xref: /netbsd/sys/netinet6/in6_pcb.c (revision e289043c)
1 /*	$NetBSD: in6_pcb.c,v 1.171 2022/10/14 19:39:32 ryo 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 1.171 2022/10/14 19:39:32 ryo 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 
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 
161 	KASSERT(so->so_proto->pr_domain->dom_family == AF_INET6);
162 
163 	in6p = pool_get(&in6pcb_pool, PR_NOWAIT);
164 	if (in6p == NULL)
165 		return (ENOBUFS);
166 	memset((void *)in6p, 0, sizeof(*in6p));
167 	in6p->in6p_af = AF_INET6;
168 	in6p->in6p_table = table;
169 	in6p->in6p_socket = so;
170 	in6p->in6p_hops = -1;	/* use kernel default */
171 	in6p->in6p_icmp6filt = NULL;
172 	in6p->in6p_portalgo = PORTALGO_DEFAULT;
173 	in6p->in6p_bindportonsend = false;
174 #if defined(IPSEC)
175 	if (ipsec_enabled) {
176 		int error = ipsec_init_pcbpolicy(so, &in6p->in6p_sp);
177 		if (error != 0) {
178 			pool_put(&in6pcb_pool, in6p);
179 			return error;
180 		}
181 		in6p->in6p_sp->sp_inph = (struct inpcb_hdr *)in6p;
182 	}
183 #endif /* IPSEC */
184 	s = splsoftnet();
185 	TAILQ_INSERT_HEAD(&table->inpt_queue, (struct inpcb_hdr*)in6p,
186 	    inph_queue);
187 	LIST_INSERT_HEAD(IN6PCBHASH_PORT(table, in6p->in6p_lport),
188 	    &in6p->in6p_head, inph_lhash);
189 	in6_pcbstate(in6p, IN6P_ATTACHED);
190 	splx(s);
191 	if (ip6_v6only)
192 		in6p->in6p_flags |= IN6P_IPV6_V6ONLY;
193 	so->so_pcb = (void *)in6p;
194 	return (0);
195 }
196 
197 /*
198  * Bind address from sin6 to in6p.
199  */
200 static int
201 in6_pcbbind_addr(struct in6pcb *in6p, struct sockaddr_in6 *sin6, struct lwp *l)
202 {
203 	int error;
204 	int s;
205 
206 	/*
207 	 * We should check the family, but old programs
208 	 * incorrectly fail to initialize it.
209 	 */
210 	if (sin6->sin6_family != AF_INET6)
211 		return (EAFNOSUPPORT);
212 
213 #ifndef INET
214 	if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr))
215 		return (EADDRNOTAVAIL);
216 #endif
217 
218 	if ((error = sa6_embedscope(sin6, ip6_use_defzone)) != 0)
219 		return (error);
220 
221 	s = pserialize_read_enter();
222 	if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
223 		if ((in6p->in6p_flags & IN6P_IPV6_V6ONLY) != 0) {
224 			error = EINVAL;
225 			goto out;
226 		}
227 		if (sin6->sin6_addr.s6_addr32[3]) {
228 			struct sockaddr_in sin;
229 
230 			memset(&sin, 0, sizeof(sin));
231 			sin.sin_len = sizeof(sin);
232 			sin.sin_family = AF_INET;
233 			bcopy(&sin6->sin6_addr.s6_addr32[3],
234 			    &sin.sin_addr, sizeof(sin.sin_addr));
235 			if (!IN_MULTICAST(sin.sin_addr.s_addr)) {
236 				struct ifaddr *ifa;
237 				ifa = ifa_ifwithaddr((struct sockaddr *)&sin);
238 				if (ifa == NULL &&
239 				    (in6p->in6p_flags & IN6P_BINDANY) == 0) {
240 					error = EADDRNOTAVAIL;
241 					goto out;
242 				}
243 			}
244 		}
245 	} else if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
246 		// succeed
247 	} else if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
248 		struct ifaddr *ifa = NULL;
249 
250 		if ((in6p->in6p_flags & IN6P_FAITH) == 0) {
251 			ifa = ifa_ifwithaddr(sin6tosa(sin6));
252 			if (ifa == NULL &&
253 			    (in6p->in6p_flags & IN6P_BINDANY) == 0) {
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 in6_addr ia6;
448 	struct ifnet *ifp = NULL;	/* outgoing interface */
449 	int error = 0;
450 	int scope_ambiguous = 0;
451 #ifdef INET
452 	struct in6_addr mapped;
453 #endif
454 	struct sockaddr_in6 tmp;
455 	struct vestigial_inpcb vestige;
456 	struct psref psref;
457 	int bound;
458 
459 	(void)&in6a;				/* XXX fool gcc */
460 
461 	if (in6p->in6p_af != AF_INET6)
462 		return (EINVAL);
463 
464 	if (sin6->sin6_len != sizeof(*sin6))
465 		return (EINVAL);
466 	if (sin6->sin6_family != AF_INET6)
467 		return (EAFNOSUPPORT);
468 	if (sin6->sin6_port == 0)
469 		return (EADDRNOTAVAIL);
470 
471 	if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr) &&
