xref: /dragonfly/sys/netinet6/in6.c (revision dadd6466)
1 /*	$FreeBSD: src/sys/netinet6/in6.c,v 1.7.2.9 2002/04/28 05:40:26 suz Exp $	*/
2 /*	$KAME: in6.c,v 1.259 2002/01/21 11:37:50 keiichi 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 "opt_inet.h"
65 #include "opt_inet6.h"
66 
67 #include <sys/param.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/systm.h>
74 #include <sys/proc.h>
75 #include <sys/priv.h>
76 #include <sys/time.h>
77 #include <sys/kernel.h>
78 #include <sys/syslog.h>
79 #include <sys/jail.h>
80 
81 #include <sys/thread2.h>
82 #include <sys/msgport2.h>
83 
84 #include <net/if.h>
85 #include <net/if_types.h>
86 #include <net/route.h>
87 #include <net/if_dl.h>
88 
89 #include <netinet/in.h>
90 #include <netinet/in_var.h>
91 #include <netinet/if_ether.h>
92 #include <netinet/in_systm.h>
93 #include <netinet/ip.h>
94 #include <netinet/in_pcb.h>
95 
96 #include <netinet/ip6.h>
97 #include <netinet6/ip6_var.h>
98 #include <netinet6/nd6.h>
99 #include <netinet6/mld6_var.h>
100 #include <netinet6/ip6_mroute.h>
101 #include <netinet6/in6_ifattach.h>
102 #include <netinet6/scope6_var.h>
103 #include <netinet6/in6_pcb.h>
104 #include <netinet6/in6_var.h>
105 
106 #include <net/net_osdep.h>
107 
108 /*
109  * Definitions of some costant IP6 addresses.
110  */
111 const struct in6_addr kin6addr_any = IN6ADDR_ANY_INIT;
112 const struct in6_addr kin6addr_loopback = IN6ADDR_LOOPBACK_INIT;
113 const struct in6_addr kin6addr_nodelocal_allnodes =
114 	IN6ADDR_NODELOCAL_ALLNODES_INIT;
115 const struct in6_addr kin6addr_linklocal_allnodes =
116 	IN6ADDR_LINKLOCAL_ALLNODES_INIT;
117 const struct in6_addr kin6addr_linklocal_allrouters =
118 	IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
119 
120 const struct in6_addr in6mask0 = IN6MASK0;
121 const struct in6_addr in6mask32 = IN6MASK32;
122 const struct in6_addr in6mask64 = IN6MASK64;
123 const struct in6_addr in6mask96 = IN6MASK96;
124 const struct in6_addr in6mask128 = IN6MASK128;
125 
126 const struct sockaddr_in6 sa6_any = {sizeof(sa6_any), AF_INET6,
127 				     0, 0, IN6ADDR_ANY_INIT, 0};
128 
129 static int in6_lifaddr_ioctl (struct socket *, u_long, caddr_t,
130 	struct ifnet *, struct thread *);
131 static int in6_ifinit (struct ifnet *, struct in6_ifaddr *,
132 			   struct sockaddr_in6 *, int);
133 static void in6_unlink_ifa (struct in6_ifaddr *, struct ifnet *);
134 static void in6_ifloop_request_callback(int, int, struct rt_addrinfo *, struct rtentry *, void *);
135 
136 struct in6_multihead in6_multihead;	/* XXX BSS initialization */
137 
138 int	(*faithprefix_p)(struct in6_addr *);
139 
140 /*
141  * Subroutine for in6_ifaddloop() and in6_ifremloop().
142  * This routine does actual work.
143  */
144 static void
145 in6_ifloop_request(int cmd, struct ifaddr *ifa)
146 {
147 	struct sockaddr_in6 all1_sa;
148         struct rt_addrinfo rtinfo;
149 	int error;
150 
151 	bzero(&all1_sa, sizeof(all1_sa));
152 	all1_sa.sin6_family = AF_INET6;
153 	all1_sa.sin6_len = sizeof(struct sockaddr_in6);
154 	all1_sa.sin6_addr = in6mask128;
155 
156 	/*
157 	 * We specify the address itself as the gateway, and set the
158 	 * RTF_LLINFO flag, so that the corresponding host route would have
159 	 * the flag, and thus applications that assume traditional behavior
160 	 * would be happy.  Note that we assume the caller of the function
161 	 * (probably implicitly) set nd6_rtrequest() to ifa->ifa_rtrequest,
162 	 * which changes the outgoing interface to the loopback interface.
163 	 */
164 	bzero(&rtinfo, sizeof(struct rt_addrinfo));
165 	rtinfo.rti_info[RTAX_DST] = ifa->ifa_addr;
166 	rtinfo.rti_info[RTAX_GATEWAY] = ifa->ifa_addr;
167 	rtinfo.rti_info[RTAX_NETMASK] = (struct sockaddr *)&all1_sa;
168 	rtinfo.rti_flags = RTF_UP|RTF_HOST|RTF_LLINFO;
169 
170 	error = rtrequest1_global(cmd, &rtinfo,
171 	    in6_ifloop_request_callback, ifa, RTREQ_PRIO_NORM);
172 	if (error != 0) {
173 		log(LOG_ERR, "in6_ifloop_request: "
174 		    "%s operation failed for %s (errno=%d)\n",
175 		    cmd == RTM_ADD ? "ADD" : "DELETE",
176 		    ip6_sprintf(&((struct in6_ifaddr *)ifa)->ia_addr.sin6_addr),
177 		    error);
178 	}
179 }
180 
181 static void
182 in6_ifloop_request_callback(int cmd, int error, struct rt_addrinfo *rtinfo,
183 			    struct rtentry *rt, void *arg)
184 {
185 	struct ifaddr *ifa = arg;
186 
187 	if (error)
188 		goto done;
189 
190 	/*
191 	 * Make sure rt_ifa be equal to IFA, the second argument of the
192 	 * function.
193 	 * We need this because when we refer to rt_ifa->ia6_flags in
194 	 * ip6_input, we assume that the rt_ifa points to the address instead
195 	 * of the loopback address.
196 	 */
197 	if (cmd == RTM_ADD && rt && ifa != rt->rt_ifa) {
198 		++rt->rt_refcnt;
199 		IFAFREE(rt->rt_ifa);
200 		IFAREF(ifa);
201 		rt->rt_ifa = ifa;
202 		--rt->rt_refcnt;
203 	}
204 
205 	/*
206 	 * Report the addition/removal of the address to the routing socket.
207 	 * XXX: since we called rtinit for a p2p interface with a destination,
208 	 *      we end up reporting twice in such a case.  Should we rather
209 	 *      omit the second report?
210 	 */
211 	if (rt) {
212 		if (mycpuid == 0)
213 			rt_newaddrmsg(cmd, ifa, error, rt);
214 		if (cmd == RTM_DELETE) {
215 			if (rt->rt_refcnt == 0) {
216 				++rt->rt_refcnt;
217 				rtfree(rt);
218 			}
219 		}
220 	}
221 done:
222 	/* no way to return any new error */
223 	;
224 }
225 
226 /*
227  * Add ownaddr as loopback rtentry.  We previously add the route only if
228  * necessary (ex. on a p2p link).  However, since we now manage addresses
229  * separately from prefixes, we should always add the route.  We can't
230  * rely on the cloning mechanism from the corresponding interface route
231  * any more.
232  */
233 void
234 in6_ifaddloop(struct ifaddr *ifa)
235 {
236 	struct rtentry *rt;
237 
238 	/* If there is no loopback entry, allocate one. */
239 	rt = rtpurelookup(ifa->ifa_addr);
240 	if (rt == NULL || !(rt->rt_flags & RTF_HOST) ||
241 	    !(rt->rt_ifp->if_flags & IFF_LOOPBACK))
242 		in6_ifloop_request(RTM_ADD, ifa);
243 	if (rt != NULL)
244 		rt->rt_refcnt--;
245 }
246 
247 /*
248  * Remove loopback rtentry of ownaddr generated by in6_ifaddloop(),
249  * if it exists.
250  */
251 void
252 in6_ifremloop(struct ifaddr *ifa)
253 {
254 	struct in6_ifaddr *ia;
255 	struct rtentry *rt;
256 	int ia_count = 0;
257 
258 	/*
259 	 * Some of BSD variants do not remove cloned routes
260 	 * from an interface direct route, when removing the direct route
261 	 * (see comments in net/net_osdep.h).  Even for variants that do remove
262 	 * cloned routes, they could fail to remove the cloned routes when
263 	 * we handle multple addresses that share a common prefix.
264 	 * So, we should remove the route corresponding to the deleted address
265 	 * regardless of the result of in6_is_ifloop_auto().
266 	 */
267 
268 	/*
269 	 * Delete the entry only if exact one ifa exists.  More than one ifa
270 	 * can exist if we assign a same single address to multiple
271 	 * (probably p2p) interfaces.
272 	 * XXX: we should avoid such a configuration in IPv6...
273 	 */
274 	for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
275 		if (IN6_ARE_ADDR_EQUAL(IFA_IN6(ifa), &ia->ia_addr.sin6_addr)) {
276 			ia_count++;
277 			if (ia_count > 1)
278 				break;
279 		}
280 	}
281 
282 	if (ia_count == 1) {
283 		/*
284 		 * Before deleting, check if a corresponding loopbacked host
285 		 * route surely exists.  With this check, we can avoid to
286 		 * delete an interface direct route whose destination is same
287 		 * as the address being removed.  This can happen when remofing
288 		 * a subnet-router anycast address on an interface attahced
289 		 * to a shared medium.
290 		 */
291 		rt = rtpurelookup(ifa->ifa_addr);
292 		if (rt != NULL && (rt->rt_flags & RTF_HOST) &&
293 		    (rt->rt_ifp->if_flags & IFF_LOOPBACK)) {
294 			rt->rt_refcnt--;
295 			in6_ifloop_request(RTM_DELETE, ifa);
296 		}
297 	}
298 }
299 
300 int
301 in6_ifindex2scopeid(int idx)
302 {
303 	struct ifnet *ifp;
304 	struct sockaddr_in6 *sin6;
305 	struct ifaddr_container *ifac;
306 
307 	if (idx < 0 || if_index < idx)
308 		return -1;
309 	ifp = ifindex2ifnet[idx];
310 
311 	TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link)
312 	{
313 		struct ifaddr *ifa = ifac->ifa;
314 
315 		if (ifa->ifa_addr->sa_family != AF_INET6)
316 			continue;
317 		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
318 		if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr))
319 			return sin6->sin6_scope_id & 0xffff;
320 	}
321 
322 	return -1;
323 }
324 
325 int
326 in6_mask2len(struct in6_addr *mask, u_char *lim0)
327 {
328 	int x = 0, y;
329 	u_char *lim = lim0, *p;
330 
331 	if (lim0 == NULL ||
332 	    lim0 - (u_char *)mask > sizeof(*mask)) /* ignore the scope_id part */
333 		lim = (u_char *)mask + sizeof(*mask);
334 	for (p = (u_char *)mask; p < lim; x++, p++) {
335 		if (*p != 0xff)
336 			break;
337 	}
338 	y = 0;
339 	if (p < lim) {
340 		for (y = 0; y < 8; y++) {
341 			if ((*p & (0x80 >> y)) == 0)
342 				break;
343 		}
344 	}
345 
346 	/*
347 	 * when the limit pointer is given, do a stricter check on the
348 	 * remaining bits.
