xref: /netbsd-src/sys/netinet6/in6_pcb.c (revision f21b7d7f2cbdd5c14b3882c4e8a3d43580d460a6)
1 /*	$NetBSD: in6_pcb.c,v 1.150 2016/09/29 12:19:47 roy Exp $	*/
2 /*	$KAME: in6_pcb.c,v 1.84 2001/02/08 18:02:08 itojun Exp $	*/
3 
4 /*
5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 /*
34  * Copyright (c) 1982, 1986, 1991, 1993
35  *	The Regents of the University of California.  All rights reserved.
36  *
37  * Redistribution and use in source and binary forms, with or without
38  * modification, are permitted provided that the following conditions
39  * are met:
40  * 1. Redistributions of source code must retain the above copyright
41  *    notice, this list of conditions and the following disclaimer.
42  * 2. Redistributions in binary form must reproduce the above copyright
43  *    notice, this list of conditions and the following disclaimer in the
44  *    documentation and/or other materials provided with the distribution.
45  * 3. Neither the name of the University nor the names of its contributors
46  *    may be used to endorse or promote products derived from this software
47  *    without specific prior written permission.
48  *
49  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
50  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
51  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
52  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
53  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
54  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
55  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
56  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
57  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
58  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
59  * SUCH DAMAGE.
60  *
61  *	@(#)in_pcb.c	8.2 (Berkeley) 1/4/94
62  */
63 
64 #include <sys/cdefs.h>
65 __KERNEL_RCSID(0, "$NetBSD: in6_pcb.c,v 1.150 2016/09/29 12:19:47 roy Exp $");
66 
67 #ifdef _KERNEL_OPT
68 #include "opt_inet.h"
69 #include "opt_ipsec.h"
70 #endif
71 
72 #include <sys/param.h>
73 #include <sys/systm.h>
74 #include <sys/mbuf.h>
75 #include <sys/protosw.h>
76 #include <sys/socket.h>
77 #include <sys/socketvar.h>
78 #include <sys/ioctl.h>
79 #include <sys/errno.h>
80 #include <sys/time.h>
81 #include <sys/proc.h>
82 #include <sys/kauth.h>
83 #include <sys/domain.h>
84 #include <sys/once.h>
85 
86 #include <net/if.h>
87 #include <net/route.h>
88 
89 #include <netinet/in.h>
90 #include <netinet/in_var.h>
91 #include <netinet/in_systm.h>
92 #include <netinet/ip.h>
93 #include <netinet/in_pcb.h>
94 #include <netinet/ip6.h>
95 #include <netinet/portalgo.h>
96 #include <netinet6/ip6_var.h>
97 #include <netinet6/in6_pcb.h>
98 #include <netinet6/scope6_var.h>
99 #include <netinet6/nd6.h>
100 
101 #include "faith.h"
102 
103 #ifdef IPSEC
104 #include <netipsec/ipsec.h>
105 #include <netipsec/ipsec6.h>
106 #include <netipsec/key.h>
107 #endif /* IPSEC */
108 
109 #include <netinet/tcp_vtw.h>
110 
111 const struct in6_addr zeroin6_addr;
112 
113 #define	IN6PCBHASH_PORT(table, lport) \
114 	&(table)->inpt_porthashtbl[ntohs(lport) & (table)->inpt_porthash]
115 #define IN6PCBHASH_BIND(table, laddr, lport) \
116 	&(table)->inpt_bindhashtbl[ \
117 	    (((laddr)->s6_addr32[0] ^ (laddr)->s6_addr32[1] ^ \
118 	      (laddr)->s6_addr32[2] ^ (laddr)->s6_addr32[3]) + ntohs(lport)) & \
119 	    (table)->inpt_bindhash]
120 #define IN6PCBHASH_CONNECT(table, faddr, fport, laddr, lport) \
121 	&(table)->inpt_bindhashtbl[ \
122 	    ((((faddr)->s6_addr32[0] ^ (faddr)->s6_addr32[1] ^ \
123 	      (faddr)->s6_addr32[2] ^ (faddr)->s6_addr32[3]) + ntohs(fport)) + \
124 	     (((laddr)->s6_addr32[0] ^ (laddr)->s6_addr32[1] ^ \
125 	      (laddr)->s6_addr32[2] ^ (laddr)->s6_addr32[3]) + \
126 	      ntohs(lport))) & (table)->inpt_bindhash]
127 
128 int ip6_anonportmin = IPV6PORT_ANONMIN;
129 int ip6_anonportmax = IPV6PORT_ANONMAX;
130 int ip6_lowportmin  = IPV6PORT_RESERVEDMIN;
131 int ip6_lowportmax  = IPV6PORT_RESERVEDMAX;
132 
133 static struct pool in6pcb_pool;
134 
135 static int
136 in6pcb_poolinit(void)
137 {
138 
139 	pool_init(&in6pcb_pool, sizeof(struct in6pcb), 0, 0, 0, "in6pcbpl",
140 	    NULL, IPL_SOFTNET);
141 	return 0;
142 }
143 
144 void
145 in6_pcbinit(struct inpcbtable *table, int bindhashsize, int connecthashsize)
146 {
147 	static ONCE_DECL(control);
148 
149 	in_pcbinit(table, bindhashsize, connecthashsize);
150 	table->inpt_lastport = (u_int16_t)ip6_anonportmax;
151 
152 	RUN_ONCE(&control, in6pcb_poolinit);
153 }
154 
155 int
156 in6_pcballoc(struct socket *so, void *v)
157 {
158 	struct inpcbtable *table = v;
159 	struct in6pcb *in6p;
160 	int s;
161 
162 	s = splnet();
163 	in6p = pool_get(&in6pcb_pool, PR_NOWAIT);
164 	splx(s);
165 	if (in6p == NULL)
166 		return (ENOBUFS);
167 	memset((void *)in6p, 0, sizeof(*in6p));
168 	in6p->in6p_af = AF_INET6;
169 	in6p->in6p_table = table;
170 	in6p->in6p_socket = so;
171 	in6p->in6p_hops = -1;	/* use kernel default */
172 	in6p->in6p_icmp6filt = NULL;
173 	in6p->in6p_portalgo = PORTALGO_DEFAULT;
174 	in6p->in6p_bindportonsend = false;
175 #if defined(IPSEC)
176 	if (ipsec_enabled) {
177 		int error = ipsec_init_pcbpolicy(so, &in6p->in6p_sp);
178 		if (error != 0) {
179 			s = splnet();
180 			pool_put(&in6pcb_pool, in6p);
181 			splx(s);
182 			return error;
183 		}
184 	}
185 #endif /* IPSEC */
186 	s = splnet();
187 	TAILQ_INSERT_HEAD(&table->inpt_queue, (struct inpcb_hdr*)in6p,
188 	    inph_queue);
189 	LIST_INSERT_HEAD(IN6PCBHASH_PORT(table, in6p->in6p_lport),
190 	    &in6p->in6p_head, inph_lhash);
191 	in6_pcbstate(in6p, IN6P_ATTACHED);
192 	splx(s);
193 	if (ip6_v6only)
194 		in6p->in6p_flags |= IN6P_IPV6_V6ONLY;
195 	so->so_pcb = (void *)in6p;
196 	return (0);
197 }
198 
199 /*
200  * Bind address from sin6 to in6p.
