xref: /dragonfly/sys/netinet6/in6.c (revision f6061ce2)
1 /*	$FreeBSD: src/sys/netinet6/in6.c,v 1.7.2.9 2002/04/28 05:40:26 suz Exp $	*/
2 /*	$DragonFly: src/sys/netinet6/in6.c,v 1.30 2008/10/03 07:59:20 hasso Exp $	*/
3 /*	$KAME: in6.c,v 1.259 2002/01/21 11:37:50 keiichi Exp $	*/
4 
5 /*
6  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
7  * All rights reserved.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  * 3. Neither the name of the project nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  */
33 
34 /*
35  * Copyright (c) 1982, 1986, 1991, 1993
36  *	The Regents of the University of California.  All rights reserved.
37  *
38  * Redistribution and use in source and binary forms, with or without
39  * modification, are permitted provided that the following conditions
40  * are met:
41  * 1. Redistributions of source code must retain the above copyright
42  *    notice, this list of conditions and the following disclaimer.
43  * 2. Redistributions in binary form must reproduce the above copyright
44  *    notice, this list of conditions and the following disclaimer in the
45  *    documentation and/or other materials provided with the distribution.
46  * 3. All advertising materials mentioning features or use of this software
47  *    must display the following acknowledgement:
48  *	This product includes software developed by the University of
49  *	California, Berkeley and its contributors.
50  * 4. Neither the name of the University nor the names of its contributors
51  *    may be used to endorse or promote products derived from this software
52  *    without specific prior written permission.
53  *
54  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
55  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
56  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
57  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
58  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
59  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
60  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
61  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
62  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
63  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
64  * SUCH DAMAGE.
65  *
66  *	@(#)in.c	8.2 (Berkeley) 11/15/93
67  */
68 
69 #include "opt_inet.h"
70 #include "opt_inet6.h"
71 
72 #include <sys/param.h>
73 #include <sys/errno.h>
74 #include <sys/malloc.h>
75 #include <sys/socket.h>
76 #include <sys/socketvar.h>
77 #include <sys/sockio.h>
78 #include <sys/systm.h>
79 #include <sys/proc.h>
80 #include <sys/priv.h>
81 #include <sys/time.h>
82 #include <sys/kernel.h>
83 #include <sys/syslog.h>
84 
85 #include <sys/thread2.h>
86 #include <sys/msgport2.h>
87 
88 #include <net/if.h>
89 #include <net/if_types.h>
90 #include <net/route.h>
91 #include <net/if_dl.h>
92 
93 #include <netinet/in.h>
94 #include <netinet/in_var.h>
95 #include <netinet/if_ether.h>
96 #include <netinet/in_systm.h>
97 #include <netinet/ip.h>
98 #include <netinet/in_pcb.h>
99 
100 #include <netinet/ip6.h>
101 #include <netinet6/ip6_var.h>
102 #include <netinet6/nd6.h>
103 #include <netinet6/mld6_var.h>
104 #include <netinet6/ip6_mroute.h>
105 #include <netinet6/in6_ifattach.h>
106 #include <netinet6/scope6_var.h>
107 #include <netinet6/in6_pcb.h>
108 #include <netinet6/in6_var.h>
109 
110 #include <net/net_osdep.h>
111 
112 /*
113  * Definitions of some costant IP6 addresses.
114  */
115 const struct in6_addr kin6addr_any = IN6ADDR_ANY_INIT;
116 const struct in6_addr kin6addr_loopback = IN6ADDR_LOOPBACK_INIT;
117 const struct in6_addr kin6addr_nodelocal_allnodes =
118 	IN6ADDR_NODELOCAL_ALLNODES_INIT;
119 const struct in6_addr kin6addr_linklocal_allnodes =
120 	IN6ADDR_LINKLOCAL_ALLNODES_INIT;
121 const struct in6_addr kin6addr_linklocal_allrouters =
122 	IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
123 
124 const struct in6_addr in6mask0 = IN6MASK0;
125 const struct in6_addr in6mask32 = IN6MASK32;
126 const struct in6_addr in6mask64 = IN6MASK64;
127 const struct in6_addr in6mask96 = IN6MASK96;
128 const struct in6_addr in6mask128 = IN6MASK128;
129 
130 const struct sockaddr_in6 sa6_any = {sizeof(sa6_any), AF_INET6,
131 				     0, 0, IN6ADDR_ANY_INIT, 0};
132 
133 static int in6_lifaddr_ioctl (struct socket *, u_long, caddr_t,
134 	struct ifnet *, struct thread *);
135 static int in6_ifinit (struct ifnet *, struct in6_ifaddr *,
136 			   struct sockaddr_in6 *, int);
137 static void in6_unlink_ifa (struct in6_ifaddr *, struct ifnet *);
138 static void in6_ifloop_request_callback(int, int, struct rt_addrinfo *, struct rtentry *, void *);
139 
140 struct in6_multihead in6_multihead;	/* XXX BSS initialization */
141 
142 int	(*faithprefix_p)(struct in6_addr *);
143 
144 /*
145  * Subroutine for in6_ifaddloop() and in6_ifremloop().
146  * This routine does actual work.
147  */
148 static void
149 in6_ifloop_request(int cmd, struct ifaddr *ifa)
150 {
151 	struct sockaddr_in6 all1_sa;
152         struct rt_addrinfo rtinfo;
153 	int error;
154 
155 	bzero(&all1_sa, sizeof(all1_sa));
156 	all1_sa.sin6_family = AF_INET6;
157 	all1_sa.sin6_len = sizeof(struct sockaddr_in6);
158 	all1_sa.sin6_addr = in6mask128;
159 
160 	/*
161 	 * We specify the address itself as the gateway, and set the
162 	 * RTF_LLINFO flag, so that the corresponding host route would have
163 	 * the flag, and thus applications that assume traditional behavior
164 	 * would be happy.  Note that we assume the caller of the function
165 	 * (probably implicitly) set nd6_rtrequest() to ifa->ifa_rtrequest,
166 	 * which changes the outgoing interface to the loopback interface.
167 	 */
168 	bzero(&rtinfo, sizeof(struct rt_addrinfo));
169 	rtinfo.rti_info[RTAX_DST] = ifa->ifa_addr;
170 	rtinfo.rti_info[RTAX_GATEWAY] = ifa->ifa_addr;
171 	rtinfo.rti_info[RTAX_NETMASK] = (struct sockaddr *)&all1_sa;
172 	rtinfo.rti_flags = RTF_UP|RTF_HOST|RTF_LLINFO;
173 
174 	error = rtrequest1_global(cmd, &rtinfo,
175 				  in6_ifloop_request_callback, ifa);
176 	if (error != 0) {
177 		log(LOG_ERR, "in6_ifloop_request: "
178 		    "%s operation failed for %s (errno=%d)\n",
179 		    cmd == RTM_ADD ? "ADD" : "DELETE",
180 		    ip6_sprintf(&((struct in6_ifaddr *)ifa)->ia_addr.sin6_addr),
181 		    error);
182 	}
183 }
184 
185 static void
186 in6_ifloop_request_callback(int cmd, int error, struct rt_addrinfo *rtinfo,
187 			    struct rtentry *rt, void *arg)
188 {
189 	struct ifaddr *ifa = arg;
190 
191 	if (error)
192 		goto done;
193 
194 	/*
195 	 * Make sure rt_ifa be equal to IFA, the second argument of the
196 	 * function.
197 	 * We need this because when we refer to rt_ifa->ia6_flags in
198 	 * ip6_input, we assume that the rt_ifa points to the address instead
199 	 * of the loopback address.
200 	 */
201 	if (cmd == RTM_ADD && rt && ifa != rt->rt_ifa) {
202 		++rt->rt_refcnt;
203 		IFAFREE(rt->rt_ifa);
204 		IFAREF(ifa);
205 		rt->rt_ifa = ifa;
206 		--rt->rt_refcnt;
207 	}
208 
209 	/*
210 	 * Report the addition/removal of the address to the routing socket.
211 	 * XXX: since we called rtinit for a p2p interface with a destination,
212 	 *      we end up reporting twice in such a case.  Should we rather
213 	 *      omit the second report?
214 	 */
215 	if (rt) {
216 		if (mycpuid == 0)
217 			rt_newaddrmsg(cmd, ifa, error, rt);
218 		if (cmd == RTM_DELETE) {
219 			if (rt->rt_refcnt == 0) {
220 				++rt->rt_refcnt;
221 				rtfree(rt);
222 			}
223 		}
224 	}
225 done:
226 	/* no way to return any new error */
227 	;
228 }
229 
230 /*
231  * Add ownaddr as loopback rtentry.  We previously add the route only if
232  * necessary (ex. on a p2p link).  However, since we now manage addresses
233  * separately from prefixes, we should always add the route.  We can't
234  * rely on the cloning mechanism from the corresponding interface route
235  * any more.
236  */
237 void
238 in6_ifaddloop(struct ifaddr *ifa)
239 {
240 	struct rtentry *rt;
241 
242 	/* If there is no loopback entry, allocate one. */
243 	rt = rtpurelookup(ifa->ifa_addr);
244 	if (rt == NULL || !(rt->rt_flags & RTF_HOST) ||
245 	    !(rt->rt_ifp->if_flags & IFF_LOOPBACK))
246 		in6_ifloop_request(RTM_ADD, ifa);
247 	if (rt != NULL)
248 		rt->rt_refcnt--;
249 }
250 
251 /*
252  * Remove loopback rtentry of ownaddr generated by in6_ifaddloop(),
253  * if it exists.
254  */
255 void
256 in6_ifremloop(struct ifaddr *ifa)
257 {
258 	struct in6_ifaddr *ia;
259 	struct rtentry *rt;
260 	int ia_count = 0;
261 
262 	/*
263 	 * Some of BSD variants do not remove cloned routes
264 	 * from an interface direct route, when removing the direct route
265 	 * (see comments in net/net_osdep.h).  Even for variants that do remove
266 	 * cloned routes, they could fail to remove the cloned routes when
267 	 * we handle multple addresses that share a common prefix.
268 	 * So, we should remove the route corresponding to the deleted address
269 	 * regardless of the result of in6_is_ifloop_auto().
270 	 */
271 
272 	/*
273 	 * Delete the entry only if exact one ifa exists.  More than one ifa
274 	 * can exist if we assign a same single address to multiple
275 	 * (probably p2p) interfaces.
276 	 * XXX: we should avoid such a configuration in IPv6...
277 	 */
278 	for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
279 		if (IN6_ARE_ADDR_EQUAL(IFA_IN6(ifa), &ia->ia_addr.sin6_addr)) {
280 			ia_count++;
281 			if (ia_count > 1)
282 				break;
283 		}
284 	}
285 
286 	if (ia_count == 1) {
287 		/*
288 		 * Before deleting, check if a corresponding loopbacked host
289 		 * route surely exists.  With this check, we can avoid to
290 		 * delete an interface direct route whose destination is same
291 		 * as the address being removed.  This can happen when remofing
292 		 * a subnet-router anycast address on an interface attahced
293 		 * to a shared medium.
294 		 */
295 		rt = rtpurelookup(ifa->ifa_addr);
296 		if (rt != NULL && (rt->rt_flags & RTF_HOST) &&
297 		    (rt->rt_ifp->if_flags & IFF_LOOPBACK)) {
298 			rt->rt_refcnt--;
299 			in6_ifloop_request(RTM_DELETE, ifa);
300 		}
301 	}
302 }
303 
304 int
305 in6_ifindex2scopeid(int idx)
306 {
307 	struct ifnet *ifp;
308 	struct sockaddr_in6 *sin6;
309 	struct ifaddr_container *ifac;
310 
311 	if (idx < 0 || if_index < idx)
312 		return -1;
313 	ifp = ifindex2ifnet[idx];
314 
315 	TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link)
316 	{
317 		struct ifaddr *ifa = ifac->ifa;
318 
319 		if (ifa->ifa_addr->sa_family != AF_INET6)
320 			continue;
321 		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
322 		if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr))
323 			return sin6->sin6_scope_id & 0xffff;
324 	}
325 
326 	return -1;
327 }
328 
329 int
330 in6_mask2len(struct in6_addr *mask, u_char *lim0)
331 {
332 	int x = 0, y;
333 	u_char *lim = lim0, *p;
334 
335 	if (lim0 == NULL ||
336 	    lim0 - (u_char *)mask > sizeof(*mask)) /* ignore the scope_id part */
337 		lim = (u_char *)mask + sizeof(*mask);
338 	for (p = (u_char *)mask; p < lim; x++, p++) {
339 		if (*p != 0xff)
340 			break;
341 	}
342 	y = 0;
343 	if (p < lim) {
344 		for (y = 0; y < 8; y++) {
345 			if ((*p & (0x80 >> y)) == 0)
346 				break;
347 		}
348 	}
349 
350 	/*
351 	 * when the limit pointer is given, do a stricter check on the
352 	 * remaining bits.
