xref: /netbsd-src/sys/netinet6/in6_pcb.c (revision 4391d5e9d4f291db41e3b3ba26a01b5e51364aae)
1 /*	$NetBSD: in6_pcb.c,v 1.123 2013/06/05 19:01:26 christos 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 1.123 2013/06/05 19:01:26 christos Exp $");
66 
67 #include "opt_inet.h"
68 #include "opt_ipsec.h"
69 
70 #include <sys/param.h>
71 #include <sys/systm.h>
72 #include <sys/malloc.h>
73 #include <sys/mbuf.h>
74 #include <sys/protosw.h>
75 #include <sys/socket.h>
76 #include <sys/socketvar.h>
77 #include <sys/ioctl.h>
78 #include <sys/errno.h>
79 #include <sys/time.h>
80 #include <sys/proc.h>
81 #include <sys/kauth.h>
82 #include <sys/domain.h>
83 #include <sys/once.h>
84 
85 #include <net/if.h>
86 #include <net/route.h>
87 
88 #include <netinet/in.h>
89 #include <netinet/in_var.h>
90 #include <netinet/in_systm.h>
91 #include <netinet/ip.h>
92 #include <netinet/in_pcb.h>
93 #include <netinet/ip6.h>
94 #include <netinet/portalgo.h>
95 #include <netinet6/ip6_var.h>
96 #include <netinet6/in6_pcb.h>
97 #include <netinet6/scope6_var.h>
98 #include <netinet6/nd6.h>
99 
100 #include "faith.h"
101 
102 #ifdef IPSEC
103 #include <netipsec/ipsec.h>
104 #include <netipsec/ipsec6.h>
105 #include <netipsec/key.h>
106 #endif /* IPSEC */
107 
108 #include <netinet/tcp_vtw.h>
109 
110 const struct in6_addr zeroin6_addr;
111 
112 #define	IN6PCBHASH_PORT(table, lport) \
113 	&(table)->inpt_porthashtbl[ntohs(lport) & (table)->inpt_porthash]
114 #define IN6PCBHASH_BIND(table, laddr, lport) \
115 	&(table)->inpt_bindhashtbl[ \
116 	    (((laddr)->s6_addr32[0] ^ (laddr)->s6_addr32[1] ^ \
117 	      (laddr)->s6_addr32[2] ^ (laddr)->s6_addr32[3]) + ntohs(lport)) & \
118 	    (table)->inpt_bindhash]
119 #define IN6PCBHASH_CONNECT(table, faddr, fport, laddr, lport) \
120 	&(table)->inpt_bindhashtbl[ \
121 	    ((((faddr)->s6_addr32[0] ^ (faddr)->s6_addr32[1] ^ \
122 	      (faddr)->s6_addr32[2] ^ (faddr)->s6_addr32[3]) + ntohs(fport)) + \
123 	     (((laddr)->s6_addr32[0] ^ (laddr)->s6_addr32[1] ^ \
124 	      (laddr)->s6_addr32[2] ^ (laddr)->s6_addr32[3]) + \
125 	      ntohs(lport))) & (table)->inpt_bindhash]
126 
127 int ip6_anonportmin = IPV6PORT_ANONMIN;
128 int ip6_anonportmax = IPV6PORT_ANONMAX;
129 int ip6_lowportmin  = IPV6PORT_RESERVEDMIN;
130 int ip6_lowportmax  = IPV6PORT_RESERVEDMAX;
131 
132 static struct pool in6pcb_pool;
133 
134 static int
135 in6pcb_poolinit(void)
136 {
137 
138 	pool_init(&in6pcb_pool, sizeof(struct in6pcb), 0, 0, 0, "in6pcbpl",
139 	    NULL, IPL_SOFTNET);
140 	return 0;
141 }
142 
143 void
144 in6_pcbinit(struct inpcbtable *table, int bindhashsize, int connecthashsize)
145 {
146 	static ONCE_DECL(control);
147 
148 	in_pcbinit(table, bindhashsize, connecthashsize);
149 	table->inpt_lastport = (u_int16_t)ip6_anonportmax;
150 
151 	RUN_ONCE(&control, in6pcb_poolinit);
152 }
153 
154 int
155 in6_pcballoc(struct socket *so, void *v)
156 {
157 	struct inpcbtable *table = v;
158 	struct in6pcb *in6p;
159 	int s;
160 #if defined(IPSEC)
161 	int error;
162 #endif
163 
164 	s = splnet();
165 	in6p = pool_get(&in6pcb_pool, PR_NOWAIT);
166 	splx(s);
167 	if (in6p == NULL)
168 		return (ENOBUFS);
169 	memset((void *)in6p, 0, sizeof(*in6p));
170 	in6p->in6p_af = AF_INET6;
171 	in6p->in6p_table = table;
172 	in6p->in6p_socket = so;
173 	in6p->in6p_hops = -1;	/* use kernel default */
174 	in6p->in6p_icmp6filt = NULL;
175 	in6p->in6p_portalgo = PORTALGO_DEFAULT;
176 	in6p->in6p_bindportonsend = false;
177 #if defined(IPSEC)
178 	error = ipsec_init_pcbpolicy(so, &in6p->in6p_sp);
179 	if (error != 0) {
180 		s = splnet();
181 		pool_put(&in6pcb_pool, in6p);
182 		splx(s);
183 		return error;
184 	}
185 #endif /* IPSEC */
186 	s = splnet();
187 	CIRCLEQ_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 
207 	/*
208 	 * We should check the family, but old programs
209 	 * incorrectly fail to intialize it.
