xref: /openbsd-src/sys/netinet6/in6.c (revision 1a8dbaac879b9f3335ad7fb25429ce63ac1d6bac)
1 /*	$OpenBSD: in6.c,v 1.242 2020/08/07 18:09:16 florian Exp $	*/
2 /*	$KAME: in6.c,v 1.372 2004/06/14 08:14:21 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.c	8.2 (Berkeley) 11/15/93
62  */
63 
64 #include "carp.h"
65 
66 #include <sys/param.h>
67 #include <sys/ioctl.h>
68 #include <sys/errno.h>
69 #include <sys/malloc.h>
70 #include <sys/socket.h>
71 #include <sys/socketvar.h>
72 #include <sys/sockio.h>
73 #include <sys/mbuf.h>
74 #include <sys/systm.h>
75 #include <sys/time.h>
76 #include <sys/kernel.h>
77 #include <sys/syslog.h>
78 
79 #include <net/if.h>
80 #include <net/if_dl.h>
81 #include <net/if_types.h>
82 #include <net/route.h>
83 
84 #include <netinet/in.h>
85 #include <netinet/if_ether.h>
86 
87 #include <netinet6/in6_var.h>
88 #include <netinet/ip6.h>
89 #include <netinet6/ip6_var.h>
90 #include <netinet6/nd6.h>
91 #include <netinet6/mld6_var.h>
92 #ifdef MROUTING
93 #include <netinet6/ip6_mroute.h>
94 #endif
95 #include <netinet6/in6_ifattach.h>
96 #if NCARP > 0
97 #include <netinet/ip_carp.h>
98 #endif
99 
100 /*
101  * Definitions of some constant IP6 addresses.
102  */
103 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
104 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
105 const struct in6_addr in6addr_intfacelocal_allnodes =
106 	IN6ADDR_INTFACELOCAL_ALLNODES_INIT;
107 const struct in6_addr in6addr_linklocal_allnodes =
108 	IN6ADDR_LINKLOCAL_ALLNODES_INIT;
109 const struct in6_addr in6addr_linklocal_allrouters =
110 	IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
111 
112 const struct in6_addr in6mask0 = IN6MASK0;
113 const struct in6_addr in6mask32 = IN6MASK32;
114 const struct in6_addr in6mask64 = IN6MASK64;
115 const struct in6_addr in6mask96 = IN6MASK96;
116 const struct in6_addr in6mask128 = IN6MASK128;
117 
118 int in6_ioctl(u_long, caddr_t, struct ifnet *, int);
119 int in6_ioctl_change_ifaddr(u_long, caddr_t, struct ifnet *);
120 int in6_ioctl_get(u_long, caddr_t, struct ifnet *);
121 int in6_check_embed_scope(struct sockaddr_in6 *, unsigned int);
122 int in6_clear_scope_id(struct sockaddr_in6 *, unsigned int);
123 int in6_ifinit(struct ifnet *, struct in6_ifaddr *, int);
124 void in6_unlink_ifa(struct in6_ifaddr *, struct ifnet *);
125 
126 const struct sockaddr_in6 sa6_any = {
127 	sizeof(sa6_any), AF_INET6, 0, 0, IN6ADDR_ANY_INIT, 0
128 };
129 
130 int
131 in6_mask2len(struct in6_addr *mask, u_char *lim0)
132 {
133 	int x = 0, y;
134 	u_char *lim = lim0, *p;
135 
136 	/* ignore the scope_id part */
137 	if (lim0 == NULL || lim0 - (u_char *)mask > sizeof(*mask))
138 		lim = (u_char *)mask + sizeof(*mask);
139 	for (p = (u_char *)mask; p < lim; x++, p++) {
140 		if (*p != 0xff)
141 			break;
142 	}
143 	y = 0;
144 	if (p < lim) {
145 		for (y = 0; y < 8; y++) {
146 			if ((*p & (0x80 >> y)) == 0)
147 				break;
148 		}
149 	}
150 
151 	/*
152 	 * when the limit pointer is given, do a stricter check on the
153 	 * remaining bits.
154 	 */
155 	if (p < lim) {
156 		if (y != 0 && (*p & (0x00ff >> y)) != 0)
157 			return (-1);
158 		for (p = p + 1; p < lim; p++)
159 			if (*p != 0)
160 				return (-1);
161 	}
162 
163 	return x * 8 + y;
164 }
165 
166 int
167 in6_nam2sin6(const struct mbuf *nam, struct sockaddr_in6 **sin6)
168 {
169 	struct sockaddr *sa = mtod(nam, struct sockaddr *);
170 
171 	if (nam->m_len < offsetof(struct sockaddr, sa_data))
172 		return EINVAL;
173 	if (sa->sa_family != AF_INET6)
174 		return EAFNOSUPPORT;
175 	if (sa->sa_len != nam->m_len)
176 		return EINVAL;
177 	if (sa->sa_len != sizeof(struct sockaddr_in6))
178 		return EINVAL;
179 	*sin6 = satosin6(sa);
180 
181 	return 0;
182 }
183 
184 int
185 in6_sa2sin6(struct sockaddr *sa, struct sockaddr_in6 **sin6)
186 {
187 	if (sa->sa_family != AF_INET6)
188 		return EAFNOSUPPORT;
189 	if (sa->sa_len != sizeof(struct sockaddr_in6))
190 		return EINVAL;
191 	*sin6 = satosin6(sa);
192 
193 	return 0;
194 }
195 
196 int
197 in6_control(struct socket *so, u_long cmd, caddr_t data, struct ifnet *ifp)
198 {
199 	int privileged;
200 	int error;
201 
202 	privileged = 0;
203 	if ((so->so_state & SS_PRIV) != 0)
204 		privileged++;
205 
206 	switch (cmd) {
207 #ifdef MROUTING
208 	case SIOCGETSGCNT_IN6:
209 	case SIOCGETMIFCNT_IN6:
210 		error = mrt6_ioctl(so, cmd, data);
211 		break;
212 #endif /* MROUTING */
213 	default:
214 		error = in6_ioctl(cmd, data, ifp, privileged);
215 		break;
216 	}
217 
218 	return error;
219 }
220 
221 int
222 in6_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp, int privileged)
223 {
224 	if (ifp == NULL)
225 		return (ENXIO);
226 
227 	switch (cmd) {
228 	case SIOCGIFINFO_IN6:
229 	case SIOCGNBRINFO_IN6:
230 		return (nd6_ioctl(cmd, data, ifp));
231 	case SIOCGIFDSTADDR_IN6:
232 	case SIOCGIFNETMASK_IN6:
233 	case SIOCGIFAFLAG_IN6:
234 	case SIOCGIFALIFETIME_IN6:
235 		return (in6_ioctl_get(cmd, data, ifp));
236 	case SIOCAIFADDR_IN6:
237 	case SIOCDIFADDR_IN6:
238 		if (!privileged)
239 			return (EPERM);
240 		return (in6_ioctl_change_ifaddr(cmd, data, ifp));
241 	case SIOCSIFADDR:
242 	case SIOCSIFDSTADDR:
243 	case SIOCSIFBRDADDR:
244 	case SIOCSIFNETMASK:
245 		/*
246 		 * Do not pass those ioctl to driver handler since they are not
247 		 * properly set up. Instead just error out.
248 		 */
249 		return (EINVAL);
250 	default:
251 		return (EOPNOTSUPP);
252 	}
253 }
254 
255 int
256 in6_ioctl_change_ifaddr(u_long cmd, caddr_t data, struct ifnet *ifp)
257 {
258 	struct	in6_ifaddr *ia6 = NULL;
259 	struct	in6_aliasreq *ifra = (struct in6_aliasreq *)data;
260 	struct	sockaddr *sa;
261 	struct	sockaddr_in6 *sa6 = NULL;
262 	int	error = 0, newifaddr = 0, plen;
263 
264 	/*
265 	 * Find address for this interface, if it exists.
