xref: /dragonfly/sys/netinet6/in6.c (revision 0dace59e)
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 
80 #include <sys/thread2.h>
81 #include <sys/msgport2.h>
82 
83 #include <net/if.h>
84 #include <net/if_types.h>
85 #include <net/route.h>
86 #include <net/if_dl.h>
87 
88 #include <netinet/in.h>
89 #include <netinet/in_var.h>
90 #include <netinet/if_ether.h>
91 #include <netinet/in_systm.h>
92 #include <netinet/ip.h>
93 #include <netinet/in_pcb.h>
94 
95 #include <netinet/ip6.h>
96 #include <netinet6/ip6_var.h>
97 #include <netinet6/nd6.h>
98 #include <netinet6/mld6_var.h>
99 #include <netinet6/ip6_mroute.h>
100 #include <netinet6/in6_ifattach.h>
101 #include <netinet6/scope6_var.h>
102 #include <netinet6/in6_pcb.h>
103 #include <netinet6/in6_var.h>
104 
105 #include <net/net_osdep.h>
106 
107 /*
108  * Definitions of some costant IP6 addresses.
109  */
110 const struct in6_addr kin6addr_any = IN6ADDR_ANY_INIT;
111 const struct in6_addr kin6addr_loopback = IN6ADDR_LOOPBACK_INIT;
112 const struct in6_addr kin6addr_nodelocal_allnodes =
113 	IN6ADDR_NODELOCAL_ALLNODES_INIT;
114 const struct in6_addr kin6addr_linklocal_allnodes =
115 	IN6ADDR_LINKLOCAL_ALLNODES_INIT;
116 const struct in6_addr kin6addr_linklocal_allrouters =
117 	IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
118 
119 const struct in6_addr in6mask0 = IN6MASK0;
120 const struct in6_addr in6mask32 = IN6MASK32;
121 const struct in6_addr in6mask64 = IN6MASK64;
122 const struct in6_addr in6mask96 = IN6MASK96;
123 const struct in6_addr in6mask128 = IN6MASK128;
124 
125 const struct sockaddr_in6 sa6_any = {sizeof(sa6_any), AF_INET6,
126 				     0, 0, IN6ADDR_ANY_INIT, 0};
127 
128 static int in6_lifaddr_ioctl (struct socket *, u_long, caddr_t,
129 	struct ifnet *, struct thread *);
130 static int in6_ifinit (struct ifnet *, struct in6_ifaddr *,
131 			   struct sockaddr_in6 *, int);
132 static void in6_unlink_ifa (struct in6_ifaddr *, struct ifnet *);
133 static void in6_ifloop_request_callback(int, int, struct rt_addrinfo *, struct rtentry *, void *);
134 
135 struct in6_multihead in6_multihead;	/* XXX BSS initialization */
136 
137 int	(*faithprefix_p)(struct in6_addr *);
138 
139 /*
140  * Subroutine for in6_ifaddloop() and in6_ifremloop().
141  * This routine does actual work.
142  */
143 static void
144 in6_ifloop_request(int cmd, struct ifaddr *ifa)
145 {
146 	struct sockaddr_in6 all1_sa;
147         struct rt_addrinfo rtinfo;
148 	int error;
149 
150 	bzero(&all1_sa, sizeof(all1_sa));
151 	all1_sa.sin6_family = AF_INET6;
152 	all1_sa.sin6_len = sizeof(struct sockaddr_in6);
153 	all1_sa.sin6_addr = in6mask128;
154 
155 	/*
156 	 * We specify the address itself as the gateway, and set the
157 	 * RTF_LLINFO flag, so that the corresponding host route would have
158 	 * the flag, and thus applications that assume traditional behavior
159 	 * would be happy.  Note that we assume the caller of the function
160 	 * (probably implicitly) set nd6_rtrequest() to ifa->ifa_rtrequest,
161 	 * which changes the outgoing interface to the loopback interface.
162 	 */
163 	bzero(&rtinfo, sizeof(struct rt_addrinfo));
164 	rtinfo.rti_info[RTAX_DST] = ifa->ifa_addr;
165 	rtinfo.rti_info[RTAX_GATEWAY] = ifa->ifa_addr;
166 	rtinfo.rti_info[RTAX_NETMASK] = (struct sockaddr *)&all1_sa;
167 	rtinfo.rti_flags = RTF_UP|RTF_HOST|RTF_LLINFO;
168 
169 	error = rtrequest1_global(cmd, &rtinfo,
170 	    in6_ifloop_request_callback, ifa, RTREQ_PRIO_NORM);
171 	if (error != 0) {
172 		log(LOG_ERR, "in6_ifloop_request: "
173 		    "%s operation failed for %s (errno=%d)\n",
174 		    cmd == RTM_ADD ? "ADD" : "DELETE",
175 		    ip6_sprintf(&((struct in6_ifaddr *)ifa)->ia_addr.sin6_addr),
176 		    error);
177 	}
178 }
179 
180 static void
181 in6_ifloop_request_callback(int cmd, int error, struct rt_addrinfo *rtinfo,
182 			    struct rtentry *rt, void *arg)
183 {
184 	struct ifaddr *ifa = arg;
185 
186 	if (error)
187 		goto done;
188 
189 	/*
190 	 * Make sure rt_ifa be equal to IFA, the second argument of the
191 	 * function.
192 	 * We need this because when we refer to rt_ifa->ia6_flags in
193 	 * ip6_input, we assume that the rt_ifa points to the address instead
194 	 * of the loopback address.
195 	 */
196 	if (cmd == RTM_ADD && rt && ifa != rt->rt_ifa) {
197 		++rt->rt_refcnt;
198 		IFAFREE(rt->rt_ifa);
199 		IFAREF(ifa);
200 		rt->rt_ifa = ifa;
201 		--rt->rt_refcnt;
202 	}
203 
204 	/*
205 	 * Report the addition/removal of the address to the routing socket.
206 	 * XXX: since we called rtinit for a p2p interface with a destination,
207 	 *      we end up reporting twice in such a case.  Should we rather
208 	 *      omit the second report?
209 	 */
210 	if (rt) {
211 		if (mycpuid == 0)
212 			rt_newaddrmsg(cmd, ifa, error, rt);
213 		if (cmd == RTM_DELETE) {
214 			if (rt->rt_refcnt == 0) {
215 				++rt->rt_refcnt;
216 				rtfree(rt);
217 			}
218 		}
219 	}
220 done:
221 	/* no way to return any new error */
222 	;
223 }
224 
225 /*
226  * Add ownaddr as loopback rtentry.  We previously add the route only if
227  * necessary (ex. on a p2p link).  However, since we now manage addresses
228  * separately from prefixes, we should always add the route.  We can't
229  * rely on the cloning mechanism from the corresponding interface route
230  * any more.
231  */
232 void
233 in6_ifaddloop(struct ifaddr *ifa)
234 {
235 	struct rtentry *rt;
236 
237 	/* If there is no loopback entry, allocate one. */
238 	rt = rtpurelookup(ifa->ifa_addr);
239 	if (rt == NULL || !(rt->rt_flags & RTF_HOST) ||
240 	    !(rt->rt_ifp->if_flags & IFF_LOOPBACK))
241 		in6_ifloop_request(RTM_ADD, ifa);
242 	if (rt != NULL)
243 		rt->rt_refcnt--;
244 }
245 
246 /*
247  * Remove loopback rtentry of ownaddr generated by in6_ifaddloop(),
248  * if it exists.
249  */
250 void
251 in6_ifremloop(struct ifaddr *ifa)
252 {
253 	struct in6_ifaddr *ia;
254 	struct rtentry *rt;
255 	int ia_count = 0;
256 
257 	/*
258 	 * Some of BSD variants do not remove cloned routes
259 	 * from an interface direct route, when removing the direct route
260 	 * (see comments in net/net_osdep.h).  Even for variants that do remove
261 	 * cloned routes, they could fail to remove the cloned routes when
262 	 * we handle multple addresses that share a common prefix.
263 	 * So, we should remove the route corresponding to the deleted address
264 	 * regardless of the result of in6_is_ifloop_auto().
265 	 */
266 
267 	/*
268 	 * Delete the entry only if exact one ifa exists.  More than one ifa
269 	 * can exist if we assign a same single address to multiple
270 	 * (probably p2p) interfaces.
271 	 * XXX: we should avoid such a configuration in IPv6...
272 	 */
273 	for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
274 		if (IN6_ARE_ADDR_EQUAL(IFA_IN6(ifa), &ia->ia_addr.sin6_addr)) {
275 			ia_count++;
276 			if (ia_count > 1)
277 				break;
278 		}
279 	}
280 
281 	if (ia_count == 1) {
282 		/*
283 		 * Before deleting, check if a corresponding loopbacked host
284 		 * route surely exists.  With this check, we can avoid to
285 		 * delete an interface direct route whose destination is same
286 		 * as the address being removed.  This can happen when remofing
287 		 * a subnet-router anycast address on an interface attahced
288 		 * to a shared medium.
289 		 */
290 		rt = rtpurelookup(ifa->ifa_addr);
291 		if (rt != NULL && (rt->rt_flags & RTF_HOST) &&
292 		    (rt->rt_ifp->if_flags & IFF_LOOPBACK)) {
293 			rt->rt_refcnt--;
294 			in6_ifloop_request(RTM_DELETE, ifa);
295 		}
296 	}
297 }
298 
299 int
300 in6_ifindex2scopeid(int idx)
301 {
302 	struct ifnet *ifp;
303 	struct sockaddr_in6 *sin6;
304 	struct ifaddr_container *ifac;
305 
306 	if (idx < 0 || if_index < idx)
307 		return -1;
308 	ifp = ifindex2ifnet[idx];
309 
310 	TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link)
311 	{
312 		struct ifaddr *ifa = ifac->ifa;
313 
314 		if (ifa->ifa_addr->sa_family != AF_INET6)
315 			continue;
316 		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
317 		if (IN6_IS_ADDR_SITELOCAL(&sin6->sin6_addr))
318 			return sin6->sin6_scope_id & 0xffff;
319 	}
320 
321 	return -1;
322 }
323 
324 int
325 in6_mask2len(struct in6_addr *mask, u_char *lim0)
326 {
327 	int x = 0, y;
328 	u_char *lim = lim0, *p;
329 
330 	if (lim0 == NULL ||
331 	    lim0 - (u_char *)mask > sizeof(*mask)) /* ignore the scope_id part */
332 		lim = (u_char *)mask + sizeof(*mask);
333 	for (p = (u_char *)mask; p < lim; x++, p++) {
334 		if (*p != 0xff)
335 			break;
336 	}
337 	y = 0;
338 	if (p < lim) {
339 		for (y = 0; y < 8; y++) {
340 			if ((*p & (0x80 >> y)) == 0)
341 				break;
342 		}
343 	}
344 
345 	/*
346 	 * when the limit pointer is given, do a stricter check on the
347 	 * remaining bits.