472 	    in6p->in6p_socket->so_type == SOCK_STREAM)
473 		return EADDRNOTAVAIL;
474 
475 	if (sin6->sin6_scope_id == 0 && !ip6_use_defzone)
476 		scope_ambiguous = 1;
477 	if ((error = sa6_embedscope(sin6, ip6_use_defzone)) != 0)
478 		return(error);
479 
480 	/* sanity check for mapped address case */
481 	if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
482 		if ((in6p->in6p_flags & IN6P_IPV6_V6ONLY) != 0)
483 			return EINVAL;
484 		if (IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr))
485 			in6p->in6p_laddr.s6_addr16[5] = htons(0xffff);
486 		if (!IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr))
487 			return EINVAL;
488 	} else
489 	{
490 		if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr))
491 			return EINVAL;
492 	}
493 
494 	/* protect *sin6 from overwrites */
495 	tmp = *sin6;
496 	sin6 = &tmp;
497 
498 	bound = curlwp_bind();
499 	/* Source address selection. */
500 	if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr) &&
501 	    in6p->in6p_laddr.s6_addr32[3] == 0) {
502 #ifdef INET
503 		struct sockaddr_in sin;
504 		struct in_ifaddr *ia4;
505 		struct psref _psref;
506 
507 		memset(&sin, 0, sizeof(sin));
508 		sin.sin_len = sizeof(sin);
509 		sin.sin_family = AF_INET;
510 		memcpy(&sin.sin_addr, &sin6->sin6_addr.s6_addr32[3],
511 			sizeof(sin.sin_addr));
512 		ia4 = in_selectsrc(&sin, &in6p->in6p_route,
513 			in6p->in6p_socket->so_options, NULL, &error, &_psref);
514 		if (ia4 == NULL) {
515 			if (error == 0)
516 				error = EADDRNOTAVAIL;
517 			curlwp_bindx(bound);
518 			return (error);
519 		}
520 		memset(&mapped, 0, sizeof(mapped));
521 		mapped.s6_addr16[5] = htons(0xffff);
522 		memcpy(&mapped.s6_addr32[3], &IA_SIN(ia4)->sin_addr,
523 		    sizeof(IA_SIN(ia4)->sin_addr));
524 		ia4_release(ia4, &_psref);
525 		in6a = &mapped;
526 #else
527 		curlwp_bindx(bound);
528 		return EADDRNOTAVAIL;
529 #endif
530 	} else {
531 		/*
532 		 * XXX: in6_selectsrc might replace the bound local address
533 		 * with the address specified by setsockopt(IPV6_PKTINFO).
534 		 * Is it the intended behavior?
535 		 */
536 		error = in6_selectsrc(sin6, in6p->in6p_outputopts,
537 		    in6p->in6p_moptions, &in6p->in6p_route, &in6p->in6p_laddr,
538 		    &ifp, &psref, &ia6);
539 		if (error == 0)
540 			in6a = &ia6;
541 		if (ifp && scope_ambiguous &&
542 		    (error = in6_setscope(&sin6->sin6_addr, ifp, NULL)) != 0) {
543 			if_put(ifp, &psref);
544 			curlwp_bindx(bound);
545 			return error;
546 		}
547 
548 		if (in6a == NULL) {
549 			if_put(ifp, &psref);
550 			curlwp_bindx(bound);
551 			if (error == 0)
552 				error = EADDRNOTAVAIL;
553 			return error;
554 		}
555 	}
556 
557 	if (ifp != NULL) {
558 		in6p->in6p_ip6.ip6_hlim = (u_int8_t)in6_selecthlim(in6p, ifp);
559 		if_put(ifp, &psref);
560 	} else
561 		in6p->in6p_ip6.ip6_hlim = (u_int8_t)in6_selecthlim_rt(in6p);
562 	curlwp_bindx(bound);
563 
564 	if (in6_pcblookup_connect(in6p->in6p_table, &sin6->sin6_addr,
565 	    sin6->sin6_port,
566 	    IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) ? in6a : &in6p->in6p_laddr,
567 				  in6p->in6p_lport, 0, &vestige)
568 		|| vestige.valid)
569 		return (EADDRINUSE);
570 	if (IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) ||
571 	    (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr) &&
572 	     in6p->in6p_laddr.s6_addr32[3] == 0))
573 	{
574 		if (in6p->in6p_lport == 0) {
575 			error = in6_pcbbind(in6p, NULL, l);
576 			if (error != 0)
577 				return error;
578 		}
579 		in6p->in6p_laddr = *in6a;
580 	}
581 	in6p->in6p_faddr = sin6->sin6_addr;
582 	in6p->in6p_fport = sin6->sin6_port;
583 
584         /* Late bind, if needed */
585 	if (in6p->in6p_bindportonsend) {
586                struct sockaddr_in6 lsin = *((const struct sockaddr_in6 *)
587 		    in6p->in6p_socket->so_proto->pr_domain->dom_sa_any);
588 		lsin.sin6_addr = in6p->in6p_laddr;
589 		lsin.sin6_port = 0;
590 