201  */
202 static int
203 in6_pcbbind_addr(struct in6pcb *in6p, struct sockaddr_in6 *sin6, struct lwp *l)
204 {
205 	int error;
206 	int s;
207 
208 	/*
209 	 * We should check the family, but old programs
210 	 * incorrectly fail to intialize it.
211 	 */
212 	if (sin6->sin6_family != AF_INET6)
213 		return (EAFNOSUPPORT);
214 
215 #ifndef INET
216 	if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr))
217 		return (EADDRNOTAVAIL);
218 #endif
219 
220 	if ((error = sa6_embedscope(sin6, ip6_use_defzone)) != 0)
221 		return (error);
222 
223 	s = pserialize_read_enter();
224 	if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
225 		if ((in6p->in6p_flags & IN6P_IPV6_V6ONLY) != 0) {
226 			error = EINVAL;
227 			goto out;
228 		}
229 		if (sin6->sin6_addr.s6_addr32[3]) {
230 			struct sockaddr_in sin;
231 
232 			memset(&sin, 0, sizeof(sin));
233 			sin.sin_len = sizeof(sin);
234 			sin.sin_family = AF_INET;
235 			bcopy(&sin6->sin6_addr.s6_addr32[3],
236 			    &sin.sin_addr, sizeof(sin.sin_addr));
237 			if (!IN_MULTICAST(sin.sin_addr.s_addr)) {
238 				struct ifaddr *ifa;
239 				ifa = ifa_ifwithaddr((struct sockaddr *)&sin);
240 				if (ifa == NULL) {
241 					error = EADDRNOTAVAIL;
242 					goto out;
243 				}
244 			}
245 		}
246 	} else if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
247 		// succeed
248 	} else if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
249 		struct ifaddr *ifa = NULL;
250 
251 		if ((in6p->in6p_flags & IN6P_FAITH) == 0) {
252 			ifa = ifa_ifwithaddr(sin6tosa(sin6));
253 			if (ifa == NULL) {
254 				error = EADDRNOTAVAIL;
255 				goto out;
256 			}
257 		}
258 
259 		/*
260 		 * bind to an anycast address might accidentally
261 		 * cause sending a packet with an anycast source
262 		 * address, so we forbid it.
263 		 *
264 		 * We should allow to bind to a deprecated address,
265 		 * since the application dare to use it.
266 		 * But, can we assume that they are careful enough
267 		 * to check if the address is deprecated or not?
268 		 * Maybe, as a safeguard, we should have a setsockopt
269 		 * flag to control the bind(2) behavior against
270 		 * deprecated addresses (default: forbid bind(2)).
271 		 */
272 		if (ifa &&
273 		    ifatoia6(ifa)->ia6_flags &
274 		    (IN6_IFF_ANYCAST | IN6_IFF_DUPLICATED)) {
275 			error = EADDRNOTAVAIL;
276 			goto out;
277 		}
278 	}
279 	in6p->in6p_laddr = sin6->sin6_addr;
280 	error = 0;
281 out:
282 	pserialize_read_exit(s);
283 	return error;
284 }
285 
286 /*
287  * Bind port from sin6 to in6p.
288  */
289 static int
290 in6_pcbbind_port(struct in6pcb *in6p, struct sockaddr_in6 *sin6, struct lwp *l)
291 {
292 	struct inpcbtable *table = in6p->in6p_table;
293 	struct socket *so = in6p->in6p_socket;
294 	int wild = 0, reuseport = (so->so_options & SO_REUSEPORT);
295 	int error;
296 
297 	if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0 &&
298 	   ((so->so_proto->pr_flags & PR_CONNREQUIRED) == 0 ||
299 	    (so->so_options & SO_ACCEPTCONN) == 0))
300 		wild = 1;
301 
302 	if (sin6->sin6_port != 0) {
303 		enum kauth_network_req req;
304 
305 #ifndef IPNOPRIVPORTS
306 		if (ntohs(sin6->sin6_port) < IPV6PORT_RESERVED)
307 			req = KAUTH_REQ_NETWORK_BIND_PRIVPORT;
308 		else
309 #endif /* IPNOPRIVPORTS */
310 			req = KAUTH_REQ_NETWORK_BIND_PORT;
311 
312 		error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_BIND,
313 		    req, so, sin6, NULL);
314 		if (error)
315 			return (EACCES);
316 	}
317 
318 	if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
319 		/*
320 		 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
321 		 * allow compepte duplication of binding if
322 		 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
323 		 * and a multicast address is bound on both
324 		 * new and duplicated sockets.