348 	 */
349 	if (p < lim) {
350 		if (y != 0 && (*p & (0x00ff >> y)) != 0)
351 			return (-1);
352 		for (p = p + 1; p < lim; p++)
353 			if (*p != 0)
354 				return (-1);
355 	}
356 
357 	return x * 8 + y;
358 }
359 
360 void
361 in6_len2mask(struct in6_addr *mask, int len)
362 {
363 	int i;
364 
365 	bzero(mask, sizeof(*mask));
366 	for (i = 0; i < len / 8; i++)
367 		mask->s6_addr8[i] = 0xff;
368 	if (len % 8)
369 		mask->s6_addr8[i] = (0xff00 >> (len % 8)) & 0xff;
370 }
371 
372 #define ifa2ia6(ifa)	((struct in6_ifaddr *)(ifa))
373 #define ia62ifa(ia6)	(&((ia6)->ia_ifa))
374 
375 void
376 in6_control_dispatch(netmsg_t msg)
377 {
378 	int error;
379 
380 	error = in6_control(msg->control.base.nm_so,
381 			    msg->control.nm_cmd,
382 			    msg->control.nm_data,
383 			    msg->control.nm_ifp,
384 			    msg->control.nm_td);
385 	lwkt_replymsg(&msg->control.base.lmsg, error);
386 }
387 
388 int
389 in6_control(struct socket *so, u_long cmd, caddr_t data,
390 	    struct ifnet *ifp, struct thread *td)
391 {
392 	struct	in6_ifreq *ifr = (struct in6_ifreq *)data;
393 	struct	in6_ifaddr *ia = NULL;
394 	struct	in6_aliasreq *ifra = (struct in6_aliasreq *)data;
395 	struct	in6_ifextra *xtra;
396 	int privileged;
397 	int error;
398 
399 	privileged = 0;
400 	if (priv_check(td, PRIV_ROOT) == 0)
401 		privileged++;
402 
403 	switch (cmd) {
404 	case SIOCGETSGCNT_IN6:
405 	case SIOCGETMIFCNT_IN6:
406 		return (mrt6_ioctl(cmd, data));
407 	}
408 
409 	switch(cmd) {
410 	case SIOCAADDRCTL_POLICY:
411 	case SIOCDADDRCTL_POLICY:
412 		if (!privileged)
413 			return (EPERM);
414 		return (in6_src_ioctl(cmd, data));
415 	}
416 
417 	if (ifp == NULL)
418 		return (EOPNOTSUPP);
419 
420 	switch (cmd) {
421 	case SIOCSNDFLUSH_IN6:
422 	case SIOCSPFXFLUSH_IN6:
423 	case SIOCSRTRFLUSH_IN6:
424 	case SIOCSDEFIFACE_IN6:
425 	case SIOCSIFINFO_FLAGS:
426 		if (!privileged)
427 			return (EPERM);
428 		/* fall through */
429 	case OSIOCGIFINFO_IN6:
430 	case SIOCGIFINFO_IN6:
431 	case SIOCGDRLST_IN6:
432 	case SIOCGPRLST_IN6:
433 	case SIOCGNBRINFO_IN6:
434 	case SIOCGDEFIFACE_IN6:
435 		return (nd6_ioctl(cmd, data, ifp));
436 	}
437 
438 	switch (cmd) {
439 	case SIOCSIFPREFIX_IN6:
440 	case SIOCDIFPREFIX_IN6:
441 	case SIOCAIFPREFIX_IN6:
442 	case SIOCCIFPREFIX_IN6:
443 	case SIOCSGIFPREFIX_IN6:
444 	case SIOCGIFPREFIX_IN6:
445 		log(LOG_NOTICE,
446 		    "prefix ioctls are now invalidated. "
447 		    "please use ifconfig.\n");
448 		return (EOPNOTSUPP);
449 	}
450 
451 	switch (cmd) {
452 	case SIOCSSCOPE6:
453 		if (!privileged)
454 			return (EPERM);
455 		return (scope6_set(ifp,
456 			(struct scope6_id *)ifr->ifr_ifru.ifru_scope_id));
457 		break;
458 	case SIOCGSCOPE6:
459 		return (scope6_get(ifp,
460 			(struct scope6_id *)ifr->ifr_ifru.ifru_scope_id));
461 		break;
462 	case SIOCGSCOPE6DEF:
463 		return (scope6_get_default((struct scope6_id *)
464 			ifr->ifr_ifru.ifru_scope_id));
465 		break;
466 	}
467 
468 	switch (cmd) {
469 	case SIOCALIFADDR:
470 	case SIOCDLIFADDR:
471 		if (!privileged)
472 			return (EPERM);
473 		/* fall through */
474 	case SIOCGLIFADDR:
475 		return in6_lifaddr_ioctl(so, cmd, data, ifp, td);
476 	}
477 
478 	/*
479 	 * Find address for this interface, if it exists.
480 	 */
481 	if (ifra->ifra_addr.sin6_family == AF_INET6) { /* XXX */
482 		struct sockaddr_in6 *sa6 =
483 			(struct sockaddr_in6 *)&ifra->ifra_addr;
484 
485 		if (IN6_IS_ADDR_LINKLOCAL(&sa6->sin6_addr)) {
486 			if (sa6->sin6_addr.s6_addr16[1] == 0) {
487 				/* link ID is not embedded by the user */
488 				sa6->sin6_addr.s6_addr16[1] =
489 					htons(ifp->if_index);
490 			} else if (sa6->sin6_addr.s6_addr16[1] !=
491 				    htons(ifp->if_index)) {
492 				return (EINVAL);	/* link ID contradicts */
493 			}
494 			if (sa6->sin6_scope_id) {
495 				if (sa6->sin6_scope_id !=
496 				    (u_int32_t)ifp->if_index)
497 					return (EINVAL);
498 				sa6->sin6_scope_id = 0; /* XXX: good way? */
499 			}
500 		}
501 		ia = in6ifa_ifpwithaddr(ifp, &ifra->ifra_addr.sin6_addr);
502 	}
503 
504 	switch (cmd) {
505 	case SIOCSIFADDR_IN6:
506 	case SIOCSIFDSTADDR_IN6:
507 	case SIOCSIFNETMASK_IN6:
508 		/*
509 		 * Since IPv6 allows a node to assign multiple addresses
510 		 * on a single interface, SIOCSIFxxx ioctls are not suitable
511 		 * and should be unused.
512 		 */
513 		/* we decided to obsolete this command (20000704) */
514 		return (EINVAL);
515 
516 	case SIOCDIFADDR_IN6:
517 		/*
518 		 * for IPv4, we look for existing in_ifaddr here to allow
519 		 * "ifconfig if0 delete" to remove first IPv4 address on the
520 		 * interface.  For IPv6, as the spec allow multiple interface
521 		 * address from the day one, we consider "remove the first one"
522 		 * semantics to be not preferable.
523 		 */
524 		if (ia == NULL)
525 			return (EADDRNOTAVAIL);
526 		/* FALLTHROUGH */
527 	case SIOCAIFADDR_IN6:
528 		/*
529 		 * We always require users to specify a valid IPv6 address for
530 		 * the corresponding operation.
531 		 */
532 		if (ifra->ifra_addr.sin6_family != AF_INET6 ||
533 		    ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6))
534 			return (EAFNOSUPPORT);
535 		if (!privileged)
536 			return (EPERM);
537 
538 		break;
539 
540 	case SIOCGIFADDR_IN6:
541 		/* This interface is basically deprecated. use SIOCGIFCONF. */
542 		/* fall through */
543 	case SIOCGIFAFLAG_IN6:
544 	case SIOCGIFNETMASK_IN6:
545 	case SIOCGIFDSTADDR_IN6:
546 	case SIOCGIFALIFETIME_IN6:
547 		/* must think again about its semantics */
548 		if (ia == NULL)
549 			return (EADDRNOTAVAIL);
550 		break;
551 	case SIOCSIFALIFETIME_IN6:
552 	    {
553 		struct in6_addrlifetime *lt;
554 
555 		if (!privileged)
556 			return (EPERM);
557 		if (ia == NULL)
558 			return (EADDRNOTAVAIL);
559 		/* sanity for overflow - beware unsigned */
560 		lt = &ifr->ifr_ifru.ifru_lifetime;
561 		if (lt->ia6t_vltime != ND6_INFINITE_LIFETIME
562 		    && lt->ia6t_vltime + time_uptime < time_uptime) {
563 			return EINVAL;
564 		}
565 		if (lt->ia6t_pltime != ND6_INFINITE_LIFETIME
566 		    && lt->ia6t_pltime + time_uptime < time_uptime) {
567 			return EINVAL;
568 		}
569 		break;
570 	    }
571 	}
572 
573 	switch (cmd) {
574 
575 	case SIOCGIFADDR_IN6:
576 		ifr->ifr_addr = ia->ia_addr;
577 		break;
578 
579 	case SIOCGIFDSTADDR_IN6:
580 		if (!(ifp->if_flags & IFF_POINTOPOINT))
581 			return (EINVAL);
582 		/*
583 		 * XXX: should we check if ifa_dstaddr is NULL and return
584 		 * an error?
585 		 */
586 		ifr->ifr_dstaddr = ia->ia_dstaddr;
587 		break;
588 
589 	case SIOCGIFNETMASK_IN6:
590 		ifr->ifr_addr = ia->ia_prefixmask;
591 		break;
592 
593 	case SIOCGIFAFLAG_IN6:
594 		ifr->ifr_ifru.ifru_flags6 = ia->ia6_flags;
595 		break;
596 
597 	case SIOCGIFSTAT_IN6:
598 		if (ifp == NULL || (xtra = ifp->if_afdata[AF_INET6]) == NULL)
599 			return EINVAL;
600 		bzero(&ifr->ifr_ifru.ifru_stat,
601 		      sizeof(ifr->ifr_ifru.ifru_stat));
602 		ifr->ifr_ifru.ifru_stat = *xtra->in6_ifstat;
603 		break;
604 
605 	case SIOCGIFSTAT_ICMP6:
606 		if (ifp == NULL || (xtra = ifp->if_afdata[AF_INET6]) == NULL)
607 			return EINVAL;
608 		bzero(&ifr->ifr_ifru.ifru_stat,
609 			sizeof(ifr->ifr_ifru.ifru_icmp6stat));
610 		ifr->ifr_ifru.ifru_icmp6stat = *xtra->icmp6_ifstat;
611 		break;
612 
613 	case SIOCSIFADDR:
614 	case SIOCSIFDSTADDR:
615 	case SIOCSIFBRDADDR:
616 	case SIOCSIFNETMASK:
617 		/*
618 		 * Do not pass those ioctl to driver handler since they are not
619 		 * properly setup.  Instead just error out.
620 		 */
621 		return (EOPNOTSUPP);
622 
623 	case SIOCGIFALIFETIME_IN6:
624 		ifr->ifr_ifru.ifru_lifetime = ia->ia6_lifetime;
625 		break;
626 
627 	case SIOCSIFALIFETIME_IN6:
628 		ia->ia6_lifetime = ifr->ifr_ifru.ifru_lifetime;
629 		/* for sanity */
630 		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
631 			ia->ia6_lifetime.ia6t_expire =
632 				time_uptime + ia->ia6_lifetime.ia6t_vltime;
633 		} else
634 			ia->ia6_lifetime.ia6t_expire = 0;
635 		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
636 			ia->ia6_lifetime.ia6t_preferred =
637 				time_uptime + ia->ia6_lifetime.ia6t_pltime;
638 		} else
639 			ia->ia6_lifetime.ia6t_preferred = 0;
640 		break;
641 
642 	case SIOCAIFADDR_IN6:
643 	{
644 		int i, error = 0, iaIsNew;
645 		struct nd_prefix pr0, *pr;
646 
647 		if (ia != NULL)
648 			iaIsNew = 0;
649 		else
650 			iaIsNew = 1;
651 
652 		/*
653 		 * first, make or update the interface address structure,
654 		 * and link it to the list.
655 		 */
656 		if ((error = in6_update_ifa(ifp, ifra, ia)) != 0)
657 			return (error);
658 
659 		/*
660 		 * then, make the prefix on-link on the interface.
661 		 * XXX: we'd rather create the prefix before the address, but
662 		 * we need at least one address to install the corresponding
663 		 * interface route, so we configure the address first.
664 		 */
665 
666 		/*
667 		 * convert mask to prefix length (prefixmask has already
668 		 * been validated in in6_update_ifa().
669 		 */
670 		bzero(&pr0, sizeof(pr0));
671 		pr0.ndpr_ifp = ifp;
672 		pr0.ndpr_plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
673 					     NULL);
674 		if (pr0.ndpr_plen == 128)
675 			break;	/* we don't need to install a host route. */
676 		pr0.ndpr_prefix = ifra->ifra_addr;
677 		pr0.ndpr_mask = ifra->ifra_prefixmask.sin6_addr;
678 		/* apply the mask for safety. */
679 		for (i = 0; i < 4; i++) {
680 			pr0.ndpr_prefix.sin6_addr.s6_addr32[i] &=
681 				ifra->ifra_prefixmask.sin6_addr.s6_addr32[i];
682 		}
683 		/*
684 		 * XXX: since we don't have an API to set prefix (not address)
685 		 * lifetimes, we just use the same lifetimes as addresses.
686 		 * The (temporarily) installed lifetimes can be overridden by
687 		 * later advertised RAs (when accept_rtadv is non 0), which is
688 		 * an intended behavior.
689 		 */
690 		pr0.ndpr_raf_onlink = 1; /* should be configurable? */
691 		pr0.ndpr_raf_auto =
692 			((ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0);
693 		pr0.ndpr_vltime = ifra->ifra_lifetime.ia6t_vltime;
694 		pr0.ndpr_pltime = ifra->ifra_lifetime.ia6t_pltime;
695 
696 		/* add the prefix if there's one. */
697 		if ((pr = nd6_prefix_lookup(&pr0)) == NULL) {
698 			/*
699 			 * nd6_prelist_add will install the corresponding
700 			 * interface route.
