m, -todrop);
2464 		tlen -= todrop;
2465 		tiflags &= ~(TH_PUSH|TH_FIN);
2466 	}
2467 
2468 	/*
2469 	 * If last ACK falls within this segment's sequence numbers,
2470 	 *  record the timestamp.
2471 	 * NOTE:
2472 	 * 1) That the test incorporates suggestions from the latest
2473 	 *    proposal of the tcplw@cray.com list (Braden 1993/04/26).
2474 	 * 2) That updating only on newer timestamps interferes with
2475 	 *    our earlier PAWS tests, so this check should be solely
2476 	 *    predicated on the sequence space of this segment.
2477 	 * 3) That we modify the segment boundary check to be
2478 	 *        Last.ACK.Sent <= SEG.SEQ + SEG.Len
2479 	 *    instead of RFC1323's
2480 	 *        Last.ACK.Sent < SEG.SEQ + SEG.Len,
2481 	 *    This modified check allows us to overcome RFC1323's
2482 	 *    limitations as described in Stevens TCP/IP Illustrated
2483 	 *    Vol. 2 p.869. In such cases, we can still calculate the
2484 	 *    RTT correctly when RCV.NXT == Last.ACK.Sent.
2485 	 */
2486 	if (opti.ts_present &&
2487 	    SEQ_LEQ(th->th_seq, tp->last_ack_sent) &&
2488 	    SEQ_LEQ(tp->last_ack_sent, th->th_seq + tlen +
2489 		    ((tiflags & (TH_SYN|TH_FIN)) != 0))) {
2490 		tp->ts_recent_age = tcp_now;
2491 		tp->ts_recent = opti.ts_val;
2492 	}
2493 
2494 	/*
2495 	 * If the RST bit is set examine the state:
2496 	 *    SYN_RECEIVED STATE:
2497 	 *	If passive open, return to LISTEN state.
2498 	 *	If active open, inform user that connection was refused.
2499 	 *    ESTABLISHED, FIN_WAIT_1, FIN_WAIT2, CLOSE_WAIT STATES:
2500 	 *	Inform user that connection was reset, and close tcb.
2501 	 *    CLOSING, LAST_ACK, TIME_WAIT STATES
2502 	 *	Close the tcb.
2503 	 */
2504 	if (tiflags & TH_RST) {
2505 		if (th->th_seq != tp->rcv_nxt)
2506 			goto dropafterack_ratelim;
2507 
2508 		switch (tp->t_state) {
2509 		case TCPS_SYN_RECEIVED:
2510 			so->so_error = ECONNREFUSED;
2511 			goto close;
2512 
2513 		case TCPS_ESTABLISHED:
2514 		case TCPS_FIN_WAIT_1:
2515 		case TCPS_FIN_WAIT_2:
2516 		case TCPS_CLOSE_WAIT:
2517 			so->so_error = ECONNRESET;
2518 		close:
2519 			tp->t_state = TCPS_CLOSED;
2520 			TCP_STATINC(TCP_STAT_DROPS);
2521 			tp = tcp_close(tp);
2522 			goto drop;
2523 
2524 		case TCPS_CLOSING:
2525 		case TCPS_LAST_ACK:
2526 		case TCPS_TIME_WAIT:
2527 			tp = tcp_close(tp);
2528 			goto drop;
2529 		}
2530 	}
2531 
2532 	/*
2533 	 * Since we've covered the SYN-SENT and SYN-RECEIVED states above
2534 	 * we must be in a synchronized state.  RFC791 states (under RST
2535 	 * generation) that any unacceptable segment (an out-of-order SYN
2536 	 * qualifies) received in a synchronized state must elicit only an
2537 	 * empty acknowledgment segment ... and the connection remains in
2538 	 * the same state.
2539 	 */
2540 	if (tiflags & TH_SYN) {
2541 		if (tp->rcv_nxt == th->th_seq) {
2542 			tcp_respond(tp, m, m, th, (tcp_seq)0, th->th_ack - 1,
2543 			    TH_ACK);
2544 			if (tcp_saveti)
2545 				m_freem(tcp_saveti);
2546 			return;
2547 		}
2548 
2549 		goto dropafterack_ratelim;
2550 	}
2551 
2552 	/*
2553 	 * If the ACK bit is off we drop the segment and return.
2554 	 */
2555 	if ((tiflags & TH_ACK) == 0) {
2556 		if (tp->t_flags & TF_ACKNOW)
2557 			goto dropafterack;
2558 		else
2559 			goto drop;
2560 	}
2561 
2562 	/*
2563 	 * Ack processing.
2564 	 */
2565 	switch (tp->t_state) {
2566 
2567 	/*
2568 	 * In SYN_RECEIVED state if the ack ACKs our SYN then enter
2569 	 * ESTABLISHED state and continue processing, otherwise
2570 	 * send an RST.
2571 	 */
2572 	case TCPS_SYN_RECEIVED:
2573 		if (SEQ_GT(tp->snd_una, th->th_ack) ||
2574 		    SEQ_GT(th->th_ack, tp->snd_max))
2575 			goto dropwithreset;
2576 		TCP_STATINC(TCP_STAT_CONNECTS);
2577 		soisconnected(so);
2578 		tcp_established(tp);
2579 		/* Do window scaling? */
2580 		if ((tp->t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) ==
2581 		    (TF_RCVD_SCALE|TF_REQ_SCALE)) {
2582 			tp->snd_scale = tp->requested_s_scale;
2583 			tp->rcv_scale = tp->request_r_scale;
2584 		}
2585 		TCP_REASS_LOCK(tp);
2586 		(void) tcp_reass(tp, NULL, NULL, &tlen);
2587 		tp->snd_wl1 = th->th_seq - 1;
2588 		/* fall into ... */
2589 
2590 	/*
2591 	 * In ESTABLISHED state: drop duplicate ACKs; ACK out of range
2592 	 * ACKs.  If the ack is in the range
2593 	 *	tp->snd_una < th->th_ack <= tp->snd_max
2594 	 * then advance tp->snd_una to th->th_ack and drop
2595 	 * data from the retransmission queue.  If this ACK reflects
2596 	 * more up to date window information we update our window information.
2597 	 */
2598 	case TCPS_ESTABLISHED:
2599 	case TCPS_FIN_WAIT_1:
2600 	case TCPS_FIN_WAIT_2:
2601 	case TCPS_CLOSE_WAIT:
2602 	case TCPS_CLOSING:
2603 	case TCPS_LAST_ACK:
2604 	case TCPS_TIME_WAIT:
2605 
2606 		if (SEQ_LEQ(th->th_ack, tp->snd_una)) {
2607 			if (tlen == 0 && !dupseg && tiwin == tp->snd_wnd) {
2608 				TCP_STATINC(TCP_STAT_RCVDUPACK);
2609 				/*
2610 				 * If we have outstanding data (other than
2611 				 * a window probe), this is a completely
2612 				 * duplicate ack (ie, window info didn't
2613 				 * change), the ack is the biggest we've
2614 				 * seen and we've seen exactly our rexmt
2615 				 * threshhold of them, assume a packet
2616 				 * has been dropped and retransmit it.
2617 				 * Kludge snd_nxt & the congestion
2618 				 * window so we send only this one
2619 				 * packet.
2620 				 */
2621 				if (TCP_TIMER_ISARMED(tp, TCPT_REXMT) == 0 ||
2622 				    th->th_ack != tp->snd_una)
2623 					tp->t_dupacks = 0;
2624 				else if (tp->t_partialacks < 0 &&
2625 					 (++tp->t_dupacks == tcprexmtthresh ||
2626 					 TCP_FACK_FASTRECOV(tp))) {
2627 					/*
2628 					 * Do the fast retransmit, and adjust
2629 					 * congestion control paramenters.
2630 					 */
2631 					if (tp->t_congctl->fast_retransmit(tp, th)) {
2632 						/* False fast retransmit */
2633 						break;
2634 					} else
2635 						goto drop;
2636 				} else if (tp->t_dupacks > tcprexmtthresh) {
2637 					tp->snd_cwnd += tp->t_segsz;
2638 					KERNEL_LOCK(1, NULL);
2639 					(void) tcp_output(tp);
2640 					KERNEL_UNLOCK_ONE(NULL);
2641 					goto drop;
2642 				}
2643 			} else {
2644 				/*
2645 				 * If the ack appears to be very old, only
2646 				 * allow data that is in-sequence.  This
2647 				 * makes it somewhat more difficult to insert
2648 				 * forged data by guessing sequence numbers.
2649 				 * Sent an ack to try to update the send
2650 				 * sequence number on the other side.
2651 				 */
2652 				if (tlen && th->th_seq != tp->rcv_nxt &&
2653 				    SEQ_LT(th->th_ack,
2654 				    tp->snd_una - tp->max_sndwnd))
2655 					goto dropafterack;
2656 			}
2657 			break;
2658 		}
2659 		/*
2660 		 * If the congestion window was inflated to account
2661 		 * for the other side's cached packets, retract it.
2662 		 */
2663 		tp->t_congctl->fast_retransmit_newack(tp, th);
2664 
2665 		if (SEQ_GT(th->th_ack, tp->snd_max)) {
2666 			TCP_STATINC(TCP_STAT_RCVACKTOOMUCH);
2667 			goto dropafterack;
2668 		}
2669 		acked = th->th_ack - tp->snd_una;
2670 		tcps = TCP_STAT_GETREF();
2671 		tcps[TCP_STAT_RCVACKPACK]++;
2672 		tcps[TCP_STAT_RCVACKBYTE] += acked;
2673 		TCP_STAT_PUTREF();
2674 
2675 		/*
2676 		 * If we have a timestamp reply, update smoothed
2677 		 * round trip time.  If no timestamp is present but
2678 		 * transmit timer is running and timed sequence
2679 		 * number was acked, update smoothed round trip time.
2680 		 * Since we now have an rtt measurement, cancel the
2681 		 * timer backoff (cf., Phil Karn's retransmit alg.).
2682 		 * Recompute the initial retransmit timer.
2683 		 */
2684 		if (ts_rtt)
2685 			tcp_xmit_timer(tp, ts_rtt - 1);
2686 		else if (tp->t_rtttime && SEQ_GT(th->th_ack, tp->t_rtseq))
2687 			tcp_xmit_timer(tp, tcp_now - tp->t_rtttime);
2688 
2689 		/*
2690 		 * If all outstanding data is acked, stop retransmit
2691 		 * timer and remember to restart (more output or persist).
2692 		 * If there is more data to be acked, restart retransmit
2693 		 * timer, using current (possibly backed-off) value.
2694 		 */
2695 		if (th->th_ack == tp->snd_max) {
2696 			TCP_TIMER_DISARM(tp, TCPT_REXMT);
2697 			needoutput = 1;
2698 		} else if (TCP_TIMER_ISARMED(tp, TCPT_PERSIST) == 0)
2699 			TCP_TIMER_ARM(tp, TCPT_REXMT, tp->t_rxtcur);
2700 
2701 		/*
2702 		 * New data has been acked, adjust the congestion window.
2703 		 */
2704 		tp->t_congctl->newack(tp, th);
2705 
2706 		nd6_hint(tp);
2707 		if (acked > so->so_snd.sb_cc) {
2708 			tp->snd_wnd -= so->so_snd.sb_cc;
2709 			sbdrop(&so->so_snd, (int)so->so_snd.sb_cc);
2710 			ourfinisacked = 1;
2711 		} else {
2712 			if (acked > (tp->t_lastoff - tp->t_inoff))
2713 				tp->t_lastm = NULL;
2714 			sbdrop(&so->so_snd, acked);
2715 			tp->t_lastoff -= acked;
2716 			tp->snd_wnd -= acked;
2717 			ourfinisacked = 0;
2718 		}
2719 		sowwakeup(so);
2720 
2721 		icmp_check(tp, th, acked);
2722 
2723 		tp->snd_una = th->th_ack;
2724 		if (SEQ_GT(tp->snd_una, tp->snd_fack))
2725 			tp->snd_fack = tp->snd_una;
2726 		if (SEQ_LT(tp->snd_nxt, tp->snd_una))
2727 			tp->snd_nxt = tp->snd_una;
2728 		if (SEQ_LT(tp->snd_high, tp->snd_una))
2729 			tp->snd_high = tp->snd_una;
2730 
2731 		switch (tp->t_state) {
2732 
2733 		/*
2734 		 * In FIN_WAIT_1 STATE in addition to the processing
2735 		 * for the ESTABLISHED state if our FIN is now acknowledged
2736 		 * then enter FIN_WAIT_2.
