xref: /netbsd-src/sys/netinet6/in6_pcb.c (revision 5bbd2a12505d72a8177929a37b5cee489d0a1cfd)
1 /*	$NetBSD: in6_pcb.c,v 1.121 2012/08/24 06:03:18 dholland 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 1.121 2012/08/24 06:03:18 dholland 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 FAST_IPSEC
103 #include <netipsec/ipsec.h>
104 #include <netipsec/ipsec6.h>
105 #include <netipsec/key.h>
106 #endif /* FAST_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(FAST_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(FAST_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 (sin6->sin6_scope_id == 0 && !ip6_use_defzone)
456 		scope_ambiguous = 1;
457 	if ((error = sa6_embedscope(sin6, ip6_use_defzone)) != 0)
458 		return(error);
459 
460 	/* sanity check for mapped address case */
461 	if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
462 		if ((in6p->in6p_flags & IN6P_IPV6_V6ONLY) != 0)
463 			return EINVAL;
464 		if (IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr))
465 			in6p->in6p_laddr.s6_addr16[5] = htons(0xffff);
466 		if (!IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr))
467 			return EINVAL;
468 	} else
469 	{
470 		if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr))
471 			return EINVAL;
472 	}
473 
474 	/* protect *sin6 from overwrites */
475 	tmp = *sin6;
476 	sin6 = &tmp;
477 
478 	/* Source address selection. */
479 	if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr) &&
480 	    in6p->in6p_laddr.s6_addr32[3] == 0) {
481 #ifdef INET
482 		struct sockaddr_in sin, *sinp;
483 
484 		memset(&sin, 0, sizeof(sin));
485 		sin.sin_len = sizeof(sin);
486 		sin.sin_family = AF_INET;
487 		memcpy(&sin.sin_addr, &sin6->sin6_addr.s6_addr32[3],
488 			sizeof(sin.sin_addr));
489 		sinp = in_selectsrc(&sin, &in6p->in6p_route,
490 			in6p->in6p_socket->so_options, NULL, &error);
491 		if (sinp == 0) {
492 			if (error == 0)
493 				error = EADDRNOTAVAIL;
494 			return (error);
495 		}
496 		memset(&mapped, 0, sizeof(mapped));
497 		mapped.s6_addr16[5] = htons(0xffff);
498 		memcpy(&mapped.s6_addr32[3], &sinp->sin_addr, sizeof(sinp->sin_addr));
499 		in6a = &mapped;
500 #else
501 		return EADDRNOTAVAIL;
502 #endif
503 	} else {
504 		/*
505 		 * XXX: in6_selectsrc might replace the bound local address
506 		 * with the address specified by setsockopt(IPV6_PKTINFO).
507 		 * Is it the intended behavior?
508 		 */
509 		in6a = in6_selectsrc(sin6, in6p->in6p_outputopts,
510 				     in6p->in6p_moptions,
511 				     &in6p->in6p_route,
512 				     &in6p->in6p_laddr, &ifp, &error);
513 		if (ifp && scope_ambiguous &&
514 		    (error = in6_setscope(&sin6->sin6_addr, ifp, NULL)) != 0) {
515 			return(error);
516 		}
517 
518 		if (in6a == 0) {
519 			if (error == 0)
520 				error = EADDRNOTAVAIL;
521 			return (error);
522 		}
523 	}
524 	if (ifp == NULL && (rt = rtcache_validate(&in6p->in6p_route)) != NULL)
525 		ifp = rt->rt_ifp;
526 
527 	in6p->in6p_ip6.ip6_hlim = (u_int8_t)in6_selecthlim(in6p, ifp);
528 
529 	if (in6_pcblookup_connect(in6p->in6p_table, &sin6->sin6_addr,
530 	    sin6->sin6_port,
531 	    IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) ? in6a : &in6p->in6p_laddr,
532 				  in6p->in6p_lport, 0, &vestige)
533 		|| vestige.valid)
534 		return (EADDRINUSE);
535 	if (IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) ||
536 	    (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr) &&
537 	     in6p->in6p_laddr.s6_addr32[3] == 0))
538 	{
539 		if (in6p->in6p_lport == 0) {