353 	 */
354 	if (p < lim) {
355 		if (y != 0 && (*p & (0x00ff >> y)) != 0)
356 			return (-1);
357 		for (p = p + 1; p < lim; p++)
358 			if (*p != 0)
359 				return (-1);
360 	}
361 
362 	return x * 8 + y;
363 }
364 
365 void
366 in6_len2mask(struct in6_addr *mask, int len)
367 {
368 	int i;
369 
370 	bzero(mask, sizeof(*mask));
371 	for (i = 0; i < len / 8; i++)
372 		mask->s6_addr8[i] = 0xff;
373 	if (len % 8)
374 		mask->s6_addr8[i] = (0xff00 >> (len % 8)) & 0xff;
375 }
376 
377 #define ifa2ia6(ifa)	((struct in6_ifaddr *)(ifa))
378 #define ia62ifa(ia6)	(&((ia6)->ia_ifa))
379 
380 void
381 in6_control_dispatch(netmsg_t msg)
382 {
383 	int error;
384 
385 	error = in6_control(msg->control.base.nm_so,
386 			    msg->control.nm_cmd,
387 			    msg->control.nm_data,
388 			    msg->control.nm_ifp,
389 			    msg->control.nm_td);
390 	lwkt_replymsg(&msg->control.base.lmsg, error);
391 }
392 
393 int
394 in6_control(struct socket *so, u_long cmd, caddr_t data,
395 	    struct ifnet *ifp, struct thread *td)
396 {
397 	struct	in6_ifreq *ifr = (struct in6_ifreq *)data;
398 	struct	in6_ifaddr *ia = NULL;
399 	struct	in6_aliasreq *ifra = (struct in6_aliasreq *)data;
400 	int privileged;
401 	int error;
402 
403 	privileged = 0;
404 	if (priv_check(td, PRIV_ROOT) == 0)
405 		privileged++;
406 
407 	switch (cmd) {
408 	case SIOCGETSGCNT_IN6:
409 	case SIOCGETMIFCNT_IN6:
410 		return (mrt6_ioctl(cmd, data));
411 	}
412 
413 	switch(cmd) {
414 	case SIOCAADDRCTL_POLICY:
415 	case SIOCDADDRCTL_POLICY:
416 		if (!privileged)
417 			return (EPERM);
418 		return (in6_src_ioctl(cmd, data));
419 	}
420 
421 	if (ifp == NULL)
422 		return (EOPNOTSUPP);
423 
424 	switch (cmd) {
425 	case SIOCSNDFLUSH_IN6:
426 	case SIOCSPFXFLUSH_IN6:
427 	case SIOCSRTRFLUSH_IN6:
428 	case SIOCSDEFIFACE_IN6:
429 	case SIOCSIFINFO_FLAGS:
430 		if (!privileged)
431 			return (EPERM);
432 		/* fall through */
433 	case OSIOCGIFINFO_IN6:
434 	case SIOCGIFINFO_IN6:
435 	case SIOCGDRLST_IN6:
436 	case SIOCGPRLST_IN6:
437 	case SIOCGNBRINFO_IN6:
438 	case SIOCGDEFIFACE_IN6:
439 		return (nd6_ioctl(cmd, data, ifp));
440 	}
441 
442 	switch (cmd) {
443 	case SIOCSIFPREFIX_IN6:
444 	case SIOCDIFPREFIX_IN6:
445 	case SIOCAIFPREFIX_IN6:
446 	case SIOCCIFPREFIX_IN6:
447 	case SIOCSGIFPREFIX_IN6:
448 	case SIOCGIFPREFIX_IN6:
449 		log(LOG_NOTICE,
450 		    "prefix ioctls are now invalidated. "
451 		    "please use ifconfig.\n");
452 		return (EOPNOTSUPP);
453 	}
454 
455 	switch (cmd) {
456 	case SIOCSSCOPE6:
457 		if (!privileged)
458 			return (EPERM);
459 		return (scope6_set(ifp,
460 			(struct scope6_id *)ifr->ifr_ifru.ifru_scope_id));
461 		break;
462 	case SIOCGSCOPE6:
463 		return (scope6_get(ifp,
464 			(struct scope6_id *)ifr->ifr_ifru.ifru_scope_id));
465 		break;
466 	case SIOCGSCOPE6DEF:
467 		return (scope6_get_default((struct scope6_id *)
468 			ifr->ifr_ifru.ifru_scope_id));
469 		break;
470 	}
471 
472 	switch (cmd) {
473 	case SIOCALIFADDR:
474 	case SIOCDLIFADDR:
475 		if (!privileged)
476 			return (EPERM);
477 		/* fall through */
478 	case SIOCGLIFADDR:
479 		return in6_lifaddr_ioctl(so, cmd, data, ifp, td);
480 	}
481 
482 	/*
483 	 * Find address for this interface, if it exists.
484 	 */
485 	if (ifra->ifra_addr.sin6_family == AF_INET6) { /* XXX */
486 		struct sockaddr_in6 *sa6 =
487 			(struct sockaddr_in6 *)&ifra->ifra_addr;
488 
489 		if (IN6_IS_ADDR_LINKLOCAL(&sa6->sin6_addr)) {
490 			if (sa6->sin6_addr.s6_addr16[1] == 0) {
491 				/* link ID is not embedded by the user */
492 				sa6->sin6_addr.s6_addr16[1] =
493 					htons(ifp->if_index);
494 			} else if (sa6->sin6_addr.s6_addr16[1] !=
495 				    htons(ifp->if_index)) {
496 				return (EINVAL);	/* link ID contradicts */
497 			}
498 			if (sa6->sin6_scope_id) {
499 				if (sa6->sin6_scope_id !=
500 				    (u_int32_t)ifp->if_index)
501 					return (EINVAL);
502 				sa6->sin6_scope_id = 0; /* XXX: good way? */
503 			}
504 		}
505 		ia = in6ifa_ifpwithaddr(ifp, &ifra->ifra_addr.sin6_addr);
506 	}
507 
508 	switch (cmd) {
509 	case SIOCSIFADDR_IN6:
510 	case SIOCSIFDSTADDR_IN6:
511 	case SIOCSIFNETMASK_IN6:
512 		/*
513 		 * Since IPv6 allows a node to assign multiple addresses
514 		 * on a single interface, SIOCSIFxxx ioctls are not suitable
515 		 * and should be unused.
516 		 */
517 		/* we decided to obsolete this command (20000704) */
518 		return (EINVAL);
519 
520 	case SIOCDIFADDR_IN6:
521 		/*
522 		 * for IPv4, we look for existing in_ifaddr here to allow
523 		 * "ifconfig if0 delete" to remove first IPv4 address on the
524 		 * interface.  For IPv6, as the spec allow multiple interface
525 		 * address from the day one, we consider "remove the first one"
526 		 * semantics to be not preferable.
527 		 */
528 		if (ia == NULL)
529 			return (EADDRNOTAVAIL);
530 		/* FALLTHROUGH */
531 	case SIOCAIFADDR_IN6:
532 		/*
533 		 * We always require users to specify a valid IPv6 address for
534 		 * the corresponding operation.
535 		 */
536 		if (ifra->ifra_addr.sin6_family != AF_INET6 ||
537 		    ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6))
538 			return (EAFNOSUPPORT);
539 		if (!privileged)
540 			return (EPERM);
541 
542 		break;
543 
544 	case SIOCGIFADDR_IN6:
545 		/* This interface is basically deprecated. use SIOCGIFCONF. */
546 		/* fall through */
547 	case SIOCGIFAFLAG_IN6:
548 	case SIOCGIFNETMASK_IN6:
549 	case SIOCGIFDSTADDR_IN6:
550 	case SIOCGIFALIFETIME_IN6:
551 		/* must think again about its semantics */
552 		if (ia == NULL)
553 			return (EADDRNOTAVAIL);
554 		break;
555 	case SIOCSIFALIFETIME_IN6:
556 	    {
557 		struct in6_addrlifetime *lt;
558 
559 		if (!privileged)
560 			return (EPERM);
561 		if (ia == NULL)
562 			return (EADDRNOTAVAIL);
563 		/* sanity for overflow - beware unsigned */
564 		lt = &ifr->ifr_ifru.ifru_lifetime;
565 		if (lt->ia6t_vltime != ND6_INFINITE_LIFETIME
566 		 && lt->ia6t_vltime + time_second < time_second) {
567 			return EINVAL;
568 		}
569 		if (lt->ia6t_pltime != ND6_INFINITE_LIFETIME
570 		 && lt->ia6t_pltime + time_second < time_second) {
571 			return EINVAL;
572 		}
573 		break;
574 	    }
575 	}
576 
577 	switch (cmd) {
578 
579 	case SIOCGIFADDR_IN6:
580 		ifr->ifr_addr = ia->ia_addr;
581 		break;
582 
583 	case SIOCGIFDSTADDR_IN6:
584 		if (!(ifp->if_flags & IFF_POINTOPOINT))
585 			return (EINVAL);
586 		/*
587 		 * XXX: should we check if ifa_dstaddr is NULL and return
588 		 * an error?
589 		 */
590 		ifr->ifr_dstaddr = ia->ia_dstaddr;
591 		break;
592 
593 	case SIOCGIFNETMASK_IN6:
594 		ifr->ifr_addr = ia->ia_prefixmask;
595 		break;
596 
597 	case SIOCGIFAFLAG_IN6:
598 		ifr->ifr_ifru.ifru_flags6 = ia->ia6_flags;
599 		break;
600 
601 	case SIOCGIFSTAT_IN6:
602 		if (ifp == NULL)
603 			return EINVAL;
604 		bzero(&ifr->ifr_ifru.ifru_stat,
605 			sizeof(ifr->ifr_ifru.ifru_stat));
606 		ifr->ifr_ifru.ifru_stat =
607 			*((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->in6_ifstat;
608 		break;
609 
610 	case SIOCGIFSTAT_ICMP6:
611 		bzero(&ifr->ifr_ifru.ifru_stat,
612 			sizeof(ifr->ifr_ifru.ifru_icmp6stat));
613 		ifr->ifr_ifru.ifru_icmp6stat =
614 			*((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->icmp6_ifstat;
615 		break;
616 
617 	case SIOCGIFALIFETIME_IN6:
618 		ifr->ifr_ifru.ifru_lifetime = ia->ia6_lifetime;
619 		break;
620 
621 	case SIOCSIFALIFETIME_IN6:
622 		ia->ia6_lifetime = ifr->ifr_ifru.ifru_lifetime;
623 		/* for sanity */
624 		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
625 			ia->ia6_lifetime.ia6t_expire =
626 				time_second + ia->ia6_lifetime.ia6t_vltime;
627 		} else
628 			ia->ia6_lifetime.ia6t_expire = 0;
629 		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
630 			ia->ia6_lifetime.ia6t_preferred =
631 				time_second + ia->ia6_lifetime.ia6t_pltime;
632 		} else
633 			ia->ia6_lifetime.ia6t_preferred = 0;
634 		break;
635 
636 	case SIOCAIFADDR_IN6:
637 	{
638 		int i, error = 0, iaIsNew;
639 		struct nd_prefix pr0, *pr;
640 
641 		if (ia != NULL)
642 			iaIsNew = 0;
643 		else
644 			iaIsNew = 1;
645 
646 		/*
647 		 * first, make or update the interface address structure,
648 		 * and link it to the list.
649 		 */
650 		if ((error = in6_update_ifa(ifp, ifra, ia)) != 0)
651 			return (error);
652 
653 		/*
654 		 * then, make the prefix on-link on the interface.
655 		 * XXX: we'd rather create the prefix before the address, but
656 		 * we need at least one address to install the corresponding
657 		 * interface route, so we configure the address first.
658 		 */
659 
660 		/*
661 		 * convert mask to prefix length (prefixmask has already
662 		 * been validated in in6_update_ifa().