2737 		 */
2738 		case TCPS_FIN_WAIT_1:
2739 			if (ourfinisacked) {
2740 				/*
2741 				 * If we can't receive any more
2742 				 * data, then closing user can proceed.
2743 				 * Starting the timer is contrary to the
2744 				 * specification, but if we don't get a FIN
2745 				 * we'll hang forever.
2746 				 */
2747 				if (so->so_state & SS_CANTRCVMORE) {
2748 					soisdisconnected(so);
2749 					if (tp->t_maxidle > 0)
2750 						TCP_TIMER_ARM(tp, TCPT_2MSL,
2751 						    tp->t_maxidle);
2752 				}
2753 				tp->t_state = TCPS_FIN_WAIT_2;
2754 			}
2755 			break;
2756 
2757 	 	/*
2758 		 * In CLOSING STATE in addition to the processing for
2759 		 * the ESTABLISHED state if the ACK acknowledges our FIN
2760 		 * then enter the TIME-WAIT state, otherwise ignore
2761 		 * the segment.
2762 		 */
2763 		case TCPS_CLOSING:
2764 			if (ourfinisacked) {
2765 				tp->t_state = TCPS_TIME_WAIT;
2766 				tcp_canceltimers(tp);
2767 				TCP_TIMER_ARM(tp, TCPT_2MSL, 2 * tp->t_msl);
2768 				soisdisconnected(so);
2769 			}
2770 			break;
2771 
2772 		/*
2773 		 * In LAST_ACK, we may still be waiting for data to drain
2774 		 * and/or to be acked, as well as for the ack of our FIN.
2775 		 * If our FIN is now acknowledged, delete the TCB,
2776 		 * enter the closed state and return.
2777 		 */
2778 		case TCPS_LAST_ACK:
2779 			if (ourfinisacked) {
2780 				tp = tcp_close(tp);
2781 				goto drop;
2782 			}
2783 			break;
2784 
2785 		/*
2786 		 * In TIME_WAIT state the only thing that should arrive
2787 		 * is a retransmission of the remote FIN.  Acknowledge
2788 		 * it and restart the finack timer.
2789 		 */
2790 		case TCPS_TIME_WAIT:
2791 			TCP_TIMER_ARM(tp, TCPT_2MSL, 2 * tp->t_msl);
2792 			goto dropafterack;
2793 		}
2794 	}
2795 
2796 step6:
2797 	/*
2798 	 * Update window information.
2799 	 * Don't look at window if no ACK: TAC's send garbage on first SYN.
2800 	 */
2801 	if ((tiflags & TH_ACK) && (SEQ_LT(tp->snd_wl1, th->th_seq) ||
2802 	    (tp->snd_wl1 == th->th_seq && (SEQ_LT(tp->snd_wl2, th->th_ack) ||
2803 	    (tp->snd_wl2 == th->th_ack && tiwin > tp->snd_wnd))))) {
2804 		/* keep track of pure window updates */
2805 		if (tlen == 0 &&
2806 		    tp->snd_wl2 == th->th_ack && tiwin > tp->snd_wnd)
2807 			TCP_STATINC(TCP_STAT_RCVWINUPD);
2808 		tp->snd_wnd = tiwin;
2809 		tp->snd_wl1 = th->th_seq;
2810 		tp->snd_wl2 = th->th_ack;
2811 		if (tp->snd_wnd > tp->max_sndwnd)
2812 			tp->max_sndwnd = tp->snd_wnd;
2813 		needoutput = 1;
2814 	}
2815 
2816 	/*
2817 	 * Process segments with URG.
2818 	 */
2819 	if ((tiflags & TH_URG) && th->th_urp &&
2820 	    TCPS_HAVERCVDFIN(tp->t_state) == 0) {
2821 		/*
2822 		 * This is a kludge, but if we receive and accept
2823 		 * random urgent pointers, we'll crash in
2824 		 * soreceive.  It's hard to imagine someone
2825 		 * actually wanting to send this much urgent data.
2826 		 */
2827 		if (th->th_urp + so->so_rcv.sb_cc > sb_max) {
2828 			th->th_urp = 0;			/* XXX */
2829 			tiflags &= ~TH_URG;		/* XXX */
2830 			goto dodata;			/* XXX */
2831 		}
2832 		/*
2833 		 * If this segment advances the known urgent pointer,
2834 		 * then mark the data stream.  This should not happen
2835 		 * in CLOSE_WAIT, CLOSING, LAST_ACK or TIME_WAIT STATES since
2836 		 * a FIN has been received from the remote side.
2837 		 * In these states we ignore the URG.
2838 		 *
2839 		 * According to RFC961 (Assigned Protocols),
2840 		 * the urgent pointer points to the last octet
2841 		 * of urgent data.  We continue, however,
2842 		 * to consider it to indicate the first octet
2843 		 * of data past the urgent section as the original
2844 		 * spec states (in one of two places).
2845 		 */
2846 		if (SEQ_GT(th->th_seq+th->th_urp, tp->rcv_up)) {
2847 			tp->rcv_up = th->th_seq + th->th_urp;
2848 			so->so_oobmark = so->so_rcv.sb_cc +
2849 			    (tp->rcv_up - tp->rcv_nxt) - 1;
2850 			if (so->so_oobmark == 0)
2851 				so->so_state |= SS_RCVATMARK;
2852 			sohasoutofband(so);
2853 			tp->t_oobflags &= ~(TCPOOB_HAVEDATA | TCPOOB_HADDATA);
2854 		}
2855 		/*
2856 		 * Remove out of band data so doesn't get presented to user.
2857 		 * This can happen independent of advancing the URG pointer,
2858 		 * but if two URG's are pending at once, some out-of-band
2859 		 * data may creep in... ick.
2860 		 */
2861 		if (th->th_urp <= (u_int16_t) tlen
2862 #ifdef SO_OOBINLINE
2863 		     && (so->so_options & SO_OOBINLINE) == 0
2864 #endif
2865 		     )
2866 			tcp_pulloutofband(so, th, m, hdroptlen);
2867 	} else
2868 		/*
2869 		 * If no out of band data is expected,
2870 		 * pull receive urgent pointer along
2871 		 * with the receive window.
2872 		 */
2873 		if (SEQ_GT(tp->rcv_nxt, tp->rcv_up))
2874 			tp->rcv_up = tp->rcv_nxt;
2875 dodata:							/* XXX */
2876 
2877 	/*
2878 	 * Process the segment text, merging it into the TCP sequencing queue,
2879 	 * and arranging for acknowledgement of receipt if necessary.
2880 	 * This process logically involves adjusting tp->rcv_wnd as data
2881 	 * is presented to the user (this happens in tcp_usrreq.c,
2882 	 * tcp_rcvd()).  If a FIN has already been received on this
2883 	 * connection then we just ignore the text.
2884 	 */
2885 	if ((tlen || (tiflags & TH_FIN)) &&
2886 	    TCPS_HAVERCVDFIN(tp->t_state) == 0) {
2887 		/*
2888 		 * Insert segment ti into reassembly queue of tcp with
2889 		 * control block tp.  Return TH_FIN if reassembly now includes
2890 		 * a segment with FIN.  The macro form does the common case
2891 		 * inline (segment is the next to be received on an
2892 		 * established connection, and the queue is empty),
2893 		 * avoiding linkage into and removal from the queue and
2894 		 * repetition of various conversions.
2895 		 * Set DELACK for segments received in order, but ack
2896 		 * immediately when segments are out of order
2897 		 * (so fast retransmit can work).
2898 		 */
2899 		/* NOTE: this was TCP_REASS() macro, but used only once */
2900 		TCP_REASS_LOCK(tp);
2901 		if (th->th_seq == tp->rcv_nxt &&
2902 		    TAILQ_FIRST(&tp->segq) == NULL &&
2903 		    tp->t_state == TCPS_ESTABLISHED) {
2904 			tcp_setup_ack(tp, th);
2905 			tp->rcv_nxt += tlen;
2906 			tiflags = th->th_flags & TH_FIN;
2907 			tcps = TCP_STAT_GETREF();
2908 			tcps[TCP_STAT_RCVPACK]++;
2909 			tcps[TCP_STAT_RCVBYTE] += tlen;
2910 			TCP_STAT_PUTREF();
2911 			nd6_hint(tp);
2912 			if (so->so_state & SS_CANTRCVMORE)
2913 				m_freem(m);
2914 			else {
2915 				m_adj(m, hdroptlen);
2916 				sbappendstream(&(so)->so_rcv, m);
2917 			}
2918 			TCP_REASS_UNLOCK(tp);
2919 			sorwakeup(so);
2920 		} else {
2921 			m_adj(m, hdroptlen);
2922 			tiflags = tcp_reass(tp, th, m, &tlen);
2923 			tp->t_flags |= TF_ACKNOW;
2924 		}
2925 
2926 		/*
2927 		 * Note the amount of data that peer has sent into
2928 		 * our window, in order to estimate the sender's
2929 		 * buffer size.
2930 		 */
2931 		len = so->so_rcv.sb_hiwat - (tp->rcv_adv - tp->rcv_nxt);
2932 	} else {
2933 		m_freem(m);
2934 		m = NULL;
2935 		tiflags &= ~TH_FIN;
2936 	}
2937 
2938 	/*
2939 	 * If FIN is received ACK the FIN and let the user know
2940 	 * that the connection is closing.  Ignore a FIN received before
2941 	 * the connection is fully established.
2942 	 */
2943 	if ((tiflags & TH_FIN) && TCPS_HAVEESTABLISHED(tp->t_state)) {
2944 		if (TCPS_HAVERCVDFIN(tp->t_state) == 0) {
2945 			socantrcvmore(so);
2946 			tp->t_flags |= TF_ACKNOW;
2947 			tp->rcv_nxt++;
2948 		}
2949 		switch (tp->t_state) {
2950 
2951 	 	/*
2952 		 * In ESTABLISHED STATE enter the CLOSE_WAIT state.
2953 		 */
2954 		case TCPS_ESTABLISHED:
2955 			tp->t_state = TCPS_CLOSE_WAIT;
2956 			break;
2957 
2958 	 	/*
2959 		 * If still in FIN_WAIT_1 STATE FIN has not been acked so
2960 		 * enter the CLOSING state.
2961 		 */
2962 		case TCPS_FIN_WAIT_1:
2963 			tp->t_state = TCPS_CLOSING;
2964 			break;
2965 
2966 	 	/*
2967 		 * In FIN_WAIT_2 state enter the TIME_WAIT state,
2968 		 * starting the time-wait timer, turning off the other
2969 		 * standard timers.
2970 		 */
2971 		case TCPS_FIN_WAIT_2:
2972 			tp->t_state = TCPS_TIME_WAIT;
2973 			tcp_canceltimers(tp);
2974 			TCP_TIMER_ARM(tp, TCPT_2MSL, 2 * tp->t_msl);
2975 			soisdisconnected(so);
2976 			break;
2977 
2978 		/*
2979 		 * In TIME_WAIT state restart the 2 MSL time_wait timer.
2980 		 */
2981 		case TCPS_TIME_WAIT:
2982 			TCP_TIMER_ARM(tp, TCPT_2MSL, 2 * tp->t_msl);
2983 			break;
2984 		}
2985 	}
2986 #ifdef TCP_DEBUG
2987 	if (so->so_options & SO_DEBUG)
2988 		tcp_trace(TA_INPUT, ostate, tp, tcp_saveti, 0);
2989 #endif
2990 
2991 	/*
2992 	 * Return any desired output.
2993 	 */
2994 	if (needoutput || (tp->t_flags & TF_ACKNOW)) {
2995 		KERNEL_LOCK(1, NULL);
2996 		(void) tcp_output(tp);
2997 		KERNEL_UNLOCK_ONE(NULL);
2998 	}
2999 	if (tcp_saveti)
3000 		m_freem(tcp_saveti);
3001 
3002 	if (tp->t_state == TCPS_TIME_WAIT
3003 	    && (so->so_state & SS_NOFDREF)
3004 	    && (tp->t_inpcb || af != AF_INET)
3005 	    && (tp->t_in6pcb || af != AF_INET6)
3006 	    && ((af == AF_INET ? tcp4_vtw_enable : tcp6_vtw_enable) & 1) != 0
3007 	    && TAILQ_EMPTY(&tp->segq)
3008 	    && vtw_add(af, tp)) {
3009 		;
3010 	}
3011 	return;
3012 
3013 badsyn:
3014 	/*
3015 	 * Received a bad SYN.  Increment counters and dropwithreset.
3016 	 */
3017 	TCP_STATINC(TCP_STAT_BADSYN);
3018 	tp = NULL;
3019 	goto dropwithreset;
3020 
3021 dropafterack:
3022 	/*
3023 	 * Generate an ACK dropping incoming segment if it occupies
3024 	 * sequence space, where the ACK reflects our state.