210 	 */
211 	if (sin6->sin6_family != AF_INET6)
212 		return (EAFNOSUPPORT);
213 
214 #ifndef INET
215 	if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr))
216 		return (EADDRNOTAVAIL);
217 #endif
218 
219 	if ((error = sa6_embedscope(sin6, ip6_use_defzone)) != 0)
220 		return (error);
221 
222 	if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
223 		if ((in6p->in6p_flags & IN6P_IPV6_V6ONLY) != 0)
224 			return (EINVAL);
225 		if (sin6->sin6_addr.s6_addr32[3]) {
226 			struct sockaddr_in sin;
227 
228 			memset(&sin, 0, sizeof(sin));
229 			sin.sin_len = sizeof(sin);
230 			sin.sin_family = AF_INET;
231 			bcopy(&sin6->sin6_addr.s6_addr32[3],
232 			    &sin.sin_addr, sizeof(sin.sin_addr));
233 			if (ifa_ifwithaddr((struct sockaddr *)&sin) == 0)
234 				return EADDRNOTAVAIL;
235 		}
236 	} else if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
237 		struct ifaddr *ia = NULL;
238 
239 		if ((in6p->in6p_flags & IN6P_FAITH) == 0 &&
240 		    (ia = ifa_ifwithaddr((struct sockaddr *)sin6)) == 0)
241 			return (EADDRNOTAVAIL);
242 
243 		/*
244 		 * bind to an anycast address might accidentally
245 		 * cause sending a packet with an anycast source
246 		 * address, so we forbid it.
247 		 *
248 		 * We should allow to bind to a deprecated address,
249 		 * since the application dare to use it.
250 		 * But, can we assume that they are careful enough
251 		 * to check if the address is deprecated or not?
252 		 * Maybe, as a safeguard, we should have a setsockopt
253 		 * flag to control the bind(2) behavior against
254 		 * deprecated addresses (default: forbid bind(2)).
255 		 */
256 		if (ia &&
257 		    ((struct in6_ifaddr *)ia)->ia6_flags &
258 		    (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY|IN6_IFF_DETACHED))
259 			return (EADDRNOTAVAIL);
260 	}
261 
262 
263 	in6p->in6p_laddr = sin6->sin6_addr;
264 
265 
266 	return (0);
267 }
268 
269 /*
270  * Bind port from sin6 to in6p.
271  */
272 static int
273 in6_pcbbind_port(struct in6pcb *in6p, struct sockaddr_in6 *sin6, struct lwp *l)
274 {
275 	struct inpcbtable *table = in6p->in6p_table;
276 	struct socket *so = in6p->in6p_socket;
277 	int wild = 0, reuseport = (so->so_options & SO_REUSEPORT);
278 	int error;
279 
280 	if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0 &&
281 	   ((so->so_proto->pr_flags & PR_CONNREQUIRED) == 0 ||
282 	    (so->so_options & SO_ACCEPTCONN) == 0))
283 		wild = 1;
284 
285 	if (sin6->sin6_port != 0) {
286 		enum kauth_network_req req;
287 
288 #ifndef IPNOPRIVPORTS
289 		if (ntohs(sin6->sin6_port) < IPV6PORT_RESERVED)
290 			req = KAUTH_REQ_NETWORK_BIND_PRIVPORT;
291 		else
292 #endif /* IPNOPRIVPORTS */
293 			req = KAUTH_REQ_NETWORK_BIND_PORT;
294 
295 		error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_BIND,
296 		    req, so, sin6, NULL);
297 		if (error)
298 			return (EACCES);
299 	}
300 
301 	if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
302 		/*
303 		 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
304 		 * allow compepte duplication of binding if
305 		 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
306 		 * and a multicast address is bound on both
307 		 * new and duplicated sockets.
308 		 */
309 		if (so->so_options & SO_REUSEADDR)
310 			reuseport = SO_REUSEADDR|SO_REUSEPORT;
311 	}
312 
313 	if (sin6->sin6_port != 0) {
314 		if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
315 #ifdef INET
316 			struct inpcb *t;
317 			struct vestigial_inpcb vestige;
318 
319 			t = in_pcblookup_port(table,
320 			    *(struct in_addr *)&sin6->sin6_addr.s6_addr32[3],
321 			    sin6->sin6_port, wild, &vestige);
322 			if (t && (reuseport & t->inp_socket->so_options) == 0)
323 				return (EADDRINUSE);
324 			if (!t
325 			    && vestige.valid
326 			    && !(reuseport && vestige.reuse_port))
327 			    return EADDRINUSE;
328 #else
329 			return (EADDRNOTAVAIL);
330 #endif
331 		}
332 
333 		{
334 			struct in6pcb *t;
335 			struct vestigial_inpcb vestige;
336 
337 			t = in6_pcblookup_port(table, &sin6->sin6_addr,
338 			    sin6->sin6_port, wild, &vestige);
339 			if (t && (reuseport & t->in6p_socket->so_options) == 0)
340 				return (EADDRINUSE);
341 			if (!t
342 			    && vestige.valid
343 			    && !(reuseport && vestige.reuse_port))
344 			    return EADDRINUSE;
345 		}
346 	}
347 
348 	if (sin6->sin6_port == 0) {
349 		int e;
350 		e = in6_pcbsetport(sin6, in6p, l);
351 		if (e != 0)
352 			return (e);
353 	} else {
354 		in6p->in6p_lport = sin6->sin6_port;
355 		in6_pcbstate(in6p, IN6P_BOUND);
356 	}
357 
358 	LIST_REMOVE(&in6p->in6p_head, inph_lhash);
359 	LIST_INSERT_HEAD(IN6PCBHASH_PORT(table, in6p->in6p_lport),
360 	    &in6p->in6p_head, inph_lhash);
361 
362 	return (0);
363 }
364 
365 int
366 in6_pcbbind(void *v, struct mbuf *nam, struct lwp *l)
367 {
368 	struct in6pcb *in6p = v;
369 	struct sockaddr_in6 lsin6;
370 	struct sockaddr_in6 *sin6 = NULL;
371 	int error;
372 
373 	if (in6p->in6p_af != AF_INET6)
374 		return (EINVAL);
375 
376 	/*
377 	 * If we already have a local port or a local address it means we're
378 	 * bounded.