540 			error = in6_pcbbind(in6p, NULL, l);
541 			if (error != 0)
542 				return error;
543 		}
544 		in6p->in6p_laddr = *in6a;
545 	}
546 	in6p->in6p_faddr = sin6->sin6_addr;
547 	in6p->in6p_fport = sin6->sin6_port;
548 
549         /* Late bind, if needed */
550 	if (in6p->in6p_bindportonsend) {
551                struct sockaddr_in6 lsin = *((const struct sockaddr_in6 *)
552 		    in6p->in6p_socket->so_proto->pr_domain->dom_sa_any);
553 		lsin.sin6_addr = in6p->in6p_laddr;
554 		lsin.sin6_port = 0;
555 
556                if ((error = in6_pcbbind_port(in6p, &lsin, l)) != 0)
557                        return error;
558 	}
559 
560 	in6_pcbstate(in6p, IN6P_CONNECTED);
561 	in6p->in6p_flowinfo &= ~IPV6_FLOWLABEL_MASK;
562 	if (ip6_auto_flowlabel)
563 		in6p->in6p_flowinfo |=
564 		    (htonl(ip6_randomflowlabel()) & IPV6_FLOWLABEL_MASK);
565 #if defined(FAST_IPSEC)
566 	if (in6p->in6p_socket->so_type == SOCK_STREAM)
567 		ipsec_pcbconn(in6p->in6p_sp);
568 #endif
569 	return (0);
570 }
571 
572 void
573 in6_pcbdisconnect(struct in6pcb *in6p)
574 {
575 	memset((void *)&in6p->in6p_faddr, 0, sizeof(in6p->in6p_faddr));
576 	in6p->in6p_fport = 0;
577 	in6_pcbstate(in6p, IN6P_BOUND);
578 	in6p->in6p_flowinfo &= ~IPV6_FLOWLABEL_MASK;
579 #if defined(FAST_IPSEC)
580 	ipsec_pcbdisconn(in6p->in6p_sp);
581 #endif
582 	if (in6p->in6p_socket->so_state & SS_NOFDREF)
583 		in6_pcbdetach(in6p);
584 }
585 
586 void
587 in6_pcbdetach(struct in6pcb *in6p)
588 {
589 	struct socket *so = in6p->in6p_socket;
590 	int s;
591 
592 	if (in6p->in6p_af != AF_INET6)
593 		return;
594 
595 #if defined(FAST_IPSEC)
596 	ipsec6_delete_pcbpolicy(in6p);
597 #endif /* IPSEC */
598 	so->so_pcb = 0;
599 	if (in6p->in6p_options)
600 		m_freem(in6p->in6p_options);
601 	if (in6p->in6p_outputopts != NULL) {
602 		ip6_clearpktopts(in6p->in6p_outputopts, -1);
603 		free(in6p->in6p_outputopts, M_IP6OPT);
604 	}
605 	rtcache_free(&in6p->in6p_route);
606 	ip6_freemoptions(in6p->in6p_moptions);
607 	s = splnet();
608 	in6_pcbstate(in6p, IN6P_ATTACHED);
609 	LIST_REMOVE(&in6p->in6p_head, inph_lhash);
610 	CIRCLEQ_REMOVE(&in6p->in6p_table->inpt_queue, &in6p->in6p_head,
611 	    inph_queue);
612 	pool_put(&in6pcb_pool, in6p);
613 	splx(s);
614 	sofree(so);				/* drops the socket's lock */
615 	mutex_enter(softnet_lock);		/* reacquire it */
616 }
617 
618 void
619 in6_setsockaddr(struct in6pcb *in6p, struct mbuf *nam)
620 {
621 	struct sockaddr_in6 *sin6;
622 
623 	if (in6p->in6p_af != AF_INET6)
624 		return;
625 
626 	nam->m_len = sizeof(*sin6);
627 	sin6 = mtod(nam, struct sockaddr_in6 *);
628 	sockaddr_in6_init(sin6, &in6p->in6p_laddr, in6p->in6p_lport, 0, 0);
629 	(void)sa6_recoverscope(sin6); /* XXX: should catch errors */
630 }
631 
632 void
633 in6_setpeeraddr(struct in6pcb *in6p, struct mbuf *nam)
634 {
635 	struct sockaddr_in6 *sin6;
636 
637 	if (in6p->in6p_af != AF_INET6)
638 		return;
639 
640 	nam->m_len = sizeof(*sin6);
641 	sin6 = mtod(nam, struct sockaddr_in6 *);
642 	sockaddr_in6_init(sin6, &in6p->in6p_faddr, in6p->in6p_fport, 0, 0);
643 	(void)sa6_recoverscope(sin6); /* XXX: should catch errors */
644 }
645 
646 /*
647  * Pass some notification to all connections of a protocol
648  * associated with address dst.  The local address and/or port numbers
649  * may be specified to limit the search.  The "usual action" will be
650  * taken, depending on the ctlinput cmd.  The caller must filter any
651  * cmds that are uninteresting (e.g., no error in the map).