349 	 */
350 	if (p < lim) {
351 		if (y != 0 && (*p & (0x00ff >> y)) != 0)
352 			return (-1);
353 		for (p = p + 1; p < lim; p++)
354 			if (*p != 0)
355 				return (-1);
356 	}
357 
358 	return x * 8 + y;
359 }
360 
361 void
362 in6_len2mask(struct in6_addr *mask, int len)
363 {
364 	int i;
365 
366 	bzero(mask, sizeof(*mask));
367 	for (i = 0; i < len / 8; i++)
368 		mask->s6_addr8[i] = 0xff;
369 	if (len % 8)
370 		mask->s6_addr8[i] = (0xff00 >> (len % 8)) & 0xff;
371 }
372 
373 #define ifa2ia6(ifa)	((struct in6_ifaddr *)(ifa))
374 #define ia62ifa(ia6)	(&((ia6)->ia_ifa))
375 
376 void
377 in6_control_dispatch(netmsg_t msg)
378 {
379 	int error;
380 
381 	error = in6_control(msg->control.base.nm_so,
382 			    msg->control.nm_cmd,
383 			    msg->control.nm_data,
384 			    msg->control.nm_ifp,
385 			    msg->control.nm_td);
386 	lwkt_replymsg(&msg->control.base.lmsg, error);
387 }
388 
389 int
390 in6_control(struct socket *so, u_long cmd, caddr_t data,
391 	    struct ifnet *ifp, struct thread *td)
392 {
393 	struct	in6_ifreq *ifr = (struct in6_ifreq *)data;
394 	struct	in6_ifaddr *ia = NULL;
395 	struct	in6_aliasreq *ifra = (struct in6_aliasreq *)data;
396 	struct	in6_ifextra *xtra;
397 	int privileged;
398 	int error;
399 
400 	privileged = 0;
401 	if (priv_check(td, PRIV_ROOT) == 0)
402 		privileged++;
403 
404 	switch (cmd) {
405 	case SIOCGETSGCNT_IN6:
406 	case SIOCGETMIFCNT_IN6:
407 		return (mrt6_ioctl(cmd, data));
408 	}
409 
410 	switch(cmd) {
411 	case SIOCAADDRCTL_POLICY:
412 	case SIOCDADDRCTL_POLICY:
413 		if (!privileged)
414 			return (EPERM);
415 		return (in6_src_ioctl(cmd, data));
416 	}
417 
418 	if (ifp == NULL)
419 		return (EOPNOTSUPP);
420 
421 	switch (cmd) {
422 	case SIOCSNDFLUSH_IN6:
423 	case SIOCSPFXFLUSH_IN6:
424 	case SIOCSRTRFLUSH_IN6:
425 	case SIOCSDEFIFACE_IN6:
426 	case SIOCSIFINFO_FLAGS:
427 		if (!privileged)
428 			return (EPERM);
429 		/* fall through */
430 	case OSIOCGIFINFO_IN6:
431 	case SIOCGIFINFO_IN6:
432 	case SIOCGDRLST_IN6:
433 	case SIOCGPRLST_IN6:
434 	case SIOCGNBRINFO_IN6:
435 	case SIOCGDEFIFACE_IN6:
436 		return (nd6_ioctl(cmd, data, ifp));
437 	}
438 
439 	switch (cmd) {
440 	case SIOCSIFPREFIX_IN6:
441 	case SIOCDIFPREFIX_IN6:
442 	case SIOCAIFPREFIX_IN6:
443 	case SIOCCIFPREFIX_IN6:
444 	case SIOCSGIFPREFIX_IN6:
445 	case SIOCGIFPREFIX_IN6:
446 		log(LOG_NOTICE,
447 		    "prefix ioctls are now invalidated. "
448 		    "please use ifconfig.\n");
449 		return (EOPNOTSUPP);
450 	}
451 
452 	switch (cmd) {
453 	case SIOCSSCOPE6:
454 		if (!privileged)
455 			return (EPERM);
456 		return (scope6_set(ifp,
457 			(struct scope6_id *)ifr->ifr_ifru.ifru_scope_id));
458 		break;
459 	case SIOCGSCOPE6:
460 		return (scope6_get(ifp,
461 			(struct scope6_id *)ifr->ifr_ifru.ifru_scope_id));
462 		break;
463 	case SIOCGSCOPE6DEF:
464 		return (scope6_get_default((struct scope6_id *)
465 			ifr->ifr_ifru.ifru_scope_id));
466 		break;
467 	}
468 
469 	switch (cmd) {
470 	case SIOCALIFADDR:
471 	case SIOCDLIFADDR:
472 		if (!privileged)
473 			return (EPERM);
474 		/* fall through */
475 	case SIOCGLIFADDR:
476 		return in6_lifaddr_ioctl(so, cmd, data, ifp, td);
477 	}
478 
479 	/*
480 	 * Find address for this interface, if it exists.
481 	 */
482 	if (ifra->ifra_addr.sin6_family == AF_INET6) { /* XXX */
483 		struct sockaddr_in6 *sa6 =
484 			(struct sockaddr_in6 *)&ifra->ifra_addr;
485 
486 		if (IN6_IS_ADDR_LINKLOCAL(&sa6->sin6_addr)) {
487 			if (sa6->sin6_addr.s6_addr16[1] == 0) {
488 				/* link ID is not embedded by the user */
489 				sa6->sin6_addr.s6_addr16[1] =
490 					htons(ifp->if_index);
491 			} else if (sa6->sin6_addr.s6_addr16[1] !=
492 				    htons(ifp->if_index)) {
493 				return (EINVAL);	/* link ID contradicts */
494 			}
495 			if (sa6->sin6_scope_id) {
496 				if (sa6->sin6_scope_id !=
497 				    (u_int32_t)ifp->if_index)
498 					return (EINVAL);
499 				sa6->sin6_scope_id = 0; /* XXX: good way? */
500 			}
501 		}
502 		ia = in6ifa_ifpwithaddr(ifp, &ifra->ifra_addr.sin6_addr);
503 	}
504 
505 	switch (cmd) {
506 	case SIOCSIFADDR_IN6:
507 	case SIOCSIFDSTADDR_IN6:
508 	case SIOCSIFNETMASK_IN6:
509 		/*
510 		 * Since IPv6 allows a node to assign multiple addresses
511 		 * on a single interface, SIOCSIFxxx ioctls are not suitable
512 		 * and should be unused.
513 		 */
514 		/* we decided to obsolete this command (20000704) */
515 		return (EINVAL);
516 
517 	case SIOCDIFADDR_IN6:
518 		/*
519 		 * for IPv4, we look for existing in_ifaddr here to allow
520 		 * "ifconfig if0 delete" to remove first IPv4 address on the
521 		 * interface.  For IPv6, as the spec allow multiple interface
522 		 * address from the day one, we consider "remove the first one"
523 		 * semantics to be not preferable.
524 		 */
525 		if (ia == NULL)
526 			return (EADDRNOTAVAIL);
527 		/* FALLTHROUGH */
528 	case SIOCAIFADDR_IN6:
529 		/*
530 		 * We always require users to specify a valid IPv6 address for
531 		 * the corresponding operation.
532 		 */
533 		if (ifra->ifra_addr.sin6_family != AF_INET6 ||
534 		    ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6))
535 			return (EAFNOSUPPORT);
536 		if (!privileged)
537 			return (EPERM);
538 
539 		break;
540 
541 	case SIOCGIFADDR_IN6:
542 		/* This interface is basically deprecated. use SIOCGIFCONF. */
543 		/* fall through */
544 	case SIOCGIFAFLAG_IN6:
545 	case SIOCGIFNETMASK_IN6:
546 	case SIOCGIFDSTADDR_IN6:
547 	case SIOCGIFALIFETIME_IN6:
548 		/* must think again about its semantics */
549 		if (ia == NULL)
550 			return (EADDRNOTAVAIL);
551 		break;
552 	case SIOCSIFALIFETIME_IN6:
553 	    {
554 		struct in6_addrlifetime *lt;
555 
556 		if (!privileged)
557 			return (EPERM);
558 		if (ia == NULL)
559 			return (EADDRNOTAVAIL);
560 		/* sanity for overflow - beware unsigned */
561 		lt = &ifr->ifr_ifru.ifru_lifetime;
562 		if (lt->ia6t_vltime != ND6_INFINITE_LIFETIME
563 		    && lt->ia6t_vltime + time_uptime < time_uptime) {
564 			return EINVAL;
565 		}
566 		if (lt->ia6t_pltime != ND6_INFINITE_LIFETIME
567 		    && lt->ia6t_pltime + time_uptime < time_uptime) {
568 			return EINVAL;
569 		}
570 		break;
571 	    }
572 	}
573 
574 	switch (cmd) {
575 
576 	case SIOCGIFADDR_IN6:
577 		ifr->ifr_addr = ia->ia_addr;
578 		break;
579 
580 	case SIOCGIFDSTADDR_IN6:
581 		if (!(ifp->if_flags & IFF_POINTOPOINT))
582 			return (EINVAL);
583 		/*
584 		 * XXX: should we check if ifa_dstaddr is NULL and return
585 		 * an error?
586 		 */
587 		ifr->ifr_dstaddr = ia->ia_dstaddr;
588 		break;
589 
590 	case SIOCGIFNETMASK_IN6:
591 		ifr->ifr_addr = ia->ia_prefixmask;
592 		break;
593 
594 	case SIOCGIFAFLAG_IN6:
595 		ifr->ifr_ifru.ifru_flags6 = ia->ia6_flags;
596 		break;
597 
598 	case SIOCGIFSTAT_IN6:
599 		if (ifp == NULL || (xtra = ifp->if_afdata[AF_INET6]) == NULL)
600 			return EINVAL;
601 		bzero(&ifr->ifr_ifru.ifru_stat,
602 		      sizeof(ifr->ifr_ifru.ifru_stat));
603 		ifr->ifr_ifru.ifru_stat = *xtra->in6_ifstat;
604 		break;
605 
606 	case SIOCGIFSTAT_ICMP6:
607 		if (ifp == NULL || (xtra = ifp->if_afdata[AF_INET6]) == NULL)
608 			return EINVAL;
609 		bzero(&ifr->ifr_ifru.ifru_stat,
610 			sizeof(ifr->ifr_ifru.ifru_icmp6stat));
611 		ifr->ifr_ifru.ifru_icmp6stat = *xtra->icmp6_ifstat;
612 		break;
613 
614 	case SIOCSIFADDR:
615 	case SIOCSIFDSTADDR:
616 	case SIOCSIFBRDADDR:
617 	case SIOCSIFNETMASK:
618 		/*
619 		 * Do not pass those ioctl to driver handler since they are not
620 		 * properly setup.  Instead just error out.
621 		 */
622 		return (EOPNOTSUPP);
623 
624 	case SIOCGIFALIFETIME_IN6:
625 		ifr->ifr_ifru.ifru_lifetime = ia->ia6_lifetime;
626 		break;
627 
628 	case SIOCSIFALIFETIME_IN6:
629 		ia->ia6_lifetime = ifr->ifr_ifru.ifru_lifetime;
630 		/* for sanity */
631 		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
632 			ia->ia6_lifetime.ia6t_expire =
633 				time_uptime + ia->ia6_lifetime.ia6t_vltime;
634 		} else
635 			ia->ia6_lifetime.ia6t_expire = 0;
636 		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
637 			ia->ia6_lifetime.ia6t_preferred =
638 				time_uptime + ia->ia6_lifetime.ia6t_pltime;
639 		} else
640 			ia->ia6_lifetime.ia6t_preferred = 0;
641 		break;
642 
643 	case SIOCAIFADDR_IN6:
644 	{
645 		int i, error = 0, iaIsNew;
646 		struct nd_prefix pr0, *pr;
647 
648 		if (ia != NULL)
649 			iaIsNew = 0;
650 		else
651 			iaIsNew = 1;
652 
653 		/*
654 		 * first, make or update the interface address structure,
655 		 * and link it to the list.
656 		 */
657 		if ((error = in6_update_ifa(ifp, ifra, ia)) != 0)
658 			return (error);
659 
660 		/*
661 		 * then, make the prefix on-link on the interface.
662 		 * XXX: we'd rather create the prefix before the address, but
663 		 * we need at least one address to install the corresponding
664 		 * interface route, so we configure the address first.
665 		 */
666 
667 		/*
668 		 * convert mask to prefix length (prefixmask has already
669 		 * been validated in in6_update_ifa().
670 		 */
671 		bzero(&pr0, sizeof(pr0));
672 		pr0.ndpr_ifp = ifp;
673 		pr0.ndpr_plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
674 					     NULL);
675 		if (pr0.ndpr_plen == 128)
676 			break;	/* we don't need to install a host route. */
677 		pr0.ndpr_prefix = ifra->ifra_addr;
678 		pr0.ndpr_mask = ifra->ifra_prefixmask.sin6_addr;
679 		/* apply the mask for safety. */
680 		for (i = 0; i < 4; i++) {
681 			pr0.ndpr_prefix.sin6_addr.s6_addr32[i] &=
682 				ifra->ifra_prefixmask.sin6_addr.s6_addr32[i];
683 		}
684 		/*
685 		 * XXX: since we don't have an API to set prefix (not address)
686 		 * lifetimes, we just use the same lifetimes as addresses.
687 		 * The (temporarily) installed lifetimes can be overridden by
688 		 * later advertised RAs (when accept_rtadv is non 0), which is
689 		 * an intended behavior.