379 	 */
380 	if (in6p->in6p_lport || !(IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) ||
381 	    (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr) &&
382 	      in6p->in6p_laddr.s6_addr32[3] == 0)))
383 		return (EINVAL);
384 
385 	if (nam != NULL) {
386 		/* We were provided a sockaddr_in6 to use. */
387 		sin6 = mtod(nam, struct sockaddr_in6 *);
388 		if (nam->m_len != sizeof(*sin6))
389 			return (EINVAL);
390 	} else {
391 		/* We always bind to *something*, even if it's "anything". */
392 		lsin6 = *((const struct sockaddr_in6 *)
393 		    in6p->in6p_socket->so_proto->pr_domain->dom_sa_any);
394 		sin6 = &lsin6;
395 	}
396 
397 	/* Bind address. */
398 	error = in6_pcbbind_addr(in6p, sin6, l);
399 	if (error)
400 		return (error);
401 
402 	/* Bind port. */
403 	error = in6_pcbbind_port(in6p, sin6, l);
404 	if (error) {
405 		/*
406 		 * Reset the address here to "any" so we don't "leak" the
407 		 * in6pcb.
408 		 */
409 		in6p->in6p_laddr = in6addr_any;
410 
411 		return (error);
412 	}
413 
414 
415 #if 0
416 	in6p->in6p_flowinfo = 0;	/* XXX */
417 #endif
418 	return (0);
419 }
420 
421 /*
422  * Connect from a socket to a specified address.
423  * Both address and port must be specified in argument sin6.
424  * If don't have a local address for this socket yet,
425  * then pick one.
426  */
427 int
428 in6_pcbconnect(void *v, struct mbuf *nam, struct lwp *l)
429 {
430 	struct rtentry *rt;
431 	struct in6pcb *in6p = v;
432 	struct in6_addr *in6a = NULL;
433 	struct sockaddr_in6 *sin6 = mtod(nam, struct sockaddr_in6 *);
434 	struct ifnet *ifp = NULL;	/* outgoing interface */
435 	int error = 0;
436 	int scope_ambiguous = 0;
437 #ifdef INET
438 	struct in6_addr mapped;
439 #endif
440 	struct sockaddr_in6 tmp;
441 	struct vestigial_inpcb vestige;
442 
443 	(void)&in6a;				/* XXX fool gcc */
444 
445 	if (in6p->in6p_af != AF_INET6)
446 		return (EINVAL);
447 
448 	if (nam->m_len != sizeof(*sin6))
449 		return (EINVAL);
450 	if (sin6->sin6_family != AF_INET6)
451 		return (EAFNOSUPPORT);
452 	if (sin6->sin6_port == 0)
453 		return (EADDRNOTAVAIL);
454 
455 	if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr) &&
456 	    in6p->in6p_socket->so_type == SOCK_STREAM)
457 		return EADDRNOTAVAIL;
458 
459 	if (sin6->sin6_scope_id == 0 && !ip6_use_defzone)
460 		scope_ambiguous = 1;
461 	if ((error = sa6_embedscope(sin6, ip6_use_defzone)) != 0)
462 		return(error);
463 
464 	/* sanity check for mapped address case */
465 	if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
466 		if ((in6p->in6p_flags & IN6P_IPV6_V6ONLY) != 0)
467 			return EINVAL;
468 		if (IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr))
469 			in6p->in6p_laddr.s6_addr16[5] = htons(0xffff);
470 		if (!IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr))
471 			return EINVAL;
472 	} else
473 	{
474 		if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr))
475 			return EINVAL;
476 	}
477 
478 	/* protect *sin6 from overwrites */
479 	tmp = *sin6;
480 	sin6 = &tmp;
481 
482 	/* Source address selection. */
483 	if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr) &&
484 	    in6p->in6p_laddr.s6_addr32[3] == 0) {
485 #ifdef INET
486 		struct sockaddr_in sin, *sinp;
487 
488 		memset(&sin, 0, sizeof(sin));
489 		sin.sin_len = sizeof(sin);
490 		sin.sin_family = AF_INET;
491 		memcpy(&sin.sin_addr, &sin6->sin6_addr.s6_addr32[3],
492 			sizeof(sin.sin_addr));
493 		sinp = in_selectsrc(&sin, &in6p->in6p_route,
494 			in6p->in6p_socket->so_options, NULL, &error);
495 		if (sinp == 0) {
496 			if (error == 0)
497 				error = EADDRNOTAVAIL;
498 			return (error);
499 		}
500 		memset(&mapped, 0, sizeof(mapped));
501 		mapped.s6_addr16[5] = htons(0xffff);
502 		memcpy(&mapped.s6_addr32[3], &sinp->sin_addr, sizeof(sinp->sin_addr));
503 		in6a = &mapped;
504 #else
505 		return EADDRNOTAVAIL;
506 #endif
507 	} else {
508 		/*
509 		 * XXX: in6_selectsrc might replace the bound local address
510 		 * with the address specified by setsockopt(IPV6_PKTINFO).
511 		 * Is it the intended behavior?