266 	 *
267 	 * In netinet code, we have checked ifra_addr in SIOCSIF*ADDR operation
268 	 * only, and used the first interface address as the target of other
269 	 * operations (without checking ifra_addr).  This was because netinet
270 	 * code/API assumed at most 1 interface address per interface.
271 	 * Since IPv6 allows a node to assign multiple addresses
272 	 * on a single interface, we almost always look and check the
273 	 * presence of ifra_addr, and reject invalid ones here.
274 	 * It also decreases duplicated code among SIOC*_IN6 operations.
275 	 *
276 	 * We always require users to specify a valid IPv6 address for
277 	 * the corresponding operation.
278 	 */
279 	switch (cmd) {
280 	case SIOCAIFADDR_IN6:
281 		sa = sin6tosa(&ifra->ifra_addr);
282 		break;
283 	case SIOCDIFADDR_IN6:
284 		sa = sin6tosa(&((struct in6_ifreq *)data)->ifr_addr);
285 		break;
286 	default:
287 		panic("%s: invalid ioctl %lu", __func__, cmd);
288 	}
289 	if (sa->sa_family == AF_INET6) {
290 		error = in6_sa2sin6(sa, &sa6);
291 		if (error)
292 			return (error);
293 	}
294 
295 	NET_LOCK();
296 
297 	if (sa6 != NULL) {
298 		error = in6_check_embed_scope(sa6, ifp->if_index);
299 		if (error)
300 			goto err;
301 		error = in6_clear_scope_id(sa6, ifp->if_index);
302 		if (error)
303 			goto err;
304 		ia6 = in6ifa_ifpwithaddr(ifp, &sa6->sin6_addr);
305 	}
306 
307 	switch (cmd) {
308 	case SIOCDIFADDR_IN6:
309 		/*
310 		 * for IPv4, we look for existing in_ifaddr here to allow
311 		 * "ifconfig if0 delete" to remove the first IPv4 address on
312 		 * the interface.  For IPv6, as the spec allows multiple
313 		 * interface address from the day one, we consider "remove the
314 		 * first one" semantics to be not preferable.
315 		 */
316 		if (ia6 == NULL) {
317 			error = EADDRNOTAVAIL;
318 			break;
319 		}
320 		in6_purgeaddr(&ia6->ia_ifa);
321 		if_addrhooks_run(ifp);
322 		break;
323 
324 	case SIOCAIFADDR_IN6:
325 		if (ifra->ifra_addr.sin6_family != AF_INET6 ||
326 		    ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6)) {
327 			error = EAFNOSUPPORT;
328 			break;
329 		}
330 
331 		/* reject read-only flags */
332 		if ((ifra->ifra_flags & IN6_IFF_DUPLICATED) != 0 ||
333 		    (ifra->ifra_flags & IN6_IFF_DETACHED) != 0 ||
334 		    (ifra->ifra_flags & IN6_IFF_DEPRECATED) != 0) {
335 			error = EINVAL;
336 			break;
337 		}
338 
339 		if (ia6 == NULL)
340 			newifaddr = 1;
341 
342 		/*
343 		 * Make the address tentative before joining multicast
344 		 * addresses, so that corresponding MLD responses would
345 		 * not have a tentative source address.
346 		 */
347 		if (newifaddr && in6if_do_dad(ifp))
348 			ifra->ifra_flags |= IN6_IFF_TENTATIVE;
349 
350 		/*
351 		 * first, make or update the interface address structure,
352 		 * and link it to the list. try to enable inet6 if there
353 		 * is no link-local yet.
354 		 */
355 		error = in6_ifattach(ifp);
356 		if (error)
357 			break;
358 		error = in6_update_ifa(ifp, ifra, ia6);
359 		if (error)
360 			break;
361 
362 		ia6 = NULL;
363 		if (sa6 != NULL)
364 			ia6 = in6ifa_ifpwithaddr(ifp, &sa6->sin6_addr);
365 		if (ia6 == NULL) {
366 			/*
367 			 * this can happen when the user specify the 0 valid
368 			 * lifetime.
369 			 */
370 			break;
371 		}
372 
373 		/* Perform DAD, if needed. */
374 		if (ia6->ia6_flags & IN6_IFF_TENTATIVE)
375 			nd6_dad_start(&ia6->ia_ifa);
376 
377 		if (!newifaddr) {
378 			if_addrhooks_run(ifp);
379 			break;
380 		}
381 
382 		plen = in6_mask2len(&ia6->ia_prefixmask.sin6_addr, NULL);
383 		if ((ifp->if_flags & IFF_LOOPBACK) || plen == 128) {
384 			if_addrhooks_run(ifp);
385 			break;	/* No need to install a connected route. */
386 		}
387 
388 		error = rt_ifa_add(&ia6->ia_ifa,
389 		    RTF_CLONING | RTF_CONNECTED | RTF_MPATH,
390 		    ia6->ia_ifa.ifa_addr, ifp->if_rdomain);
391 		if (error) {
392 			in6_purgeaddr(&ia6->ia_ifa);
393 			break;
394 		}
395 		if_addrhooks_run(ifp);
396 		break;
397 	}
398 
399 err:
400 	NET_UNLOCK();
401 	return (error);
402 }
403 
404 int
405 in6_ioctl_get(u_long cmd, caddr_t data, struct ifnet *ifp)
406 {
407 	struct	in6_ifreq *ifr = (struct in6_ifreq *)data;
408 	struct	in6_ifaddr *ia6 = NULL;
409 	struct	sockaddr *sa;
410 	struct	sockaddr_in6 *sa6 = NULL;
411 	int	error = 0;
412 
413 	sa = sin6tosa(&ifr->ifr_addr);
414 	if (sa->sa_family == AF_INET6) {
415 		sa->sa_len = sizeof(struct sockaddr_in6);
416 		error = in6_sa2sin6(sa, &sa6);
417 		if (error)
418 			return (error);
419 	}
420 
421 	NET_RLOCK_IN_IOCTL();
422 
423 	if (sa6 != NULL) {
424 		error = in6_check_embed_scope(sa6, ifp->if_index);
425 		if (error)
426 			goto err;
427 		error = in6_clear_scope_id(sa6, ifp->if_index);
428 		if (error)
429 			goto err;
430 		ia6 = in6ifa_ifpwithaddr(ifp, &sa6->sin6_addr);
431 	}
432 
433 	/* must think again about its semantics */
434 	if (ia6 == NULL) {
435 		error = EADDRNOTAVAIL;
436 		goto err;
437 	}
438 
439 	switch (cmd) {
440 	case SIOCGIFDSTADDR_IN6:
441 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
442 			error = EINVAL;
443 			break;
444 		}
445 		/*
446 		 * XXX: should we check if ifa_dstaddr is NULL and return
447 		 * an error?
448 		 */
449 		ifr->ifr_dstaddr = ia6->ia_dstaddr;
450 		break;
451 
452 	case SIOCGIFNETMASK_IN6:
453 		ifr->ifr_addr = ia6->ia_prefixmask;
454 		break;
455 
456 	case SIOCGIFAFLAG_IN6:
457 		ifr->ifr_ifru.ifru_flags6 = ia6->ia6_flags;
458 		break;
459 
460 	case SIOCGIFALIFETIME_IN6:
461 		ifr->ifr_ifru.ifru_lifetime = ia6->ia6_lifetime;
462 		if (ia6->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
463 			time_t expire, maxexpire;
464 			struct in6_addrlifetime *retlt =
465 			    &ifr->ifr_ifru.ifru_lifetime;
466 
467 			/*
468 			 * XXX: adjust expiration time assuming time_t is
469 			 * signed.