591                if ((error = in6_pcbbind_port(in6p, &lsin, l)) != 0)
592                        return error;
593 	}
594 
595 	in6_pcbstate(in6p, IN6P_CONNECTED);
596 	in6p->in6p_flowinfo &= ~IPV6_FLOWLABEL_MASK;
597 	if (ip6_auto_flowlabel)
598 		in6p->in6p_flowinfo |=
599 		    (htonl(ip6_randomflowlabel()) & IPV6_FLOWLABEL_MASK);
600 #if defined(IPSEC)
601 	if (ipsec_enabled && in6p->in6p_socket->so_type == SOCK_STREAM)
602 		ipsec_pcbconn(in6p->in6p_sp);
603 #endif
604 	return (0);
605 }
606 
607 void
608 in6_pcbdisconnect(struct in6pcb *in6p)
609 {
610 	memset((void *)&in6p->in6p_faddr, 0, sizeof(in6p->in6p_faddr));
611 	in6p->in6p_fport = 0;
612 	in6_pcbstate(in6p, IN6P_BOUND);
613 	in6p->in6p_flowinfo &= ~IPV6_FLOWLABEL_MASK;
614 #if defined(IPSEC)
615 	if (ipsec_enabled)
616 		ipsec_pcbdisconn(in6p->in6p_sp);
617 #endif
618 	if (in6p->in6p_socket->so_state & SS_NOFDREF)
619 		in6_pcbdetach(in6p);
620 }
621 
622 void
623 in6_pcbdetach(struct in6pcb *in6p)
624 {
625 	struct socket *so = in6p->in6p_socket;
626 	int s;
627 
628 	if (in6p->in6p_af != AF_INET6)
629 		return;
630 
631 #if defined(IPSEC)
632 	if (ipsec_enabled)
633 		ipsec_delete_pcbpolicy(in6p);
634 #endif
635 	so->so_pcb = NULL;
636 
637 	s = splsoftnet();
638 	in6_pcbstate(in6p, IN6P_ATTACHED);
639 	LIST_REMOVE(&in6p->in6p_head, inph_lhash);
640 	TAILQ_REMOVE(&in6p->in6p_table->inpt_queue, &in6p->in6p_head,
641 	    inph_queue);
642 	splx(s);
643 
644 	if (in6p->in6p_options) {
645 		m_freem(in6p->in6p_options);
646 	}
647 	if (in6p->in6p_outputopts != NULL) {
648 		ip6_clearpktopts(in6p->in6p_outputopts, -1);
649 		free(in6p->in6p_outputopts, M_IP6OPT);
650 	}
651 	rtcache_free(&in6p->in6p_route);
652 	ip6_freemoptions(in6p->in6p_moptions);
653 	ip_freemoptions(in6p->in6p_v4moptions);
654 	sofree(so);				/* drops the socket's lock */
655 
656 	pool_put(&in6pcb_pool, in6p);
657 	mutex_enter(softnet_lock);		/* reacquire it */
658 }
659 
660 void
661 in6_setsockaddr(struct in6pcb *in6p, struct sockaddr_in6 *sin6)
662 {
663 
664 	if (in6p->in6p_af != AF_INET6)
665 		return;
666 
667 	sockaddr_in6_init(sin6, &in6p->in6p_laddr, in6p->in6p_lport, 0, 0);
668 	(void)sa6_recoverscope(sin6); /* XXX: should catch errors */
669 }
670 
671 void
672 in6_setpeeraddr(struct in6pcb *in6p, struct sockaddr_in6 *sin6)
673 {
674 
675 	if (in6p->in6p_af != AF_INET6)
676 		return;
677 
678 	sockaddr_in6_init(sin6, &in6p->in6p_faddr, in6p->in6p_fport, 0, 0);
679 	(void)sa6_recoverscope(sin6); /* XXX: should catch errors */
680 }
681 
682 /*
683  * Pass some notification to all connections of a protocol
684  * associated with address dst.  The local address and/or port numbers
685  * may be specified to limit the search.  The "usual action" will be
686  * taken, depending on the ctlinput cmd.  The caller must filter any
687  * cmds that are uninteresting (e.g., no error in the map).
688  * Call the protocol specific routine (if any) to report
689  * any errors for each matching socket.
690  *
691  * Must be called at splsoftnet.
692  *
693  * Note: src (4th arg) carries the flowlabel value on the original IPv6
694  * header, in sin6_flowinfo member.
695  */
696 int
697 in6_pcbnotify(struct inpcbtable *table, const struct sockaddr *dst,
698     u_int fport_arg, const struct sockaddr *src, u_int lport_arg, int cmd,
699     void *cmdarg, void (*notify)(struct in6pcb *, int))
700 {
701 	struct inpcb_hdr *inph;
702 	struct sockaddr_in6 sa6_src;
703 	const struct sockaddr_in6 *sa6_dst;
704 	u_int16_t fport = fport_arg, lport = lport_arg;
705 	int errno;
706 	int nmatch = 0;
707 	u_int32_t flowinfo;
708 
709 	if ((unsigned)cmd >= PRC_NCMDS || dst->sa_family != AF_INET6)
710 		return 0;
711 
712 	sa6_dst = (const struct sockaddr_in6 *)dst;
713 	if (IN6_IS_ADDR_UNSPECIFIED(&sa6_dst->sin6_addr))
714 		return 0;
715 
716 	/*
717 	 * note that src can be NULL when we get notify by local fragmentation.
