xref: /freebsd/sys/netinet6/in6.c (revision a02aba5f3c73d7ed377f88327fedd11f70f23353)
1 /*-
2  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. Neither the name of the project nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	$KAME: in6.c,v 1.259 2002/01/21 11:37:50 keiichi Exp $
30  */
31 
32 /*-
33  * Copyright (c) 1982, 1986, 1991, 1993
34  *	The Regents of the University of California.  All rights reserved.
35  *
36  * Redistribution and use in source and binary forms, with or without
37  * modification, are permitted provided that the following conditions
38  * are met:
39  * 1. Redistributions of source code must retain the above copyright
40  *    notice, this list of conditions and the following disclaimer.
41  * 2. Redistributions in binary form must reproduce the above copyright
42  *    notice, this list of conditions and the following disclaimer in the
43  *    documentation and/or other materials provided with the distribution.
44  * 4. Neither the name of the University nor the names of its contributors
45  *    may be used to endorse or promote products derived from this software
46  *    without specific prior written permission.
47  *
48  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58  * SUCH DAMAGE.
59  *
60  *	@(#)in.c	8.2 (Berkeley) 11/15/93
61  */
62 
63 #include <sys/cdefs.h>
64 __FBSDID("$FreeBSD$");
65 
66 #include "opt_compat.h"
67 #include "opt_inet.h"
68 #include "opt_inet6.h"
69 
70 #include <sys/param.h>
71 #include <sys/errno.h>
72 #include <sys/jail.h>
73 #include <sys/malloc.h>
74 #include <sys/socket.h>
75 #include <sys/socketvar.h>
76 #include <sys/sockio.h>
77 #include <sys/systm.h>
78 #include <sys/priv.h>
79 #include <sys/proc.h>
80 #include <sys/time.h>
81 #include <sys/kernel.h>
82 #include <sys/syslog.h>
83 
84 #include <net/if.h>
85 #include <net/if_var.h>
86 #include <net/if_types.h>
87 #include <net/route.h>
88 #include <net/if_dl.h>
89 #include <net/vnet.h>
90 
91 #include <netinet/in.h>
92 #include <netinet/in_var.h>
93 #include <net/if_llatbl.h>
94 #include <netinet/if_ether.h>
95 #include <netinet/in_systm.h>
96 #include <netinet/ip.h>
97 #include <netinet/in_pcb.h>
98 
99 #include <netinet/ip6.h>
100 #include <netinet6/ip6_var.h>
101 #include <netinet6/nd6.h>
102 #include <netinet6/mld6_var.h>
103 #include <netinet6/ip6_mroute.h>
104 #include <netinet6/in6_ifattach.h>
105 #include <netinet6/scope6_var.h>
106 #include <netinet6/in6_pcb.h>
107 
108 /*
109  * Definitions of some costant IP6 addresses.
110  */
111 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
112 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
113 const struct in6_addr in6addr_nodelocal_allnodes =
114 	IN6ADDR_NODELOCAL_ALLNODES_INIT;
115 const struct in6_addr in6addr_linklocal_allnodes =
116 	IN6ADDR_LINKLOCAL_ALLNODES_INIT;
117 const struct in6_addr in6addr_linklocal_allrouters =
118 	IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
119 const struct in6_addr in6addr_linklocal_allv2routers =
120 	IN6ADDR_LINKLOCAL_ALLV2ROUTERS_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 =
129 	{ sizeof(sa6_any), AF_INET6, 0, 0, IN6ADDR_ANY_INIT, 0 };
130 
131 static int in6_lifaddr_ioctl __P((struct socket *, u_long, caddr_t,
132 	struct ifnet *, struct thread *));
133 static int in6_ifinit __P((struct ifnet *, struct in6_ifaddr *,
134 	struct sockaddr_in6 *, int));
135 static void in6_unlink_ifa(struct in6_ifaddr *, struct ifnet *);
136 
137 int	(*faithprefix_p)(struct in6_addr *);
138 
139 
140 
141 int
142 in6_mask2len(struct in6_addr *mask, u_char *lim0)
143 {
144 	int x = 0, y;
145 	u_char *lim = lim0, *p;
146 
147 	/* ignore the scope_id part */
148 	if (lim0 == NULL || lim0 - (u_char *)mask > sizeof(*mask))
149 		lim = (u_char *)mask + sizeof(*mask);
150 	for (p = (u_char *)mask; p < lim; x++, p++) {
151 		if (*p != 0xff)
152 			break;
153 	}
154 	y = 0;
155 	if (p < lim) {
156 		for (y = 0; y < 8; y++) {
157 			if ((*p & (0x80 >> y)) == 0)
158 				break;
159 		}
160 	}
161 
162 	/*
163 	 * when the limit pointer is given, do a stricter check on the
164 	 * remaining bits.