325 		 */
326 		if (so->so_options & (SO_REUSEADDR | SO_REUSEPORT))
327 			reuseport = SO_REUSEADDR|SO_REUSEPORT;
328 	}
329 
330 	if (sin6->sin6_port != 0) {
331 		if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
332 #ifdef INET
333 			struct inpcb *t;
334 			struct vestigial_inpcb vestige;
335 
336 			t = in_pcblookup_port(table,
337 			    *(struct in_addr *)&sin6->sin6_addr.s6_addr32[3],
338 			    sin6->sin6_port, wild, &vestige);
339 			if (t && (reuseport & t->inp_socket->so_options) == 0)
340 				return (EADDRINUSE);
341 			if (!t
342 			    && vestige.valid
343 			    && !(reuseport && vestige.reuse_port))
344 			    return EADDRINUSE;
345 #else
346 			return (EADDRNOTAVAIL);
347 #endif
348 		}
349 
350 		{
351 			struct in6pcb *t;
352 			struct vestigial_inpcb vestige;
353 
354 			t = in6_pcblookup_port(table, &sin6->sin6_addr,
355 			    sin6->sin6_port, wild, &vestige);
356 			if (t && (reuseport & t->in6p_socket->so_options) == 0)
357 				return (EADDRINUSE);
358 			if (!t
359 			    && vestige.valid
360 			    && !(reuseport && vestige.reuse_port))
361 			    return EADDRINUSE;
362 		}
363 	}
364 
365 	if (sin6->sin6_port == 0) {
366 		int e;
367 		e = in6_pcbsetport(sin6, in6p, l);
368 		if (e != 0)
369 			return (e);
370 	} else {
371 		in6p->in6p_lport = sin6->sin6_port;
372 		in6_pcbstate(in6p, IN6P_BOUND);
373 	}
374 
375 	LIST_REMOVE(&in6p->in6p_head, inph_lhash);
376 	LIST_INSERT_HEAD(IN6PCBHASH_PORT(table, in6p->in6p_lport),
377 	    &in6p->in6p_head, inph_lhash);
378 
379 	return (0);
380 }
381 
382 int
383 in6_pcbbind(void *v, struct sockaddr_in6 *sin6, struct lwp *l)
384 {
385 	struct in6pcb *in6p = v;
386 	struct sockaddr_in6 lsin6;
387 	int error;
388 
389 	if (in6p->in6p_af != AF_INET6)
390 		return (EINVAL);
391 
392 	/*
393 	 * If we already have a local port or a local address it means we're
394 	 * bounded.
395 	 */
396 	if (in6p->in6p_lport || !(IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) ||
397 	    (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr) &&
398 	      in6p->in6p_laddr.s6_addr32[3] == 0)))
399 		return (EINVAL);
400 
401 	if (NULL != sin6) {
402 		/* We were provided a sockaddr_in6 to use. */
403 		if (sin6->sin6_len != sizeof(*sin6))
404 			return (EINVAL);
405 	} else {
406 		/* We always bind to *something*, even if it's "anything". */
407 		lsin6 = *((const struct sockaddr_in6 *)
408 		    in6p->in6p_socket->so_proto->pr_domain->dom_sa_any);
409 		sin6 = &lsin6;
410 	}
411 
412 	/* Bind address. */
413 	error = in6_pcbbind_addr(in6p, sin6, l);
414 	if (error)
415 		return (error);
416 
417 	/* Bind port. */
418 	error = in6_pcbbind_port(in6p, sin6, l);
419 	if (error) {
420 		/*
421 		 * Reset the address here to "any" so we don't "leak" the
422 		 * in6pcb.
423 		 */
424 		in6p->in6p_laddr = in6addr_any;
425 
426 		return (error);
427 	}
428 
429 
430 #if 0
431 	in6p->in6p_flowinfo = 0;	/* XXX */
432 #endif
433 	return (0);
434 }
435 
436 /*
437  * Connect from a socket to a specified address.
438  * Both address and port must be specified in argument sin6.
439  * If don't have a local address for this socket yet,
440  * then pick one.
441  */
442 int
443 in6_pcbconnect(void *v, struct sockaddr_in6 *sin6, struct lwp *l)
444 {
445 	struct in6pcb *in6p = v;
446 	struct in6_addr *in6a = NULL;
447 	struct ifnet *ifp = NULL;	/* outgoing interface */
448 	int error = 0;
449 	int scope_ambiguous = 0;
450 #ifdef INET
451 	struct in6_addr mapped;
452 #endif
453 	struct sockaddr_in6 tmp;
454 	struct vestigial_inpcb vestige;
455 	struct psref psref;
456 	int bound;
457 
458 	(void)&in6a;				/* XXX fool gcc */
459 
460 	if (in6p->in6p_af != AF_INET6)
461 		return (EINVAL);
462 
463 	if (sin6->sin6_len != sizeof(*sin6))
464 		return (EINVAL);
465 	if (sin6->sin6_family != AF_INET6)
466 		return (EAFNOSUPPORT);
467 	if (sin6->sin6_port == 0)
468 		return (EADDRNOTAVAIL);
469 
470 	if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr) &&
471 	    in6p->in6p_socket->so_type == SOCK_STREAM)
472 		return EADDRNOTAVAIL;
473 
474 	if (sin6->sin6_scope_id == 0 && !ip6_use_defzone)
475 		scope_ambiguous = 1;
476 	if ((error = sa6_embedscope(sin6, ip6_use_defzone)) != 0)
477 		return(error);
478 
479 	/* sanity check for mapped address case */
480 	if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
481 		if ((in6p->in6p_flags & IN6P_IPV6_V6ONLY) != 0)
482 			return EINVAL;
483 		if (IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr))
484 			in6p->in6p_laddr.s6_addr16[5] = htons(0xffff);
485 		if (!IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr))
486 			return EINVAL;
487 	} else
488 	{
489 		if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr))
490 			return EINVAL;
491 	}
492 
493 	/* protect *sin6 from overwrites */
494 	tmp = *sin6;
495 	sin6 = &tmp;
496 
497 	bound = curlwp_bind();
498 	/* Source address selection. */
499 	if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr) &&
500 	    in6p->in6p_laddr.s6_addr32[3] == 0) {
501 #ifdef INET
502 		struct sockaddr_in sin;
503 		struct in_ifaddr *ia4;
504 		struct psref _psref;
505 
506 		memset(&sin, 0, sizeof(sin));
507 		sin.sin_len = sizeof(sin);
508 		sin.sin_family = AF_INET;
509 		memcpy(&sin.sin_addr, &sin6->sin6_addr.s6_addr32[3],
510 			sizeof(sin.sin_addr));
511 		ia4 = in_selectsrc(&sin, &in6p->in6p_route,
512 			in6p->in6p_socket->so_options, NULL, &error, &_psref);
513 		if (ia4 == NULL) {
514 			if (error == 0)
515 				error = EADDRNOTAVAIL;
516 			return (error);
517 		}
518 		memset(&mapped, 0, sizeof(mapped));
519 		mapped.s6_addr16[5] = htons(0xffff);
520 		memcpy(&mapped.s6_addr32[3], &IA_SIN(ia4)->sin_addr,
521 		    sizeof(IA_SIN(ia4)->sin_addr));
522 		ia4_release(ia4, &_psref);
523 		in6a = &mapped;
524 #else
525 		return EADDRNOTAVAIL;
526 #endif
527 	} else {
528 		/*
529 		 * XXX: in6_selectsrc might replace the bound local address
530 		 * with the address specified by setsockopt(IPV6_PKTINFO).