470 			 */
471 			maxexpire =
472 			    (time_t)~(1ULL << ((sizeof(maxexpire) * 8) - 1));
473 			if (ia6->ia6_lifetime.ia6t_vltime <
474 			    maxexpire - ia6->ia6_updatetime) {
475 				expire = ia6->ia6_updatetime +
476 				    ia6->ia6_lifetime.ia6t_vltime;
477 				if (expire != 0) {
478 					expire -= getuptime();
479 					expire += gettime();
480 				}
481 				retlt->ia6t_expire = expire;
482 			} else
483 				retlt->ia6t_expire = maxexpire;
484 		}
485 		if (ia6->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
486 			time_t expire, maxexpire;
487 			struct in6_addrlifetime *retlt =
488 			    &ifr->ifr_ifru.ifru_lifetime;
489 
490 			/*
491 			 * XXX: adjust expiration time assuming time_t is
492 			 * signed.
493 			 */
494 			maxexpire =
495 			    (time_t)~(1ULL << ((sizeof(maxexpire) * 8) - 1));
496 			if (ia6->ia6_lifetime.ia6t_pltime <
497 			    maxexpire - ia6->ia6_updatetime) {
498 				expire = ia6->ia6_updatetime +
499 				    ia6->ia6_lifetime.ia6t_pltime;
500 				if (expire != 0) {
501 					expire -= getuptime();
502 					expire += gettime();
503 				}
504 				retlt->ia6t_preferred = expire;
505 			} else
506 				retlt->ia6t_preferred = maxexpire;
507 		}
508 		break;
509 
510 	default:
511 		panic("%s: invalid ioctl %lu", __func__, cmd);
512 	}
513 
514 err:
515 	NET_RUNLOCK_IN_IOCTL();
516 	return (error);
517 }
518 
519 int
520 in6_check_embed_scope(struct sockaddr_in6 *sa6, unsigned int ifidx)
521 {
522 	if (IN6_IS_ADDR_LINKLOCAL(&sa6->sin6_addr)) {
523 		if (sa6->sin6_addr.s6_addr16[1] == 0) {
524 			/* link ID is not embedded by the user */
525 			sa6->sin6_addr.s6_addr16[1] = htons(ifidx);
526 		} else if (sa6->sin6_addr.s6_addr16[1] != htons(ifidx))
527 			return EINVAL;	/* link ID contradicts */
528 	}
529 	return 0;
530 }
531 
532 int
533 in6_clear_scope_id(struct sockaddr_in6 *sa6, unsigned int ifidx)
534 {
535 	if (IN6_IS_ADDR_LINKLOCAL(&sa6->sin6_addr)) {
536 		if (sa6->sin6_scope_id) {
537 			if (sa6->sin6_scope_id != (u_int32_t)ifidx)
538 				return EINVAL;
539 			sa6->sin6_scope_id = 0; /* XXX: good way? */
540 		}
541 	}
542 	return 0;
543 }
544 
545 /*
546  * Update parameters of an IPv6 interface address.
547  * If necessary, a new entry is created and linked into address chains.
548  * This function is separated from in6_control().
549  */
550 int
551 in6_update_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
552     struct in6_ifaddr *ia6)
553 {
554 	int error = 0, hostIsNew = 0, plen = -1;
555 	struct sockaddr_in6 dst6;
556 	struct in6_addrlifetime *lt;
557 	struct in6_multi_mship *imm;
558 	struct rtentry *rt;
559 	char addr[INET6_ADDRSTRLEN];
560 
561 	NET_ASSERT_LOCKED();
562 
563 	/* Validate parameters */
564 	if (ifp == NULL || ifra == NULL) /* this maybe redundant */
565 		return (EINVAL);
566 
567 	/*
568 	 * The destination address for a p2p link must have a family
569 	 * of AF_UNSPEC or AF_INET6.
570 	 */
571 	if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
572 	    ifra->ifra_dstaddr.sin6_family != AF_INET6 &&
573 	    ifra->ifra_dstaddr.sin6_family != AF_UNSPEC)
574 		return (EAFNOSUPPORT);
575 
576 	/*
577 	 * validate ifra_prefixmask.  don't check sin6_family, netmask
578 	 * does not carry fields other than sin6_len.
579 	 */
580 	if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6))
581 		return (EINVAL);
582 	/*
583 	 * Because the IPv6 address architecture is classless, we require
584 	 * users to specify a (non 0) prefix length (mask) for a new address.
585 	 * We also require the prefix (when specified) mask is valid, and thus
586 	 * reject a non-consecutive mask.
587 	 */
588 	if (ia6 == NULL && ifra->ifra_prefixmask.sin6_len == 0)
589 		return (EINVAL);
590 	if (ifra->ifra_prefixmask.sin6_len != 0) {
591 		plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
592 		    (u_char *)&ifra->ifra_prefixmask +
593 		    ifra->ifra_prefixmask.sin6_len);
594 		if (plen <= 0)
595 			return (EINVAL);
596 	} else {
597 		/*
598 		 * In this case, ia6 must not be NULL.  We just use its prefix
599 		 * length.
600 		 */
601 		plen = in6_mask2len(&ia6->ia_prefixmask.sin6_addr, NULL);
602 	}
603 	/*
604 	 * If the destination address on a p2p interface is specified,
605 	 * and the address is a scoped one, validate/set the scope
606 	 * zone identifier.
607 	 */
608 	dst6 = ifra->ifra_dstaddr;
609 	if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) != 0 &&
610 	    (dst6.sin6_family == AF_INET6)) {
611 		error = in6_check_embed_scope(&dst6, ifp->if_index);
612 		if (error)
613 			return error;
614 	}
615 	/*
616 	 * The destination address can be specified only for a p2p or a
617 	 * loopback interface.  If specified, the corresponding prefix length
618 	 * must be 128.
619 	 */
620 	if (ifra->ifra_dstaddr.sin6_family == AF_INET6) {
621 		if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) == 0)
622 			return (EINVAL);
623 		if (plen != 128)
624 			return (EINVAL);
625 	}
626 	/* lifetime consistency check */
627 	lt = &ifra->ifra_lifetime;
628 	if (lt->ia6t_pltime > lt->ia6t_vltime)
629 		return (EINVAL);
630 	if (lt->ia6t_vltime == 0) {
631 		/*
632 		 * the following log might be noisy, but this is a typical
633 		 * configuration mistake or a tool's bug.
634 		 */
635 		nd6log((LOG_INFO, "%s: valid lifetime is 0 for %s\n", __func__,
636 		    inet_ntop(AF_INET6, &ifra->ifra_addr.sin6_addr,
637 		    addr, sizeof(addr))));
638 
639 		if (ia6 == NULL)
640 			return (0); /* there's nothing to do */
641 	}
642 
643 	/*
644 	 * If this is a new address, allocate a new ifaddr and link it
645 	 * into chains.
646 	 */
647 	if (ia6 == NULL) {
648 		hostIsNew = 1;
649 		ia6 = malloc(sizeof(*ia6), M_IFADDR, M_WAITOK | M_ZERO);
650 		LIST_INIT(&ia6->ia6_memberships);
651 		/* Initialize the address and masks, and put time stamp */
652 		ia6->ia_ifa.ifa_addr = sin6tosa(&ia6->ia_addr);
653 		ia6->ia_addr.sin6_family = AF_INET6;
654 		ia6->ia_addr.sin6_len = sizeof(ia6->ia_addr);
655 		ia6->ia6_updatetime = getuptime();
656 		if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
657 			/*
658 			 * XXX: some functions expect that ifa_dstaddr is not
659 			 * NULL for p2p interfaces.