512 		 */
513 		in6a = in6_selectsrc(sin6, in6p->in6p_outputopts,
514 				     in6p->in6p_moptions,
515 				     &in6p->in6p_route,
516 				     &in6p->in6p_laddr, &ifp, &error);
517 		if (ifp && scope_ambiguous &&
518 		    (error = in6_setscope(&sin6->sin6_addr, ifp, NULL)) != 0) {
519 			return(error);
520 		}
521 
522 		if (in6a == 0) {
523 			if (error == 0)
524 				error = EADDRNOTAVAIL;
525 			return (error);
526 		}
527 	}
528 	if (ifp == NULL && (rt = rtcache_validate(&in6p->in6p_route)) != NULL)
529 		ifp = rt->rt_ifp;
530 
531 	in6p->in6p_ip6.ip6_hlim = (u_int8_t)in6_selecthlim(in6p, ifp);
532 
533 	if (in6_pcblookup_connect(in6p->in6p_table, &sin6->sin6_addr,
534 	    sin6->sin6_port,
535 	    IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) ? in6a : &in6p->in6p_laddr,
536 				  in6p->in6p_lport, 0, &vestige)
537 		|| vestige.valid)
538 		return (EADDRINUSE);
539 	if (IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) ||
540 	    (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr) &&
541 	     in6p->in6p_laddr.s6_addr32[3] == 0))
542 	{
543 		if (in6p->in6p_lport == 0) {
544 			error = in6_pcbbind(in6p, NULL, l);
545 			if (error != 0)
546 				return error;
547 		}
548 		in6p->in6p_laddr = *in6a;
549 	}
550 	in6p->in6p_faddr = sin6->sin6_addr;
551 	in6p->in6p_fport = sin6->sin6_port;
552 
553         /* Late bind, if needed */
554 	if (in6p->in6p_bindportonsend) {
555                struct sockaddr_in6 lsin = *((const struct sockaddr_in6 *)
556 		    in6p->in6p_socket->so_proto->pr_domain->dom_sa_any);
557 		lsin.sin6_addr = in6p->in6p_laddr;
558 		lsin.sin6_port = 0;
559 
560                if ((error = in6_pcbbind_port(in6p, &lsin, l)) != 0)
561                        return error;
562 	}
563 
564 	in6_pcbstate(in6p, IN6P_CONNECTED);
565 	in6p->in6p_flowinfo &= ~IPV6_FLOWLABEL_MASK;
566 	if (ip6_auto_flowlabel)
567 		in6p->in6p_flowinfo |=
568 		    (htonl(ip6_randomflowlabel()) & IPV6_FLOWLABEL_MASK);
569 #if defined(IPSEC)
570 	if (in6p->in6p_socket->so_type == SOCK_STREAM)
571 		ipsec_pcbconn(in6p->in6p_sp);
572 #endif
573 	return (0);
574 }
575 
576 void
577 in6_pcbdisconnect(struct in6pcb *in6p)
578 {
579 	memset((void *)&in6p->in6p_faddr, 0, sizeof(in6p->in6p_faddr));
580 	in6p->in6p_fport = 0;
581 	in6_pcbstate(in6p, IN6P_BOUND);
582 	in6p->in6p_flowinfo &= ~IPV6_FLOWLABEL_MASK;
583 #if defined(IPSEC)
584 	ipsec_pcbdisconn(in6p->in6p_sp);
585 #endif
586 	if (in6p->in6p_socket->so_state & SS_NOFDREF)
587 		in6_pcbdetach(in6p);
588 }
589 
590 void
591 in6_pcbdetach(struct in6pcb *in6p)
592 {
593 	struct socket *so = in6p->in6p_socket;
594 	int s;
595 
596 	if (in6p->in6p_af != AF_INET6)
597 		return;
598 
599 #if defined(IPSEC)
600 	ipsec6_delete_pcbpolicy(in6p);
601 #endif /* IPSEC */
602 	so->so_pcb = 0;
603 	if (in6p->in6p_options)
604 		m_freem(in6p->in6p_options);
605 	if (in6p->in6p_outputopts != NULL) {
606 		ip6_clearpktopts(in6p->in6p_outputopts, -1);
607 		free(in6p->in6p_outputopts, M_IP6OPT);
608 	}
609 	rtcache_free(&in6p->in6p_route);
610 	ip6_freemoptions(in6p->in6p_moptions);
611 	s = splnet();
612 	in6_pcbstate(in6p, IN6P_ATTACHED);
613 	LIST_REMOVE(&in6p->in6p_head, inph_lhash);
614 	CIRCLEQ_REMOVE(&in6p->in6p_table->inpt_queue, &in6p->in6p_head,
615 	    inph_queue);
616 	pool_put(&in6pcb_pool, in6p);
617 	splx(s);
618 	sofree(so);				/* drops the socket's lock */
619 	mutex_enter(softnet_lock);		/* reacquire it */
620 }
621 
622 void
623 in6_setsockaddr(struct in6pcb *in6p, struct mbuf *nam)
624 {
625 	struct sockaddr_in6 *sin6;
626 
627 	if (in6p->in6p_af != AF_INET6)
628 		return;
629 
630 	nam->m_len = sizeof(*sin6);
631 	sin6 = mtod(nam, struct sockaddr_in6 *);
632 	sockaddr_in6_init(sin6, &in6p->in6p_laddr, in6p->in6p_lport, 0, 0);
633 	(void)sa6_recoverscope(sin6); /* XXX: should catch errors */
634 }
635 
636 void
637 in6_setpeeraddr(struct in6pcb *in6p, struct mbuf *nam)
638 {
639 	struct sockaddr_in6 *sin6;
640 
641 	if (in6p->in6p_af != AF_INET6)
642 		return;
643 
644 	nam->m_len = sizeof(*sin6);
645 	sin6 = mtod(nam, struct sockaddr_in6 *);
646 	sockaddr_in6_init(sin6, &in6p->in6p_faddr, in6p->in6p_fport, 0, 0);
647 	(void)sa6_recoverscope(sin6); /* XXX: should catch errors */
648 }
649 
650 /*
651  * Pass some notification to all connections of a protocol
652  * associated with address dst.  The local address and/or port numbers
653  * may be specified to limit the search.  The "usual action" will be
654  * taken, depending on the ctlinput cmd.  The caller must filter any
655  * cmds that are uninteresting (e.g., no error in the map).
656  * Call the protocol specific routine (if any) to report
657  * any errors for each matching socket.
658  *
659  * Must be called at splsoftnet.
660  *
661  * Note: src (4th arg) carries the flowlabel value on the original IPv6
662  * header, in sin6_flowinfo member.
663  */
664 int
665 in6_pcbnotify(struct inpcbtable *table, const struct sockaddr *dst,
666     u_int fport_arg, const struct sockaddr *src, u_int lport_arg, int cmd,
667     void *cmdarg, void (*notify)(struct in6pcb *, int))
668 {
669 	struct rtentry *rt;
670 	struct in6pcb *in6p, *nin6p;
671 	struct sockaddr_in6 sa6_src;
672 	const struct sockaddr_in6 *sa6_dst;
673 	u_int16_t fport = fport_arg, lport = lport_arg;
674 	int errno;
675 	int nmatch = 0;
676 	u_int32_t flowinfo;
677 
678 	if ((unsigned)cmd >= PRC_NCMDS || dst->sa_family != AF_INET6)
679 		return 0;
680 
681 	sa6_dst = (const struct sockaddr_in6 *)dst;
682 	if (IN6_IS_ADDR_UNSPECIFIED(&sa6_dst->sin6_addr))
683 		return 0;
684 
685 	/*
686 	 * note that src can be NULL when we get notify by local fragmentation.