3025 	 */
3026 	if (tiflags & TH_RST)
3027 		goto drop;
3028 	goto dropafterack2;
3029 
3030 dropafterack_ratelim:
3031 	/*
3032 	 * We may want to rate-limit ACKs against SYN/RST attack.
3033 	 */
3034 	if (ppsratecheck(&tcp_ackdrop_ppslim_last, &tcp_ackdrop_ppslim_count,
3035 	    tcp_ackdrop_ppslim) == 0) {
3036 		/* XXX stat */
3037 		goto drop;
3038 	}
3039 	/* ...fall into dropafterack2... */
3040 
3041 dropafterack2:
3042 	m_freem(m);
3043 	tp->t_flags |= TF_ACKNOW;
3044 	KERNEL_LOCK(1, NULL);
3045 	(void) tcp_output(tp);
3046 	KERNEL_UNLOCK_ONE(NULL);
3047 	if (tcp_saveti)
3048 		m_freem(tcp_saveti);
3049 	return;
3050 
3051 dropwithreset_ratelim:
3052 	/*
3053 	 * We may want to rate-limit RSTs in certain situations,
3054 	 * particularly if we are sending an RST in response to
3055 	 * an attempt to connect to or otherwise communicate with
3056 	 * a port for which we have no socket.
3057 	 */
3058 	if (ppsratecheck(&tcp_rst_ppslim_last, &tcp_rst_ppslim_count,
3059 	    tcp_rst_ppslim) == 0) {
3060 		/* XXX stat */
3061 		goto drop;
3062 	}
3063 	/* ...fall into dropwithreset... */
3064 
3065 dropwithreset:
3066 	/*
3067 	 * Generate a RST, dropping incoming segment.
3068 	 * Make ACK acceptable to originator of segment.
3069 	 */
3070 	if (tiflags & TH_RST)
3071 		goto drop;
3072 
3073 	switch (af) {
3074 #ifdef INET6
3075 	case AF_INET6:
3076 		/* For following calls to tcp_respond */
3077 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
3078 			goto drop;
3079 		break;
3080 #endif /* INET6 */
3081 	case AF_INET:
3082 		if (IN_MULTICAST(ip->ip_dst.s_addr) ||
3083 		    in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif))
3084 			goto drop;
3085 	}
3086 
3087 	if (tiflags & TH_ACK)
3088 		(void)tcp_respond(tp, m, m, th, (tcp_seq)0, th->th_ack, TH_RST);
3089 	else {
3090 		if (tiflags & TH_SYN)
3091 			tlen++;
3092 		(void)tcp_respond(tp, m, m, th, th->th_seq + tlen, (tcp_seq)0,
3093 		    TH_RST|TH_ACK);
3094 	}
3095 	if (tcp_saveti)
3096 		m_freem(tcp_saveti);
3097 	return;
3098 
3099 badcsum:
3100 drop:
3101 	/*
3102 	 * Drop space held by incoming segment and return.
3103 	 */
3104 	if (tp) {
3105 		if (tp->t_inpcb)
3106 			so = tp->t_inpcb->inp_socket;
3107 #ifdef INET6
3108 		else if (tp->t_in6pcb)
3109 			so = tp->t_in6pcb->in6p_socket;
3110 #endif
3111 		else
3112 			so = NULL;
3113 #ifdef TCP_DEBUG
3114 		if (so && (so->so_options & SO_DEBUG) != 0)
3115 			tcp_trace(TA_DROP, ostate, tp, tcp_saveti, 0);
3116 #endif
3117 	}
3118 	if (tcp_saveti)
3119 		m_freem(tcp_saveti);
3120 	m_freem(m);
3121 	return;
3122 }
3123 
3124 #ifdef TCP_SIGNATURE
3125 int
3126 tcp_signature_apply(void *fstate, void *data, u_int len)
3127 {
3128 
3129 	MD5Update(fstate, (u_char *)data, len);
3130 	return (0);
3131 }
3132 
3133 struct secasvar *
3134 tcp_signature_getsav(struct mbuf *m, struct tcphdr *th)
3135 {
3136 	struct ip *ip;
3137 	struct ip6_hdr *ip6;
3138 
3139 	ip = mtod(m, struct ip *);
3140 	switch (ip->ip_v) {
3141 	case 4:
3142 		ip = mtod(m, struct ip *);
3143 		ip6 = NULL;
3144 		break;
3145 	case 6:
3146 		ip = NULL;
3147 		ip6 = mtod(m, struct ip6_hdr *);
3148 		break;
3149 	default:
3150 		return (NULL);
3151 	}
3152 
3153 #ifdef IPSEC
3154 	if (ipsec_used) {
3155 		union sockaddr_union dst;
3156 		/* Extract the destination from the IP header in the mbuf. */
3157 		memset(&dst, 0, sizeof(union sockaddr_union));
3158 		if (ip != NULL) {
3159 			dst.sa.sa_len = sizeof(struct sockaddr_in);
3160 			dst.sa.sa_family = AF_INET;
3161 			dst.sin.sin_addr = ip->ip_dst;
3162 		} else {
3163 			dst.sa.sa_len = sizeof(struct sockaddr_in6);
3164 			dst.sa.sa_family = AF_INET6;
3165 			dst.sin6.sin6_addr = ip6->ip6_dst;
3166 		}
3167 
3168 		/*
3169 		 * Look up an SADB entry which matches the address of the peer.
3170 		 */
3171 		return KEY_ALLOCSA(&dst, IPPROTO_TCP, htonl(TCP_SIG_SPI), 0, 0);
3172 	}
3173 	return NULL;
3174 #else
3175 	if (ip)
3176 		return key_allocsa(AF_INET, (void *)&ip->ip_src,
3177 		    (void *)&ip->ip_dst, IPPROTO_TCP,
3178 		    htonl(TCP_SIG_SPI), 0, 0);
3179 	else
3180 		return key_allocsa(AF_INET6, (void *)&ip6->ip6_src,
3181 		    (void *)&ip6->ip6_dst, IPPROTO_TCP,
3182 		    htonl(TCP_SIG_SPI), 0, 0);
3183 #endif
3184 }
3185 
3186 int
3187 tcp_signature(struct mbuf *m, struct tcphdr *th, int thoff,
3188     struct secasvar *sav, char *sig)
3189 {
3190 	MD5_CTX ctx;
3191 	struct ip *ip;
3192 	struct ipovly *ipovly;
3193 	struct ip6_hdr *ip6;
3194 	struct ippseudo ippseudo;
3195 	struct ip6_hdr_pseudo ip6pseudo;
3196 	struct tcphdr th0;
3197 	int l, tcphdrlen;
3198 
3199 	if (sav == NULL)
3200 		return (-1);
3201 
3202 	tcphdrlen = th->th_off * 4;
3203 
3204 	switch (mtod(m, struct ip *)->ip_v) {
3205 	case 4:
3206 		ip = mtod(m, struct ip *);
3207 		ip6 = NULL;
3208 		break;
3209 	case 6:
3210 		ip = NULL;
3211 		ip6 = mtod(m, struct ip6_hdr *);
3212 		break;
3213 	default:
3214 		return (-1);
3215 	}
3216 
3217 	MD5Init(&ctx);
3218 
3219 	if (ip) {
3220 		memset(&ippseudo, 0, sizeof(ippseudo));
3221 		ipovly = (struct ipovly *)ip;
3222 		ippseudo.ippseudo_src = ipovly->ih_src;
3223 		ippseudo.ippseudo_dst = ipovly->ih_dst;
3224 		ippseudo.ippseudo_pad = 0;
3225 		ippseudo.ippseudo_p = IPPROTO_TCP;
3226 		ippseudo.ippseudo_len = htons(m->m_pkthdr.len - thoff);
3227 		MD5Update(&ctx, (char *)&ippseudo, sizeof(ippseudo));
3228 	} else {
3229 		memset(&ip6pseudo, 0, sizeof(ip6pseudo));
3230 		ip6pseudo.ip6ph_src = ip6->ip6_src;
3231 		in6_clearscope(&ip6pseudo.ip6ph_src);
3232 		ip6pseudo.ip6ph_dst = ip6->ip6_dst;
3233 		in6_clearscope(&ip6pseudo.ip6ph_dst);
3234 		ip6pseudo.ip6ph_len = htons(m->m_pkthdr.len - thoff);
3235 		ip6pseudo.ip6ph_nxt = IPPROTO_TCP;
3236 		MD5Update(&ctx, (char *)&ip6pseudo, sizeof(ip6pseudo));
3237 	}
3238 
3239 	th0 = *th;
3240 	th0.th_sum = 0;
3241 	MD5Update(&ctx, (char *)&th0, sizeof(th0));
3242 
3243 	l = m->m_pkthdr.len - thoff - tcphdrlen;
3244 	if (l > 0)
3245 		m_apply(m, thoff + tcphdrlen,
3246 		    m->m_pkthdr.len - thoff - tcphdrlen,
3247 		    tcp_signature_apply, &ctx);
3248 
3249 	MD5Update(&ctx, _KEYBUF(sav->key_auth), _KEYLEN(sav->key_auth));
3250 	MD5Final(sig, &ctx);
3251 
3252 	return (0);
3253 }
3254 #endif
3255 
3256 /*
3257  * tcp_dooptions: parse and process tcp options.
3258  *
3259  * returns -1 if this segment should be dropped.  (eg. wrong signature)
3260  * otherwise returns 0.
3261  */
3262 
3263 static int
3264 tcp_dooptions(struct tcpcb *tp, const u_char *cp, int cnt,
3265     struct tcphdr *th,
3266     struct mbuf *m, int toff, struct tcp_opt_info *oi)
3267 {
3268 	u_int16_t mss;
3269 	int opt, optlen = 0;
3270 #ifdef TCP_SIGNATURE
3271 	void *sigp = NULL;
3272 	char sigbuf[TCP_SIGLEN];
3273 	struct secasvar *sav = NULL;
3274 #endif
3275 
3276 	for (; cp && cnt > 0; cnt -= optlen, cp += optlen) {
3277 		opt = cp[0];
3278 		if (opt == TCPOPT_EOL)
3279 			break;
3280 		if (opt == TCPOPT_NOP)
3281 			optlen = 1;
3282 		else {
3283 			if (cnt < 2)
3284 				break;
3285 			optlen = cp[1];
3286 			if (optlen < 2 || optlen > cnt)
3287 				break;
3288 		}
3289 		switch (opt) {
3290 
3291 		default:
3292 			continue;
3293 
3294 		case TCPOPT_MAXSEG:
3295 			if (optlen != TCPOLEN_MAXSEG)
3296 				continue;
3297 			if (!(th->th_flags & TH_SYN))
3298 				continue;
3299 			if (TCPS_HAVERCVDSYN(tp->t_state))
3300 				continue;
3301 			bcopy(cp + 2, &mss, sizeof(mss));
3302 			oi->maxseg = ntohs(mss);
3303 			break;
3304 
3305 		case TCPOPT_WINDOW:
3306 			if (optlen != TCPOLEN_WINDOW)
3307 				continue;
3308 			if (!(th->th_flags & TH_SYN))
3309 				continue;
3310 			if (TCPS_HAVERCVDSYN(tp->t_state))
3311 				continue;
3312 			tp->t_flags |= TF_RCVD_SCALE;
3313 			tp->requested_s_scale = cp[2];
3314 			if (tp->requested_s_scale > TCP_MAX_WINSHIFT) {
3315 				char buf[INET6_ADDRSTRLEN];
3316 				struct ip *ip = mtod(m, struct ip *);
3317 #ifdef INET6
3318 				struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
3319 #endif
3320 				if (ip)
3321 					in_print(buf, sizeof(buf),
3322 					    &ip->ip_src);
3323 #ifdef INET6
3324 				else if (ip6)
3325 					in6_print(buf, sizeof(buf),
3326 					    &ip6->ip6_src);
3327 #endif
3328 				else
3329 					strlcpy(buf, "(unknown)", sizeof(buf));
3330 				log(LOG_ERR, "TCP: invalid wscale %d from %s, "
3331 				    "assuming %d\n",
3332 				    tp->requested_s_scale, buf,
3333 				    TCP_MAX_WINSHIFT);
3334 				tp->requested_s_scale = TCP_MAX_WINSHIFT;
3335 			}
3336 			break;
3337 
3338 		case TCPOPT_TIMESTAMP:
3339 			if (optlen != TCPOLEN_TIMESTAMP)
3340 				continue;
3341 			oi->ts_present = 1;
3342 			bcopy(cp + 2, &oi->ts_val, sizeof(oi->ts_val));
3343 			NTOHL(oi->ts_val);
3344 			bcopy(cp + 6, &oi->ts_ecr, sizeof(oi->ts_ecr));
3345 			NTOHL(oi->ts_ecr);
3346 
3347 			if (!(th->th_flags & TH_SYN))
3348 				continue;
3349 			if (TCPS_HAVERCVDSYN(tp->t_state))
3350 				continue;
3351 			/*
3352 			 * A timestamp received in a SYN makes
3353 			 * it ok to send timestamp requests and replies.