663 		 */
664 		bzero(&pr0, sizeof(pr0));
665 		pr0.ndpr_ifp = ifp;
666 		pr0.ndpr_plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
667 					     NULL);
668 		if (pr0.ndpr_plen == 128)
669 			break;	/* we don't need to install a host route. */
670 		pr0.ndpr_prefix = ifra->ifra_addr;
671 		pr0.ndpr_mask = ifra->ifra_prefixmask.sin6_addr;
672 		/* apply the mask for safety. */
673 		for (i = 0; i < 4; i++) {
674 			pr0.ndpr_prefix.sin6_addr.s6_addr32[i] &=
675 				ifra->ifra_prefixmask.sin6_addr.s6_addr32[i];
676 		}
677 		/*
678 		 * XXX: since we don't have an API to set prefix (not address)
679 		 * lifetimes, we just use the same lifetimes as addresses.
680 		 * The (temporarily) installed lifetimes can be overridden by
681 		 * later advertised RAs (when accept_rtadv is non 0), which is
682 		 * an intended behavior.
683 		 */
684 		pr0.ndpr_raf_onlink = 1; /* should be configurable? */
685 		pr0.ndpr_raf_auto =
686 			((ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0);
687 		pr0.ndpr_vltime = ifra->ifra_lifetime.ia6t_vltime;
688 		pr0.ndpr_pltime = ifra->ifra_lifetime.ia6t_pltime;
689 
690 		/* add the prefix if there's one. */
691 		if ((pr = nd6_prefix_lookup(&pr0)) == NULL) {
692 			/*
693 			 * nd6_prelist_add will install the corresponding
694 			 * interface route.
695 			 */
696 			if ((error = nd6_prelist_add(&pr0, NULL, &pr)) != 0)
697 				return (error);
698 			if (pr == NULL) {
699 				log(LOG_ERR, "nd6_prelist_add succeeded but "
700 				    "no prefix\n");
701 				return (EINVAL); /* XXX panic here? */
702 			}
703 		}
704 		if ((ia = in6ifa_ifpwithaddr(ifp, &ifra->ifra_addr.sin6_addr))
705 		    == NULL) {
706 		    	/* XXX: this should not happen! */
707 			log(LOG_ERR, "in6_control: addition succeeded, but"
708 			    " no ifaddr\n");
709 		} else {
710 			if ((ia->ia6_flags & IN6_IFF_AUTOCONF) &&
711 			    ia->ia6_ndpr == NULL) { /* new autoconfed addr */
712 				ia->ia6_ndpr = pr;
713 				pr->ndpr_refcnt++;
714 
715 				/*
716 				 * If this is the first autoconf address from
717 				 * the prefix, create a temporary address
718 				 * as well (when specified).
719 				 */
720 				if (ip6_use_tempaddr &&
721 				    pr->ndpr_refcnt == 1) {
722 					int e;
723 					if ((e = in6_tmpifadd(ia, 1)) != 0) {
724 						log(LOG_NOTICE, "in6_control: "
725 						    "failed to create a "
726 						    "temporary address, "
727 						    "errno=%d\n",
728 						    e);
729 					}
730 				}
731 			}
732 
733 			/*
734 			 * this might affect the status of autoconfigured
735 			 * addresses, that is, this address might make
736 			 * other addresses detached.
737 			 */
738 			pfxlist_onlink_check();
739 		}
740 		if (error == 0 && ia) {
741 			EVENTHANDLER_INVOKE(ifaddr_event, ifp,
742 			iaIsNew ? IFADDR_EVENT_ADD : IFADDR_EVENT_CHANGE,
743 			&ia->ia_ifa);
744 		}
745 		break;
746 	}
747 
748 	case SIOCDIFADDR_IN6:
749 	{
750 		int i = 0;
751 		struct nd_prefix pr0, *pr;
752 
753 		/*
754 		 * If the address being deleted is the only one that owns
755 		 * the corresponding prefix, expire the prefix as well.
756 		 * XXX: theoretically, we don't have to warry about such
757 		 * relationship, since we separate the address management
758 		 * and the prefix management.  We do this, however, to provide
759 		 * as much backward compatibility as possible in terms of
760 		 * the ioctl operation.
761 		 */
762 		bzero(&pr0, sizeof(pr0));
763 		pr0.ndpr_ifp = ifp;
764 		pr0.ndpr_plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr,
765 					     NULL);
766 		if (pr0.ndpr_plen == 128)
767 			goto purgeaddr;
768 		pr0.ndpr_prefix = ia->ia_addr;
769 		pr0.ndpr_mask = ia->ia_prefixmask.sin6_addr;
770 		for (i = 0; i < 4; i++) {
771 			pr0.ndpr_prefix.sin6_addr.s6_addr32[i] &=
772 				ia->ia_prefixmask.sin6_addr.s6_addr32[i];
773 		}
774 		/*
775 		 * The logic of the following condition is a bit complicated.
776 		 * We expire the prefix when
777 		 * 1. the address obeys autoconfiguration and it is the
778 		 *    only owner of the associated prefix, or
779 		 * 2. the address does not obey autoconf and there is no
780 		 *    other owner of the prefix.
781 		 */
782 		if ((pr = nd6_prefix_lookup(&pr0)) != NULL &&
783 		    (((ia->ia6_flags & IN6_IFF_AUTOCONF) &&
784 		       pr->ndpr_refcnt == 1) ||
785 		     (!(ia->ia6_flags & IN6_IFF_AUTOCONF) &&
786 		      pr->ndpr_refcnt == 0))) {
787 			pr->ndpr_expire = 1; /* XXX: just for expiration */
788 		}
789 
790 purgeaddr:
791 		EVENTHANDLER_INVOKE(ifaddr_event, ifp, IFADDR_EVENT_DELETE,
792 				    &ia->ia_ifa);
793 		in6_purgeaddr(&ia->ia_ifa);
794 		break;
795 	}
796 
797 	default:
798 		if (ifp == NULL || ifp->if_ioctl == 0)
799 			return (EOPNOTSUPP);
800 		ifnet_serialize_all(ifp);
801 		error = ifp->if_ioctl(ifp, cmd, data, td->td_proc->p_ucred);
802 		ifnet_deserialize_all(ifp);
803 		return (error);
804 	}
805 
806 	return (0);
807 }
808 
809 /*
810  * Update parameters of an IPv6 interface address.
811  * If necessary, a new entry is created and linked into address chains.
812  * This function is separated from in6_control().
813  * XXX: should this be performed under splnet()?
814  */
815 int
816 in6_update_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
817 	       struct in6_ifaddr *ia)
818 {
819 	int error = 0, hostIsNew = 0, plen = -1;
820 	struct in6_ifaddr *oia;
821 	struct sockaddr_in6 dst6;
822 	struct in6_addrlifetime *lt;
823 
824 	/* Validate parameters */
825 	if (ifp == NULL || ifra == NULL) /* this maybe redundant */
826 		return (EINVAL);
827 
828 	/*
829 	 * The destination address for a p2p link must have a family
830 	 * of AF_UNSPEC or AF_INET6.
831 	 */
832 	if ((ifp->if_flags & IFF_POINTOPOINT) &&
833 	    ifra->ifra_dstaddr.sin6_family != AF_INET6 &&
834 	    ifra->ifra_dstaddr.sin6_family != AF_UNSPEC)
835 		return (EAFNOSUPPORT);
836 	/*
837 	 * validate ifra_prefixmask.  don't check sin6_family, netmask
838 	 * does not carry fields other than sin6_len.
839 	 */
840 	if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6))
841 		return (EINVAL);
842 	/*
843 	 * Because the IPv6 address architecture is classless, we require
844 	 * users to specify a (non 0) prefix length (mask) for a new address.
845 	 * We also require the prefix (when specified) mask is valid, and thus
846 	 * reject a non-consecutive mask.
847 	 */
848 	if (ia == NULL && ifra->ifra_prefixmask.sin6_len == 0)
849 		return (EINVAL);
850 	if (ifra->ifra_prefixmask.sin6_len != 0) {
851 		plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
852 				    (u_char *)&ifra->ifra_prefixmask +
853 				    ifra->ifra_prefixmask.sin6_len);
854 		if (plen <= 0)
855 			return (EINVAL);
856 	}
857 	else {
858 		/*
859 		 * In this case, ia must not be NULL.  We just use its prefix
860 		 * length.
861 		 */
862 		plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
863 	}
864 	/*
865 	 * If the destination address on a p2p interface is specified,
866 	 * and the address is a scoped one, validate/set the scope
867 	 * zone identifier.
868 	 */
869 	dst6 = ifra->ifra_dstaddr;
870 	if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) &&
871 	    (dst6.sin6_family == AF_INET6)) {
872 		int scopeid;
873 
874 		if ((error = in6_recoverscope(&dst6,
875 					      &ifra->ifra_dstaddr.sin6_addr,
876 					      ifp)) != 0)
877 			return (error);
878 		scopeid = in6_addr2scopeid(ifp, &dst6.sin6_addr);
879 		if (dst6.sin6_scope_id == 0) /* user omit to specify the ID. */
880 			dst6.sin6_scope_id = scopeid;
881 		else if (dst6.sin6_scope_id != scopeid)
882 			return (EINVAL); /* scope ID mismatch. */
883 		if ((error = in6_embedscope(&dst6.sin6_addr, &dst6, NULL, NULL))
884 		    != 0)
885 			return (error);
886 		dst6.sin6_scope_id = 0; /* XXX */
887 	}
888 	/*
889 	 * The destination address can be specified only for a p2p or a
890 	 * loopback interface.  If specified, the corresponding prefix length
891 	 * must be 128.
892 	 */
893 	if (ifra->ifra_dstaddr.sin6_family == AF_INET6) {
894 		if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) == 0) {
895 			/* XXX: noisy message */
896 			log(LOG_INFO, "in6_update_ifa: a destination can be "
897 			    "specified for a p2p or a loopback IF only\n");
898 			return (EINVAL);
899 		}
900 		if (plen != 128) {
901 			/*
902 			 * The following message seems noisy, but we dare to
903 			 * add it for diagnosis.
904 			 */
905 			log(LOG_INFO, "in6_update_ifa: prefixlen must be 128 "
906 			    "when dstaddr is specified\n");
907 			return (EINVAL);
908 		}
909 	}
910 	/* lifetime consistency check */
911 	lt = &ifra->ifra_lifetime;
912 	if (lt->ia6t_vltime != ND6_INFINITE_LIFETIME
913 	    && lt->ia6t_vltime + time_second < time_second) {
914 		return EINVAL;
915 	}
916 	if (lt->ia6t_vltime == 0) {
917 		/*
918 		 * the following log might be noisy, but this is a typical
919 		 * configuration mistake or a tool's bug.
920 		 */
921 		log(LOG_INFO,
922 		    "in6_update_ifa: valid lifetime is 0 for %s\n",
923 		    ip6_sprintf(&ifra->ifra_addr.sin6_addr));
924 	}
925 	if (lt->ia6t_pltime != ND6_INFINITE_LIFETIME
926 	    && lt->ia6t_pltime + time_second < time_second) {
927 		return EINVAL;
928 	}
929 
930 	/*
931 	 * If this is a new address, allocate a new ifaddr and link it
932 	 * into chains.
933 	 */
934 	if (ia == NULL) {
935 		hostIsNew = 1;
936 		/*
937 		 * When in6_update_ifa() is called in a process of a received
938 		 * RA, it is called under splnet().  So, we should call malloc
939 		 * with M_NOWAIT.
940 		 */
941 		ia = ifa_create(sizeof(*ia), M_NOWAIT);
942 		if (ia == NULL)
943 			return (ENOBUFS);
944 		/* Initialize the address and masks */
945 		ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
946 		ia->ia_addr.sin6_family = AF_INET6;
947 		ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
948 		if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
949 			/*
950 			 * XXX: some functions expect that ifa_dstaddr is not
951 			 * NULL for p2p interfaces.
952 			 */
953 			ia->ia_ifa.ifa_dstaddr
954 				= (struct sockaddr *)&ia->ia_dstaddr;
955 		} else {
956 			ia->ia_ifa.ifa_dstaddr = NULL;
957 		}
958 		ia->ia_ifa.ifa_netmask
959 			= (struct sockaddr *)&ia->ia_prefixmask;
960 
961 		ia->ia_ifp = ifp;
962 		if ((oia = in6_ifaddr) != NULL) {
963 			for ( ; oia->ia_next; oia = oia->ia_next)
964 				continue;
965 			oia->ia_next = ia;
966 		} else
967 			in6_ifaddr = ia;
968 
969 		ifa_iflink(&ia->ia_ifa, ifp, 1);
970 	}
971 
972 	/* set prefix mask */
973 	if (ifra->ifra_prefixmask.sin6_len) {
974 		/*
975 		 * We prohibit changing the prefix length of an existing
976 		 * address, because
977 		 * + such an operation should be rare in IPv6, and
978 		 * + the operation would confuse prefix management.