531 		 * Is it the intended behavior?
532 		 */
533 		in6a = in6_selectsrc(sin6, in6p->in6p_outputopts,
534 				     in6p->in6p_moptions,
535 				     &in6p->in6p_route,
536 				     &in6p->in6p_laddr, &ifp, &psref, &error);
537 		if (ifp && scope_ambiguous &&
538 		    (error = in6_setscope(&sin6->sin6_addr, ifp, NULL)) != 0) {
539 			if_put(ifp, &psref);
540 			curlwp_bindx(bound);
541 			return(error);
542 		}
543 
544 		if (in6a == NULL) {
545 			if_put(ifp, &psref);
546 			curlwp_bindx(bound);
547 			if (error == 0)
548 				error = EADDRNOTAVAIL;
549 			return (error);
550 		}
551 	}
552 
553 	if (ifp != NULL) {
554 		in6p->in6p_ip6.ip6_hlim = (u_int8_t)in6_selecthlim(in6p, ifp);
555 		if_put(ifp, &psref);
556 	} else
557 		in6p->in6p_ip6.ip6_hlim = (u_int8_t)in6_selecthlim_rt(in6p);
558 	curlwp_bindx(bound);
559 
560 	if (in6_pcblookup_connect(in6p->in6p_table, &sin6->sin6_addr,
561 	    sin6->sin6_port,
562 	    IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) ? in6a : &in6p->in6p_laddr,
563 				  in6p->in6p_lport, 0, &vestige)
564 		|| vestige.valid)
565 		return (EADDRINUSE);
566 	if (IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) ||
567 	    (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr) &&
568 	     in6p->in6p_laddr.s6_addr32[3] == 0))
569 	{
570 		if (in6p->in6p_lport == 0) {
571 			error = in6_pcbbind(in6p, NULL, l);
572 			if (error != 0)
573 				return error;
574 		}
575 		in6p->in6p_laddr = *in6a;
576 	}
577 	in6p->in6p_faddr = sin6->sin6_addr;
578 	in6p->in6p_fport = sin6->sin6_port;
579 
580         /* Late bind, if needed */
581 	if (in6p->in6p_bindportonsend) {
582                struct sockaddr_in6 lsin = *((const struct sockaddr_in6 *)
583 		    in6p->in6p_socket->so_proto->pr_domain->dom_sa_any);
584 		lsin.sin6_addr = in6p->in6p_laddr;
585 		lsin.sin6_port = 0;
586 
587                if ((error = in6_pcbbind_port(in6p, &lsin, l)) != 0)
588                        return error;
589 	}
590 
591 	in6_pcbstate(in6p, IN6P_CONNECTED);
592 	in6p->in6p_flowinfo &= ~IPV6_FLOWLABEL_MASK;
593 	if (ip6_auto_flowlabel)
594 		in6p->in6p_flowinfo |=
595 		    (htonl(ip6_randomflowlabel()) & IPV6_FLOWLABEL_MASK);
596 #if defined(IPSEC)
597 	if (ipsec_enabled && in6p->in6p_socket->so_type == SOCK_STREAM)
598 		ipsec_pcbconn(in6p->in6p_sp);
599 #endif
600 	return (0);
601 }
602 
603 void
604 in6_pcbdisconnect(struct in6pcb *in6p)
605 {
606 	memset((void *)&in6p->in6p_faddr, 0, sizeof(in6p->in6p_faddr));
607 	in6p->in6p_fport = 0;
608 	in6_pcbstate(in6p, IN6P_BOUND);
609 	in6p->in6p_flowinfo &= ~IPV6_FLOWLABEL_MASK;
610 #if defined(IPSEC)
611 	if (ipsec_enabled)
612 		ipsec_pcbdisconn(in6p->in6p_sp);
613 #endif
614 	if (in6p->in6p_socket->so_state & SS_NOFDREF)
615 		in6_pcbdetach(in6p);
616 }
617 
618 void
619 in6_pcbdetach(struct in6pcb *in6p)
620 {
621 	struct socket *so = in6p->in6p_socket;
622 	int s;
623 
624 	if (in6p->in6p_af != AF_INET6)
625 		return;
626 
627 #if defined(IPSEC)
628 	if (ipsec_enabled)
629 		ipsec6_delete_pcbpolicy(in6p);
630 #endif
631 	so->so_pcb = NULL;
632 
633 	s = splnet();
634 	in6_pcbstate(in6p, IN6P_ATTACHED);
635 	LIST_REMOVE(&in6p->in6p_head, inph_lhash);
636 	TAILQ_REMOVE(&in6p->in6p_table->inpt_queue, &in6p->in6p_head,
637 	    inph_queue);
638 	splx(s);
639 
640 	if (in6p->in6p_options) {
641 		m_freem(in6p->in6p_options);
642 	}
643 	if (in6p->in6p_outputopts != NULL) {
644 		ip6_clearpktopts(in6p->in6p_outputopts, -1);
645 		free(in6p->in6p_outputopts, M_IP6OPT);
646 	}
647 	rtcache_free(&in6p->in6p_route);
648 	ip6_freemoptions(in6p->in6p_moptions);
649 	ip_freemoptions(in6p->in6p_v4moptions);
650 	sofree(so);				/* drops the socket's lock */
651 
652 	pool_put(&in6pcb_pool, in6p);
653 	mutex_enter(softnet_lock);		/* reacquire it */
654 }
655 
656 void
657 in6_setsockaddr(struct in6pcb *in6p, struct sockaddr_in6 *sin6)
658 {
659 
660 	if (in6p->in6p_af != AF_INET6)
661 		return;
662 
663 	sockaddr_in6_init(sin6, &in6p->in6p_laddr, in6p->in6p_lport, 0, 0);
664 	(void)sa6_recoverscope(sin6); /* XXX: should catch errors */
665 }
666 
667 void
668 in6_setpeeraddr(struct in6pcb *in6p, struct sockaddr_in6 *sin6)
669 {
670 
671 	if (in6p->in6p_af != AF_INET6)
672 		return;
673 
674 	sockaddr_in6_init(sin6, &in6p->in6p_faddr, in6p->in6p_fport, 0, 0);
675 	(void)sa6_recoverscope(sin6); /* XXX: should catch errors */
676 }
677 
678 /*
679  * Pass some notification to all connections of a protocol
680  * associated with address dst.  The local address and/or port numbers
681  * may be specified to limit the search.  The "usual action" will be
682  * taken, depending on the ctlinput cmd.  The caller must filter any
683  * cmds that are uninteresting (e.g., no error in the map).