3354 			 */
3355 			tp->t_flags |= TF_RCVD_TSTMP;
3356 			tp->ts_recent = oi->ts_val;
3357 			tp->ts_recent_age = tcp_now;
3358                         break;
3359 
3360 		case TCPOPT_SACK_PERMITTED:
3361 			if (optlen != TCPOLEN_SACK_PERMITTED)
3362 				continue;
3363 			if (!(th->th_flags & TH_SYN))
3364 				continue;
3365 			if (TCPS_HAVERCVDSYN(tp->t_state))
3366 				continue;
3367 			if (tcp_do_sack) {
3368 				tp->t_flags |= TF_SACK_PERMIT;
3369 				tp->t_flags |= TF_WILL_SACK;
3370 			}
3371 			break;
3372 
3373 		case TCPOPT_SACK:
3374 			tcp_sack_option(tp, th, cp, optlen);
3375 			break;
3376 #ifdef TCP_SIGNATURE
3377 		case TCPOPT_SIGNATURE:
3378 			if (optlen != TCPOLEN_SIGNATURE)
3379 				continue;
3380 			if (sigp && memcmp(sigp, cp + 2, TCP_SIGLEN))
3381 				return (-1);
3382 
3383 			sigp = sigbuf;
3384 			memcpy(sigbuf, cp + 2, TCP_SIGLEN);
3385 			tp->t_flags |= TF_SIGNATURE;
3386 			break;
3387 #endif
3388 		}
3389 	}
3390 
3391 #ifndef TCP_SIGNATURE
3392 	return 0;
3393 #else
3394 	if (tp->t_flags & TF_SIGNATURE) {
3395 
3396 		sav = tcp_signature_getsav(m, th);
3397 
3398 		if (sav == NULL && tp->t_state == TCPS_LISTEN)
3399 			return (-1);
3400 	}
3401 
3402 	if ((sigp ? TF_SIGNATURE : 0) ^ (tp->t_flags & TF_SIGNATURE))
3403 		goto out;
3404 
3405 	if (sigp) {
3406 		char sig[TCP_SIGLEN];
3407 
3408 		tcp_fields_to_net(th);
3409 		if (tcp_signature(m, th, toff, sav, sig) < 0) {
3410 			tcp_fields_to_host(th);
3411 			goto out;
3412 		}
3413 		tcp_fields_to_host(th);
3414 
3415 		if (memcmp(sig, sigp, TCP_SIGLEN)) {
3416 			TCP_STATINC(TCP_STAT_BADSIG);
3417 			goto out;
3418 		} else
3419 			TCP_STATINC(TCP_STAT_GOODSIG);
3420 
3421 		key_sa_recordxfer(sav, m);
3422 		KEY_FREESAV(&sav);
3423 	}
3424 	return 0;
3425 out:
3426 	if (sav != NULL)
3427 		KEY_FREESAV(&sav);
3428 	return -1;
3429 #endif
3430 }
3431 
3432 /*
3433  * Pull out of band byte out of a segment so
3434  * it doesn't appear in the user's data queue.
3435  * It is still reflected in the segment length for
3436  * sequencing purposes.
3437  */
3438 void
3439 tcp_pulloutofband(struct socket *so, struct tcphdr *th,
3440     struct mbuf *m, int off)
3441 {
3442 	int cnt = off + th->th_urp - 1;
3443 
3444 	while (cnt >= 0) {
3445 		if (m->m_len > cnt) {
3446 			char *cp = mtod(m, char *) + cnt;
3447 			struct tcpcb *tp = sototcpcb(so);
3448 
3449 			tp->t_iobc = *cp;
3450 			tp->t_oobflags |= TCPOOB_HAVEDATA;
3451 			bcopy(cp+1, cp, (unsigned)(m->m_len - cnt - 1));
3452 			m->m_len--;
3453 			return;
3454 		}
3455 		cnt -= m->m_len;
3456 		m = m->m_next;
3457 		if (m == 0)
3458 			break;
3459 	}
3460 	panic("tcp_pulloutofband");
3461 }
3462 
3463 /*
3464  * Collect new round-trip time estimate
3465  * and update averages and current timeout.
3466  *
3467  * rtt is in units of slow ticks (typically 500 ms) -- essentially the
3468  * difference of two timestamps.
3469  */
3470 void
3471 tcp_xmit_timer(struct tcpcb *tp, uint32_t rtt)
3472 {
3473 	int32_t delta;
3474 
3475 	TCP_STATINC(TCP_STAT_RTTUPDATED);
3476 	if (tp->t_srtt != 0) {
3477 		/*
3478 		 * Compute the amount to add to srtt for smoothing,
3479 		 * *alpha, or 2^(-TCP_RTT_SHIFT).  Because
3480 		 * srtt is stored in 1/32 slow ticks, we conceptually
3481 		 * shift left 5 bits, subtract srtt to get the
3482 		 * diference, and then shift right by TCP_RTT_SHIFT
3483 		 * (3) to obtain 1/8 of the difference.
3484 		 */
3485 		delta = (rtt << 2) - (tp->t_srtt >> TCP_RTT_SHIFT);
3486 		/*
3487 		 * This can never happen, because delta's lowest
3488 		 * possible value is 1/8 of t_srtt.  But if it does,
3489 		 * set srtt to some reasonable value, here chosen
3490 		 * as 1/8 tick.
3491 		 */
3492 		if ((tp->t_srtt += delta) <= 0)
3493 			tp->t_srtt = 1 << 2;
3494 		/*
3495 		 * RFC2988 requires that rttvar be updated first.
3496 		 * This code is compliant because "delta" is the old
3497 		 * srtt minus the new observation (scaled).
3498 		 *
3499 		 * RFC2988 says:
3500 		 *   rttvar = (1-beta) * rttvar + beta * |srtt-observed|
3501 		 *
3502 		 * delta is in units of 1/32 ticks, and has then been
3503 		 * divided by 8.  This is equivalent to being in 1/16s
3504 		 * units and divided by 4.  Subtract from it 1/4 of
3505 		 * the existing rttvar to form the (signed) amount to
3506 		 * adjust.
3507 		 */
3508 		if (delta < 0)
3509 			delta = -delta;
3510 		delta -= (tp->t_rttvar >> TCP_RTTVAR_SHIFT);
3511 		/*
3512 		 * As with srtt, this should never happen.  There is
3513 		 * no support in RFC2988 for this operation.  But 1/4s
3514 		 * as rttvar when faced with something arguably wrong
3515 		 * is ok.
3516 		 */
3517 		if ((tp->t_rttvar += delta) <= 0)
3518 			tp->t_rttvar = 1 << 2;
3519 
3520 		/*
3521 		 * If srtt exceeds .01 second, ensure we use the 'remote' MSL
3522 		 * Problem is: it doesn't work.  Disabled by defaulting
3523 		 * tcp_rttlocal to 0; see corresponding code in
3524 		 * tcp_subr that selects local vs remote in a different way.
3525 		 *
3526 		 * The static branch prediction hint here should be removed
3527 		 * when the rtt estimator is fixed and the rtt_enable code
3528 		 * is turned back on.
3529 		 */
3530 		if (__predict_false(tcp_rttlocal) && tcp_msl_enable
3531 		    && tp->t_srtt > tcp_msl_remote_threshold
3532 		    && tp->t_msl  < tcp_msl_remote) {
3533 			tp->t_msl = tcp_msl_remote;
3534 		}
3535 	} else {
3536 		/*
3537 		 * This is the first measurement.  Per RFC2988, 2.2,
3538 		 * set rtt=R and srtt=R/2.
3539 		 * For srtt, storage representation is 1/32 ticks,
3540 		 * so shift left by 5.
3541 		 * For rttvar, storage representation is 1/16 ticks,
3542 		 * So shift left by 4, but then right by 1 to halve.
3543 		 */
3544 		tp->t_srtt = rtt << (TCP_RTT_SHIFT + 2);
3545 		tp->t_rttvar = rtt << (TCP_RTTVAR_SHIFT + 2 - 1);
3546 	}
3547 	tp->t_rtttime = 0;
3548 	tp->t_rxtshift = 0;
3549 
3550 	/*
3551 	 * the retransmit should happen at rtt + 4 * rttvar.
3552 	 * Because of the way we do the smoothing, srtt and rttvar
3553 	 * will each average +1/2 tick of bias.  When we compute
3554 	 * the retransmit timer, we want 1/2 tick of rounding and
3555 	 * 1 extra tick because of +-1/2 tick uncertainty in the
3556 	 * firing of the timer.  The bias will give us exactly the
3557 	 * 1.5 tick we need.  But, because the bias is
3558 	 * statistical, we have to test that we don't drop below
3559 	 * the minimum feasible timer (which is 2 ticks).
3560 	 */
3561 	TCPT_RANGESET(tp->t_rxtcur, TCP_REXMTVAL(tp),
3562 	    max(tp->t_rttmin, rtt + 2), TCPTV_REXMTMAX);
3563 
3564 	/*
3565 	 * We received an ack for a packet that wasn't retransmitted;
3566 	 * it is probably safe to discard any error indications we've
3567 	 * received recently.  This isn't quite right, but close enough
3568 	 * for now (a route might have failed after we sent a segment,
3569 	 * and the return path might not be symmetrical).
3570 	 */
3571 	tp->t_softerror = 0;
3572 }
3573 
3574 
3575 /*
3576  * TCP compressed state engine.  Currently used to hold compressed
3577  * state for SYN_RECEIVED.
3578  */
3579 
3580 u_long	syn_cache_count;
3581 u_int32_t syn_hash1, syn_hash2;
3582 
3583 #define SYN_HASH(sa, sp, dp) \
3584 	((((sa)->s_addr^syn_hash1)*(((((u_int32_t)(dp))<<16) + \
3585 				     ((u_int32_t)(sp)))^syn_hash2)))
3586 #ifndef INET6
3587 #define	SYN_HASHALL(hash, src, dst) \
3588 do {									\
3589 	hash = SYN_HASH(&((const struct sockaddr_in *)(src))->sin_addr,	\
3590 		((const struct sockaddr_in *)(src))->sin_port,		\
3591 		((const struct sockaddr_in *)(dst))->sin_port);		\
3592 } while (/*CONSTCOND*/ 0)
3593 #else
3594 #define SYN_HASH6(sa, sp, dp) \
3595 	((((sa)->s6_addr32[0] ^ (sa)->s6_addr32[3] ^ syn_hash1) * \
3596 	  (((((u_int32_t)(dp))<<16) + ((u_int32_t)(sp)))^syn_hash2)) \
3597 	 & 0x7fffffff)
3598 
3599 #define SYN_HASHALL(hash, src, dst) \
3600 do {									\
3601 	switch ((src)->sa_family) {					\
3602 	case AF_INET:							\
3603 		hash = SYN_HASH(&((const struct sockaddr_in *)(src))->sin_addr, \
3604 			((const struct sockaddr_in *)(src))->sin_port,	\
3605 			((const struct sockaddr_in *)(dst))->sin_port);	\
3606 		break;							\
3607 	case AF_INET6:							\
3608 		hash = SYN_HASH6(&((const struct sockaddr_in6 *)(src))->sin6_addr, \
3609 			((const struct sockaddr_in6 *)(src))->sin6_port,	\
3610 			((const struct sockaddr_in6 *)(dst))->sin6_port);	\
3611 		break;							\
3612 	default:							\
3613 		hash = 0;						\
3614 	}								\
3615 } while (/*CONSTCOND*/0)
3616 #endif /* INET6 */
3617 
3618 static struct pool syn_cache_pool;
3619 
3620 /*
3621  * We don't estimate RTT with SYNs, so each packet starts with the default
3622  * RTT and each timer step has a fixed timeout value.
