xref: /dragonfly/sys/netinet6/in6.c (revision 6700dd34)
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 #include <net/netmsg2.h>
89 #include <net/netisr2.h>
90 
91 #include <netinet/in.h>
92 #include <netinet/in_var.h>
93 #include <netinet/if_ether.h>
94 #include <netinet/in_systm.h>
95 #include <netinet/ip.h>
96 #include <netinet/in_pcb.h>
97 
98 #include <netinet/ip6.h>
99 #include <netinet6/ip6_var.h>
100 #include <netinet6/nd6.h>
101 #include <netinet6/mld6_var.h>
102 #include <netinet6/ip6_mroute.h>
103 #include <netinet6/in6_ifattach.h>
104 #include <netinet6/scope6_var.h>
105 #include <netinet6/in6_pcb.h>
106 #include <netinet6/in6_var.h>
107 
108 #include <net/net_osdep.h>
109 
110 /*
111  * Definitions of some costant IP6 addresses.
112  */
113 const struct in6_addr kin6addr_any = IN6ADDR_ANY_INIT;
114 const struct in6_addr kin6addr_loopback = IN6ADDR_LOOPBACK_INIT;
115 const struct in6_addr kin6addr_nodelocal_allnodes =
116 	IN6ADDR_NODELOCAL_ALLNODES_INIT;
117 const struct in6_addr kin6addr_linklocal_allnodes =
118 	IN6ADDR_LINKLOCAL_ALLNODES_INIT;
119 const struct in6_addr kin6addr_linklocal_allrouters =
120 	IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
121 
122 const struct in6_addr in6mask0 = IN6MASK0;
123 const struct in6_addr in6mask32 = IN6MASK32;
124 const struct in6_addr in6mask64 = IN6MASK64;
125 const struct in6_addr in6mask96 = IN6MASK96;
126 const struct in6_addr in6mask128 = IN6MASK128;
127 
128 const struct sockaddr_in6 sa6_any = {sizeof(sa6_any), AF_INET6,
129 				     0, 0, IN6ADDR_ANY_INIT, 0};
130 
131 static int in6_lifaddr_ioctl (u_long, caddr_t, struct ifnet *,
132 	     struct thread *);
133 static int in6_ifinit (struct ifnet *, struct in6_ifaddr *,
134 			   struct sockaddr_in6 *, int);
135 static void in6_unlink_ifa (struct in6_ifaddr *, struct ifnet *);
136 static void in6_ifloop_request_callback(int, int, struct rt_addrinfo *, struct rtentry *, void *);
137 
138 static void	in6_control_internal_dispatch(netmsg_t);
139 static int	in6_control_internal(u_long, caddr_t, struct ifnet *,
140 		    struct thread *);
141 
142 struct in6_multihead in6_multihead;	/* XXX BSS initialization */
143 
144 int	(*faithprefix_p)(struct in6_addr *);
145 
146 /*
147  * Subroutine for in6_ifaddloop() and in6_ifremloop().
148  * This routine does actual work.
149  */
150 static void
151 in6_ifloop_request(int cmd, struct ifaddr *ifa)
152 {
153 	struct sockaddr_in6 all1_sa;
154         struct rt_addrinfo rtinfo;
155 	int error;
156 
157 	bzero(&all1_sa, sizeof(all1_sa));
158 	all1_sa.sin6_family = AF_INET6;
159 	all1_sa.sin6_len = sizeof(struct sockaddr_in6);
160 	all1_sa.sin6_addr = in6mask128;
161 
162 	/*
163 	 * We specify the address itself as the gateway, and set the
164 	 * RTF_LLINFO flag, so that the corresponding host route would have
165 	 * the flag, and thus applications that assume traditional behavior
166 	 * would be happy.  Note that we assume the caller of the function
167 	 * (probably implicitly) set nd6_rtrequest() to ifa->ifa_rtrequest,
168 	 * which changes the outgoing interface to the loopback interface.
169 	 */
170 	bzero(&rtinfo, sizeof(struct rt_addrinfo));
171 	rtinfo.rti_info[RTAX_DST] = ifa->ifa_addr;
172 	rtinfo.rti_info[RTAX_GATEWAY] = ifa->ifa_addr;
173 	rtinfo.rti_info[RTAX_NETMASK] = (struct sockaddr *)&all1_sa;
174 	rtinfo.rti_flags = RTF_UP|RTF_HOST|RTF_LLINFO;
175 
176 	error = rtrequest1_global(cmd, &rtinfo,
177 	    in6_ifloop_request_callback, ifa, RTREQ_PRIO_NORM);
178 	if (error != 0) {
179 		log(LOG_ERR, "in6_ifloop_request: "
180 		    "%s operation failed for %s (errno=%d)\n",
181 		    cmd == RTM_ADD ? "ADD" : "DELETE",
182 		    ip6_sprintf(&((struct in6_ifaddr *)ifa)->ia_addr.sin6_addr),
183 		    error);
184 	}
185 }
186 
187 static void
188 in6_ifloop_request_callback(int cmd, int error, struct rt_addrinfo *rtinfo,
189 			    struct rtentry *rt, void *arg)
190 {
191 	struct ifaddr *ifa = arg;
192 
193 	if (error)
194 		goto done;
195 
196 	/*
197 	 * Make sure rt_ifa be equal to IFA, the second argument of the
198 	 * function.
199 	 * We need this because when we refer to rt_ifa->ia6_flags in
200 	 * ip6_input, we assume that the rt_ifa points to the address instead
201 	 * of the loopback address.
202 	 */
203 	if (cmd == RTM_ADD && rt && ifa != rt->rt_ifa) {
204 		++rt->rt_refcnt;
205 		IFAFREE(rt->rt_ifa);
206 		IFAREF(ifa);
207 		rt->rt_ifa = ifa;
208 		--rt->rt_refcnt;
209 	}
210 
211 	/*
212 	 * Report the addition/removal of the address to the routing socket.
213 	 * XXX: since we called rtinit for a p2p interface with a destination,
214 	 *      we end up reporting twice in such a case.  Should we rather
215 	 *      omit the second report?
216 	 */
217 	if (rt) {
218 		if (mycpuid == 0)
219 			rt_newaddrmsg(cmd, ifa, error, rt);
220 		if (cmd == RTM_DELETE) {
221 			if (rt->rt_refcnt == 0) {
222 				++rt->rt_refcnt;
223 				rtfree(rt);
224 			}
225 		}
226 	}
227 done:
228 	/* no way to return any new error */
229 	;
230 }
231 
232 /*
233  * Add ownaddr as loopback rtentry.  We previously add the route only if
234  * necessary (ex. on a p2p link).  However, since we now manage addresses
235  * separately from prefixes, we should always add the route.  We can't
236  * rely on the cloning mechanism from the corresponding interface route
237  * any more.
238  */
239 void
240 in6_ifaddloop(struct ifaddr *ifa)
241 {
242 	struct rtentry *rt;
243 
244 	/* If there is no loopback entry, allocate one. */
245 	rt = rtpurelookup(ifa->ifa_addr);
246 	if (rt == NULL || !(rt->rt_flags & RTF_HOST) ||
247 	    !(rt->rt_ifp->if_flags & IFF_LOOPBACK))
248 		in6_ifloop_request(RTM_ADD, ifa);
249 	if (rt != NULL)
250 		rt->rt_refcnt--;
251 }
252 
253 /*
254  * Remove loopback rtentry of ownaddr generated by in6_ifaddloop(),
255  * if it exists.
256  */
257 void
258 in6_ifremloop(struct ifaddr *ifa)
259 {
260 	struct in6_ifaddr *ia;
261 	struct rtentry *rt;
262 	int ia_count = 0;
263 
264 	/*
265 	 * Some of BSD variants do not remove cloned routes
266 	 * from an interface direct route, when removing the direct route
267 	 * (see comments in net/net_osdep.h).  Even for variants that do remove
268 	 * cloned routes, they could fail to remove the cloned routes when
269 	 * we handle multple addresses that share a common prefix.
270 	 * So, we should remove the route corresponding to the deleted address
271 	 * regardless of the result of in6_is_ifloop_auto().
272 	 */
273 
274 	/*
275 	 * Delete the entry only if exact one ifa exists.  More than one ifa
276 	 * can exist if we assign a same single address to multiple
277 	 * (probably p2p) interfaces.
278 	 * XXX: we should avoid such a configuration in IPv6...
279 	 */
280 	for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
281 		if (IN6_ARE_ADDR_EQUAL(IFA_IN6(ifa), &ia->ia_addr.sin6_addr)) {
282 			ia_count++;
283 			if (ia_count > 1)
284 				break;
285 		}
286 	}
287 
288 	if (ia_count == 1) {
289 		/*
290 		 * Before deleting, check if a corresponding loopbacked host
291 		 * route surely exists.  With this check, we can avoid to
292 		 * delete an interface direct route whose destination is same
293 		 * as the address being removed.  This can happen when remofing
294 		 * a subnet-router anycast address on an interface attahced
295 		 * to a shared medium.
296 		 */
297 		rt = rtpurelookup(ifa->ifa_addr);
298 		if (rt != NULL && (rt->rt_flags & RTF_HOST) &&
299 		    (rt->rt_ifp->if_flags & IFF_LOOPBACK)) {
300 			rt->rt_refcnt--;
301 			in6_ifloop_request(RTM_DELETE, ifa);
302 		}
303 	}
304 }
305 
306 int
307 in6_mask2len(const struct in6_addr *mask, const u_char *lim0)
308 {
309 	int x = 0, y;
310 	const u_char *lim = lim0, *p;
311 
312 	if (lim0 == NULL ||
313 	    lim0 - (const u_char *)mask > sizeof(*mask)) {
314 		/* Ignore the scope_id part */
315 		lim = (const u_char *)mask + sizeof(*mask);
316 	}
317 	for (p = (const u_char *)mask; p < lim; x++, p++) {
318 		if (*p != 0xff)
319 			break;
320 	}
321 	y = 0;
322 	if (p < lim) {
323 		for (y = 0; y < 8; y++) {
324 			if ((*p & (0x80 >> y)) == 0)
325 				break;
326 		}
327 	}
328 
329 	/*
330 	 * When the limit pointer is given, do a stricter check on the
331 	 * remaining bits.
332 	 */
333 	if (p < lim) {
334 		if (y != 0 && (*p & (0x00ff >> y)) != 0)
335 			return (-1);
336 		for (p = p + 1; p < lim; p++)
337 			if (*p != 0)
338 				return (-1);
339 	}
340 
341 	return x * 8 + y;
342 }
343 
344 #define ifa2ia6(ifa)	((struct in6_ifaddr *)(ifa))
345 #define ia62ifa(ia6)	(&((ia6)->ia_ifa))
346 
347 void
348 in6_control_dispatch(netmsg_t msg)
349 {
350 	int error;
351 
352 	error = in6_control(msg->control.nm_cmd,
353 			    msg->control.nm_data,
354 			    msg->control.nm_ifp,
355 			    msg->control.nm_td);
356 	lwkt_replymsg(&msg->control.base.lmsg, error);
357 }
358 
359 int
360 in6_control(u_long cmd, caddr_t data, struct ifnet *ifp, struct thread *td)
361 {
362 	struct netmsg_pru_control msg;
363 
364 	switch (cmd) {
365 	case SIOCSIFPREFIX_IN6:
366 	case SIOCDIFPREFIX_IN6:
367 	case SIOCAIFPREFIX_IN6:
368 	case SIOCCIFPREFIX_IN6:
369 	case SIOCSGIFPREFIX_IN6:
370 	case SIOCGIFPREFIX_IN6:
371 		log(LOG_NOTICE, "prefix ioctls are now invalidated. "
372 		    "please use ifconfig.\n");
373 		return (EOPNOTSUPP);
374 
375 	case SIOCSIFADDR_IN6:
376 	case SIOCSIFDSTADDR_IN6:
377 	case SIOCSIFNETMASK_IN6:
378 		/*
379 		 * Since IPv6 allows a node to assign multiple addresses
380 		 * on a single interface, SIOCSIFxxx ioctls are not suitable
381 		 * and should be unused.