165 	 */
166 	if (p < lim) {
167 		if (y != 0 && (*p & (0x00ff >> y)) != 0)
168 			return (-1);
169 		for (p = p + 1; p < lim; p++)
170 			if (*p != 0)
171 				return (-1);
172 	}
173 
174 	return x * 8 + y;
175 }
176 
177 #define ifa2ia6(ifa)	((struct in6_ifaddr *)(ifa))
178 #define ia62ifa(ia6)	(&((ia6)->ia_ifa))
179 
180 #ifdef COMPAT_FREEBSD32
181 struct in6_ndifreq32 {
182         char ifname[IFNAMSIZ];
183         uint32_t ifindex;
184 };
185 #define	SIOCGDEFIFACE32_IN6     _IOWR('i', 86, struct in6_ndifreq32)
186 #endif
187 
188 int
189 in6_control(struct socket *so, u_long cmd, caddr_t data,
190     struct ifnet *ifp, struct thread *td)
191 {
192 	struct	in6_ifreq *ifr = (struct in6_ifreq *)data;
193 	struct	in6_ifaddr *ia = NULL;
194 	struct	in6_aliasreq *ifra = (struct in6_aliasreq *)data;
195 	struct sockaddr_in6 *sa6;
196 	int error;
197 
198 	switch (cmd) {
199 	case SIOCGETSGCNT_IN6:
200 	case SIOCGETMIFCNT_IN6:
201 		return (mrt6_ioctl ? mrt6_ioctl(cmd, data) : EOPNOTSUPP);
202 	}
203 
204 	switch(cmd) {
205 	case SIOCAADDRCTL_POLICY:
206 	case SIOCDADDRCTL_POLICY:
207 		if (td != NULL) {
208 			error = priv_check(td, PRIV_NETINET_ADDRCTRL6);
209 			if (error)
210 				return (error);
211 		}
212 		return (in6_src_ioctl(cmd, data));
213 	}
214 
215 	if (ifp == NULL)
216 		return (EOPNOTSUPP);
217 
218 	switch (cmd) {
219 	case SIOCSNDFLUSH_IN6:
220 	case SIOCSPFXFLUSH_IN6:
221 	case SIOCSRTRFLUSH_IN6:
222 	case SIOCSDEFIFACE_IN6:
223 	case SIOCSIFINFO_FLAGS:
224 	case SIOCSIFINFO_IN6:
225 		if (td != NULL) {
226 			error = priv_check(td, PRIV_NETINET_ND6);
227 			if (error)
228 				return (error);
229 		}
230 		/* FALLTHROUGH */
231 	case OSIOCGIFINFO_IN6:
232 	case SIOCGIFINFO_IN6:
233 	case SIOCGDRLST_IN6:
234 	case SIOCGPRLST_IN6:
235 	case SIOCGNBRINFO_IN6:
236 	case SIOCGDEFIFACE_IN6:
237 		return (nd6_ioctl(cmd, data, ifp));
238 
239 #ifdef COMPAT_FREEBSD32
240 	case SIOCGDEFIFACE32_IN6:
241 		{
242 			struct in6_ndifreq ndif;
243 			struct in6_ndifreq32 *ndif32;
244 
245 			error = nd6_ioctl(SIOCGDEFIFACE_IN6, (caddr_t)&ndif,
246 			    ifp);
247 			if (error)
248 				return (error);
249 			ndif32 = (struct in6_ndifreq32 *)data;
250 			ndif32->ifindex = ndif.ifindex;
251 			return (0);
252 		}
253 #endif
254 	}
255 
256 	switch (cmd) {
257 	case SIOCSIFPREFIX_IN6:
258 	case SIOCDIFPREFIX_IN6:
259 	case SIOCAIFPREFIX_IN6:
260 	case SIOCCIFPREFIX_IN6:
261 	case SIOCSGIFPREFIX_IN6:
262 	case SIOCGIFPREFIX_IN6:
263 		log(LOG_NOTICE,
264 		    "prefix ioctls are now invalidated. "
265 		    "please use ifconfig.\n");
266 		return (EOPNOTSUPP);
267 	}
268 
269 	switch (cmd) {
270 	case SIOCSSCOPE6:
271 		if (td != NULL) {
272 			error = priv_check(td, PRIV_NETINET_SCOPE6);
273 			if (error)
274 				return (error);
275 		}
276 		return (scope6_set(ifp,
277 		    (struct scope6_id *)ifr->ifr_ifru.ifru_scope_id));
278 	case SIOCGSCOPE6:
279 		return (scope6_get(ifp,
280 		    (struct scope6_id *)ifr->ifr_ifru.ifru_scope_id));
281 	case SIOCGSCOPE6DEF:
282 		return (scope6_get_default((struct scope6_id *)
283 		    ifr->ifr_ifru.ifru_scope_id));
284 	}
285 
286 	switch (cmd) {
287 	case SIOCALIFADDR:
288 		if (td != NULL) {
289 			error = priv_check(td, PRIV_NET_ADDIFADDR);
290 			if (error)
291 				return (error);
292 		}
293 		return in6_lifaddr_ioctl(so, cmd, data, ifp, td);
294 
295 	case SIOCDLIFADDR:
296 		if (td != NULL) {
297 			error = priv_check(td, PRIV_NET_DELIFADDR);
298 			if (error)
299 				return (error);
300 		}
301 		/* FALLTHROUGH */
302 	case SIOCGLIFADDR:
303 		return in6_lifaddr_ioctl(so, cmd, data, ifp, td);
304 	}
305 
306 	/*
307 	 * Find address for this interface, if it exists.