3623  */
3624 #define	SYN_CACHE_TIMER_ARM(sc)						\
3625 do {									\
3626 	TCPT_RANGESET((sc)->sc_rxtcur,					\
3627 	    TCPTV_SRTTDFLT * tcp_backoff[(sc)->sc_rxtshift], TCPTV_MIN,	\
3628 	    TCPTV_REXMTMAX);						\
3629 	callout_reset(&(sc)->sc_timer,					\
3630 	    (sc)->sc_rxtcur * (hz / PR_SLOWHZ), syn_cache_timer, (sc));	\
3631 } while (/*CONSTCOND*/0)
3632 
3633 #define	SYN_CACHE_TIMESTAMP(sc)	(tcp_now - (sc)->sc_timebase)
3634 
3635 static inline void
3636 syn_cache_rm(struct syn_cache *sc)
3637 {
3638 	TAILQ_REMOVE(&tcp_syn_cache[sc->sc_bucketidx].sch_bucket,
3639 	    sc, sc_bucketq);
3640 	sc->sc_tp = NULL;
3641 	LIST_REMOVE(sc, sc_tpq);
3642 	tcp_syn_cache[sc->sc_bucketidx].sch_length--;
3643 	callout_stop(&sc->sc_timer);
3644 	syn_cache_count--;
3645 }
3646 
3647 static inline void
3648 syn_cache_put(struct syn_cache *sc)
3649 {
3650 	if (sc->sc_ipopts)
3651 		(void) m_free(sc->sc_ipopts);
3652 	rtcache_free(&sc->sc_route);
3653 	sc->sc_flags |= SCF_DEAD;
3654 	if (!callout_invoking(&sc->sc_timer))
3655 		callout_schedule(&(sc)->sc_timer, 1);
3656 }
3657 
3658 void
3659 syn_cache_init(void)
3660 {
3661 	int i;
3662 
3663 	pool_init(&syn_cache_pool, sizeof(struct syn_cache), 0, 0, 0,
3664 	    "synpl", NULL, IPL_SOFTNET);
3665 
3666 	/* Initialize the hash buckets. */
3667 	for (i = 0; i < tcp_syn_cache_size; i++)
3668 		TAILQ_INIT(&tcp_syn_cache[i].sch_bucket);
3669 }
3670 
3671 void
3672 syn_cache_insert(struct syn_cache *sc, struct tcpcb *tp)
3673 {
3674 	struct syn_cache_head *scp;
3675 	struct syn_cache *sc2;
3676 	int s;
3677 
3678 	/*
3679 	 * If there are no entries in the hash table, reinitialize
3680 	 * the hash secrets.
3681 	 */
3682 	if (syn_cache_count == 0) {
3683 		syn_hash1 = cprng_fast32();
3684 		syn_hash2 = cprng_fast32();
3685 	}
3686 
3687 	SYN_HASHALL(sc->sc_hash, &sc->sc_src.sa, &sc->sc_dst.sa);
3688 	sc->sc_bucketidx = sc->sc_hash % tcp_syn_cache_size;
3689 	scp = &tcp_syn_cache[sc->sc_bucketidx];
3690 
3691 	/*
3692 	 * Make sure that we don't overflow the per-bucket
3693 	 * limit or the total cache size limit.
3694 	 */
3695 	s = splsoftnet();
3696 	if (scp->sch_length >= tcp_syn_bucket_limit) {
3697 		TCP_STATINC(TCP_STAT_SC_BUCKETOVERFLOW);
3698 		/*
3699 		 * The bucket is full.  Toss the oldest element in the
3700 		 * bucket.  This will be the first entry in the bucket.
3701 		 */
3702 		sc2 = TAILQ_FIRST(&scp->sch_bucket);
3703 #ifdef DIAGNOSTIC
3704 		/*
3705 		 * This should never happen; we should always find an
3706 		 * entry in our bucket.
3707 		 */
3708 		if (sc2 == NULL)
3709 			panic("syn_cache_insert: bucketoverflow: impossible");
3710 #endif
3711 		syn_cache_rm(sc2);
3712 		syn_cache_put(sc2);	/* calls pool_put but see spl above */
3713 	} else if (syn_cache_count >= tcp_syn_cache_limit) {
3714 		struct syn_cache_head *scp2, *sce;
3715 
3716 		TCP_STATINC(TCP_STAT_SC_OVERFLOWED);
3717 		/*
3718 		 * The cache is full.  Toss the oldest entry in the
3719 		 * first non-empty bucket we can find.
3720 		 *
3721 		 * XXX We would really like to toss the oldest
3722 		 * entry in the cache, but we hope that this
3723 		 * condition doesn't happen very often.
3724 		 */
3725 		scp2 = scp;
3726 		if (TAILQ_EMPTY(&scp2->sch_bucket)) {
3727 			sce = &tcp_syn_cache[tcp_syn_cache_size];
3728 			for (++scp2; scp2 != scp; scp2++) {
3729 				if (scp2 >= sce)
3730 					scp2 = &tcp_syn_cache[0];
3731 				if (! TAILQ_EMPTY(&scp2->sch_bucket))
3732 					break;
3733 			}
3734 #ifdef DIAGNOSTIC
3735 			/*
3736 			 * This should never happen; we should always find a
3737 			 * non-empty bucket.
3738 			 */
3739 			if (scp2 == scp)
3740 				panic("syn_cache_insert: cacheoverflow: "
3741 				    "impossible");
3742 #endif
3743 		}
3744 		sc2 = TAILQ_FIRST(&scp2->sch_bucket);
3745 		syn_cache_rm(sc2);
3746 		syn_cache_put(sc2);	/* calls pool_put but see spl above */
3747 	}
3748 
3749 	/*
3750 	 * Initialize the entry's timer.
3751 	 */
3752 	sc->sc_rxttot = 0;
3753 	sc->sc_rxtshift = 0;
3754 	SYN_CACHE_TIMER_ARM(sc);
3755 
3756 	/* Link it from tcpcb entry */
3757 	LIST_INSERT_HEAD(&tp->t_sc, sc, sc_tpq);
3758 
3759 	/* Put it into the bucket. */
3760 	TAILQ_INSERT_TAIL(&scp->sch_bucket, sc, sc_bucketq);
3761 	scp->sch_length++;
3762 	syn_cache_count++;
3763 
3764 	TCP_STATINC(TCP_STAT_SC_ADDED);
3765 	splx(s);
3766 }
3767 
3768 /*
3769  * Walk the timer queues, looking for SYN,ACKs that need to be retransmitted.
3770  * If we have retransmitted an entry the maximum number of times, expire
3771  * that entry.
3772  */
3773 void
3774 syn_cache_timer(void *arg)
3775 {
3776 	struct syn_cache *sc = arg;
3777 
3778 	mutex_enter(softnet_lock);
3779 	KERNEL_LOCK(1, NULL);
3780 	callout_ack(&sc->sc_timer);
3781 
3782 	if (__predict_false(sc->sc_flags & SCF_DEAD)) {
3783 		TCP_STATINC(TCP_STAT_SC_DELAYED_FREE);
3784 		callout_destroy(&sc->sc_timer);
3785 		pool_put(&syn_cache_pool, sc);
3786 		KERNEL_UNLOCK_ONE(NULL);
3787 		mutex_exit(softnet_lock);
3788 		return;
3789 	}
3790 
3791 	if (__predict_false(sc->sc_rxtshift == TCP_MAXRXTSHIFT)) {
3792 		/* Drop it -- too many retransmissions. */
3793 		goto dropit;
3794 	}
3795 
3796 	/*
3797 	 * Compute the total amount of time this entry has
3798 	 * been on a queue.  If this entry has been on longer
3799 	 * than the keep alive timer would allow, expire it.
3800 	 */
3801 	sc->sc_rxttot += sc->sc_rxtcur;
3802 	if (sc->sc_rxttot >= tcp_keepinit)
3803 		goto dropit;
3804 
3805 	TCP_STATINC(TCP_STAT_SC_RETRANSMITTED);
3806 	(void) syn_cache_respond(sc, NULL);
3807 
3808 	/* Advance the timer back-off. */
3809 	sc->sc_rxtshift++;
3810 	SYN_CACHE_TIMER_ARM(sc);
3811 
3812 	KERNEL_UNLOCK_ONE(NULL);
3813 	mutex_exit(softnet_lock);
3814 	return;
3815 
3816  dropit:
3817 	TCP_STATINC(TCP_STAT_SC_TIMED_OUT);
3818 	syn_cache_rm(sc);
3819 	if (sc->sc_ipopts)
3820 		(void) m_free(sc->sc_ipopts);
3821 	rtcache_free(&sc->sc_route);
3822 	callout_destroy(&sc->sc_timer);
3823 	pool_put(&syn_cache_pool, sc);
3824 	KERNEL_UNLOCK_ONE(NULL);
3825 	mutex_exit(softnet_lock);
3826 }
3827 
3828 /*
3829  * Remove syn cache created by the specified tcb entry,
3830  * because this does not make sense to keep them
3831  * (if there's no tcb entry, syn cache entry will never be used)
3832  */
3833 void
3834 syn_cache_cleanup(struct tcpcb *tp)
3835 {
3836 	struct syn_cache *sc, *nsc;
3837 	int s;
3838 
3839 	s = splsoftnet();
3840 
3841 	for (sc = LIST_FIRST(&tp->t_sc); sc != NULL; sc = nsc) {
3842 		nsc = LIST_NEXT(sc, sc_tpq);
3843 
3844 #ifdef DIAGNOSTIC
3845 		if (sc->sc_tp != tp)
3846 			panic("invalid sc_tp in syn_cache_cleanup");
3847 #endif
3848 		syn_cache_rm(sc);
3849 		syn_cache_put(sc);	/* calls pool_put but see spl above */
3850 	}
3851 	/* just for safety */
3852 	LIST_INIT(&tp->t_sc);
3853 
3854 	splx(s);
3855 }
3856 
3857 /*
3858  * Find an entry in the syn cache.
3859  */
3860 struct syn_cache *
3861 syn_cache_lookup(const struct sockaddr *src, const struct sockaddr *dst,
3862     struct syn_cache_head **headp)
3863 {
3864 	struct syn_cache *sc;
3865 	struct syn_cache_head *scp;
3866 	u_int32_t hash;
3867 	int s;
3868 
3869 	SYN_HASHALL(hash, src, dst);
3870 
3871 	scp = &tcp_syn_cache[hash % tcp_syn_cache_size];
3872 	*headp = scp;
3873 	s = splsoftnet();
3874 	for (sc = TAILQ_FIRST(&scp->sch_bucket); sc != NULL;
3875 	     sc = TAILQ_NEXT(sc, sc_bucketq)) {
3876 		if (sc->sc_hash != hash)
3877 			continue;
3878 		if (!memcmp(&sc->sc_src, src, src->sa_len) &&
3879 		    !memcmp(&sc->sc_dst, dst, dst->sa_len)) {
3880 			splx(s);
3881 			return (sc);
3882 		}
3883 	}
3884 	splx(s);
3885 	return (NULL);
3886 }
3887 
3888 /*
3889  * This function gets called when we receive an ACK for a
3890  * socket in the LISTEN state.  We look up the connection
3891  * in the syn cache, and if its there, we pull it out of
3892  * the cache and turn it into a full-blown connection in
3893  * the SYN-RECEIVED state.
3894  *
3895  * The return values may not be immediately obvious, and their effects
3896  * can be subtle, so here they are:
3897  *
3898  *	NULL	SYN was not found in cache; caller should drop the
3899  *		packet and send an RST.
3900  *
3901  *	-1	We were unable to create the new connection, and are
3902  *		aborting it.  An ACK,RST is being sent to the peer
3903  *		(unless we got screwey sequence numbners; see below),
3904  *		because the 3-way handshake has been completed.  Caller
3905  *		should not free the mbuf, since we may be using it.  If
3906  *		we are not, we will free it.
3907  *
3908  *	Otherwise, the return value is a pointer to the new socket
3909  *	associated with the connection.
3910  */
3911 struct socket *
3912 syn_cache_get(struct sockaddr *src, struct sockaddr *dst,
3913     struct tcphdr *th, unsigned int hlen, unsigned int tlen,
3914     struct socket *so, struct mbuf *m)
3915 {
3916 	struct syn_cache *sc;
3917 	struct syn_cache_head *scp;
3918 	struct inpcb *inp = NULL;
3919 #ifdef INET6
3920 	struct in6pcb *in6p = NULL;
3921 #endif
3922 	struct tcpcb *tp = 0;
3923 	struct mbuf *am;
3924 	int s;
3925 	struct socket *oso;
3926 
3927 	s = splsoftnet();
3928 	if ((sc = syn_cache_lookup(src, dst, &scp)) == NULL) {
3929 		splx(s);
3930 		return (NULL);
3931 	}
3932 
3933 	/*
3934 	 * Verify the sequence and ack numbers.  Try getting the correct
3935 	 * response again.