687 	 */
688 	sa6_src = (src == NULL) ? sa6_any : *(const struct sockaddr_in6 *)src;
689 	flowinfo = sa6_src.sin6_flowinfo;
690 
691 	/*
692 	 * Redirects go to all references to the destination,
693 	 * and use in6_rtchange to invalidate the route cache.
694 	 * Dead host indications: also use in6_rtchange to invalidate
695 	 * the cache, and deliver the error to all the sockets.
696 	 * Otherwise, if we have knowledge of the local port and address,
697 	 * deliver only to that socket.
698 	 */
699 	if (PRC_IS_REDIRECT(cmd) || cmd == PRC_HOSTDEAD) {
700 		fport = 0;
701 		lport = 0;
702 		memset((void *)&sa6_src.sin6_addr, 0, sizeof(sa6_src.sin6_addr));
703 
704 		if (cmd != PRC_HOSTDEAD)
705 			notify = in6_rtchange;
706 	}
707 
708 	errno = inet6ctlerrmap[cmd];
709 	for (in6p = (struct in6pcb *)CIRCLEQ_FIRST(&table->inpt_queue);
710 	    in6p != (void *)&table->inpt_queue;
711 	    in6p = nin6p) {
712 		nin6p = (struct in6pcb *)CIRCLEQ_NEXT(in6p, in6p_queue);
713 
714 		if (in6p->in6p_af != AF_INET6)
715 			continue;
716 
717 		/*
718 		 * Under the following condition, notify of redirects
719 		 * to the pcb, without making address matches against inpcb.
720 		 * - redirect notification is arrived.
721 		 * - the inpcb is unconnected.
722 		 * - the inpcb is caching !RTF_HOST routing entry.
723 		 * - the ICMPv6 notification is from the gateway cached in the
724 		 *   inpcb.  i.e. ICMPv6 notification is from nexthop gateway
725 		 *   the inpcb used very recently.
726 		 *
727 		 * This is to improve interaction between netbsd/openbsd
728 		 * redirect handling code, and inpcb route cache code.
729 		 * without the clause, !RTF_HOST routing entry (which carries
730 		 * gateway used by inpcb right before the ICMPv6 redirect)
731 		 * will be cached forever in unconnected inpcb.
732 		 *
733 		 * There still is a question regarding to what is TRT:
734 		 * - On bsdi/freebsd, RTF_HOST (cloned) routing entry will be
735 		 *   generated on packet output.  inpcb will always cache
736 		 *   RTF_HOST routing entry so there's no need for the clause
737 		 *   (ICMPv6 redirect will update RTF_HOST routing entry,
738 		 *   and inpcb is caching it already).
739 		 *   However, bsdi/freebsd are vulnerable to local DoS attacks
740 		 *   due to the cloned routing entries.
741 		 * - Specwise, "destination cache" is mentioned in RFC2461.
742 		 *   Jinmei says that it implies bsdi/freebsd behavior, itojun
743 		 *   is not really convinced.
744 		 * - Having hiwat/lowat on # of cloned host route (redirect/
745 		 *   pmtud) may be a good idea.  netbsd/openbsd has it.  see
746 		 *   icmp6_mtudisc_update().
747 		 */
748 		if ((PRC_IS_REDIRECT(cmd) || cmd == PRC_HOSTDEAD) &&
749 		    IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
750 		    (rt = rtcache_validate(&in6p->in6p_route)) != NULL &&
751 		    !(rt->rt_flags & RTF_HOST)) {
752 			const struct sockaddr_in6 *dst6;
753 
754 			dst6 = (const struct sockaddr_in6 *)
755 			    rtcache_getdst(&in6p->in6p_route);
756 			if (dst6 == NULL)
757 				;
758 			else if (IN6_ARE_ADDR_EQUAL(&dst6->sin6_addr,
759 			    &sa6_dst->sin6_addr))
760 				goto do_notify;
761 		}
762 
763 		/*
764 		 * If the error designates a new path MTU for a destination
765 		 * and the application (associated with this socket) wanted to
766 		 * know the value, notify. Note that we notify for all
767 		 * disconnected sockets if the corresponding application
768 		 * wanted. This is because some UDP applications keep sending
769 		 * sockets disconnected.
770 		 * XXX: should we avoid to notify the value to TCP sockets?
771 		 */
772 		if (cmd == PRC_MSGSIZE && (in6p->in6p_flags & IN6P_MTU) != 0 &&
773 		    (IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) ||
774 		     IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &sa6_dst->sin6_addr))) {
775 			ip6_notify_pmtu(in6p, (const struct sockaddr_in6 *)dst,
776 					(u_int32_t *)cmdarg);
777 		}
778 
779 		/*
780 		 * Detect if we should notify the error. If no source and
781 		 * destination ports are specified, but non-zero flowinfo and
782 		 * local address match, notify the error. This is the case
783 		 * when the error is delivered with an encrypted buffer
784 		 * by ESP. Otherwise, just compare addresses and ports
785 		 * as usual.
786 		 */
787 		if (lport == 0 && fport == 0 && flowinfo &&
788 		    in6p->in6p_socket != NULL &&
789 		    flowinfo == (in6p->in6p_flowinfo & IPV6_FLOWLABEL_MASK) &&
790 		    IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &sa6_src.sin6_addr))
791 			goto do_notify;
792 		else if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr,
793 					     &sa6_dst->sin6_addr) ||
794 		    in6p->in6p_socket == 0 ||
795 		    (lport && in6p->in6p_lport != lport) ||
796 		    (!IN6_IS_ADDR_UNSPECIFIED(&sa6_src.sin6_addr) &&
797 		     !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr,
798 					 &sa6_src.sin6_addr)) ||
799 		    (fport && in6p->in6p_fport != fport))
800 			continue;
801 
802 	  do_notify:
803 		if (notify)
804 			(*notify)(in6p, errno);
805 		nmatch++;
806 	}
807 	return nmatch;
808 }
809 
810 void
811 in6_pcbpurgeif0(struct inpcbtable *table, struct ifnet *ifp)
812 {
813 	struct in6pcb *in6p, *nin6p;
814 	struct ip6_moptions *im6o;
815 	struct in6_multi_mship *imm, *nimm;
816 
817 	for (in6p = (struct in6pcb *)CIRCLEQ_FIRST(&table->inpt_queue);
818 	    in6p != (void *)&table->inpt_queue;
819 	    in6p = nin6p) {
820 		nin6p = (struct in6pcb *)CIRCLEQ_NEXT(in6p, in6p_queue);
821 		if (in6p->in6p_af != AF_INET6)
822 			continue;
823 
824 		im6o = in6p->in6p_moptions;
825 		if (im6o) {
826 			/*
827 			 * Unselect the outgoing interface if it is being
828 			 * detached.