308 	 *
309 	 * In netinet code, we have checked ifra_addr in SIOCSIF*ADDR operation
310 	 * only, and used the first interface address as the target of other
311 	 * operations (without checking ifra_addr).  This was because netinet
312 	 * code/API assumed at most 1 interface address per interface.
313 	 * Since IPv6 allows a node to assign multiple addresses
314 	 * on a single interface, we almost always look and check the
315 	 * presence of ifra_addr, and reject invalid ones here.
316 	 * It also decreases duplicated code among SIOC*_IN6 operations.
317 	 */
318 	switch (cmd) {
319 	case SIOCAIFADDR_IN6:
320 	case SIOCSIFPHYADDR_IN6:
321 		sa6 = &ifra->ifra_addr;
322 		break;
323 	case SIOCSIFADDR_IN6:
324 	case SIOCGIFADDR_IN6:
325 	case SIOCSIFDSTADDR_IN6:
326 	case SIOCSIFNETMASK_IN6:
327 	case SIOCGIFDSTADDR_IN6:
328 	case SIOCGIFNETMASK_IN6:
329 	case SIOCDIFADDR_IN6:
330 	case SIOCGIFPSRCADDR_IN6:
331 	case SIOCGIFPDSTADDR_IN6:
332 	case SIOCGIFAFLAG_IN6:
333 	case SIOCSNDFLUSH_IN6:
334 	case SIOCSPFXFLUSH_IN6:
335 	case SIOCSRTRFLUSH_IN6:
336 	case SIOCGIFALIFETIME_IN6:
337 	case SIOCSIFALIFETIME_IN6:
338 	case SIOCGIFSTAT_IN6:
339 	case SIOCGIFSTAT_ICMP6:
340 		sa6 = &ifr->ifr_addr;
341 		break;
342 	default:
343 		sa6 = NULL;
344 		break;
345 	}
346 	if (sa6 && sa6->sin6_family == AF_INET6) {
347 		if (sa6->sin6_scope_id != 0)
348 			error = sa6_embedscope(sa6, 0);
349 		else
350 			error = in6_setscope(&sa6->sin6_addr, ifp, NULL);
351 		if (error != 0)
352 			return (error);
353 		if (td != NULL && (error = prison_check_ip6(td->td_ucred,
354 		    &sa6->sin6_addr)) != 0)
355 			return (error);
356 		ia = in6ifa_ifpwithaddr(ifp, &sa6->sin6_addr);
357 	} else
358 		ia = NULL;
359 
360 	switch (cmd) {
361 	case SIOCSIFADDR_IN6:
362 	case SIOCSIFDSTADDR_IN6:
363 	case SIOCSIFNETMASK_IN6:
364 		/*
365 		 * Since IPv6 allows a node to assign multiple addresses
366 		 * on a single interface, SIOCSIFxxx ioctls are deprecated.
367 		 */
368 		/* we decided to obsolete this command (20000704) */
369 		error = EINVAL;
370 		goto out;
371 
372 	case SIOCDIFADDR_IN6:
373 		/*
374 		 * for IPv4, we look for existing in_ifaddr here to allow
375 		 * "ifconfig if0 delete" to remove the first IPv4 address on
376 		 * the interface.  For IPv6, as the spec allows multiple
377 		 * interface address from the day one, we consider "remove the
378 		 * first one" semantics to be not preferable.
379 		 */
380 		if (ia == NULL) {
381 			error = EADDRNOTAVAIL;
382 			goto out;
383 		}
384 		/* FALLTHROUGH */
385 	case SIOCAIFADDR_IN6:
386 		/*
387 		 * We always require users to specify a valid IPv6 address for
388 		 * the corresponding operation.
389 		 */
390 		if (ifra->ifra_addr.sin6_family != AF_INET6 ||
391 		    ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6)) {
392 			error = EAFNOSUPPORT;
393 			goto out;
394 		}
395 
396 		if (td != NULL) {
397 			error = priv_check(td, (cmd == SIOCDIFADDR_IN6) ?