382 		 */
383 		/* We decided to obsolete this command (20000704) */
384 		return (EINVAL);
385 
386 	case SIOCSIFADDR:
387 	case SIOCSIFDSTADDR:
388 	case SIOCSIFBRDADDR:
389 	case SIOCSIFNETMASK:
390 		/*
391 		 * Do not pass those ioctl to driver handler since they are not
392 		 * properly setup.  Instead just error out.
393 		 */
394 		return (EOPNOTSUPP);
395 
396 	/* mroute */
397 	case SIOCGETSGCNT_IN6:
398 	case SIOCGETMIFCNT_IN6:
399 	/* srcsel policy */
400 	case SIOCAADDRCTL_POLICY:
401 	case SIOCDADDRCTL_POLICY:
402 	/* nd6 */
403 	case SIOCSNDFLUSH_IN6:
404 	case SIOCSPFXFLUSH_IN6:
405 	case SIOCSRTRFLUSH_IN6:
406 	case SIOCSDEFIFACE_IN6:
407 	case SIOCSIFINFO_FLAGS:
408 	case OSIOCGIFINFO_IN6:
409 	case SIOCGIFINFO_IN6:
410 	case SIOCGDRLST_IN6:
411 	case SIOCGPRLST_IN6:
412 	case SIOCGNBRINFO_IN6:
413 	case SIOCGDEFIFACE_IN6:
414 	/* scope6 */
415 	case SIOCSSCOPE6:
416 	case SIOCGSCOPE6:
417 	case SIOCGSCOPE6DEF:
418 	/* change address */
419 	case SIOCALIFADDR:
420 	case SIOCDLIFADDR:
421 	case SIOCSIFALIFETIME_IN6:
422 	case SIOCAIFADDR_IN6:
423 	case SIOCDIFADDR_IN6:
424 		/*
425 		 * Dispatch these SIOCs to netisr0.
426 		 */
427 		netmsg_init(&msg.base, NULL, &curthread->td_msgport, 0,
428 		    in6_control_internal_dispatch);
429 		msg.nm_cmd = cmd;
430 		msg.nm_data = data;
431 		msg.nm_ifp = ifp;
432 		msg.nm_td = td;
433 		lwkt_domsg(netisr_cpuport(0), &msg.base.lmsg, 0);
434 		return msg.base.lmsg.ms_error;
435 
436 	default:
437 		return in6_control_internal(cmd, data, ifp, td);
438 	}
439 }
440 
441 static void
442 in6_control_internal_dispatch(netmsg_t msg)
443 {
444 	int error;
445 
446 	error = in6_control_internal(msg->control.nm_cmd, msg->control.nm_data,
447 	    msg->control.nm_ifp, msg->control.nm_td);
448 	lwkt_replymsg(&msg->lmsg, error);
449 }
450 
451 static int
452 in6_control_internal(u_long cmd, caddr_t data, struct ifnet *ifp,
453     struct thread *td)
454 {
455 	struct in6_ifreq *ifr = (struct in6_ifreq *)data;
456 	struct in6_ifaddr *ia = NULL;
457 	struct in6_aliasreq *ifra = (struct in6_aliasreq *)data;
458 	struct in6_ifextra *xtra;
459 	boolean_t privileged;
460 	int error;
461 
462 	privileged = FALSE;
463 	if (priv_check(td, PRIV_ROOT) == 0)
464 		privileged = TRUE;
465 
466 	switch (cmd) {
467 	case SIOCALIFADDR:
468 	case SIOCDLIFADDR:
469 		if (!privileged)
470 			return (EPERM);
471 		/* FALLTHROUGH */
472 	case SIOCGLIFADDR:
473 		if (ifp == NULL)
474 			return (EOPNOTSUPP);
475 		return in6_lifaddr_ioctl(cmd, data, ifp, td);
476 	}
477 
478 	switch (cmd) {
479 	case SIOCGETSGCNT_IN6:
480 	case SIOCGETMIFCNT_IN6:
481 		return (mrt6_ioctl(cmd, data));
482 	}
483 
484 	switch(cmd) {
485 	case SIOCAADDRCTL_POLICY:
486 	case SIOCDADDRCTL_POLICY:
487 		if (!privileged)
488 			return (EPERM);
489 		return (in6_src_ioctl(cmd, data));
490 	}
491 
492 	if (ifp == NULL)
493 		return (EOPNOTSUPP);
494 
495 	switch (cmd) {
496 	case SIOCSNDFLUSH_IN6:
497 	case SIOCSPFXFLUSH_IN6:
498 	case SIOCSRTRFLUSH_IN6:
499 	case SIOCSDEFIFACE_IN6:
500 	case SIOCSIFINFO_FLAGS:
501 		if (!privileged)
502 			return (EPERM);
503 		/* FALLTHROUGH */
504 	case OSIOCGIFINFO_IN6:
505 	case SIOCGIFINFO_IN6:
506 	case SIOCGDRLST_IN6:
507 	case SIOCGPRLST_IN6:
508 	case SIOCGNBRINFO_IN6:
509 	case SIOCGDEFIFACE_IN6:
510 		return (nd6_ioctl(cmd, data, ifp));
511 	}
512 
513 	switch (cmd) {
514 	case SIOCSSCOPE6:
515 		if (!privileged)
516 			return (EPERM);
517 		return (scope6_set(ifp,
518 			(struct scope6_id *)ifr->ifr_ifru.ifru_scope_id));
519 
520 	case SIOCGSCOPE6:
521 		return (scope6_get(ifp,
522 			(struct scope6_id *)ifr->ifr_ifru.ifru_scope_id));
523 
524 	case SIOCGSCOPE6DEF:
525 		return (scope6_get_default((struct scope6_id *)
526 			ifr->ifr_ifru.ifru_scope_id));
527 	}
528 
529 	/*
530 	 * Find address for this interface, if it exists.
531 	 */
532 	if (ifra->ifra_addr.sin6_family == AF_INET6) { /* XXX */
533 		struct sockaddr_in6 *sa6 =
534 		    (struct sockaddr_in6 *)&ifra->ifra_addr;
535 
536 		if (IN6_IS_ADDR_LINKLOCAL(&sa6->sin6_addr)) {
537 			if (sa6->sin6_addr.s6_addr16[1] == 0) {
538 				/* Link ID is not embedded by the user */
539 				sa6->sin6_addr.s6_addr16[1] =
540 				    htons(ifp->if_index);
541 			} else if (sa6->sin6_addr.s6_addr16[1] !=
542 			    htons(ifp->if_index)) {
543 				/* Link ID contradicts */
544 				return (EINVAL);
545 			}
546 			if (sa6->sin6_scope_id) {
547 				if (sa6->sin6_scope_id !=
548 				    (u_int32_t)ifp->if_index)
549 					return (EINVAL);
550 				sa6->sin6_scope_id = 0; /* XXX: good way? */
551 			}
552 		}
553 		ia = in6ifa_ifpwithaddr(ifp, &ifra->ifra_addr.sin6_addr);
554 	}
555 
556 	switch (cmd) {
557 	case SIOCDIFADDR_IN6:
558 		/*
559 		 * For IPv4, we look for existing in_ifaddr here to allow
560 		 * "ifconfig if0 delete" to remove first IPv4 address on the
561 		 * interface.  For IPv6, as the spec allow multiple interface
562 		 * address from the day one, we consider "remove the first one"
563 		 * semantics to be not preferable.
564 		 */
565 		if (ia == NULL)
566 			return (EADDRNOTAVAIL);
567 		/* FALLTHROUGH */
568 	case SIOCAIFADDR_IN6:
569 		/*
570 		 * We always require users to specify a valid IPv6 address for
571 		 * the corresponding operation.
572 		 */
573 		if (ifra->ifra_addr.sin6_family != AF_INET6 ||
574 		    ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6))
575 			return (EAFNOSUPPORT);
576 		if (!privileged)
577 			return (EPERM);
578 		break;
579 
580 	case SIOCGIFADDR_IN6:
581 		/* This interface is basically deprecated.  Use SIOCGIFCONF. */
582 		/* FALLTHROUGH */
583 	case SIOCGIFAFLAG_IN6:
584 	case SIOCGIFNETMASK_IN6:
585 	case SIOCGIFDSTADDR_IN6:
586 	case SIOCGIFALIFETIME_IN6:
587 		/* Must think again about its semantics */
588 		if (ia == NULL)
589 			return (EADDRNOTAVAIL);
590 		break;
591 
592 	case SIOCSIFALIFETIME_IN6:
593 	    {
594 		const struct in6_addrlifetime *lt;
595 
596 		if (!privileged)
597 			return (EPERM);
598 		if (ia == NULL)
599 			return (EADDRNOTAVAIL);
600 		/* Sanity for overflow - beware unsigned */
601 		lt = &ifr->ifr_ifru.ifru_lifetime;
602 		if (lt->ia6t_vltime != ND6_INFINITE_LIFETIME &&
603 		    lt->ia6t_vltime + time_uptime < time_uptime)
604 			return EINVAL;
605 		if (lt->ia6t_pltime != ND6_INFINITE_LIFETIME &&
606 		    lt->ia6t_pltime + time_uptime < time_uptime)
607 			return EINVAL;
608 		break;
609 	    }
610 	}
611 
612 	switch (cmd) {
613 	case SIOCGIFADDR_IN6:
614 		ifr->ifr_addr = ia->ia_addr;
615 		break;
616 
617 	case SIOCGIFDSTADDR_IN6:
618 		if (!(ifp->if_flags & IFF_POINTOPOINT))
619 			return (EINVAL);
620 		/*
621 		 * XXX: Should we check if ifa_dstaddr is NULL and return
622 		 * an error?