660 			 */
661 			ia6->ia_ifa.ifa_dstaddr = sin6tosa(&ia6->ia_dstaddr);
662 		} else {
663 			ia6->ia_ifa.ifa_dstaddr = NULL;
664 		}
665 		ia6->ia_ifa.ifa_netmask = sin6tosa(&ia6->ia_prefixmask);
666 
667 		ia6->ia_ifp = ifp;
668 		ia6->ia_addr = ifra->ifra_addr;
669 		ifa_add(ifp, &ia6->ia_ifa);
670 	}
671 
672 	/* set prefix mask */
673 	if (ifra->ifra_prefixmask.sin6_len) {
674 		/*
675 		 * We prohibit changing the prefix length of an existing
676 		 * address, because
677 		 * + such an operation should be rare in IPv6, and
678 		 * + the operation would confuse prefix management.
679 		 */
680 		if (ia6->ia_prefixmask.sin6_len &&
681 		    in6_mask2len(&ia6->ia_prefixmask.sin6_addr, NULL) != plen) {
682 			error = EINVAL;
683 			goto unlink;
684 		}
685 		ia6->ia_prefixmask = ifra->ifra_prefixmask;
686 	}
687 
688 	/*
689 	 * If a new destination address is specified, scrub the old one and
690 	 * install the new destination.  Note that the interface must be
691 	 * p2p or loopback (see the check above.)
692 	 */
693 	if ((ifp->if_flags & IFF_POINTOPOINT) && dst6.sin6_family == AF_INET6 &&
694 	    !IN6_ARE_ADDR_EQUAL(&dst6.sin6_addr, &ia6->ia_dstaddr.sin6_addr)) {
695 		struct ifaddr *ifa = &ia6->ia_ifa;
696 
697 		if ((ia6->ia_flags & IFA_ROUTE) != 0 &&
698 		    rt_ifa_del(ifa, RTF_HOST, ifa->ifa_dstaddr,
699 		     ifp->if_rdomain) != 0) {
700 			nd6log((LOG_ERR, "%s: failed to remove a route "
701 			    "to the old destination: %s\n", __func__,
702 			    inet_ntop(AF_INET6, &ia6->ia_addr.sin6_addr,
703 			    addr, sizeof(addr))));
704 			/* proceed anyway... */
705 		} else
706 			ia6->ia_flags &= ~IFA_ROUTE;
707 		ia6->ia_dstaddr = dst6;
708 	}
709 
710 	/*
711 	 * Set lifetimes.  We do not refer to ia6t_expire and ia6t_preferred
712 	 * to see if the address is deprecated or invalidated, but initialize
713 	 * these members for applications.
714 	 */
715 	ia6->ia6_updatetime = getuptime();
716 	ia6->ia6_lifetime = ifra->ifra_lifetime;
717 	if (ia6->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
718 		ia6->ia6_lifetime.ia6t_expire =
719 		    getuptime() + ia6->ia6_lifetime.ia6t_vltime;
720 	} else
721 		ia6->ia6_lifetime.ia6t_expire = 0;
722 	if (ia6->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
723 		ia6->ia6_lifetime.ia6t_preferred =
724 		    getuptime() + ia6->ia6_lifetime.ia6t_pltime;
725 	} else
726 		ia6->ia6_lifetime.ia6t_preferred = 0;
727 
728 	/* reset the interface and routing table appropriately. */
729 	if ((error = in6_ifinit(ifp, ia6, hostIsNew)) != 0)
730 		goto unlink;
731 
732 	/* re-run DAD */
733 	if (ia6->ia6_flags & (IN6_IFF_TENTATIVE|IN6_IFF_DUPLICATED))
734 		ifra->ifra_flags |= IN6_IFF_TENTATIVE;
735 	/*
736 	 * configure address flags.
737 	 */
738 	ia6->ia6_flags = ifra->ifra_flags;
739 
740 	nd6_expire_timer_update(ia6);
741 
742 	/*
743 	 * We are done if we have simply modified an existing address.
744 	 */
745 	if (!hostIsNew)
746 		return (error);
747 
748 	/*
749 	 * Beyond this point, we should call in6_purgeaddr upon an error,
750 	 * not just go to unlink.
751 	 */
752 
753 	/* join necessary multiast groups */
754 	if ((ifp->if_flags & IFF_MULTICAST) != 0) {
755 		struct sockaddr_in6 mltaddr, mltmask;
756 
757 		/* join solicited multicast addr for new host id */
758 		struct sockaddr_in6 llsol;
759 
760 		bzero(&llsol, sizeof(llsol));
761 		llsol.sin6_family = AF_INET6;
762 		llsol.sin6_len = sizeof(llsol);
763 		llsol.sin6_addr.s6_addr16[0] = htons(0xff02);
764 		llsol.sin6_addr.s6_addr16[1] = htons(ifp->if_index);
765 		llsol.sin6_addr.s6_addr32[1] = 0;
766 		llsol.sin6_addr.s6_addr32[2] = htonl(1);
767 		llsol.sin6_addr.s6_addr32[3] =
768 		    ifra->ifra_addr.sin6_addr.s6_addr32[3];
769 		llsol.sin6_addr.s6_addr8[12] = 0xff;
770 		imm = in6_joingroup(ifp, &llsol.sin6_addr, &error);
771 		if (!imm)
772 			goto cleanup;
773 		LIST_INSERT_HEAD(&ia6->ia6_memberships, imm, i6mm_chain);
774 
775 		bzero(&mltmask, sizeof(mltmask));
776 		mltmask.sin6_len = sizeof(struct sockaddr_in6);
777 		mltmask.sin6_family = AF_INET6;
778 		mltmask.sin6_addr = in6mask32;
779 
780 		/*
781 		 * join link-local all-nodes address
782 		 */
783 		bzero(&mltaddr, sizeof(mltaddr));
784 		mltaddr.sin6_len = sizeof(struct sockaddr_in6);
785 		mltaddr.sin6_family = AF_INET6;
786 		mltaddr.sin6_addr = in6addr_linklocal_allnodes;
787 		mltaddr.sin6_addr.s6_addr16[1] = htons(ifp->if_index);
788 		mltaddr.sin6_scope_id = 0;
789 
790 		/*
791 		 * XXX: do we really need this automatic routes?
792 		 * We should probably reconsider this stuff.  Most applications
793 		 * actually do not need the routes, since they usually specify
794 		 * the outgoing interface.
795 		 */
796 		rt = rtalloc(sin6tosa(&mltaddr), 0, ifp->if_rdomain);
797 		if (rt) {
798 			/* 32bit came from "mltmask" */
799 			if (memcmp(&mltaddr.sin6_addr,
800 			    &satosin6(rt_key(rt))->sin6_addr,
801 			    32 / 8)) {
802 				rtfree(rt);
803 				rt = NULL;
804 			}
805 		}
806 		if (!rt) {
807 			struct rt_addrinfo info;
808 
809 			bzero(&info, sizeof(info));
810 			info.rti_ifa = &ia6->ia_ifa;
811 			info.rti_info[RTAX_DST] = sin6tosa(&mltaddr);
812 			info.rti_info[RTAX_GATEWAY] = sin6tosa(&ia6->ia_addr);
813 			info.rti_info[RTAX_NETMASK] = sin6tosa(&mltmask);
814 			info.rti_info[RTAX_IFA] = sin6tosa(&ia6->ia_addr);
815 			info.rti_flags = RTF_MULTICAST;
816 			error = rtrequest(RTM_ADD, &info, RTP_CONNECTED, NULL,
817 			    ifp->if_rdomain);
818 			if (error)
819 				goto cleanup;
820 		} else {
821 			rtfree(rt);
822 		}
823 		imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error);
824 		if (!imm)
825 			goto cleanup;
826 		LIST_INSERT_HEAD(&ia6->ia6_memberships, imm, i6mm_chain);
827 
828 		/*
829 		 * join interface-local all-nodes address.