652  * Call the protocol specific routine (if any) to report
653  * any errors for each matching socket.
654  *
655  * Must be called at splsoftnet.
656  *
657  * Note: src (4th arg) carries the flowlabel value on the original IPv6
658  * header, in sin6_flowinfo member.
659  */
660 int
661 in6_pcbnotify(struct inpcbtable *table, const struct sockaddr *dst,
662     u_int fport_arg, const struct sockaddr *src, u_int lport_arg, int cmd,
663     void *cmdarg, void (*notify)(struct in6pcb *, int))
664 {
665 	struct rtentry *rt;
666 	struct in6pcb *in6p, *nin6p;
667 	struct sockaddr_in6 sa6_src;
668 	const struct sockaddr_in6 *sa6_dst;
669 	u_int16_t fport = fport_arg, lport = lport_arg;
670 	int errno;
671 	int nmatch = 0;
672 	u_int32_t flowinfo;
673 
674 	if ((unsigned)cmd >= PRC_NCMDS || dst->sa_family != AF_INET6)
675 		return 0;
676 
677 	sa6_dst = (const struct sockaddr_in6 *)dst;
678 	if (IN6_IS_ADDR_UNSPECIFIED(&sa6_dst->sin6_addr))
679 		return 0;
680 
681 	/*
682 	 * note that src can be NULL when we get notify by local fragmentation.
683 	 */
684 	sa6_src = (src == NULL) ? sa6_any : *(const struct sockaddr_in6 *)src;
685 	flowinfo = sa6_src.sin6_flowinfo;
686 
687 	/*
688 	 * Redirects go to all references to the destination,
689 	 * and use in6_rtchange to invalidate the route cache.
690 	 * Dead host indications: also use in6_rtchange to invalidate
691 	 * the cache, and deliver the error to all the sockets.
692 	 * Otherwise, if we have knowledge of the local port and address,
693 	 * deliver only to that socket.
694 	 */
695 	if (PRC_IS_REDIRECT(cmd) || cmd == PRC_HOSTDEAD) {
696 		fport = 0;
697 		lport = 0;
698 		memset((void *)&sa6_src.sin6_addr, 0, sizeof(sa6_src.sin6_addr));
699 
700 		if (cmd != PRC_HOSTDEAD)
701 			notify = in6_rtchange;
702 	}
703 
704 	errno = inet6ctlerrmap[cmd];
705 	for (in6p = (struct in6pcb *)CIRCLEQ_FIRST(&table->inpt_queue);
706 	    in6p != (void *)&table->inpt_queue;
707 	    in6p = nin6p) {
708 		nin6p = (struct in6pcb *)CIRCLEQ_NEXT(in6p, in6p_queue);
709 
710 		if (in6p->in6p_af != AF_INET6)
711 			continue;
712 
713 		/*
714 		 * Under the following condition, notify of redirects
715 		 * to the pcb, without making address matches against inpcb.
716 		 * - redirect notification is arrived.
717 		 * - the inpcb is unconnected.
718 		 * - the inpcb is caching !RTF_HOST routing entry.
719 		 * - the ICMPv6 notification is from the gateway cached in the
720 		 *   inpcb.  i.e. ICMPv6 notification is from nexthop gateway
721 		 *   the inpcb used very recently.
722 		 *
723 		 * This is to improve interaction between netbsd/openbsd
724 		 * redirect handling code, and inpcb route cache code.
725 		 * without the clause, !RTF_HOST routing entry (which carries
726 		 * gateway used by inpcb right before the ICMPv6 redirect)
727 		 * will be cached forever in unconnected inpcb.
728 		 *
729 		 * There still is a question regarding to what is TRT:
730 		 * - On bsdi/freebsd, RTF_HOST (cloned) routing entry will be
731 		 *   generated on packet output.  inpcb will always cache
732 		 *   RTF_HOST routing entry so there's no need for the clause
733 		 *   (ICMPv6 redirect will update RTF_HOST routing entry,
734 		 *   and inpcb is caching it already).