829 			 */
830 			if (im6o->im6o_multicast_ifp == ifp)
831 				im6o->im6o_multicast_ifp = NULL;
832 
833 			/*
834 			 * Drop multicast group membership if we joined
835 			 * through the interface being detached.
836 			 * XXX controversial - is it really legal for kernel
837 			 * to force this?
838 			 */
839 			for (imm = im6o->im6o_memberships.lh_first;
840 			     imm != NULL; imm = nimm) {
841 				nimm = imm->i6mm_chain.le_next;
842 				if (imm->i6mm_maddr->in6m_ifp == ifp) {
843 					LIST_REMOVE(imm, i6mm_chain);
844 					in6_leavegroup(imm);
845 				}
846 			}
847 		}
848 	}
849 }
850 
851 void
852 in6_pcbpurgeif(struct inpcbtable *table, struct ifnet *ifp)
853 {
854 	struct rtentry *rt;
855 	struct in6pcb *in6p, *nin6p;
856 
857 	for (in6p = (struct in6pcb *)CIRCLEQ_FIRST(&table->inpt_queue);
858 	    in6p != (void *)&table->inpt_queue;
859 	    in6p = nin6p) {
860 		nin6p = (struct in6pcb *)CIRCLEQ_NEXT(in6p, in6p_queue);
861 		if (in6p->in6p_af != AF_INET6)
862 			continue;
863 		if ((rt = rtcache_validate(&in6p->in6p_route)) != NULL &&
864 		    rt->rt_ifp == ifp)
865 			in6_rtchange(in6p, 0);
866 	}
867 }
868 
869 /*
870  * Check for alternatives when higher level complains
871  * about service problems.  For now, invalidate cached
872  * routing information.  If the route was created dynamically
873  * (by a redirect), time to try a default gateway again.
874  */
875 void
876 in6_losing(struct in6pcb *in6p)
877 {
878 	struct rtentry *rt;
879 	struct rt_addrinfo info;
880 
881 	if (in6p->in6p_af != AF_INET6)
882 		return;
883 
884 	if ((rt = rtcache_validate(&in6p->in6p_route)) == NULL)
885 		return;
886 
887 	memset(&info, 0, sizeof(info));
888 	info.rti_info[RTAX_DST] = rtcache_getdst(&in6p->in6p_route);
889 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
890 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
891 	rt_missmsg(RTM_LOSING, &info, rt->rt_flags, 0);
892 	if (rt->rt_flags & RTF_DYNAMIC) {
893 		(void)rtrequest(RTM_DELETE, rt_getkey(rt),
894 		    rt->rt_gateway, rt_mask(rt), rt->rt_flags, NULL);
895 	}
896 	/*
897 	 * A new route can be allocated
898 	 * the next time output is attempted.
899 	 */
900 	rtcache_free(&in6p->in6p_route);
901 }
902 
903 /*
904  * After a routing change, flush old routing.  A new route can be
905  * allocated the next time output is attempted.
906  */
907 void
908 in6_rtchange(struct in6pcb *in6p, int errno)
909 {
910 	if (in6p->in6p_af != AF_INET6)
911 		return;
912 
913 	rtcache_free(&in6p->in6p_route);
914 	/*
915 	 * A new route can be allocated the next time
916 	 * output is attempted.
917 	 */
918 }
919 
920 struct in6pcb *
921 in6_pcblookup_port(struct inpcbtable *table, struct in6_addr *laddr6,
922 		   u_int lport_arg, int lookup_wildcard, struct vestigial_inpcb *vp)
923 {
924 	struct inpcbhead *head;
925 	struct inpcb_hdr *inph;
926 	struct in6pcb *in6p, *match = 0;
927 	int matchwild = 3, wildcard;
928 	u_int16_t lport = lport_arg;
929 
930 	if (vp)
931 		vp->valid = 0;
932 
933 	head = IN6PCBHASH_PORT(table, lport);
934 	LIST_FOREACH(inph, head, inph_lhash) {
935 		in6p = (struct in6pcb *)inph;
936 		if (in6p->in6p_af != AF_INET6)
937 			continue;
938 
939 		if (in6p->in6p_lport != lport)
940 			continue;
941 		wildcard = 0;
942 		if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_faddr)) {
943 			if ((in6p->in6p_flags & IN6P_IPV6_V6ONLY) != 0)
944 				continue;
945 		}
946 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr))
947 			wildcard++;
948 		if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr)) {
949 			if ((in6p->in6p_flags & IN6P_IPV6_V6ONLY) != 0)
950 				continue;
951 			if (!IN6_IS_ADDR_V4MAPPED(laddr6))
952 				continue;
953 
954 			/* duplicate of IPv4 logic */
955 			wildcard = 0;
956 			if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_faddr) &&
957 			    in6p->in6p_faddr.s6_addr32[3])
958 				wildcard++;
959 			if (!in6p->in6p_laddr.s6_addr32[3]) {
960 				if (laddr6->s6_addr32[3])
961 					wildcard++;
962 			} else {
963 				if (!laddr6->s6_addr32[3])
964 					wildcard++;
965 				else {
966 					if (in6p->in6p_laddr.s6_addr32[3] !=
967 					    laddr6->s6_addr32[3])
968 						continue;
969 				}
970 			}
971 		} else if (IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
972 			if (IN6_IS_ADDR_V4MAPPED(laddr6)) {
973 				if ((in6p->in6p_flags & IN6P_IPV6_V6ONLY) != 0)