718 	 */
719 	sa6_src = (src == NULL) ? sa6_any : *(const struct sockaddr_in6 *)src;
720 	flowinfo = sa6_src.sin6_flowinfo;
721 
722 	/*
723 	 * Redirects go to all references to the destination,
724 	 * and use in6_rtchange to invalidate the route cache.
725 	 * Dead host indications: also use in6_rtchange to invalidate
726 	 * the cache, and deliver the error to all the sockets.
727 	 * Otherwise, if we have knowledge of the local port and address,
728 	 * deliver only to that socket.
729 	 */
730 	if (PRC_IS_REDIRECT(cmd) || cmd == PRC_HOSTDEAD) {
731 		fport = 0;
732 		lport = 0;
733 		memset((void *)&sa6_src.sin6_addr, 0, sizeof(sa6_src.sin6_addr));
734 
735 		if (cmd != PRC_HOSTDEAD)
736 			notify = in6_rtchange;
737 	}
738 
739 	errno = inet6ctlerrmap[cmd];
740 	TAILQ_FOREACH(inph, &table->inpt_queue, inph_queue) {
741 		struct in6pcb *in6p = (struct in6pcb *)inph;
742 		struct rtentry *rt = NULL;
743 
744 		if (in6p->in6p_af != AF_INET6)
745 			continue;
746 
747 		/*
748 		 * Under the following condition, notify of redirects
749 		 * to the pcb, without making address matches against inpcb.
750 		 * - redirect notification is arrived.
751 		 * - the inpcb is unconnected.
752 		 * - the inpcb is caching !RTF_HOST routing entry.
753 		 * - the ICMPv6 notification is from the gateway cached in the
754 		 *   inpcb.  i.e. ICMPv6 notification is from nexthop gateway
755 		 *   the inpcb used very recently.
756 		 *
757 		 * This is to improve interaction between netbsd/openbsd
758 		 * redirect handling code, and inpcb route cache code.
759 		 * without the clause, !RTF_HOST routing entry (which carries
760 		 * gateway used by inpcb right before the ICMPv6 redirect)
761 		 * will be cached forever in unconnected inpcb.
762 		 *
763 		 * There still is a question regarding to what is TRT:
764 		 * - On bsdi/freebsd, RTF_HOST (cloned) routing entry will be
765 		 *   generated on packet output.  inpcb will always cache
766 		 *   RTF_HOST routing entry so there's no need for the clause
767 		 *   (ICMPv6 redirect will update RTF_HOST routing entry,
768 		 *   and inpcb is caching it already).
769 		 *   However, bsdi/freebsd are vulnerable to local DoS attacks
770 		 *   due to the cloned routing entries.
771 		 * - Specwise, "destination cache" is mentioned in RFC2461.
772 		 *   Jinmei says that it implies bsdi/freebsd behavior, itojun
773 		 *   is not really convinced.
774 		 * - Having hiwat/lowat on # of cloned host route (redirect/
775 		 *   pmtud) may be a good idea.  netbsd/openbsd has it.  see
776 		 *   icmp6_mtudisc_update().
777 		 */
778 		if ((PRC_IS_REDIRECT(cmd) || cmd == PRC_HOSTDEAD) &&
779 		    IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
780 		    (rt = rtcache_validate(&in6p->in6p_route)) != NULL &&
781 		    !(rt->rt_flags & RTF_HOST)) {
782 			const struct sockaddr_in6 *dst6;
783 
784 			dst6 = (const struct sockaddr_in6 *)
785 			    rtcache_getdst(&in6p->in6p_route);
786 			if (dst6 == NULL)
787 				;
788 			else if (IN6_ARE_ADDR_EQUAL(&dst6->sin6_addr,
789 			    &sa6_dst->sin6_addr)) {
790 				rtcache_unref(rt, &in6p->in6p_route);
791 				goto do_notify;
792 			}
793 		}
794 		rtcache_unref(rt, &in6p->in6p_route);
795 
796 		/*
797 		 * If the error designates a new path MTU for a destination
798 		 * and the application (associated with this socket) wanted to
799 		 * know the value, notify. Note that we notify for all
800 		 * disconnected sockets if the corresponding application
801 		 * wanted. This is because some UDP applications keep sending
802 		 * sockets disconnected.
803 		 * XXX: should we avoid to notify the value to TCP sockets?
804 		 */
805 		if (cmd == PRC_MSGSIZE && (in6p->in6p_flags & IN6P_MTU) != 0 &&
806 		    (IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) ||
807 		     IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &sa6_dst->sin6_addr))) {
808 			ip6_notify_pmtu(in6p, (const struct sockaddr_in6 *)dst,
809 					(u_int32_t *)cmdarg);
810 		}
811 
812 		/*
813 		 * Detect if we should notify the error. If no source and
814 		 * destination ports are specified, but non-zero flowinfo and
815 		 * local address match, notify the error. This is the case
816 		 * when the error is delivered with an encrypted buffer
817 		 * by ESP. Otherwise, just compare addresses and ports
818 		 * as usual.