623 		 */
624 		ifr->ifr_dstaddr = ia->ia_dstaddr;
625 		break;
626 
627 	case SIOCGIFNETMASK_IN6:
628 		ifr->ifr_addr = ia->ia_prefixmask;
629 		break;
630 
631 	case SIOCGIFAFLAG_IN6:
632 		ifr->ifr_ifru.ifru_flags6 = ia->ia6_flags;
633 		break;
634 
635 	case SIOCGIFSTAT_IN6:
636 		if ((xtra = ifp->if_afdata[AF_INET6]) == NULL)
637 			return EINVAL;
638 		bzero(&ifr->ifr_ifru.ifru_stat,
639 		    sizeof(ifr->ifr_ifru.ifru_stat));
640 		ifr->ifr_ifru.ifru_stat = *xtra->in6_ifstat;
641 		break;
642 
643 	case SIOCGIFSTAT_ICMP6:
644 		if ((xtra = ifp->if_afdata[AF_INET6]) == NULL)
645 			return EINVAL;
646 		bzero(&ifr->ifr_ifru.ifru_stat,
647 		    sizeof(ifr->ifr_ifru.ifru_icmp6stat));
648 		ifr->ifr_ifru.ifru_icmp6stat = *xtra->icmp6_ifstat;
649 		break;
650 
651 	case SIOCGIFALIFETIME_IN6:
652 		ifr->ifr_ifru.ifru_lifetime = ia->ia6_lifetime;
653 		break;
654 
655 	case SIOCSIFALIFETIME_IN6:
656 		ia->ia6_lifetime = ifr->ifr_ifru.ifru_lifetime;
657 		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
658 			ia->ia6_lifetime.ia6t_expire =
659 			    time_uptime + ia->ia6_lifetime.ia6t_vltime;
660 		} else {
661 			ia->ia6_lifetime.ia6t_expire = 0;
662 		}
663 		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
664 			ia->ia6_lifetime.ia6t_preferred =
665 			    time_uptime + ia->ia6_lifetime.ia6t_pltime;
666 		} else {
667 			ia->ia6_lifetime.ia6t_preferred = 0;
668 		}
669 		break;
670 
671 	case SIOCAIFADDR_IN6:
672 	{
673 		int i, error = 0, iaIsNew;
674 		struct nd_prefix pr0, *pr;
675 
676 		if (ia != NULL)
677 			iaIsNew = 0;
678 		else
679 			iaIsNew = 1;
680 
681 		/*
682 		 * First, make or update the interface address structure,
683 		 * and link it to the list.
684 		 */
685 		if ((error = in6_update_ifa(ifp, ifra, ia)) != 0)
686 			return (error);
687 
688 		/*
689 		 * Then, make the prefix on-link on the interface.
690 		 * XXX: We'd rather create the prefix before the address, but
691 		 * we need at least one address to install the corresponding
692 		 * interface route, so we configure the address first.
693 		 */
694 
695 		/*
696 		 * Convert mask to prefix length (prefixmask has already
697 		 * been validated in in6_update_ifa().
698 		 */
699 		bzero(&pr0, sizeof(pr0));
700 		pr0.ndpr_ifp = ifp;
701 		pr0.ndpr_plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
702 		    NULL);
703 		if (pr0.ndpr_plen == 128)
704 			break;	/* no need to install a host route. */
705 		pr0.ndpr_prefix = ifra->ifra_addr;
706 		pr0.ndpr_mask = ifra->ifra_prefixmask.sin6_addr;
707 		/* Apply the mask for safety. */
708 		for (i = 0; i < 4; i++) {
709 			pr0.ndpr_prefix.sin6_addr.s6_addr32[i] &=
710 			    ifra->ifra_prefixmask.sin6_addr.s6_addr32[i];
711 		}
712 		/*
713 		 * XXX: Since we don't have an API to set prefix (not address)
714 		 * lifetimes, we just use the same lifetimes as addresses.
715 		 * The (temporarily) installed lifetimes can be overridden by
716 		 * later advertised RAs (when accept_rtadv is non 0), which is
717 		 * an intended behavior.
718 		 */
719 		pr0.ndpr_raf_onlink = 1; /* should be configurable? */
720 		pr0.ndpr_raf_auto =
721 		    ((ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0);
722 		pr0.ndpr_vltime = ifra->ifra_lifetime.ia6t_vltime;
723 		pr0.ndpr_pltime = ifra->ifra_lifetime.ia6t_pltime;
724 
725 		/* Add the prefix if there's one. */
726 		if ((pr = nd6_prefix_lookup(&pr0)) == NULL) {
727 			/*
728 			 * nd6_prelist_add will install the corresponding
729 			 * interface route.
730 			 */
731 			if ((error = nd6_prelist_add(&pr0, NULL, &pr)) != 0)
732 				return (error);
733 			if (pr == NULL) {
734 				log(LOG_ERR, "nd6_prelist_add succeeded but "
735 				    "no prefix\n");
736 				return (EINVAL); /* XXX panic here? */
737 			}
738 		}
739 
740 		ia = in6ifa_ifpwithaddr(ifp, &ifra->ifra_addr.sin6_addr);
741 		if (ia == NULL) {
742 		    	/* XXX: This should not happen! */
743 			log(LOG_ERR, "in6_control: addition succeeded, but"
744 			    " no ifaddr\n");
745 		} else {
746 			if ((ia->ia6_flags & IN6_IFF_AUTOCONF) &&
747 			    ia->ia6_ndpr == NULL) {
748 				/*
749 				 * New autoconf address
750 				 */
751 				ia->ia6_ndpr = pr;
752 				pr->ndpr_refcnt++;
753 
754 				/*
755 				 * If this is the first autoconf address from
756 				 * the prefix, create a temporary address
757 				 * as well (when specified).
758 				 */
759 				if (ip6_use_tempaddr && pr->ndpr_refcnt == 1) {
760 					int e;
761 
762 					if ((e = in6_tmpifadd(ia, 1)) != 0) {
763 						log(LOG_NOTICE, "in6_control: "
764 						    "failed to create a "
765 						    "temporary address, "
766 						    "errno=%d\n", e);
767 					}
768 				}
769 			}
770 
771 			/*
772 			 * This might affect the status of autoconfigured
773 			 * addresses, that is, this address might make
774 			 * other addresses detached.
775 			 */
776 			pfxlist_onlink_check();
777 		}
778 		if (error == 0 && ia) {
779 			EVENTHANDLER_INVOKE(ifaddr_event, ifp,
780 			    iaIsNew ? IFADDR_EVENT_ADD : IFADDR_EVENT_CHANGE,
781 			    &ia->ia_ifa);
782 		}
783 		break;
784 	}
785 
786 	case SIOCDIFADDR_IN6:
787 	{
788 		int i = 0;
789 		struct nd_prefix pr0, *pr;
790 
791 		/*
792 		 * If the address being deleted is the only one that owns
793 		 * the corresponding prefix, expire the prefix as well.
794 		 * XXX: Theoretically, we don't have to warry about such
795 		 * relationship, since we separate the address management
796 		 * and the prefix management.  We do this, however, to provide
797 		 * as much backward compatibility as possible in terms of
798 		 * the ioctl operation.
799 		 */
800 		bzero(&pr0, sizeof(pr0));
801 		pr0.ndpr_ifp = ifp;
802 		pr0.ndpr_plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr,
803 		    NULL);
804 		if (pr0.ndpr_plen == 128)
805 			goto purgeaddr;
806 		pr0.ndpr_prefix = ia->ia_addr;
807 		pr0.ndpr_mask = ia->ia_prefixmask.sin6_addr;
808 		for (i = 0; i < 4; i++) {
809 			pr0.ndpr_prefix.sin6_addr.s6_addr32[i] &=
810 			    ia->ia_prefixmask.sin6_addr.s6_addr32[i];
811 		}
812 		/*
813 		 * The logic of the following condition is a bit complicated.
814 		 * We expire the prefix when
815 		 * 1. The address obeys autoconfiguration and it is the
816 		 *    only owner of the associated prefix, or
817 		 * 2. The address does not obey autoconf and there is no
818 		 *    other owner of the prefix.
819 		 */
820 		if ((pr = nd6_prefix_lookup(&pr0)) != NULL &&
821 		    (((ia->ia6_flags & IN6_IFF_AUTOCONF) &&
822 		      pr->ndpr_refcnt == 1) ||
823 		     (!(ia->ia6_flags & IN6_IFF_AUTOCONF) &&
824 		      pr->ndpr_refcnt == 0)))
825 			pr->ndpr_expire = 1; /* XXX: just for expiration */
826 
827 purgeaddr:
828 		EVENTHANDLER_INVOKE(ifaddr_event, ifp, IFADDR_EVENT_DELETE,
829 		    &ia->ia_ifa);
830 		in6_purgeaddr(&ia->ia_ifa);
831 		break;
832 	}
833 
834 	default:
835 		if (ifp->if_ioctl == NULL)
836 			return (EOPNOTSUPP);
837 		ifnet_serialize_all(ifp);
838 		error = ifp->if_ioctl(ifp, cmd, data, td->td_proc->p_ucred);
839 		ifnet_deserialize_all(ifp);
840 		return (error);
841 	}
842 
843 	return (0);
844 }
845 
846 /*
847  * Update parameters of an IPv6 interface address.
848  * If necessary, a new entry is created and linked into address chains.
849  * This function is separated from in6_control().
850  * XXX: should this be performed under splnet()?
851  */
852 int
853 in6_update_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
854 	       struct in6_ifaddr *ia)
855 {
856 	int error = 0, hostIsNew = 0, plen = -1;
857 	struct in6_ifaddr *oia;
858 	struct sockaddr_in6 dst6;
859 	struct in6_addrlifetime *lt;
860 
861 	/* Validate parameters */
862 	if (ifp == NULL || ifra == NULL) /* this maybe redundant */
863 		return (EINVAL);
864 
865 	/*
866 	 * The destination address for a p2p link must have a family
867 	 * of AF_UNSPEC or AF_INET6.
868 	 */
869 	if ((ifp->if_flags & IFF_POINTOPOINT) &&
870 	    ifra->ifra_dstaddr.sin6_family != AF_INET6 &&
871 	    ifra->ifra_dstaddr.sin6_family != AF_UNSPEC)
872 		return (EAFNOSUPPORT);
873 	/*
874 	 * validate ifra_prefixmask.  don't check sin6_family, netmask
875 	 * does not carry fields other than sin6_len.
876 	 */
877 	if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6))
878 		return (EINVAL);
879 	/*
880 	 * Because the IPv6 address architecture is classless, we require
881 	 * users to specify a (non 0) prefix length (mask) for a new address.
882 	 * We also require the prefix (when specified) mask is valid, and thus
883 	 * reject a non-consecutive mask.
884 	 */
885 	if (ia == NULL && ifra->ifra_prefixmask.sin6_len == 0)
886 		return (EINVAL);
887 	if (ifra->ifra_prefixmask.sin6_len != 0) {
888 		plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
889 				    (u_char *)&ifra->ifra_prefixmask +
890 				    ifra->ifra_prefixmask.sin6_len);
891 		if (plen <= 0)
892 			return (EINVAL);
893 	}
894 	else {
895 		/*
896 		 * In this case, ia must not be NULL.  We just use its prefix
897 		 * length.
898 		 */
899 		plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
900 	}
901 	/*
902 	 * If the destination address on a p2p interface is specified,
903 	 * and the address is a scoped one, validate/set the scope
904 	 * zone identifier.
905 	 */
906 	dst6 = ifra->ifra_dstaddr;
907 	if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) &&
908 	    (dst6.sin6_family == AF_INET6)) {
909 		int scopeid;
910 
911 		if ((error = in6_recoverscope(&dst6,
912 					      &ifra->ifra_dstaddr.sin6_addr,
913 					      ifp)) != 0)
914 			return (error);
915 		if (in6_addr2zoneid(ifp, &dst6.sin6_addr, &scopeid))
916 			return (EINVAL);
917 		if (dst6.sin6_scope_id == 0) /* user omit to specify the ID. */
918 			dst6.sin6_scope_id = scopeid;
919 		else if (dst6.sin6_scope_id != scopeid)
920 			return (EINVAL); /* scope ID mismatch. */
921 		if ((error = in6_embedscope(&dst6.sin6_addr, &dst6, NULL, NULL))
922 		    != 0)
923 			return (error);
924 		dst6.sin6_scope_id = 0; /* XXX */
925 	}
926 	/*
927 	 * The destination address can be specified only for a p2p or a
928 	 * loopback interface.  If specified, the corresponding prefix length
929 	 * must be 128.