701 			 */
702 			if ((error = nd6_prelist_add(&pr0, NULL, &pr)) != 0)
703 				return (error);
704 			if (pr == NULL) {
705 				log(LOG_ERR, "nd6_prelist_add succeeded but "
706 				    "no prefix\n");
707 				return (EINVAL); /* XXX panic here? */
708 			}
709 		}
710 		if ((ia = in6ifa_ifpwithaddr(ifp, &ifra->ifra_addr.sin6_addr))
711 		    == NULL) {
712 		    	/* XXX: this should not happen! */
713 			log(LOG_ERR, "in6_control: addition succeeded, but"
714 			    " no ifaddr\n");
715 		} else {
716 			if ((ia->ia6_flags & IN6_IFF_AUTOCONF) &&
717 			    ia->ia6_ndpr == NULL) { /* new autoconfed addr */
718 				ia->ia6_ndpr = pr;
719 				pr->ndpr_refcnt++;
720 
721 				/*
722 				 * If this is the first autoconf address from
723 				 * the prefix, create a temporary address
724 				 * as well (when specified).
725 				 */
726 				if (ip6_use_tempaddr &&
727 				    pr->ndpr_refcnt == 1) {
728 					int e;
729 					if ((e = in6_tmpifadd(ia, 1)) != 0) {
730 						log(LOG_NOTICE, "in6_control: "
731 						    "failed to create a "
732 						    "temporary address, "
733 						    "errno=%d\n",
734 						    e);
735 					}
736 				}
737 			}
738 
739 			/*
740 			 * this might affect the status of autoconfigured
741 			 * addresses, that is, this address might make
742 			 * other addresses detached.
743 			 */
744 			pfxlist_onlink_check();
745 		}
746 		if (error == 0 && ia) {
747 			EVENTHANDLER_INVOKE(ifaddr_event, ifp,
748 			iaIsNew ? IFADDR_EVENT_ADD : IFADDR_EVENT_CHANGE,
749 			&ia->ia_ifa);
750 		}
751 		break;
752 	}
753 
754 	case SIOCDIFADDR_IN6:
755 	{
756 		int i = 0;
757 		struct nd_prefix pr0, *pr;
758 
759 		/*
760 		 * If the address being deleted is the only one that owns
761 		 * the corresponding prefix, expire the prefix as well.
762 		 * XXX: theoretically, we don't have to warry about such
763 		 * relationship, since we separate the address management
764 		 * and the prefix management.  We do this, however, to provide
765 		 * as much backward compatibility as possible in terms of
766 		 * the ioctl operation.
767 		 */
768 		bzero(&pr0, sizeof(pr0));
769 		pr0.ndpr_ifp = ifp;
770 		pr0.ndpr_plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr,
771 					     NULL);
772 		if (pr0.ndpr_plen == 128)
773 			goto purgeaddr;
774 		pr0.ndpr_prefix = ia->ia_addr;
775 		pr0.ndpr_mask = ia->ia_prefixmask.sin6_addr;
776 		for (i = 0; i < 4; i++) {
777 			pr0.ndpr_prefix.sin6_addr.s6_addr32[i] &=
778 				ia->ia_prefixmask.sin6_addr.s6_addr32[i];
779 		}
780 		/*
781 		 * The logic of the following condition is a bit complicated.
782 		 * We expire the prefix when
783 		 * 1. the address obeys autoconfiguration and it is the
784 		 *    only owner of the associated prefix, or
785 		 * 2. the address does not obey autoconf and there is no
786 		 *    other owner of the prefix.
787 		 */
788 		if ((pr = nd6_prefix_lookup(&pr0)) != NULL &&
789 		    (((ia->ia6_flags & IN6_IFF_AUTOCONF) &&
790 		       pr->ndpr_refcnt == 1) ||
791 		     (!(ia->ia6_flags & IN6_IFF_AUTOCONF) &&
792 		      pr->ndpr_refcnt == 0))) {
793 			pr->ndpr_expire = 1; /* XXX: just for expiration */
794 		}
795 
796 purgeaddr:
797 		EVENTHANDLER_INVOKE(ifaddr_event, ifp, IFADDR_EVENT_DELETE,
798 				    &ia->ia_ifa);
799 		in6_purgeaddr(&ia->ia_ifa);
800 		break;
801 	}
802 
803 	default:
804 		if (ifp == NULL || ifp->if_ioctl == NULL)
805 			return (EOPNOTSUPP);
806 		ifnet_serialize_all(ifp);
807 		error = ifp->if_ioctl(ifp, cmd, data, td->td_proc->p_ucred);
808 		ifnet_deserialize_all(ifp);
809 		return (error);
810 	}
811 
812 	return (0);
813 }
814 
815 /*
816  * Update parameters of an IPv6 interface address.
817  * If necessary, a new entry is created and linked into address chains.
818  * This function is separated from in6_control().
819  * XXX: should this be performed under splnet()?
820  */
821 int
822 in6_update_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
823 	       struct in6_ifaddr *ia)
824 {
825 	int error = 0, hostIsNew = 0, plen = -1;
826 	struct in6_ifaddr *oia;
827 	struct sockaddr_in6 dst6;
828 	struct in6_addrlifetime *lt;
829 
830 	/* Validate parameters */
831 	if (ifp == NULL || ifra == NULL) /* this maybe redundant */
832 		return (EINVAL);
833 
834 	/*
835 	 * The destination address for a p2p link must have a family
836 	 * of AF_UNSPEC or AF_INET6.
837 	 */
838 	if ((ifp->if_flags & IFF_POINTOPOINT) &&
839 	    ifra->ifra_dstaddr.sin6_family != AF_INET6 &&
840 	    ifra->ifra_dstaddr.sin6_family != AF_UNSPEC)
841 		return (EAFNOSUPPORT);
842 	/*
843 	 * validate ifra_prefixmask.  don't check sin6_family, netmask
844 	 * does not carry fields other than sin6_len.
845 	 */
846 	if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6))
847 		return (EINVAL);
848 	/*
849 	 * Because the IPv6 address architecture is classless, we require
850 	 * users to specify a (non 0) prefix length (mask) for a new address.
851 	 * We also require the prefix (when specified) mask is valid, and thus
852 	 * reject a non-consecutive mask.
853 	 */
854 	if (ia == NULL && ifra->ifra_prefixmask.sin6_len == 0)
855 		return (EINVAL);
856 	if (ifra->ifra_prefixmask.sin6_len != 0) {
857 		plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
858 				    (u_char *)&ifra->ifra_prefixmask +
859 				    ifra->ifra_prefixmask.sin6_len);
860 		if (plen <= 0)
861 			return (EINVAL);
862 	}
863 	else {
864 		/*
865 		 * In this case, ia must not be NULL.  We just use its prefix
866 		 * length.
867 		 */
868 		plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
869 	}
870 	/*
871 	 * If the destination address on a p2p interface is specified,
872 	 * and the address is a scoped one, validate/set the scope
873 	 * zone identifier.
874 	 */
875 	dst6 = ifra->ifra_dstaddr;
876 	if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) &&
877 	    (dst6.sin6_family == AF_INET6)) {
878 		int scopeid;
879 
880 		if ((error = in6_recoverscope(&dst6,
881 					      &ifra->ifra_dstaddr.sin6_addr,
882 					      ifp)) != 0)
883 			return (error);
884 		scopeid = in6_addr2scopeid(ifp, &dst6.sin6_addr);
885 		if (dst6.sin6_scope_id == 0) /* user omit to specify the ID. */
886 			dst6.sin6_scope_id = scopeid;
887 		else if (dst6.sin6_scope_id != scopeid)
888 			return (EINVAL); /* scope ID mismatch. */
889 		if ((error = in6_embedscope(&dst6.sin6_addr, &dst6, NULL, NULL))
890 		    != 0)
891 			return (error);
892 		dst6.sin6_scope_id = 0; /* XXX */
893 	}
894 	/*
895 	 * The destination address can be specified only for a p2p or a
896 	 * loopback interface.  If specified, the corresponding prefix length
897 	 * must be 128.
898 	 */
899 	if (ifra->ifra_dstaddr.sin6_family == AF_INET6) {
900 		if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) == 0) {
901 			/* XXX: noisy message */
902 			log(LOG_INFO, "in6_update_ifa: a destination can be "
903 			    "specified for a p2p or a loopback IF only\n");
904 			return (EINVAL);
905 		}
906 		if (plen != 128) {
907 			/*
908 			 * The following message seems noisy, but we dare to
909 			 * add it for diagnosis.
910 			 */
911 			log(LOG_INFO, "in6_update_ifa: prefixlen must be 128 "
912 			    "when dstaddr is specified\n");
913 			return (EINVAL);
914 		}
915 	}
916 	/* lifetime consistency check */
917 	lt = &ifra->ifra_lifetime;
918 	if (lt->ia6t_vltime != ND6_INFINITE_LIFETIME
919 	    && lt->ia6t_vltime + time_uptime < time_uptime) {
920 		return EINVAL;
921 	}
922 	if (lt->ia6t_vltime == 0) {
923 		/*
924 		 * the following log might be noisy, but this is a typical
925 		 * configuration mistake or a tool's bug.
926 		 */
927 		log(LOG_INFO,
928 		    "in6_update_ifa: valid lifetime is 0 for %s\n",
929 		    ip6_sprintf(&ifra->ifra_addr.sin6_addr));
930 	}
931 	if (lt->ia6t_pltime != ND6_INFINITE_LIFETIME
932 	    && lt->ia6t_pltime + time_uptime < time_uptime) {
933 		return EINVAL;
934 	}
935 
936 	/*
937 	 * If this is a new address, allocate a new ifaddr and link it
938 	 * into chains.
939 	 */
940 	if (ia == NULL) {
941 		hostIsNew = 1;
942 		/*
943 		 * When in6_update_ifa() is called in a process of a received
944 		 * RA, it is called under splnet().  So, we should call malloc
945 		 * with M_NOWAIT.
946 		 */
947 		ia = ifa_create(sizeof(*ia), M_NOWAIT);
948 		if (ia == NULL)
949 			return (ENOBUFS);
950 		/* Initialize the address and masks */
951 		ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
952 		ia->ia_addr.sin6_family = AF_INET6;
953 		ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
954 		if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
955 			/*
956 			 * XXX: some functions expect that ifa_dstaddr is not
957 			 * NULL for p2p interfaces.
958 			 */
959 			ia->ia_ifa.ifa_dstaddr
960 				= (struct sockaddr *)&ia->ia_dstaddr;
961 		} else {
962 			ia->ia_ifa.ifa_dstaddr = NULL;
963 		}
964 		ia->ia_ifa.ifa_netmask
965 			= (struct sockaddr *)&ia->ia_prefixmask;
966 
967 		ia->ia_ifp = ifp;
968 		if ((oia = in6_ifaddr) != NULL) {
969 			for ( ; oia->ia_next; oia = oia->ia_next)
970 				continue;
971 			oia->ia_next = ia;
972 		} else
973 			in6_ifaddr = ia;
974 
975 		ifa_iflink(&ia->ia_ifa, ifp, 1);
976 	}
977 
978 	/* set prefix mask */
979 	if (ifra->ifra_prefixmask.sin6_len) {
980 		/*
981 		 * We prohibit changing the prefix length of an existing
982 		 * address, because
983 		 * + such an operation should be rare in IPv6, and
984 		 * + the operation would confuse prefix management.
985 		 */
986 		if (ia->ia_prefixmask.sin6_len &&
987 		    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) != plen) {
988 			log(LOG_INFO, "in6_update_ifa: the prefix length of an"
989 			    " existing (%s) address should not be changed\n",
990 			    ip6_sprintf(&ia->ia_addr.sin6_addr));
991 			error = EINVAL;
992 			goto unlink;
993 		}
994 		ia->ia_prefixmask = ifra->ifra_prefixmask;
995 	}
996 
997 	/*
998 	 * If a new destination address is specified, scrub the old one and
999 	 * install the new destination.  Note that the interface must be
1000 	 * p2p or loopback (see the check above.)