974 					continue;
975 			}
976 			if (!IN6_IS_ADDR_UNSPECIFIED(laddr6))
977 				wildcard++;
978 		} else {
979 			if (IN6_IS_ADDR_V4MAPPED(laddr6)) {
980 				if ((in6p->in6p_flags & IN6P_IPV6_V6ONLY) != 0)
981 					continue;
982 			}
983 			if (IN6_IS_ADDR_UNSPECIFIED(laddr6))
984 				wildcard++;
985 			else {
986 				if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr,
987 				    laddr6))
988 					continue;
989 			}
990 		}
991 		if (wildcard && !lookup_wildcard)
992 			continue;
993 		if (wildcard < matchwild) {
994 			match = in6p;
995 			matchwild = wildcard;
996 			if (matchwild == 0)
997 				break;
998 		}
999 	}
1000 	if (match && matchwild == 0)
1001 		return match;
1002 
1003 	if (vp && table->vestige && table->vestige->init_ports6) {
1004 		struct vestigial_inpcb better;
1005 		void *state;
1006 
1007 		state = (*table->vestige->init_ports6)(laddr6,
1008 						       lport_arg,
1009 						       lookup_wildcard);
1010 		while (table->vestige
1011 		       && (*table->vestige->next_port6)(state, vp)) {
1012 
1013 			if (vp->lport != lport)
1014 				continue;
1015 			wildcard = 0;
1016 			if (!IN6_IS_ADDR_UNSPECIFIED(&vp->faddr.v6))
1017 				wildcard++;
1018 			if (IN6_IS_ADDR_UNSPECIFIED(&vp->laddr.v6)) {
1019 				if (!IN6_IS_ADDR_UNSPECIFIED(laddr6))
1020 					wildcard++;
1021 			} else {
1022 				if (IN6_IS_ADDR_V4MAPPED(laddr6)) {
1023 					if (vp->v6only)
1024 						continue;
1025 				}
1026 				if (IN6_IS_ADDR_UNSPECIFIED(laddr6))
1027 					wildcard++;
1028 				else {
1029 					if (!IN6_ARE_ADDR_EQUAL(&vp->laddr.v6, laddr6))
1030 						continue;
1031 				}
1032 			}
1033 			if (wildcard && !lookup_wildcard)
1034 				continue;
1035 			if (wildcard < matchwild) {
1036 				better = *vp;
1037 				match  = (void*)&better;
1038 
1039 				matchwild = wildcard;
1040 				if (matchwild == 0)
1041 					break;
1042 			}
1043 		}
1044 
1045 		if (match) {
1046 			if (match != (void*)&better)
1047 				return match;
1048 			else {
1049 				*vp = better;
1050 				return 0;
1051 			}
1052 		}
1053 	}
1054 	return (match);
1055 }
1056 
1057 /*
1058  * WARNING: return value (rtentry) could be IPv4 one if in6pcb is connected to
1059  * IPv4 mapped address.
1060  */
1061 struct rtentry *
1062 in6_pcbrtentry(struct in6pcb *in6p)
1063 {
1064 	struct rtentry *rt;
1065 	struct route *ro;
1066 	union {
1067 		const struct sockaddr *sa;
1068 		const struct sockaddr_in6 *sa6;
1069 #ifdef INET
1070 		const struct sockaddr_in *sa4;
1071 #endif
1072 	} cdst;
1073 
1074 	ro = &in6p->in6p_route;
1075 
1076 	if (in6p->in6p_af != AF_INET6)
1077 		return (NULL);
1078 
1079 	cdst.sa = rtcache_getdst(ro);
1080 	if (cdst.sa == NULL)
1081 		;
1082 #ifdef INET
1083 	else if (cdst.sa->sa_family == AF_INET) {
1084 		KASSERT(IN6_IS_ADDR_V4MAPPED(&in6p->in6p_faddr));
1085 		if (cdst.sa4->sin_addr.s_addr != in6p->in6p_faddr.s6_addr32[3])
1086 			rtcache_free(ro);
1087 	}
1088 #endif
1089 	else {
1090 		if (!IN6_ARE_ADDR_EQUAL(&cdst.sa6->sin6_addr,
1091 					&in6p->in6p_faddr))
1092 			rtcache_free(ro);
1093 	}
1094 	if ((rt = rtcache_validate(ro)) == NULL)
1095 		rt = rtcache_update(ro, 1);
1096 #ifdef INET
1097 	if (rt == NULL && IN6_IS_ADDR_V4MAPPED(&in6p->in6p_faddr)) {
1098 		union {
1099 			struct sockaddr		dst;
1100 			struct sockaddr_in	dst4;
1101 		} u;
1102 		struct in_addr addr;
1103 
1104 		addr.s_addr = in6p->in6p_faddr.s6_addr32[3];
1105 
1106 		sockaddr_in_init(&u.dst4, &addr, 0);
1107 		rtcache_setdst(ro, &u.dst);
1108 
1109 		rt = rtcache_init(ro);
1110 	} else
1111 #endif
1112 	if (rt == NULL && !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) {
1113 		union {
1114 			struct sockaddr		dst;
1115 			struct sockaddr_in6	dst6;
1116 		} u;
1117 
1118 		sockaddr_in6_init(&u.dst6, &in6p->in6p_faddr, 0, 0, 0);
1119 		rtcache_setdst(ro, &u.dst);
1120 
1121 		rt = rtcache_init(ro);
1122 	}
1123 	return rt;
1124 }
1125 
1126 struct in6pcb *
1127 in6_pcblookup_connect(struct inpcbtable *table, const struct in6_addr *faddr6,
1128 		      u_int fport_arg, const struct in6_addr *laddr6, u_int lport_arg,
1129 		      int faith,
1130 		      struct vestigial_inpcb *vp)
1131 {
1132 	struct inpcbhead *head;
1133 	struct inpcb_hdr *inph;
1134 	struct in6pcb *in6p;
1135 	u_int16_t fport = fport_arg, lport = lport_arg;
1136 
1137 	if (vp)
1138 		vp->valid = 0;
1139 