398 			    PRIV_NET_DELIFADDR : PRIV_NET_ADDIFADDR);
399 			if (error)
400 				goto out;
401 		}
402 		break;
403 
404 	case SIOCGIFADDR_IN6:
405 		/* This interface is basically deprecated. use SIOCGIFCONF. */
406 		/* FALLTHROUGH */
407 	case SIOCGIFAFLAG_IN6:
408 	case SIOCGIFNETMASK_IN6:
409 	case SIOCGIFDSTADDR_IN6:
410 	case SIOCGIFALIFETIME_IN6:
411 		/* must think again about its semantics */
412 		if (ia == NULL) {
413 			error = EADDRNOTAVAIL;
414 			goto out;
415 		}
416 		break;
417 
418 	case SIOCSIFALIFETIME_IN6:
419 	    {
420 		struct in6_addrlifetime *lt;
421 
422 		if (td != NULL) {
423 			error = priv_check(td, PRIV_NETINET_ALIFETIME6);
424 			if (error)
425 				goto out;
426 		}
427 		if (ia == NULL) {
428 			error = EADDRNOTAVAIL;
429 			goto out;
430 		}
431 		/* sanity for overflow - beware unsigned */
432 		lt = &ifr->ifr_ifru.ifru_lifetime;
433 		if (lt->ia6t_vltime != ND6_INFINITE_LIFETIME &&
434 		    lt->ia6t_vltime + time_second < time_second) {
435 			error = EINVAL;
436 			goto out;
437 		}
438 		if (lt->ia6t_pltime != ND6_INFINITE_LIFETIME &&
439 		    lt->ia6t_pltime + time_second < time_second) {
440 			error = EINVAL;
441 			goto out;
442 		}
443 		break;
444 	    }
445 	}
446 
447 	switch (cmd) {
448 	case SIOCGIFADDR_IN6:
449 		ifr->ifr_addr = ia->ia_addr;
450 		if ((error = sa6_recoverscope(&ifr->ifr_addr)) != 0)
451 			goto out;
452 		break;
453 
454 	case SIOCGIFDSTADDR_IN6:
455 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
456 			error = EINVAL;
457 			goto out;
458 		}
459 		/*
460 		 * XXX: should we check if ifa_dstaddr is NULL and return
461 		 * an error?
462 		 */
463 		ifr->ifr_dstaddr = ia->ia_dstaddr;
464 		if ((error = sa6_recoverscope(&ifr->ifr_dstaddr)) != 0)
465 			goto out;
466 		break;
467 
468 	case SIOCGIFNETMASK_IN6:
469 		ifr->ifr_addr = ia->ia_prefixmask;
470 		break;
471 
472 	case SIOCGIFAFLAG_IN6:
473 		ifr->ifr_ifru.ifru_flags6 = ia->ia6_flags;
474 		break;
475 
476 	case SIOCGIFSTAT_IN6:
477 		if (ifp == NULL) {
478 			error = EINVAL;
479 			goto out;
480 		}
481 		bzero(&ifr->ifr_ifru.ifru_stat,
482 		    sizeof(ifr->ifr_ifru.ifru_stat));
483 		ifr->ifr_ifru.ifru_stat =
484 		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->in6_ifstat;
485 		break;
486 
487 	case SIOCGIFSTAT_ICMP6:
488 		if (ifp == NULL) {
489 			error = EINVAL;
490 			goto out;
491 		}
492 		bzero(&ifr->ifr_ifru.ifru_icmp6stat,
493 		    sizeof(ifr->ifr_ifru.ifru_icmp6stat));
494 		ifr->ifr_ifru.ifru_icmp6stat =
495 		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->icmp6_ifstat;
496 		break;
497 
498 	case SIOCGIFALIFETIME_IN6:
499 		ifr->ifr_ifru.ifru_lifetime = ia->ia6_lifetime;
500 		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
501 			time_t maxexpire;
502 			struct in6_addrlifetime *retlt =
503 			    &ifr->ifr_ifru.ifru_lifetime;
504 
505 			/*
506 			 * XXX: adjust expiration time assuming time_t is
507 			 * signed.
508 			 */
509 			maxexpire = (-1) &
510 			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
511 			if (ia->ia6_lifetime.ia6t_vltime <
512 			    maxexpire - ia->ia6_updatetime) {
513 				retlt->ia6t_expire = ia->ia6_updatetime +
514 				    ia->ia6_lifetime.ia6t_vltime;
515 			} else
516 				retlt->ia6t_expire = maxexpire;
517 		}
518 		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
519 			time_t maxexpire;
520 			struct in6_addrlifetime *retlt =
521 			    &ifr->ifr_ifru.ifru_lifetime;
522 
523 			/*
524 			 * XXX: adjust expiration time assuming time_t is
525 			 * signed.
526 			 */
527 			maxexpire = (-1) &
528 			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
529 			if (ia->ia6_lifetime.ia6t_pltime <
530 			    maxexpire - ia->ia6_updatetime) {
531 				retlt->ia6t_preferred = ia->ia6_updatetime +
532 				    ia->ia6_lifetime.ia6t_pltime;
533 			} else
534 				retlt->ia6t_preferred = maxexpire;
535 		}
536 		break;
537 
538 	case SIOCSIFALIFETIME_IN6:
539 		ia->ia6_lifetime = ifr->ifr_ifru.ifru_lifetime;
540 		/* for sanity */
541 		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
542 			ia->ia6_lifetime.ia6t_expire =
543 				time_second + ia->ia6_lifetime.ia6t_vltime;
544 		} else
545 			ia->ia6_lifetime.ia6t_expire = 0;
546 		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
547 			ia->ia6_lifetime.ia6t_preferred =
548 				time_second + ia->ia6_lifetime.ia6t_pltime;
549 		} else
550 			ia->ia6_lifetime.ia6t_preferred = 0;
551 		break;
552 
553 	case SIOCAIFADDR_IN6:
554 	{
555 		int i;
556 		struct nd_prefixctl pr0;
557 		struct nd_prefix *pr;
558 
559 		/*
560 		 * first, make or update the interface address structure,
561 		 * and link it to the list.