819 		 */
820 		if (lport == 0 && fport == 0 && flowinfo &&
821 		    in6p->in6p_socket != NULL &&
822 		    flowinfo == (in6p->in6p_flowinfo & IPV6_FLOWLABEL_MASK) &&
823 		    IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &sa6_src.sin6_addr))
824 			goto do_notify;
825 		else if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr,
826 					     &sa6_dst->sin6_addr) ||
827 		    in6p->in6p_socket == NULL ||
828 		    (lport && in6p->in6p_lport != lport) ||
829 		    (!IN6_IS_ADDR_UNSPECIFIED(&sa6_src.sin6_addr) &&
830 		     !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr,
831 					 &sa6_src.sin6_addr)) ||
832 		    (fport && in6p->in6p_fport != fport))
833 			continue;
834 
835 	  do_notify:
836 		if (notify)
837 			(*notify)(in6p, errno);
838 		nmatch++;
839 	}
840 	return nmatch;
841 }
842 
843 void
844 in6_pcbpurgeif0(struct inpcbtable *table, struct ifnet *ifp)
845 {
846 	struct inpcb_hdr *inph;
847 	struct ip6_moptions *im6o;
848 	struct in6_multi_mship *imm, *nimm;
849 
850 	KASSERT(ifp != NULL);
851 
852 	TAILQ_FOREACH(inph, &table->inpt_queue, inph_queue) {
853 		struct in6pcb *in6p = (struct in6pcb *)inph;
854 		bool need_unlock = false;
855 		if (in6p->in6p_af != AF_INET6)
856 			continue;
857 
858 		/* The caller holds either one of in6ps' lock */
859 		if (!in6p_locked(in6p)) {
860 			in6p_lock(in6p);
861 			need_unlock = true;
862 		}
863 		im6o = in6p->in6p_moptions;
864 		if (im6o) {
865 			/*
866 			 * Unselect the outgoing interface if it is being
867 			 * detached.
868 			 */
869 			if (im6o->im6o_multicast_if_index == ifp->if_index)
870 				im6o->im6o_multicast_if_index = 0;
871 
872 			/*
873 			 * Drop multicast group membership if we joined
874 			 * through the interface being detached.
875 			 * XXX controversial - is it really legal for kernel
876 			 * to force this?
877 			 */
878 			LIST_FOREACH_SAFE(imm, &im6o->im6o_memberships,
879 			    i6mm_chain, nimm) {
880 				if (imm->i6mm_maddr->in6m_ifp == ifp) {
881 					LIST_REMOVE(imm, i6mm_chain);
882 					in6_leavegroup(imm);
883 				}
884 			}
885 		}
886 
887 		in_purgeifmcast(in6p->in6p_v4moptions, ifp);
888 
889 		if (need_unlock)
890 			in6p_unlock(in6p);
891 	}
892 }
893 
894 void
895 in6_pcbpurgeif(struct inpcbtable *table, struct ifnet *ifp)
896 {
897 	struct rtentry *rt;
898 	struct inpcb_hdr *inph;
899 
900 	TAILQ_FOREACH(inph, &table->inpt_queue, inph_queue) {
901 		struct in6pcb *in6p = (struct in6pcb *)inph;
902 		if (in6p->in6p_af != AF_INET6)
903 			continue;
904 		if ((rt = rtcache_validate(&in6p->in6p_route)) != NULL &&
905 		    rt->rt_ifp == ifp) {
906 			rtcache_unref(rt, &in6p->in6p_route);
907 			in6_rtchange(in6p, 0);
908 		} else
909 			rtcache_unref(rt, &in6p->in6p_route);
910 	}
911 }
912 
913 /*
914  * Check for alternatives when higher level complains
915  * about service problems.  For now, invalidate cached
916  * routing information.  If the route was created dynamically
917  * (by a redirect), time to try a default gateway again.
918  */
919 void
920 in6_losing(struct in6pcb *in6p)
921 {
922 	struct rtentry *rt;
923 	struct rt_addrinfo info;
924 
925 	if (in6p->in6p_af != AF_INET6)
926 		return;
927 
928 	if ((rt = rtcache_validate(&in6p->in6p_route)) == NULL)
929 		return;
930 
931 	memset(&info, 0, sizeof(info));
932 	info.rti_info[RTAX_DST] = rtcache_getdst(&in6p->in6p_route);
933 	info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
934 	info.rti_info[RTAX_NETMASK] = rt_mask(rt);
935 	rt_missmsg(RTM_LOSING, &info, rt->rt_flags, 0);
936 	if (rt->rt_flags & RTF_DYNAMIC) {
937 		int error;
938 		struct rtentry *nrt;
939 
940 		error = rtrequest(RTM_DELETE, rt_getkey(rt),
941 		    rt->rt_gateway, rt_mask(rt), rt->rt_flags, &nrt);
942 		rtcache_unref(rt, &in6p->in6p_route);
943 		if (error == 0) {
944 			rt_newmsg_dynamic(RTM_DELETE, nrt);
945 			rt_free(nrt);
946 		}
947 	} else
948 		rtcache_unref(rt, &in6p->in6p_route);
949 	/*
950 	 * A new route can be allocated
951 	 * the next time output is attempted.