684  * Call the protocol specific routine (if any) to report
685  * any errors for each matching socket.
686  *
687  * Must be called at splsoftnet.
688  *
689  * Note: src (4th arg) carries the flowlabel value on the original IPv6
690  * header, in sin6_flowinfo member.
691  */
692 int
693 in6_pcbnotify(struct inpcbtable *table, const struct sockaddr *dst,
694     u_int fport_arg, const struct sockaddr *src, u_int lport_arg, int cmd,
695     void *cmdarg, void (*notify)(struct in6pcb *, int))
696 {
697 	struct rtentry *rt;
698 	struct inpcb_hdr *inph, *ninph;
699 	struct sockaddr_in6 sa6_src;
700 	const struct sockaddr_in6 *sa6_dst;
701 	u_int16_t fport = fport_arg, lport = lport_arg;
702 	int errno;
703 	int nmatch = 0;
704 	u_int32_t flowinfo;
705 
706 	if ((unsigned)cmd >= PRC_NCMDS || dst->sa_family != AF_INET6)
707 		return 0;
708 
709 	sa6_dst = (const struct sockaddr_in6 *)dst;
710 	if (IN6_IS_ADDR_UNSPECIFIED(&sa6_dst->sin6_addr))
711 		return 0;
712 
713 	/*
714 	 * note that src can be NULL when we get notify by local fragmentation.
715 	 */
716 	sa6_src = (src == NULL) ? sa6_any : *(const struct sockaddr_in6 *)src;
717 	flowinfo = sa6_src.sin6_flowinfo;
718 
719 	/*
720 	 * Redirects go to all references to the destination,
721 	 * and use in6_rtchange to invalidate the route cache.
722 	 * Dead host indications: also use in6_rtchange to invalidate
723 	 * the cache, and deliver the error to all the sockets.
724 	 * Otherwise, if we have knowledge of the local port and address,
725 	 * deliver only to that socket.
726 	 */
727 	if (PRC_IS_REDIRECT(cmd) || cmd == PRC_HOSTDEAD) {
728 		fport = 0;
729 		lport = 0;
730 		memset((void *)&sa6_src.sin6_addr, 0, sizeof(sa6_src.sin6_addr));
731 
732 		if (cmd != PRC_HOSTDEAD)
733 			notify = in6_rtchange;
734 	}
735 
736 	errno = inet6ctlerrmap[cmd];
737 	TAILQ_FOREACH_SAFE(inph, &table->inpt_queue, inph_queue, ninph) {
738 		struct in6pcb *in6p = (struct in6pcb *)inph;
739 		if (in6p->in6p_af != AF_INET6)
740 			continue;
741 
742 		/*
743 		 * Under the following condition, notify of redirects
744 		 * to the pcb, without making address matches against inpcb.
745 		 * - redirect notification is arrived.
746 		 * - the inpcb is unconnected.
747 		 * - the inpcb is caching !RTF_HOST routing entry.
748 		 * - the ICMPv6 notification is from the gateway cached in the
749 		 *   inpcb.  i.e. ICMPv6 notification is from nexthop gateway
750 		 *   the inpcb used very recently.
751 		 *
752 		 * This is to improve interaction between netbsd/openbsd
753 		 * redirect handling code, and inpcb route cache code.
754 		 * without the clause, !RTF_HOST routing entry (which carries
755 		 * gateway used by inpcb right before the ICMPv6 redirect)
756 		 * will be cached forever in unconnected inpcb.
757 		 *
758 		 * There still is a question regarding to what is TRT:
759 		 * - On bsdi/freebsd, RTF_HOST (cloned) routing entry will be
760 		 *   generated on packet output.  inpcb will always cache
761 		 *   RTF_HOST routing entry so there's no need for the clause
762 		 *   (ICMPv6 redirect will update RTF_HOST routing entry,
763 		 *   and inpcb is caching it already).
764 		 *   However, bsdi/freebsd are vulnerable to local DoS attacks
765 		 *   due to the cloned routing entries.
766 		 * - Specwise, "destination cache" is mentioned in RFC2461.
767 		 *   Jinmei says that it implies bsdi/freebsd behavior, itojun
768 		 *   is not really convinced.
769 		 * - Having hiwat/lowat on # of cloned host route (redirect/
770 		 *   pmtud) may be a good idea.  netbsd/openbsd has it.  see
771 		 *   icmp6_mtudisc_update().
772 		 */
773 		if ((PRC_IS_REDIRECT(cmd) || cmd == PRC_HOSTDEAD) &&
774 		    IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
775 		    (rt = rtcache_validate(&in6p->in6p_route)) != NULL &&
776 		    !(rt->rt_flags & RTF_HOST)) {
777 			const struct sockaddr_in6 *dst6;
778 
779 			dst6 = (const struct sockaddr_in6 *)
780 			    rtcache_getdst(&in6p->in6p_route);
781 			if (dst6 == NULL)
782 				;
783 			else if (IN6_ARE_ADDR_EQUAL(&dst6->sin6_addr,
784 			    &sa6_dst->sin6_addr))
785 				goto do_notify;
786 		}
787 
788 		/*
789 		 * If the error designates a new path MTU for a destination
790 		 * and the application (associated with this socket) wanted to
791 		 * know the value, notify. Note that we notify for all
792 		 * disconnected sockets if the corresponding application
793 		 * wanted. This is because some UDP applications keep sending
794 		 * sockets disconnected.
795 		 * XXX: should we avoid to notify the value to TCP sockets?
796 		 */
797 		if (cmd == PRC_MSGSIZE && (in6p->in6p_flags & IN6P_MTU) != 0 &&
798 		    (IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) ||
799 		     IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &sa6_dst->sin6_addr))) {
800 			ip6_notify_pmtu(in6p, (const struct sockaddr_in6 *)dst,
801 					(u_int32_t *)cmdarg);
802 		}
803 
804 		/*
805 		 * Detect if we should notify the error. If no source and
806 		 * destination ports are specified, but non-zero flowinfo and
807 		 * local address match, notify the error. This is the case
808 		 * when the error is delivered with an encrypted buffer
809 		 * by ESP. Otherwise, just compare addresses and ports
810 		 * as usual.