735 		 *   However, bsdi/freebsd are vulnerable to local DoS attacks
736 		 *   due to the cloned routing entries.
737 		 * - Specwise, "destination cache" is mentioned in RFC2461.
738 		 *   Jinmei says that it implies bsdi/freebsd behavior, itojun
739 		 *   is not really convinced.
740 		 * - Having hiwat/lowat on # of cloned host route (redirect/
741 		 *   pmtud) may be a good idea.  netbsd/openbsd has it.  see
742 		 *   icmp6_mtudisc_update().
743 		 */
744 		if ((PRC_IS_REDIRECT(cmd) || cmd == PRC_HOSTDEAD) &&
745 		    IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
746 		    (rt = rtcache_validate(&in6p->in6p_route)) != NULL &&
747 		    !(rt->rt_flags & RTF_HOST)) {
748 			const struct sockaddr_in6 *dst6;
749 
750 			dst6 = (const struct sockaddr_in6 *)
751 			    rtcache_getdst(&in6p->in6p_route);
752 			if (dst6 == NULL)
753 				;
754 			else if (IN6_ARE_ADDR_EQUAL(&dst6->sin6_addr,
755 			    &sa6_dst->sin6_addr))
756 				goto do_notify;
757 		}
758 
759 		/*
760 		 * If the error designates a new path MTU for a destination
761 		 * and the application (associated with this socket) wanted to
762 		 * know the value, notify. Note that we notify for all
763 		 * disconnected sockets if the corresponding application
764 		 * wanted. This is because some UDP applications keep sending
765 		 * sockets disconnected.
766 		 * XXX: should we avoid to notify the value to TCP sockets?
767 		 */
768 		if (cmd == PRC_MSGSIZE && (in6p->in6p_flags & IN6P_MTU) != 0 &&
769 		    (IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) ||
770 		     IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &sa6_dst->sin6_addr))) {
771 			ip6_notify_pmtu(in6p, (const struct sockaddr_in6 *)dst,
772 					(u_int32_t *)cmdarg);
773 		}
774 
775 		/*
776 		 * Detect if we should notify the error. If no source and
777 		 * destination ports are specified, but non-zero flowinfo and
778 		 * local address match, notify the error. This is the case
779 		 * when the error is delivered with an encrypted buffer
780 		 * by ESP. Otherwise, just compare addresses and ports
781 		 * as usual.
782 		 */
783 		if (lport == 0 && fport == 0 && flowinfo &&
784 		    in6p->in6p_socket != NULL &&
785 		    flowinfo == (in6p->in6p_flowinfo & IPV6_FLOWLABEL_MASK) &&
786 		    IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &sa6_src.sin6_addr))
787 			goto do_notify;
788 		else if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr,
789 					     &sa6_dst->sin6_addr) ||
790 		    in6p->in6p_socket == 0 ||
791 		    (lport && in6p->in6p_lport != lport) ||
792 		    (!IN6_IS_ADDR_UNSPECIFIED(&sa6_src.sin6_addr) &&
793 		     !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr,
794 					 &sa6_src.sin6_addr)) ||
795 		    (fport && in6p->in6p_fport != fport))
796 			continue;
797 
798 	  do_notify:
799 		if (notify)
800 			(*notify)(in6p, errno);
801 		nmatch++;
802 	}
803 	return nmatch;
804 }
805 
806 void
807 in6_pcbpurgeif0(struct inpcbtable *table, struct ifnet *ifp)
808 {
809 	struct in6pcb *in6p, *nin6p;
810 	struct ip6_moptions *im6o;
811 	struct in6_multi_mship *imm, *nimm;
812 
813 	for (in6p = (struct in6pcb *)CIRCLEQ_FIRST(&table->inpt_queue);
814 	    in6p != (void *)&table->inpt_queue;
815 	    in6p = nin6p) {
816 		nin6p = (struct in6pcb *)CIRCLEQ_NEXT(in6p, in6p_queue);
817 		if (in6p->in6p_af != AF_INET6)
818 			continue;
819 
820 		im6o = in6p->in6p_moptions;
821 		if (im6o) {
822 			/*
823 			 * Unselect the outgoing interface if it is being
824 			 * detached.