930 	 */
931 	if (ifra->ifra_dstaddr.sin6_family == AF_INET6) {
932 		if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) == 0) {
933 			/* XXX: noisy message */
934 			log(LOG_INFO, "in6_update_ifa: a destination can be "
935 			    "specified for a p2p or a loopback IF only\n");
936 			return (EINVAL);
937 		}
938 		if (plen != 128) {
939 			/*
940 			 * The following message seems noisy, but we dare to
941 			 * add it for diagnosis.
942 			 */
943 			log(LOG_INFO, "in6_update_ifa: prefixlen must be 128 "
944 			    "when dstaddr is specified\n");
945 			return (EINVAL);
946 		}
947 	}
948 	/* lifetime consistency check */
949 	lt = &ifra->ifra_lifetime;
950 	if (lt->ia6t_vltime != ND6_INFINITE_LIFETIME
951 	    && lt->ia6t_vltime + time_uptime < time_uptime) {
952 		return EINVAL;
953 	}
954 	if (lt->ia6t_vltime == 0) {
955 		/*
956 		 * the following log might be noisy, but this is a typical
957 		 * configuration mistake or a tool's bug.
958 		 */
959 		log(LOG_INFO,
960 		    "in6_update_ifa: valid lifetime is 0 for %s\n",
961 		    ip6_sprintf(&ifra->ifra_addr.sin6_addr));
962 	}
963 	if (lt->ia6t_pltime != ND6_INFINITE_LIFETIME
964 	    && lt->ia6t_pltime + time_uptime < time_uptime) {
965 		return EINVAL;
966 	}
967 
968 	/*
969 	 * If this is a new address, allocate a new ifaddr and link it
970 	 * into chains.
971 	 */
972 	if (ia == NULL) {
973 		hostIsNew = 1;
974 		ia = ifa_create(sizeof(*ia));
975 
976 		/* Initialize the address and masks */
977 		ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
978 		ia->ia_addr.sin6_family = AF_INET6;
979 		ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
980 		if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
981 			/*
982 			 * XXX: some functions expect that ifa_dstaddr is not
983 			 * NULL for p2p interfaces.
984 			 */
985 			ia->ia_ifa.ifa_dstaddr
986 				= (struct sockaddr *)&ia->ia_dstaddr;
987 		} else {
988 			ia->ia_ifa.ifa_dstaddr = NULL;
989 		}
990 		ia->ia_ifa.ifa_netmask
991 			= (struct sockaddr *)&ia->ia_prefixmask;
992 
993 		ia->ia_ifp = ifp;
994 		if ((oia = in6_ifaddr) != NULL) {
995 			for ( ; oia->ia_next; oia = oia->ia_next)
996 				continue;
997 			oia->ia_next = ia;
998 		} else
999 			in6_ifaddr = ia;
1000 
1001 		ifa_iflink(&ia->ia_ifa, ifp, 1);
1002 	}
1003 
1004 	/* set prefix mask */
1005 	if (ifra->ifra_prefixmask.sin6_len) {
1006 		/*
1007 		 * We prohibit changing the prefix length of an existing
1008 		 * address, because
1009 		 * + such an operation should be rare in IPv6, and
1010 		 * + the operation would confuse prefix management.
1011 		 */
1012 		if (ia->ia_prefixmask.sin6_len &&
1013 		    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) != plen) {
1014 			log(LOG_INFO, "in6_update_ifa: the prefix length of an"
1015 			    " existing (%s) address should not be changed\n",
1016 			    ip6_sprintf(&ia->ia_addr.sin6_addr));
1017 			error = EINVAL;
1018 			goto unlink;
1019 		}
1020 		ia->ia_prefixmask = ifra->ifra_prefixmask;
1021 	}
1022 
1023 	/*
1024 	 * If a new destination address is specified, scrub the old one and
1025 	 * install the new destination.  Note that the interface must be
1026 	 * p2p or loopback (see the check above.)
1027 	 */
1028 	if (dst6.sin6_family == AF_INET6 &&
1029 	    !IN6_ARE_ADDR_EQUAL(&dst6.sin6_addr,
1030 				&ia->ia_dstaddr.sin6_addr)) {
1031 		int e;
1032 
1033 		if ((ia->ia_flags & IFA_ROUTE) &&
1034 		    (e = rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST))
1035 		    != 0) {
1036 			log(LOG_ERR, "in6_update_ifa: failed to remove "
1037 			    "a route to the old destination: %s\n",
1038 			    ip6_sprintf(&ia->ia_addr.sin6_addr));
1039 			/* proceed anyway... */
1040 		}
1041 		else
1042 			ia->ia_flags &= ~IFA_ROUTE;
1043 		ia->ia_dstaddr = dst6;
1044 	}
1045 
1046 	/* reset the interface and routing table appropriately. */
1047 	if ((error = in6_ifinit(ifp, ia, &ifra->ifra_addr, hostIsNew)) != 0)
1048 		goto unlink;
1049 
1050 	/*
1051 	 * Beyond this point, we should call in6_purgeaddr upon an error,
1052 	 * not just go to unlink.
1053 	 */
1054 
1055 	if (ifp->if_flags & IFF_MULTICAST) {
1056 		struct sockaddr_in6 mltaddr, mltmask;
1057 		struct in6_multi *in6m;
1058 
1059 		if (hostIsNew) {
1060 			/*
1061 			 * join solicited multicast addr for new host id
1062 			 */
1063 			struct in6_addr llsol;
1064 			bzero(&llsol, sizeof(struct in6_addr));
1065 			llsol.s6_addr16[0] = htons(0xff02);
1066 			llsol.s6_addr16[1] = htons(ifp->if_index);
1067 			llsol.s6_addr32[1] = 0;
1068 			llsol.s6_addr32[2] = htonl(1);
1069 			llsol.s6_addr32[3] =
1070 				ifra->ifra_addr.sin6_addr.s6_addr32[3];
1071 			llsol.s6_addr8[12] = 0xff;
1072 			in6_addmulti(&llsol, ifp, &error);
1073 			if (error != 0) {
1074 				log(LOG_WARNING,
1075 				    "in6_update_ifa: addmulti failed for "
1076 				    "%s on %s (errno=%d)\n",
1077 				    ip6_sprintf(&llsol), if_name(ifp),
1078 				    error);
1079 				in6_purgeaddr((struct ifaddr *)ia);
1080 				return (error);
1081 			}
1082 		}
1083 
1084 		bzero(&mltmask, sizeof(mltmask));
1085 		mltmask.sin6_len = sizeof(struct sockaddr_in6);
1086 		mltmask.sin6_family = AF_INET6;
1087 		mltmask.sin6_addr = in6mask32;
1088 
1089 		/*
1090 		 * join link-local all-nodes address
1091 		 */
1092 		bzero(&mltaddr, sizeof(mltaddr));
1093 		mltaddr.sin6_len = sizeof(struct sockaddr_in6);
1094 		mltaddr.sin6_family = AF_INET6;
1095 		mltaddr.sin6_addr = kin6addr_linklocal_allnodes;
1096 		mltaddr.sin6_addr.s6_addr16[1] = htons(ifp->if_index);
1097 
1098 		in6m = IN6_LOOKUP_MULTI(&mltaddr.sin6_addr, ifp);
1099 		if (in6m == NULL) {
1100 			rtrequest_global(RTM_ADD,
1101 				  (struct sockaddr *)&mltaddr,
1102 				  (struct sockaddr *)&ia->ia_addr,
1103 				  (struct sockaddr *)&mltmask,
1104 				  RTF_UP|RTF_CLONING);  /* xxx */
1105 			in6_addmulti(&mltaddr.sin6_addr, ifp, &error);
1106 			if (error != 0) {
1107 				log(LOG_WARNING,
1108 				    "in6_update_ifa: addmulti failed for "
1109 				    "%s on %s (errno=%d)\n",
1110 				    ip6_sprintf(&mltaddr.sin6_addr),
1111 				    if_name(ifp), error);
1112 			}
1113 		}
1114 
1115 		/*
1116 		 * join node information group address
1117 		 */
1118 #define hostnamelen	strlen(hostname)
1119 		if (in6_nigroup(ifp, hostname, hostnamelen, &mltaddr.sin6_addr)
1120 		    == 0) {
1121 			in6m = IN6_LOOKUP_MULTI(&mltaddr.sin6_addr, ifp);
1122 			if (in6m == NULL && ia != NULL) {
1123 				in6_addmulti(&mltaddr.sin6_addr, ifp, &error);
1124 				if (error != 0) {
1125 					log(LOG_WARNING, "in6_update_ifa: "
1126 					    "addmulti failed for "
1127 					    "%s on %s (errno=%d)\n",
1128 					    ip6_sprintf(&mltaddr.sin6_addr),
1129 					    if_name(ifp), error);
1130 				}
1131 			}
1132 		}
1133 #undef hostnamelen
1134 
1135 		/*
1136 		 * join node-local all-nodes address, on loopback.
1137 		 * XXX: since "node-local" is obsoleted by interface-local,
1138 		 *      we have to join the group on every interface with
1139 		 *      some interface-boundary restriction.
1140 		 */
1141 		if (ifp->if_flags & IFF_LOOPBACK) {
1142 			struct in6_ifaddr *ia_loop;
1143 
1144 			struct in6_addr loop6 = kin6addr_loopback;
1145 			ia_loop = in6ifa_ifpwithaddr(ifp, &loop6);
1146 
1147 			mltaddr.sin6_addr = kin6addr_nodelocal_allnodes;
1148 
1149 			in6m = IN6_LOOKUP_MULTI(&mltaddr.sin6_addr, ifp);
1150 			if (in6m == NULL && ia_loop != NULL) {
1151 				rtrequest_global(RTM_ADD,
1152 					  (struct sockaddr *)&mltaddr,
1153 					  (struct sockaddr *)&ia_loop->ia_addr,
1154 					  (struct sockaddr *)&mltmask,
1155 					  RTF_UP);
1156 				in6_addmulti(&mltaddr.sin6_addr, ifp, &error);
1157 				if (error != 0) {
1158 					log(LOG_WARNING, "in6_update_ifa: "
1159 					    "addmulti failed for %s on %s "
1160 					    "(errno=%d)\n",
1161 					    ip6_sprintf(&mltaddr.sin6_addr),
1162 					    if_name(ifp), error);
1163 				}
1164 			}
1165 		}
1166 	}
1167 
1168 	ia->ia6_flags = ifra->ifra_flags;
1169 	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/*safety*/
1170 	ia->ia6_flags &= ~IN6_IFF_NODAD;	/* Mobile IPv6 */
1171 
1172 	ia->ia6_lifetime = ifra->ifra_lifetime;
1173 	/* for sanity */
1174 	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1175 		ia->ia6_lifetime.ia6t_expire =
1176 			time_uptime + ia->ia6_lifetime.ia6t_vltime;
1177 	} else
1178 		ia->ia6_lifetime.ia6t_expire = 0;
1179 	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1180 		ia->ia6_lifetime.ia6t_preferred =
1181 			time_uptime + ia->ia6_lifetime.ia6t_pltime;
1182 	} else
1183 		ia->ia6_lifetime.ia6t_preferred = 0;
1184 
1185 	/*
1186 	 * Perform DAD, if needed.
1187 	 * XXX It may be of use, if we can administratively
1188 	 * disable DAD.
1189 	 */
1190 	if (in6if_do_dad(ifp) && !(ifra->ifra_flags & IN6_IFF_NODAD)) {
1191 		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1192 		nd6_dad_start((struct ifaddr *)ia, NULL);
1193 	}
1194 
1195 	return (error);
1196 
1197 unlink:
1198 	/*
1199 	 * XXX: if a change of an existing address failed, keep the entry
1200 	 * anyway.