830 		 * (ff01::1%ifN, and ff01::%ifN/32)
831 		 */
832 		bzero(&mltaddr, sizeof(mltaddr));
833 		mltaddr.sin6_len = sizeof(struct sockaddr_in6);
834 		mltaddr.sin6_family = AF_INET6;
835 		mltaddr.sin6_addr = in6addr_intfacelocal_allnodes;
836 		mltaddr.sin6_addr.s6_addr16[1] = htons(ifp->if_index);
837 		mltaddr.sin6_scope_id = 0;
838 
839 		/* XXX: again, do we really need the route? */
840 		rt = rtalloc(sin6tosa(&mltaddr), 0, ifp->if_rdomain);
841 		if (rt) {
842 			/* 32bit came from "mltmask" */
843 			if (memcmp(&mltaddr.sin6_addr,
844 			    &satosin6(rt_key(rt))->sin6_addr,
845 			    32 / 8)) {
846 				rtfree(rt);
847 				rt = NULL;
848 			}
849 		}
850 		if (!rt) {
851 			struct rt_addrinfo info;
852 
853 			bzero(&info, sizeof(info));
854 			info.rti_ifa = &ia6->ia_ifa;
855 			info.rti_info[RTAX_DST] = sin6tosa(&mltaddr);
856 			info.rti_info[RTAX_GATEWAY] = sin6tosa(&ia6->ia_addr);
857 			info.rti_info[RTAX_NETMASK] = sin6tosa(&mltmask);
858 			info.rti_info[RTAX_IFA] = sin6tosa(&ia6->ia_addr);
859 			info.rti_flags = RTF_MULTICAST;
860 			error = rtrequest(RTM_ADD, &info, RTP_CONNECTED, NULL,
861 			    ifp->if_rdomain);
862 			if (error)
863 				goto cleanup;
864 		} else {
865 			rtfree(rt);
866 		}
867 		imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error);
868 		if (!imm)
869 			goto cleanup;
870 		LIST_INSERT_HEAD(&ia6->ia6_memberships, imm, i6mm_chain);
871 	}
872 
873 	return (error);
874 
875   unlink:
876 	/*
877 	 * XXX: if a change of an existing address failed, keep the entry
878 	 * anyway.
879 	 */
880 	if (hostIsNew)
881 		in6_unlink_ifa(ia6, ifp);
882 	return (error);
883 
884   cleanup:
885 	in6_purgeaddr(&ia6->ia_ifa);
886 	return error;
887 }
888 
889 void
890 in6_purgeaddr(struct ifaddr *ifa)
891 {
892 	struct ifnet *ifp = ifa->ifa_ifp;
893 	struct in6_ifaddr *ia6 = ifatoia6(ifa);
894 	struct in6_multi_mship *imm;
895 
896 	/* stop DAD processing */
897 	nd6_dad_stop(ifa);
898 
899 	/*
900 	 * delete route to the destination of the address being purged.
901 	 * The interface must be p2p or loopback in this case.
902 	 */
903 	if ((ifp->if_flags & IFF_POINTOPOINT) && (ia6->ia_flags & IFA_ROUTE) &&
904 	    ia6->ia_dstaddr.sin6_len != 0) {
905 		int e;
906 
907 		e = rt_ifa_del(ifa, RTF_HOST, ifa->ifa_dstaddr,
908 		    ifp->if_rdomain);
909 		if (e != 0) {
910 			char addr[INET6_ADDRSTRLEN];
911 			log(LOG_ERR, "in6_purgeaddr: failed to remove "
912 			    "a route to the p2p destination: %s on %s, "
913 			    "errno=%d\n",
914 			    inet_ntop(AF_INET6, &ia6->ia_addr.sin6_addr,
915 				addr, sizeof(addr)),
916 			    ifp->if_xname, e);
917 			/* proceed anyway... */
918 		} else
919 			ia6->ia_flags &= ~IFA_ROUTE;
920 	}
921 
922 	/* Remove ownaddr's loopback rtentry, if it exists. */
923 	rt_ifa_dellocal(&(ia6->ia_ifa));
924 
925 	/*
926 	 * leave from multicast groups we have joined for the interface
927 	 */
928 	while (!LIST_EMPTY(&ia6->ia6_memberships)) {
929 		imm = LIST_FIRST(&ia6->ia6_memberships);
930 		LIST_REMOVE(imm, i6mm_chain);
931 		in6_leavegroup(imm);
932 	}
933 
934 	in6_unlink_ifa(ia6, ifp);
935 }
936 
937 void
938 in6_unlink_ifa(struct in6_ifaddr *ia6, struct ifnet *ifp)
939 {
940 	struct ifaddr *ifa = &ia6->ia_ifa;
941 	extern int ifatrash;
942 	int plen;
943 
944 	NET_ASSERT_LOCKED();
945 
946 	/* Release the reference to the base prefix. */
947 	plen = in6_mask2len(&ia6->ia_prefixmask.sin6_addr, NULL);
948 	if ((ifp->if_flags & IFF_LOOPBACK) == 0 && plen != 128) {
949 		rt_ifa_del(ifa, RTF_CLONING | RTF_CONNECTED,
950 		    ifa->ifa_addr, ifp->if_rdomain);
951 	}
952 
953 	rt_ifa_purge(ifa);
954 	ifa_del(ifp, ifa);
955 
956 	ifatrash++;
957 	ia6->ia_ifp = NULL;
958 	ifafree(&ia6->ia_ifa);
959 }
960 
961 /*
962  * Initialize an interface's intetnet6 address
963  * and routing table entry.
964  */
965 int
966 in6_ifinit(struct ifnet *ifp, struct in6_ifaddr *ia6, int newhost)
967 {
968 	int	error = 0, plen, ifacount = 0;
969 	struct ifaddr *ifa;
970 
971 	NET_ASSERT_LOCKED();
972 
973 	/*
974 	 * Give the interface a chance to initialize
975 	 * if this is its first address (or it is a CARP interface)
976 	 * and to validate the address if necessary.
977 	 */
978 	TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
979 		if (ifa->ifa_addr->sa_family != AF_INET6)
980 			continue;
981 		ifacount++;
982 	}
983 
984 	if ((ifacount <= 1 || ifp->if_type == IFT_CARP ||
985 	    (ifp->if_flags & (IFF_LOOPBACK|IFF_POINTOPOINT))) &&
986 	    (error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia6))) {
987 		return (error);
988 	}
989 
990 	ia6->ia_ifa.ifa_metric = ifp->if_metric;
991 
992 	/* we could do in(6)_socktrim here, but just omit it at this moment. */
993 
994 	/*
995 	 * Special case:
996 	 * If the destination address is specified for a point-to-point
997 	 * interface, install a route to the destination as an interface
998 	 * direct route.
999 	 */
1000 	plen = in6_mask2len(&ia6->ia_prefixmask.sin6_addr, NULL); /* XXX */
1001 	if ((ifp->if_flags & IFF_POINTOPOINT) && plen == 128 &&
1002 	    ia6->ia_dstaddr.sin6_family == AF_INET6) {
1003 		ifa = &ia6->ia_ifa;
1004 		error = rt_ifa_add(ifa, RTF_HOST | RTF_MPATH,
1005 		    ifa->ifa_dstaddr, ifp->if_rdomain);
1006 		if (error != 0)
1007 			return (error);
1008 		ia6->ia_flags |= IFA_ROUTE;
1009 	}
1010 
1011 	if (newhost)
1012 		error = rt_ifa_addlocal(&(ia6->ia_ifa));
1013 
1014 	return (error);
1015 }
1016 
1017 /*
1018  * Add an address to the list of IP6 multicast addresses for a
1019  * given interface.