811 		 */
812 		if (lport == 0 && fport == 0 && flowinfo &&
813 		    in6p->in6p_socket != NULL &&
814 		    flowinfo == (in6p->in6p_flowinfo & IPV6_FLOWLABEL_MASK) &&
815 		    IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &sa6_src.sin6_addr))
816 			goto do_notify;
817 		else if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr,
818 					     &sa6_dst->sin6_addr) ||
819 		    in6p->in6p_socket == NULL ||
820 		    (lport && in6p->in6p_lport != lport) ||
821 		    (!IN6_IS_ADDR_UNSPECIFIED(&sa6_src.sin6_addr) &&
822 		     !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr,
823 					 &sa6_src.sin6_addr)) ||
824 		    (fport && in6p->in6p_fport != fport))
825 			continue;
826 
827 	  do_notify:
828 		if (notify)
829 			(*notify)(in6p, errno);
830 		nmatch++;
831 	}
832 	return nmatch;
833 }
834 
835 void
836 in6_pcbpurgeif0(struct inpcbtable *table, struct ifnet *ifp)
837 {
838 	struct inpcb_hdr *inph, *ninph;
839 	struct ip6_moptions *im6o;
840 	struct in6_multi_mship *imm, *nimm;
841 
842 	KASSERT(ifp != NULL);
843 
844 	TAILQ_FOREACH_SAFE(inph, &table->inpt_queue, inph_queue, ninph) {
845 		struct in6pcb *in6p = (struct in6pcb *)inph;
846 		if (in6p->in6p_af != AF_INET6)
847 			continue;
848 
849 		im6o = in6p->in6p_moptions;
850 		if (im6o) {
851 			/*
852 			 * Unselect the outgoing interface if it is being
853 			 * detached.
854 			 */
855 			if (im6o->im6o_multicast_if_index == ifp->if_index)
856 				im6o->im6o_multicast_if_index = 0;
857 
858 			/*
859 			 * Drop multicast group membership if we joined
860 			 * through the interface being detached.
861 			 * XXX controversial - is it really legal for kernel
862 			 * to force this?
863 			 */
864 			for (imm = im6o->im6o_memberships.lh_first;
865 			     imm != NULL; imm = nimm) {
866 				nimm = imm->i6mm_chain.le_next;
867 				if (imm->i6mm_maddr->in6m_ifp == ifp) {
868 					LIST_REMOVE(imm, i6mm_chain);
869 					in6_leavegroup(imm);
870 				}
871 			}
872 		}
873 		in_purgeifmcast(in6p->in6p_v4moptions, ifp);
874 	}
875 }
876 
877 void
878 in6_pcbpurgeif(struct inpcbtable *table, struct ifnet *ifp)
879 {
880 	struct rtentry *rt;
881 	struct inpcb_hdr *inph, *ninph;
882 
883 	TAILQ_FOREACH_SAFE(inph, &table->inpt_queue, inph_queue, ninph) {
884 		struct in6pcb *in6p = (struct in6pcb *)inph;
885 		if (in6p->in6p_af != AF_INET6)
886 			continue;
887 		if ((rt = rtcache_validate(&in6p->in6p_route)) != NULL &&
888 		    rt->rt_ifp == ifp)
889 			in6_rtchange(in6p, 0);
890 	}
891 }
892 
893 /*
894  * Check for alternatives when higher level complains
895  * about service problems.  For now, invalidate cached
896  * routing information.  If the route was created dynamically
897  * (by a redirect), time to try a default gateway again.
898  */
899 void
900 in6_losing(struct in6pcb *in6p)
901 {
902 	struct rtentry *rt;
903 	struct rt_addrinfo info;
904 
905 	if (in6p->in6p_af != AF_INET6)
906 		return;
907 
908 	if ((rt = rtcache_validate(&in6p->in6p_route)) == NULL)
909 		return;
910 
911 	memset(&info, 0, sizeof(info));
912 	info.rti_info[RTAX_DST] = rtcache_getdst(&in6p->in6p_route);
913 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
914 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
915 	rt_missmsg(RTM_LOSING, &info, rt->rt_flags, 0);
916 	if (rt->rt_flags & RTF_DYNAMIC) {
917 		(void)rtrequest(RTM_DELETE, rt_getkey(rt),
918 		    rt->rt_gateway, rt_mask(rt), rt->rt_flags, NULL);
919 	}
920 	/*
921 	 * A new route can be allocated
922 	 * the next time output is attempted.
923 	 */
924 	rtcache_free(&in6p->in6p_route);
925 }
926 
927 /*
928  * After a routing change, flush old routing.  A new route can be
929  * allocated the next time output is attempted.
930  */
931 void
932 in6_rtchange(struct in6pcb *in6p, int errno)
933 {
934 	if (in6p->in6p_af != AF_INET6)
935 		return;
936 
937 	rtcache_free(&in6p->in6p_route);
938 	/*
939 	 * A new route can be allocated the next time
940 	 * output is attempted.