825 			 */
826 			if (im6o->im6o_multicast_ifp == ifp)
827 				im6o->im6o_multicast_ifp = NULL;
828 
829 			/*
830 			 * Drop multicast group membership if we joined
831 			 * through the interface being detached.
832 			 * XXX controversial - is it really legal for kernel
833 			 * to force this?
834 			 */
835 			for (imm = im6o->im6o_memberships.lh_first;
836 			     imm != NULL; imm = nimm) {
837 				nimm = imm->i6mm_chain.le_next;
838 				if (imm->i6mm_maddr->in6m_ifp == ifp) {
839 					LIST_REMOVE(imm, i6mm_chain);
840 					in6_leavegroup(imm);
841 				}
842 			}
843 		}
844 	}
845 }
846 
847 void
848 in6_pcbpurgeif(struct inpcbtable *table, struct ifnet *ifp)
849 {
850 	struct rtentry *rt;
851 	struct in6pcb *in6p, *nin6p;
852 
853 	for (in6p = (struct in6pcb *)CIRCLEQ_FIRST(&table->inpt_queue);
854 	    in6p != (void *)&table->inpt_queue;
855 	    in6p = nin6p) {
856 		nin6p = (struct in6pcb *)CIRCLEQ_NEXT(in6p, in6p_queue);
857 		if (in6p->in6p_af != AF_INET6)
858 			continue;
859 		if ((rt = rtcache_validate(&in6p->in6p_route)) != NULL &&
860 		    rt->rt_ifp == ifp)
861 			in6_rtchange(in6p, 0);
862 	}
863 }
864 
865 /*
866  * Check for alternatives when higher level complains
867  * about service problems.  For now, invalidate cached
868  * routing information.  If the route was created dynamically
869  * (by a redirect), time to try a default gateway again.
870  */
871 void
872 in6_losing(struct in6pcb *in6p)
873 {
874 	struct rtentry *rt;
875 	struct rt_addrinfo info;
876 
877 	if (in6p->in6p_af != AF_INET6)
878 		return;
879 
880 	if ((rt = rtcache_validate(&in6p->in6p_route)) == NULL)
881 		return;
882 
883 	memset(&info, 0, sizeof(info));
884 	info.rti_info[RTAX_DST] = rtcache_getdst(&in6p->in6p_route);
885 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
886 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
887 	rt_missmsg(RTM_LOSING, &info, rt->rt_flags, 0);
888 	if (rt->rt_flags & RTF_DYNAMIC) {
889 		(void)rtrequest(RTM_DELETE, rt_getkey(rt),
890 		    rt->rt_gateway, rt_mask(rt), rt->rt_flags, NULL);
891 	}
892 	/*
893 	 * A new route can be allocated
894 	 * the next time output is attempted.
895 	 */
896 	rtcache_free(&in6p->in6p_route);
897 }
898 
899 /*
900  * After a routing change, flush old routing.  A new route can be
901  * allocated the next time output is attempted.
902  */
903 void
904 in6_rtchange(struct in6pcb *in6p, int errno)
905 {
906 	if (in6p->in6p_af != AF_INET6)
907 		return;
908 
909 	rtcache_free(&in6p->in6p_route);
910 	/*
911 	 * A new route can be allocated the next time
912 	 * output is attempted.