3936 	 */
3937 	if ((th->th_ack != sc->sc_iss + 1) ||
3938 	    SEQ_LEQ(th->th_seq, sc->sc_irs) ||
3939 	    SEQ_GT(th->th_seq, sc->sc_irs + 1 + sc->sc_win)) {
3940 		(void) syn_cache_respond(sc, m);
3941 		splx(s);
3942 		return ((struct socket *)(-1));
3943 	}
3944 
3945 	/* Remove this cache entry */
3946 	syn_cache_rm(sc);
3947 	splx(s);
3948 
3949 	/*
3950 	 * Ok, create the full blown connection, and set things up
3951 	 * as they would have been set up if we had created the
3952 	 * connection when the SYN arrived.  If we can't create
3953 	 * the connection, abort it.
3954 	 */
3955 	/*
3956 	 * inp still has the OLD in_pcb stuff, set the
3957 	 * v6-related flags on the new guy, too.   This is
3958 	 * done particularly for the case where an AF_INET6
3959 	 * socket is bound only to a port, and a v4 connection
3960 	 * comes in on that port.
3961 	 * we also copy the flowinfo from the original pcb
3962 	 * to the new one.
3963 	 */
3964 	oso = so;
3965 	so = sonewconn(so, true);
3966 	if (so == NULL)
3967 		goto resetandabort;
3968 
3969 	switch (so->so_proto->pr_domain->dom_family) {
3970 #ifdef INET
3971 	case AF_INET:
3972 		inp = sotoinpcb(so);
3973 		break;
3974 #endif
3975 #ifdef INET6
3976 	case AF_INET6:
3977 		in6p = sotoin6pcb(so);
3978 		break;
3979 #endif
3980 	}
3981 	switch (src->sa_family) {
3982 #ifdef INET
3983 	case AF_INET:
3984 		if (inp) {
3985 			inp->inp_laddr = ((struct sockaddr_in *)dst)->sin_addr;
3986 			inp->inp_lport = ((struct sockaddr_in *)dst)->sin_port;
3987 			inp->inp_options = ip_srcroute();
3988 			in_pcbstate(inp, INP_BOUND);
3989 			if (inp->inp_options == NULL) {
3990 				inp->inp_options = sc->sc_ipopts;
3991 				sc->sc_ipopts = NULL;
3992 			}
3993 		}
3994 #ifdef INET6
3995 		else if (in6p) {
3996 			/* IPv4 packet to AF_INET6 socket */
3997 			memset(&in6p->in6p_laddr, 0, sizeof(in6p->in6p_laddr));
3998 			in6p->in6p_laddr.s6_addr16[5] = htons(0xffff);
3999 			bcopy(&((struct sockaddr_in *)dst)->sin_addr,
4000 				&in6p->in6p_laddr.s6_addr32[3],
4001 				sizeof(((struct sockaddr_in *)dst)->sin_addr));
4002 			in6p->in6p_lport = ((struct sockaddr_in *)dst)->sin_port;
4003 			in6totcpcb(in6p)->t_family = AF_INET;
4004 			if (sotoin6pcb(oso)->in6p_flags & IN6P_IPV6_V6ONLY)
4005 				in6p->in6p_flags |= IN6P_IPV6_V6ONLY;
4006 			else
4007 				in6p->in6p_flags &= ~IN6P_IPV6_V6ONLY;
4008 			in6_pcbstate(in6p, IN6P_BOUND);
4009 		}
4010 #endif
4011 		break;
4012 #endif
4013 #ifdef INET6
4014 	case AF_INET6:
4015 		if (in6p) {
4016 			in6p->in6p_laddr = ((struct sockaddr_in6 *)dst)->sin6_addr;
4017 			in6p->in6p_lport = ((struct sockaddr_in6 *)dst)->sin6_port;
4018 			in6_pcbstate(in6p, IN6P_BOUND);
4019 		}
4020 		break;
4021 #endif
4022 	}
4023 #ifdef INET6
4024 	if (in6p && in6totcpcb(in6p)->t_family == AF_INET6 && sotoinpcb(oso)) {
4025 		struct in6pcb *oin6p = sotoin6pcb(oso);
4026 		/* inherit socket options from the listening socket */
4027 		in6p->in6p_flags |= (oin6p->in6p_flags & IN6P_CONTROLOPTS);
4028 		if (in6p->in6p_flags & IN6P_CONTROLOPTS) {
4029 			m_freem(in6p->in6p_options);
4030 			in6p->in6p_options = 0;
4031 		}
4032 		ip6_savecontrol(in6p, &in6p->in6p_options,
4033 			mtod(m, struct ip6_hdr *), m);
4034 	}
4035 #endif
4036 
4037 #if defined(IPSEC)
4038 	if (ipsec_used) {
4039 		/*
4040 		 * we make a copy of policy, instead of sharing the policy, for
4041 		 * better behavior in terms of SA lookup and dead SA removal.
4042 		 */
4043 		if (inp) {
4044 			/* copy old policy into new socket's */
4045 			if (ipsec_copy_pcbpolicy(sotoinpcb(oso)->inp_sp,
4046 			    inp->inp_sp))
4047 				printf("tcp_input: could not copy policy\n");
4048 		}
4049 #ifdef INET6
4050 		else if (in6p) {
4051 			/* copy old policy into new socket's */
4052 			if (ipsec_copy_pcbpolicy(sotoin6pcb(oso)->in6p_sp,
4053 			    in6p->in6p_sp))
4054 				printf("tcp_input: could not copy policy\n");
4055 		}
4056 #endif
4057 	}
4058 #endif
4059 
4060 	/*
4061 	 * Give the new socket our cached route reference.
4062 	 */
4063 	if (inp) {
4064 		rtcache_copy(&inp->inp_route, &sc->sc_route);
4065 		rtcache_free(&sc->sc_route);
4066 	}
4067 #ifdef INET6
4068 	else {
4069 		rtcache_copy(&in6p->in6p_route, &sc->sc_route);
4070 		rtcache_free(&sc->sc_route);
4071 	}
4072 #endif
4073 
4074 	am = m_get(M_DONTWAIT, MT_SONAME);	/* XXX */
4075 	if (am == NULL)
4076 		goto resetandabort;
4077 	MCLAIM(am, &tcp_mowner);
4078 	am->m_len = src->sa_len;
4079 	bcopy(src, mtod(am, void *), src->sa_len);
4080 	if (inp) {
4081 		if (in_pcbconnect(inp, am, &lwp0)) {
4082 			(void) m_free(am);
4083 			goto resetandabort;
4084 		}
4085 	}
4086 #ifdef INET6
4087 	else if (in6p) {
4088 		if (src->sa_family == AF_INET) {
4089 			/* IPv4 packet to AF_INET6 socket */
4090 			struct sockaddr_in6 *sin6;
4091 			sin6 = mtod(am, struct sockaddr_in6 *);
4092 			am->m_len = sizeof(*sin6);
4093 			memset(sin6, 0, sizeof(*sin6));
4094 			sin6->sin6_family = AF_INET6;
4095 			sin6->sin6_len = sizeof(*sin6);
4096 			sin6->sin6_port = ((struct sockaddr_in *)src)->sin_port;
4097 			sin6->sin6_addr.s6_addr16[5] = htons(0xffff);
4098 			bcopy(&((struct sockaddr_in *)src)->sin_addr,
4099 				&sin6->sin6_addr.s6_addr32[3],
4100 				sizeof(sin6->sin6_addr.s6_addr32[3]));
4101 		}
4102 		if (in6_pcbconnect(in6p, am, NULL)) {
4103 			(void) m_free(am);
4104 			goto resetandabort;
4105 		}
4106 	}
4107 #endif
4108 	else {
4109 		(void) m_free(am);
4110 		goto resetandabort;
4111 	}
4112 	(void) m_free(am);
4113 
4114 	if (inp)
4115 		tp = intotcpcb(inp);
4116 #ifdef INET6
4117 	else if (in6p)
4118 		tp = in6totcpcb(in6p);
4119 #endif
4120 	else
4121 		tp = NULL;
4122 	tp->t_flags = sototcpcb(oso)->t_flags & TF_NODELAY;
4123 	if (sc->sc_request_r_scale != 15) {
4124 		tp->requested_s_scale = sc->sc_requested_s_scale;
4125 		tp->request_r_scale = sc->sc_request_r_scale;
4126 		tp->snd_scale = sc->sc_requested_s_scale;
4127 		tp->rcv_scale = sc->sc_request_r_scale;
4128 		tp->t_flags |= TF_REQ_SCALE|TF_RCVD_SCALE;
4129 	}
4130 	if (sc->sc_flags & SCF_TIMESTAMP)
4131 		tp->t_flags |= TF_REQ_TSTMP|TF_RCVD_TSTMP;
4132 	tp->ts_timebase = sc->sc_timebase;
4133 
4134 	tp->t_template = tcp_template(tp);
4135 	if (tp->t_template == 0) {
4136 		tp = tcp_drop(tp, ENOBUFS);	/* destroys socket */
4137 		so = NULL;
4138 		m_freem(m);
4139 		goto abort;
4140 	}
4141 
4142 	tp->iss = sc->sc_iss;
4143 	tp->irs = sc->sc_irs;
4144 	tcp_sendseqinit(tp);
4145 	tcp_rcvseqinit(tp);
4146 	tp->t_state = TCPS_SYN_RECEIVED;
4147 	TCP_TIMER_ARM(tp, TCPT_KEEP, tp->t_keepinit);
4148 	TCP_STATINC(TCP_STAT_ACCEPTS);
4149 
4150 	if ((sc->sc_flags & SCF_SACK_PERMIT) && tcp_do_sack)
4151 		tp->t_flags |= TF_WILL_SACK;
4152 
4153 	if ((sc->sc_flags & SCF_ECN_PERMIT) && tcp_do_ecn)
4154 		tp->t_flags |= TF_ECN_PERMIT;
4155 
4156 #ifdef TCP_SIGNATURE
4157 	if (sc->sc_flags & SCF_SIGNATURE)
4158 		tp->t_flags |= TF_SIGNATURE;
4159 #endif
4160 
4161 	/* Initialize tp->t_ourmss before we deal with the peer's! */
4162 	tp->t_ourmss = sc->sc_ourmaxseg;
4163 	tcp_mss_from_peer(tp, sc->sc_peermaxseg);
4164 
4165 	/*
4166 	 * Initialize the initial congestion window.  If we
4167 	 * had to retransmit the SYN,ACK, we must initialize cwnd
4168 	 * to 1 segment (i.e. the Loss Window).
4169 	 */
4170 	if (sc->sc_rxtshift)
4171 		tp->snd_cwnd = tp->t_peermss;
4172 	else {
4173 		int ss = tcp_init_win;
4174 #ifdef INET
4175 		if (inp != NULL && in_localaddr(inp->inp_faddr))
4176 			ss = tcp_init_win_local;
4177 #endif
4178 #ifdef INET6
4179 		if (in6p != NULL && in6_localaddr(&in6p->in6p_faddr))
4180 			ss = tcp_init_win_local;
4181 #endif
4182 		tp->snd_cwnd = TCP_INITIAL_WINDOW(ss, tp->t_peermss);
4183 	}
4184 
4185 	tcp_rmx_rtt(tp);
4186 	tp->snd_wl1 = sc->sc_irs;
4187 	tp->rcv_up = sc->sc_irs + 1;
4188 
4189 	/*
4190 	 * This is what whould have happened in tcp_output() when
4191 	 * the SYN,ACK was sent.
4192 	 */
4193 	tp->snd_up = tp->snd_una;
4194 	tp->snd_max = tp->snd_nxt = tp->iss+1;
4195 	TCP_TIMER_ARM(tp, TCPT_REXMT, tp->t_rxtcur);
4196 	if (sc->sc_win > 0 && SEQ_GT(tp->rcv_nxt + sc->sc_win, tp->rcv_adv))
4197 		tp->rcv_adv = tp->rcv_nxt + sc->sc_win;
4198 	tp->last_ack_sent = tp->rcv_nxt;
4199 	tp->t_partialacks = -1;
4200 	tp->t_dupacks = 0;
4201 
4202 	TCP_STATINC(TCP_STAT_SC_COMPLETED);
4203 	s = splsoftnet();
4204 	syn_cache_put(sc);
4205 	splx(s);
4206 	return (so);
4207 
4208 resetandabort:
4209 	(void)tcp_respond(NULL, m, m, th, (tcp_seq)0, th->th_ack, TH_RST);
4210 abort:
4211 	if (so != NULL) {
4212 		(void) soqremque(so, 1);
4213 		(void) soabort(so);
4214 		mutex_enter(softnet_lock);
4215 	}
4216 	s = splsoftnet();
4217 	syn_cache_put(sc);
4218 	splx(s);
4219 	TCP_STATINC(TCP_STAT_SC_ABORTED);
4220 	return ((struct socket *)(-1));
4221 }
4222 
4223 /*
4224  * This function is called when we get a RST for a
4225  * non-existent connection, so that we can see if the
4226  * connection is in the syn cache.  If it is, zap it.