941 	 */
942 }
943 
944 struct in6pcb *
945 in6_pcblookup_port(struct inpcbtable *table, struct in6_addr *laddr6,
946 		   u_int lport_arg, int lookup_wildcard, struct vestigial_inpcb *vp)
947 {
948 	struct inpcbhead *head;
949 	struct inpcb_hdr *inph;
950 	struct in6pcb *in6p, *match = NULL;
951 	int matchwild = 3, wildcard;
952 	u_int16_t lport = lport_arg;
953 
954 	if (vp)
955 		vp->valid = 0;
956 
957 	head = IN6PCBHASH_PORT(table, lport);
958 	LIST_FOREACH(inph, head, inph_lhash) {
959 		in6p = (struct in6pcb *)inph;
960 		if (in6p->in6p_af != AF_INET6)
961 			continue;
962 
963 		if (in6p->in6p_lport != lport)
964 			continue;
965 		wildcard = 0;
966 		if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_faddr)) {
967 			if ((in6p->in6p_flags & IN6P_IPV6_V6ONLY) != 0)
968 				continue;
969 		}
970 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr))
971 			wildcard++;
972 		if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr)) {
973 			if ((in6p->in6p_flags & IN6P_IPV6_V6ONLY) != 0)
974 				continue;
975 			if (!IN6_IS_ADDR_V4MAPPED(laddr6))
976 				continue;
977 
978 			/* duplicate of IPv4 logic */
979 			wildcard = 0;
980 			if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_faddr) &&
981 			    in6p->in6p_faddr.s6_addr32[3])
982 				wildcard++;
983 			if (!in6p->in6p_laddr.s6_addr32[3]) {
984 				if (laddr6->s6_addr32[3])
985 					wildcard++;
986 			} else {
987 				if (!laddr6->s6_addr32[3])
988 					wildcard++;
989 				else {
990 					if (in6p->in6p_laddr.s6_addr32[3] !=
991 					    laddr6->s6_addr32[3])
992 						continue;
993 				}
994 			}
995 		} else if (IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
996 			if (IN6_IS_ADDR_V4MAPPED(laddr6)) {
997 				if ((in6p->in6p_flags & IN6P_IPV6_V6ONLY) != 0)
998 					continue;
999 			}
1000 			if (!IN6_IS_ADDR_UNSPECIFIED(laddr6))
1001 				wildcard++;
1002 		} else {
1003 			if (IN6_IS_ADDR_V4MAPPED(laddr6)) {
1004 				if ((in6p->in6p_flags & IN6P_IPV6_V6ONLY) != 0)
1005 					continue;
1006 			}
1007 			if (IN6_IS_ADDR_UNSPECIFIED(laddr6))
1008 				wildcard++;
1009 			else {
1010 				if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr,
1011 				    laddr6))
1012 					continue;
1013 			}
1014 		}
1015 		if (wildcard && !lookup_wildcard)
1016 			continue;
1017 		if (wildcard < matchwild) {
1018 			match = in6p;
1019 			matchwild = wildcard;
1020 			if (matchwild == 0)
1021 				break;
1022 		}
1023 	}
1024 	if (match && matchwild == 0)
1025 		return match;
1026 
1027 	if (vp && table->vestige && table->vestige->init_ports6) {
1028 		struct vestigial_inpcb better;
1029 		void *state;
1030 
1031 		state = (*table->vestige->init_ports6)(laddr6,
1032 						       lport_arg,
1033 						       lookup_wildcard);
1034 		while (table->vestige
1035 		       && (*table->vestige->next_port6)(state, vp)) {
1036 
1037 			if (vp->lport != lport)
1038 				continue;
1039 			wildcard = 0;
1040 			if (!IN6_IS_ADDR_UNSPECIFIED(&vp->faddr.v6))
1041 				wildcard++;
1042 			if (IN6_IS_ADDR_UNSPECIFIED(&vp->laddr.v6)) {
1043 				if (!IN6_IS_ADDR_UNSPECIFIED(laddr6))
1044 					wildcard++;
1045 			} else {
1046 				if (IN6_IS_ADDR_V4MAPPED(laddr6)) {
1047 					if (vp->v6only)
1048 						continue;
1049 				}
1050 				if (IN6_IS_ADDR_UNSPECIFIED(laddr6))
1051 					wildcard++;
1052 				else {
1053 					if (!IN6_ARE_ADDR_EQUAL(&vp->laddr.v6, laddr6))
1054 						continue;
1055 				}
1056 			}
1057 			if (wildcard && !lookup_wildcard)
1058 				continue;
1059 			if (wildcard < matchwild) {
1060 				better = *vp;
1061 				match  = (void*)&better;
1062 
1063 				matchwild = wildcard;
1064 				if (matchwild == 0)
1065 					break;
1066 			}
1067 		}
1068 
1069 		if (match) {
1070 			if (match != (void*)&better)
1071 				return match;
1072 			else {
1073 				*vp = better;
1074 				return 0;
1075 			}
1076 		}
1077 	}
1078 	return (match);
1079 }
1080 
1081 /*
1082  * WARNING: return value (rtentry) could be IPv4 one if in6pcb is connected to
1083  * IPv4 mapped address.
1084  */
1085 struct rtentry *
1086 in6_pcbrtentry(struct in6pcb *in6p)
1087 {
1088 	struct rtentry *rt;
1089 	struct route *ro;
1090 	union {
1091 		const struct sockaddr *sa;
1092 		const struct sockaddr_in6 *sa6;
1093 #ifdef INET
1094 		const struct sockaddr_in *sa4;
1095 #endif
1096 	} cdst;
1097 
1098 	ro = &in6p->in6p_route;
1099 
1100 	if (in6p->in6p_af != AF_INET6)
1101 		return (NULL);
1102 
1103 	cdst.sa = rtcache_getdst(ro);
1104 	if (cdst.sa == NULL)
1105 		;
1106 #ifdef INET
1107 	else if (cdst.sa->sa_family == AF_INET) {
1108 		KASSERT(IN6_IS_ADDR_V4MAPPED(&in6p->in6p_faddr));
1109 		if (cdst.sa4->sin_addr.s_addr != in6p->in6p_faddr.s6_addr32[3])
1110 			rtcache_free(ro);
1111 	}
1112 #endif
1113 	else {
1114 		if (!IN6_ARE_ADDR_EQUAL(&cdst.sa6->sin6_addr,
1115 					&in6p->in6p_faddr))
1116 			rtcache_free(ro);
1117 	}
1118 	if ((rt = rtcache_validate(ro)) == NULL)
1119 		rt = rtcache_update(ro, 1);
1120 #ifdef INET
1121 	if (rt == NULL && IN6_IS_ADDR_V4MAPPED(&in6p->in6p_faddr)) {
1122 		union {
1123 			struct sockaddr		dst;
1124 			struct sockaddr_in	dst4;
1125 		} u;
1126 		struct in_addr addr;
1127 
1128 		addr.s_addr = in6p->in6p_faddr.s6_addr32[3];
1129 
1130 		sockaddr_in_init(&u.dst4, &addr, 0);
1131 		if (rtcache_setdst(ro, &u.dst) != 0)