562 		 */
563 		if ((error = in6_update_ifa(ifp, ifra, ia, 0)) != 0)
564 			goto out;
565 		if (ia != NULL)
566 			ifa_free(&ia->ia_ifa);
567 		if ((ia = in6ifa_ifpwithaddr(ifp, &ifra->ifra_addr.sin6_addr))
568 		    == NULL) {
569 			/*
570 			 * this can happen when the user specify the 0 valid
571 			 * lifetime.
572 			 */
573 			break;
574 		}
575 
576 		/*
577 		 * then, make the prefix on-link on the interface.
578 		 * XXX: we'd rather create the prefix before the address, but
579 		 * we need at least one address to install the corresponding
580 		 * interface route, so we configure the address first.
581 		 */
582 
583 		/*
584 		 * convert mask to prefix length (prefixmask has already
585 		 * been validated in in6_update_ifa().
586 		 */
587 		bzero(&pr0, sizeof(pr0));
588 		pr0.ndpr_ifp = ifp;
589 		pr0.ndpr_plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
590 		    NULL);
591 		if (pr0.ndpr_plen == 128) {
592 			break;	/* we don't need to install a host route. */
593 		}
594 		pr0.ndpr_prefix = ifra->ifra_addr;
595 		/* apply the mask for safety. */
596 		for (i = 0; i < 4; i++) {
597 			pr0.ndpr_prefix.sin6_addr.s6_addr32[i] &=
598 			    ifra->ifra_prefixmask.sin6_addr.s6_addr32[i];
599 		}
600 		/*
601 		 * XXX: since we don't have an API to set prefix (not address)
602 		 * lifetimes, we just use the same lifetimes as addresses.
603 		 * The (temporarily) installed lifetimes can be overridden by
604 		 * later advertised RAs (when accept_rtadv is non 0), which is
605 		 * an intended behavior.
606 		 */
607 		pr0.ndpr_raf_onlink = 1; /* should be configurable? */
608 		pr0.ndpr_raf_auto =
609 		    ((ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0);
610 		pr0.ndpr_vltime = ifra->ifra_lifetime.ia6t_vltime;
611 		pr0.ndpr_pltime = ifra->ifra_lifetime.ia6t_pltime;
612 
613 		/* add the prefix if not yet. */
614 		if ((pr = nd6_prefix_lookup(&pr0)) == NULL) {
615 			/*
616 			 * nd6_prelist_add will install the corresponding
617 			 * interface route.
618 			 */
619 			if ((error = nd6_prelist_add(&pr0, NULL, &pr)) != 0)
620 				goto out;
621 			if (pr == NULL) {
622 				log(LOG_ERR, "nd6_prelist_add succeeded but "
623 				    "no prefix\n");
624 				error = EINVAL;
625 				goto out;
626 			}
627 		}
628 
629 		/* relate the address to the prefix */
630 		if (ia->ia6_ndpr == NULL) {
631 			ia->ia6_ndpr = pr;
632 			pr->ndpr_refcnt++;
633 
634 			/*
635 			 * If this is the first autoconf address from the
636 			 * prefix, create a temporary address as well
637 			 * (when required).
638 			 */
639 			if ((ia->ia6_flags & IN6_IFF_AUTOCONF) &&
640 			    V_ip6_use_tempaddr && pr->ndpr_refcnt == 1) {
641 				int e;
642 				if ((e = in6_tmpifadd(ia, 1, 0)) != 0) {
643 					log(LOG_NOTICE, "in6_control: failed "
644 					    "to create a temporary address, "
645 					    "errno=%d\n", e);
646 				}
647 			}
648 		}
649 
650 		/*
651 		 * this might affect the status of autoconfigured addresses,
652 		 * that is, this address might make other addresses detached.
653 		 */
654 		pfxlist_onlink_check();
655 		if (error == 0 && ia) {
656 			if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) {
657 				/*
658 				 * Try to clear the flag when a new
659 				 * IPv6 address is added onto an
660 				 * IFDISABLED interface and it
661 				 * succeeds.
662 				 */
663 				struct in6_ndireq nd;
664 
665 				memset(&nd, 0, sizeof(nd));
666 				nd.ndi.flags = ND_IFINFO(ifp)->flags;
667 				nd.ndi.flags &= ~ND6_IFF_IFDISABLED;
668 				if (nd6_ioctl(SIOCSIFINFO_FLAGS,
669 				    (caddr_t)&nd, ifp) < 0)
670 					log(LOG_NOTICE, "SIOCAIFADDR_IN6: "
671 					    "SIOCSIFINFO_FLAGS for -ifdisabled "
672 					    "failed.");
673 				/*
674 				 * Ignore failure of clearing the flag
675 				 * intentionally.  The failure means
676 				 * address duplication was detected.