979 		 */
980 		if (ia->ia_prefixmask.sin6_len &&
981 		    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) != plen) {
982 			log(LOG_INFO, "in6_update_ifa: the prefix length of an"
983 			    " existing (%s) address should not be changed\n",
984 			    ip6_sprintf(&ia->ia_addr.sin6_addr));
985 			error = EINVAL;
986 			goto unlink;
987 		}
988 		ia->ia_prefixmask = ifra->ifra_prefixmask;
989 	}
990 
991 	/*
992 	 * If a new destination address is specified, scrub the old one and
993 	 * install the new destination.  Note that the interface must be
994 	 * p2p or loopback (see the check above.)
995 	 */
996 	if (dst6.sin6_family == AF_INET6 &&
997 	    !IN6_ARE_ADDR_EQUAL(&dst6.sin6_addr,
998 				&ia->ia_dstaddr.sin6_addr)) {
999 		int e;
1000 
1001 		if ((ia->ia_flags & IFA_ROUTE) &&
1002 		    (e = rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST))
1003 		    != 0) {
1004 			log(LOG_ERR, "in6_update_ifa: failed to remove "
1005 			    "a route to the old destination: %s\n",
1006 			    ip6_sprintf(&ia->ia_addr.sin6_addr));
1007 			/* proceed anyway... */
1008 		}
1009 		else
1010 			ia->ia_flags &= ~IFA_ROUTE;
1011 		ia->ia_dstaddr = dst6;
1012 	}
1013 
1014 	/* reset the interface and routing table appropriately. */
1015 	if ((error = in6_ifinit(ifp, ia, &ifra->ifra_addr, hostIsNew)) != 0)
1016 		goto unlink;
1017 
1018 	/*
1019 	 * Beyond this point, we should call in6_purgeaddr upon an error,
1020 	 * not just go to unlink.
1021 	 */
1022 
1023 #if 0				/* disable this mechanism for now */
1024 	/* update prefix list */
1025 	if (hostIsNew &&
1026 	    (ifra->ifra_flags & IN6_IFF_NOPFX) == 0) { /* XXX */
1027 		int iilen;
1028 
1029 		iilen = (sizeof(ia->ia_prefixmask.sin6_addr) << 3) - plen;
1030 		if ((error = in6_prefix_add_ifid(iilen, ia)) != 0) {
1031 			in6_purgeaddr((struct ifaddr *)ia);
1032 			return (error);
1033 		}
1034 	}
1035 #endif
1036 
1037 	if (ifp->if_flags & IFF_MULTICAST) {
1038 		struct sockaddr_in6 mltaddr, mltmask;
1039 		struct in6_multi *in6m;
1040 
1041 		if (hostIsNew) {
1042 			/*
1043 			 * join solicited multicast addr for new host id
1044 			 */
1045 			struct in6_addr llsol;
1046 			bzero(&llsol, sizeof(struct in6_addr));
1047 			llsol.s6_addr16[0] = htons(0xff02);
1048 			llsol.s6_addr16[1] = htons(ifp->if_index);
1049 			llsol.s6_addr32[1] = 0;
1050 			llsol.s6_addr32[2] = htonl(1);
1051 			llsol.s6_addr32[3] =
1052 				ifra->ifra_addr.sin6_addr.s6_addr32[3];
1053 			llsol.s6_addr8[12] = 0xff;
1054 			in6_addmulti(&llsol, ifp, &error);
1055 			if (error != 0) {
1056 				log(LOG_WARNING,
1057 				    "in6_update_ifa: addmulti failed for "
1058 				    "%s on %s (errno=%d)\n",
1059 				    ip6_sprintf(&llsol), if_name(ifp),
1060 				    error);
1061 				in6_purgeaddr((struct ifaddr *)ia);
1062 				return (error);
1063 			}
1064 		}
1065 
1066 		bzero(&mltmask, sizeof(mltmask));
1067 		mltmask.sin6_len = sizeof(struct sockaddr_in6);
1068 		mltmask.sin6_family = AF_INET6;
1069 		mltmask.sin6_addr = in6mask32;
1070 
1071 		/*
1072 		 * join link-local all-nodes address
1073 		 */
1074 		bzero(&mltaddr, sizeof(mltaddr));
1075 		mltaddr.sin6_len = sizeof(struct sockaddr_in6);
1076 		mltaddr.sin6_family = AF_INET6;
1077 		mltaddr.sin6_addr = kin6addr_linklocal_allnodes;
1078 		mltaddr.sin6_addr.s6_addr16[1] = htons(ifp->if_index);
1079 
1080 		IN6_LOOKUP_MULTI(mltaddr.sin6_addr, ifp, in6m);
1081 		if (in6m == NULL) {
1082 			rtrequest_global(RTM_ADD,
1083 				  (struct sockaddr *)&mltaddr,
1084 				  (struct sockaddr *)&ia->ia_addr,
1085 				  (struct sockaddr *)&mltmask,
1086 				  RTF_UP|RTF_CLONING);  /* xxx */
1087 			in6_addmulti(&mltaddr.sin6_addr, ifp, &error);
1088 			if (error != 0) {
1089 				log(LOG_WARNING,
1090 				    "in6_update_ifa: addmulti failed for "
1091 				    "%s on %s (errno=%d)\n",
1092 				    ip6_sprintf(&mltaddr.sin6_addr),
1093 				    if_name(ifp), error);
1094 			}
1095 		}
1096 
1097 		/*
1098 		 * join node information group address
1099 		 */
1100 #define hostnamelen	strlen(hostname)
1101 		if (in6_nigroup(ifp, hostname, hostnamelen, &mltaddr.sin6_addr)
1102 		    == 0) {
1103 			IN6_LOOKUP_MULTI(mltaddr.sin6_addr, ifp, in6m);
1104 			if (in6m == NULL && ia != NULL) {
1105 				in6_addmulti(&mltaddr.sin6_addr, ifp, &error);
1106 				if (error != 0) {
1107 					log(LOG_WARNING, "in6_update_ifa: "
1108 					    "addmulti failed for "
1109 					    "%s on %s (errno=%d)\n",
1110 					    ip6_sprintf(&mltaddr.sin6_addr),
1111 					    if_name(ifp), error);
1112 				}
1113 			}
1114 		}
1115 #undef hostnamelen
1116 
1117 		/*
1118 		 * join node-local all-nodes address, on loopback.
1119 		 * XXX: since "node-local" is obsoleted by interface-local,
1120 		 *      we have to join the group on every interface with
1121 		 *      some interface-boundary restriction.
1122 		 */
1123 		if (ifp->if_flags & IFF_LOOPBACK) {
1124 			struct in6_ifaddr *ia_loop;
1125 
1126 			struct in6_addr loop6 = kin6addr_loopback;
1127 			ia_loop = in6ifa_ifpwithaddr(ifp, &loop6);
1128 
1129 			mltaddr.sin6_addr = kin6addr_nodelocal_allnodes;
1130 
1131 			IN6_LOOKUP_MULTI(mltaddr.sin6_addr, ifp, in6m);
1132 			if (in6m == NULL && ia_loop != NULL) {
1133 				rtrequest_global(RTM_ADD,
1134 					  (struct sockaddr *)&mltaddr,
1135 					  (struct sockaddr *)&ia_loop->ia_addr,
1136 					  (struct sockaddr *)&mltmask,
1137 					  RTF_UP);
1138 				in6_addmulti(&mltaddr.sin6_addr, ifp, &error);
1139 				if (error != 0) {
1140 					log(LOG_WARNING, "in6_update_ifa: "
1141 					    "addmulti failed for %s on %s "
1142 					    "(errno=%d)\n",
1143 					    ip6_sprintf(&mltaddr.sin6_addr),
1144 					    if_name(ifp), error);
1145 				}
1146 			}
1147 		}
1148 	}
1149 
1150 	ia->ia6_flags = ifra->ifra_flags;
1151 	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/*safety*/
1152 	ia->ia6_flags &= ~IN6_IFF_NODAD;	/* Mobile IPv6 */
1153 
1154 	ia->ia6_lifetime = ifra->ifra_lifetime;
1155 	/* for sanity */
1156 	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1157 		ia->ia6_lifetime.ia6t_expire =
1158 			time_second + ia->ia6_lifetime.ia6t_vltime;
1159 	} else
1160 		ia->ia6_lifetime.ia6t_expire = 0;
1161 	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1162 		ia->ia6_lifetime.ia6t_preferred =
1163 			time_second + ia->ia6_lifetime.ia6t_pltime;
1164 	} else
1165 		ia->ia6_lifetime.ia6t_preferred = 0;
1166 
1167 	/*
1168 	 * Perform DAD, if needed.
1169 	 * XXX It may be of use, if we can administratively
1170 	 * disable DAD.
1171 	 */
1172 	if (in6if_do_dad(ifp) && !(ifra->ifra_flags & IN6_IFF_NODAD)) {
1173 		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1174 		nd6_dad_start((struct ifaddr *)ia, NULL);
1175 	}
1176 
1177 	return (error);
1178 
1179 unlink:
1180 	/*
1181 	 * XXX: if a change of an existing address failed, keep the entry
1182 	 * anyway.
1183 	 */
1184 	if (hostIsNew)
1185 		in6_unlink_ifa(ia, ifp);
1186 	return (error);
1187 }
1188 
1189 void
1190 in6_purgeaddr(struct ifaddr *ifa)
1191 {
1192 	struct ifnet *ifp = ifa->ifa_ifp;
1193 	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1194 
1195 	/* stop DAD processing */
1196 	nd6_dad_stop(ifa);
1197 
1198 	/*
1199 	 * delete route to the destination of the address being purged.
1200 	 * The interface must be p2p or loopback in this case.
1201 	 */
1202 	if ((ia->ia_flags & IFA_ROUTE) && ia->ia_dstaddr.sin6_len != 0) {
1203 		int e;
1204 
1205 		if ((e = rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST))
1206 		    != 0) {
1207 			log(LOG_ERR, "in6_purgeaddr: failed to remove "
1208 			    "a route to the p2p destination: %s on %s, "
1209 			    "errno=%d\n",
1210 			    ip6_sprintf(&ia->ia_addr.sin6_addr), if_name(ifp),
1211 			    e);
1212 			/* proceed anyway... */
1213 		}
1214 		else
1215 			ia->ia_flags &= ~IFA_ROUTE;
1216 	}
1217 
1218 	/* Remove ownaddr's loopback rtentry, if it exists. */
1219 	in6_ifremloop(&(ia->ia_ifa));
1220 
1221 	if (ifp->if_flags & IFF_MULTICAST) {
1222 		/*
1223 		 * delete solicited multicast addr for deleting host id
1224 		 */
1225 		struct in6_multi *in6m;
1226 		struct in6_addr llsol;
1227 		bzero(&llsol, sizeof(struct in6_addr));
1228 		llsol.s6_addr16[0] = htons(0xff02);
1229 		llsol.s6_addr16[1] = htons(ifp->if_index);
1230 		llsol.s6_addr32[1] = 0;
1231 		llsol.s6_addr32[2] = htonl(1);
1232 		llsol.s6_addr32[3] =
1233 			ia->ia_addr.sin6_addr.s6_addr32[3];
1234 		llsol.s6_addr8[12] = 0xff;
1235 
1236 		IN6_LOOKUP_MULTI(llsol, ifp, in6m);
1237 		if (in6m)
1238 			in6_delmulti(in6m);
1239 	}
1240 
1241 	in6_unlink_ifa(ia, ifp);
1242 }
1243 
1244 static void
1245 in6_unlink_ifa(struct in6_ifaddr *ia, struct ifnet *ifp)
1246 {
1247 	int plen, iilen;
1248 	struct in6_ifaddr *oia;
1249 
1250 	crit_enter();
1251 
1252 	ifa_ifunlink(&ia->ia_ifa, ifp);
1253 
1254 	oia = ia;
1255 	if (oia == (ia = in6_ifaddr))
1256 		in6_ifaddr = ia->ia_next;
1257 	else {
1258 		while (ia->ia_next && (ia->ia_next != oia))
1259 			ia = ia->ia_next;
1260 		if (ia->ia_next)
1261 			ia->ia_next = oia->ia_next;
1262 		else {
1263 			/* search failed */
1264 			kprintf("Couldn't unlink in6_ifaddr from in6_ifaddr\n");
1265 		}
1266 	}
1267 
1268 	if (oia->ia6_ifpr) {	/* check for safety */
1269 		plen = in6_mask2len(&oia->ia_prefixmask.sin6_addr, NULL);
1270 		iilen = (sizeof(oia->ia_prefixmask.sin6_addr) << 3) - plen;
1271 		in6_prefix_remove_ifid(iilen, oia);
1272 	}
1273 
1274 	/*
1275 	 * When an autoconfigured address is being removed, release the
1276 	 * reference to the base prefix.  Also, since the release might
1277 	 * affect the status of other (detached) addresses, call
1278 	 * pfxlist_onlink_check().