952 	 */
953 	rtcache_free(&in6p->in6p_route);
954 }
955 
956 /*
957  * After a routing change, flush old routing.  A new route can be
958  * allocated the next time output is attempted.
959  */
960 void
961 in6_rtchange(struct in6pcb *in6p, int errno)
962 {
963 	if (in6p->in6p_af != AF_INET6)
964 		return;
965 
966 	rtcache_free(&in6p->in6p_route);
967 	/*
968 	 * A new route can be allocated the next time
969 	 * output is attempted.
970 	 */
971 }
972 
973 struct in6pcb *
974 in6_pcblookup_port(struct inpcbtable *table, struct in6_addr *laddr6,
975 		   u_int lport_arg, int lookup_wildcard, struct vestigial_inpcb *vp)
976 {
977 	struct inpcbhead *head;
978 	struct inpcb_hdr *inph;
979 	struct in6pcb *in6p, *match = NULL;
980 	int matchwild = 3, wildcard;
981 	u_int16_t lport = lport_arg;
982 
983 	if (vp)
984 		vp->valid = 0;
985 
986 	head = IN6PCBHASH_PORT(table, lport);
987 	LIST_FOREACH(inph, head, inph_lhash) {
988 		in6p = (struct in6pcb *)inph;
989 		if (in6p->in6p_af != AF_INET6)
990 			continue;
991 
992 		if (in6p->in6p_lport != lport)
993 			continue;
994 		wildcard = 0;
995 		if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_faddr)) {
996 			if ((in6p->in6p_flags & IN6P_IPV6_V6ONLY) != 0)
997 				continue;
998 		}
999 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr))
1000 			wildcard++;
1001 		if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr)) {
1002 			if ((in6p->in6p_flags & IN6P_IPV6_V6ONLY) != 0)
1003 				continue;
1004 			if (!IN6_IS_ADDR_V4MAPPED(laddr6))
1005 				continue;
1006 
1007 			/* duplicate of IPv4 logic */
1008 			wildcard = 0;
1009 			if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_faddr) &&
1010 			    in6p->in6p_faddr.s6_addr32[3])
1011 				wildcard++;
1012 			if (!in6p->in6p_laddr.s6_addr32[3]) {
1013 				if (laddr6->s6_addr32[3])
1014 					wildcard++;
1015 			} else {
1016 				if (!laddr6->s6_addr32[3])
1017 					wildcard++;
1018 				else {
1019 					if (in6p->in6p_laddr.s6_addr32[3] !=
1020 					    laddr6->s6_addr32[3])
1021 						continue;
1022 				}
1023 			}
1024 		} else if (IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1025 			if (IN6_IS_ADDR_V4MAPPED(laddr6)) {
1026 				if ((in6p->in6p_flags & IN6P_IPV6_V6ONLY) != 0)
1027 					continue;
1028 			}
1029 			if (!IN6_IS_ADDR_UNSPECIFIED(laddr6))
1030 				wildcard++;
1031 		} else {
1032 			if (IN6_IS_ADDR_V4MAPPED(laddr6)) {
1033 				if ((in6p->in6p_flags & IN6P_IPV6_V6ONLY) != 0)
1034 					continue;
1035 			}
1036 			if (IN6_IS_ADDR_UNSPECIFIED(laddr6))
1037 				wildcard++;
1038 			else {
1039 				if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr,
1040 				    laddr6))
1041 					continue;
1042 			}
1043 		}
1044 		if (wildcard && !lookup_wildcard)
1045 			continue;
1046 		if (wildcard < matchwild) {
1047 			match = in6p;
1048 			matchwild = wildcard;
1049 			if (matchwild == 0)
1050 				break;
1051 		}
1052 	}
1053 	if (match && matchwild == 0)
1054 		return match;
1055 
1056 	if (vp && table->vestige && table->vestige->init_ports6) {
1057 		struct vestigial_inpcb better;
1058 		bool has_better = false;
1059 		void *state;
1060 
1061 		state = (*table->vestige->init_ports6)(laddr6,
1062 						       lport_arg,
1063 						       lookup_wildcard);
1064 		while (table->vestige
1065 		       && (*table->vestige->next_port6)(state, vp)) {
1066 
1067 			if (vp->lport != lport)
1068 				continue;
1069 			wildcard = 0;
1070 			if (!IN6_IS_ADDR_UNSPECIFIED(&vp->faddr.v6))
1071 				wildcard++;
1072 			if (IN6_IS_ADDR_UNSPECIFIED(&vp->laddr.v6)) {
1073 				if (!IN6_IS_ADDR_UNSPECIFIED(laddr6))
1074 					wildcard++;
1075 			} else {
1076 				if (IN6_IS_ADDR_V4MAPPED(laddr6)) {
1077 					if (vp->v6only)
1078 						continue;
1079 				}
1080 				if (IN6_IS_ADDR_UNSPECIFIED(laddr6))
1081 					wildcard++;
1082 				else {
1083 					if (!IN6_ARE_ADDR_EQUAL(&vp->laddr.v6, laddr6))
1084 						continue;
1085 				}
1086 			}
1087 			if (wildcard && !lookup_wildcard)
1088 				continue;
1089 			if (wildcard < matchwild) {
1090 				better = *vp;
1091 				has_better = true;
1092 
1093 				matchwild = wildcard;
1094 				if (matchwild == 0)
1095 					break;
1096 			}
1097 		}
1098 
1099 		if (has_better) {
1100 			*vp = better;
1101 			return 0;
1102 		}
1103 	}
1104 	return (match);
1105 }
1106 
1107 /*
1108  * WARNING: return value (rtentry) could be IPv4 one if in6pcb is connected to
1109  * IPv4 mapped address.