1020  */
1021 struct in6_multi *
1022 in6_addmulti(struct in6_addr *maddr6, struct ifnet *ifp, int *errorp)
1023 {
1024 	struct	in6_ifreq ifr;
1025 	struct	in6_multi *in6m;
1026 
1027 	NET_ASSERT_LOCKED();
1028 
1029 	*errorp = 0;
1030 	/*
1031 	 * See if address already in list.
1032 	 */
1033 	IN6_LOOKUP_MULTI(*maddr6, ifp, in6m);
1034 	if (in6m != NULL) {
1035 		/*
1036 		 * Found it; just increment the refrence count.
1037 		 */
1038 		in6m->in6m_refcnt++;
1039 	} else {
1040 		/*
1041 		 * New address; allocate a new multicast record
1042 		 * and link it into the interface's multicast list.
1043 		 */
1044 		in6m = malloc(sizeof(*in6m), M_IPMADDR, M_NOWAIT | M_ZERO);
1045 		if (in6m == NULL) {
1046 			*errorp = ENOBUFS;
1047 			return (NULL);
1048 		}
1049 
1050 		in6m->in6m_sin.sin6_len = sizeof(struct sockaddr_in6);
1051 		in6m->in6m_sin.sin6_family = AF_INET6;
1052 		in6m->in6m_sin.sin6_addr = *maddr6;
1053 		in6m->in6m_refcnt = 1;
1054 		in6m->in6m_ifidx = ifp->if_index;
1055 		in6m->in6m_ifma.ifma_addr = sin6tosa(&in6m->in6m_sin);
1056 
1057 		/*
1058 		 * Ask the network driver to update its multicast reception
1059 		 * filter appropriately for the new address.
1060 		 */
1061 		memcpy(&ifr.ifr_addr, &in6m->in6m_sin, sizeof(in6m->in6m_sin));
1062 		*errorp = (*ifp->if_ioctl)(ifp, SIOCADDMULTI, (caddr_t)&ifr);
1063 		if (*errorp) {
1064 			free(in6m, M_IPMADDR, sizeof(*in6m));
1065 			return (NULL);
1066 		}
1067 
1068 		TAILQ_INSERT_HEAD(&ifp->if_maddrlist, &in6m->in6m_ifma,
1069 		    ifma_list);
1070 
1071 		/*
1072 		 * Let MLD6 know that we have joined a new IP6 multicast
1073 		 * group.
1074 		 */
1075 		mld6_start_listening(in6m);
1076 	}
1077 
1078 	return (in6m);
1079 }
1080 
1081 /*
1082  * Delete a multicast address record.
1083  */
1084 void
1085 in6_delmulti(struct in6_multi *in6m)
1086 {
1087 	struct	in6_ifreq ifr;
1088 	struct	ifnet *ifp;
1089 
1090 	NET_ASSERT_LOCKED();
1091 
1092 	if (--in6m->in6m_refcnt == 0) {
1093 		/*
1094 		 * No remaining claims to this record; let MLD6 know
1095 		 * that we are leaving the multicast group.
1096 		 */
1097 		mld6_stop_listening(in6m);
1098 		ifp = if_get(in6m->in6m_ifidx);
1099 
1100 		/*
1101 		 * Notify the network driver to update its multicast
1102 		 * reception filter.
1103 		 */
1104 		if (ifp != NULL) {
1105 			bzero(&ifr.ifr_addr, sizeof(struct sockaddr_in6));
1106 			ifr.ifr_addr.sin6_len = sizeof(struct sockaddr_in6);
1107 			ifr.ifr_addr.sin6_family = AF_INET6;
1108 			ifr.ifr_addr.sin6_addr = in6m->in6m_addr;
1109 			KERNEL_LOCK();
1110 			(*ifp->if_ioctl)(ifp, SIOCDELMULTI, (caddr_t)&ifr);
1111 			KERNEL_UNLOCK();
1112 
1113 			TAILQ_REMOVE(&ifp->if_maddrlist, &in6m->in6m_ifma,
1114 			    ifma_list);
1115 		}
1116 		if_put(ifp);
1117 
1118 		free(in6m, M_IPMADDR, sizeof(*in6m));
1119 	}
1120 }
1121 
1122 /*
1123  * Return 1 if the multicast group represented by ``maddr6'' has been
1124  * joined by interface ``ifp'', 0 otherwise.
1125  */
1126 int
1127 in6_hasmulti(struct in6_addr *maddr6, struct ifnet *ifp)
1128 {
1129 	struct in6_multi *in6m;
1130 	int joined;
1131 
1132 	IN6_LOOKUP_MULTI(*maddr6, ifp, in6m);
1133 	joined = (in6m != NULL);
1134 
1135 	return (joined);
1136 }
1137 
1138 struct in6_multi_mship *
1139 in6_joingroup(struct ifnet *ifp, struct in6_addr *addr, int *errorp)
1140 {
1141 	struct in6_multi_mship *imm;
1142 
1143 	imm = malloc(sizeof(*imm), M_IPMADDR, M_NOWAIT);
1144 	if (!imm) {
1145 		*errorp = ENOBUFS;
1146 		return NULL;
1147 	}
1148 	imm->i6mm_maddr = in6_addmulti(addr, ifp, errorp);
1149 	if (!imm->i6mm_maddr) {
1150 		/* *errorp is alrady set */
1151 		free(imm, M_IPMADDR, sizeof(*imm));
1152 		return NULL;
1153 	}
1154 	return imm;
1155 }
1156 
1157 void
1158 in6_leavegroup(struct in6_multi_mship *imm)
1159 {
1160 
1161 	if (imm->i6mm_maddr)
1162 		in6_delmulti(imm->i6mm_maddr);
1163 	free(imm,  M_IPMADDR, sizeof(*imm));
1164 }
1165 
1166 /*
1167  * Find an IPv6 interface link-local address specific to an interface.
1168  */
1169 struct in6_ifaddr *
1170 in6ifa_ifpforlinklocal(struct ifnet *ifp, int ignoreflags)
1171 {
1172 	struct ifaddr *ifa;
1173 
1174 	TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1175 		if (ifa->ifa_addr->sa_family != AF_INET6)
1176 			continue;
1177 		if (IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa))) {
1178 			if ((ifatoia6(ifa)->ia6_flags & ignoreflags) != 0)
1179 				continue;
1180 			break;
1181 		}
1182 	}
1183 
1184 	return (ifatoia6(ifa));
1185 }
1186 
1187 
1188 /*
1189  * find the internet address corresponding to a given interface and address.
1190  */
1191 struct in6_ifaddr *
1192 in6ifa_ifpwithaddr(struct ifnet *ifp, struct in6_addr *addr)
1193 {
1194 	struct ifaddr *ifa;
1195 
1196 	TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1197 		if (ifa->ifa_addr->sa_family != AF_INET6)
1198 			continue;
1199 		if (IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa)))
1200 			break;
1201 	}
1202 
1203 	return (ifatoia6(ifa));
1204 }
1205 
1206 /*
1207  * Get a scope of the address. Node-local, link-local, site-local or global.