690 		 */
691 		pr0.ndpr_raf_onlink = 1; /* should be configurable? */
692 		pr0.ndpr_raf_auto =
693 			((ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0);
694 		pr0.ndpr_vltime = ifra->ifra_lifetime.ia6t_vltime;
695 		pr0.ndpr_pltime = ifra->ifra_lifetime.ia6t_pltime;
696 
697 		/* add the prefix if there's one. */
698 		if ((pr = nd6_prefix_lookup(&pr0)) == NULL) {
699 			/*
700 			 * nd6_prelist_add will install the corresponding
701 			 * interface route.
702 			 */
703 			if ((error = nd6_prelist_add(&pr0, NULL, &pr)) != 0)
704 				return (error);
705 			if (pr == NULL) {
706 				log(LOG_ERR, "nd6_prelist_add succeeded but "
707 				    "no prefix\n");
708 				return (EINVAL); /* XXX panic here? */
709 			}
710 		}
711 		if ((ia = in6ifa_ifpwithaddr(ifp, &ifra->ifra_addr.sin6_addr))
712 		    == NULL) {
713 		    	/* XXX: this should not happen! */
714 			log(LOG_ERR, "in6_control: addition succeeded, but"
715 			    " no ifaddr\n");
716 		} else {
717 			if ((ia->ia6_flags & IN6_IFF_AUTOCONF) &&
718 			    ia->ia6_ndpr == NULL) { /* new autoconfed addr */
719 				ia->ia6_ndpr = pr;
720 				pr->ndpr_refcnt++;
721 
722 				/*
723 				 * If this is the first autoconf address from
724 				 * the prefix, create a temporary address
725 				 * as well (when specified).
726 				 */
727 				if (ip6_use_tempaddr &&
728 				    pr->ndpr_refcnt == 1) {
729 					int e;
730 					if ((e = in6_tmpifadd(ia, 1)) != 0) {
731 						log(LOG_NOTICE, "in6_control: "
732 						    "failed to create a "
733 						    "temporary address, "
734 						    "errno=%d\n",
735 						    e);
736 					}
737 				}
738 			}
739 
740 			/*
741 			 * this might affect the status of autoconfigured
742 			 * addresses, that is, this address might make
743 			 * other addresses detached.
744 			 */
745 			pfxlist_onlink_check();
746 		}
747 		if (error == 0 && ia) {
748 			EVENTHANDLER_INVOKE(ifaddr_event, ifp,
749 			iaIsNew ? IFADDR_EVENT_ADD : IFADDR_EVENT_CHANGE,
750 			&ia->ia_ifa);
751 		}
752 		break;
753 	}
754 
755 	case SIOCDIFADDR_IN6:
756 	{
757 		int i = 0;
758 		struct nd_prefix pr0, *pr;
759 
760 		/*
761 		 * If the address being deleted is the only one that owns
762 		 * the corresponding prefix, expire the prefix as well.
763 		 * XXX: theoretically, we don't have to warry about such
764 		 * relationship, since we separate the address management
765 		 * and the prefix management.  We do this, however, to provide
766 		 * as much backward compatibility as possible in terms of
767 		 * the ioctl operation.
768 		 */
769 		bzero(&pr0, sizeof(pr0));
770 		pr0.ndpr_ifp = ifp;
771 		pr0.ndpr_plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr,
772 					     NULL);
773 		if (pr0.ndpr_plen == 128)
774 			goto purgeaddr;
775 		pr0.ndpr_prefix = ia->ia_addr;
776 		pr0.ndpr_mask = ia->ia_prefixmask.sin6_addr;
777 		for (i = 0; i < 4; i++) {
778 			pr0.ndpr_prefix.sin6_addr.s6_addr32[i] &=
779 				ia->ia_prefixmask.sin6_addr.s6_addr32[i];
780 		}
781 		/*
782 		 * The logic of the following condition is a bit complicated.
783 		 * We expire the prefix when
784 		 * 1. the address obeys autoconfiguration and it is the
785 		 *    only owner of the associated prefix, or
786 		 * 2. the address does not obey autoconf and there is no
787 		 *    other owner of the prefix.
788 		 */
789 		if ((pr = nd6_prefix_lookup(&pr0)) != NULL &&
790 		    (((ia->ia6_flags & IN6_IFF_AUTOCONF) &&
791 		       pr->ndpr_refcnt == 1) ||
792 		     (!(ia->ia6_flags & IN6_IFF_AUTOCONF) &&
793 		      pr->ndpr_refcnt == 0))) {
794 			pr->ndpr_expire = 1; /* XXX: just for expiration */
795 		}
796 
797 purgeaddr:
798 		EVENTHANDLER_INVOKE(ifaddr_event, ifp, IFADDR_EVENT_DELETE,
799 				    &ia->ia_ifa);
800 		in6_purgeaddr(&ia->ia_ifa);
801 		break;
802 	}
803 
804 	default:
805 		if (ifp == NULL || ifp->if_ioctl == NULL)
806 			return (EOPNOTSUPP);
807 		ifnet_serialize_all(ifp);
808 		error = ifp->if_ioctl(ifp, cmd, data, td->td_proc->p_ucred);
809 		ifnet_deserialize_all(ifp);
810 		return (error);
811 	}
812 
813 	return (0);
814 }
815 
816 /*
817  * Update parameters of an IPv6 interface address.
818  * If necessary, a new entry is created and linked into address chains.
819  * This function is separated from in6_control().
820  * XXX: should this be performed under splnet()?
821  */
822 int
823 in6_update_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
824 	       struct in6_ifaddr *ia)
825 {
826 	int error = 0, hostIsNew = 0, plen = -1;
827 	struct in6_ifaddr *oia;
828 	struct sockaddr_in6 dst6;
829 	struct in6_addrlifetime *lt;
830 
831 	/* Validate parameters */
832 	if (ifp == NULL || ifra == NULL) /* this maybe redundant */
833 		return (EINVAL);
834 
835 	/*
836 	 * The destination address for a p2p link must have a family
837 	 * of AF_UNSPEC or AF_INET6.
838 	 */
839 	if ((ifp->if_flags & IFF_POINTOPOINT) &&
840 	    ifra->ifra_dstaddr.sin6_family != AF_INET6 &&
841 	    ifra->ifra_dstaddr.sin6_family != AF_UNSPEC)
842 		return (EAFNOSUPPORT);
843 	/*
844 	 * validate ifra_prefixmask.  don't check sin6_family, netmask
845 	 * does not carry fields other than sin6_len.
846 	 */
847 	if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6))
848 		return (EINVAL);
849 	/*
850 	 * Because the IPv6 address architecture is classless, we require
851 	 * users to specify a (non 0) prefix length (mask) for a new address.
852 	 * We also require the prefix (when specified) mask is valid, and thus
853 	 * reject a non-consecutive mask.
854 	 */
855 	if (ia == NULL && ifra->ifra_prefixmask.sin6_len == 0)
856 		return (EINVAL);
857 	if (ifra->ifra_prefixmask.sin6_len != 0) {
858 		plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
859 				    (u_char *)&ifra->ifra_prefixmask +
860 				    ifra->ifra_prefixmask.sin6_len);
861 		if (plen <= 0)
862 			return (EINVAL);
863 	}
864 	else {
865 		/*
866 		 * In this case, ia must not be NULL.  We just use its prefix
867 		 * length.
868 		 */
869 		plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
870 	}
871 	/*
872 	 * If the destination address on a p2p interface is specified,
873 	 * and the address is a scoped one, validate/set the scope
874 	 * zone identifier.
875 	 */
876 	dst6 = ifra->ifra_dstaddr;
877 	if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) &&
878 	    (dst6.sin6_family == AF_INET6)) {
879 		int scopeid;
880 
881 		if ((error = in6_recoverscope(&dst6,
882 					      &ifra->ifra_dstaddr.sin6_addr,
883 					      ifp)) != 0)
884 			return (error);
885 		scopeid = in6_addr2scopeid(ifp, &dst6.sin6_addr);
886 		if (dst6.sin6_scope_id == 0) /* user omit to specify the ID. */
887 			dst6.sin6_scope_id = scopeid;
888 		else if (dst6.sin6_scope_id != scopeid)
889 			return (EINVAL); /* scope ID mismatch. */
890 		if ((error = in6_embedscope(&dst6.sin6_addr, &dst6, NULL, NULL))
891 		    != 0)
892 			return (error);
893 		dst6.sin6_scope_id = 0; /* XXX */
894 	}
895 	/*
896 	 * The destination address can be specified only for a p2p or a
897 	 * loopback interface.  If specified, the corresponding prefix length
898 	 * must be 128.
899 	 */
900 	if (ifra->ifra_dstaddr.sin6_family == AF_INET6) {
901 		if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) == 0) {
902 			/* XXX: noisy message */
903 			log(LOG_INFO, "in6_update_ifa: a destination can be "
904 			    "specified for a p2p or a loopback IF only\n");
905 			return (EINVAL);
906 		}
907 		if (plen != 128) {
908 			/*
909 			 * The following message seems noisy, but we dare to
910 			 * add it for diagnosis.
911 			 */
912 			log(LOG_INFO, "in6_update_ifa: prefixlen must be 128 "
913 			    "when dstaddr is specified\n");
914 			return (EINVAL);
915 		}
916 	}
917 	/* lifetime consistency check */
918 	lt = &ifra->ifra_lifetime;
919 	if (lt->ia6t_vltime != ND6_INFINITE_LIFETIME
920 	    && lt->ia6t_vltime + time_uptime < time_uptime) {
921 		return EINVAL;
922 	}
923 	if (lt->ia6t_vltime == 0) {
924 		/*
925 		 * the following log might be noisy, but this is a typical
926 		 * configuration mistake or a tool's bug.
927 		 */
928 		log(LOG_INFO,
929 		    "in6_update_ifa: valid lifetime is 0 for %s\n",
930 		    ip6_sprintf(&ifra->ifra_addr.sin6_addr));
931 	}
932 	if (lt->ia6t_pltime != ND6_INFINITE_LIFETIME
933 	    && lt->ia6t_pltime + time_uptime < time_uptime) {
934 		return EINVAL;
935 	}
936 
937 	/*
938 	 * If this is a new address, allocate a new ifaddr and link it
939 	 * into chains.
940 	 */
941 	if (ia == NULL) {
942 		hostIsNew = 1;
943 		/*
944 		 * When in6_update_ifa() is called in a process of a received
945 		 * RA, it is called under splnet().  So, we should call malloc
946 		 * with M_NOWAIT.
947 		 */
948 		ia = ifa_create(sizeof(*ia), M_NOWAIT);
949 		if (ia == NULL)
950 			return (ENOBUFS);
951 		/* Initialize the address and masks */
952 		ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
953 		ia->ia_addr.sin6_family = AF_INET6;
954 		ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
955 		if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
956 			/*
957 			 * XXX: some functions expect that ifa_dstaddr is not
958 			 * NULL for p2p interfaces.
959 			 */
960 			ia->ia_ifa.ifa_dstaddr
961 				= (struct sockaddr *)&ia->ia_dstaddr;
962 		} else {
963 			ia->ia_ifa.ifa_dstaddr = NULL;
964 		}
965 		ia->ia_ifa.ifa_netmask
966 			= (struct sockaddr *)&ia->ia_prefixmask;
967 
968 		ia->ia_ifp = ifp;
969 		if ((oia = in6_ifaddr) != NULL) {
970 			for ( ; oia->ia_next; oia = oia->ia_next)
971 				continue;
972 			oia->ia_next = ia;
973 		} else
974 			in6_ifaddr = ia;
975 
976 		ifa_iflink(&ia->ia_ifa, ifp, 1);
977 	}
978 
979 	/* set prefix mask */
980 	if (ifra->ifra_prefixmask.sin6_len) {
981 		/*
982 		 * We prohibit changing the prefix length of an existing
983 		 * address, because
984 		 * + such an operation should be rare in IPv6, and
985 		 * + the operation would confuse prefix management.
986 		 */
987 		if (ia->ia_prefixmask.sin6_len &&
988 		    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) != plen) {
989 			log(LOG_INFO, "in6_update_ifa: the prefix length of an"
990 			    " existing (%s) address should not be changed\n",
991 			    ip6_sprintf(&ia->ia_addr.sin6_addr));
992 			error = EINVAL;
993 			goto unlink;
994 		}
995 		ia->ia_prefixmask = ifra->ifra_prefixmask;
996 	}
997 
998 	/*
999 	 * If a new destination address is specified, scrub the old one and
1000 	 * install the new destination.  Note that the interface must be
1001 	 * p2p or loopback (see the check above.)