4227  */
4228 
4229 void
4230 syn_cache_reset(struct sockaddr *src, struct sockaddr *dst, struct tcphdr *th)
4231 {
4232 	struct syn_cache *sc;
4233 	struct syn_cache_head *scp;
4234 	int s = splsoftnet();
4235 
4236 	if ((sc = syn_cache_lookup(src, dst, &scp)) == NULL) {
4237 		splx(s);
4238 		return;
4239 	}
4240 	if (SEQ_LT(th->th_seq, sc->sc_irs) ||
4241 	    SEQ_GT(th->th_seq, sc->sc_irs+1)) {
4242 		splx(s);
4243 		return;
4244 	}
4245 	syn_cache_rm(sc);
4246 	TCP_STATINC(TCP_STAT_SC_RESET);
4247 	syn_cache_put(sc);	/* calls pool_put but see spl above */
4248 	splx(s);
4249 }
4250 
4251 void
4252 syn_cache_unreach(const struct sockaddr *src, const struct sockaddr *dst,
4253     struct tcphdr *th)
4254 {
4255 	struct syn_cache *sc;
4256 	struct syn_cache_head *scp;
4257 	int s;
4258 
4259 	s = splsoftnet();
4260 	if ((sc = syn_cache_lookup(src, dst, &scp)) == NULL) {
4261 		splx(s);
4262 		return;
4263 	}
4264 	/* If the sequence number != sc_iss, then it's a bogus ICMP msg */
4265 	if (ntohl (th->th_seq) != sc->sc_iss) {
4266 		splx(s);
4267 		return;
4268 	}
4269 
4270 	/*
4271 	 * If we've retransmitted 3 times and this is our second error,
4272 	 * we remove the entry.  Otherwise, we allow it to continue on.
4273 	 * This prevents us from incorrectly nuking an entry during a
4274 	 * spurious network outage.
4275 	 *
4276 	 * See tcp_notify().
4277 	 */
4278 	if ((sc->sc_flags & SCF_UNREACH) == 0 || sc->sc_rxtshift < 3) {
4279 		sc->sc_flags |= SCF_UNREACH;
4280 		splx(s);
4281 		return;
4282 	}
4283 
4284 	syn_cache_rm(sc);
4285 	TCP_STATINC(TCP_STAT_SC_UNREACH);
4286 	syn_cache_put(sc);	/* calls pool_put but see spl above */
4287 	splx(s);
4288 }
4289 
4290 /*
4291  * Given a LISTEN socket and an inbound SYN request, add
4292  * this to the syn cache, and send back a segment:
4293  *	<SEQ=ISS><ACK=RCV_NXT><CTL=SYN,ACK>
4294  * to the source.
4295  *
4296  * IMPORTANT NOTE: We do _NOT_ ACK data that might accompany the SYN.
4297  * Doing so would require that we hold onto the data and deliver it
4298  * to the application.  However, if we are the target of a SYN-flood
4299  * DoS attack, an attacker could send data which would eventually
4300  * consume all available buffer space if it were ACKed.  By not ACKing
4301  * the data, we avoid this DoS scenario.
4302  */
4303 
4304 int
4305 syn_cache_add(struct sockaddr *src, struct sockaddr *dst, struct tcphdr *th,
4306     unsigned int hlen, struct socket *so, struct mbuf *m, u_char *optp,
4307     int optlen, struct tcp_opt_info *oi)
4308 {
4309 	struct tcpcb tb, *tp;
4310 	long win;
4311 	struct syn_cache *sc;
4312 	struct syn_cache_head *scp;
4313 	struct mbuf *ipopts;
4314 	struct tcp_opt_info opti;
4315 	int s;
4316 
4317 	tp = sototcpcb(so);
4318 
4319 	memset(&opti, 0, sizeof(opti));
4320 
4321 	/*
4322 	 * RFC1122 4.2.3.10, p. 104: discard bcast/mcast SYN
4323 	 *
4324 	 * Note this check is performed in tcp_input() very early on.
4325 	 */
4326 
4327 	/*
4328 	 * Initialize some local state.
4329 	 */
4330 	win = sbspace(&so->so_rcv);
4331 	if (win > TCP_MAXWIN)
4332 		win = TCP_MAXWIN;
4333 
4334 	switch (src->sa_family) {
4335 #ifdef INET
4336 	case AF_INET:
4337 		/*
4338 		 * Remember the IP options, if any.
4339 		 */
4340 		ipopts = ip_srcroute();
4341 		break;
4342 #endif
4343 	default:
4344 		ipopts = NULL;
4345 	}
4346 
4347 #ifdef TCP_SIGNATURE
4348 	if (optp || (tp->t_flags & TF_SIGNATURE))
4349 #else
4350 	if (optp)
4351 #endif
4352 	{
4353 		tb.t_flags = tcp_do_rfc1323 ? (TF_REQ_SCALE|TF_REQ_TSTMP) : 0;
4354 #ifdef TCP_SIGNATURE
4355 		tb.t_flags |= (tp->t_flags & TF_SIGNATURE);
4356 #endif
4357 		tb.t_state = TCPS_LISTEN;
4358 		if (tcp_dooptions(&tb, optp, optlen, th, m, m->m_pkthdr.len -
4359 		    sizeof(struct tcphdr) - optlen - hlen, oi) < 0)
4360 			return (0);
4361 	} else
4362 		tb.t_flags = 0;
4363 
4364 	/*
4365 	 * See if we already have an entry for this connection.
4366 	 * If we do, resend the SYN,ACK.  We do not count this
4367 	 * as a retransmission (XXX though maybe we should).
4368 	 */
4369 	if ((sc = syn_cache_lookup(src, dst, &scp)) != NULL) {
4370 		TCP_STATINC(TCP_STAT_SC_DUPESYN);
4371 		if (ipopts) {
4372 			/*
4373 			 * If we were remembering a previous source route,
4374 			 * forget it and use the new one we've been given.
4375 			 */
4376 			if (sc->sc_ipopts)
4377 				(void) m_free(sc->sc_ipopts);
4378 			sc->sc_ipopts = ipopts;
4379 		}
4380 		sc->sc_timestamp = tb.ts_recent;
4381 		if (syn_cache_respond(sc, m) == 0) {
4382 			uint64_t *tcps = TCP_STAT_GETREF();
4383 			tcps[TCP_STAT_SNDACKS]++;
4384 			tcps[TCP_STAT_SNDTOTAL]++;
4385 			TCP_STAT_PUTREF();
4386 		}
4387 		return (1);
4388 	}
4389 
4390 	s = splsoftnet();
4391 	sc = pool_get(&syn_cache_pool, PR_NOWAIT);
4392 	splx(s);
4393 	if (sc == NULL) {
4394 		if (ipopts)
4395 			(void) m_free(ipopts);
4396 		return (0);
4397 	}
4398 
4399 	/*
4400 	 * Fill in the cache, and put the necessary IP and TCP
4401 	 * options into the reply.
4402 	 */
4403 	memset(sc, 0, sizeof(struct syn_cache));
4404 	callout_init(&sc->sc_timer, CALLOUT_MPSAFE);
4405 	bcopy(src, &sc->sc_src, src->sa_len);
4406 	bcopy(dst, &sc->sc_dst, dst->sa_len);
4407 	sc->sc_flags = 0;
4408 	sc->sc_ipopts = ipopts;
4409 	sc->sc_irs = th->th_seq;
4410 	switch (src->sa_family) {
4411 #ifdef INET
4412 	case AF_INET:
4413 	    {
4414 		struct sockaddr_in *srcin = (void *) src;
4415 		struct sockaddr_in *dstin = (void *) dst;
4416 
4417 		sc->sc_iss = tcp_new_iss1(&dstin->sin_addr,
4418 		    &srcin->sin_addr, dstin->sin_port,
4419 		    srcin->sin_port, sizeof(dstin->sin_addr), 0);
4420 		break;
4421 	    }
4422 #endif /* INET */
4423 #ifdef INET6
4424 	case AF_INET6:
4425 	    {
4426 		struct sockaddr_in6 *srcin6 = (void *) src;
4427 		struct sockaddr_in6 *dstin6 = (void *) dst;
4428 
4429 		sc->sc_iss = tcp_new_iss1(&dstin6->sin6_addr,
4430 		    &srcin6->sin6_addr, dstin6->sin6_port,
4431 		    srcin6->sin6_port, sizeof(dstin6->sin6_addr), 0);
4432 		break;
4433 	    }
4434 #endif /* INET6 */
4435 	}
4436 	sc->sc_peermaxseg = oi->maxseg;
4437 	sc->sc_ourmaxseg = tcp_mss_to_advertise(m->m_flags & M_PKTHDR ?
4438 						m->m_pkthdr.rcvif : NULL,
4439 						sc->sc_src.sa.sa_family);
4440 	sc->sc_win = win;
4441 	sc->sc_timebase = tcp_now - 1;	/* see tcp_newtcpcb() */
4442 	sc->sc_timestamp = tb.ts_recent;
4443 	if ((tb.t_flags & (TF_REQ_TSTMP|TF_RCVD_TSTMP)) ==
4444 	    (TF_REQ_TSTMP|TF_RCVD_TSTMP))
4445 		sc->sc_flags |= SCF_TIMESTAMP;
4446 	if ((tb.t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) ==
4447 	    (TF_RCVD_SCALE|TF_REQ_SCALE)) {
4448 		sc->sc_requested_s_scale = tb.requested_s_scale;
4449 		sc->sc_request_r_scale = 0;
4450 		/*
4451 		 * Pick the smallest possible scaling factor that
4452 		 * will still allow us to scale up to sb_max.
4453 		 *
4454 		 * We do this because there are broken firewalls that
4455 		 * will corrupt the window scale option, leading to
4456 		 * the other endpoint believing that our advertised
4457 		 * window is unscaled.  At scale factors larger than
4458 		 * 5 the unscaled window will drop below 1500 bytes,
4459 		 * leading to serious problems when traversing these
4460 		 * broken firewalls.
4461 		 *
4462 		 * With the default sbmax of 256K, a scale factor
4463 		 * of 3 will be chosen by this algorithm.  Those who
4464 		 * choose a larger sbmax should watch out
4465 		 * for the compatiblity problems mentioned above.
4466 		 *
4467 		 * RFC1323: The Window field in a SYN (i.e., a <SYN>
4468 		 * or <SYN,ACK>) segment itself is never scaled.
4469 		 */
4470 		while (sc->sc_request_r_scale < TCP_MAX_WINSHIFT &&
4471 		    (TCP_MAXWIN << sc->sc_request_r_scale) < sb_max)
4472 			sc->sc_request_r_scale++;
4473 	} else {
4474 		sc->sc_requested_s_scale = 15;
4475 		sc->sc_request_r_scale = 15;
4476 	}
4477 	if ((tb.t_flags & TF_SACK_PERMIT) && tcp_do_sack)
4478 		sc->sc_flags |= SCF_SACK_PERMIT;
4479 
4480 	/*
4481 	 * ECN setup packet recieved.
4482 	 */
4483 	if ((th->th_flags & (TH_ECE|TH_CWR)) && tcp_do_ecn)
4484 		sc->sc_flags |= SCF_ECN_PERMIT;
4485 
4486 #ifdef TCP_SIGNATURE
4487 	if (tb.t_flags & TF_SIGNATURE)
4488 		sc->sc_flags |= SCF_SIGNATURE;
4489 #endif
4490 	sc->sc_tp = tp;
4491 	if (syn_cache_respond(sc, m) == 0) {
4492 		uint64_t *tcps = TCP_STAT_GETREF();
4493 		tcps[TCP_STAT_SNDACKS]++;
4494 		tcps[TCP_STAT_SNDTOTAL]++;
4495 		TCP_STAT_PUTREF();
4496 		syn_cache_insert(sc, tp);
4497 	} else {
4498 		s = splsoftnet();
4499 		/*
4500 		 * syn_cache_put() will try to schedule the timer, so
4501 		 * we need to initialize it
4502 		 */
4503 		SYN_CACHE_TIMER_ARM(sc);
4504 		syn_cache_put(sc);
4505 		splx(s);
4506 		TCP_STATINC(TCP_STAT_SC_DROPPED);
4507 	}
4508 	return (1);
4509 }
4510 
4511 /*
4512  * syn_cache_respond: (re)send SYN+ACK.
4513  *
4514  * returns 0 on success.  otherwise returns an errno, typically ENOBUFS.