1140 	head = IN6PCBHASH_CONNECT(table, faddr6, fport, laddr6, lport);
1141 	LIST_FOREACH(inph, head, inph_hash) {
1142 		in6p = (struct in6pcb *)inph;
1143 		if (in6p->in6p_af != AF_INET6)
1144 			continue;
1145 
1146 		/* find exact match on both source and dest */
1147 		if (in6p->in6p_fport != fport)
1148 			continue;
1149 		if (in6p->in6p_lport != lport)
1150 			continue;
1151 		if (IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr))
1152 			continue;
1153 		if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, faddr6))
1154 			continue;
1155 		if (IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr))
1156 			continue;
1157 		if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, laddr6))
1158 			continue;
1159 		if ((IN6_IS_ADDR_V4MAPPED(laddr6) ||
1160 		     IN6_IS_ADDR_V4MAPPED(faddr6)) &&
1161 		    (in6p->in6p_flags & IN6P_IPV6_V6ONLY))
1162 			continue;
1163 		return in6p;
1164 	}
1165 	if (vp && table->vestige) {
1166 		if ((*table->vestige->lookup6)(faddr6, fport_arg,
1167 					       laddr6, lport_arg, vp))
1168 			return 0;
1169 	}
1170 
1171 	return NULL;
1172 }
1173 
1174 struct in6pcb *
1175 in6_pcblookup_bind(struct inpcbtable *table, const struct in6_addr *laddr6,
1176 	u_int lport_arg, int faith)
1177 {
1178 	struct inpcbhead *head;
1179 	struct inpcb_hdr *inph;
1180 	struct in6pcb *in6p;
1181 	u_int16_t lport = lport_arg;
1182 #ifdef INET
1183 	struct in6_addr zero_mapped;
1184 #endif
1185 
1186 	head = IN6PCBHASH_BIND(table, laddr6, lport);
1187 	LIST_FOREACH(inph, head, inph_hash) {
1188 		in6p = (struct in6pcb *)inph;
1189 		if (in6p->in6p_af != AF_INET6)
1190 			continue;
1191 
1192 		if (faith && (in6p->in6p_flags & IN6P_FAITH) == 0)
1193 			continue;
1194 		if (in6p->in6p_fport != 0)
1195 			continue;
1196 		if (in6p->in6p_lport != lport)
1197 			continue;
1198 		if (IN6_IS_ADDR_V4MAPPED(laddr6) &&
1199 		    (in6p->in6p_flags & IN6P_IPV6_V6ONLY) != 0)
1200 			continue;
1201 		if (IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, laddr6))
1202 			goto out;
1203 	}
1204 #ifdef INET
1205 	if (IN6_IS_ADDR_V4MAPPED(laddr6)) {
1206 		memset(&zero_mapped, 0, sizeof(zero_mapped));
1207 		zero_mapped.s6_addr16[5] = 0xffff;
1208 		head = IN6PCBHASH_BIND(table, &zero_mapped, lport);
1209 		LIST_FOREACH(inph, head, inph_hash) {
1210 			in6p = (struct in6pcb *)inph;
1211 			if (in6p->in6p_af != AF_INET6)
1212 				continue;
1213 
1214 			if (faith && (in6p->in6p_flags & IN6P_FAITH) == 0)
1215 				continue;
1216 			if (in6p->in6p_fport != 0)
1217 				continue;
1218 			if (in6p->in6p_lport != lport)
1219 				continue;
1220 			if ((in6p->in6p_flags & IN6P_IPV6_V6ONLY) != 0)
1221 				continue;
1222 			if (IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &zero_mapped))
1223 				goto out;
1224 		}
1225 	}
1226 #endif
1227 	head = IN6PCBHASH_BIND(table, &zeroin6_addr, lport);
1228 	LIST_FOREACH(inph, head, inph_hash) {
1229 		in6p = (struct in6pcb *)inph;
1230 		if (in6p->in6p_af != AF_INET6)
1231 			continue;
1232 
1233 		if (faith && (in6p->in6p_flags & IN6P_FAITH) == 0)
1234 			continue;
1235 		if (in6p->in6p_fport != 0)
1236 			continue;
1237 		if (in6p->in6p_lport != lport)
1238 			continue;
1239 		if (IN6_IS_ADDR_V4MAPPED(laddr6) &&
1240 		    (in6p->in6p_flags & IN6P_IPV6_V6ONLY) != 0)
1241 			continue;
1242 		if (IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &zeroin6_addr))
1243 			goto out;
1244 	}
1245 	return (NULL);
1246 
1247 out:
1248 	inph = &in6p->in6p_head;
1249 	if (inph != LIST_FIRST(head)) {
1250 		LIST_REMOVE(inph, inph_hash);
1251 		LIST_INSERT_HEAD(head, inph, inph_hash);
1252 	}
1253 	return in6p;
1254 }
1255 
1256 void
1257 in6_pcbstate(struct in6pcb *in6p, int state)
1258 {
1259 
1260 	if (in6p->in6p_af != AF_INET6)
1261 		return;
1262 
1263 	if (in6p->in6p_state > IN6P_ATTACHED)
1264 		LIST_REMOVE(&in6p->in6p_head, inph_hash);
1265 
1266 	switch (state) {
1267 	case IN6P_BOUND:
1268 		LIST_INSERT_HEAD(IN6PCBHASH_BIND(in6p->in6p_table,
1269 		    &in6p->in6p_laddr, in6p->in6p_lport), &in6p->in6p_head,
1270 		    inph_hash);
1271 		break;
1272 	case IN6P_CONNECTED:
1273 		LIST_INSERT_HEAD(IN6PCBHASH_CONNECT(in6p->in6p_table,
1274 		    &in6p->in6p_faddr, in6p->in6p_fport,
1275 		    &in6p->in6p_laddr, in6p->in6p_lport), &in6p->in6p_head,
1276 		    inph_hash);
1277 		break;
1278 	}
1279 
1280 	in6p->in6p_state = state;
1281 }
1282