1002 	 */
1003 	if (dst6.sin6_family == AF_INET6 &&
1004 	    !IN6_ARE_ADDR_EQUAL(&dst6.sin6_addr,
1005 				&ia->ia_dstaddr.sin6_addr)) {
1006 		int e;
1007 
1008 		if ((ia->ia_flags & IFA_ROUTE) &&
1009 		    (e = rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST))
1010 		    != 0) {
1011 			log(LOG_ERR, "in6_update_ifa: failed to remove "
1012 			    "a route to the old destination: %s\n",
1013 			    ip6_sprintf(&ia->ia_addr.sin6_addr));
1014 			/* proceed anyway... */
1015 		}
1016 		else
1017 			ia->ia_flags &= ~IFA_ROUTE;
1018 		ia->ia_dstaddr = dst6;
1019 	}
1020 
1021 	/* reset the interface and routing table appropriately. */
1022 	if ((error = in6_ifinit(ifp, ia, &ifra->ifra_addr, hostIsNew)) != 0)
1023 		goto unlink;
1024 
1025 	/*
1026 	 * Beyond this point, we should call in6_purgeaddr upon an error,
1027 	 * not just go to unlink.
1028 	 */
1029 
1030 #if 0				/* disable this mechanism for now */
1031 	/* update prefix list */
1032 	if (hostIsNew &&
1033 	    (ifra->ifra_flags & IN6_IFF_NOPFX) == 0) { /* XXX */
1034 		int iilen;
1035 
1036 		iilen = (sizeof(ia->ia_prefixmask.sin6_addr) << 3) - plen;
1037 		if ((error = in6_prefix_add_ifid(iilen, ia)) != 0) {
1038 			in6_purgeaddr((struct ifaddr *)ia);
1039 			return (error);
1040 		}
1041 	}
1042 #endif
1043 
1044 	if (ifp->if_flags & IFF_MULTICAST) {
1045 		struct sockaddr_in6 mltaddr, mltmask;
1046 		struct in6_multi *in6m;
1047 
1048 		if (hostIsNew) {
1049 			/*
1050 			 * join solicited multicast addr for new host id
1051 			 */
1052 			struct in6_addr llsol;
1053 			bzero(&llsol, sizeof(struct in6_addr));
1054 			llsol.s6_addr16[0] = htons(0xff02);
1055 			llsol.s6_addr16[1] = htons(ifp->if_index);
1056 			llsol.s6_addr32[1] = 0;
1057 			llsol.s6_addr32[2] = htonl(1);
1058 			llsol.s6_addr32[3] =
1059 				ifra->ifra_addr.sin6_addr.s6_addr32[3];
1060 			llsol.s6_addr8[12] = 0xff;
1061 			in6_addmulti(&llsol, ifp, &error);
1062 			if (error != 0) {
1063 				log(LOG_WARNING,
1064 				    "in6_update_ifa: addmulti failed for "
1065 				    "%s on %s (errno=%d)\n",
1066 				    ip6_sprintf(&llsol), if_name(ifp),
1067 				    error);
1068 				in6_purgeaddr((struct ifaddr *)ia);
1069 				return (error);
1070 			}
1071 		}
1072 
1073 		bzero(&mltmask, sizeof(mltmask));
1074 		mltmask.sin6_len = sizeof(struct sockaddr_in6);
1075 		mltmask.sin6_family = AF_INET6;
1076 		mltmask.sin6_addr = in6mask32;
1077 
1078 		/*
1079 		 * join link-local all-nodes address
1080 		 */
1081 		bzero(&mltaddr, sizeof(mltaddr));
1082 		mltaddr.sin6_len = sizeof(struct sockaddr_in6);
1083 		mltaddr.sin6_family = AF_INET6;
1084 		mltaddr.sin6_addr = kin6addr_linklocal_allnodes;
1085 		mltaddr.sin6_addr.s6_addr16[1] = htons(ifp->if_index);
1086 
1087 		IN6_LOOKUP_MULTI(mltaddr.sin6_addr, ifp, in6m);
1088 		if (in6m == NULL) {
1089 			rtrequest_global(RTM_ADD,
1090 				  (struct sockaddr *)&mltaddr,
1091 				  (struct sockaddr *)&ia->ia_addr,
1092 				  (struct sockaddr *)&mltmask,
1093 				  RTF_UP|RTF_CLONING);  /* xxx */
1094 			in6_addmulti(&mltaddr.sin6_addr, ifp, &error);
1095 			if (error != 0) {
1096 				log(LOG_WARNING,
1097 				    "in6_update_ifa: addmulti failed for "
1098 				    "%s on %s (errno=%d)\n",
1099 				    ip6_sprintf(&mltaddr.sin6_addr),
1100 				    if_name(ifp), error);
1101 			}
1102 		}
1103 
1104 		/*
1105 		 * join node information group address
1106 		 */
1107 #define hostnamelen	strlen(hostname)
1108 		if (in6_nigroup(ifp, hostname, hostnamelen, &mltaddr.sin6_addr)
1109 		    == 0) {
1110 			IN6_LOOKUP_MULTI(mltaddr.sin6_addr, ifp, in6m);
1111 			if (in6m == NULL && ia != NULL) {
1112 				in6_addmulti(&mltaddr.sin6_addr, ifp, &error);
1113 				if (error != 0) {
1114 					log(LOG_WARNING, "in6_update_ifa: "
1115 					    "addmulti failed for "
1116 					    "%s on %s (errno=%d)\n",
1117 					    ip6_sprintf(&mltaddr.sin6_addr),
1118 					    if_name(ifp), error);
1119 				}
1120 			}
1121 		}
1122 #undef hostnamelen
1123 
1124 		/*
1125 		 * join node-local all-nodes address, on loopback.
1126 		 * XXX: since "node-local" is obsoleted by interface-local,
1127 		 *      we have to join the group on every interface with
1128 		 *      some interface-boundary restriction.
1129 		 */
1130 		if (ifp->if_flags & IFF_LOOPBACK) {
1131 			struct in6_ifaddr *ia_loop;
1132 
1133 			struct in6_addr loop6 = kin6addr_loopback;
1134 			ia_loop = in6ifa_ifpwithaddr(ifp, &loop6);
1135 
1136 			mltaddr.sin6_addr = kin6addr_nodelocal_allnodes;
1137 
1138 			IN6_LOOKUP_MULTI(mltaddr.sin6_addr, ifp, in6m);
1139 			if (in6m == NULL && ia_loop != NULL) {
1140 				rtrequest_global(RTM_ADD,
1141 					  (struct sockaddr *)&mltaddr,
1142 					  (struct sockaddr *)&ia_loop->ia_addr,
1143 					  (struct sockaddr *)&mltmask,
1144 					  RTF_UP);
1145 				in6_addmulti(&mltaddr.sin6_addr, ifp, &error);
1146 				if (error != 0) {
1147 					log(LOG_WARNING, "in6_update_ifa: "
1148 					    "addmulti failed for %s on %s "
1149 					    "(errno=%d)\n",
1150 					    ip6_sprintf(&mltaddr.sin6_addr),
1151 					    if_name(ifp), error);
1152 				}
1153 			}
1154 		}
1155 	}
1156 
1157 	ia->ia6_flags = ifra->ifra_flags;
1158 	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/*safety*/
1159 	ia->ia6_flags &= ~IN6_IFF_NODAD;	/* Mobile IPv6 */
1160 
1161 	ia->ia6_lifetime = ifra->ifra_lifetime;
1162 	/* for sanity */
1163 	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1164 		ia->ia6_lifetime.ia6t_expire =
1165 			time_uptime + ia->ia6_lifetime.ia6t_vltime;
1166 	} else
1167 		ia->ia6_lifetime.ia6t_expire = 0;
1168 	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1169 		ia->ia6_lifetime.ia6t_preferred =
1170 			time_uptime + ia->ia6_lifetime.ia6t_pltime;
1171 	} else
1172 		ia->ia6_lifetime.ia6t_preferred = 0;
1173 
1174 	/*
1175 	 * Perform DAD, if needed.
1176 	 * XXX It may be of use, if we can administratively
1177 	 * disable DAD.
1178 	 */
1179 	if (in6if_do_dad(ifp) && !(ifra->ifra_flags & IN6_IFF_NODAD)) {
1180 		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1181 		nd6_dad_start((struct ifaddr *)ia, NULL);
1182 	}
1183 
1184 	return (error);
1185 
1186 unlink:
1187 	/*
1188 	 * XXX: if a change of an existing address failed, keep the entry
1189 	 * anyway.
1190 	 */
1191 	if (hostIsNew)
1192 		in6_unlink_ifa(ia, ifp);
1193 	return (error);
1194 }
1195 
1196 void
1197 in6_purgeaddr(struct ifaddr *ifa)
1198 {
1199 	struct ifnet *ifp = ifa->ifa_ifp;
1200 	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1201 
1202 	/* stop DAD processing */
1203 	nd6_dad_stop(ifa);
1204 
1205 	/*
1206 	 * delete route to the destination of the address being purged.
1207 	 * The interface must be p2p or loopback in this case.
1208 	 */
1209 	if ((ia->ia_flags & IFA_ROUTE) && ia->ia_dstaddr.sin6_len != 0) {
1210 		int e;
1211 
1212 		if ((e = rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST))
1213 		    != 0) {
1214 			log(LOG_ERR, "in6_purgeaddr: failed to remove "
1215 			    "a route to the p2p destination: %s on %s, "
1216 			    "errno=%d\n",
1217 			    ip6_sprintf(&ia->ia_addr.sin6_addr), if_name(ifp),
1218 			    e);
1219 			/* proceed anyway... */
1220 		}
1221 		else
1222 			ia->ia_flags &= ~IFA_ROUTE;
1223 	}
1224 
1225 	/* Remove ownaddr's loopback rtentry, if it exists. */
1226 	in6_ifremloop(&(ia->ia_ifa));
1227 
1228 	if (ifp->if_flags & IFF_MULTICAST) {
1229 		/*
1230 		 * delete solicited multicast addr for deleting host id
1231 		 */
1232 		struct in6_multi *in6m;
1233 		struct in6_addr llsol;
1234 		bzero(&llsol, sizeof(struct in6_addr));
1235 		llsol.s6_addr16[0] = htons(0xff02);
1236 		llsol.s6_addr16[1] = htons(ifp->if_index);
1237 		llsol.s6_addr32[1] = 0;
1238 		llsol.s6_addr32[2] = htonl(1);
1239 		llsol.s6_addr32[3] =
1240 			ia->ia_addr.sin6_addr.s6_addr32[3];
1241 		llsol.s6_addr8[12] = 0xff;
1242 
1243 		IN6_LOOKUP_MULTI(llsol, ifp, in6m);
1244 		if (in6m)
1245 			in6_delmulti(in6m);
1246 	}
1247 
1248 	in6_unlink_ifa(ia, ifp);
1249 }
1250 
1251 static void
1252 in6_unlink_ifa(struct in6_ifaddr *ia, struct ifnet *ifp)
1253 {
1254 	int plen, iilen;
1255 	struct in6_ifaddr *oia;
1256 
1257 	crit_enter();
1258 
1259 	ifa_ifunlink(&ia->ia_ifa, ifp);
1260 
1261 	oia = ia;
1262 	if (oia == (ia = in6_ifaddr))
1263 		in6_ifaddr = ia->ia_next;
1264 	else {
1265 		while (ia->ia_next && (ia->ia_next != oia))
1266 			ia = ia->ia_next;
1267 		if (ia->ia_next)
1268 			ia->ia_next = oia->ia_next;
1269 		else {
1270 			/* search failed */
1271 			kprintf("Couldn't unlink in6_ifaddr from in6_ifaddr\n");
1272 		}
1273 	}
1274 
1275 	if (oia->ia6_ifpr) {	/* check for safety */
1276 		plen = in6_mask2len(&oia->ia_prefixmask.sin6_addr, NULL);
1277 		iilen = (sizeof(oia->ia_prefixmask.sin6_addr) << 3) - plen;
1278 		in6_prefix_remove_ifid(iilen, oia);
1279 	}
1280 
1281 	/*
1282 	 * When an autoconfigured address is being removed, release the
1283 	 * reference to the base prefix.  Also, since the release might
1284 	 * affect the status of other (detached) addresses, call
1285 	 * pfxlist_onlink_check().