913 	 */
914 }
915 
916 struct in6pcb *
917 in6_pcblookup_port(struct inpcbtable *table, struct in6_addr *laddr6,
918 		   u_int lport_arg, int lookup_wildcard, struct vestigial_inpcb *vp)
919 {
920 	struct inpcbhead *head;
921 	struct inpcb_hdr *inph;
922 	struct in6pcb *in6p, *match = 0;
923 	int matchwild = 3, wildcard;
924 	u_int16_t lport = lport_arg;
925 
926 	if (vp)
927 		vp->valid = 0;
928 
929 	head = IN6PCBHASH_PORT(table, lport);
930 	LIST_FOREACH(inph, head, inph_lhash) {
931 		in6p = (struct in6pcb *)inph;
932 		if (in6p->in6p_af != AF_INET6)
933 			continue;
934 
935 		if (in6p->in6p_lport != lport)
936 			continue;
937 		wildcard = 0;
938 		if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_faddr)) {
939 			if ((in6p->in6p_flags & IN6P_IPV6_V6ONLY) != 0)
940 				continue;
941 		}
942 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr))
943 			wildcard++;
944 		if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr)) {
945 			if ((in6p->in6p_flags & IN6P_IPV6_V6ONLY) != 0)
946 				continue;
947 			if (!IN6_IS_ADDR_V4MAPPED(laddr6))
948 				continue;
949 
950 			/* duplicate of IPv4 logic */
951 			wildcard = 0;
952 			if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_faddr) &&
953 			    in6p->in6p_faddr.s6_addr32[3])
954 				wildcard++;
955 			if (!in6p->in6p_laddr.s6_addr32[3]) {
956 				if (laddr6->s6_addr32[3])
957 					wildcard++;
958 			} else {
959 				if (!laddr6->s6_addr32[3])
960 					wildcard++;
961 				else {
962 					if (in6p->in6p_laddr.s6_addr32[3] !=
963 					    laddr6->s6_addr32[3])
964 						continue;
965 				}
966 			}
967 		} else if (IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
968 			if (IN6_IS_ADDR_V4MAPPED(laddr6)) {
969 				if ((in6p->in6p_flags & IN6P_IPV6_V6ONLY) != 0)
970 					continue;
971 			}
972 			if (!IN6_IS_ADDR_UNSPECIFIED(laddr6))
973 				wildcard++;
974 		} else {
975 			if (IN6_IS_ADDR_V4MAPPED(laddr6)) {
976 				if ((in6p->in6p_flags & IN6P_IPV6_V6ONLY) != 0)
977 					continue;
978 			}
979 			if (IN6_IS_ADDR_UNSPECIFIED(laddr6))
980 				wildcard++;
981 			else {
982 				if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr,
983 				    laddr6))
984 					continue;
985 			}
986 		}
987 		if (wildcard && !lookup_wildcard)
988 			continue;
989 		if (wildcard < matchwild) {
990 			match = in6p;
991 			matchwild = wildcard;
992 			if (matchwild == 0)
993 				break;
994 		}
995 	}
996 	if (match && matchwild == 0)
997 		return match;
998 
999 	if (vp && table->vestige && table->vestige->init_ports6) {
1000 		struct vestigial_inpcb better;
1001 		void *state;
1002 
1003 		state = (*table->vestige->init_ports6)(laddr6,
1004 						       lport_arg,
1005 						       lookup_wildcard);
1006 		while (table->vestige
1007 		       && (*table->vestige->next_port6)(state, vp)) {
1008 
1009 			if (vp->lport != lport)
1010 				continue;
1011 			wildcard = 0;
1012 			if (!IN6_IS_ADDR_UNSPECIFIED(&vp->faddr.v6))
1013 				wildcard++;
1014 			if (IN6_IS_ADDR_UNSPECIFIED(&vp->laddr.v6)) {
1015 				if (!IN6_IS_ADDR_UNSPECIFIED(laddr6))
1016 					wildcard++;
1017 			} else {
1018 				if (IN6_IS_ADDR_V4MAPPED(laddr6)) {
1019 					if (vp->v6only)
1020 						continue;
1021 				}
1022 				if (IN6_IS_ADDR_UNSPECIFIED(laddr6))
1023 					wildcard++;
1024 				else {
1025 					if (!IN6_ARE_ADDR_EQUAL(&vp->laddr.v6, laddr6))
1026 						continue;
1027 				}
1028 			}
1029 			if (wildcard && !lookup_wildcard)
1030 				continue;
1031 			if (wildcard < matchwild) {
1032 				better = *vp;
1033 				match  = (void*)&better;
1034 
1035 				matchwild = wildcard;
1036 				if (matchwild == 0)
1037 					break;
1038 			}
1039 		}
1040 
1041 		if (match) {
1042 			if (match != (void*)&better)
1043 				return match;
1044 			else {
1045 				*vp = better;
1046 				return 0;
1047 			}
1048 		}
1049 	}
1050 	return (match);
1051 }
1052 
1053 /*
1054  * WARNING: return value (rtentry) could be IPv4 one if in6pcb is connected to
1055  * IPv4 mapped address.