1279 	 */
1280 	if (oia->ia6_flags & IN6_IFF_AUTOCONF) {
1281 		if (oia->ia6_ndpr == NULL) {
1282 			log(LOG_NOTICE, "in6_unlink_ifa: autoconf'ed address "
1283 			    "%p has no prefix\n", oia);
1284 		} else {
1285 			oia->ia6_ndpr->ndpr_refcnt--;
1286 			oia->ia6_flags &= ~IN6_IFF_AUTOCONF;
1287 			oia->ia6_ndpr = NULL;
1288 		}
1289 
1290 		pfxlist_onlink_check();
1291 	}
1292 
1293 	/*
1294 	 * release another refcnt for the link from in6_ifaddr.
1295 	 * Note that we should decrement the refcnt at least once for all *BSD.
1296 	 */
1297 	ifa_destroy(&oia->ia_ifa);
1298 
1299 	crit_exit();
1300 }
1301 
1302 void
1303 in6_purgeif(struct ifnet *ifp)
1304 {
1305 	struct ifaddr_container *ifac, *next;
1306 
1307 	TAILQ_FOREACH_MUTABLE(ifac, &ifp->if_addrheads[mycpuid],
1308 			      ifa_link, next) {
1309 		if (ifac->ifa->ifa_addr->sa_family != AF_INET6)
1310 			continue;
1311 		in6_purgeaddr(ifac->ifa);
1312 	}
1313 
1314 	in6_ifdetach(ifp);
1315 }
1316 
1317 /*
1318  * SIOC[GAD]LIFADDR.
1319  *	SIOCGLIFADDR: get first address. (?)
1320  *	SIOCGLIFADDR with IFLR_PREFIX:
1321  *		get first address that matches the specified prefix.
1322  *	SIOCALIFADDR: add the specified address.
1323  *	SIOCALIFADDR with IFLR_PREFIX:
1324  *		add the specified prefix, filling hostid part from
1325  *		the first link-local address.  prefixlen must be <= 64.
1326  *	SIOCDLIFADDR: delete the specified address.
1327  *	SIOCDLIFADDR with IFLR_PREFIX:
1328  *		delete the first address that matches the specified prefix.
1329  * return values:
1330  *	EINVAL on invalid parameters
1331  *	EADDRNOTAVAIL on prefix match failed/specified address not found
1332  *	other values may be returned from in6_ioctl()
1333  *
1334  * NOTE: SIOCALIFADDR(with IFLR_PREFIX set) allows prefixlen less than 64.
1335  * this is to accomodate address naming scheme other than RFC2374,
1336  * in the future.
1337  * RFC2373 defines interface id to be 64bit, but it allows non-RFC2374
1338  * address encoding scheme. (see figure on page 8)
1339  */
1340 static int
1341 in6_lifaddr_ioctl(struct socket *so, u_long cmd, caddr_t data,
1342 		  struct ifnet *ifp, struct thread *td)
1343 {
1344 	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
1345 	struct sockaddr *sa;
1346 
1347 	/* sanity checks */
1348 	if (!data || !ifp) {
1349 		panic("invalid argument to in6_lifaddr_ioctl");
1350 		/*NOTRECHED*/
1351 	}
1352 
1353 	switch (cmd) {
1354 	case SIOCGLIFADDR:
1355 		/* address must be specified on GET with IFLR_PREFIX */
1356 		if (!(iflr->flags & IFLR_PREFIX))
1357 			break;
1358 		/* FALLTHROUGH */
1359 	case SIOCALIFADDR:
1360 	case SIOCDLIFADDR:
1361 		/* address must be specified on ADD and DELETE */
1362 		sa = (struct sockaddr *)&iflr->addr;
1363 		if (sa->sa_family != AF_INET6)
1364 			return EINVAL;
1365 		if (sa->sa_len != sizeof(struct sockaddr_in6))
1366 			return EINVAL;
1367 		/* XXX need improvement */
1368 		sa = (struct sockaddr *)&iflr->dstaddr;
1369 		if (sa->sa_family && sa->sa_family != AF_INET6)
1370 			return EINVAL;
1371 		if (sa->sa_len && sa->sa_len != sizeof(struct sockaddr_in6))
1372 			return EINVAL;
1373 		break;
1374 	default: /* shouldn't happen */
1375 #if 0
1376 		panic("invalid cmd to in6_lifaddr_ioctl");
1377 		/* NOTREACHED */
1378 #else
1379 		return EOPNOTSUPP;
1380 #endif
1381 	}
1382 	if (sizeof(struct in6_addr) * 8 < iflr->prefixlen)
1383 		return EINVAL;
1384 
1385 	switch (cmd) {
1386 	case SIOCALIFADDR:
1387 	    {
1388 		struct in6_aliasreq ifra;
1389 		struct in6_addr *hostid = NULL;
1390 		int prefixlen;
1391 
1392 		if (iflr->flags & IFLR_PREFIX) {
1393 			struct ifaddr *ifa;
1394 			struct sockaddr_in6 *sin6;
1395 
1396 			/*
1397 			 * hostid is to fill in the hostid part of the
1398 			 * address.  hostid points to the first link-local
1399 			 * address attached to the interface.
1400 			 */
1401 			ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0);
1402 			if (!ifa)
1403 				return EADDRNOTAVAIL;
1404 			hostid = IFA_IN6(ifa);
1405 
1406 		 	/* prefixlen must be <= 64. */
1407 			if (64 < iflr->prefixlen)
1408 				return EINVAL;
1409 			prefixlen = iflr->prefixlen;
1410 
1411 			/* hostid part must be zero. */
1412 			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1413 			if (sin6->sin6_addr.s6_addr32[2] != 0
1414 			 || sin6->sin6_addr.s6_addr32[3] != 0) {
1415 				return EINVAL;
1416 			}
1417 		} else
1418 			prefixlen = iflr->prefixlen;
1419 
1420 		/* copy args to in6_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
1421 		bzero(&ifra, sizeof(ifra));
1422 		bcopy(iflr->iflr_name, ifra.ifra_name,
1423 			sizeof(ifra.ifra_name));
1424 
1425 		bcopy(&iflr->addr, &ifra.ifra_addr,
1426 			((struct sockaddr *)&iflr->addr)->sa_len);
1427 		if (hostid) {
1428 			/* fill in hostid part */
1429 			ifra.ifra_addr.sin6_addr.s6_addr32[2] =
1430 				hostid->s6_addr32[2];
1431 			ifra.ifra_addr.sin6_addr.s6_addr32[3] =
1432 				hostid->s6_addr32[3];
1433 		}
1434 
1435 		if (((struct sockaddr *)&iflr->dstaddr)->sa_family) {	/*XXX*/
1436 			bcopy(&iflr->dstaddr, &ifra.ifra_dstaddr,
1437 				((struct sockaddr *)&iflr->dstaddr)->sa_len);
1438 			if (hostid) {
1439 				ifra.ifra_dstaddr.sin6_addr.s6_addr32[2] =
1440 					hostid->s6_addr32[2];
1441 				ifra.ifra_dstaddr.sin6_addr.s6_addr32[3] =
1442 					hostid->s6_addr32[3];
1443 			}
1444 		}
1445 
1446 		ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
1447 		in6_len2mask(&ifra.ifra_prefixmask.sin6_addr, prefixlen);
1448 
1449 		ifra.ifra_flags = iflr->flags & ~IFLR_PREFIX;
1450 		return in6_control(so, SIOCAIFADDR_IN6, (caddr_t)&ifra, ifp, td);
1451 	    }
1452 	case SIOCGLIFADDR:
1453 	case SIOCDLIFADDR:
1454 	    {
1455 		struct ifaddr_container *ifac;
1456 		struct in6_ifaddr *ia;
1457 		struct in6_addr mask, candidate, match;
1458 		struct sockaddr_in6 *sin6;
1459 		int cmp;
1460 
1461 		bzero(&mask, sizeof(mask));
1462 		if (iflr->flags & IFLR_PREFIX) {
1463 			/* lookup a prefix rather than address. */
1464 			in6_len2mask(&mask, iflr->prefixlen);
1465 
1466 			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1467 			bcopy(&sin6->sin6_addr, &match, sizeof(match));
1468 			match.s6_addr32[0] &= mask.s6_addr32[0];
1469 			match.s6_addr32[1] &= mask.s6_addr32[1];
1470 			match.s6_addr32[2] &= mask.s6_addr32[2];
1471 			match.s6_addr32[3] &= mask.s6_addr32[3];
1472 
1473 			/* if you set extra bits, that's wrong */
1474 			if (bcmp(&match, &sin6->sin6_addr, sizeof(match)))
1475 				return EINVAL;
1476 
1477 			cmp = 1;
1478 		} else {
1479 			if (cmd == SIOCGLIFADDR) {
1480 				/* on getting an address, take the 1st match */
1481 				cmp = 0;	/* XXX */
1482 			} else {
1483 				/* on deleting an address, do exact match */
1484 				in6_len2mask(&mask, 128);
1485 				sin6 = (struct sockaddr_in6 *)&iflr->addr;
1486 				bcopy(&sin6->sin6_addr, &match, sizeof(match));
1487 
1488 				cmp = 1;
1489 			}
1490 		}
1491 
1492 		TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
1493 			struct ifaddr *ifa = ifac->ifa;
1494 
1495 			if (ifa->ifa_addr->sa_family != AF_INET6)
1496 				continue;
1497 			if (!cmp)
1498 				break;
1499 
1500 			bcopy(IFA_IN6(ifa), &candidate, sizeof(candidate));
1501 			/*
1502 			 * XXX: this is adhoc, but is necessary to allow
1503 			 * a user to specify fe80::/64 (not /10) for a
1504 			 * link-local address.
1505 			 */
1506 			if (IN6_IS_ADDR_LINKLOCAL(&candidate))
1507 				candidate.s6_addr16[1] = 0;
1508 			candidate.s6_addr32[0] &= mask.s6_addr32[0];
1509 			candidate.s6_addr32[1] &= mask.s6_addr32[1];
1510 			candidate.s6_addr32[2] &= mask.s6_addr32[2];
1511 			candidate.s6_addr32[3] &= mask.s6_addr32[3];
1512 			if (IN6_ARE_ADDR_EQUAL(&candidate, &match))
1513 				break;
1514 		}
1515 		if (ifac == NULL)
1516 			return EADDRNOTAVAIL;
1517 		ia = ifa2ia6(ifac->ifa);
1518 
1519 		if (cmd == SIOCGLIFADDR) {
1520 			struct sockaddr_in6 *s6;
1521 
1522 			/* fill in the if_laddrreq structure */
1523 			bcopy(&ia->ia_addr, &iflr->addr, ia->ia_addr.sin6_len);
1524 			s6 = (struct sockaddr_in6 *)&iflr->addr;
1525 			if (IN6_IS_ADDR_LINKLOCAL(&s6->sin6_addr)) {
1526 				s6->sin6_addr.s6_addr16[1] = 0;
1527 				s6->sin6_scope_id =
1528 					in6_addr2scopeid(ifp, &s6->sin6_addr);
1529 			}
1530 			if (ifp->if_flags & IFF_POINTOPOINT) {
1531 				bcopy(&ia->ia_dstaddr, &iflr->dstaddr,
1532 					ia->ia_dstaddr.sin6_len);
1533 				s6 = (struct sockaddr_in6 *)&iflr->dstaddr;
1534 				if (IN6_IS_ADDR_LINKLOCAL(&s6->sin6_addr)) {
1535 					s6->sin6_addr.s6_addr16[1] = 0;
1536 					s6->sin6_scope_id =
1537 						in6_addr2scopeid(ifp,
1538 								 &s6->sin6_addr);
1539 				}
1540 			} else
1541 				bzero(&iflr->dstaddr, sizeof(iflr->dstaddr));
1542 
1543 			iflr->prefixlen =
1544 				in6_mask2len(&ia->ia_prefixmask.sin6_addr,
1545 					     NULL);
1546 
1547 			iflr->flags = ia->ia6_flags;	/* XXX */
1548 
1549 			return 0;
1550 		} else {
1551 			struct in6_aliasreq ifra;
1552 
1553 			/* fill in6_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
1554 			bzero(&ifra, sizeof(ifra));
1555 			bcopy(iflr->iflr_name, ifra.ifra_name,
1556 			      sizeof(ifra.ifra_name));
1557 
1558 			bcopy(&ia->ia_addr, &ifra.ifra_addr,
1559 			      ia->ia_addr.sin6_len);
1560 			if (ifp->if_flags & IFF_POINTOPOINT)
1561 				bcopy(&ia->ia_dstaddr, &ifra.ifra_dstaddr,
1562 				      ia->ia_dstaddr.sin6_len);
1563 			else
1564 				bzero(&ifra.ifra_dstaddr,
1565 				      sizeof(ifra.ifra_dstaddr));
1566 			bcopy(&ia->ia_prefixmask, &ifra.ifra_dstaddr,
1567 			      ia->ia_prefixmask.sin6_len);
1568 
1569 			ifra.ifra_flags = ia->ia6_flags;
1570 			return in6_control(so, SIOCDIFADDR_IN6, (caddr_t)&ifra,
1571 				ifp, td);
1572 		}
1573 	    }
1574 	}
1575 
1576 	return EOPNOTSUPP;	/* just for safety */
1577 }
1578 
1579 /*
1580  * Initialize an interface's intetnet6 address
1581  * and routing table entry.