677 				 */
678 			}
679 			EVENTHANDLER_INVOKE(ifaddr_event, ifp);
680 		}
681 		break;
682 	}
683 
684 	case SIOCDIFADDR_IN6:
685 	{
686 		struct nd_prefix *pr;
687 
688 		/*
689 		 * If the address being deleted is the only one that owns
690 		 * the corresponding prefix, expire the prefix as well.
691 		 * XXX: theoretically, we don't have to worry about such
692 		 * relationship, since we separate the address management
693 		 * and the prefix management.  We do this, however, to provide
694 		 * as much backward compatibility as possible in terms of
695 		 * the ioctl operation.
696 		 * Note that in6_purgeaddr() will decrement ndpr_refcnt.
697 		 */
698 		pr = ia->ia6_ndpr;
699 		in6_purgeaddr(&ia->ia_ifa);
700 		if (pr && pr->ndpr_refcnt == 0)
701 			prelist_remove(pr);
702 		EVENTHANDLER_INVOKE(ifaddr_event, ifp);
703 		break;
704 	}
705 
706 	default:
707 		if (ifp == NULL || ifp->if_ioctl == 0) {
708 			error = EOPNOTSUPP;
709 			goto out;
710 		}
711 		error = (*ifp->if_ioctl)(ifp, cmd, data);
712 		goto out;
713 	}
714 
715 	error = 0;
716 out:
717 	if (ia != NULL)
718 		ifa_free(&ia->ia_ifa);
719 	return (error);
720 }
721 
722 /*
723  * Update parameters of an IPv6 interface address.
724  * If necessary, a new entry is created and linked into address chains.
725  * This function is separated from in6_control().
726  * XXX: should this be performed under splnet()?
727  */
728 int
729 in6_update_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
730     struct in6_ifaddr *ia, int flags)
731 {
732 	int error = 0, hostIsNew = 0, plen = -1;
733 	struct sockaddr_in6 dst6;
734 	struct in6_addrlifetime *lt;
735 	struct in6_multi_mship *imm;
736 	struct in6_multi *in6m_sol;
737 	struct rtentry *rt;
738 	int delay;
739 	char ip6buf[INET6_ADDRSTRLEN];
740 
741 	/* Validate parameters */
742 	if (ifp == NULL || ifra == NULL) /* this maybe redundant */
743 		return (EINVAL);
744 
745 	/*
746 	 * The destination address for a p2p link must have a family
747 	 * of AF_UNSPEC or AF_INET6.
748 	 */
749 	if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
750 	    ifra->ifra_dstaddr.sin6_family != AF_INET6 &&
751 	    ifra->ifra_dstaddr.sin6_family != AF_UNSPEC)
752 		return (EAFNOSUPPORT);
753 	/*
754 	 * validate ifra_prefixmask.  don't check sin6_family, netmask
755 	 * does not carry fields other than sin6_len.
756 	 */
757 	if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6))
758 		return (EINVAL);
759 	/*
760 	 * Because the IPv6 address architecture is classless, we require
761 	 * users to specify a (non 0) prefix length (mask) for a new address.
762 	 * We also require the prefix (when specified) mask is valid, and thus
763 	 * reject a non-consecutive mask.
764 	 */
765 	if (ia == NULL && ifra->ifra_prefixmask.sin6_len == 0)
766 		return (EINVAL);
767 	if (ifra->ifra_prefixmask.sin6_len != 0) {
768 		plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
769 		    (u_char *)&ifra->ifra_prefixmask +
770 		    ifra->ifra_prefixmask.sin6_len);
771 		if (plen <= 0)
772 			return (EINVAL);
773 	} else {
774 		/*
775 		 * In this case, ia must not be NULL.  We just use its prefix
776 		 * length.
777 		 */
778 		plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
779 	}
780 	/*
781 	 * If the destination address on a p2p interface is specified,
782 	 * and the address is a scoped one, validate/set the scope
783 	 * zone identifier.
784 	 */
785 	dst6 = ifra->ifra_dstaddr;
786 	if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) != 0 &&
787 	    (dst6.sin6_family == AF_INET6)) {
788 		struct in6_addr in6_tmp;
789 		u_int32_t zoneid;
790 
791 		in6_tmp = dst6.sin6_addr;
792 		if (in6_setscope(&in6_tmp, ifp, &zoneid))
793 			return (EINVAL); /* XXX: should be impossible */
794 
795 		if (dst6.sin6_scope_id != 0) {
796 			if (dst6.sin6_scope_id != zoneid)
797 				return (EINVAL);
798 		} else		/* user omit to specify the ID. */
799 			dst6.sin6_scope_id = zoneid;
800 
801 		/* convert into the internal form */
802 		if (sa6_embedscope(&dst6, 0))
803 			return (EINVAL); /* XXX: should be impossible */
804 	}
805 	/*
806 	 * The destination address can be specified only for a p2p or a
807 	 * loopback interface.  If specified, the corresponding prefix length
808 	 * must be 128.