4515  */
4516 
4517 int
4518 syn_cache_respond(struct syn_cache *sc, struct mbuf *m)
4519 {
4520 #ifdef INET6
4521 	struct rtentry *rt;
4522 #endif
4523 	struct route *ro;
4524 	u_int8_t *optp;
4525 	int optlen, error;
4526 	u_int16_t tlen;
4527 	struct ip *ip = NULL;
4528 #ifdef INET6
4529 	struct ip6_hdr *ip6 = NULL;
4530 #endif
4531 	struct tcpcb *tp = NULL;
4532 	struct tcphdr *th;
4533 	u_int hlen;
4534 	struct socket *so;
4535 
4536 	ro = &sc->sc_route;
4537 	switch (sc->sc_src.sa.sa_family) {
4538 	case AF_INET:
4539 		hlen = sizeof(struct ip);
4540 		break;
4541 #ifdef INET6
4542 	case AF_INET6:
4543 		hlen = sizeof(struct ip6_hdr);
4544 		break;
4545 #endif
4546 	default:
4547 		if (m)
4548 			m_freem(m);
4549 		return (EAFNOSUPPORT);
4550 	}
4551 
4552 	/* Compute the size of the TCP options. */
4553 	optlen = 4 + (sc->sc_request_r_scale != 15 ? 4 : 0) +
4554 	    ((sc->sc_flags & SCF_SACK_PERMIT) ? (TCPOLEN_SACK_PERMITTED + 2) : 0) +
4555 #ifdef TCP_SIGNATURE
4556 	    ((sc->sc_flags & SCF_SIGNATURE) ? (TCPOLEN_SIGNATURE + 2) : 0) +
4557 #endif
4558 	    ((sc->sc_flags & SCF_TIMESTAMP) ? TCPOLEN_TSTAMP_APPA : 0);
4559 
4560 	tlen = hlen + sizeof(struct tcphdr) + optlen;
4561 
4562 	/*
4563 	 * Create the IP+TCP header from scratch.
4564 	 */
4565 	if (m)
4566 		m_freem(m);
4567 #ifdef DIAGNOSTIC
4568 	if (max_linkhdr + tlen > MCLBYTES)
4569 		return (ENOBUFS);
4570 #endif
4571 	MGETHDR(m, M_DONTWAIT, MT_DATA);
4572 	if (m && (max_linkhdr + tlen) > MHLEN) {
4573 		MCLGET(m, M_DONTWAIT);
4574 		if ((m->m_flags & M_EXT) == 0) {
4575 			m_freem(m);
4576 			m = NULL;
4577 		}
4578 	}
4579 	if (m == NULL)
4580 		return (ENOBUFS);
4581 	MCLAIM(m, &tcp_tx_mowner);
4582 
4583 	/* Fixup the mbuf. */
4584 	m->m_data += max_linkhdr;
4585 	m->m_len = m->m_pkthdr.len = tlen;
4586 	if (sc->sc_tp) {
4587 		tp = sc->sc_tp;
4588 		if (tp->t_inpcb)
4589 			so = tp->t_inpcb->inp_socket;
4590 #ifdef INET6
4591 		else if (tp->t_in6pcb)
4592 			so = tp->t_in6pcb->in6p_socket;
4593 #endif
4594 		else
4595 			so = NULL;
4596 	} else
4597 		so = NULL;
4598 	m->m_pkthdr.rcvif = NULL;
4599 	memset(mtod(m, u_char *), 0, tlen);
4600 
4601 	switch (sc->sc_src.sa.sa_family) {
4602 	case AF_INET:
4603 		ip = mtod(m, struct ip *);
4604 		ip->ip_v = 4;
4605 		ip->ip_dst = sc->sc_src.sin.sin_addr;
4606 		ip->ip_src = sc->sc_dst.sin.sin_addr;
4607 		ip->ip_p = IPPROTO_TCP;
4608 		th = (struct tcphdr *)(ip + 1);
4609 		th->th_dport = sc->sc_src.sin.sin_port;
4610 		th->th_sport = sc->sc_dst.sin.sin_port;
4611 		break;
4612 #ifdef INET6
4613 	case AF_INET6:
4614 		ip6 = mtod(m, struct ip6_hdr *);
4615 		ip6->ip6_vfc = IPV6_VERSION;
4616 		ip6->ip6_dst = sc->sc_src.sin6.sin6_addr;
4617 		ip6->ip6_src = sc->sc_dst.sin6.sin6_addr;
4618 		ip6->ip6_nxt = IPPROTO_TCP;
4619 		/* ip6_plen will be updated in ip6_output() */
4620 		th = (struct tcphdr *)(ip6 + 1);
4621 		th->th_dport = sc->sc_src.sin6.sin6_port;
4622 		th->th_sport = sc->sc_dst.sin6.sin6_port;
4623 		break;
4624 #endif
4625 	default:
4626 		th = NULL;
4627 	}
4628 
4629 	th->th_seq = htonl(sc->sc_iss);
4630 	th->th_ack = htonl(sc->sc_irs + 1);
4631 	th->th_off = (sizeof(struct tcphdr) + optlen) >> 2;
4632 	th->th_flags = TH_SYN|TH_ACK;
4633 	th->th_win = htons(sc->sc_win);
4634 	/* th_sum already 0 */
4635 	/* th_urp already 0 */
4636 
4637 	/* Tack on the TCP options. */
4638 	optp = (u_int8_t *)(th + 1);
4639 	*optp++ = TCPOPT_MAXSEG;
4640 	*optp++ = 4;
4641 	*optp++ = (sc->sc_ourmaxseg >> 8) & 0xff;
4642 	*optp++ = sc->sc_ourmaxseg & 0xff;
4643 
4644 	if (sc->sc_request_r_scale != 15) {
4645 		*((u_int32_t *)optp) = htonl(TCPOPT_NOP << 24 |
4646 		    TCPOPT_WINDOW << 16 | TCPOLEN_WINDOW << 8 |
4647 		    sc->sc_request_r_scale);
4648 		optp += 4;
4649 	}
4650 
4651 	if (sc->sc_flags & SCF_TIMESTAMP) {
4652 		u_int32_t *lp = (u_int32_t *)(optp);
4653 		/* Form timestamp option as shown in appendix A of RFC 1323. */
4654 		*lp++ = htonl(TCPOPT_TSTAMP_HDR);
4655 		*lp++ = htonl(SYN_CACHE_TIMESTAMP(sc));
4656 		*lp   = htonl(sc->sc_timestamp);
4657 		optp += TCPOLEN_TSTAMP_APPA;
4658 	}
4659 
4660 	if (sc->sc_flags & SCF_SACK_PERMIT) {
4661 		u_int8_t *p = optp;
4662 
4663 		/* Let the peer know that we will SACK. */
4664 		p[0] = TCPOPT_SACK_PERMITTED;
4665 		p[1] = 2;
4666 		p[2] = TCPOPT_NOP;
4667 		p[3] = TCPOPT_NOP;
4668 		optp += 4;
4669 	}
4670 
4671 	/*
4672 	 * Send ECN SYN-ACK setup packet.
4673 	 * Routes can be asymetric, so, even if we receive a packet
4674 	 * with ECE and CWR set, we must not assume no one will block
4675 	 * the ECE packet we are about to send.
4676 	 */
4677 	if ((sc->sc_flags & SCF_ECN_PERMIT) && tp &&
4678 	    SEQ_GEQ(tp->snd_nxt, tp->snd_max)) {
4679 		th->th_flags |= TH_ECE;
4680 		TCP_STATINC(TCP_STAT_ECN_SHS);
4681 
4682 		/*
4683 		 * draft-ietf-tcpm-ecnsyn-00.txt
4684 		 *
4685 		 * "[...] a TCP node MAY respond to an ECN-setup
4686 		 * SYN packet by setting ECT in the responding
4687 		 * ECN-setup SYN/ACK packet, indicating to routers
4688 		 * that the SYN/ACK packet is ECN-Capable.
4689 		 * This allows a congested router along the path
4690 		 * to mark the packet instead of dropping the
4691 		 * packet as an indication of congestion."
4692 		 *
4693 		 * "[...] There can be a great benefit in setting
4694 		 * an ECN-capable codepoint in SYN/ACK packets [...]
4695 		 * Congestion is  most likely to occur in
4696 		 * the server-to-client direction.  As a result,
4697 		 * setting an ECN-capable codepoint in SYN/ACK
4698 		 * packets can reduce the occurence of three-second
4699 		 * retransmit timeouts resulting from the drop
4700 		 * of SYN/ACK packets."
4701 		 *
4702 		 * Page 4 and 6, January 2006.
4703 		 */
4704 
4705 		switch (sc->sc_src.sa.sa_family) {
4706 #ifdef INET
4707 		case AF_INET:
4708 			ip->ip_tos |= IPTOS_ECN_ECT0;
4709 			break;
4710 #endif
4711 #ifdef INET6
4712 		case AF_INET6:
4713 			ip6->ip6_flow |= htonl(IPTOS_ECN_ECT0 << 20);
4714 			break;
4715 #endif
4716 		}
4717 		TCP_STATINC(TCP_STAT_ECN_ECT);
4718 	}
4719 
4720 #ifdef TCP_SIGNATURE
4721 	if (sc->sc_flags & SCF_SIGNATURE) {
4722 		struct secasvar *sav;
4723 		u_int8_t *sigp;
4724 
4725 		sav = tcp_signature_getsav(m, th);
4726 
4727 		if (sav == NULL) {
4728 			if (m)
4729 				m_freem(m);
4730 			return (EPERM);
4731 		}
4732 
4733 		*optp++ = TCPOPT_SIGNATURE;
4734 		*optp++ = TCPOLEN_SIGNATURE;
4735 		sigp = optp;
4736 		memset(optp, 0, TCP_SIGLEN);
4737 		optp += TCP_SIGLEN;
4738 		*optp++ = TCPOPT_NOP;
4739 		*optp++ = TCPOPT_EOL;
4740 
4741 		(void)tcp_signature(m, th, hlen, sav, sigp);
4742 
4743 		key_sa_recordxfer(sav, m);
4744 		KEY_FREESAV(&sav);
4745 	}
4746 #endif
4747 
4748 	/* Compute the packet's checksum. */
4749 	switch (sc->sc_src.sa.sa_family) {
4750 	case AF_INET:
4751 		ip->ip_len = htons(tlen - hlen);
4752 		th->th_sum = 0;
4753 		th->th_sum = in4_cksum(m, IPPROTO_TCP, hlen, tlen - hlen);
4754 		break;
4755 #ifdef INET6
4756 	case AF_INET6:
4757 		ip6->ip6_plen = htons(tlen - hlen);
4758 		th->th_sum = 0;
4759 		th->th_sum = in6_cksum(m, IPPROTO_TCP, hlen, tlen - hlen);
4760 		break;
4761 #endif
4762 	}
4763 
4764 	/*
4765 	 * Fill in some straggling IP bits.  Note the stack expects
4766 	 * ip_len to be in host order, for convenience.
4767 	 */
4768 	switch (sc->sc_src.sa.sa_family) {
4769 #ifdef INET
4770 	case AF_INET:
4771 		ip->ip_len = htons(tlen);
4772 		ip->ip_ttl = ip_defttl;
4773 		/* XXX tos? */
4774 		break;
4775 #endif
4776 #ifdef INET6
4777 	case AF_INET6:
4778 		ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
4779 		ip6->ip6_vfc |= IPV6_VERSION;
4780 		ip6->ip6_plen = htons(tlen - hlen);
4781 		/* ip6_hlim will be initialized afterwards */
4782 		/* XXX flowlabel? */
4783 		break;
4784 #endif
4785 	}
4786 
4787 	/* XXX use IPsec policy on listening socket, on SYN ACK */
4788 	tp = sc->sc_tp;
4789 
4790 	switch (sc->sc_src.sa.sa_family) {
4791 #ifdef INET
4792 	case AF_INET:
4793 		error = ip_output(m, sc->sc_ipopts, ro,
4794 		    (ip_mtudisc ? IP_MTUDISC : 0),
4795 		    NULL, so);
4796 		break;
4797 #endif
4798 #ifdef INET6
4799 	case AF_INET6:
4800 		ip6->ip6_hlim = in6_selecthlim(NULL,
4801 				(rt = rtcache_validate(ro)) != NULL ? rt->rt_ifp
4802 				                                    : NULL);
4803 
4804 		error = ip6_output(m, NULL /*XXX*/, ro, 0, NULL, so, NULL);
4805 		break;
4806 #endif
4807 	default:
4808 		error = EAFNOSUPPORT;
4809 		break;
4810 	}
4811 	return (error);
4812 }
4813