1286 	 */
1287 	if (oia->ia6_flags & IN6_IFF_AUTOCONF) {
1288 		if (oia->ia6_ndpr == NULL) {
1289 			log(LOG_NOTICE, "in6_unlink_ifa: autoconf'ed address "
1290 			    "%p has no prefix\n", oia);
1291 		} else {
1292 			oia->ia6_ndpr->ndpr_refcnt--;
1293 			oia->ia6_flags &= ~IN6_IFF_AUTOCONF;
1294 			oia->ia6_ndpr = NULL;
1295 		}
1296 
1297 		pfxlist_onlink_check();
1298 	}
1299 
1300 	/*
1301 	 * release another refcnt for the link from in6_ifaddr.
1302 	 * Note that we should decrement the refcnt at least once for all *BSD.
1303 	 */
1304 	ifa_destroy(&oia->ia_ifa);
1305 
1306 	crit_exit();
1307 }
1308 
1309 void
1310 in6_purgeif(struct ifnet *ifp)
1311 {
1312 	struct ifaddr_container *ifac, *next;
1313 
1314 	TAILQ_FOREACH_MUTABLE(ifac, &ifp->if_addrheads[mycpuid],
1315 			      ifa_link, next) {
1316 		if (ifac->ifa->ifa_addr->sa_family != AF_INET6)
1317 			continue;
1318 		in6_purgeaddr(ifac->ifa);
1319 	}
1320 
1321 	in6_ifdetach(ifp);
1322 }
1323 
1324 /*
1325  * SIOC[GAD]LIFADDR.
1326  *	SIOCGLIFADDR: get first address. (?)
1327  *	SIOCGLIFADDR with IFLR_PREFIX:
1328  *		get first address that matches the specified prefix.
1329  *	SIOCALIFADDR: add the specified address.
1330  *	SIOCALIFADDR with IFLR_PREFIX:
1331  *		add the specified prefix, filling hostid part from
1332  *		the first link-local address.  prefixlen must be <= 64.
1333  *	SIOCDLIFADDR: delete the specified address.
1334  *	SIOCDLIFADDR with IFLR_PREFIX:
1335  *		delete the first address that matches the specified prefix.
1336  * return values:
1337  *	EINVAL on invalid parameters
1338  *	EADDRNOTAVAIL on prefix match failed/specified address not found
1339  *	other values may be returned from in6_ioctl()
1340  *
1341  * NOTE: SIOCALIFADDR(with IFLR_PREFIX set) allows prefixlen less than 64.
1342  * this is to accomodate address naming scheme other than RFC2374,
1343  * in the future.
1344  * RFC2373 defines interface id to be 64bit, but it allows non-RFC2374
1345  * address encoding scheme. (see figure on page 8)
1346  */
1347 static int
1348 in6_lifaddr_ioctl(struct socket *so, u_long cmd, caddr_t data,
1349 		  struct ifnet *ifp, struct thread *td)
1350 {
1351 	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
1352 	struct sockaddr *sa;
1353 
1354 	/* sanity checks */
1355 	if (!data || !ifp) {
1356 		panic("invalid argument to in6_lifaddr_ioctl");
1357 		/*NOTRECHED*/
1358 	}
1359 
1360 	switch (cmd) {
1361 	case SIOCGLIFADDR:
1362 		/* address must be specified on GET with IFLR_PREFIX */
1363 		if (!(iflr->flags & IFLR_PREFIX))
1364 			break;
1365 		/* FALLTHROUGH */
1366 	case SIOCALIFADDR:
1367 	case SIOCDLIFADDR:
1368 		/* address must be specified on ADD and DELETE */
1369 		sa = (struct sockaddr *)&iflr->addr;
1370 		if (sa->sa_family != AF_INET6)
1371 			return EINVAL;
1372 		if (sa->sa_len != sizeof(struct sockaddr_in6))
1373 			return EINVAL;
1374 		/* XXX need improvement */
1375 		sa = (struct sockaddr *)&iflr->dstaddr;
1376 		if (sa->sa_family && sa->sa_family != AF_INET6)
1377 			return EINVAL;
1378 		if (sa->sa_len && sa->sa_len != sizeof(struct sockaddr_in6))
1379 			return EINVAL;
1380 		break;
1381 	default: /* shouldn't happen */
1382 #if 0
1383 		panic("invalid cmd to in6_lifaddr_ioctl");
1384 		/* NOTREACHED */
1385 #else
1386 		return EOPNOTSUPP;
1387 #endif
1388 	}
1389 	if (sizeof(struct in6_addr) * 8 < iflr->prefixlen)
1390 		return EINVAL;
1391 
1392 	switch (cmd) {
1393 	case SIOCALIFADDR:
1394 	    {
1395 		struct in6_aliasreq ifra;
1396 		struct in6_addr *hostid = NULL;
1397 		int prefixlen;
1398 
1399 		if (iflr->flags & IFLR_PREFIX) {
1400 			struct ifaddr *ifa;
1401 			struct sockaddr_in6 *sin6;
1402 
1403 			/*
1404 			 * hostid is to fill in the hostid part of the
1405 			 * address.  hostid points to the first link-local
1406 			 * address attached to the interface.
1407 			 */
1408 			ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0);
1409 			if (!ifa)
1410 				return EADDRNOTAVAIL;
1411 			hostid = IFA_IN6(ifa);
1412 
1413 		 	/* prefixlen must be <= 64. */
1414 			if (64 < iflr->prefixlen)
1415 				return EINVAL;
1416 			prefixlen = iflr->prefixlen;
1417 
1418 			/* hostid part must be zero. */
1419 			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1420 			if (sin6->sin6_addr.s6_addr32[2] != 0
1421 			 || sin6->sin6_addr.s6_addr32[3] != 0) {
1422 				return EINVAL;
1423 			}
1424 		} else
1425 			prefixlen = iflr->prefixlen;
1426 
1427 		/* copy args to in6_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
1428 		bzero(&ifra, sizeof(ifra));
1429 		bcopy(iflr->iflr_name, ifra.ifra_name,
1430 			sizeof(ifra.ifra_name));
1431 
1432 		bcopy(&iflr->addr, &ifra.ifra_addr,
1433 			((struct sockaddr *)&iflr->addr)->sa_len);
1434 		if (hostid) {
1435 			/* fill in hostid part */
1436 			ifra.ifra_addr.sin6_addr.s6_addr32[2] =
1437 				hostid->s6_addr32[2];
1438 			ifra.ifra_addr.sin6_addr.s6_addr32[3] =
1439 				hostid->s6_addr32[3];
1440 		}
1441 
1442 		if (((struct sockaddr *)&iflr->dstaddr)->sa_family) {	/*XXX*/
1443 			bcopy(&iflr->dstaddr, &ifra.ifra_dstaddr,
1444 				((struct sockaddr *)&iflr->dstaddr)->sa_len);
1445 			if (hostid) {
1446 				ifra.ifra_dstaddr.sin6_addr.s6_addr32[2] =
1447 					hostid->s6_addr32[2];
1448 				ifra.ifra_dstaddr.sin6_addr.s6_addr32[3] =
1449 					hostid->s6_addr32[3];
1450 			}
1451 		}
1452 
1453 		ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
1454 		in6_len2mask(&ifra.ifra_prefixmask.sin6_addr, prefixlen);
1455 
1456 		ifra.ifra_flags = iflr->flags & ~IFLR_PREFIX;
1457 		return in6_control(so, SIOCAIFADDR_IN6, (caddr_t)&ifra, ifp, td);
1458 	    }
1459 	case SIOCGLIFADDR:
1460 	case SIOCDLIFADDR:
1461 	    {
1462 		struct ifaddr_container *ifac;
1463 		struct in6_ifaddr *ia;
1464 		struct in6_addr mask, candidate, match;
1465 		struct sockaddr_in6 *sin6;
1466 		int cmp;
1467 
1468 		bzero(&mask, sizeof(mask));
1469 		if (iflr->flags & IFLR_PREFIX) {
1470 			/* lookup a prefix rather than address. */
1471 			in6_len2mask(&mask, iflr->prefixlen);
1472 
1473 			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1474 			bcopy(&sin6->sin6_addr, &match, sizeof(match));
1475 			match.s6_addr32[0] &= mask.s6_addr32[0];
1476 			match.s6_addr32[1] &= mask.s6_addr32[1];
1477 			match.s6_addr32[2] &= mask.s6_addr32[2];
1478 			match.s6_addr32[3] &= mask.s6_addr32[3];
1479 
1480 			/* if you set extra bits, that's wrong */
1481 			if (bcmp(&match, &sin6->sin6_addr, sizeof(match)))
1482 				return EINVAL;
1483 
1484 			cmp = 1;
1485 		} else {
1486 			if (cmd == SIOCGLIFADDR) {
1487 				/* on getting an address, take the 1st match */
1488 				cmp = 0;	/* XXX */
1489 			} else {
1490 				/* on deleting an address, do exact match */
1491 				in6_len2mask(&mask, 128);
1492 				sin6 = (struct sockaddr_in6 *)&iflr->addr;
1493 				bcopy(&sin6->sin6_addr, &match, sizeof(match));
1494 
1495 				cmp = 1;
1496 			}
1497 		}
1498 
1499 		TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
1500 			struct ifaddr *ifa = ifac->ifa;
1501 
1502 			if (ifa->ifa_addr->sa_family != AF_INET6)
1503 				continue;
1504 			if (!cmp)
1505 				break;
1506 
1507 			bcopy(IFA_IN6(ifa), &candidate, sizeof(candidate));
1508 			/*
1509 			 * XXX: this is adhoc, but is necessary to allow
1510 			 * a user to specify fe80::/64 (not /10) for a
1511 			 * link-local address.
1512 			 */
1513 			if (IN6_IS_ADDR_LINKLOCAL(&candidate))
1514 				candidate.s6_addr16[1] = 0;
1515 			candidate.s6_addr32[0] &= mask.s6_addr32[0];
1516 			candidate.s6_addr32[1] &= mask.s6_addr32[1];
1517 			candidate.s6_addr32[2] &= mask.s6_addr32[2];
1518 			candidate.s6_addr32[3] &= mask.s6_addr32[3];
1519 			if (IN6_ARE_ADDR_EQUAL(&candidate, &match))
1520 				break;
1521 		}
1522 		if (ifac == NULL)
1523 			return EADDRNOTAVAIL;
1524 		ia = ifa2ia6(ifac->ifa);
1525 
1526 		if (cmd == SIOCGLIFADDR) {
1527 			struct sockaddr_in6 *s6;
1528 
1529 			/* fill in the if_laddrreq structure */
1530 			bcopy(&ia->ia_addr, &iflr->addr, ia->ia_addr.sin6_len);
1531 			s6 = (struct sockaddr_in6 *)&iflr->addr;
1532 			if (IN6_IS_ADDR_LINKLOCAL(&s6->sin6_addr)) {
1533 				s6->sin6_addr.s6_addr16[1] = 0;
1534 				s6->sin6_scope_id =
1535 					in6_addr2scopeid(ifp, &s6->sin6_addr);
1536 			}
1537 			if (ifp->if_flags & IFF_POINTOPOINT) {
1538 				bcopy(&ia->ia_dstaddr, &iflr->dstaddr,
1539 					ia->ia_dstaddr.sin6_len);
1540 				s6 = (struct sockaddr_in6 *)&iflr->dstaddr;
1541 				if (IN6_IS_ADDR_LINKLOCAL(&s6->sin6_addr)) {
1542 					s6->sin6_addr.s6_addr16[1] = 0;
1543 					s6->sin6_scope_id =
1544 						in6_addr2scopeid(ifp,
1545 								 &s6->sin6_addr);
1546 				}
1547 			} else
1548 				bzero(&iflr->dstaddr, sizeof(iflr->dstaddr));
1549 
1550 			iflr->prefixlen =
1551 				in6_mask2len(&ia->ia_prefixmask.sin6_addr,
1552 					     NULL);
1553 
1554 			iflr->flags = ia->ia6_flags;	/* XXX */
1555 
1556 			return 0;
1557 		} else {
1558 			struct in6_aliasreq ifra;
1559 
1560 			/* fill in6_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
1561 			bzero(&ifra, sizeof(ifra));
1562 			bcopy(iflr->iflr_name, ifra.ifra_name,
1563 			      sizeof(ifra.ifra_name));
1564 
1565 			bcopy(&ia->ia_addr, &ifra.ifra_addr,
1566 			      ia->ia_addr.sin6_len);
1567 			if (ifp->if_flags & IFF_POINTOPOINT)
1568 				bcopy(&ia->ia_dstaddr, &ifra.ifra_dstaddr,
1569 				      ia->ia_dstaddr.sin6_len);
1570 			else
1571 				bzero(&ifra.ifra_dstaddr,
1572 				      sizeof(ifra.ifra_dstaddr));
1573 			bcopy(&ia->ia_prefixmask, &ifra.ifra_dstaddr,
1574 			      ia->ia_prefixmask.sin6_len);
1575 
1576 			ifra.ifra_flags = ia->ia6_flags;
1577 			return in6_control(so, SIOCDIFADDR_IN6, (caddr_t)&ifra,
1578 				ifp, td);
1579 		}
1580 	    }
1581 	}
1582 
1583 	return EOPNOTSUPP;	/* just for safety */
1584 }
1585 
1586 /*
1587  * Initialize an interface's intetnet6 address
1588  * and routing table entry.