809 	 */
810 	if (ifra->ifra_dstaddr.sin6_family == AF_INET6) {
811 		if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) == 0) {
812 			/* XXX: noisy message */
813 			nd6log((LOG_INFO, "in6_update_ifa: a destination can "
814 			    "be specified for a p2p or a loopback IF only\n"));
815 			return (EINVAL);
816 		}
817 		if (plen != 128) {
818 			nd6log((LOG_INFO, "in6_update_ifa: prefixlen should "
819 			    "be 128 when dstaddr is specified\n"));
820 			return (EINVAL);
821 		}
822 	}
823 	/* lifetime consistency check */
824 	lt = &ifra->ifra_lifetime;
825 	if (lt->ia6t_pltime > lt->ia6t_vltime)
826 		return (EINVAL);
827 	if (lt->ia6t_vltime == 0) {
828 		/*
829 		 * the following log might be noisy, but this is a typical
830 		 * configuration mistake or a tool's bug.
831 		 */
832 		nd6log((LOG_INFO,
833 		    "in6_update_ifa: valid lifetime is 0 for %s\n",
834 		    ip6_sprintf(ip6buf, &ifra->ifra_addr.sin6_addr)));
835 
836 		if (ia == NULL)
837 			return (0); /* there's nothing to do */
838 	}
839 
840 	/*
841 	 * If this is a new address, allocate a new ifaddr and link it
842 	 * into chains.
843 	 */
844 	if (ia == NULL) {
845 		hostIsNew = 1;
846 		/*
847 		 * When in6_update_ifa() is called in a process of a received
848 		 * RA, it is called under an interrupt context.  So, we should
849 		 * call malloc with M_NOWAIT.
850 		 */
851 		ia = (struct in6_ifaddr *) malloc(sizeof(*ia), M_IFADDR,
852 		    M_NOWAIT);
853 		if (ia == NULL)
854 			return (ENOBUFS);
855 		bzero((caddr_t)ia, sizeof(*ia));
856 		ifa_init(&ia->ia_ifa);
857 		LIST_INIT(&ia->ia6_memberships);
858 		/* Initialize the address and masks, and put time stamp */
859 		ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
860 		ia->ia_addr.sin6_family = AF_INET6;
861 		ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
862 		ia->ia6_createtime = time_second;
863 		if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
864 			/*
865 			 * XXX: some functions expect that ifa_dstaddr is not
866 			 * NULL for p2p interfaces.
867 			 */
868 			ia->ia_ifa.ifa_dstaddr =
869 			    (struct sockaddr *)&ia->ia_dstaddr;
870 		} else {
871 			ia->ia_ifa.ifa_dstaddr = NULL;
872 		}
873 		ia->ia_ifa.ifa_netmask = (struct sockaddr *)&ia->ia_prefixmask;
874 		ia->ia_ifp = ifp;
875 		ifa_ref(&ia->ia_ifa);			/* if_addrhead */
876 		IF_ADDR_LOCK(ifp);
877 		TAILQ_INSERT_TAIL(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
878 		IF_ADDR_UNLOCK(ifp);
879 
880 		ifa_ref(&ia->ia_ifa);			/* in6_ifaddrhead */
881 		IN6_IFADDR_WLOCK();
882 		TAILQ_INSERT_TAIL(&V_in6_ifaddrhead, ia, ia_link);
883 		IN6_IFADDR_WUNLOCK();
884 	}
885 
886 	/* update timestamp */
887 	ia->ia6_updatetime = time_second;
888 
889 	/* set prefix mask */
890 	if (ifra->ifra_prefixmask.sin6_len) {
891 		/*
892 		 * We prohibit changing the prefix length of an existing
893 		 * address, because
894 		 * + such an operation should be rare in IPv6, and
895 		 * + the operation would confuse prefix management.
896 		 */
897 		if (ia->ia_prefixmask.sin6_len &&
898 		    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) != plen) {
899 			nd6log((LOG_INFO, "in6_update_ifa: the prefix length of an"
900 			    " existing (%s) address should not be changed\n",
901 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
902 			error = EINVAL;
903 			goto unlink;
904 		}
905 		ia->ia_prefixmask = ifra->ifra_prefixmask;
906 	}
907 
908 	/*
909 	 * If a new destination address is specified, scrub the old one and
910 	 * install the new destination.  Note that the interface must be
911 	 * p2p or loopback (see the check above.)