1001 	 */
1002 	if (dst6.sin6_family == AF_INET6 &&
1003 	    !IN6_ARE_ADDR_EQUAL(&dst6.sin6_addr,
1004 				&ia->ia_dstaddr.sin6_addr)) {
1005 		int e;
1006 
1007 		if ((ia->ia_flags & IFA_ROUTE) &&
1008 		    (e = rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST))
1009 		    != 0) {
1010 			log(LOG_ERR, "in6_update_ifa: failed to remove "
1011 			    "a route to the old destination: %s\n",
1012 			    ip6_sprintf(&ia->ia_addr.sin6_addr));
1013 			/* proceed anyway... */
1014 		}
1015 		else
1016 			ia->ia_flags &= ~IFA_ROUTE;
1017 		ia->ia_dstaddr = dst6;
1018 	}
1019 
1020 	/* reset the interface and routing table appropriately. */
1021 	if ((error = in6_ifinit(ifp, ia, &ifra->ifra_addr, hostIsNew)) != 0)
1022 		goto unlink;
1023 
1024 	/*
1025 	 * Beyond this point, we should call in6_purgeaddr upon an error,
1026 	 * not just go to unlink.
1027 	 */
1028 
1029 #if 0				/* disable this mechanism for now */
1030 	/* update prefix list */
1031 	if (hostIsNew &&
1032 	    (ifra->ifra_flags & IN6_IFF_NOPFX) == 0) { /* XXX */
1033 		int iilen;
1034 
1035 		iilen = (sizeof(ia->ia_prefixmask.sin6_addr) << 3) - plen;
1036 		if ((error = in6_prefix_add_ifid(iilen, ia)) != 0) {
1037 			in6_purgeaddr((struct ifaddr *)ia);
1038 			return (error);
1039 		}
1040 	}
1041 #endif
1042 
1043 	if (ifp->if_flags & IFF_MULTICAST) {
1044 		struct sockaddr_in6 mltaddr, mltmask;
1045 		struct in6_multi *in6m;
1046 
1047 		if (hostIsNew) {
1048 			/*
1049 			 * join solicited multicast addr for new host id
1050 			 */
1051 			struct in6_addr llsol;
1052 			bzero(&llsol, sizeof(struct in6_addr));
1053 			llsol.s6_addr16[0] = htons(0xff02);
1054 			llsol.s6_addr16[1] = htons(ifp->if_index);
1055 			llsol.s6_addr32[1] = 0;
1056 			llsol.s6_addr32[2] = htonl(1);
1057 			llsol.s6_addr32[3] =
1058 				ifra->ifra_addr.sin6_addr.s6_addr32[3];
1059 			llsol.s6_addr8[12] = 0xff;
1060 			in6_addmulti(&llsol, ifp, &error);
1061 			if (error != 0) {
1062 				log(LOG_WARNING,
1063 				    "in6_update_ifa: addmulti failed for "
1064 				    "%s on %s (errno=%d)\n",
1065 				    ip6_sprintf(&llsol), if_name(ifp),
1066 				    error);
1067 				in6_purgeaddr((struct ifaddr *)ia);
1068 				return (error);
1069 			}
1070 		}
1071 
1072 		bzero(&mltmask, sizeof(mltmask));
1073 		mltmask.sin6_len = sizeof(struct sockaddr_in6);
1074 		mltmask.sin6_family = AF_INET6;
1075 		mltmask.sin6_addr = in6mask32;
1076 
1077 		/*
1078 		 * join link-local all-nodes address
1079 		 */
1080 		bzero(&mltaddr, sizeof(mltaddr));
1081 		mltaddr.sin6_len = sizeof(struct sockaddr_in6);
1082 		mltaddr.sin6_family = AF_INET6;
1083 		mltaddr.sin6_addr = kin6addr_linklocal_allnodes;
1084 		mltaddr.sin6_addr.s6_addr16[1] = htons(ifp->if_index);
1085 
1086 		IN6_LOOKUP_MULTI(mltaddr.sin6_addr, ifp, in6m);
1087 		if (in6m == NULL) {
1088 			rtrequest_global(RTM_ADD,
1089 				  (struct sockaddr *)&mltaddr,
1090 				  (struct sockaddr *)&ia->ia_addr,
1091 				  (struct sockaddr *)&mltmask,
1092 				  RTF_UP|RTF_CLONING);  /* xxx */
1093 			in6_addmulti(&mltaddr.sin6_addr, ifp, &error);
1094 			if (error != 0) {
1095 				log(LOG_WARNING,
1096 				    "in6_update_ifa: addmulti failed for "
1097 				    "%s on %s (errno=%d)\n",
1098 				    ip6_sprintf(&mltaddr.sin6_addr),
1099 				    if_name(ifp), error);
1100 			}
1101 		}
1102 
1103 		/*
1104 		 * join node information group address
1105 		 */
1106 #define hostnamelen	strlen(hostname)
1107 		if (in6_nigroup(ifp, hostname, hostnamelen, &mltaddr.sin6_addr)
1108 		    == 0) {
1109 			IN6_LOOKUP_MULTI(mltaddr.sin6_addr, ifp, in6m);
1110 			if (in6m == NULL && ia != NULL) {
1111 				in6_addmulti(&mltaddr.sin6_addr, ifp, &error);
1112 				if (error != 0) {
1113 					log(LOG_WARNING, "in6_update_ifa: "
1114 					    "addmulti failed for "
1115 					    "%s on %s (errno=%d)\n",
1116 					    ip6_sprintf(&mltaddr.sin6_addr),
1117 					    if_name(ifp), error);
1118 				}
1119 			}
1120 		}
1121 #undef hostnamelen
1122 
1123 		/*
1124 		 * join node-local all-nodes address, on loopback.
1125 		 * XXX: since "node-local" is obsoleted by interface-local,
1126 		 *      we have to join the group on every interface with
1127 		 *      some interface-boundary restriction.
1128 		 */
1129 		if (ifp->if_flags & IFF_LOOPBACK) {
1130 			struct in6_ifaddr *ia_loop;
1131 
1132 			struct in6_addr loop6 = kin6addr_loopback;
1133 			ia_loop = in6ifa_ifpwithaddr(ifp, &loop6);
1134 
1135 			mltaddr.sin6_addr = kin6addr_nodelocal_allnodes;
1136 
1137 			IN6_LOOKUP_MULTI(mltaddr.sin6_addr, ifp, in6m);
1138 			if (in6m == NULL && ia_loop != NULL) {
1139 				rtrequest_global(RTM_ADD,
1140 					  (struct sockaddr *)&mltaddr,
1141 					  (struct sockaddr *)&ia_loop->ia_addr,
1142 					  (struct sockaddr *)&mltmask,
1143 					  RTF_UP);
1144 				in6_addmulti(&mltaddr.sin6_addr, ifp, &error);
1145 				if (error != 0) {
1146 					log(LOG_WARNING, "in6_update_ifa: "
1147 					    "addmulti failed for %s on %s "
1148 					    "(errno=%d)\n",
1149 					    ip6_sprintf(&mltaddr.sin6_addr),
1150 					    if_name(ifp), error);
1151 				}
1152 			}
1153 		}
1154 	}
1155 
1156 	ia->ia6_flags = ifra->ifra_flags;
1157 	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/*safety*/
1158 	ia->ia6_flags &= ~IN6_IFF_NODAD;	/* Mobile IPv6 */
1159 
1160 	ia->ia6_lifetime = ifra->ifra_lifetime;
1161 	/* for sanity */
1162 	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1163 		ia->ia6_lifetime.ia6t_expire =
1164 			time_uptime + ia->ia6_lifetime.ia6t_vltime;
1165 	} else
1166 		ia->ia6_lifetime.ia6t_expire = 0;
1167 	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1168 		ia->ia6_lifetime.ia6t_preferred =
1169 			time_uptime + ia->ia6_lifetime.ia6t_pltime;
1170 	} else
1171 		ia->ia6_lifetime.ia6t_preferred = 0;
1172 
1173 	/*
1174 	 * Perform DAD, if needed.
1175 	 * XXX It may be of use, if we can administratively
1176 	 * disable DAD.
1177 	 */
1178 	if (in6if_do_dad(ifp) && !(ifra->ifra_flags & IN6_IFF_NODAD)) {
1179 		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1180 		nd6_dad_start((struct ifaddr *)ia, NULL);
1181 	}
1182 
1183 	return (error);
1184 
1185 unlink:
1186 	/*
1187 	 * XXX: if a change of an existing address failed, keep the entry
1188 	 * anyway.
1189 	 */
1190 	if (hostIsNew)
1191 		in6_unlink_ifa(ia, ifp);
1192 	return (error);
1193 }
1194 
1195 void
1196 in6_purgeaddr(struct ifaddr *ifa)
1197 {
1198 	struct ifnet *ifp = ifa->ifa_ifp;
1199 	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1200 
1201 	/* stop DAD processing */
1202 	nd6_dad_stop(ifa);
1203 
1204 	/*
1205 	 * delete route to the destination of the address being purged.
1206 	 * The interface must be p2p or loopback in this case.
1207 	 */
1208 	if ((ia->ia_flags & IFA_ROUTE) && ia->ia_dstaddr.sin6_len != 0) {
1209 		int e;
1210 
1211 		if ((e = rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST))
1212 		    != 0) {
1213 			log(LOG_ERR, "in6_purgeaddr: failed to remove "
1214 			    "a route to the p2p destination: %s on %s, "
1215 			    "errno=%d\n",
1216 			    ip6_sprintf(&ia->ia_addr.sin6_addr), if_name(ifp),
1217 			    e);
1218 			/* proceed anyway... */
1219 		}
1220 		else
1221 			ia->ia_flags &= ~IFA_ROUTE;
1222 	}
1223 
1224 	/* Remove ownaddr's loopback rtentry, if it exists. */
1225 	in6_ifremloop(&(ia->ia_ifa));
1226 
1227 	if (ifp->if_flags & IFF_MULTICAST) {
1228 		/*
1229 		 * delete solicited multicast addr for deleting host id
1230 		 */
1231 		struct in6_multi *in6m;
1232 		struct in6_addr llsol;
1233 		bzero(&llsol, sizeof(struct in6_addr));
1234 		llsol.s6_addr16[0] = htons(0xff02);
1235 		llsol.s6_addr16[1] = htons(ifp->if_index);
1236 		llsol.s6_addr32[1] = 0;
1237 		llsol.s6_addr32[2] = htonl(1);
1238 		llsol.s6_addr32[3] =
1239 			ia->ia_addr.sin6_addr.s6_addr32[3];
1240 		llsol.s6_addr8[12] = 0xff;
1241 
1242 		IN6_LOOKUP_MULTI(llsol, ifp, in6m);
1243 		if (in6m)
1244 			in6_delmulti(in6m);
1245 	}
1246 
1247 	in6_unlink_ifa(ia, ifp);
1248 }
1249 
1250 static void
1251 in6_unlink_ifa(struct in6_ifaddr *ia, struct ifnet *ifp)
1252 {
1253 	int plen, iilen;
1254 	struct in6_ifaddr *oia;
1255 
1256 	crit_enter();
1257 
1258 	ifa_ifunlink(&ia->ia_ifa, ifp);
1259 
1260 	oia = ia;
1261 	if (oia == (ia = in6_ifaddr))
1262 		in6_ifaddr = ia->ia_next;
1263 	else {
1264 		while (ia->ia_next && (ia->ia_next != oia))
1265 			ia = ia->ia_next;
1266 		if (ia->ia_next)
1267 			ia->ia_next = oia->ia_next;
1268 		else {
1269 			/* search failed */
1270 			kprintf("Couldn't unlink in6_ifaddr from in6_ifaddr\n");
1271 		}
1272 	}
1273 
1274 	if (oia->ia6_ifpr) {	/* check for safety */
1275 		plen = in6_mask2len(&oia->ia_prefixmask.sin6_addr, NULL);
1276 		iilen = (sizeof(oia->ia_prefixmask.sin6_addr) << 3) - plen;
1277 		in6_prefix_remove_ifid(iilen, oia);
1278 	}
1279 
1280 	/*
1281 	 * When an autoconfigured address is being removed, release the
1282 	 * reference to the base prefix.  Also, since the release might
1283 	 * affect the status of other (detached) addresses, call
1284 	 * pfxlist_onlink_check().