1589  */
1590 static int
1591 in6_ifinit(struct ifnet *ifp, struct in6_ifaddr *ia, struct sockaddr_in6 *sin6,
1592 	   int newhost)
1593 {
1594 	int	error = 0, plen, ifacount = 0;
1595 	struct ifaddr_container *ifac;
1596 
1597 	/*
1598 	 * Give the interface a chance to initialize
1599 	 * if this is its first address,
1600 	 * and to validate the address if necessary.
1601 	 */
1602 	TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
1603 		if (ifac->ifa->ifa_addr == NULL)
1604 			continue;	/* just for safety */
1605 		if (ifac->ifa->ifa_addr->sa_family != AF_INET6)
1606 			continue;
1607 		ifacount++;
1608 	}
1609 
1610 	ifnet_serialize_all(ifp);
1611 
1612 	ia->ia_addr = *sin6;
1613 
1614 	if (ifacount <= 1 && ifp->if_ioctl &&
1615 	    (error = ifp->if_ioctl(ifp, SIOCSIFADDR, (caddr_t)ia, NULL))) {
1616 		ifnet_deserialize_all(ifp);
1617 		return (error);
1618 	}
1619 
1620 	ifnet_deserialize_all(ifp);
1621 
1622 	ia->ia_ifa.ifa_metric = ifp->if_metric;
1623 
1624 	/* we could do in(6)_socktrim here, but just omit it at this moment. */
1625 
1626 	/*
1627 	 * Special case:
1628 	 * If the destination address is specified for a point-to-point
1629 	 * interface, install a route to the destination as an interface
1630 	 * direct route.
1631 	 */
1632 	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1633 	if (plen == 128 && ia->ia_dstaddr.sin6_family == AF_INET6) {
1634 		if ((error = rtinit(&(ia->ia_ifa), (int)RTM_ADD,
1635 				    RTF_UP | RTF_HOST)) != 0)
1636 			return (error);
1637 		ia->ia_flags |= IFA_ROUTE;
1638 	}
1639 	if (plen < 128) {
1640 		/*
1641 		 * The RTF_CLONING flag is necessary for in6_is_ifloop_auto().
1642 		 */
1643 		ia->ia_ifa.ifa_flags |= RTF_CLONING;
1644 	}
1645 
1646 	/* Add ownaddr as loopback rtentry, if necessary (ex. on p2p link). */
1647 	if (newhost) {
1648 		/* set the rtrequest function to create llinfo */
1649 		ia->ia_ifa.ifa_rtrequest = nd6_rtrequest;
1650 		in6_ifaddloop(&(ia->ia_ifa));
1651 	}
1652 
1653 	return (error);
1654 }
1655 
1656 struct in6_multi_mship *
1657 in6_joingroup(struct ifnet *ifp, struct in6_addr *addr, int *errorp)
1658 {
1659        struct in6_multi_mship *imm;
1660 
1661        imm = kmalloc(sizeof(*imm), M_IPMADDR, M_NOWAIT);
1662        if (!imm) {
1663                *errorp = ENOBUFS;
1664                return NULL;
1665        }
1666        imm->i6mm_maddr = in6_addmulti(addr, ifp, errorp);
1667        if (!imm->i6mm_maddr) {
1668                /* *errorp is alrady set */
1669                kfree(imm, M_IPMADDR);
1670                return NULL;
1671        }
1672        return imm;
1673 }
1674 
1675 int
1676 in6_leavegroup(struct in6_multi_mship *imm)
1677 {
1678 
1679        if (imm->i6mm_maddr)
1680                in6_delmulti(imm->i6mm_maddr);
1681        kfree(imm,  M_IPMADDR);
1682        return 0;
1683 }
1684 
1685 /*
1686  * Add an address to the list of IP6 multicast addresses for a
1687  * given interface.
1688  */
1689 struct	in6_multi *
1690 in6_addmulti(struct in6_addr *maddr6, struct ifnet *ifp, int *errorp)
1691 {
1692 	struct	in6_multi *in6m;
1693 	struct sockaddr_in6 sin6;
1694 	struct ifmultiaddr *ifma;
1695 
1696 	*errorp = 0;
1697 
1698 	crit_enter();
1699 
1700 	/*
1701 	 * Call generic routine to add membership or increment
1702 	 * refcount.  It wants addresses in the form of a sockaddr,
1703 	 * so we build one here (being careful to zero the unused bytes).
1704 	 */
1705 	bzero(&sin6, sizeof sin6);
1706 	sin6.sin6_family = AF_INET6;
1707 	sin6.sin6_len = sizeof sin6;
1708 	sin6.sin6_addr = *maddr6;
1709 	*errorp = if_addmulti(ifp, (struct sockaddr *)&sin6, &ifma);
1710 	if (*errorp) {
1711 		crit_exit();
1712 		return 0;
1713 	}
1714 
1715 	/*
1716 	 * If ifma->ifma_protospec is null, then if_addmulti() created
1717 	 * a new record.  Otherwise, we are done.
1718 	 */
1719 	if (ifma->ifma_protospec != NULL) {
1720 		crit_exit();
1721 		return ifma->ifma_protospec;
1722 	}
1723 
1724 	/* XXX - if_addmulti uses M_WAITOK.  Can this really be called
1725 	   at interrupt time?  If so, need to fix if_addmulti. XXX */
1726 	in6m = (struct in6_multi *)kmalloc(sizeof(*in6m), M_IPMADDR, M_NOWAIT);
1727 	if (in6m == NULL) {
1728 		crit_exit();
1729 		return (NULL);
1730 	}
1731 
1732 	bzero(in6m, sizeof *in6m);
1733 	in6m->in6m_addr = *maddr6;
1734 	in6m->in6m_ifp = ifp;
1735 	in6m->in6m_ifma = ifma;
1736 	ifma->ifma_protospec = in6m;
1737 	LIST_INSERT_HEAD(&in6_multihead, in6m, in6m_entry);
1738 
1739 	/*
1740 	 * Let MLD6 know that we have joined a new IP6 multicast
1741 	 * group.
1742 	 */
1743 	mld6_start_listening(in6m);
1744 	crit_exit();
1745 	return (in6m);
1746 }
1747 
1748 /*
1749  * Delete a multicast address record.
1750  */
1751 void
1752 in6_delmulti(struct in6_multi *in6m)
1753 {
1754 	struct ifmultiaddr *ifma = in6m->in6m_ifma;
1755 
1756 	crit_enter();
1757 
1758 	if (ifma->ifma_refcount == 1) {
1759 		/*
1760 		 * No remaining claims to this record; let MLD6 know
1761 		 * that we are leaving the multicast group.
1762 		 */
1763 		mld6_stop_listening(in6m);
1764 		ifma->ifma_protospec = NULL;
1765 		LIST_REMOVE(in6m, in6m_entry);
1766 		kfree(in6m, M_IPMADDR);
1767 	}
1768 	/* XXX - should be separate API for when we have an ifma? */
1769 	if_delmulti(ifma->ifma_ifp, ifma->ifma_addr);
1770 	crit_exit();
1771 }
1772 
1773 /*
1774  * Find an IPv6 interface link-local address specific to an interface.
1775  */
1776 struct in6_ifaddr *
1777 in6ifa_ifpforlinklocal(struct ifnet *ifp, int ignoreflags)
1778 {
1779 	struct ifaddr_container *ifac;
1780 
1781 	TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
1782 		struct ifaddr *ifa = ifac->ifa;
1783 
1784 		if (ifa->ifa_addr == NULL)
1785 			continue;	/* just for safety */
1786 		if (ifa->ifa_addr->sa_family != AF_INET6)
1787 			continue;
1788 		if (IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa))) {
1789 			if ((((struct in6_ifaddr *)ifa)->ia6_flags &
1790 			     ignoreflags) != 0)
1791 				continue;
1792 			break;
1793 		}
1794 	}
1795 	if (ifac != NULL)
1796 		return ((struct in6_ifaddr *)(ifac->ifa));
1797 	else
1798 		return (NULL);
1799 }
1800 
1801 
1802 /*
1803  * find the internet address corresponding to a given interface and address.
1804  */
1805 struct in6_ifaddr *
1806 in6ifa_ifpwithaddr(struct ifnet *ifp, struct in6_addr *addr)
1807 {
1808 	struct ifaddr_container *ifac;
1809 
1810 	TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
1811 		struct ifaddr *ifa = ifac->ifa;
1812 
1813 		if (ifa->ifa_addr == NULL)
1814 			continue;	/* just for safety */
1815 		if (ifa->ifa_addr->sa_family != AF_INET6)
1816 			continue;
1817 		if (IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa)))
1818 			break;
1819 	}
1820 	if (ifac != NULL)
1821 		return ((struct in6_ifaddr *)(ifac->ifa));
1822 	else
1823 		return (NULL);
1824 }
1825 
1826 /*
1827  * find the internet address on a given interface corresponding to a neighbor's
1828  * address.
1829  */
1830 struct in6_ifaddr *
1831 in6ifa_ifplocaladdr(const struct ifnet *ifp, const struct in6_addr *addr)
1832 {
1833 	struct ifaddr *ifa;
1834 	struct in6_ifaddr *ia;
1835 	struct ifaddr_container *ifac;
1836 
1837 	TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
1838 		ifa = ifac->ifa;
1839 
1840 		if (ifa->ifa_addr == NULL)
1841 			continue;	/* just for safety */
1842 		if (ifa->ifa_addr->sa_family != AF_INET6)
1843 			continue;
1844 		ia = (struct in6_ifaddr *)ifa;
1845 		if (IN6_ARE_MASKED_ADDR_EQUAL(addr,
1846 				&ia->ia_addr.sin6_addr,
1847 				&ia->ia_prefixmask.sin6_addr))
1848 			return ia;
1849 	}
1850 
1851 	return NULL;
1852 }
1853 
1854 /*
1855  * Convert IP6 address to printable (loggable) representation.
1856  */
1857 static char digits[] = "0123456789abcdef";
1858 static int ip6round = 0;
1859 char *
1860 ip6_sprintf(const struct in6_addr *addr)
1861 {
1862 	static char ip6buf[8][48];
1863 	int i;
1864 	char *cp;
1865 	const u_short *a = (const u_short *)addr;
1866 	const u_char *d;
1867 	int dcolon = 0;
1868 
1869 	ip6round = (ip6round + 1) & 7;
1870 	cp = ip6buf[ip6round];
1871 
1872 	for (i = 0; i < 8; i++) {
1873 		if (dcolon == 1) {
1874 			if (*a == 0) {
1875 				if (i == 7)
1876 					*cp++ = ':';
1877 				a++;
1878 				continue;
1879 			} else
1880 				dcolon = 2;
1881 		}
1882 		if (*a == 0) {
1883 			if (dcolon == 0 && *(a + 1) == 0) {
1884 				if (i == 0)
1885 					*cp++ = ':';
1886 				*cp++ = ':';
1887 				dcolon = 1;
1888 			} else {
1889 				*cp++ = '0';
1890 				*cp++ = ':';
1891 			}
1892 			a++;
1893 			continue;
1894 		}
1895 		d = (const u_char *)a;
1896 		*cp++ = digits[*d >> 4];
1897 		*cp++ = digits[*d++ & 0xf];
1898 		*cp++ = digits[*d >> 4];
1899 		*cp++ = digits[*d & 0xf];
1900 		*cp++ = ':';
1901 		a++;
1902 	}
1903 	*--cp = 0;
1904 	return (ip6buf[ip6round]);
1905 }
1906 
1907 int
1908 in6_localaddr(struct in6_addr *in6)
1909 {
1910 	struct in6_ifaddr *ia;
1911 
1912 	if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
1913 		return 1;
1914 
1915 	for (ia = in6_ifaddr; ia; ia = ia->ia_next)
1916 		if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
1917 					      &ia->ia_prefixmask.sin6_addr))
1918 			return 1;
1919 
1920 	return (0);
1921 }
1922 
1923 int
1924 in6_is_addr_deprecated(struct sockaddr_in6 *sa6)
1925 {
1926 	struct in6_ifaddr *ia;
1927 
1928 	for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
1929 		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
1930 				       &sa6->sin6_addr) &&
1931 		    (ia->ia6_flags & IN6_IFF_DEPRECATED))
1932 			return (1); /* true */
1933 
1934 		/* XXX: do we still have to go thru the rest of the list? */
1935 	}
1936 
1937 	return (0);		/* false */
1938 }
1939 
1940 /*
1941  * return length of part which dst and src are equal
1942  * hard coding...