1201 	 */
1202 	if (hostIsNew)
1203 		in6_unlink_ifa(ia, ifp);
1204 	return (error);
1205 }
1206 
1207 void
1208 in6_purgeaddr(struct ifaddr *ifa)
1209 {
1210 	struct ifnet *ifp = ifa->ifa_ifp;
1211 	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1212 
1213 	/* stop DAD processing */
1214 	nd6_dad_stop(ifa);
1215 
1216 	/*
1217 	 * delete route to the destination of the address being purged.
1218 	 * The interface must be p2p or loopback in this case.
1219 	 */
1220 	if ((ia->ia_flags & IFA_ROUTE) && ia->ia_dstaddr.sin6_len != 0) {
1221 		int e;
1222 
1223 		if ((e = rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST))
1224 		    != 0) {
1225 			log(LOG_ERR, "in6_purgeaddr: failed to remove "
1226 			    "a route to the p2p destination: %s on %s, "
1227 			    "errno=%d\n",
1228 			    ip6_sprintf(&ia->ia_addr.sin6_addr), if_name(ifp),
1229 			    e);
1230 			/* proceed anyway... */
1231 		}
1232 		else
1233 			ia->ia_flags &= ~IFA_ROUTE;
1234 	}
1235 
1236 	/* Remove ownaddr's loopback rtentry, if it exists. */
1237 	in6_ifremloop(&(ia->ia_ifa));
1238 
1239 	if (ifp->if_flags & IFF_MULTICAST) {
1240 		/*
1241 		 * delete solicited multicast addr for deleting host id
1242 		 */
1243 		struct in6_multi *in6m;
1244 		struct in6_addr llsol;
1245 		bzero(&llsol, sizeof(struct in6_addr));
1246 		llsol.s6_addr16[0] = htons(0xff02);
1247 		llsol.s6_addr16[1] = htons(ifp->if_index);
1248 		llsol.s6_addr32[1] = 0;
1249 		llsol.s6_addr32[2] = htonl(1);
1250 		llsol.s6_addr32[3] =
1251 			ia->ia_addr.sin6_addr.s6_addr32[3];
1252 		llsol.s6_addr8[12] = 0xff;
1253 
1254 		in6m = IN6_LOOKUP_MULTI(&llsol, ifp);
1255 		if (in6m)
1256 			in6_delmulti(in6m);
1257 	}
1258 
1259 	in6_unlink_ifa(ia, ifp);
1260 }
1261 
1262 static void
1263 in6_unlink_ifa(struct in6_ifaddr *ia, struct ifnet *ifp)
1264 {
1265 	struct in6_ifaddr *oia;
1266 
1267 	crit_enter();
1268 
1269 	ifa_ifunlink(&ia->ia_ifa, ifp);
1270 
1271 	oia = ia;
1272 	if (oia == (ia = in6_ifaddr))
1273 		in6_ifaddr = ia->ia_next;
1274 	else {
1275 		while (ia->ia_next && (ia->ia_next != oia))
1276 			ia = ia->ia_next;
1277 		if (ia->ia_next)
1278 			ia->ia_next = oia->ia_next;
1279 		else {
1280 			/* search failed */
1281 			kprintf("Couldn't unlink in6_ifaddr from in6_ifaddr\n");
1282 		}
1283 	}
1284 
1285 	/*
1286 	 * When an autoconfigured address is being removed, release the
1287 	 * reference to the base prefix.  Also, since the release might
1288 	 * affect the status of other (detached) addresses, call
1289 	 * pfxlist_onlink_check().
1290 	 */
1291 	if (oia->ia6_flags & IN6_IFF_AUTOCONF) {
1292 		if (oia->ia6_ndpr == NULL) {
1293 			log(LOG_NOTICE, "in6_unlink_ifa: autoconf'ed address "
1294 			    "%p has no prefix\n", oia);
1295 		} else {
1296 			oia->ia6_ndpr->ndpr_refcnt--;
1297 			oia->ia6_flags &= ~IN6_IFF_AUTOCONF;
1298 			oia->ia6_ndpr = NULL;
1299 		}
1300 
1301 		pfxlist_onlink_check();
1302 	}
1303 
1304 	/*
1305 	 * release another refcnt for the link from in6_ifaddr.
1306 	 * Note that we should decrement the refcnt at least once for all *BSD.
1307 	 */
1308 	ifa_destroy(&oia->ia_ifa);
1309 
1310 	crit_exit();
1311 }
1312 
1313 void
1314 in6_purgeif(struct ifnet *ifp)
1315 {
1316 	struct ifaddr_container *ifac, *next;
1317 
1318 	TAILQ_FOREACH_MUTABLE(ifac, &ifp->if_addrheads[mycpuid],
1319 			      ifa_link, next) {
1320 		if (ifac->ifa->ifa_addr->sa_family != AF_INET6)
1321 			continue;
1322 		in6_purgeaddr(ifac->ifa);
1323 	}
1324 
1325 	in6_ifdetach(ifp);
1326 }
1327 
1328 /*
1329  * SIOC[GAD]LIFADDR.
1330  *	SIOCGLIFADDR: get first address. (?)
1331  *	SIOCGLIFADDR with IFLR_PREFIX:
1332  *		get first address that matches the specified prefix.
1333  *	SIOCALIFADDR: add the specified address.
1334  *	SIOCALIFADDR with IFLR_PREFIX:
1335  *		add the specified prefix, filling hostid part from
1336  *		the first link-local address.  prefixlen must be <= 64.
1337  *	SIOCDLIFADDR: delete the specified address.
1338  *	SIOCDLIFADDR with IFLR_PREFIX:
1339  *		delete the first address that matches the specified prefix.
1340  * return values:
1341  *	EINVAL on invalid parameters
1342  *	EADDRNOTAVAIL on prefix match failed/specified address not found
1343  *	other values may be returned from in6_ioctl()
1344  *
1345  * NOTE: SIOCALIFADDR(with IFLR_PREFIX set) allows prefixlen less than 64.
1346  * this is to accomodate address naming scheme other than RFC2374,
1347  * in the future.
1348  * RFC2373 defines interface id to be 64bit, but it allows non-RFC2374
1349  * address encoding scheme. (see figure on page 8)
1350  */
1351 static int
1352 in6_lifaddr_ioctl(u_long cmd, caddr_t data, struct ifnet *ifp,
1353     struct thread *td)
1354 {
1355 	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
1356 	struct sockaddr *sa;
1357 
1358 	/* sanity checks */
1359 	if (!data || !ifp) {
1360 		panic("invalid argument to in6_lifaddr_ioctl");
1361 		/*NOTRECHED*/
1362 	}
1363 
1364 	switch (cmd) {
1365 	case SIOCGLIFADDR:
1366 		/* address must be specified on GET with IFLR_PREFIX */
1367 		if (!(iflr->flags & IFLR_PREFIX))
1368 			break;
1369 		/* FALLTHROUGH */
1370 	case SIOCALIFADDR:
1371 	case SIOCDLIFADDR:
1372 		/* address must be specified on ADD and DELETE */
1373 		sa = (struct sockaddr *)&iflr->addr;
1374 		if (sa->sa_family != AF_INET6)
1375 			return EINVAL;
1376 		if (sa->sa_len != sizeof(struct sockaddr_in6))
1377 			return EINVAL;
1378 		/* XXX need improvement */
1379 		sa = (struct sockaddr *)&iflr->dstaddr;
1380 		if (sa->sa_family && sa->sa_family != AF_INET6)
1381 			return EINVAL;
1382 		if (sa->sa_len && sa->sa_len != sizeof(struct sockaddr_in6))
1383 			return EINVAL;
1384 		break;
1385 	default: /* shouldn't happen */
1386 #if 0
1387 		panic("invalid cmd to in6_lifaddr_ioctl");
1388 		/* NOTREACHED */
1389 #else
1390 		return EOPNOTSUPP;
1391 #endif
1392 	}
1393 	if (sizeof(struct in6_addr) * 8 < iflr->prefixlen)
1394 		return EINVAL;
1395 
1396 	switch (cmd) {
1397 	case SIOCALIFADDR:
1398 	    {
1399 		struct in6_aliasreq ifra;
1400 		struct in6_addr *hostid = NULL;
1401 		int prefixlen;
1402 
1403 		if (iflr->flags & IFLR_PREFIX) {
1404 			struct ifaddr *ifa;
1405 			struct sockaddr_in6 *sin6;
1406 
1407 			/*
1408 			 * hostid is to fill in the hostid part of the
1409 			 * address.  hostid points to the first link-local
1410 			 * address attached to the interface.
1411 			 */
1412 			ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0);
1413 			if (!ifa)
1414 				return EADDRNOTAVAIL;
1415 			hostid = IFA_IN6(ifa);
1416 
1417 		 	/* prefixlen must be <= 64. */
1418 			if (64 < iflr->prefixlen)
1419 				return EINVAL;
1420 			prefixlen = iflr->prefixlen;
1421 
1422 			/* hostid part must be zero. */
1423 			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1424 			if (sin6->sin6_addr.s6_addr32[2] != 0
1425 			 || sin6->sin6_addr.s6_addr32[3] != 0) {
1426 				return EINVAL;
1427 			}
1428 		} else
1429 			prefixlen = iflr->prefixlen;
1430 
1431 		/* copy args to in6_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
1432 		bzero(&ifra, sizeof(ifra));
1433 		bcopy(iflr->iflr_name, ifra.ifra_name,
1434 			sizeof(ifra.ifra_name));
1435 
1436 		bcopy(&iflr->addr, &ifra.ifra_addr,
1437 			((struct sockaddr *)&iflr->addr)->sa_len);
1438 		if (hostid) {
1439 			/* fill in hostid part */
1440 			ifra.ifra_addr.sin6_addr.s6_addr32[2] =
1441 				hostid->s6_addr32[2];
1442 			ifra.ifra_addr.sin6_addr.s6_addr32[3] =
1443 				hostid->s6_addr32[3];
1444 		}
1445 
1446 		if (((struct sockaddr *)&iflr->dstaddr)->sa_family) {	/*XXX*/
1447 			bcopy(&iflr->dstaddr, &ifra.ifra_dstaddr,
1448 				((struct sockaddr *)&iflr->dstaddr)->sa_len);
1449 			if (hostid) {
1450 				ifra.ifra_dstaddr.sin6_addr.s6_addr32[2] =
1451 					hostid->s6_addr32[2];
1452 				ifra.ifra_dstaddr.sin6_addr.s6_addr32[3] =
1453 					hostid->s6_addr32[3];
1454 			}
1455 		}
1456 
1457 		ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
1458 		in6_prefixlen2mask(&ifra.ifra_prefixmask.sin6_addr, prefixlen);
1459 
1460 		ifra.ifra_flags = iflr->flags & ~IFLR_PREFIX;
1461 		return in6_control_internal(SIOCAIFADDR_IN6, (caddr_t)&ifra,
1462 		    ifp, td);
1463 	    }
1464 	case SIOCGLIFADDR:
1465 	case SIOCDLIFADDR:
1466 	    {
1467 		struct ifaddr_container *ifac;
1468 		struct in6_ifaddr *ia;
1469 		struct in6_addr mask, candidate, match;
1470 		struct sockaddr_in6 *sin6;
1471 		int cmp;
1472 
1473 		bzero(&mask, sizeof(mask));
1474 		if (iflr->flags & IFLR_PREFIX) {
1475 			/* lookup a prefix rather than address. */
1476 			in6_prefixlen2mask(&mask, iflr->prefixlen);
1477 
1478 			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1479 			bcopy(&sin6->sin6_addr, &match, sizeof(match));
1480 			match.s6_addr32[0] &= mask.s6_addr32[0];
1481 			match.s6_addr32[1] &= mask.s6_addr32[1];
1482 			match.s6_addr32[2] &= mask.s6_addr32[2];
1483 			match.s6_addr32[3] &= mask.s6_addr32[3];
1484 
1485 			/* if you set extra bits, that's wrong */
1486 			if (bcmp(&match, &sin6->sin6_addr, sizeof(match)))
1487 				return EINVAL;
1488 
1489 			cmp = 1;
1490 		} else {
1491 			if (cmd == SIOCGLIFADDR) {
1492 				/* on getting an address, take the 1st match */
1493 				cmp = 0;	/* XXX */
1494 			} else {
1495 				/* on deleting an address, do exact match */
1496 				in6_prefixlen2mask(&mask, 128);
1497 				sin6 = (struct sockaddr_in6 *)&iflr->addr;
1498 				bcopy(&sin6->sin6_addr, &match, sizeof(match));
1499 
1500 				cmp = 1;
1501 			}
1502 		}
1503 
1504 		TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
1505 			struct ifaddr *ifa = ifac->ifa;
1506 
1507 			if (ifa->ifa_addr->sa_family != AF_INET6)
1508 				continue;
1509 			if (!cmp)
1510 				break;
1511 
1512 			bcopy(IFA_IN6(ifa), &candidate, sizeof(candidate));
1513 			/*
1514 			 * XXX: this is adhoc, but is necessary to allow
1515 			 * a user to specify fe80::/64 (not /10) for a
1516 			 * link-local address.