1285 	 */
1286 	if (oia->ia6_flags & IN6_IFF_AUTOCONF) {
1287 		if (oia->ia6_ndpr == NULL) {
1288 			log(LOG_NOTICE, "in6_unlink_ifa: autoconf'ed address "
1289 			    "%p has no prefix\n", oia);
1290 		} else {
1291 			oia->ia6_ndpr->ndpr_refcnt--;
1292 			oia->ia6_flags &= ~IN6_IFF_AUTOCONF;
1293 			oia->ia6_ndpr = NULL;
1294 		}
1295 
1296 		pfxlist_onlink_check();
1297 	}
1298 
1299 	/*
1300 	 * release another refcnt for the link from in6_ifaddr.
1301 	 * Note that we should decrement the refcnt at least once for all *BSD.
1302 	 */
1303 	ifa_destroy(&oia->ia_ifa);
1304 
1305 	crit_exit();
1306 }
1307 
1308 void
1309 in6_purgeif(struct ifnet *ifp)
1310 {
1311 	struct ifaddr_container *ifac, *next;
1312 
1313 	TAILQ_FOREACH_MUTABLE(ifac, &ifp->if_addrheads[mycpuid],
1314 			      ifa_link, next) {
1315 		if (ifac->ifa->ifa_addr->sa_family != AF_INET6)
1316 			continue;
1317 		in6_purgeaddr(ifac->ifa);
1318 	}
1319 
1320 	in6_ifdetach(ifp);
1321 }
1322 
1323 /*
1324  * SIOC[GAD]LIFADDR.
1325  *	SIOCGLIFADDR: get first address. (?)
1326  *	SIOCGLIFADDR with IFLR_PREFIX:
1327  *		get first address that matches the specified prefix.
1328  *	SIOCALIFADDR: add the specified address.
1329  *	SIOCALIFADDR with IFLR_PREFIX:
1330  *		add the specified prefix, filling hostid part from
1331  *		the first link-local address.  prefixlen must be <= 64.
1332  *	SIOCDLIFADDR: delete the specified address.
1333  *	SIOCDLIFADDR with IFLR_PREFIX:
1334  *		delete the first address that matches the specified prefix.
1335  * return values:
1336  *	EINVAL on invalid parameters
1337  *	EADDRNOTAVAIL on prefix match failed/specified address not found
1338  *	other values may be returned from in6_ioctl()
1339  *
1340  * NOTE: SIOCALIFADDR(with IFLR_PREFIX set) allows prefixlen less than 64.
1341  * this is to accomodate address naming scheme other than RFC2374,
1342  * in the future.
1343  * RFC2373 defines interface id to be 64bit, but it allows non-RFC2374
1344  * address encoding scheme. (see figure on page 8)
1345  */
1346 static int
1347 in6_lifaddr_ioctl(struct socket *so, u_long cmd, caddr_t data,
1348 		  struct ifnet *ifp, struct thread *td)
1349 {
1350 	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
1351 	struct sockaddr *sa;
1352 
1353 	/* sanity checks */
1354 	if (!data || !ifp) {
1355 		panic("invalid argument to in6_lifaddr_ioctl");
1356 		/*NOTRECHED*/
1357 	}
1358 
1359 	switch (cmd) {
1360 	case SIOCGLIFADDR:
1361 		/* address must be specified on GET with IFLR_PREFIX */
1362 		if (!(iflr->flags & IFLR_PREFIX))
1363 			break;
1364 		/* FALLTHROUGH */
1365 	case SIOCALIFADDR:
1366 	case SIOCDLIFADDR:
1367 		/* address must be specified on ADD and DELETE */
1368 		sa = (struct sockaddr *)&iflr->addr;
1369 		if (sa->sa_family != AF_INET6)
1370 			return EINVAL;
1371 		if (sa->sa_len != sizeof(struct sockaddr_in6))
1372 			return EINVAL;
1373 		/* XXX need improvement */
1374 		sa = (struct sockaddr *)&iflr->dstaddr;
1375 		if (sa->sa_family && sa->sa_family != AF_INET6)
1376 			return EINVAL;
1377 		if (sa->sa_len && sa->sa_len != sizeof(struct sockaddr_in6))
1378 			return EINVAL;
1379 		break;
1380 	default: /* shouldn't happen */
1381 #if 0
1382 		panic("invalid cmd to in6_lifaddr_ioctl");
1383 		/* NOTREACHED */
1384 #else
1385 		return EOPNOTSUPP;
1386 #endif
1387 	}
1388 	if (sizeof(struct in6_addr) * 8 < iflr->prefixlen)
1389 		return EINVAL;
1390 
1391 	switch (cmd) {
1392 	case SIOCALIFADDR:
1393 	    {
1394 		struct in6_aliasreq ifra;
1395 		struct in6_addr *hostid = NULL;
1396 		int prefixlen;
1397 
1398 		if (iflr->flags & IFLR_PREFIX) {
1399 			struct ifaddr *ifa;
1400 			struct sockaddr_in6 *sin6;
1401 
1402 			/*
1403 			 * hostid is to fill in the hostid part of the
1404 			 * address.  hostid points to the first link-local
1405 			 * address attached to the interface.
1406 			 */
1407 			ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0);
1408 			if (!ifa)
1409 				return EADDRNOTAVAIL;
1410 			hostid = IFA_IN6(ifa);
1411 
1412 		 	/* prefixlen must be <= 64. */
1413 			if (64 < iflr->prefixlen)
1414 				return EINVAL;
1415 			prefixlen = iflr->prefixlen;
1416 
1417 			/* hostid part must be zero. */
1418 			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1419 			if (sin6->sin6_addr.s6_addr32[2] != 0
1420 			 || sin6->sin6_addr.s6_addr32[3] != 0) {
1421 				return EINVAL;
1422 			}
1423 		} else
1424 			prefixlen = iflr->prefixlen;
1425 
1426 		/* copy args to in6_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
1427 		bzero(&ifra, sizeof(ifra));
1428 		bcopy(iflr->iflr_name, ifra.ifra_name,
1429 			sizeof(ifra.ifra_name));
1430 
1431 		bcopy(&iflr->addr, &ifra.ifra_addr,
1432 			((struct sockaddr *)&iflr->addr)->sa_len);
1433 		if (hostid) {
1434 			/* fill in hostid part */
1435 			ifra.ifra_addr.sin6_addr.s6_addr32[2] =
1436 				hostid->s6_addr32[2];
1437 			ifra.ifra_addr.sin6_addr.s6_addr32[3] =
1438 				hostid->s6_addr32[3];
1439 		}
1440 
1441 		if (((struct sockaddr *)&iflr->dstaddr)->sa_family) {	/*XXX*/
1442 			bcopy(&iflr->dstaddr, &ifra.ifra_dstaddr,
1443 				((struct sockaddr *)&iflr->dstaddr)->sa_len);
1444 			if (hostid) {
1445 				ifra.ifra_dstaddr.sin6_addr.s6_addr32[2] =
1446 					hostid->s6_addr32[2];
1447 				ifra.ifra_dstaddr.sin6_addr.s6_addr32[3] =
1448 					hostid->s6_addr32[3];
1449 			}
1450 		}
1451 
1452 		ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
1453 		in6_len2mask(&ifra.ifra_prefixmask.sin6_addr, prefixlen);
1454 
1455 		ifra.ifra_flags = iflr->flags & ~IFLR_PREFIX;
1456 		return in6_control(so, SIOCAIFADDR_IN6, (caddr_t)&ifra, ifp, td);
1457 	    }
1458 	case SIOCGLIFADDR:
1459 	case SIOCDLIFADDR:
1460 	    {
1461 		struct ifaddr_container *ifac;
1462 		struct in6_ifaddr *ia;
1463 		struct in6_addr mask, candidate, match;
1464 		struct sockaddr_in6 *sin6;
1465 		int cmp;
1466 
1467 		bzero(&mask, sizeof(mask));
1468 		if (iflr->flags & IFLR_PREFIX) {
1469 			/* lookup a prefix rather than address. */
1470 			in6_len2mask(&mask, iflr->prefixlen);
1471 
1472 			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1473 			bcopy(&sin6->sin6_addr, &match, sizeof(match));
1474 			match.s6_addr32[0] &= mask.s6_addr32[0];
1475 			match.s6_addr32[1] &= mask.s6_addr32[1];
1476 			match.s6_addr32[2] &= mask.s6_addr32[2];
1477 			match.s6_addr32[3] &= mask.s6_addr32[3];
1478 
1479 			/* if you set extra bits, that's wrong */
1480 			if (bcmp(&match, &sin6->sin6_addr, sizeof(match)))
1481 				return EINVAL;
1482 
1483 			cmp = 1;
1484 		} else {
1485 			if (cmd == SIOCGLIFADDR) {
1486 				/* on getting an address, take the 1st match */
1487 				cmp = 0;	/* XXX */
1488 			} else {
1489 				/* on deleting an address, do exact match */
1490 				in6_len2mask(&mask, 128);
1491 				sin6 = (struct sockaddr_in6 *)&iflr->addr;
1492 				bcopy(&sin6->sin6_addr, &match, sizeof(match));
1493 
1494 				cmp = 1;
1495 			}
1496 		}
1497 
1498 		TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
1499 			struct ifaddr *ifa = ifac->ifa;
1500 
1501 			if (ifa->ifa_addr->sa_family != AF_INET6)
1502 				continue;
1503 			if (!cmp)
1504 				break;
1505 
1506 			bcopy(IFA_IN6(ifa), &candidate, sizeof(candidate));
1507 			/*
1508 			 * XXX: this is adhoc, but is necessary to allow
1509 			 * a user to specify fe80::/64 (not /10) for a
1510 			 * link-local address.
1511 			 */
1512 			if (IN6_IS_ADDR_LINKLOCAL(&candidate))
1513 				candidate.s6_addr16[1] = 0;
1514 			candidate.s6_addr32[0] &= mask.s6_addr32[0];
1515 			candidate.s6_addr32[1] &= mask.s6_addr32[1];
1516 			candidate.s6_addr32[2] &= mask.s6_addr32[2];
1517 			candidate.s6_addr32[3] &= mask.s6_addr32[3];
1518 			if (IN6_ARE_ADDR_EQUAL(&candidate, &match))
1519 				break;
1520 		}
1521 		if (ifac == NULL)
1522 			return EADDRNOTAVAIL;
1523 		ia = ifa2ia6(ifac->ifa);
1524 
1525 		if (cmd == SIOCGLIFADDR) {
1526 			struct sockaddr_in6 *s6;
1527 
1528 			/* fill in the if_laddrreq structure */
1529 			bcopy(&ia->ia_addr, &iflr->addr, ia->ia_addr.sin6_len);
1530 			s6 = (struct sockaddr_in6 *)&iflr->addr;
1531 			if (IN6_IS_ADDR_LINKLOCAL(&s6->sin6_addr)) {
1532 				s6->sin6_addr.s6_addr16[1] = 0;
1533 				s6->sin6_scope_id =
1534 					in6_addr2scopeid(ifp, &s6->sin6_addr);
1535 			}
1536 			if (ifp->if_flags & IFF_POINTOPOINT) {
1537 				bcopy(&ia->ia_dstaddr, &iflr->dstaddr,
1538 					ia->ia_dstaddr.sin6_len);
1539 				s6 = (struct sockaddr_in6 *)&iflr->dstaddr;
1540 				if (IN6_IS_ADDR_LINKLOCAL(&s6->sin6_addr)) {
1541 					s6->sin6_addr.s6_addr16[1] = 0;
1542 					s6->sin6_scope_id =
1543 						in6_addr2scopeid(ifp,
1544 								 &s6->sin6_addr);
1545 				}
1546 			} else
1547 				bzero(&iflr->dstaddr, sizeof(iflr->dstaddr));
1548 
1549 			iflr->prefixlen =
1550 				in6_mask2len(&ia->ia_prefixmask.sin6_addr,
1551 					     NULL);
1552 
1553 			iflr->flags = ia->ia6_flags;	/* XXX */
1554 
1555 			return 0;
1556 		} else {
1557 			struct in6_aliasreq ifra;
1558 
1559 			/* fill in6_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
1560 			bzero(&ifra, sizeof(ifra));
1561 			bcopy(iflr->iflr_name, ifra.ifra_name,
1562 			      sizeof(ifra.ifra_name));
1563 
1564 			bcopy(&ia->ia_addr, &ifra.ifra_addr,
1565 			      ia->ia_addr.sin6_len);
1566 			if (ifp->if_flags & IFF_POINTOPOINT)
1567 				bcopy(&ia->ia_dstaddr, &ifra.ifra_dstaddr,
1568 				      ia->ia_dstaddr.sin6_len);
1569 			else
1570 				bzero(&ifra.ifra_dstaddr,
1571 				      sizeof(ifra.ifra_dstaddr));
1572 			bcopy(&ia->ia_prefixmask, &ifra.ifra_dstaddr,
1573 			      ia->ia_prefixmask.sin6_len);
1574 
1575 			ifra.ifra_flags = ia->ia6_flags;
1576 			return in6_control(so, SIOCDIFADDR_IN6, (caddr_t)&ifra,
1577 				ifp, td);
1578 		}
1579 	    }
1580 	}
1581 
1582 	return EOPNOTSUPP;	/* just for safety */
1583 }
1584 
1585 /*
1586  * Initialize an interface's intetnet6 address
1587  * and routing table entry.