1582  */
1583 static int
1584 in6_ifinit(struct ifnet *ifp, struct in6_ifaddr *ia, struct sockaddr_in6 *sin6,
1585 	   int newhost)
1586 {
1587 	int	error = 0, plen, ifacount = 0;
1588 	struct ifaddr_container *ifac;
1589 
1590 	/*
1591 	 * Give the interface a chance to initialize
1592 	 * if this is its first address,
1593 	 * and to validate the address if necessary.
1594 	 */
1595 	TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
1596 		if (ifac->ifa->ifa_addr == NULL)
1597 			continue;	/* just for safety */
1598 		if (ifac->ifa->ifa_addr->sa_family != AF_INET6)
1599 			continue;
1600 		ifacount++;
1601 	}
1602 
1603 	ifnet_serialize_all(ifp);
1604 
1605 	ia->ia_addr = *sin6;
1606 
1607 	if (ifacount <= 1 && ifp->if_ioctl &&
1608 	    (error = ifp->if_ioctl(ifp, SIOCSIFADDR, (caddr_t)ia, NULL))) {
1609 		ifnet_deserialize_all(ifp);
1610 		return (error);
1611 	}
1612 
1613 	ifnet_deserialize_all(ifp);
1614 
1615 	ia->ia_ifa.ifa_metric = ifp->if_metric;
1616 
1617 	/* we could do in(6)_socktrim here, but just omit it at this moment. */
1618 
1619 	/*
1620 	 * Special case:
1621 	 * If the destination address is specified for a point-to-point
1622 	 * interface, install a route to the destination as an interface
1623 	 * direct route.
1624 	 */
1625 	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1626 	if (plen == 128 && ia->ia_dstaddr.sin6_family == AF_INET6) {
1627 		if ((error = rtinit(&(ia->ia_ifa), (int)RTM_ADD,
1628 				    RTF_UP | RTF_HOST)) != 0)
1629 			return (error);
1630 		ia->ia_flags |= IFA_ROUTE;
1631 	}
1632 	if (plen < 128) {
1633 		/*
1634 		 * The RTF_CLONING flag is necessary for in6_is_ifloop_auto().
1635 		 */
1636 		ia->ia_ifa.ifa_flags |= RTF_CLONING;
1637 	}
1638 
1639 	/* Add ownaddr as loopback rtentry, if necessary (ex. on p2p link). */
1640 	if (newhost) {
1641 		/* set the rtrequest function to create llinfo */
1642 		ia->ia_ifa.ifa_rtrequest = nd6_rtrequest;
1643 		in6_ifaddloop(&(ia->ia_ifa));
1644 	}
1645 
1646 	return (error);
1647 }
1648 
1649 struct in6_multi_mship *
1650 in6_joingroup(struct ifnet *ifp, struct in6_addr *addr, int *errorp)
1651 {
1652        struct in6_multi_mship *imm;
1653 
1654        imm = kmalloc(sizeof(*imm), M_IPMADDR, M_NOWAIT);
1655        if (!imm) {
1656                *errorp = ENOBUFS;
1657                return NULL;
1658        }
1659        imm->i6mm_maddr = in6_addmulti(addr, ifp, errorp);
1660        if (!imm->i6mm_maddr) {
1661                /* *errorp is alrady set */
1662                kfree(imm, M_IPMADDR);
1663                return NULL;
1664        }
1665        return imm;
1666 }
1667 
1668 int
1669 in6_leavegroup(struct in6_multi_mship *imm)
1670 {
1671 
1672        if (imm->i6mm_maddr)
1673                in6_delmulti(imm->i6mm_maddr);
1674        kfree(imm,  M_IPMADDR);
1675        return 0;
1676 }
1677 
1678 /*
1679  * Add an address to the list of IP6 multicast addresses for a
1680  * given interface.
1681  */
1682 struct	in6_multi *
1683 in6_addmulti(struct in6_addr *maddr6, struct ifnet *ifp, int *errorp)
1684 {
1685 	struct	in6_multi *in6m;
1686 	struct sockaddr_in6 sin6;
1687 	struct ifmultiaddr *ifma;
1688 
1689 	*errorp = 0;
1690 
1691 	crit_enter();
1692 
1693 	/*
1694 	 * Call generic routine to add membership or increment
1695 	 * refcount.  It wants addresses in the form of a sockaddr,
1696 	 * so we build one here (being careful to zero the unused bytes).
1697 	 */
1698 	bzero(&sin6, sizeof sin6);
1699 	sin6.sin6_family = AF_INET6;
1700 	sin6.sin6_len = sizeof sin6;
1701 	sin6.sin6_addr = *maddr6;
1702 	*errorp = if_addmulti(ifp, (struct sockaddr *)&sin6, &ifma);
1703 	if (*errorp) {
1704 		crit_exit();
1705 		return 0;
1706 	}
1707 
1708 	/*
1709 	 * If ifma->ifma_protospec is null, then if_addmulti() created
1710 	 * a new record.  Otherwise, we are done.
1711 	 */
1712 	if (ifma->ifma_protospec != 0) {
1713 		crit_exit();
1714 		return ifma->ifma_protospec;
1715 	}
1716 
1717 	/* XXX - if_addmulti uses M_WAITOK.  Can this really be called
1718 	   at interrupt time?  If so, need to fix if_addmulti. XXX */
1719 	in6m = (struct in6_multi *)kmalloc(sizeof(*in6m), M_IPMADDR, M_NOWAIT);
1720 	if (in6m == NULL) {
1721 		crit_exit();
1722 		return (NULL);
1723 	}
1724 
1725 	bzero(in6m, sizeof *in6m);
1726 	in6m->in6m_addr = *maddr6;
1727 	in6m->in6m_ifp = ifp;
1728 	in6m->in6m_ifma = ifma;
1729 	ifma->ifma_protospec = in6m;
1730 	LIST_INSERT_HEAD(&in6_multihead, in6m, in6m_entry);
1731 
1732 	/*
1733 	 * Let MLD6 know that we have joined a new IP6 multicast
1734 	 * group.
1735 	 */
1736 	mld6_start_listening(in6m);
1737 	crit_exit();
1738 	return (in6m);
1739 }
1740 
1741 /*
1742  * Delete a multicast address record.
1743  */
1744 void
1745 in6_delmulti(struct in6_multi *in6m)
1746 {
1747 	struct ifmultiaddr *ifma = in6m->in6m_ifma;
1748 
1749 	crit_enter();
1750 
1751 	if (ifma->ifma_refcount == 1) {
1752 		/*
1753 		 * No remaining claims to this record; let MLD6 know
1754 		 * that we are leaving the multicast group.
1755 		 */
1756 		mld6_stop_listening(in6m);
1757 		ifma->ifma_protospec = 0;
1758 		LIST_REMOVE(in6m, in6m_entry);
1759 		kfree(in6m, M_IPMADDR);
1760 	}
1761 	/* XXX - should be separate API for when we have an ifma? */
1762 	if_delmulti(ifma->ifma_ifp, ifma->ifma_addr);
1763 	crit_exit();
1764 }
1765 
1766 /*
1767  * Find an IPv6 interface link-local address specific to an interface.
1768  */
1769 struct in6_ifaddr *
1770 in6ifa_ifpforlinklocal(struct ifnet *ifp, int ignoreflags)
1771 {
1772 	struct ifaddr_container *ifac;
1773 
1774 	TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
1775 		struct ifaddr *ifa = ifac->ifa;
1776 
1777 		if (ifa->ifa_addr == NULL)
1778 			continue;	/* just for safety */
1779 		if (ifa->ifa_addr->sa_family != AF_INET6)
1780 			continue;
1781 		if (IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa))) {
1782 			if ((((struct in6_ifaddr *)ifa)->ia6_flags &
1783 			     ignoreflags) != 0)
1784 				continue;
1785 			break;
1786 		}
1787 	}
1788 	if (ifac != NULL)
1789 		return ((struct in6_ifaddr *)(ifac->ifa));
1790 	else
1791 		return (NULL);
1792 }
1793 
1794 
1795 /*
1796  * find the internet address corresponding to a given interface and address.
1797  */
1798 struct in6_ifaddr *
1799 in6ifa_ifpwithaddr(struct ifnet *ifp, struct in6_addr *addr)
1800 {
1801 	struct ifaddr_container *ifac;
1802 
1803 	TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
1804 		struct ifaddr *ifa = ifac->ifa;
1805 
1806 		if (ifa->ifa_addr == NULL)
1807 			continue;	/* just for safety */
1808 		if (ifa->ifa_addr->sa_family != AF_INET6)
1809 			continue;
1810 		if (IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa)))
1811 			break;
1812 	}
1813 	if (ifac != NULL)
1814 		return ((struct in6_ifaddr *)(ifac->ifa));
1815 	else
1816 		return (NULL);
1817 }
1818 
1819 /*
1820  * find the internet address on a given interface corresponding to a neighbor's
1821  * address.
1822  */
1823 struct in6_ifaddr *
1824 in6ifa_ifplocaladdr(const struct ifnet *ifp, const struct in6_addr *addr)
1825 {
1826 	struct ifaddr *ifa;
1827 	struct in6_ifaddr *ia;
1828 	struct ifaddr_container *ifac;
1829 
1830 	TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
1831 		ifa = ifac->ifa;
1832 
1833 		if (ifa->ifa_addr == NULL)
1834 			continue;	/* just for safety */
1835 		if (ifa->ifa_addr->sa_family != AF_INET6)
1836 			continue;
1837 		ia = (struct in6_ifaddr *)ifa;
1838 		if (IN6_ARE_MASKED_ADDR_EQUAL(addr,
1839 				&ia->ia_addr.sin6_addr,
1840 				&ia->ia_prefixmask.sin6_addr))
1841 			return ia;
1842 	}
1843 
1844 	return NULL;
1845 }
1846 
1847 /*
1848  * Convert IP6 address to printable (loggable) representation.