1517 			 */
1518 			if (IN6_IS_ADDR_LINKLOCAL(&candidate))
1519 				candidate.s6_addr16[1] = 0;
1520 			candidate.s6_addr32[0] &= mask.s6_addr32[0];
1521 			candidate.s6_addr32[1] &= mask.s6_addr32[1];
1522 			candidate.s6_addr32[2] &= mask.s6_addr32[2];
1523 			candidate.s6_addr32[3] &= mask.s6_addr32[3];
1524 			if (IN6_ARE_ADDR_EQUAL(&candidate, &match))
1525 				break;
1526 		}
1527 		if (ifac == NULL)
1528 			return EADDRNOTAVAIL;
1529 		ia = ifa2ia6(ifac->ifa);
1530 
1531 		if (cmd == SIOCGLIFADDR) {
1532 			struct sockaddr_in6 *s6;
1533 
1534 			/* fill in the if_laddrreq structure */
1535 			bcopy(&ia->ia_addr, &iflr->addr, ia->ia_addr.sin6_len);
1536 			s6 = (struct sockaddr_in6 *)&iflr->addr;
1537 			if (IN6_IS_ADDR_LINKLOCAL(&s6->sin6_addr)) {
1538 				s6->sin6_addr.s6_addr16[1] = 0;
1539 				if (in6_addr2zoneid(ifp, &s6->sin6_addr,
1540 				    &s6->sin6_scope_id))
1541 					return (EINVAL);/* XXX */
1542 			}
1543 			if (ifp->if_flags & IFF_POINTOPOINT) {
1544 				bcopy(&ia->ia_dstaddr, &iflr->dstaddr,
1545 					ia->ia_dstaddr.sin6_len);
1546 				s6 = (struct sockaddr_in6 *)&iflr->dstaddr;
1547 				if (IN6_IS_ADDR_LINKLOCAL(&s6->sin6_addr)) {
1548 					s6->sin6_addr.s6_addr16[1] = 0;
1549 					if (in6_addr2zoneid(ifp,
1550 					    &s6->sin6_addr, &s6->sin6_scope_id))
1551 						return (EINVAL); /* EINVAL */
1552 				}
1553 			} else
1554 				bzero(&iflr->dstaddr, sizeof(iflr->dstaddr));
1555 
1556 			iflr->prefixlen =
1557 				in6_mask2len(&ia->ia_prefixmask.sin6_addr,
1558 					     NULL);
1559 
1560 			iflr->flags = ia->ia6_flags;	/* XXX */
1561 
1562 			return 0;
1563 		} else {
1564 			struct in6_aliasreq ifra;
1565 
1566 			/* fill in6_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
1567 			bzero(&ifra, sizeof(ifra));
1568 			bcopy(iflr->iflr_name, ifra.ifra_name,
1569 			      sizeof(ifra.ifra_name));
1570 
1571 			bcopy(&ia->ia_addr, &ifra.ifra_addr,
1572 			      ia->ia_addr.sin6_len);
1573 			if (ifp->if_flags & IFF_POINTOPOINT)
1574 				bcopy(&ia->ia_dstaddr, &ifra.ifra_dstaddr,
1575 				      ia->ia_dstaddr.sin6_len);
1576 			else
1577 				bzero(&ifra.ifra_dstaddr,
1578 				      sizeof(ifra.ifra_dstaddr));
1579 			bcopy(&ia->ia_prefixmask, &ifra.ifra_dstaddr,
1580 			      ia->ia_prefixmask.sin6_len);
1581 
1582 			ifra.ifra_flags = ia->ia6_flags;
1583 			return in6_control_internal(SIOCDIFADDR_IN6,
1584 			    (caddr_t)&ifra, ifp, td);
1585 		}
1586 	    }
1587 	}
1588 
1589 	return EOPNOTSUPP;	/* just for safety */
1590 }
1591 
1592 /*
1593  * Initialize an interface's intetnet6 address
1594  * and routing table entry.
1595  */
1596 static int
1597 in6_ifinit(struct ifnet *ifp, struct in6_ifaddr *ia, struct sockaddr_in6 *sin6,
1598 	   int newhost)
1599 {
1600 	int error = 0, plen;
1601 
1602 	ia->ia_addr = *sin6;
1603 
1604 	if (ifp->if_ioctl != NULL) {
1605 		ifnet_serialize_all(ifp);
1606 		error = ifp->if_ioctl(ifp, SIOCSIFADDR, (caddr_t)ia, NULL);
1607 		ifnet_deserialize_all(ifp);
1608 		if (error)
1609 			return (error);
1610 	}
1611 
1612 	ia->ia_ifa.ifa_metric = ifp->if_metric;
1613 
1614 	/* we could do in(6)_socktrim here, but just omit it at this moment. */
1615 
1616 	/*
1617 	 * Special case:
1618 	 * If the destination address is specified for a point-to-point
1619 	 * interface, install a route to the destination as an interface
1620 	 * direct route.
1621 	 */
1622 	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1623 	if (plen == 128 && ia->ia_dstaddr.sin6_family == AF_INET6) {
1624 		if ((error = rtinit(&(ia->ia_ifa), (int)RTM_ADD,
1625 				    RTF_UP | RTF_HOST)) != 0)
1626 			return (error);
1627 		ia->ia_flags |= IFA_ROUTE;
1628 	}
1629 	if (plen < 128) {
1630 		/*
1631 		 * The RTF_CLONING flag is necessary for in6_is_ifloop_auto().
1632 		 */
1633 		ia->ia_ifa.ifa_flags |= RTF_CLONING;
1634 	}
1635 
1636 	/* Add ownaddr as loopback rtentry, if necessary (ex. on p2p link). */
1637 	if (newhost) {
1638 		/* set the rtrequest function to create llinfo */
1639 		ia->ia_ifa.ifa_rtrequest = nd6_rtrequest;
1640 		in6_ifaddloop(&(ia->ia_ifa));
1641 	}
1642 
1643 	return (error);
1644 }
1645 
1646 struct in6_multi_mship *
1647 in6_joingroup(struct ifnet *ifp, struct in6_addr *addr, int *errorp)
1648 {
1649        struct in6_multi_mship *imm;
1650 
1651        imm = kmalloc(sizeof(*imm), M_IPMADDR, M_NOWAIT);
1652        if (!imm) {
1653                *errorp = ENOBUFS;
1654                return NULL;
1655        }
1656        imm->i6mm_maddr = in6_addmulti(addr, ifp, errorp);
1657        if (!imm->i6mm_maddr) {
1658                /* *errorp is alrady set */
1659                kfree(imm, M_IPMADDR);
1660                return NULL;
1661        }
1662        return imm;
1663 }
1664 
1665 int
1666 in6_leavegroup(struct in6_multi_mship *imm)
1667 {
1668 
1669        if (imm->i6mm_maddr)
1670                in6_delmulti(imm->i6mm_maddr);
1671        kfree(imm,  M_IPMADDR);
1672        return 0;
1673 }
1674 
1675 /*
1676  * Add an address to the list of IP6 multicast addresses for a
1677  * given interface.
1678  */
1679 struct	in6_multi *
1680 in6_addmulti(struct in6_addr *maddr6, struct ifnet *ifp, int *errorp)
1681 {
1682 	struct	in6_multi *in6m;
1683 	struct sockaddr_in6 sin6;
1684 	struct ifmultiaddr *ifma;
1685 
1686 	*errorp = 0;
1687 
1688 	crit_enter();
1689 
1690 	/*
1691 	 * Call generic routine to add membership or increment
1692 	 * refcount.  It wants addresses in the form of a sockaddr,
1693 	 * so we build one here (being careful to zero the unused bytes).
1694 	 */
1695 	bzero(&sin6, sizeof sin6);
1696 	sin6.sin6_family = AF_INET6;
1697 	sin6.sin6_len = sizeof sin6;
1698 	sin6.sin6_addr = *maddr6;
1699 	*errorp = if_addmulti(ifp, (struct sockaddr *)&sin6, &ifma);
1700 	if (*errorp) {
1701 		crit_exit();
1702 		return 0;
1703 	}
1704 
1705 	/*
1706 	 * If ifma->ifma_protospec is null, then if_addmulti() created
1707 	 * a new record.  Otherwise, we are done.
1708 	 */
1709 	if (ifma->ifma_protospec != NULL) {
1710 		crit_exit();
1711 		return ifma->ifma_protospec;
1712 	}
1713 
1714 	in6m = kmalloc(sizeof(*in6m), M_IPMADDR, M_INTWAIT | M_ZERO);
1715 	in6m->in6m_addr = *maddr6;
1716 	in6m->in6m_ifp = ifp;
1717 	in6m->in6m_ifma = ifma;
1718 	ifma->ifma_protospec = in6m;
1719 	LIST_INSERT_HEAD(&in6_multihead, in6m, in6m_entry);
1720 
1721 	/*
1722 	 * Let MLD6 know that we have joined a new IP6 multicast
1723 	 * group.
1724 	 */
1725 	mld6_start_listening(in6m);
1726 	crit_exit();
1727 	return (in6m);
1728 }
1729 
1730 /*
1731  * Delete a multicast address record.
1732  */
1733 void
1734 in6_delmulti(struct in6_multi *in6m)
1735 {
1736 	struct ifmultiaddr *ifma = in6m->in6m_ifma;
1737 
1738 	crit_enter();
1739 
1740 	if (ifma->ifma_refcount == 1) {
1741 		/*
1742 		 * No remaining claims to this record; let MLD6 know
1743 		 * that we are leaving the multicast group.