1588  */
1589 static int
1590 in6_ifinit(struct ifnet *ifp, struct in6_ifaddr *ia, struct sockaddr_in6 *sin6,
1591 	   int newhost)
1592 {
1593 	int	error = 0, plen, ifacount = 0;
1594 	struct ifaddr_container *ifac;
1595 
1596 	/*
1597 	 * Give the interface a chance to initialize
1598 	 * if this is its first address,
1599 	 * and to validate the address if necessary.
1600 	 */
1601 	TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
1602 		if (ifac->ifa->ifa_addr == NULL)
1603 			continue;	/* just for safety */
1604 		if (ifac->ifa->ifa_addr->sa_family != AF_INET6)
1605 			continue;
1606 		ifacount++;
1607 	}
1608 
1609 	ifnet_serialize_all(ifp);
1610 
1611 	ia->ia_addr = *sin6;
1612 
1613 	if (ifacount <= 1 && ifp->if_ioctl &&
1614 	    (error = ifp->if_ioctl(ifp, SIOCSIFADDR, (caddr_t)ia, NULL))) {
1615 		ifnet_deserialize_all(ifp);
1616 		return (error);
1617 	}
1618 
1619 	ifnet_deserialize_all(ifp);
1620 
1621 	ia->ia_ifa.ifa_metric = ifp->if_metric;
1622 
1623 	/* we could do in(6)_socktrim here, but just omit it at this moment. */
1624 
1625 	/*
1626 	 * Special case:
1627 	 * If the destination address is specified for a point-to-point
1628 	 * interface, install a route to the destination as an interface
1629 	 * direct route.
1630 	 */
1631 	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1632 	if (plen == 128 && ia->ia_dstaddr.sin6_family == AF_INET6) {
1633 		if ((error = rtinit(&(ia->ia_ifa), (int)RTM_ADD,
1634 				    RTF_UP | RTF_HOST)) != 0)
1635 			return (error);
1636 		ia->ia_flags |= IFA_ROUTE;
1637 	}
1638 	if (plen < 128) {
1639 		/*
1640 		 * The RTF_CLONING flag is necessary for in6_is_ifloop_auto().
1641 		 */
1642 		ia->ia_ifa.ifa_flags |= RTF_CLONING;
1643 	}
1644 
1645 	/* Add ownaddr as loopback rtentry, if necessary (ex. on p2p link). */
1646 	if (newhost) {
1647 		/* set the rtrequest function to create llinfo */
1648 		ia->ia_ifa.ifa_rtrequest = nd6_rtrequest;
1649 		in6_ifaddloop(&(ia->ia_ifa));
1650 	}
1651 
1652 	return (error);
1653 }
1654 
1655 struct in6_multi_mship *
1656 in6_joingroup(struct ifnet *ifp, struct in6_addr *addr, int *errorp)
1657 {
1658        struct in6_multi_mship *imm;
1659 
1660        imm = kmalloc(sizeof(*imm), M_IPMADDR, M_NOWAIT);
1661        if (!imm) {
1662                *errorp = ENOBUFS;
1663                return NULL;
1664        }
1665        imm->i6mm_maddr = in6_addmulti(addr, ifp, errorp);
1666        if (!imm->i6mm_maddr) {
1667                /* *errorp is alrady set */
1668                kfree(imm, M_IPMADDR);
1669                return NULL;
1670        }
1671        return imm;
1672 }
1673 
1674 int
1675 in6_leavegroup(struct in6_multi_mship *imm)
1676 {
1677 
1678        if (imm->i6mm_maddr)
1679                in6_delmulti(imm->i6mm_maddr);
1680        kfree(imm,  M_IPMADDR);
1681        return 0;
1682 }
1683 
1684 /*
1685  * Add an address to the list of IP6 multicast addresses for a
1686  * given interface.
1687  */
1688 struct	in6_multi *
1689 in6_addmulti(struct in6_addr *maddr6, struct ifnet *ifp, int *errorp)
1690 {
1691 	struct	in6_multi *in6m;
1692 	struct sockaddr_in6 sin6;
1693 	struct ifmultiaddr *ifma;
1694 
1695 	*errorp = 0;
1696 
1697 	crit_enter();
1698 
1699 	/*
1700 	 * Call generic routine to add membership or increment
1701 	 * refcount.  It wants addresses in the form of a sockaddr,
1702 	 * so we build one here (being careful to zero the unused bytes).
1703 	 */
1704 	bzero(&sin6, sizeof sin6);
1705 	sin6.sin6_family = AF_INET6;
1706 	sin6.sin6_len = sizeof sin6;
1707 	sin6.sin6_addr = *maddr6;
1708 	*errorp = if_addmulti(ifp, (struct sockaddr *)&sin6, &ifma);
1709 	if (*errorp) {
1710 		crit_exit();
1711 		return 0;
1712 	}
1713 
1714 	/*
1715 	 * If ifma->ifma_protospec is null, then if_addmulti() created
1716 	 * a new record.  Otherwise, we are done.
1717 	 */
1718 	if (ifma->ifma_protospec != NULL) {
1719 		crit_exit();
1720 		return ifma->ifma_protospec;
1721 	}
1722 
1723 	/* XXX - if_addmulti uses M_WAITOK.  Can this really be called
1724 	   at interrupt time?  If so, need to fix if_addmulti. XXX */
1725 	in6m = (struct in6_multi *)kmalloc(sizeof(*in6m), M_IPMADDR, M_NOWAIT);
1726 	if (in6m == NULL) {
1727 		crit_exit();
1728 		return (NULL);
1729 	}
1730 
1731 	bzero(in6m, sizeof *in6m);
1732 	in6m->in6m_addr = *maddr6;
1733 	in6m->in6m_ifp = ifp;
1734 	in6m->in6m_ifma = ifma;
1735 	ifma->ifma_protospec = in6m;
1736 	LIST_INSERT_HEAD(&in6_multihead, in6m, in6m_entry);
1737 
1738 	/*
1739 	 * Let MLD6 know that we have joined a new IP6 multicast
1740 	 * group.
1741 	 */
1742 	mld6_start_listening(in6m);
1743 	crit_exit();
1744 	return (in6m);
1745 }
1746 
1747 /*
1748  * Delete a multicast address record.
1749  */
1750 void
1751 in6_delmulti(struct in6_multi *in6m)
1752 {
1753 	struct ifmultiaddr *ifma = in6m->in6m_ifma;
1754 
1755 	crit_enter();
1756 
1757 	if (ifma->ifma_refcount == 1) {
1758 		/*
1759 		 * No remaining claims to this record; let MLD6 know
1760 		 * that we are leaving the multicast group.
1761 		 */
1762 		mld6_stop_listening(in6m);
1763 		ifma->ifma_protospec = NULL;
1764 		LIST_REMOVE(in6m, in6m_entry);
1765 		kfree(in6m, M_IPMADDR);
1766 	}
1767 	/* XXX - should be separate API for when we have an ifma? */
1768 	if_delmulti(ifma->ifma_ifp, ifma->ifma_addr);
1769 	crit_exit();
1770 }
1771 
1772 /*
1773  * Find an IPv6 interface link-local address specific to an interface.
1774  */
1775 struct in6_ifaddr *
1776 in6ifa_ifpforlinklocal(struct ifnet *ifp, int ignoreflags)
1777 {
1778 	struct ifaddr_container *ifac;
1779 
1780 	TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
1781 		struct ifaddr *ifa = ifac->ifa;
1782 
1783 		if (ifa->ifa_addr == NULL)
1784 			continue;	/* just for safety */
1785 		if (ifa->ifa_addr->sa_family != AF_INET6)
1786 			continue;
1787 		if (IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa))) {
1788 			if ((((struct in6_ifaddr *)ifa)->ia6_flags &
1789 			     ignoreflags) != 0)
1790 				continue;
1791 			break;
1792 		}
1793 	}
1794 	if (ifac != NULL)
1795 		return ((struct in6_ifaddr *)(ifac->ifa));
1796 	else
1797 		return (NULL);
1798 }
1799 
1800 
1801 /*
1802  * find the internet address corresponding to a given interface and address.
1803  */
1804 struct in6_ifaddr *
1805 in6ifa_ifpwithaddr(struct ifnet *ifp, struct in6_addr *addr)
1806 {
1807 	struct ifaddr_container *ifac;
1808 
1809 	TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
1810 		struct ifaddr *ifa = ifac->ifa;
1811 
1812 		if (ifa->ifa_addr == NULL)
1813 			continue;	/* just for safety */
1814 		if (ifa->ifa_addr->sa_family != AF_INET6)
1815 			continue;
1816 		if (IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa)))
1817 			break;
1818 	}
1819 	if (ifac != NULL)
1820 		return ((struct in6_ifaddr *)(ifac->ifa));
1821 	else
1822 		return (NULL);
1823 }
1824 
1825 /*
1826  * find the internet address on a given interface corresponding to a neighbor's
1827  * address.
1828  */
1829 struct in6_ifaddr *
1830 in6ifa_ifplocaladdr(const struct ifnet *ifp, const struct in6_addr *addr)
1831 {
1832 	struct ifaddr *ifa;
1833 	struct in6_ifaddr *ia;
1834 	struct ifaddr_container *ifac;
1835 
1836 	TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
1837 		ifa = ifac->ifa;
1838 
1839 		if (ifa->ifa_addr == NULL)
1840 			continue;	/* just for safety */
1841 		if (ifa->ifa_addr->sa_family != AF_INET6)
1842 			continue;
1843 		ia = (struct in6_ifaddr *)ifa;
1844 		if (IN6_ARE_MASKED_ADDR_EQUAL(addr,
1845 				&ia->ia_addr.sin6_addr,
1846 				&ia->ia_prefixmask.sin6_addr))
1847 			return ia;
1848 	}
1849 
1850 	return NULL;
1851 }
1852 
1853 /*
1854  * Convert IP6 address to printable (loggable) representation.
1855  */
1856 static char digits[] = "0123456789abcdef";
1857 static int ip6round = 0;
1858 char *
1859 ip6_sprintf(const struct in6_addr *addr)
1860 {
1861 	static char ip6buf[8][48];
1862 	int i;
1863 	char *cp;
1864 	const u_short *a = (const u_short *)addr;
1865 	const u_char *d;
1866 	int dcolon = 0;
1867 
1868 	ip6round = (ip6round + 1) & 7;
1869 	cp = ip6buf[ip6round];
1870 
1871 	for (i = 0; i < 8; i++) {
1872 		if (dcolon == 1) {
1873 			if (*a == 0) {
1874 				if (i == 7)
1875 					*cp++ = ':';
1876 				a++;
1877 				continue;
1878 			} else
1879 				dcolon = 2;
1880 		}
1881 		if (*a == 0) {
1882 			if (dcolon == 0 && *(a + 1) == 0) {
1883 				if (i == 0)
1884 					*cp++ = ':';
1885 				*cp++ = ':';
1886 				dcolon = 1;
1887 			} else {
1888 				*cp++ = '0';
1889 				*cp++ = ':';
1890 			}
1891 			a++;
1892 			continue;
1893 		}
1894 		d = (const u_char *)a;
1895 		*cp++ = digits[*d >> 4];
1896 		*cp++ = digits[*d++ & 0xf];
1897 		*cp++ = digits[*d >> 4];
1898 		*cp++ = digits[*d & 0xf];
1899 		*cp++ = ':';
1900 		a++;
1901 	}
1902 	*--cp = 0;
1903 	return (ip6buf[ip6round]);
1904 }
1905 
1906 int
1907 in6_localaddr(struct in6_addr *in6)
1908 {
1909 	struct in6_ifaddr *ia;
1910 
1911 	if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
1912 		return 1;
1913 
1914 	for (ia = in6_ifaddr; ia; ia = ia->ia_next)
1915 		if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
1916 					      &ia->ia_prefixmask.sin6_addr))
1917 			return 1;
1918 
1919 	return (0);
1920 }
1921 
1922 int
1923 in6_is_addr_deprecated(struct sockaddr_in6 *sa6)
1924 {
1925 	struct in6_ifaddr *ia;
1926 
1927 	for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
1928 		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
1929 				       &sa6->sin6_addr) &&
1930 		    (ia->ia6_flags & IN6_IFF_DEPRECATED))
1931 			return (1); /* true */
1932 
1933 		/* XXX: do we still have to go thru the rest of the list? */
1934 	}
1935 
1936 	return (0);		/* false */
1937 }
1938 
1939 /*
1940  * return length of part which dst and src are equal
1941  * hard coding...