1849  */
1850 static char digits[] = "0123456789abcdef";
1851 static int ip6round = 0;
1852 char *
1853 ip6_sprintf(const struct in6_addr *addr)
1854 {
1855 	static char ip6buf[8][48];
1856 	int i;
1857 	char *cp;
1858 	const u_short *a = (const u_short *)addr;
1859 	const u_char *d;
1860 	int dcolon = 0;
1861 
1862 	ip6round = (ip6round + 1) & 7;
1863 	cp = ip6buf[ip6round];
1864 
1865 	for (i = 0; i < 8; i++) {
1866 		if (dcolon == 1) {
1867 			if (*a == 0) {
1868 				if (i == 7)
1869 					*cp++ = ':';
1870 				a++;
1871 				continue;
1872 			} else
1873 				dcolon = 2;
1874 		}
1875 		if (*a == 0) {
1876 			if (dcolon == 0 && *(a + 1) == 0) {
1877 				if (i == 0)
1878 					*cp++ = ':';
1879 				*cp++ = ':';
1880 				dcolon = 1;
1881 			} else {
1882 				*cp++ = '0';
1883 				*cp++ = ':';
1884 			}
1885 			a++;
1886 			continue;
1887 		}
1888 		d = (const u_char *)a;
1889 		*cp++ = digits[*d >> 4];
1890 		*cp++ = digits[*d++ & 0xf];
1891 		*cp++ = digits[*d >> 4];
1892 		*cp++ = digits[*d & 0xf];
1893 		*cp++ = ':';
1894 		a++;
1895 	}
1896 	*--cp = 0;
1897 	return (ip6buf[ip6round]);
1898 }
1899 
1900 int
1901 in6_localaddr(struct in6_addr *in6)
1902 {
1903 	struct in6_ifaddr *ia;
1904 
1905 	if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
1906 		return 1;
1907 
1908 	for (ia = in6_ifaddr; ia; ia = ia->ia_next)
1909 		if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
1910 					      &ia->ia_prefixmask.sin6_addr))
1911 			return 1;
1912 
1913 	return (0);
1914 }
1915 
1916 int
1917 in6_is_addr_deprecated(struct sockaddr_in6 *sa6)
1918 {
1919 	struct in6_ifaddr *ia;
1920 
1921 	for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
1922 		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
1923 				       &sa6->sin6_addr) &&
1924 		    (ia->ia6_flags & IN6_IFF_DEPRECATED))
1925 			return (1); /* true */
1926 
1927 		/* XXX: do we still have to go thru the rest of the list? */
1928 	}
1929 
1930 	return (0);		/* false */
1931 }
1932 
1933 /*
1934  * return length of part which dst and src are equal
1935  * hard coding...
1936  */
1937 int
1938 in6_matchlen(struct in6_addr *src, struct in6_addr *dst)
1939 {
1940 	int match = 0;
1941 	u_char *s = (u_char *)src, *d = (u_char *)dst;
1942 	u_char *lim = s + 16, r;
1943 
1944 	while (s < lim)
1945 		if ((r = (*d++ ^ *s++)) != 0) {
1946 			while (r < 128) {
1947 				match++;
1948 				r <<= 1;
1949 			}
1950 			break;
1951 		} else
1952 			match += 8;
1953 	return match;
1954 }
1955 
1956 /* XXX: to be scope conscious */
1957 int
1958 in6_are_prefix_equal(struct in6_addr *p1, struct in6_addr *p2, int len)
1959 {
1960 	int bytelen, bitlen;
1961 
1962 	/* sanity check */
1963 	if (0 > len || len > 128) {
1964 		log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n",
1965 		    len);
1966 		return (0);
1967 	}
1968 
1969 	bytelen = len / 8;
1970 	bitlen = len % 8;
1971 
1972 	if (bcmp(&p1->s6_addr, &p2->s6_addr, bytelen))
1973 		return (0);
1974 	if (p1->s6_addr[bytelen] >> (8 - bitlen) !=
1975 	    p2->s6_addr[bytelen] >> (8 - bitlen))
1976 		return (0);
1977 
1978 	return (1);
1979 }
1980 
1981 void
1982 in6_prefixlen2mask(struct in6_addr *maskp, int len)
1983 {
1984 	u_char maskarray[8] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
1985 	int bytelen, bitlen, i;
1986 
1987 	/* sanity check */
1988 	if (0 > len || len > 128) {
1989 		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
1990 		    len);
1991 		return;
1992 	}
1993 
1994 	bzero(maskp, sizeof(*maskp));
1995 	bytelen = len / 8;
1996 	bitlen = len % 8;
1997 	for (i = 0; i < bytelen; i++)
1998 		maskp->s6_addr[i] = 0xff;
1999 	if (bitlen)
2000 		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
2001 }
2002 
2003 /*
2004  * return the best address out of the same scope
2005  */
2006 struct in6_ifaddr *
2007 in6_ifawithscope(struct ifnet *oifp, struct in6_addr *dst)
2008 {
2009 	int dst_scope =	in6_addrscope(dst), src_scope, best_scope = 0;
2010 	int blen = -1;
2011 	struct ifnet *ifp;
2012 	struct in6_ifaddr *ifa_best = NULL;
2013 
2014 	if (oifp == NULL) {
2015 #if 0
2016 		kprintf("in6_ifawithscope: output interface is not specified\n");
2017 #endif
2018 		return (NULL);
2019 	}
2020 
2021 	/*
2022 	 * We search for all addresses on all interfaces from the beginning.
2023 	 * Comparing an interface with the outgoing interface will be done
2024 	 * only at the final stage of tiebreaking.
2025 	 */
2026 	for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list))
2027 	{
2028 		struct ifaddr_container *ifac;
2029 
2030 		/*
2031 		 * We can never take an address that breaks the scope zone
2032 		 * of the destination.
2033 		 */
2034 		if (in6_addr2scopeid(ifp, dst) != in6_addr2scopeid(oifp, dst))
2035 			continue;
2036 
2037 		TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
2038 			int tlen = -1, dscopecmp, bscopecmp, matchcmp;
2039 			struct ifaddr *ifa = ifac->ifa;
2040 
2041 			if (ifa->ifa_addr->sa_family != AF_INET6)
2042 				continue;
2043 
2044 			src_scope = in6_addrscope(IFA_IN6(ifa));
2045 
2046 			/*
2047 			 * Don't use an address before completing DAD
2048 			 * nor a duplicated address.
2049 			 */
2050 			if (((struct in6_ifaddr *)ifa)->ia6_flags &
2051 			    IN6_IFF_NOTREADY)
2052 				continue;
2053 
2054 			/* XXX: is there any case to allow anycasts? */
2055 			if (((struct in6_ifaddr *)ifa)->ia6_flags &
2056 			    IN6_IFF_ANYCAST)
2057 				continue;
2058 
2059 			if (((struct in6_ifaddr *)ifa)->ia6_flags &
2060 			    IN6_IFF_DETACHED)
2061 				continue;
2062 
2063 			/*
2064 			 * If this is the first address we find,
2065 			 * keep it anyway.
2066 			 */
2067 			if (ifa_best == NULL)
2068 				goto replace;
2069 
2070 			/*
2071 			 * ifa_best is never NULL beyond this line except
2072 			 * within the block labeled "replace".
2073 			 */
2074 
2075 			/*
2076 			 * If ifa_best has a smaller scope than dst and
2077 			 * the current address has a larger one than
2078 			 * (or equal to) dst, always replace ifa_best.
2079 			 * Also, if the current address has a smaller scope
2080 			 * than dst, ignore it unless ifa_best also has a
2081 			 * smaller scope.
2082 			 * Consequently, after the two if-clause below,
2083 			 * the followings must be satisfied:
2084 			 * (scope(src) < scope(dst) &&
2085 			 *  scope(best) < scope(dst))
2086 			 *  OR
2087 			 * (scope(best) >= scope(dst) &&
2088 			 *  scope(src) >= scope(dst))
2089 			 */
2090 			if (IN6_ARE_SCOPE_CMP(best_scope, dst_scope) < 0 &&
2091 			    IN6_ARE_SCOPE_CMP(src_scope, dst_scope) >= 0)
2092 				goto replace; /* (A) */
2093 			if (IN6_ARE_SCOPE_CMP(src_scope, dst_scope) < 0 &&
2094 			    IN6_ARE_SCOPE_CMP(best_scope, dst_scope) >= 0)
2095 				continue; /* (B) */
2096 
2097 			/*
2098 			 * A deprecated address SHOULD NOT be used in new
2099 			 * communications if an alternate (non-deprecated)
2100 			 * address is available and has sufficient scope.
2101 			 * RFC 2462, Section 5.5.4.
2102 			 */
2103 			if (((struct in6_ifaddr *)ifa)->ia6_flags &
2104 			    IN6_IFF_DEPRECATED) {
2105 				/*
2106 				 * Ignore any deprecated addresses if
2107 				 * specified by configuration.
2108 				 */
2109 				if (!ip6_use_deprecated)
2110 					continue;
2111 
2112 				/*
2113 				 * If we have already found a non-deprecated
2114 				 * candidate, just ignore deprecated addresses.
2115 				 */
2116 				if (!(ifa_best->ia6_flags & IN6_IFF_DEPRECATED))
2117 					continue;
2118 			}
2119 
2120 			/*
2121 			 * A non-deprecated address is always preferred
2122 			 * to a deprecated one regardless of scopes and
2123 			 * address matching (Note invariants ensured by the
2124 			 * conditions (A) and (B) above.)
2125 			 */
2126 			if ((ifa_best->ia6_flags & IN6_IFF_DEPRECATED) &&
2127 			    !(((struct in6_ifaddr *)ifa)->ia6_flags &
2128 			     IN6_IFF_DEPRECATED))
2129 				goto replace;
2130 
2131 			/*
2132 			 * When we use temporary addresses described in
2133 			 * RFC 3041, we prefer temporary addresses to
2134 			 * public autoconf addresses.  Again, note the
2135 			 * invariants from (A) and (B).  Also note that we
2136 			 * don't have any preference between static addresses
2137 			 * and autoconf addresses (despite of whether or not
2138 			 * the latter is temporary or public.)
2139 			 */
2140 			if (ip6_use_tempaddr) {
2141 				struct in6_ifaddr *ifat;
2142 
2143 				ifat = (struct in6_ifaddr *)ifa;
2144 				if ((ifa_best->ia6_flags &
2145 				     (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY))
2146 				     == IN6_IFF_AUTOCONF &&
2147 				    (ifat->ia6_flags &
2148 				     (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY))
2149 				     == (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY)) {
2150 					goto replace;
2151 				}
2152 				if ((ifa_best->ia6_flags &
2153 				     (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY))
2154 				    == (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY) &&
2155 				    (ifat->ia6_flags &
2156 				     (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY))
2157 				     == IN6_IFF_AUTOCONF) {
2158 					continue;
2159 				}
2160 			}
2161 
2162 			/*
2163 			 * At this point, we have two cases:
2164 			 * 1. we are looking at a non-deprecated address,
2165 			 *    and ifa_best is also non-deprecated.
2166 			 * 2. we are looking at a deprecated address,
2167 			 *    and ifa_best is also deprecated.
2168 			 * Also, we do not have to consider a case where
2169 			 * the scope of if_best is larger(smaller) than dst and
2170 			 * the scope of the current address is smaller(larger)
2171 			 * than dst. Such a case has already been covered.
2172 			 * Tiebreaking is done according to the following
2173 			 * items:
2174 			 * - the scope comparison between the address and
2175 			 *   dst (dscopecmp)
2176 			 * - the scope comparison between the address and
2177 			 *   ifa_best (bscopecmp)
2178 			 * - if the address match dst longer than ifa_best
2179 			 *   (matchcmp)
2180 			 * - if the address is on the outgoing I/F (outI/F)
2181 			 *
2182 			 * Roughly speaking, the selection policy is
2183 			 * - the most important item is scope. The same scope
2184 			 *   is best. Then search for a larger scope.
2185 			 *   Smaller scopes are the last resort.
2186 			 * - A deprecated address is chosen only when we have
2187 			 *   no address that has an enough scope, but is
2188 			 *   prefered to any addresses of smaller scopes
2189 			 *   (this must be already done above.)
2190 			 * - addresses on the outgoing I/F are preferred to
2191 			 *   ones on other interfaces if none of above
2192 			 *   tiebreaks.  In the table below, the column "bI"
2193 			 *   means if the best_ifa is on the outgoing
2194 			 *   interface, and the column "sI" means if the ifa
2195 			 *   is on the outgoing interface.
2196 			 * - If there is no other reasons to choose one,
2197 			 *   longest address match against dst is considered.
2198 			 *
2199 			 * The precise decision table is as follows:
2200 			 * dscopecmp bscopecmp    match  bI oI | replace?
2201 			 *       N/A     equal      N/A   Y  N |   No (1)
2202 			 *       N/A     equal      N/A   N  Y |  Yes (2)
2203 			 *       N/A     equal   larger    N/A |  Yes (3)
2204 			 *       N/A     equal  !larger    N/A |   No (4)
2205 			 *    larger    larger      N/A    N/A |   No (5)
2206 			 *    larger   smaller      N/A    N/A |  Yes (6)
2207 			 *   smaller    larger      N/A    N/A |  Yes (7)
2208 			 *   smaller   smaller      N/A    N/A |   No (8)
2209 			 *     equal   smaller      N/A    N/A |  Yes (9)
2210 			 *     equal    larger       (already done at A above)
2211 			 */
2212 			dscopecmp = IN6_ARE_SCOPE_CMP(src_scope, dst_scope);
2213 			bscopecmp = IN6_ARE_SCOPE_CMP(src_scope, best_scope);
2214 
2215 			if (bscopecmp == 0) {
2216 				struct ifnet *bifp = ifa_best->ia_ifp;
2217 
2218 				if (bifp == oifp && ifp != oifp) /* (1) */
2219 					continue;
2220 				if (bifp != oifp && ifp == oifp) /* (2) */
2221 					goto replace;
2222 
2223 				/*
2224 				 * Both bifp and ifp are on the outgoing
2225 				 * interface, or both two are on a different
2226 				 * interface from the outgoing I/F.