1744 		 */
1745 		mld6_stop_listening(in6m);
1746 		ifma->ifma_protospec = NULL;
1747 		LIST_REMOVE(in6m, in6m_entry);
1748 		kfree(in6m, M_IPMADDR);
1749 	}
1750 	/* XXX - should be separate API for when we have an ifma? */
1751 	if_delmulti(ifma->ifma_ifp, ifma->ifma_addr);
1752 	crit_exit();
1753 }
1754 
1755 /*
1756  * Find an IPv6 interface link-local address specific to an interface.
1757  */
1758 struct in6_ifaddr *
1759 in6ifa_ifpforlinklocal(struct ifnet *ifp, int ignoreflags)
1760 {
1761 	const struct ifaddr_container *ifac;
1762 
1763 	TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
1764 		struct ifaddr *ifa = ifac->ifa;
1765 
1766 		if (ifa->ifa_addr == NULL)
1767 			continue;	/* just for safety */
1768 		if (ifa->ifa_addr->sa_family != AF_INET6)
1769 			continue;
1770 		if (IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa))) {
1771 			if ((((struct in6_ifaddr *)ifa)->ia6_flags &
1772 			     ignoreflags) != 0)
1773 				continue;
1774 			return (struct in6_ifaddr *)ifa;
1775 		}
1776 	}
1777 	return NULL;
1778 }
1779 
1780 
1781 /*
1782  * find the internet address corresponding to a given interface and address.
1783  */
1784 struct in6_ifaddr *
1785 in6ifa_ifpwithaddr(struct ifnet *ifp, struct in6_addr *addr)
1786 {
1787 	const struct ifaddr_container *ifac;
1788 
1789 	TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
1790 		struct ifaddr *ifa = ifac->ifa;
1791 
1792 		if (ifa->ifa_addr == NULL)
1793 			continue;	/* just for safety */
1794 		if (ifa->ifa_addr->sa_family != AF_INET6)
1795 			continue;
1796 		if (IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa)))
1797 			return (struct in6_ifaddr *)ifa;
1798 	}
1799 	return NULL;
1800 }
1801 
1802 /*
1803  * Find a link-local scoped address on ifp and return it if any.
1804  */
1805 struct in6_ifaddr *
1806 in6ifa_llaonifp(struct ifnet *ifp)
1807 {
1808 	const struct ifaddr_container *ifac;
1809 
1810 	TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
1811 		const struct sockaddr_in6 *sin6;
1812 		struct ifaddr *ifa = ifac->ifa;
1813 
1814 		if (ifa->ifa_addr->sa_family != AF_INET6)
1815 			continue;
1816 		sin6 = (const struct sockaddr_in6 *)ifa->ifa_addr;
1817 		if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr) ||
1818 		    /* XXX why are mcast addresses ifp address list? */
1819 		    IN6_IS_ADDR_MC_INTFACELOCAL(&sin6->sin6_addr) ||
1820 		    IN6_IS_ADDR_MC_NODELOCAL(&sin6->sin6_addr))
1821 			return (struct in6_ifaddr *)ifa;
1822 	}
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 in6_ifaddr *ifa_best = NULL;
2019 	u_int32_t dstzone, odstzone;
2020 	int jailed = 0;
2021 	const struct ifnet_array *arr;
2022 	int i;
2023 
2024 	if(cred && cred->cr_prison)
2025 		jailed = 1;
2026 
2027 	if (oifp == NULL)
2028 		return (NULL);
2029 
2030 	if (in6_addr2zoneid(oifp, dst, &odstzone))
2031 		return (NULL);
2032 
2033 	/*
2034 	 * We search for all addresses on all interfaces from the beginning.
2035 	 * Comparing an interface with the outgoing interface will be done
2036 	 * only at the final stage of tiebreaking.
2037 	 */
2038 	arr = ifnet_array_get();
2039 	for (i = 0; i < arr->ifnet_count; ++i) {
2040 		struct ifnet *ifp = arr->ifnet_arr[i];
2041 		struct ifaddr_container *ifac;
2042 
2043 		/*
2044 		 * We can never take an address that breaks the scope zone
2045 		 * of the destination.
2046 		 */
2047 		if (ifp->if_afdata[AF_INET6] == NULL)
2048 			continue;
2049 		if (in6_addr2zoneid(ifp, dst, &dstzone) || dstzone != odstzone)
2050 			continue;
2051 
2052 		TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
2053 			int tlen = -1, dscopecmp, bscopecmp, matchcmp;
2054 			struct ifaddr *ifa = ifac->ifa;
2055 
2056 			if (ifa->ifa_addr->sa_family != AF_INET6)
2057 				continue;
2058 
2059 			src_scope = in6_addrscope(IFA_IN6(ifa));
2060 
2061 			/*
2062 			 * Don't use an address before completing DAD
2063 			 * nor a duplicated address.
2064 			 */
2065 			if (((struct in6_ifaddr *)ifa)->ia6_flags &
2066 			    IN6_IFF_NOTREADY)
2067 				continue;
2068 
2069 			/* XXX: is there any case to allow anycasts? */
2070 			if (((struct in6_ifaddr *)ifa)->ia6_flags &
2071 			    IN6_IFF_ANYCAST)
2072 				continue;
2073 
2074 			if (((struct in6_ifaddr *)ifa)->ia6_flags &
2075 			    IN6_IFF_DETACHED)
2076 				continue;
2077 
2078 			/* Skip adresses not valid for current jail */
2079 			if (jailed &&
2080 			    !(jailed_ip(cred->cr_prison, (struct sockaddr *)(ifa->ifa_addr)) != 0))
2081 				continue;
2082 
2083 			/*
2084 			 * If this is the first address we find,
2085 			 * keep it anyway.
2086 			 */
2087 			if (ifa_best == NULL)
2088 				goto replace;
2089 
2090 			/*
2091 			 * ifa_best is never NULL beyond this line except
2092 			 * within the block labeled "replace".
2093 			 */
2094 
2095 			/*
2096 			 * If ifa_best has a smaller scope than dst and
2097 			 * the current address has a larger one than
2098 			 * (or equal to) dst, always replace ifa_best.
2099 			 * Also, if the current address has a smaller scope
2100 			 * than dst, ignore it unless ifa_best also has a
2101 			 * smaller scope.
2102 			 * Consequently, after the two if-clause below,
2103 			 * the followings must be satisfied:
2104 			 * (scope(src) < scope(dst) &&
2105 			 *  scope(best) < scope(dst))
2106 			 *  OR
2107 			 * (scope(best) >= scope(dst) &&
2108 			 *  scope(src) >= scope(dst))
2109 			 */
2110 			if (IN6_ARE_SCOPE_CMP(best_scope, dst_scope) < 0 &&
2111 			    IN6_ARE_SCOPE_CMP(src_scope, dst_scope) >= 0)
2112 				goto replace; /* (A) */
2113 			if (IN6_ARE_SCOPE_CMP(src_scope, dst_scope) < 0 &&
2114 			    IN6_ARE_SCOPE_CMP(best_scope, dst_scope) >= 0)
2115 				continue; /* (B) */
2116 
2117 			/*
2118 			 * A deprecated address SHOULD NOT be used in new
2119 			 * communications if an alternate (non-deprecated)
2120 			 * address is available and has sufficient scope.
2121 			 * RFC 2462, Section 5.5.4.
2122 			 */
2123 			if (((struct in6_ifaddr *)ifa)->ia6_flags &
2124 			    IN6_IFF_DEPRECATED) {
2125 				/*
2126 				 * Ignore any deprecated addresses if
2127 				 * specified by configuration.
2128 				 */
2129 				if (!ip6_use_deprecated)
2130 					continue;
2131 
2132 				/*
2133 				 * If we have already found a non-deprecated
2134 				 * candidate, just ignore deprecated addresses.
2135 				 */
2136 				if (!(ifa_best->ia6_flags & IN6_IFF_DEPRECATED))
2137 					continue;
2138 			}
2139 
2140 			/*
2141 			 * A non-deprecated address is always preferred
2142 			 * to a deprecated one regardless of scopes and
2143 			 * address matching (Note invariants ensured by the
2144 			 * conditions (A) and (B) above.)
2145 			 */
2146 			if ((ifa_best->ia6_flags & IN6_IFF_DEPRECATED) &&
2147 			    !(((struct in6_ifaddr *)ifa)->ia6_flags &
2148 			     IN6_IFF_DEPRECATED))
2149 				goto replace;
2150 
2151 			/*
2152 			 * When we use temporary addresses described in
2153 			 * RFC 3041, we prefer temporary addresses to
2154 			 * public autoconf addresses.  Again, note the
2155 			 * invariants from (A) and (B).  Also note that we
2156 			 * don't have any preference between static addresses
2157 			 * and autoconf addresses (despite of whether or not
2158 			 * the latter is temporary or public.)
2159 			 */
2160 			if (ip6_use_tempaddr) {
2161 				struct in6_ifaddr *ifat;
2162 
2163 				ifat = (struct in6_ifaddr *)ifa;
2164 				if ((ifa_best->ia6_flags &
2165 				     (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY))
2166 				     == IN6_IFF_AUTOCONF &&
2167 				    (ifat->ia6_flags &
2168 				     (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY))
2169 				     == (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY)) {
2170 					goto replace;
2171 				}
2172 				if ((ifa_best->ia6_flags &
2173 				     (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY))
2174 				    == (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY) &&
2175 				    (ifat->ia6_flags &
2176 				     (IN6_IFF_AUTOCONF|IN6_IFF_TEMPORARY))
2177 				     == IN6_IFF_AUTOCONF) {
2178 					continue;
2179 				}
2180 			}
2181 
2182 			/*
2183 			 * At this point, we have two cases:
2184 			 * 1. we are looking at a non-deprecated address,
2185 			 *    and ifa_best is also non-deprecated.
2186 			 * 2. we are looking at a deprecated address,
2187 			 *    and ifa_best is also deprecated.
2188 			 * Also, we do not have to consider a case where
2189 			 * the scope of if_best is larger(smaller) than dst and
2190 			 * the scope of the current address is smaller(larger)
2191 			 * than dst. Such a case has already been covered.
2192 			 * Tiebreaking is done according to the following
2193 			 * items:
2194 			 * - the scope comparison between the address and
2195 			 *   dst (dscopecmp)
2196 			 * - the scope comparison between the address and
2197 			 *   ifa_best (bscopecmp)
2198 			 * - if the address match dst longer than ifa_best
2199 			 *   (matchcmp)
2200 			 * - if the address is on the outgoing I/F (outI/F)
2201 			 *
2202 			 * Roughly speaking, the selection policy is
2203 			 * - the most important item is scope. The same scope
2204 			 *   is best. Then search for a larger scope.
2205 			 *   Smaller scopes are the last resort.
2206 			 * - A deprecated address is chosen only when we have
2207 			 *   no address that has an enough scope, but is
2208 			 *   prefered to any addresses of smaller scopes
2209 			 *   (this must be already done above.)
2210 			 * - addresses on the outgoing I/F are preferred to
2211 			 *   ones on other interfaces if none of above
2212 			 *   tiebreaks.  In the table below, the column "bI"
2213 			 *   means if the best_ifa is on the outgoing
2214 			 *   interface, and the column "sI" means if the ifa
2215 			 *   is on the outgoing interface.
2216 			 * - If there is no other reasons to choose one,
2217 			 *   longest address match against dst is considered.
2218 			 *
2219 			 * The precise decision table is as follows:
2220 			 * dscopecmp bscopecmp    match  bI oI | replace?