1942  */
1943 int
1944 in6_matchlen(struct in6_addr *src, struct in6_addr *dst)
1945 {
1946 	int match = 0;
1947 	u_char *s = (u_char *)src, *d = (u_char *)dst;
1948 	u_char *lim = s + 16, r;
1949 
1950 	while (s < lim)
1951 		if ((r = (*d++ ^ *s++)) != 0) {
1952 			while (r < 128) {
1953 				match++;
1954 				r <<= 1;
1955 			}
1956 			break;
1957 		} else
1958 			match += 8;
1959 	return match;
1960 }
1961 
1962 /* XXX: to be scope conscious */
1963 int
1964 in6_are_prefix_equal(struct in6_addr *p1, struct in6_addr *p2, int len)
1965 {
1966 	int bytelen, bitlen;
1967 
1968 	/* sanity check */
1969 	if (0 > len || len > 128) {
1970 		log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n",
1971 		    len);
1972 		return (0);
1973 	}
1974 
1975 	bytelen = len / 8;
1976 	bitlen = len % 8;
1977 
1978 	if (bcmp(&p1->s6_addr, &p2->s6_addr, bytelen))
1979 		return (0);
1980 	if (p1->s6_addr[bytelen] >> (8 - bitlen) !=
1981 	    p2->s6_addr[bytelen] >> (8 - bitlen))
1982 		return (0);
1983 
1984 	return (1);
1985 }
1986 
1987 void
1988 in6_prefixlen2mask(struct in6_addr *maskp, int len)
1989 {
1990 	u_char maskarray[8] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
1991 	int bytelen, bitlen, i;
1992 
1993 	/* sanity check */
1994 	if (0 > len || len > 128) {
1995 		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
1996 		    len);
1997 		return;
1998 	}
1999 
2000 	bzero(maskp, sizeof(*maskp));
2001 	bytelen = len / 8;
2002 	bitlen = len % 8;
2003 	for (i = 0; i < bytelen; i++)
2004 		maskp->s6_addr[i] = 0xff;
2005 	if (bitlen)
2006 		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
2007 }
2008 
2009 /*
2010  * return the best address out of the same scope
2011  */
2012 struct in6_ifaddr *
2013 in6_ifawithscope(struct ifnet *oifp, struct in6_addr *dst)
2014 {
2015 	int dst_scope =	in6_addrscope(dst), src_scope, best_scope = 0;
2016 	int blen = -1;
2017 	struct ifnet *ifp;
2018 	struct in6_ifaddr *ifa_best = NULL;
2019 
2020 	if (oifp == NULL) {
2021 #if 0
2022 		kprintf("in6_ifawithscope: output interface is not specified\n");
2023 #endif
2024 		return (NULL);
2025 	}
2026 
2027 	/*
2028 	 * We search for all addresses on all interfaces from the beginning.
2029 	 * Comparing an interface with the outgoing interface will be done
2030 	 * only at the final stage of tiebreaking.
2031 	 */
2032 	for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list))
2033 	{
2034 		struct ifaddr_container *ifac;
2035 
2036 		/*
2037 		 * We can never take an address that breaks the scope zone
2038 		 * of the destination.
2039 		 */
2040 		if (ifp->if_afdata[AF_INET6] == NULL)
2041 			continue;
2042 		if (in6_addr2scopeid(ifp, dst) != in6_addr2scopeid(oifp, dst))
2043 			continue;
2044 
2045 		TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
2046 			int tlen = -1, dscopecmp, bscopecmp, matchcmp;
2047 			struct ifaddr *ifa = ifac->ifa;
2048 
2049 			if (ifa->ifa_addr->sa_family != AF_INET6)
2050 				continue;
2051 
2052 			src_scope = in6_addrscope(IFA_IN6(ifa));
2053 
2054 			/*
2055 			 * Don't use an address before completing DAD
2056 			 * nor a duplicated address.
2057 			 */
2058 			if (((struct in6_ifaddr *)ifa)->ia6_flags &
2059 			    IN6_IFF_NOTREADY)
2060 				continue;
2061 
2062 			/* XXX: is there any case to allow anycasts? */
2063 			if (((struct in6_ifaddr *)ifa)->ia6_flags &
2064 			    IN6_IFF_ANYCAST)
2065 				continue;
2066 
2067 			if (((struct in6_ifaddr *)ifa)->ia6_flags &
2068 			    IN6_IFF_DETACHED)
2069 				continue;
2070 
2071 			/*
2072 			 * If this is the first address we find,
2073 			 * keep it anyway.
2074 			 */
2075 			if (ifa_best == NULL)
2076 				goto replace;
2077 
2078 			/*
2079 			 * ifa_best is never NULL beyond this line except
2080 			 * within the block labeled "replace".
2081 			 */
2082 
2083 			/*
2084 			 * If ifa_best has a smaller scope than dst and
2085 			 * the current address has a larger one than
2086 			 * (or equal to) dst, always replace ifa_best.
2087 			 * Also, if the current address has a smaller scope
2088 			 * than dst, ignore it unless ifa_best also has a
2089 			 * smaller scope.
2090 			 * Consequently, after the two if-clause below,
2091 			 * the followings must be satisfied:
2092 			 * (scope(src) < scope(dst) &&
2093 			 *  scope(best) < scope(dst))
2094 			 *  OR
2095 			 * (scope(best) >= scope(dst) &&
2096 			 *  scope(src) >= scope(dst))
2097 			 */
2098 			if (IN6_ARE_SCOPE_CMP(best_scope, dst_scope) < 0 &&
2099 			    IN6_ARE_SCOPE_CMP(src_scope, dst_scope) >= 0)
2100 				goto replace; /* (A) */
2101 			if (IN6_ARE_SCOPE_CMP(src_scope, dst_scope) < 0 &&
2102 			    IN6_ARE_SCOPE_CMP(best_scope, dst_scope) >= 0)
2103 				continue; /* (B) */
2104 
2105 			/*
2106 			 * A deprecated address SHOULD NOT be used in new
2107 			 * communications if an alternate (non-deprecated)
2108 			 * address is available and has sufficient scope.
2109 			 * RFC 2462, Section 5.5.4.
2110 			 */
2111 			if (((struct in6_ifaddr *)ifa)->ia6_flags &
2112 			    IN6_IFF_DEPRECATED) {
2113 				/*
2114 				 * Ignore any deprecated addresses if
2115 				 * specified by configuration.
2116 				 */
2117 				if (!ip6_use_deprecated)
2118 					continue;
2119 
2120 				/*
2121 				 * If we have already found a non-deprecated
2122 				 * candidate, just ignore deprecated addresses.
2123 				 */
2124 				if (!(ifa_best->ia6_flags & IN6_IFF_DEPRECATED))
2125 					continue;
2126 			}
2127 
2128 			/*
2129 			 * A non-deprecated address is always preferred
2130 			 * to a deprecated one regardless of scopes and
2131 			 * address matching (Note invariants ensured by the
2132 			 * conditions (A) and (B) above.)
2133 			 */
2134 			if ((ifa_best->ia6_flags & IN6_IFF_DEPRECATED) &&
2135 			    !(((struct in6_ifaddr *)ifa)->ia6_flags &
2136 			     IN6_IFF_DEPRECATED))
2137 				goto replace;
2138 
2139 			/*
2140 			 * When we use temporary addresses described in
2141 			 * RFC 3041, we prefer temporary addresses to
2142 			 * public autoconf addresses.  Again, note the
2143 			 * invariants from (A) and (B).  Also note that we
2144 			 * don't have any preference between static addresses
2145 			 * and autoconf addresses (despite of whether or not
2146 			 * the latter is temporary or public.)
2147 			 */
2148 			if (ip6_use_tempaddr) {
2149 				struct in6_ifaddr *ifat;
2150 
2151 				ifat = (struct in6_ifaddr *)ifa;
2152 				if ((ifa_best->ia6_flags &
2153 				     (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY))
2154 				     == IN6_IFF_AUTOCONF &&
2155 				    (ifat->ia6_flags &
2156 				     (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY))
2157 				     == (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY)) {
2158 					goto replace;
2159 				}
2160 				if ((ifa_best->ia6_flags &
2161 				     (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY))
2162 				    == (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY) &&
2163 				    (ifat->ia6_flags &
2164 				     (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY))
2165 				     == IN6_IFF_AUTOCONF) {
2166 					continue;
2167 				}
2168 			}
2169 
2170 			/*
2171 			 * At this point, we have two cases:
2172 			 * 1. we are looking at a non-deprecated address,
2173 			 *    and ifa_best is also non-deprecated.
2174 			 * 2. we are looking at a deprecated address,
2175 			 *    and ifa_best is also deprecated.
2176 			 * Also, we do not have to consider a case where
2177 			 * the scope of if_best is larger(smaller) than dst and
2178 			 * the scope of the current address is smaller(larger)
2179 			 * than dst. Such a case has already been covered.
2180 			 * Tiebreaking is done according to the following
2181 			 * items:
2182 			 * - the scope comparison between the address and
2183 			 *   dst (dscopecmp)
2184 			 * - the scope comparison between the address and
2185 			 *   ifa_best (bscopecmp)
2186 			 * - if the address match dst longer than ifa_best
2187 			 *   (matchcmp)
2188 			 * - if the address is on the outgoing I/F (outI/F)
2189 			 *
2190 			 * Roughly speaking, the selection policy is
2191 			 * - the most important item is scope. The same scope
2192 			 *   is best. Then search for a larger scope.
2193 			 *   Smaller scopes are the last resort.
2194 			 * - A deprecated address is chosen only when we have
2195 			 *   no address that has an enough scope, but is
2196 			 *   prefered to any addresses of smaller scopes
2197 			 *   (this must be already done above.)
2198 			 * - addresses on the outgoing I/F are preferred to
2199 			 *   ones on other interfaces if none of above
2200 			 *   tiebreaks.  In the table below, the column "bI"
2201 			 *   means if the best_ifa is on the outgoing
2202 			 *   interface, and the column "sI" means if the ifa
2203 			 *   is on the outgoing interface.
2204 			 * - If there is no other reasons to choose one,
2205 			 *   longest address match against dst is considered.
2206 			 *
2207 			 * The precise decision table is as follows:
2208 			 * dscopecmp bscopecmp    match  bI oI | replace?
2209 			 *       N/A     equal      N/A   Y  N |   No (1)
2210 			 *       N/A     equal      N/A   N  Y |  Yes (2)
2211 			 *       N/A     equal   larger    N/A |  Yes (3)
2212 			 *       N/A     equal  !larger    N/A |   No (4)
2213 			 *    larger    larger      N/A    N/A |   No (5)
2214 			 *    larger   smaller      N/A    N/A |  Yes (6)
2215 			 *   smaller    larger      N/A    N/A |  Yes (7)
2216 			 *   smaller   smaller      N/A    N/A |   No (8)
2217 			 *     equal   smaller      N/A    N/A |  Yes (9)
2218 			 *     equal    larger       (already done at A above)
2219 			 */
2220 			dscopecmp = IN6_ARE_SCOPE_CMP(src_scope, dst_scope);
2221 			bscopecmp = IN6_ARE_SCOPE_CMP(src_scope, best_scope);
2222 
2223 			if (bscopecmp == 0) {
2224 				struct ifnet *bifp = ifa_best->ia_ifp;
2225 
2226 				if (bifp == oifp && ifp != oifp) /* (1) */
2227 					continue;
2228 				if (bifp != oifp && ifp == oifp) /* (2) */
2229 					goto replace;
2230 
2231 				/*
2232 				 * Both bifp and ifp are on the outgoing
2233 				 * interface, or both two are on a different
2234 				 * interface from the outgoing I/F.