1110  */
1111 struct rtentry *
1112 in6_pcbrtentry(struct in6pcb *in6p)
1113 {
1114 	struct rtentry *rt;
1115 	struct route *ro;
1116 	union {
1117 		const struct sockaddr *sa;
1118 		const struct sockaddr_in6 *sa6;
1119 #ifdef INET
1120 		const struct sockaddr_in *sa4;
1121 #endif
1122 	} cdst;
1123 
1124 	ro = &in6p->in6p_route;
1125 
1126 	if (in6p->in6p_af != AF_INET6)
1127 		return (NULL);
1128 
1129 	cdst.sa = rtcache_getdst(ro);
1130 	if (cdst.sa == NULL)
1131 		;
1132 #ifdef INET
1133 	else if (cdst.sa->sa_family == AF_INET) {
1134 		KASSERT(IN6_IS_ADDR_V4MAPPED(&in6p->in6p_faddr));
1135 		if (cdst.sa4->sin_addr.s_addr != in6p->in6p_faddr.s6_addr32[3])
1136 			rtcache_free(ro);
1137 	}
1138 #endif
1139 	else {
1140 		if (!IN6_ARE_ADDR_EQUAL(&cdst.sa6->sin6_addr,
1141 					&in6p->in6p_faddr))
1142 			rtcache_free(ro);
1143 	}
1144 	if ((rt = rtcache_validate(ro)) == NULL)
1145 		rt = rtcache_update(ro, 1);
1146 #ifdef INET
1147 	if (rt == NULL && IN6_IS_ADDR_V4MAPPED(&in6p->in6p_faddr)) {
1148 		union {
1149 			struct sockaddr		dst;
1150 			struct sockaddr_in	dst4;
1151 		} u;
1152 		struct in_addr addr;
1153 
1154 		addr.s_addr = in6p->in6p_faddr.s6_addr32[3];
1155 
1156 		sockaddr_in_init(&u.dst4, &addr, 0);
1157 		if (rtcache_setdst(ro, &u.dst) != 0)
1158 			return NULL;
1159 
1160 		rt = rtcache_init(ro);
1161 	} else
1162 #endif
1163 	if (rt == NULL && !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) {
1164 		union {
1165 			struct sockaddr		dst;
1166 			struct sockaddr_in6	dst6;
1167 		} u;
1168 
1169 		sockaddr_in6_init(&u.dst6, &in6p->in6p_faddr, 0, 0, 0);
1170 		if (rtcache_setdst(ro, &u.dst) != 0)
1171 			return NULL;
1172 
1173 		rt = rtcache_init(ro);
1174 	}
1175 	return rt;
1176 }
1177 
1178 void
1179 in6_pcbrtentry_unref(struct rtentry *rt, struct in6pcb *in6p)
1180 {
1181 
1182 	rtcache_unref(rt, &in6p->in6p_route);
1183 }
1184 
1185 struct in6pcb *
1186 in6_pcblookup_connect(struct inpcbtable *table, const struct in6_addr *faddr6,
1187 		      u_int fport_arg, const struct in6_addr *laddr6, u_int lport_arg,
1188 		      int faith,
1189 		      struct vestigial_inpcb *vp)
1190 {
1191 	struct inpcbhead *head;
1192 	struct inpcb_hdr *inph;
1193 	struct in6pcb *in6p;
1194 	u_int16_t fport = fport_arg, lport = lport_arg;
1195 
1196 	if (vp)
1197 		vp->valid = 0;
1198 
1199 	head = IN6PCBHASH_CONNECT(table, faddr6, fport, laddr6, lport);
1200 	LIST_FOREACH(inph, head, inph_hash) {
1201 		in6p = (struct in6pcb *)inph;
1202 		if (in6p->in6p_af != AF_INET6)
1203 			continue;
1204 
1205 		/* find exact match on both source and dest */
1206 		if (in6p->in6p_fport != fport)
1207 			continue;
1208 		if (in6p->in6p_lport != lport)
1209 			continue;
1210 		if (IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr))
1211 			continue;
1212 		if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, faddr6))
1213 			continue;
1214 		if (IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr))
1215 			continue;
1216 		if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, laddr6))
1217 			continue;
1218 		if ((IN6_IS_ADDR_V4MAPPED(laddr6) ||
1219 		     IN6_IS_ADDR_V4MAPPED(faddr6)) &&
1220 		    (in6p->in6p_flags & IN6P_IPV6_V6ONLY))
1221 			continue;
1222 		return in6p;
1223 	}
1224 	if (vp && table->vestige) {
1225 		if ((*table->vestige->lookup6)(faddr6, fport_arg,