1943  */
1944 int
1945 in6_matchlen(struct in6_addr *src, struct in6_addr *dst)
1946 {
1947 	int match = 0;
1948 	u_char *s = (u_char *)src, *d = (u_char *)dst;
1949 	u_char *lim = s + 16, r;
1950 
1951 	while (s < lim)
1952 		if ((r = (*d++ ^ *s++)) != 0) {
1953 			while (r < 128) {
1954 				match++;
1955 				r <<= 1;
1956 			}
1957 			break;
1958 		} else
1959 			match += 8;
1960 	return match;
1961 }
1962 
1963 /* XXX: to be scope conscious */
1964 int
1965 in6_are_prefix_equal(struct in6_addr *p1, struct in6_addr *p2, int len)
1966 {
1967 	int bytelen, bitlen;
1968 
1969 	/* sanity check */
1970 	if (0 > len || len > 128) {
1971 		log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n",
1972 		    len);
1973 		return (0);
1974 	}
1975 
1976 	bytelen = len / 8;
1977 	bitlen = len % 8;
1978 
1979 	if (bcmp(&p1->s6_addr, &p2->s6_addr, bytelen))
1980 		return (0);
1981 	if (p1->s6_addr[bytelen] >> (8 - bitlen) !=
1982 	    p2->s6_addr[bytelen] >> (8 - bitlen))
1983 		return (0);
1984 
1985 	return (1);
1986 }
1987 
1988 void
1989 in6_prefixlen2mask(struct in6_addr *maskp, int len)
1990 {
1991 	u_char maskarray[8] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
1992 	int bytelen, bitlen, i;
1993 
1994 	/* sanity check */
1995 	if (0 > len || len > 128) {
1996 		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
1997 		    len);
1998 		return;
1999 	}
2000 
2001 	bzero(maskp, sizeof(*maskp));
2002 	bytelen = len / 8;
2003 	bitlen = len % 8;
2004 	for (i = 0; i < bytelen; i++)
2005 		maskp->s6_addr[i] = 0xff;
2006 	if (bitlen)
2007 		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
2008 }
2009 
2010 /*
2011  * return the best address out of the same scope
2012  */
2013 struct in6_ifaddr *
2014 in6_ifawithscope(struct ifnet *oifp, struct in6_addr *dst, struct ucred *cred)
2015 {
2016 	int dst_scope =	in6_addrscope(dst), src_scope, best_scope = 0;
2017 	int blen = -1;
2018 	struct ifnet *ifp;
2019 	struct in6_ifaddr *ifa_best = NULL;
2020 	int jailed = 0;
2021 
2022 	if(cred && cred->cr_prison)
2023 		jailed = 1;
2024 
2025 	if (oifp == NULL) {
2026 #if 0
2027 		kprintf("in6_ifawithscope: output interface is not specified\n");
2028 #endif
2029 		return (NULL);
2030 	}
2031 
2032 	/*
2033 	 * We search for all addresses on all interfaces from the beginning.
2034 	 * Comparing an interface with the outgoing interface will be done
2035 	 * only at the final stage of tiebreaking.
2036 	 */
2037 	for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list))
2038 	{
2039 		struct ifaddr_container *ifac;
2040 
2041 		/*
2042 		 * We can never take an address that breaks the scope zone
2043 		 * of the destination.
2044 		 */
2045 		if (ifp->if_afdata[AF_INET6] == NULL)
2046 			continue;
2047 		if (in6_addr2scopeid(ifp, dst) != in6_addr2scopeid(oifp, dst))
2048 			continue;
2049 
2050 		TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
2051 			int tlen = -1, dscopecmp, bscopecmp, matchcmp;
2052 			struct ifaddr *ifa = ifac->ifa;
2053 
2054 			if (ifa->ifa_addr->sa_family != AF_INET6)
2055 				continue;
2056 
2057 			src_scope = in6_addrscope(IFA_IN6(ifa));
2058 
2059 			/*
2060 			 * Don't use an address before completing DAD
2061 			 * nor a duplicated address.
2062 			 */
2063 			if (((struct in6_ifaddr *)ifa)->ia6_flags &
2064 			    IN6_IFF_NOTREADY)
2065 				continue;
2066 
2067 			/* XXX: is there any case to allow anycasts? */
2068 			if (((struct in6_ifaddr *)ifa)->ia6_flags &
2069 			    IN6_IFF_ANYCAST)
2070 				continue;
2071 
2072 			if (((struct in6_ifaddr *)ifa)->ia6_flags &
2073 			    IN6_IFF_DETACHED)
2074 				continue;
2075 
2076 			/* Skip adresses not valid for current jail */
2077 			if (jailed &&
2078 			    !jailed_ip(cred->cr_prison, (struct sockaddr *)(ifa->ifa_addr)) != 0)
2079 				continue;
2080 
2081 			/*
2082 			 * If this is the first address we find,
2083 			 * keep it anyway.
2084 			 */
2085 			if (ifa_best == NULL)
2086 				goto replace;
2087 
2088 			/*
2089 			 * ifa_best is never NULL beyond this line except
2090 			 * within the block labeled "replace".
2091 			 */
2092 
2093 			/*
2094 			 * If ifa_best has a smaller scope than dst and
2095 			 * the current address has a larger one than
2096 			 * (or equal to) dst, always replace ifa_best.
2097 			 * Also, if the current address has a smaller scope
2098 			 * than dst, ignore it unless ifa_best also has a
2099 			 * smaller scope.
2100 			 * Consequently, after the two if-clause below,
2101 			 * the followings must be satisfied:
2102 			 * (scope(src) < scope(dst) &&
2103 			 *  scope(best) < scope(dst))
2104 			 *  OR
2105 			 * (scope(best) >= scope(dst) &&
2106 			 *  scope(src) >= scope(dst))
2107 			 */
2108 			if (IN6_ARE_SCOPE_CMP(best_scope, dst_scope) < 0 &&
2109 			    IN6_ARE_SCOPE_CMP(src_scope, dst_scope) >= 0)
2110 				goto replace; /* (A) */
2111 			if (IN6_ARE_SCOPE_CMP(src_scope, dst_scope) < 0 &&
2112 			    IN6_ARE_SCOPE_CMP(best_scope, dst_scope) >= 0)
2113 				continue; /* (B) */
2114 
2115 			/*
2116 			 * A deprecated address SHOULD NOT be used in new
2117 			 * communications if an alternate (non-deprecated)
2118 			 * address is available and has sufficient scope.
2119 			 * RFC 2462, Section 5.5.4.
2120 			 */
2121 			if (((struct in6_ifaddr *)ifa)->ia6_flags &
2122 			    IN6_IFF_DEPRECATED) {
2123 				/*
2124 				 * Ignore any deprecated addresses if
2125 				 * specified by configuration.
2126 				 */
2127 				if (!ip6_use_deprecated)
2128 					continue;
2129 
2130 				/*
2131 				 * If we have already found a non-deprecated
2132 				 * candidate, just ignore deprecated addresses.
2133 				 */
2134 				if (!(ifa_best->ia6_flags & IN6_IFF_DEPRECATED))
2135 					continue;
2136 			}
2137 
2138 			/*
2139 			 * A non-deprecated address is always preferred
2140 			 * to a deprecated one regardless of scopes and
2141 			 * address matching (Note invariants ensured by the
2142 			 * conditions (A) and (B) above.)
2143 			 */
2144 			if ((ifa_best->ia6_flags & IN6_IFF_DEPRECATED) &&
2145 			    !(((struct in6_ifaddr *)ifa)->ia6_flags &
2146 			     IN6_IFF_DEPRECATED))
2147 				goto replace;
2148 
2149 			/*
2150 			 * When we use temporary addresses described in
2151 			 * RFC 3041, we prefer temporary addresses to
2152 			 * public autoconf addresses.  Again, note the
2153 			 * invariants from (A) and (B).  Also note that we
2154 			 * don't have any preference between static addresses
2155 			 * and autoconf addresses (despite of whether or not
2156 			 * the latter is temporary or public.)
2157 			 */
2158 			if (ip6_use_tempaddr) {
2159 				struct in6_ifaddr *ifat;
2160 
2161 				ifat = (struct in6_ifaddr *)ifa;
2162 				if ((ifa_best->ia6_flags &
2163 				     (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY))
2164 				     == IN6_IFF_AUTOCONF &&
2165 				    (ifat->ia6_flags &
2166 				     (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY))
2167 				     == (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY)) {
2168 					goto replace;
2169 				}
2170 				if ((ifa_best->ia6_flags &
2171 				     (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY))
2172 				    == (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY) &&
2173 				    (ifat->ia6_flags &
2174 				     (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY))
2175 				     == IN6_IFF_AUTOCONF) {
2176 					continue;
2177 				}
2178 			}
2179 
2180 			/*
2181 			 * At this point, we have two cases:
2182 			 * 1. we are looking at a non-deprecated address,
2183 			 *    and ifa_best is also non-deprecated.
2184 			 * 2. we are looking at a deprecated address,
2185 			 *    and ifa_best is also deprecated.
2186 			 * Also, we do not have to consider a case where
2187 			 * the scope of if_best is larger(smaller) than dst and
2188 			 * the scope of the current address is smaller(larger)
2189 			 * than dst. Such a case has already been covered.
2190 			 * Tiebreaking is done according to the following
2191 			 * items:
2192 			 * - the scope comparison between the address and
2193 			 *   dst (dscopecmp)
2194 			 * - the scope comparison between the address and
2195 			 *   ifa_best (bscopecmp)
2196 			 * - if the address match dst longer than ifa_best
2197 			 *   (matchcmp)
2198 			 * - if the address is on the outgoing I/F (outI/F)
2199 			 *
2200 			 * Roughly speaking, the selection policy is
2201 			 * - the most important item is scope. The same scope
2202 			 *   is best. Then search for a larger scope.
2203 			 *   Smaller scopes are the last resort.
2204 			 * - A deprecated address is chosen only when we have
2205 			 *   no address that has an enough scope, but is
2206 			 *   prefered to any addresses of smaller scopes
2207 			 *   (this must be already done above.)
2208 			 * - addresses on the outgoing I/F are preferred to
2209 			 *   ones on other interfaces if none of above
2210 			 *   tiebreaks.  In the table below, the column "bI"
2211 			 *   means if the best_ifa is on the outgoing
2212 			 *   interface, and the column "sI" means if the ifa
2213 			 *   is on the outgoing interface.
2214 			 * - If there is no other reasons to choose one,
2215 			 *   longest address match against dst is considered.
2216 			 *
2217 			 * The precise decision table is as follows:
2218 			 * dscopecmp bscopecmp    match  bI oI | replace?
2219 			 *       N/A     equal      N/A   Y  N |   No (1)
2220 			 *       N/A     equal      N/A   N  Y |  Yes (2)
2221 			 *       N/A     equal   larger    N/A |  Yes (3)
2222 			 *       N/A     equal  !larger    N/A |   No (4)
2223 			 *    larger    larger      N/A    N/A |   No (5)
2224 			 *    larger   smaller      N/A    N/A |  Yes (6)
2225 			 *   smaller    larger      N/A    N/A |  Yes (7)
2226 			 *   smaller   smaller      N/A    N/A |   No (8)
2227 			 *     equal   smaller      N/A    N/A |  Yes (9)
2228 			 *     equal    larger       (already done at A above)
2229 			 */
2230 			dscopecmp = IN6_ARE_SCOPE_CMP(src_scope, dst_scope);
2231 			bscopecmp = IN6_ARE_SCOPE_CMP(src_scope, best_scope);
2232 
2233 			if (bscopecmp == 0) {
2234 				struct ifnet *bifp = ifa_best->ia_ifp;
2235 
2236 				if (bifp == oifp && ifp != oifp) /* (1) */
2237 					continue;
2238 				if (bifp != oifp && ifp == oifp) /* (2) */
2239 					goto replace;
2240 
2241 				/*
2242 				 * Both bifp and ifp are on the outgoing
2243 				 * interface, or both two are on a different
2244 				 * interface from the outgoing I/F.