2221 			 *       N/A     equal      N/A   Y  N |   No (1)
2222 			 *       N/A     equal      N/A   N  Y |  Yes (2)
2223 			 *       N/A     equal   larger    N/A |  Yes (3)
2224 			 *       N/A     equal  !larger    N/A |   No (4)
2225 			 *    larger    larger      N/A    N/A |   No (5)
2226 			 *    larger   smaller      N/A    N/A |  Yes (6)
2227 			 *   smaller    larger      N/A    N/A |  Yes (7)
2228 			 *   smaller   smaller      N/A    N/A |   No (8)
2229 			 *     equal   smaller      N/A    N/A |  Yes (9)
2230 			 *     equal    larger       (already done at A above)
2231 			 */
2232 			dscopecmp = IN6_ARE_SCOPE_CMP(src_scope, dst_scope);
2233 			bscopecmp = IN6_ARE_SCOPE_CMP(src_scope, best_scope);
2234 
2235 			if (bscopecmp == 0) {
2236 				struct ifnet *bifp = ifa_best->ia_ifp;
2237 
2238 				if (bifp == oifp && ifp != oifp) /* (1) */
2239 					continue;
2240 				if (bifp != oifp && ifp == oifp) /* (2) */
2241 					goto replace;
2242 
2243 				/*
2244 				 * Both bifp and ifp are on the outgoing
2245 				 * interface, or both two are on a different
2246 				 * interface from the outgoing I/F.
2247 				 * now we need address matching against dst
2248 				 * for tiebreaking.
2249 				 */
2250 				tlen = in6_matchlen(IFA_IN6(ifa), dst);
2251 				matchcmp = tlen - blen;
2252 				if (matchcmp > 0) /* (3) */
2253 					goto replace;
2254 				continue; /* (4) */
2255 			}
2256 			if (dscopecmp > 0) {
2257 				if (bscopecmp > 0) /* (5) */
2258 					continue;
2259 				goto replace; /* (6) */
2260 			}
2261 			if (dscopecmp < 0) {
2262 				if (bscopecmp > 0) /* (7) */
2263 					goto replace;
2264 				continue; /* (8) */
2265 			}
2266 
2267 			/* now dscopecmp must be 0 */
2268 			if (bscopecmp < 0)
2269 				goto replace; /* (9) */
2270 
2271 replace:
2272 			ifa_best = (struct in6_ifaddr *)ifa;
2273 			blen = tlen >= 0 ? tlen :
2274 				in6_matchlen(IFA_IN6(ifa), dst);
2275 			best_scope = in6_addrscope(&ifa_best->ia_addr.sin6_addr);
2276 		}
2277 	}
2278 
2279 	/* count statistics for future improvements */
2280 	if (ifa_best == NULL)
2281 		ip6stat.ip6s_sources_none++;
2282 	else {
2283 		if (oifp == ifa_best->ia_ifp)
2284 			ip6stat.ip6s_sources_sameif[best_scope]++;
2285 		else
2286 			ip6stat.ip6s_sources_otherif[best_scope]++;
2287 
2288 		if (best_scope == dst_scope)
2289 			ip6stat.ip6s_sources_samescope[best_scope]++;
2290 		else
2291 			ip6stat.ip6s_sources_otherscope[best_scope]++;
2292 
2293 		if (ifa_best->ia6_flags & IN6_IFF_DEPRECATED)
2294 			ip6stat.ip6s_sources_deprecated[best_scope]++;
2295 	}
2296 
2297 	return (ifa_best);
2298 }
2299 
2300 /*
2301  * return the best address out of the same scope. if no address was
2302  * found, return the first valid address from designated IF.
2303  */
2304 struct in6_ifaddr *
2305 in6_ifawithifp(struct ifnet *ifp, struct in6_addr *dst)
2306 {
2307 	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
2308 	struct ifaddr_container *ifac;
2309 	struct in6_ifaddr *besta = NULL;
2310 	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
2311 
2312 	dep[0] = dep[1] = NULL;
2313 
2314 	/*
2315 	 * We first look for addresses in the same scope.
2316 	 * If there is one, return it.
2317 	 * If two or more, return one which matches the dst longest.
2318 	 * If none, return one of global addresses assigned other ifs.
2319 	 */
2320 	TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
2321 		struct ifaddr *ifa = ifac->ifa;
2322 
2323 		if (ifa->ifa_addr->sa_family != AF_INET6)
2324 			continue;
2325 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
2326 			continue; /* XXX: is there any case to allow anycast? */
2327 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
2328 			continue; /* don't use this interface */
2329 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
2330 			continue;
2331 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2332 			if (ip6_use_deprecated)
2333 				dep[0] = (struct in6_ifaddr *)ifa;
2334 			continue;
2335 		}
2336 
2337 		if (dst_scope == in6_addrscope(IFA_IN6(ifa))) {
2338 			/*
2339 			 * call in6_matchlen() as few as possible
2340 			 */
2341 			if (besta) {
2342 				if (blen == -1)
2343 					blen = in6_matchlen(&besta->ia_addr.sin6_addr, dst);
2344 				tlen = in6_matchlen(IFA_IN6(ifa), dst);
2345 				if (tlen > blen) {
2346 					blen = tlen;
2347 					besta = (struct in6_ifaddr *)ifa;
2348 				}
2349 			} else
2350 				besta = (struct in6_ifaddr *)ifa;
2351 		}
2352 	}
2353 	if (besta)
2354 		return (besta);
2355 
2356 	TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
2357 		struct ifaddr *ifa = ifac->ifa;
2358 
2359 		if (ifa->ifa_addr->sa_family != AF_INET6)
2360 			continue;
2361 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
2362 			continue; /* XXX: is there any case to allow anycast? */
2363 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
2364 			continue; /* don't use this interface */
2365 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
2366 			continue;
2367 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2368 			if (ip6_use_deprecated)
2369 				dep[1] = (struct in6_ifaddr *)ifa;
2370 			continue;
2371 		}
2372 
2373 		return (struct in6_ifaddr *)ifa;
2374 	}
2375 
2376 	/* use the last-resort values, that are, deprecated addresses */
2377 	if (dep[0])
2378 		return dep[0];
2379 	if (dep[1])
2380 		return dep[1];
2381 
2382 	return NULL;
2383 }
2384 
2385 /*
2386  * perform DAD when interface becomes IFF_UP.
2387  */
2388 static void
2389 in6_if_up_dispatch(netmsg_t nmsg)
2390 {
2391 	struct ifnet *ifp = nmsg->lmsg.u.ms_resultp;
2392 	struct ifaddr_container *ifac;
2393 	struct in6_ifaddr *ia;
2394 	int dad_delay;		/* delay ticks before DAD output */
2395 
2396 	ASSERT_NETISR0;
2397 
2398 	/*
2399 	 * special cases, like 6to4, are handled in in6_ifattach
2400 	 */
2401 	in6_ifattach(ifp, NULL);
2402 
2403 	dad_delay = 0;
2404 	TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
2405 		struct ifaddr *ifa = ifac->ifa;
2406 
2407 		if (ifa->ifa_addr->sa_family != AF_INET6)
2408 			continue;
2409 		ia = (struct in6_ifaddr *)ifa;
2410 		if (ia->ia6_flags & IN6_IFF_TENTATIVE)
2411 			nd6_dad_start(ifa, &dad_delay);
2412 	}
2413 
2414 	netisr_replymsg(&nmsg->base, 0);
2415 }
2416 
2417 void
2418 in6_if_up(struct ifnet *ifp)
2419 {
2420 	struct netmsg_base nmsg;
2421 
2422 	netmsg_init(&nmsg, NULL, &curthread->td_msgport, 0, in6_if_up_dispatch);
2423 	nmsg.lmsg.u.ms_resultp = ifp;
2424 	netisr_domsg(&nmsg, 0);
2425 }
2426 
2427 int
2428 in6if_do_dad(struct ifnet *ifp)
2429 {
2430 	if (ifp->if_flags & IFF_LOOPBACK)
2431 		return (0);
2432 
2433 	switch (ifp->if_type) {
2434 #ifdef IFT_DUMMY
2435 	case IFT_DUMMY:
2436 #endif
2437 	case IFT_FAITH:
2438 		/*
2439 		 * These interfaces do not have the IFF_LOOPBACK flag,
2440 		 * but loop packets back.  We do not have to do DAD on such
2441 		 * interfaces.  We should even omit it, because loop-backed
2442 		 * NS would confuse the DAD procedure.
2443 		 */
2444 		return (0);
2445 	default:
2446 		/*
2447 		 * Our DAD routine requires the interface up and running.
2448 		 * However, some interfaces can be up before the RUNNING
2449 		 * status.  Additionaly, users may try to assign addresses
2450 		 * before the interface becomes up (or running).
2451 		 * We simply skip DAD in such a case as a work around.
2452 		 * XXX: we should rather mark "tentative" on such addresses,
2453 		 * and do DAD after the interface becomes ready.
2454 		 */
2455 		if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) !=
2456 		    (IFF_UP|IFF_RUNNING))
2457 			return (0);
2458 
2459 		return (1);
2460 	}
2461 }
2462 
2463 /*
2464  * Calculate max IPv6 MTU through all the interfaces and store it
2465  * to in6_maxmtu.
2466  */
2467 void
2468 in6_setmaxmtu(void)
2469 {
2470 	unsigned long maxmtu = 0;
2471 	const struct ifnet_array *arr;
2472 	int i;
2473 
2474 	ASSERT_NETISR0;
2475 
2476 	arr = ifnet_array_get();
2477 	for (i = 0; i < arr->ifnet_count; ++i) {
2478 		struct ifnet *ifp = arr->ifnet_arr[i];
2479 
2480 		/* this function can be called during ifnet initialization */
2481 		if (ifp->if_afdata[AF_INET6] == NULL)
2482 			continue;
2483 		if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
2484 		    IN6_LINKMTU(ifp) > maxmtu)
2485 			maxmtu = IN6_LINKMTU(ifp);
2486 	}
2487 	if (maxmtu)	/* update only when maxmtu is positive */
2488 		in6_maxmtu = maxmtu;
2489 }
2490 
2491 void *
2492 in6_domifattach(struct ifnet *ifp)
2493 {
2494 	struct in6_ifextra *ext;
2495 
2496 	ext = (struct in6_ifextra *)kmalloc(sizeof(*ext), M_IFADDR, M_WAITOK);
2497 	bzero(ext, sizeof(*ext));
2498 
2499 	ext->in6_ifstat = (struct in6_ifstat *)kmalloc(sizeof(struct in6_ifstat),
2500 		M_IFADDR, M_WAITOK);
2501 	bzero(ext->in6_ifstat, sizeof(*ext->in6_ifstat));
2502 
2503 	ext->icmp6_ifstat =
2504 		(struct icmp6_ifstat *)kmalloc(sizeof(struct icmp6_ifstat),
2505 			M_IFADDR, M_WAITOK);
2506 	bzero(ext->icmp6_ifstat, sizeof(*ext->icmp6_ifstat));
2507 
2508 	ext->nd_ifinfo = nd6_ifattach(ifp);
2509 	ext->scope6_id = scope6_ifattach(ifp);
2510 	return ext;
2511 }
2512 
2513 void
2514 in6_domifdetach(struct ifnet *ifp, void *aux)
2515 {
2516 	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
2517 	scope6_ifdetach(ext->scope6_id);
2518 	nd6_ifdetach(ext->nd_ifinfo);
2519 	kfree(ext->in6_ifstat, M_IFADDR);
2520 	kfree(ext->icmp6_ifstat, M_IFADDR);
2521 	kfree(ext, M_IFADDR);
2522 }
2523