1226 					       laddr6, lport_arg, vp))
1227 			return NULL;
1228 	}
1229 
1230 	return NULL;
1231 }
1232 
1233 struct in6pcb *
1234 in6_pcblookup_bind(struct inpcbtable *table, const struct in6_addr *laddr6,
1235 	u_int lport_arg, int faith)
1236 {
1237 	struct inpcbhead *head;
1238 	struct inpcb_hdr *inph;
1239 	struct in6pcb *in6p;
1240 	u_int16_t lport = lport_arg;
1241 #ifdef INET
1242 	struct in6_addr zero_mapped;
1243 #endif
1244 
1245 	head = IN6PCBHASH_BIND(table, laddr6, lport);
1246 	LIST_FOREACH(inph, head, inph_hash) {
1247 		in6p = (struct in6pcb *)inph;
1248 		if (in6p->in6p_af != AF_INET6)
1249 			continue;
1250 
1251 		if (faith && (in6p->in6p_flags & IN6P_FAITH) == 0)
1252 			continue;
1253 		if (in6p->in6p_fport != 0)
1254 			continue;
1255 		if (in6p->in6p_lport != lport)
1256 			continue;
1257 		if (IN6_IS_ADDR_V4MAPPED(laddr6) &&
1258 		    (in6p->in6p_flags & IN6P_IPV6_V6ONLY) != 0)
1259 			continue;
1260 		if (IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, laddr6))
1261 			goto out;
1262 	}
1263 #ifdef INET
1264 	if (IN6_IS_ADDR_V4MAPPED(laddr6)) {
1265 		memset(&zero_mapped, 0, sizeof(zero_mapped));
1266 		zero_mapped.s6_addr16[5] = 0xffff;
1267 		head = IN6PCBHASH_BIND(table, &zero_mapped, lport);
1268 		LIST_FOREACH(inph, head, inph_hash) {
1269 			in6p = (struct in6pcb *)inph;
1270 			if (in6p->in6p_af != AF_INET6)
1271 				continue;
1272 
1273 			if (faith && (in6p->in6p_flags & IN6P_FAITH) == 0)
1274 				continue;
1275 			if (in6p->in6p_fport != 0)
1276 				continue;
1277 			if (in6p->in6p_lport != lport)
1278 				continue;
1279 			if ((in6p->in6p_flags & IN6P_IPV6_V6ONLY) != 0)
1280 				continue;
1281 			if (IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &zero_mapped))
1282 				goto out;
1283 		}
1284 	}
1285 #endif
1286 	head = IN6PCBHASH_BIND(table, &zeroin6_addr, lport);
1287 	LIST_FOREACH(inph, head, inph_hash) {
1288 		in6p = (struct in6pcb *)inph;
1289 		if (in6p->in6p_af != AF_INET6)
1290 			continue;
1291 
1292 		if (faith && (in6p->in6p_flags & IN6P_FAITH) == 0)
1293 			continue;
1294 		if (in6p->in6p_fport != 0)
1295 			continue;
1296 		if (in6p->in6p_lport != lport)
1297 			continue;
1298 		if (IN6_IS_ADDR_V4MAPPED(laddr6) &&
1299 		    (in6p->in6p_flags & IN6P_IPV6_V6ONLY) != 0)
1300 			continue;
1301 		if (IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &zeroin6_addr))
1302 			goto out;
1303 	}
1304 	return (NULL);
1305 
1306 out:
1307 	inph = &in6p->in6p_head;
1308 	if (inph != LIST_FIRST(head)) {
1309 		LIST_REMOVE(inph, inph_hash);
1310 		LIST_INSERT_HEAD(head, inph, inph_hash);
1311 	}
1312 	return in6p;
1313 }
1314 
1315 void
1316 in6_pcbstate(struct in6pcb *in6p, int state)
1317 {
1318 
1319 	if (in6p->in6p_af != AF_INET6)
1320 		return;
1321 
1322 	if (in6p->in6p_state > IN6P_ATTACHED)
1323 		LIST_REMOVE(&in6p->in6p_head, inph_hash);
1324 
1325 	switch (state) {
1326 	case IN6P_BOUND:
1327 		LIST_INSERT_HEAD(IN6PCBHASH_BIND(in6p->in6p_table,
1328 		    &in6p->in6p_laddr, in6p->in6p_lport), &in6p->in6p_head,
1329 		    inph_hash);
1330 		break;
1331 	case IN6P_CONNECTED:
1332 		LIST_INSERT_HEAD(IN6PCBHASH_CONNECT(in6p->in6p_table,
1333 		    &in6p->in6p_faddr, in6p->in6p_fport,
1334 		    &in6p->in6p_laddr, in6p->in6p_lport), &in6p->in6p_head,
1335 		    inph_hash);
1336 		break;
1337 	}
1338 
1339 	in6p->in6p_state = state;
1340 }
1341