2227 				 * now we need address matching against dst
2228 				 * for tiebreaking.
2229 				 */
2230 				tlen = in6_matchlen(IFA_IN6(ifa), dst);
2231 				matchcmp = tlen - blen;
2232 				if (matchcmp > 0) /* (3) */
2233 					goto replace;
2234 				continue; /* (4) */
2235 			}
2236 			if (dscopecmp > 0) {
2237 				if (bscopecmp > 0) /* (5) */
2238 					continue;
2239 				goto replace; /* (6) */
2240 			}
2241 			if (dscopecmp < 0) {
2242 				if (bscopecmp > 0) /* (7) */
2243 					goto replace;
2244 				continue; /* (8) */
2245 			}
2246 
2247 			/* now dscopecmp must be 0 */
2248 			if (bscopecmp < 0)
2249 				goto replace; /* (9) */
2250 
2251 		  replace:
2252 			ifa_best = (struct in6_ifaddr *)ifa;
2253 			blen = tlen >= 0 ? tlen :
2254 				in6_matchlen(IFA_IN6(ifa), dst);
2255 			best_scope = in6_addrscope(&ifa_best->ia_addr.sin6_addr);
2256 		}
2257 	}
2258 
2259 	/* count statistics for future improvements */
2260 	if (ifa_best == NULL)
2261 		ip6stat.ip6s_sources_none++;
2262 	else {
2263 		if (oifp == ifa_best->ia_ifp)
2264 			ip6stat.ip6s_sources_sameif[best_scope]++;
2265 		else
2266 			ip6stat.ip6s_sources_otherif[best_scope]++;
2267 
2268 		if (best_scope == dst_scope)
2269 			ip6stat.ip6s_sources_samescope[best_scope]++;
2270 		else
2271 			ip6stat.ip6s_sources_otherscope[best_scope]++;
2272 
2273 		if (ifa_best->ia6_flags & IN6_IFF_DEPRECATED)
2274 			ip6stat.ip6s_sources_deprecated[best_scope]++;
2275 	}
2276 
2277 	return (ifa_best);
2278 }
2279 
2280 /*
2281  * return the best address out of the same scope. if no address was
2282  * found, return the first valid address from designated IF.
2283  */
2284 struct in6_ifaddr *
2285 in6_ifawithifp(struct ifnet *ifp, struct in6_addr *dst)
2286 {
2287 	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
2288 	struct ifaddr_container *ifac;
2289 	struct in6_ifaddr *besta = 0;
2290 	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
2291 
2292 	dep[0] = dep[1] = NULL;
2293 
2294 	/*
2295 	 * We first look for addresses in the same scope.
2296 	 * If there is one, return it.
2297 	 * If two or more, return one which matches the dst longest.
2298 	 * If none, return one of global addresses assigned other ifs.
2299 	 */
2300 	TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
2301 		struct ifaddr *ifa = ifac->ifa;
2302 
2303 		if (ifa->ifa_addr->sa_family != AF_INET6)
2304 			continue;
2305 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
2306 			continue; /* XXX: is there any case to allow anycast? */
2307 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
2308 			continue; /* don't use this interface */
2309 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
2310 			continue;
2311 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2312 			if (ip6_use_deprecated)
2313 				dep[0] = (struct in6_ifaddr *)ifa;
2314 			continue;
2315 		}
2316 
2317 		if (dst_scope == in6_addrscope(IFA_IN6(ifa))) {
2318 			/*
2319 			 * call in6_matchlen() as few as possible
2320 			 */
2321 			if (besta) {
2322 				if (blen == -1)
2323 					blen = in6_matchlen(&besta->ia_addr.sin6_addr, dst);
2324 				tlen = in6_matchlen(IFA_IN6(ifa), dst);
2325 				if (tlen > blen) {
2326 					blen = tlen;
2327 					besta = (struct in6_ifaddr *)ifa;
2328 				}
2329 			} else
2330 				besta = (struct in6_ifaddr *)ifa;
2331 		}
2332 	}
2333 	if (besta)
2334 		return (besta);
2335 
2336 	TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
2337 		struct ifaddr *ifa = ifac->ifa;
2338 
2339 		if (ifa->ifa_addr->sa_family != AF_INET6)
2340 			continue;
2341 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
2342 			continue; /* XXX: is there any case to allow anycast? */
2343 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
2344 			continue; /* don't use this interface */
2345 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
2346 			continue;
2347 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2348 			if (ip6_use_deprecated)
2349 				dep[1] = (struct in6_ifaddr *)ifa;
2350 			continue;
2351 		}
2352 
2353 		return (struct in6_ifaddr *)ifa;
2354 	}
2355 
2356 	/* use the last-resort values, that are, deprecated addresses */
2357 	if (dep[0])
2358 		return dep[0];
2359 	if (dep[1])
2360 		return dep[1];
2361 
2362 	return NULL;
2363 }
2364 
2365 /*
2366  * perform DAD when interface becomes IFF_UP.
2367  */
2368 void
2369 in6_if_up(struct ifnet *ifp)
2370 {
2371 	struct ifaddr_container *ifac;
2372 	struct in6_ifaddr *ia;
2373 	int dad_delay;		/* delay ticks before DAD output */
2374 
2375 	/*
2376 	 * special cases, like 6to4, are handled in in6_ifattach
2377 	 */
2378 	in6_ifattach(ifp, NULL);
2379 
2380 	dad_delay = 0;
2381 	TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
2382 		struct ifaddr *ifa = ifac->ifa;
2383 
2384 		if (ifa->ifa_addr->sa_family != AF_INET6)
2385 			continue;
2386 		ia = (struct in6_ifaddr *)ifa;
2387 		if (ia->ia6_flags & IN6_IFF_TENTATIVE)
2388 			nd6_dad_start(ifa, &dad_delay);
2389 	}
2390 }
2391 
2392 int
2393 in6if_do_dad(struct ifnet *ifp)
2394 {
2395 	if (ifp->if_flags & IFF_LOOPBACK)
2396 		return (0);
2397 
2398 	switch (ifp->if_type) {
2399 #ifdef IFT_DUMMY
2400 	case IFT_DUMMY:
2401 #endif
2402 	case IFT_FAITH:
2403 		/*
2404 		 * These interfaces do not have the IFF_LOOPBACK flag,
2405 		 * but loop packets back.  We do not have to do DAD on such
2406 		 * interfaces.  We should even omit it, because loop-backed
2407 		 * NS would confuse the DAD procedure.
2408 		 */
2409 		return (0);
2410 	default:
2411 		/*
2412 		 * Our DAD routine requires the interface up and running.
2413 		 * However, some interfaces can be up before the RUNNING
2414 		 * status.  Additionaly, users may try to assign addresses
2415 		 * before the interface becomes up (or running).
2416 		 * We simply skip DAD in such a case as a work around.
2417 		 * XXX: we should rather mark "tentative" on such addresses,
2418 		 * and do DAD after the interface becomes ready.
2419 		 */
2420 		if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) !=
2421 		    (IFF_UP|IFF_RUNNING))
2422 			return (0);
2423 
2424 		return (1);
2425 	}
2426 }
2427 
2428 /*
2429  * Calculate max IPv6 MTU through all the interfaces and store it
2430  * to in6_maxmtu.
2431  */
2432 void
2433 in6_setmaxmtu(void)
2434 {
2435 	unsigned long maxmtu = 0;
2436 	struct ifnet *ifp;
2437 
2438 	for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list))
2439 	{
2440 		/* this function can be called during ifnet initialization */
2441 		if (!ifp->if_afdata[AF_INET6])
2442 			continue;
2443 		if (!(ifp->if_flags & IFF_LOOPBACK) &&
2444 		    ND_IFINFO(ifp)->linkmtu > maxmtu)
2445 			maxmtu =  ND_IFINFO(ifp)->linkmtu;
2446 	}
2447 	if (maxmtu)	/* update only when maxmtu is positive */
2448 		in6_maxmtu = maxmtu;
2449 }
2450 
2451 void *
2452 in6_domifattach(struct ifnet *ifp)
2453 {
2454 	struct in6_ifextra *ext;
2455 
2456 	ext = (struct in6_ifextra *)kmalloc(sizeof(*ext), M_IFADDR, M_WAITOK);
2457 	bzero(ext, sizeof(*ext));
2458 
2459 	ext->in6_ifstat = (struct in6_ifstat *)kmalloc(sizeof(struct in6_ifstat),
2460 		M_IFADDR, M_WAITOK);
2461 	bzero(ext->in6_ifstat, sizeof(*ext->in6_ifstat));
2462 
2463 	ext->icmp6_ifstat =
2464 		(struct icmp6_ifstat *)kmalloc(sizeof(struct icmp6_ifstat),
2465 			M_IFADDR, M_WAITOK);
2466 	bzero(ext->icmp6_ifstat, sizeof(*ext->icmp6_ifstat));
2467 
2468 	ext->nd_ifinfo = nd6_ifattach(ifp);
2469 	ext->scope6_id = scope6_ifattach(ifp);
2470 	return ext;
2471 }
2472 
2473 void
2474 in6_domifdetach(struct ifnet *ifp, void *aux)
2475 {
2476 	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
2477 	scope6_ifdetach(ext->scope6_id);
2478 	nd6_ifdetach(ext->nd_ifinfo);
2479 	kfree(ext->in6_ifstat, M_IFADDR);
2480 	kfree(ext->icmp6_ifstat, M_IFADDR);
2481 	kfree(ext, M_IFADDR);
2482 }
2483 
2484 /*
2485  * Convert sockaddr_in6 to sockaddr_in.  Original sockaddr_in6 must be
2486  * v4 mapped addr or v4 compat addr
2487  */
2488 void
2489 in6_sin6_2_sin(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2490 {
2491 	bzero(sin, sizeof(*sin));
2492 	sin->sin_len = sizeof(struct sockaddr_in);
2493 	sin->sin_family = AF_INET;
2494 	sin->sin_port = sin6->sin6_port;
2495 	sin->sin_addr.s_addr = sin6->sin6_addr.s6_addr32[3];
2496 }
2497 
2498 /* Convert sockaddr_in to sockaddr_in6 in v4 mapped addr format. */
2499 void
2500 in6_sin_2_v4mapsin6(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2501 {
2502 	bzero(sin6, sizeof(*sin6));
2503 	sin6->sin6_len = sizeof(struct sockaddr_in6);
2504 	sin6->sin6_family = AF_INET6;
2505 	sin6->sin6_port = sin->sin_port;
2506 	sin6->sin6_addr.s6_addr32[0] = 0;
2507 	sin6->sin6_addr.s6_addr32[1] = 0;
2508 	sin6->sin6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
2509 	sin6->sin6_addr.s6_addr32[3] = sin->sin_addr.s_addr;
2510 }
2511 
2512 /* Convert sockaddr_in6 into sockaddr_in. */
2513 void
2514 in6_sin6_2_sin_in_sock(struct sockaddr *nam)
2515 {
2516 	struct sockaddr_in *sin_p;
2517 	struct sockaddr_in6 sin6;
2518 
2519 	/*
2520 	 * Save original sockaddr_in6 addr and convert it
2521 	 * to sockaddr_in.
2522 	 */
2523 	sin6 = *(struct sockaddr_in6 *)nam;
2524 	sin_p = (struct sockaddr_in *)nam;
2525 	in6_sin6_2_sin(sin_p, &sin6);
2526 }
2527 
2528 /* Convert sockaddr_in into sockaddr_in6 in v4 mapped addr format. */
2529 void
2530 in6_sin_2_v4mapsin6_in_sock(struct sockaddr **nam)
2531 {
2532 	struct sockaddr_in *sin_p;
2533 	struct sockaddr_in6 *sin6_p;
2534 
2535 	MALLOC(sin6_p, struct sockaddr_in6 *, sizeof *sin6_p, M_SONAME,
2536 	       M_WAITOK);
2537 	sin_p = (struct sockaddr_in *)*nam;
2538 	in6_sin_2_v4mapsin6(sin_p, sin6_p);
2539 	FREE(*nam, M_SONAME);
2540 	*nam = (struct sockaddr *)sin6_p;
2541 }
2542