1208  */
1209 int
1210 in6_addrscope(struct in6_addr *addr)
1211 {
1212 	int scope;
1213 
1214 	if (addr->s6_addr8[0] == 0xfe) {
1215 		scope = addr->s6_addr8[1] & 0xc0;
1216 
1217 		switch (scope) {
1218 		case 0x80:
1219 			return __IPV6_ADDR_SCOPE_LINKLOCAL;
1220 			break;
1221 		case 0xc0:
1222 			return __IPV6_ADDR_SCOPE_SITELOCAL;
1223 			break;
1224 		default:
1225 			return __IPV6_ADDR_SCOPE_GLOBAL; /* just in case */
1226 			break;
1227 		}
1228 	}
1229 
1230 
1231 	if (addr->s6_addr8[0] == 0xff) {
1232 		scope = addr->s6_addr8[1] & 0x0f;
1233 
1234 		/*
1235 		 * due to other scope such as reserved,
1236 		 * return scope doesn't work.
1237 		 */
1238 		switch (scope) {
1239 		case __IPV6_ADDR_SCOPE_INTFACELOCAL:
1240 			return __IPV6_ADDR_SCOPE_INTFACELOCAL;
1241 			break;
1242 		case __IPV6_ADDR_SCOPE_LINKLOCAL:
1243 			return __IPV6_ADDR_SCOPE_LINKLOCAL;
1244 			break;
1245 		case __IPV6_ADDR_SCOPE_SITELOCAL:
1246 			return __IPV6_ADDR_SCOPE_SITELOCAL;
1247 			break;
1248 		default:
1249 			return __IPV6_ADDR_SCOPE_GLOBAL;
1250 			break;
1251 		}
1252 	}
1253 
1254 	if (bcmp(&in6addr_loopback, addr, sizeof(*addr) - 1) == 0) {
1255 		if (addr->s6_addr8[15] == 1) /* loopback */
1256 			return __IPV6_ADDR_SCOPE_INTFACELOCAL;
1257 		if (addr->s6_addr8[15] == 0) /* unspecified */
1258 			return __IPV6_ADDR_SCOPE_LINKLOCAL;
1259 	}
1260 
1261 	return __IPV6_ADDR_SCOPE_GLOBAL;
1262 }
1263 
1264 int
1265 in6_addr2scopeid(unsigned int ifidx, struct in6_addr *addr)
1266 {
1267 	int scope = in6_addrscope(addr);
1268 
1269 	switch(scope) {
1270 	case __IPV6_ADDR_SCOPE_INTFACELOCAL:
1271 	case __IPV6_ADDR_SCOPE_LINKLOCAL:
1272 		/* XXX: we do not distinguish between a link and an I/F. */
1273 		return (ifidx);
1274 
1275 	case __IPV6_ADDR_SCOPE_SITELOCAL:
1276 		return (0);	/* XXX: invalid. */
1277 
1278 	default:
1279 		return (0);	/* XXX: treat as global. */
1280 	}
1281 }
1282 
1283 /*
1284  * return length of part which dst and src are equal
1285  * hard coding...
1286  */
1287 int
1288 in6_matchlen(struct in6_addr *src, struct in6_addr *dst)
1289 {
1290 	int match = 0;
1291 	u_char *s = (u_char *)src, *d = (u_char *)dst;
1292 	u_char *lim = s + 16, r;
1293 
1294 	while (s < lim)
1295 		if ((r = (*d++ ^ *s++)) != 0) {
1296 			while (r < 128) {
1297 				match++;
1298 				r <<= 1;
1299 			}
1300 			break;
1301 		} else
1302 			match += 8;
1303 	return match;
1304 }
1305 
1306 void
1307 in6_prefixlen2mask(struct in6_addr *maskp, int len)
1308 {
1309 	u_char maskarray[8] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
1310 	int bytelen, bitlen, i;
1311 
1312 	/* sanity check */
1313 	if (0 > len || len > 128) {
1314 		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
1315 		    len);
1316 		return;
1317 	}
1318 
1319 	bzero(maskp, sizeof(*maskp));
1320 	bytelen = len / 8;
1321 	bitlen = len % 8;
1322 	for (i = 0; i < bytelen; i++)
1323 		maskp->s6_addr[i] = 0xff;
1324 	/* len == 128 is ok because bitlen == 0 then */
1325 	if (bitlen)
1326 		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
1327 }
1328 
1329 /*
1330  * return the best address out of the same scope
1331  */
1332 struct in6_ifaddr *
1333 in6_ifawithscope(struct ifnet *oifp, struct in6_addr *dst, u_int rdomain)
1334 {
1335 	int dst_scope =	in6_addrscope(dst), src_scope, best_scope = 0;
1336 	int blen = -1;
1337 	struct ifaddr *ifa;
1338 	struct ifnet *ifp;
1339 	struct in6_ifaddr *ia6_best = NULL;
1340 
1341 	if (oifp == NULL) {
1342 		printf("%s: output interface is not specified\n", __func__);
1343 		return (NULL);
1344 	}
1345 
1346 	/* We search for all addresses on all interfaces from the beginning. */
1347 	TAILQ_FOREACH(ifp, &ifnet, if_list) {
1348 		if (ifp->if_rdomain != rdomain)
1349 			continue;
1350 #if NCARP > 0
1351 		/*
1352 		 * Never use a carp address of an interface which is not
1353 		 * the master.
1354 		 */
1355 		if (ifp->if_type == IFT_CARP && !carp_iamatch(ifp))
1356 			continue;
1357 #endif
1358 
1359 		/*
1360 		 * We can never take an address that breaks the scope zone
1361 		 * of the destination.
1362 		 */
1363 		if (in6_addr2scopeid(ifp->if_index, dst) !=
1364 		    in6_addr2scopeid(oifp->if_index, dst))
1365 			continue;
1366 
1367 		TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1368 			int tlen = -1;
1369 
1370 			if (ifa->ifa_addr->sa_family != AF_INET6)
1371 				continue;
1372 
1373 			src_scope = in6_addrscope(IFA_IN6(ifa));
1374 
1375 			/*
1376 			 * Don't use an address before completing DAD
1377 			 * nor a duplicated address.
1378 			 */
1379 			if (ifatoia6(ifa)->ia6_flags &
1380 			    (IN6_IFF_TENTATIVE|IN6_IFF_DUPLICATED))
1381 				continue;
1382 
1383 			/*
1384 			 * RFC 6724 allows anycast addresses as source address
1385 			 * because the restriction was removed in RFC 4291.
1386 			 * However RFC 4443 states that ICMPv6 responses
1387 			 * MUST use a unicast source address.
1388 			 *
1389 			 * XXX Skip anycast addresses for now since
1390 			 * icmp6_reflect() uses this function for source
1391 			 * address selection.
1392 			 */
1393 			if (ifatoia6(ifa)->ia6_flags & IN6_IFF_ANYCAST)
1394 				continue;
1395 
1396 			if (ifatoia6(ifa)->ia6_flags & IN6_IFF_DETACHED)
1397 				continue;
1398 
1399 			/*
1400 			 * If this is the first address we find,
1401 			 * keep it anyway.
1402 			 */
1403 			if (ia6_best == NULL)
1404 				goto replace;
1405 
1406 			/*
1407 			 * ia6_best is never NULL beyond this line except
1408 			 * within the block labeled "replace".
1409 			 */
1410 
1411 			/*
1412 			 * Rule 2: Prefer appropriate scope.
1413 			 * Find the address with the smallest scope that is
1414 			 * bigger (or equal) to the scope of the destination
1415 			 * address.
1416 			 * Accept an address with smaller scope than the
1417 			 * destination if non exists with bigger scope.