1056  */
1057 struct rtentry *
1058 in6_pcbrtentry(struct in6pcb *in6p)
1059 {
1060 	struct rtentry *rt;
1061 	struct route *ro;
1062 	union {
1063 		const struct sockaddr *sa;
1064 		const struct sockaddr_in6 *sa6;
1065 #ifdef INET
1066 		const struct sockaddr_in *sa4;
1067 #endif
1068 	} cdst;
1069 
1070 	ro = &in6p->in6p_route;
1071 
1072 	if (in6p->in6p_af != AF_INET6)
1073 		return (NULL);
1074 
1075 	cdst.sa = rtcache_getdst(ro);
1076 	if (cdst.sa == NULL)
1077 		;
1078 #ifdef INET
1079 	else if (cdst.sa->sa_family == AF_INET) {
1080 		KASSERT(IN6_IS_ADDR_V4MAPPED(&in6p->in6p_faddr));
1081 		if (cdst.sa4->sin_addr.s_addr != in6p->in6p_faddr.s6_addr32[3])
1082 			rtcache_free(ro);
1083 	}
1084 #endif
1085 	else {
1086 		if (!IN6_ARE_ADDR_EQUAL(&cdst.sa6->sin6_addr,
1087 					&in6p->in6p_faddr))
1088 			rtcache_free(ro);
1089 	}
1090 	if ((rt = rtcache_validate(ro)) == NULL)
1091 		rt = rtcache_update(ro, 1);
1092 #ifdef INET
1093 	if (rt == NULL && IN6_IS_ADDR_V4MAPPED(&in6p->in6p_faddr)) {
1094 		union {
1095 			struct sockaddr		dst;
1096 			struct sockaddr_in	dst4;
1097 		} u;
1098 		struct in_addr addr;
1099 
1100 		addr.s_addr = in6p->in6p_faddr.s6_addr32[3];
1101 
1102 		sockaddr_in_init(&u.dst4, &addr, 0);
1103 		rtcache_setdst(ro, &u.dst);
1104 
1105 		rt = rtcache_init(ro);
1106 	} else
1107 #endif
1108 	if (rt == NULL && !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) {
1109 		union {
1110 			struct sockaddr		dst;
1111 			struct sockaddr_in6	dst6;
1112 		} u;
1113 
1114 		sockaddr_in6_init(&u.dst6, &in6p->in6p_faddr, 0, 0, 0);
1115 		rtcache_setdst(ro, &u.dst);
1116 
1117 		rt = rtcache_init(ro);
1118 	}
1119 	return rt;
1120 }
1121 
1122 struct in6pcb *
1123 in6_pcblookup_connect(struct inpcbtable *table, const struct in6_addr *faddr6,
1124 		      u_int fport_arg, const struct in6_addr *laddr6, u_int lport_arg,
1125 		      int faith,
1126 		      struct vestigial_inpcb *vp)
1127 {
1128 	struct inpcbhead *head;
1129 	struct inpcb_hdr *inph;
1130 	struct in6pcb *in6p;
1131 	u_int16_t fport = fport_arg, lport = lport_arg;
1132 
1133 	if (vp)
1134 		vp->valid = 0;
1135 
1136 	head = IN6PCBHASH_CONNECT(table, faddr6, fport, laddr6, lport);
1137 	LIST_FOREACH(inph, head, inph_hash) {
1138 		in6p = (struct in6pcb *)inph;
1139 		if (in6p->in6p_af != AF_INET6)
1140 			continue;
1141 
1142 		/* find exact match on both source and dest */
1143 		if (in6p->in6p_fport != fport)
1144 			continue;
1145 		if (in6p->in6p_lport != lport)
1146 			continue;
1147 		if (IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr))
1148 			continue;
1149 		if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, faddr6))
1150 			continue;
1151 		if (IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr))
1152 			continue;
1153 		if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, laddr6))
1154 			continue;
1155 		if ((IN6_IS_ADDR_V4MAPPED(laddr6) ||
1156 		     IN6_IS_ADDR_V4MAPPED(faddr6)) &&
1157 		    (in6p->in6p_flags & IN6P_IPV6_V6ONLY))
1158 			continue;
1159 		return in6p;
1160 	}
1161 	if (vp && table->vestige) {
1162 		if ((*table->vestige->lookup6)(faddr6, fport_arg,
1163 					       laddr6, lport_arg, vp))
1164 			return 0;
1165 	}
1166 
1167 	return NULL;
1168 }
1169 