1132 			return NULL;
1133 
1134 		rt = rtcache_init(ro);
1135 	} else
1136 #endif
1137 	if (rt == NULL && !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) {
1138 		union {
1139 			struct sockaddr		dst;
1140 			struct sockaddr_in6	dst6;
1141 		} u;
1142 
1143 		sockaddr_in6_init(&u.dst6, &in6p->in6p_faddr, 0, 0, 0);
1144 		if (rtcache_setdst(ro, &u.dst) != 0)
1145 			return NULL;
1146 
1147 		rt = rtcache_init(ro);
1148 	}
1149 	return rt;
1150 }
1151 
1152 struct in6pcb *
1153 in6_pcblookup_connect(struct inpcbtable *table, const struct in6_addr *faddr6,
1154 		      u_int fport_arg, const struct in6_addr *laddr6, u_int lport_arg,
1155 		      int faith,
1156 		      struct vestigial_inpcb *vp)
1157 {
1158 	struct inpcbhead *head;
1159 	struct inpcb_hdr *inph;
1160 	struct in6pcb *in6p;
1161 	u_int16_t fport = fport_arg, lport = lport_arg;
1162 
1163 	if (vp)
1164 		vp->valid = 0;
1165 
1166 	head = IN6PCBHASH_CONNECT(table, faddr6, fport, laddr6, lport);
1167 	LIST_FOREACH(inph, head, inph_hash) {
1168 		in6p = (struct in6pcb *)inph;
1169 		if (in6p->in6p_af != AF_INET6)
1170 			continue;
1171 
1172 		/* find exact match on both source and dest */
1173 		if (in6p->in6p_fport != fport)
1174 			continue;
1175 		if (in6p->in6p_lport != lport)
1176 			continue;
1177 		if (IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr))
1178 			continue;
1179 		if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, faddr6))
1180 			continue;
1181 		if (IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr))
1182 			continue;
1183 		if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, laddr6))
1184 			continue;
1185 		if ((IN6_IS_ADDR_V4MAPPED(laddr6) ||
1186 		     IN6_IS_ADDR_V4MAPPED(faddr6)) &&
1187 		    (in6p->in6p_flags & IN6P_IPV6_V6ONLY))
1188 			continue;
1189 		return in6p;
1190 	}
1191 	if (vp && table->vestige) {
1192 		if ((*table->vestige->lookup6)(faddr6, fport_arg,
1193 					       laddr6, lport_arg, vp))
1194 			return NULL;
1195 	}
1196 
1197 	return NULL;
1198 }
1199 
1200 struct in6pcb *
1201 in6_pcblookup_bind(struct inpcbtable *table, const struct in6_addr *laddr6,
1202 	u_int lport_arg, int faith)
1203 {
1204 	struct inpcbhead *head;
1205 	struct inpcb_hdr *inph;
1206 	struct in6pcb *in6p;
1207 	u_int16_t lport = lport_arg;
1208 #ifdef INET
1209 	struct in6_addr zero_mapped;
1210 #endif
1211 
1212 	head = IN6PCBHASH_BIND(table, laddr6, lport);
1213 	LIST_FOREACH(inph, head, inph_hash) {
1214 		in6p = (struct in6pcb *)inph;
1215 		if (in6p->in6p_af != AF_INET6)
1216 			continue;
1217 
1218 		if (faith && (in6p->in6p_flags & IN6P_FAITH) == 0)
1219 			continue;
1220 		if (in6p->in6p_fport != 0)
1221 			continue;
1222 		if (in6p->in6p_lport != lport)
1223 			continue;
1224 		if (IN6_IS_ADDR_V4MAPPED(laddr6) &&
1225 		    (in6p->in6p_flags & IN6P_IPV6_V6ONLY) != 0)
1226 			continue;
1227 		if (IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, laddr6))
1228 			goto out;
1229 	}
1230 #ifdef INET
1231 	if (IN6_IS_ADDR_V4MAPPED(laddr6)) {
1232 		memset(&zero_mapped, 0, sizeof(zero_mapped));
1233 		zero_mapped.s6_addr16[5] = 0xffff;
1234 		head = IN6PCBHASH_BIND(table, &zero_mapped, lport);
1235 		LIST_FOREACH(inph, head, inph_hash) {
1236 			in6p = (struct in6pcb *)inph;
1237 			if (in6p->in6p_af != AF_INET6)
1238 				continue;
1239 
1240 			if (faith && (in6p->in6p_flags & IN6P_FAITH) == 0)
1241 				continue;
1242 			if (in6p->in6p_fport != 0)
1243 				continue;
1244 			if (in6p->in6p_lport != lport)
1245 				continue;
1246 			if ((in6p->in6p_flags & IN6P_IPV6_V6ONLY) != 0)
1247 				continue;
1248 			if (IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &zero_mapped))
1249 				goto out;
1250 		}
1251 	}
1252 #endif
1253 	head = IN6PCBHASH_BIND(table, &zeroin6_addr, lport);
1254 	LIST_FOREACH(inph, head, inph_hash) {
1255 		in6p = (struct in6pcb *)inph;
1256 		if (in6p->in6p_af != AF_INET6)
1257 			continue;
1258 
1259 		if (faith && (in6p->in6p_flags & IN6P_FAITH) == 0)
1260 			continue;
1261 		if (in6p->in6p_fport != 0)
1262 			continue;
1263 		if (in6p->in6p_lport != lport)
1264 			continue;
1265 		if (IN6_IS_ADDR_V4MAPPED(laddr6) &&
1266 		    (in6p->in6p_flags & IN6P_IPV6_V6ONLY) != 0)
1267 			continue;
1268 		if (IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &zeroin6_addr))
1269 			goto out;
1270 	}
1271 	return (NULL);
1272 
1273 out:
1274 	inph = &in6p->in6p_head;
1275 	if (inph != LIST_FIRST(head)) {
1276 		LIST_REMOVE(inph, inph_hash);
1277 		LIST_INSERT_HEAD(head, inph, inph_hash);
1278 	}
1279 	return in6p;
1280 }
1281 
1282 void
1283 in6_pcbstate(struct in6pcb *in6p, int state)
1284 {
1285 
1286 	if (in6p->in6p_af != AF_INET6)
1287 		return;
1288 
1289 	if (in6p->in6p_state > IN6P_ATTACHED)
1290 		LIST_REMOVE(&in6p->in6p_head, inph_hash);
1291 
1292 	switch (state) {
1293 	case IN6P_BOUND:
1294 		LIST_INSERT_HEAD(IN6PCBHASH_BIND(in6p->in6p_table,
1295 		    &in6p->in6p_laddr, in6p->in6p_lport), &in6p->in6p_head,
1296 		    inph_hash);
1297 		break;
1298 	case IN6P_CONNECTED:
1299 		LIST_INSERT_HEAD(IN6PCBHASH_CONNECT(in6p->in6p_table,
1300 		    &in6p->in6p_faddr, in6p->in6p_fport,
1301 		    &in6p->in6p_laddr, in6p->in6p_lport), &in6p->in6p_head,
1302 		    inph_hash);
1303 		break;
1304 	}
1305 
1306 	in6p->in6p_state = state;
1307 }
1308