2235 				 * now we need address matching against dst
2236 				 * for tiebreaking.
2237 				 */
2238 				tlen = in6_matchlen(IFA_IN6(ifa), dst);
2239 				matchcmp = tlen - blen;
2240 				if (matchcmp > 0) /* (3) */
2241 					goto replace;
2242 				continue; /* (4) */
2243 			}
2244 			if (dscopecmp > 0) {
2245 				if (bscopecmp > 0) /* (5) */
2246 					continue;
2247 				goto replace; /* (6) */
2248 			}
2249 			if (dscopecmp < 0) {
2250 				if (bscopecmp > 0) /* (7) */
2251 					goto replace;
2252 				continue; /* (8) */
2253 			}
2254 
2255 			/* now dscopecmp must be 0 */
2256 			if (bscopecmp < 0)
2257 				goto replace; /* (9) */
2258 
2259 replace:
2260 			ifa_best = (struct in6_ifaddr *)ifa;
2261 			blen = tlen >= 0 ? tlen :
2262 				in6_matchlen(IFA_IN6(ifa), dst);
2263 			best_scope = in6_addrscope(&ifa_best->ia_addr.sin6_addr);
2264 		}
2265 	}
2266 
2267 	/* count statistics for future improvements */
2268 	if (ifa_best == NULL)
2269 		ip6stat.ip6s_sources_none++;
2270 	else {
2271 		if (oifp == ifa_best->ia_ifp)
2272 			ip6stat.ip6s_sources_sameif[best_scope]++;
2273 		else
2274 			ip6stat.ip6s_sources_otherif[best_scope]++;
2275 
2276 		if (best_scope == dst_scope)
2277 			ip6stat.ip6s_sources_samescope[best_scope]++;
2278 		else
2279 			ip6stat.ip6s_sources_otherscope[best_scope]++;
2280 
2281 		if (ifa_best->ia6_flags & IN6_IFF_DEPRECATED)
2282 			ip6stat.ip6s_sources_deprecated[best_scope]++;
2283 	}
2284 
2285 	return (ifa_best);
2286 }
2287 
2288 /*
2289  * return the best address out of the same scope. if no address was
2290  * found, return the first valid address from designated IF.
2291  */
2292 struct in6_ifaddr *
2293 in6_ifawithifp(struct ifnet *ifp, struct in6_addr *dst)
2294 {
2295 	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
2296 	struct ifaddr_container *ifac;
2297 	struct in6_ifaddr *besta = NULL;
2298 	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
2299 
2300 	dep[0] = dep[1] = NULL;
2301 
2302 	/*
2303 	 * We first look for addresses in the same scope.
2304 	 * If there is one, return it.
2305 	 * If two or more, return one which matches the dst longest.
2306 	 * If none, return one of global addresses assigned other ifs.
2307 	 */
2308 	TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
2309 		struct ifaddr *ifa = ifac->ifa;
2310 
2311 		if (ifa->ifa_addr->sa_family != AF_INET6)
2312 			continue;
2313 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
2314 			continue; /* XXX: is there any case to allow anycast? */
2315 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
2316 			continue; /* don't use this interface */
2317 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
2318 			continue;
2319 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2320 			if (ip6_use_deprecated)
2321 				dep[0] = (struct in6_ifaddr *)ifa;
2322 			continue;
2323 		}
2324 
2325 		if (dst_scope == in6_addrscope(IFA_IN6(ifa))) {
2326 			/*
2327 			 * call in6_matchlen() as few as possible
2328 			 */
2329 			if (besta) {
2330 				if (blen == -1)
2331 					blen = in6_matchlen(&besta->ia_addr.sin6_addr, dst);
2332 				tlen = in6_matchlen(IFA_IN6(ifa), dst);
2333 				if (tlen > blen) {
2334 					blen = tlen;
2335 					besta = (struct in6_ifaddr *)ifa;
2336 				}
2337 			} else
2338 				besta = (struct in6_ifaddr *)ifa;
2339 		}
2340 	}
2341 	if (besta)
2342 		return (besta);
2343 
2344 	TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
2345 		struct ifaddr *ifa = ifac->ifa;
2346 
2347 		if (ifa->ifa_addr->sa_family != AF_INET6)
2348 			continue;
2349 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
2350 			continue; /* XXX: is there any case to allow anycast? */
2351 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
2352 			continue; /* don't use this interface */
2353 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
2354 			continue;
2355 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2356 			if (ip6_use_deprecated)
2357 				dep[1] = (struct in6_ifaddr *)ifa;
2358 			continue;
2359 		}
2360 
2361 		return (struct in6_ifaddr *)ifa;
2362 	}
2363 
2364 	/* use the last-resort values, that are, deprecated addresses */
2365 	if (dep[0])
2366 		return dep[0];
2367 	if (dep[1])
2368 		return dep[1];
2369 
2370 	return NULL;
2371 }
2372 
2373 /*
2374  * perform DAD when interface becomes IFF_UP.
2375  */
2376 void
2377 in6_if_up(struct ifnet *ifp)
2378 {
2379 	struct ifaddr_container *ifac;
2380 	struct in6_ifaddr *ia;
2381 	int dad_delay;		/* delay ticks before DAD output */
2382 
2383 	/*
2384 	 * special cases, like 6to4, are handled in in6_ifattach
2385 	 */
2386 	in6_ifattach(ifp, NULL);
2387 
2388 	dad_delay = 0;
2389 	TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
2390 		struct ifaddr *ifa = ifac->ifa;
2391 
2392 		if (ifa->ifa_addr->sa_family != AF_INET6)
2393 			continue;
2394 		ia = (struct in6_ifaddr *)ifa;
2395 		if (ia->ia6_flags & IN6_IFF_TENTATIVE)
2396 			nd6_dad_start(ifa, &dad_delay);
2397 	}
2398 }
2399 
2400 int
2401 in6if_do_dad(struct ifnet *ifp)
2402 {
2403 	if (ifp->if_flags & IFF_LOOPBACK)
2404 		return (0);
2405 
2406 	switch (ifp->if_type) {
2407 #ifdef IFT_DUMMY
2408 	case IFT_DUMMY:
2409 #endif
2410 	case IFT_FAITH:
2411 		/*
2412 		 * These interfaces do not have the IFF_LOOPBACK flag,
2413 		 * but loop packets back.  We do not have to do DAD on such
2414 		 * interfaces.  We should even omit it, because loop-backed
2415 		 * NS would confuse the DAD procedure.
2416 		 */
2417 		return (0);
2418 	default:
2419 		/*
2420 		 * Our DAD routine requires the interface up and running.
2421 		 * However, some interfaces can be up before the RUNNING
2422 		 * status.  Additionaly, users may try to assign addresses
2423 		 * before the interface becomes up (or running).
2424 		 * We simply skip DAD in such a case as a work around.
2425 		 * XXX: we should rather mark "tentative" on such addresses,
2426 		 * and do DAD after the interface becomes ready.
2427 		 */
2428 		if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) !=
2429 		    (IFF_UP|IFF_RUNNING))
2430 			return (0);
2431 
2432 		return (1);
2433 	}
2434 }
2435 
2436 /*
2437  * Calculate max IPv6 MTU through all the interfaces and store it
2438  * to in6_maxmtu.
2439  */
2440 void
2441 in6_setmaxmtu(void)
2442 {
2443 	unsigned long maxmtu = 0;
2444 	struct ifnet *ifp;
2445 
2446 	for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list))
2447 	{
2448 		/* this function can be called during ifnet initialization */
2449 		if (ifp->if_afdata[AF_INET6] == NULL)
2450 			continue;
2451 		if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
2452 		    ND_IFINFO(ifp)->linkmtu > maxmtu)
2453 			maxmtu =  ND_IFINFO(ifp)->linkmtu;
2454 	}
2455 	if (maxmtu)	/* update only when maxmtu is positive */
2456 		in6_maxmtu = maxmtu;
2457 }
2458 
2459 void *
2460 in6_domifattach(struct ifnet *ifp)
2461 {
2462 	struct in6_ifextra *ext;
2463 
2464 	ext = (struct in6_ifextra *)kmalloc(sizeof(*ext), M_IFADDR, M_WAITOK);
2465 	bzero(ext, sizeof(*ext));
2466 
2467 	ext->in6_ifstat = (struct in6_ifstat *)kmalloc(sizeof(struct in6_ifstat),
2468 		M_IFADDR, M_WAITOK);
2469 	bzero(ext->in6_ifstat, sizeof(*ext->in6_ifstat));
2470 
2471 	ext->icmp6_ifstat =
2472 		(struct icmp6_ifstat *)kmalloc(sizeof(struct icmp6_ifstat),
2473 			M_IFADDR, M_WAITOK);
2474 	bzero(ext->icmp6_ifstat, sizeof(*ext->icmp6_ifstat));
2475 
2476 	ext->nd_ifinfo = nd6_ifattach(ifp);
2477 	ext->scope6_id = scope6_ifattach(ifp);
2478 	return ext;
2479 }
2480 
2481 void
2482 in6_domifdetach(struct ifnet *ifp, void *aux)
2483 {
2484 	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
2485 	scope6_ifdetach(ext->scope6_id);
2486 	nd6_ifdetach(ext->nd_ifinfo);
2487 	kfree(ext->in6_ifstat, M_IFADDR);
2488 	kfree(ext->icmp6_ifstat, M_IFADDR);
2489 	kfree(ext, M_IFADDR);
2490 }
2491 
2492 /*
2493  * Convert sockaddr_in6 to sockaddr_in.  Original sockaddr_in6 must be
2494  * v4 mapped addr or v4 compat addr
2495  */
2496 void
2497 in6_sin6_2_sin(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2498 {
2499 	bzero(sin, sizeof(*sin));
2500 	sin->sin_len = sizeof(struct sockaddr_in);
2501 	sin->sin_family = AF_INET;
2502 	sin->sin_port = sin6->sin6_port;
2503 	sin->sin_addr.s_addr = sin6->sin6_addr.s6_addr32[3];
2504 }
2505 
2506 /* Convert sockaddr_in to sockaddr_in6 in v4 mapped addr format. */
2507 void
2508 in6_sin_2_v4mapsin6(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2509 {
2510 	bzero(sin6, sizeof(*sin6));
2511 	sin6->sin6_len = sizeof(struct sockaddr_in6);
2512 	sin6->sin6_family = AF_INET6;
2513 	sin6->sin6_port = sin->sin_port;
2514 	sin6->sin6_addr.s6_addr32[0] = 0;
2515 	sin6->sin6_addr.s6_addr32[1] = 0;
2516 	sin6->sin6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
2517 	sin6->sin6_addr.s6_addr32[3] = sin->sin_addr.s_addr;
2518 }
2519 
2520 /* Convert sockaddr_in6 into sockaddr_in. */
2521 void
2522 in6_sin6_2_sin_in_sock(struct sockaddr *nam)
2523 {
2524 	struct sockaddr_in *sin_p;
2525 	struct sockaddr_in6 sin6;
2526 
2527 	/*
2528 	 * Save original sockaddr_in6 addr and convert it
2529 	 * to sockaddr_in.
2530 	 */
2531 	sin6 = *(struct sockaddr_in6 *)nam;
2532 	sin_p = (struct sockaddr_in *)nam;
2533 	in6_sin6_2_sin(sin_p, &sin6);
2534 }
2535 
2536 /* Convert sockaddr_in into sockaddr_in6 in v4 mapped addr format. */
2537 void
2538 in6_sin_2_v4mapsin6_in_sock(struct sockaddr **nam)
2539 {
2540 	struct sockaddr_in *sin_p;
2541 	struct sockaddr_in6 *sin6_p;
2542 
2543 	sin6_p = kmalloc(sizeof *sin6_p, M_SONAME, M_WAITOK);
2544 	sin_p = (struct sockaddr_in *)*nam;
2545 	in6_sin_2_v4mapsin6(sin_p, sin6_p);
2546 	kfree(*nam, M_SONAME);
2547 	*nam = (struct sockaddr *)sin6_p;
2548 }
2549