1418 			 */
1419 			if (best_scope < src_scope) {
1420 				if (best_scope < dst_scope)
1421 					goto replace;
1422 				else
1423 					continue;
1424 			} else if (src_scope < best_scope) {
1425 				if (src_scope < dst_scope)
1426 					continue;
1427 				else
1428 					goto replace;
1429 			}
1430 
1431 			/* Rule 3: Avoid deprecated addresses. */
1432 			if (ifatoia6(ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
1433 				/*
1434 				 * Ignore any deprecated addresses if
1435 				 * specified by configuration.
1436 				 */
1437 				if (!ip6_use_deprecated)
1438 					continue;
1439 
1440 				/*
1441 				 * If we have already found a non-deprecated
1442 				 * candidate, just ignore deprecated addresses.
1443 				 */
1444 				if ((ia6_best->ia6_flags & IN6_IFF_DEPRECATED)
1445 				    == 0)
1446 					continue;
1447 			} else if ((ia6_best->ia6_flags & IN6_IFF_DEPRECATED))
1448 				goto replace;
1449 
1450 			/*
1451 			 * Rule 4: Prefer home addresses.
1452 			 * We do not support home addresses.
1453 			 */
1454 
1455 			/* Rule 5: Prefer outgoing interface */
1456 			if (ia6_best->ia_ifp == oifp && ifp != oifp)
1457 				continue;
1458 			if (ia6_best->ia_ifp != oifp && ifp == oifp)
1459 				goto replace;
1460 
1461 			/*
1462 			 * Rule 5.5: Prefer addresses in a prefix advertised
1463 			 * by the next-hop.
1464 			 * We do not track this information.
1465 			 */
1466 
1467 			/*
1468 			 * Rule 6: Prefer matching label.
1469 			 * We do not implement policy tables.
1470 			 */
1471 
1472 			/* Rule 7: Prefer temporary addresses. */
1473 			if ((ia6_best->ia6_flags & IN6_IFF_TEMPORARY) &&
1474 			    !(ifatoia6(ifa)->ia6_flags & IN6_IFF_TEMPORARY))
1475 				continue;
1476 			if (!(ia6_best->ia6_flags & IN6_IFF_TEMPORARY) &&
1477 			    (ifatoia6(ifa)->ia6_flags & IN6_IFF_TEMPORARY))
1478 				goto replace;
1479 
1480 			/* Rule 8: Use longest matching prefix. */
1481 			tlen = in6_matchlen(IFA_IN6(ifa), dst);
1482 			if (tlen > blen) {
1483 #if NCARP > 0
1484 				/*
1485 				 * Don't let carp interfaces win a tie against
1486 				 * the output interface based on matchlen.
1487 				 * We should only use a carp address if no
1488 				 * other interface has a usable address.
1489 				 * Otherwise, when communicating from a carp
1490 				 * master to a carp backup, the backup system
1491 				 * won't respond since the carp address is also
1492 				 * configured as a local address on the backup.
1493 				 * Note that carp interfaces in backup state
1494 				 * were already skipped above.
1495 				 */
1496 				if (ifp->if_type == IFT_CARP &&
1497 				    oifp->if_type != IFT_CARP)
1498 					continue;
1499 #endif
1500 				goto replace;
1501 			} else if (tlen < blen)
1502 				continue;
1503 
1504 			/*
1505 			 * If the eight rules fail to choose a single address,
1506 			 * the tiebreaker is implementation-specific.
1507 			 */
1508 
1509 			 /* Prefer address with highest pltime. */
1510 			if (ia6_best->ia6_updatetime +
1511 			    ia6_best->ia6_lifetime.ia6t_pltime <
1512 			    ifatoia6(ifa)->ia6_updatetime +
1513 			    ifatoia6(ifa)->ia6_lifetime.ia6t_pltime)
1514 				goto replace;
1515 			else if (ia6_best->ia6_updatetime +
1516 			    ia6_best->ia6_lifetime.ia6t_pltime >
1517 			    ifatoia6(ifa)->ia6_updatetime +
1518 			    ifatoia6(ifa)->ia6_lifetime.ia6t_pltime)
1519 				continue;
1520 
1521 			/* Prefer address with highest vltime. */
1522 			if (ia6_best->ia6_updatetime +
1523 			    ia6_best->ia6_lifetime.ia6t_vltime <
1524 			    ifatoia6(ifa)->ia6_updatetime +
1525 			    ifatoia6(ifa)->ia6_lifetime.ia6t_vltime)
1526 				goto replace;
1527 			else if (ia6_best->ia6_updatetime +
1528 			    ia6_best->ia6_lifetime.ia6t_vltime >
1529 			    ifatoia6(ifa)->ia6_updatetime +
1530 			    ifatoia6(ifa)->ia6_lifetime.ia6t_vltime)
1531 				continue;
1532 
1533 			continue;
1534 		  replace:
1535 			ia6_best = ifatoia6(ifa);
1536 			blen = tlen >= 0 ? tlen :
1537 			    in6_matchlen(IFA_IN6(ifa), dst);
1538 			best_scope =
1539 			    in6_addrscope(&ia6_best->ia_addr.sin6_addr);
1540 		}
1541 	}
1542 
1543 	/* count statistics for future improvements */
1544 	if (ia6_best == NULL)
1545 		ip6stat_inc(ip6s_sources_none);
1546 	else {
1547 		if (oifp == ia6_best->ia_ifp)
1548 			ip6stat_inc(ip6s_sources_sameif + best_scope);
1549 		else
1550 			ip6stat_inc(ip6s_sources_otherif + best_scope);
1551 
1552 		if (best_scope == dst_scope)
1553 			ip6stat_inc(ip6s_sources_samescope + best_scope);
1554 		else
1555 			ip6stat_inc(ip6s_sources_otherscope + best_scope);
1556 
1557 		if ((ia6_best->ia6_flags & IN6_IFF_DEPRECATED) != 0)
1558 			ip6stat_inc(ip6s_sources_deprecated + best_scope);
1559 	}
1560 
1561 	return (ia6_best);
1562 }
1563 
1564 int
1565 in6if_do_dad(struct ifnet *ifp)
1566 {
1567 	if ((ifp->if_flags & IFF_LOOPBACK) != 0)
1568 		return (0);
1569 
1570 	switch (ifp->if_type) {
1571 #if NCARP > 0
1572 	case IFT_CARP:
1573 		/*
1574 		 * XXX: DAD does not work currently on carp(4)
1575 		 * so disable it for now.
1576 		 */
1577 		return (0);
1578 #endif
1579 	default:
1580 		/*
1581 		 * Our DAD routine requires the interface up and running.
1582 		 * However, some interfaces can be up before the RUNNING
1583 		 * status.  Additionaly, users may try to assign addresses
1584 		 * before the interface becomes up (or running).
1585 		 * We simply skip DAD in such a case as a work around.
1586 		 * XXX: we should rather mark "tentative" on such addresses,
1587 		 * and do DAD after the interface becomes ready.
1588 		 */
1589 		if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) !=
1590 		    (IFF_UP|IFF_RUNNING))
1591 			return (0);
1592 
1593 		return (1);
1594 	}
1595 }
1596 
1597 void *
1598 in6_domifattach(struct ifnet *ifp)
1599 {
1600 	struct in6_ifextra *ext;
1601 
1602 	ext = malloc(sizeof(*ext), M_IFADDR, M_WAITOK | M_ZERO);
1603 
1604 	ext->nd_ifinfo = nd6_ifattach(ifp);
1605 	ext->nprefixes = 0;
1606 	ext->ndefrouters = 0;
1607 	return ext;
1608 }
1609 
1610 void
1611 in6_domifdetach(struct ifnet *ifp, void *aux)
1612 {
1613 	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
1614 
1615 	nd6_ifdetach(ext->nd_ifinfo);
1616 	free(ext, M_IFADDR, sizeof(*ext));
1617 }
1618