2245 				 * now we need address matching against dst
2246 				 * for tiebreaking.
2247 				 */
2248 				tlen = in6_matchlen(IFA_IN6(ifa), dst);
2249 				matchcmp = tlen - blen;
2250 				if (matchcmp > 0) /* (3) */
2251 					goto replace;
2252 				continue; /* (4) */
2253 			}
2254 			if (dscopecmp > 0) {
2255 				if (bscopecmp > 0) /* (5) */
2256 					continue;
2257 				goto replace; /* (6) */
2258 			}
2259 			if (dscopecmp < 0) {
2260 				if (bscopecmp > 0) /* (7) */
2261 					goto replace;
2262 				continue; /* (8) */
2263 			}
2264 
2265 			/* now dscopecmp must be 0 */
2266 			if (bscopecmp < 0)
2267 				goto replace; /* (9) */
2268 
2269 replace:
2270 			ifa_best = (struct in6_ifaddr *)ifa;
2271 			blen = tlen >= 0 ? tlen :
2272 				in6_matchlen(IFA_IN6(ifa), dst);
2273 			best_scope = in6_addrscope(&ifa_best->ia_addr.sin6_addr);
2274 		}
2275 	}
2276 
2277 	/* count statistics for future improvements */
2278 	if (ifa_best == NULL)
2279 		ip6stat.ip6s_sources_none++;
2280 	else {
2281 		if (oifp == ifa_best->ia_ifp)
2282 			ip6stat.ip6s_sources_sameif[best_scope]++;
2283 		else
2284 			ip6stat.ip6s_sources_otherif[best_scope]++;
2285 
2286 		if (best_scope == dst_scope)
2287 			ip6stat.ip6s_sources_samescope[best_scope]++;
2288 		else
2289 			ip6stat.ip6s_sources_otherscope[best_scope]++;
2290 
2291 		if (ifa_best->ia6_flags & IN6_IFF_DEPRECATED)
2292 			ip6stat.ip6s_sources_deprecated[best_scope]++;
2293 	}
2294 
2295 	return (ifa_best);
2296 }
2297 
2298 /*
2299  * return the best address out of the same scope. if no address was
2300  * found, return the first valid address from designated IF.
2301  */
2302 struct in6_ifaddr *
2303 in6_ifawithifp(struct ifnet *ifp, struct in6_addr *dst)
2304 {
2305 	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
2306 	struct ifaddr_container *ifac;
2307 	struct in6_ifaddr *besta = NULL;
2308 	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
2309 
2310 	dep[0] = dep[1] = NULL;
2311 
2312 	/*
2313 	 * We first look for addresses in the same scope.
2314 	 * If there is one, return it.
2315 	 * If two or more, return one which matches the dst longest.
2316 	 * If none, return one of global addresses assigned other ifs.
2317 	 */
2318 	TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
2319 		struct ifaddr *ifa = ifac->ifa;
2320 
2321 		if (ifa->ifa_addr->sa_family != AF_INET6)
2322 			continue;
2323 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
2324 			continue; /* XXX: is there any case to allow anycast? */
2325 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
2326 			continue; /* don't use this interface */
2327 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
2328 			continue;
2329 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2330 			if (ip6_use_deprecated)
2331 				dep[0] = (struct in6_ifaddr *)ifa;
2332 			continue;
2333 		}
2334 
2335 		if (dst_scope == in6_addrscope(IFA_IN6(ifa))) {
2336 			/*
2337 			 * call in6_matchlen() as few as possible
2338 			 */
2339 			if (besta) {
2340 				if (blen == -1)
2341 					blen = in6_matchlen(&besta->ia_addr.sin6_addr, dst);
2342 				tlen = in6_matchlen(IFA_IN6(ifa), dst);
2343 				if (tlen > blen) {
2344 					blen = tlen;
2345 					besta = (struct in6_ifaddr *)ifa;
2346 				}
2347 			} else
2348 				besta = (struct in6_ifaddr *)ifa;
2349 		}
2350 	}
2351 	if (besta)
2352 		return (besta);
2353 
2354 	TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
2355 		struct ifaddr *ifa = ifac->ifa;
2356 
2357 		if (ifa->ifa_addr->sa_family != AF_INET6)
2358 			continue;
2359 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
2360 			continue; /* XXX: is there any case to allow anycast? */
2361 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
2362 			continue; /* don't use this interface */
2363 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
2364 			continue;
2365 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2366 			if (ip6_use_deprecated)
2367 				dep[1] = (struct in6_ifaddr *)ifa;
2368 			continue;
2369 		}
2370 
2371 		return (struct in6_ifaddr *)ifa;
2372 	}
2373 
2374 	/* use the last-resort values, that are, deprecated addresses */
2375 	if (dep[0])
2376 		return dep[0];
2377 	if (dep[1])
2378 		return dep[1];
2379 
2380 	return NULL;
2381 }
2382 
2383 /*
2384  * perform DAD when interface becomes IFF_UP.
2385  */
2386 void
2387 in6_if_up(struct ifnet *ifp)
2388 {
2389 	struct ifaddr_container *ifac;
2390 	struct in6_ifaddr *ia;
2391 	int dad_delay;		/* delay ticks before DAD output */
2392 
2393 	/*
2394 	 * special cases, like 6to4, are handled in in6_ifattach
2395 	 */
2396 	in6_ifattach(ifp, NULL);
2397 
2398 	dad_delay = 0;
2399 	TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
2400 		struct ifaddr *ifa = ifac->ifa;
2401 
2402 		if (ifa->ifa_addr->sa_family != AF_INET6)
2403 			continue;
2404 		ia = (struct in6_ifaddr *)ifa;
2405 		if (ia->ia6_flags & IN6_IFF_TENTATIVE)
2406 			nd6_dad_start(ifa, &dad_delay);
2407 	}
2408 }
2409 
2410 int
2411 in6if_do_dad(struct ifnet *ifp)
2412 {
2413 	if (ifp->if_flags & IFF_LOOPBACK)
2414 		return (0);
2415 
2416 	switch (ifp->if_type) {
2417 #ifdef IFT_DUMMY
2418 	case IFT_DUMMY:
2419 #endif
2420 	case IFT_FAITH:
2421 		/*
2422 		 * These interfaces do not have the IFF_LOOPBACK flag,
2423 		 * but loop packets back.  We do not have to do DAD on such
2424 		 * interfaces.  We should even omit it, because loop-backed
2425 		 * NS would confuse the DAD procedure.
2426 		 */
2427 		return (0);
2428 	default:
2429 		/*
2430 		 * Our DAD routine requires the interface up and running.
2431 		 * However, some interfaces can be up before the RUNNING
2432 		 * status.  Additionaly, users may try to assign addresses
2433 		 * before the interface becomes up (or running).
2434 		 * We simply skip DAD in such a case as a work around.
2435 		 * XXX: we should rather mark "tentative" on such addresses,
2436 		 * and do DAD after the interface becomes ready.
2437 		 */
2438 		if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) !=
2439 		    (IFF_UP|IFF_RUNNING))
2440 			return (0);
2441 
2442 		return (1);
2443 	}
2444 }
2445 
2446 /*
2447  * Calculate max IPv6 MTU through all the interfaces and store it
2448  * to in6_maxmtu.
2449  */
2450 void
2451 in6_setmaxmtu(void)
2452 {
2453 	unsigned long maxmtu = 0;
2454 	struct ifnet *ifp;
2455 
2456 	for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list))
2457 	{
2458 		/* this function can be called during ifnet initialization */
2459 		if (ifp->if_afdata[AF_INET6] == NULL)
2460 			continue;
2461 		if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
2462 		    ND_IFINFO(ifp)->linkmtu > maxmtu)
2463 			maxmtu =  ND_IFINFO(ifp)->linkmtu;
2464 	}
2465 	if (maxmtu)	/* update only when maxmtu is positive */
2466 		in6_maxmtu = maxmtu;
2467 }
2468 
2469 void *
2470 in6_domifattach(struct ifnet *ifp)
2471 {
2472 	struct in6_ifextra *ext;
2473 
2474 	ext = (struct in6_ifextra *)kmalloc(sizeof(*ext), M_IFADDR, M_WAITOK);
2475 	bzero(ext, sizeof(*ext));
2476 
2477 	ext->in6_ifstat = (struct in6_ifstat *)kmalloc(sizeof(struct in6_ifstat),
2478 		M_IFADDR, M_WAITOK);
2479 	bzero(ext->in6_ifstat, sizeof(*ext->in6_ifstat));
2480 
2481 	ext->icmp6_ifstat =
2482 		(struct icmp6_ifstat *)kmalloc(sizeof(struct icmp6_ifstat),
2483 			M_IFADDR, M_WAITOK);
2484 	bzero(ext->icmp6_ifstat, sizeof(*ext->icmp6_ifstat));
2485 
2486 	ext->nd_ifinfo = nd6_ifattach(ifp);
2487 	ext->scope6_id = scope6_ifattach(ifp);
2488 	return ext;
2489 }
2490 
2491 void
2492 in6_domifdetach(struct ifnet *ifp, void *aux)
2493 {
2494 	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
2495 	scope6_ifdetach(ext->scope6_id);
2496 	nd6_ifdetach(ext->nd_ifinfo);
2497 	kfree(ext->in6_ifstat, M_IFADDR);
2498 	kfree(ext->icmp6_ifstat, M_IFADDR);
2499 	kfree(ext, M_IFADDR);
2500 }
2501 
2502 /*
2503  * Convert sockaddr_in6 to sockaddr_in.  Original sockaddr_in6 must be
2504  * v4 mapped addr or v4 compat addr
2505  */
2506 void
2507 in6_sin6_2_sin(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2508 {
2509 	bzero(sin, sizeof(*sin));
2510 	sin->sin_len = sizeof(struct sockaddr_in);
2511 	sin->sin_family = AF_INET;
2512 	sin->sin_port = sin6->sin6_port;
2513 	sin->sin_addr.s_addr = sin6->sin6_addr.s6_addr32[3];
2514 }
2515 
2516 /* Convert sockaddr_in to sockaddr_in6 in v4 mapped addr format. */
2517 void
2518 in6_sin_2_v4mapsin6(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2519 {
2520 	bzero(sin6, sizeof(*sin6));
2521 	sin6->sin6_len = sizeof(struct sockaddr_in6);
2522 	sin6->sin6_family = AF_INET6;
2523 	sin6->sin6_port = sin->sin_port;
2524 	sin6->sin6_addr.s6_addr32[0] = 0;
2525 	sin6->sin6_addr.s6_addr32[1] = 0;
2526 	sin6->sin6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
2527 	sin6->sin6_addr.s6_addr32[3] = sin->sin_addr.s_addr;
2528 }
2529 
2530 /* Convert sockaddr_in6 into sockaddr_in. */
2531 void
2532 in6_sin6_2_sin_in_sock(struct sockaddr *nam)
2533 {
2534 	struct sockaddr_in *sin_p;
2535 	struct sockaddr_in6 sin6;
2536 
2537 	/*
2538 	 * Save original sockaddr_in6 addr and convert it
2539 	 * to sockaddr_in.
2540 	 */
2541 	sin6 = *(struct sockaddr_in6 *)nam;
2542 	sin_p = (struct sockaddr_in *)nam;
2543 	in6_sin6_2_sin(sin_p, &sin6);
2544 }
2545 
2546 /* Convert sockaddr_in into sockaddr_in6 in v4 mapped addr format. */
2547 void
2548 in6_sin_2_v4mapsin6_in_sock(struct sockaddr **nam)
2549 {
2550 	struct sockaddr_in *sin_p;
2551 	struct sockaddr_in6 *sin6_p;
2552 
2553 	sin6_p = kmalloc(sizeof *sin6_p, M_SONAME, M_WAITOK);
2554 	sin_p = (struct sockaddr_in *)*nam;
2555 	in6_sin_2_v4mapsin6(sin_p, sin6_p);
2556 	kfree(*nam, M_SONAME);
2557 	*nam = (struct sockaddr *)sin6_p;
2558 }
2559