1170 struct in6pcb *
1171 in6_pcblookup_bind(struct inpcbtable *table, const struct in6_addr *laddr6,
1172 	u_int lport_arg, int faith)
1173 {
1174 	struct inpcbhead *head;
1175 	struct inpcb_hdr *inph;
1176 	struct in6pcb *in6p;
1177 	u_int16_t lport = lport_arg;
1178 #ifdef INET
1179 	struct in6_addr zero_mapped;
1180 #endif
1181 
1182 	head = IN6PCBHASH_BIND(table, laddr6, lport);
1183 	LIST_FOREACH(inph, head, inph_hash) {
1184 		in6p = (struct in6pcb *)inph;
1185 		if (in6p->in6p_af != AF_INET6)
1186 			continue;
1187 
1188 		if (faith && (in6p->in6p_flags & IN6P_FAITH) == 0)
1189 			continue;
1190 		if (in6p->in6p_fport != 0)
1191 			continue;
1192 		if (in6p->in6p_lport != lport)
1193 			continue;
1194 		if (IN6_IS_ADDR_V4MAPPED(laddr6) &&
1195 		    (in6p->in6p_flags & IN6P_IPV6_V6ONLY) != 0)
1196 			continue;
1197 		if (IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, laddr6))
1198 			goto out;
1199 	}
1200 #ifdef INET
1201 	if (IN6_IS_ADDR_V4MAPPED(laddr6)) {
1202 		memset(&zero_mapped, 0, sizeof(zero_mapped));
1203 		zero_mapped.s6_addr16[5] = 0xffff;
1204 		head = IN6PCBHASH_BIND(table, &zero_mapped, lport);
1205 		LIST_FOREACH(inph, head, inph_hash) {
1206 			in6p = (struct in6pcb *)inph;
1207 			if (in6p->in6p_af != AF_INET6)
1208 				continue;
1209 
1210 			if (faith && (in6p->in6p_flags & IN6P_FAITH) == 0)
1211 				continue;
1212 			if (in6p->in6p_fport != 0)
1213 				continue;
1214 			if (in6p->in6p_lport != lport)
1215 				continue;
1216 			if ((in6p->in6p_flags & IN6P_IPV6_V6ONLY) != 0)
1217 				continue;
1218 			if (IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &zero_mapped))
1219 				goto out;
1220 		}
1221 	}
1222 #endif
1223 	head = IN6PCBHASH_BIND(table, &zeroin6_addr, lport);
1224 	LIST_FOREACH(inph, head, inph_hash) {
1225 		in6p = (struct in6pcb *)inph;
1226 		if (in6p->in6p_af != AF_INET6)
1227 			continue;
1228 
1229 		if (faith && (in6p->in6p_flags & IN6P_FAITH) == 0)
1230 			continue;
1231 		if (in6p->in6p_fport != 0)
1232 			continue;
1233 		if (in6p->in6p_lport != lport)
1234 			continue;
1235 		if (IN6_IS_ADDR_V4MAPPED(laddr6) &&
1236 		    (in6p->in6p_flags & IN6P_IPV6_V6ONLY) != 0)
1237 			continue;
1238 		if (IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &zeroin6_addr))
1239 			goto out;
1240 	}
1241 	return (NULL);
1242 
1243 out:
1244 	inph = &in6p->in6p_head;
1245 	if (inph != LIST_FIRST(head)) {
1246 		LIST_REMOVE(inph, inph_hash);
1247 		LIST_INSERT_HEAD(head, inph, inph_hash);
1248 	}
1249 	return in6p;
1250 }
1251 
1252 void
1253 in6_pcbstate(struct in6pcb *in6p, int state)
1254 {
1255 
1256 	if (in6p->in6p_af != AF_INET6)
1257 		return;
1258 
1259 	if (in6p->in6p_state > IN6P_ATTACHED)
1260 		LIST_REMOVE(&in6p->in6p_head, inph_hash);
1261 
1262 	switch (state) {
1263 	case IN6P_BOUND:
1264 		LIST_INSERT_HEAD(IN6PCBHASH_BIND(in6p->in6p_table,
1265 		    &in6p->in6p_laddr, in6p->in6p_lport), &in6p->in6p_head,
1266 		    inph_hash);
1267 		break;
1268 	case IN6P_CONNECTED:
1269 		LIST_INSERT_HEAD(IN6PCBHASH_CONNECT(in6p->in6p_table,
1270 		    &in6p->in6p_faddr, in6p->in6p_fport,
1271 		    &in6p->in6p_laddr, in6p->in6p_lport), &in6p->in6p_head,
1272 		    inph_hash);
1273 		break;
1274 	}
1275 
1276 	in6p->in6p_state = state;
1277 }
1278