912 	 */
913 	if (dst6.sin6_family == AF_INET6 &&
914 	    !IN6_ARE_ADDR_EQUAL(&dst6.sin6_addr, &ia->ia_dstaddr.sin6_addr)) {
915 		int e;
916 
917 		if ((ia->ia_flags & IFA_ROUTE) != 0 &&
918 		    (e = rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST)) != 0) {
919 			nd6log((LOG_ERR, "in6_update_ifa: failed to remove "
920 			    "a route to the old destination: %s\n",
921 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
922 			/* proceed anyway... */
923 		} else
924 			ia->ia_flags &= ~IFA_ROUTE;
925 		ia->ia_dstaddr = dst6;
926 	}
927 
928 	/*
929 	 * Set lifetimes.  We do not refer to ia6t_expire and ia6t_preferred
930 	 * to see if the address is deprecated or invalidated, but initialize
931 	 * these members for applications.
932 	 */
933 	ia->ia6_lifetime = ifra->ifra_lifetime;
934 	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
935 		ia->ia6_lifetime.ia6t_expire =
936 		    time_second + ia->ia6_lifetime.ia6t_vltime;
937 	} else
938 		ia->ia6_lifetime.ia6t_expire = 0;
939 	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
940 		ia->ia6_lifetime.ia6t_preferred =
941 		    time_second + ia->ia6_lifetime.ia6t_pltime;
942 	} else
943 		ia->ia6_lifetime.ia6t_preferred = 0;
944 
945 	/* reset the interface and routing table appropriately. */
946 	if ((error = in6_ifinit(ifp, ia, &ifra->ifra_addr, hostIsNew)) != 0)
947 		goto unlink;
948 
949 	/*
950 	 * configure address flags.
951 	 */
952 	ia->ia6_flags = ifra->ifra_flags;
953 	/*
954 	 * backward compatibility - if IN6_IFF_DEPRECATED is set from the
955 	 * userland, make it deprecated.
956 	 */
957 	if ((ifra->ifra_flags & IN6_IFF_DEPRECATED) != 0) {
958 		ia->ia6_lifetime.ia6t_pltime = 0;
959 		ia->ia6_lifetime.ia6t_preferred = time_second;
960 	}
961 	/*
962 	 * Make the address tentative before joining multicast addresses,
963 	 * so that corresponding MLD responses would not have a tentative
964 	 * source address.
965 	 */
966 	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/* safety */
967 	if (hostIsNew && in6if_do_dad(ifp))
968 		ia->ia6_flags |= IN6_IFF_TENTATIVE;
969 
970 	/* DAD should be performed after ND6_IFF_IFDISABLED is cleared. */
971 	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)
972 		ia->ia6_flags |= IN6_IFF_TENTATIVE;
973 
974 	/*
975 	 * We are done if we have simply modified an existing address.
976 	 */
977 	if (!hostIsNew)
978 		return (error);
979 
980 	/*
981 	 * Beyond this point, we should call in6_purgeaddr upon an error,
982 	 * not just go to unlink.
983 	 */
984 
985 	/* Join necessary multicast groups */
986 	in6m_sol = NULL;
987 	if ((ifp->if_flags & IFF_MULTICAST) != 0) {
988 		struct sockaddr_in6 mltaddr, mltmask;
989 		struct in6_addr llsol;
990 
991 		/* join solicited multicast addr for new host id */
992 		bzero(&llsol, sizeof(struct in6_addr));
993 		llsol.s6_addr32[0] = IPV6_ADDR_INT32_MLL;
994 		llsol.s6_addr32[1] = 0;
995 		llsol.s6_addr32[2] = htonl(1);
996 		llsol.s6_addr32[3] = ifra->ifra_addr.sin6_addr.s6_addr32[3];
997 		llsol.s6_addr8[12] = 0xff;
998 		if ((error = in6_setscope(&llsol, ifp, NULL)) != 0) {
999 			/* XXX: should not happen */
1000 			log(LOG_ERR, "in6_update_ifa: "
1001 			    "in6_setscope failed\n");
1002 			goto cleanup;
1003 		}
1004 		delay = 0;
1005 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1006 			/*
1007 			 * We need a random delay for DAD on the address
1008 			 * being configured.  It also means delaying
1009 			 * transmission of the corresponding MLD report to
1010 			 * avoid report collision.
1011 			 * [RFC 4861, Section 6.3.7]
1012 			 */
1013 			delay = arc4random() %
1014 			    (MAX_RTR_SOLICITATION_DELAY * hz);
1015 		}
1016 		imm = in6_joingroup(ifp, &llsol, &error, delay);
1017 		if (imm == NULL) {
1018 			nd6log((LOG_WARNING,
1019 			    "in6_update_ifa: addmulti failed for "
1020 			    "%s on %s (errno=%d)\n",
1021 			    ip6_sprintf(ip6buf, &llsol), if_name(ifp),
1022 			    error));
1023 			goto cleanup;
1024 		}
1025 		LIST_INSERT_HEAD(&ia->ia6_memberships,
1026 		    imm, i6mm_chain);
1027 		in6m_sol = imm->i6mm_maddr;
1028 
1029 		bzero(&mltmask, sizeof(mltmask));
1030 		mltmask.sin6_len = sizeof(struct sockaddr_in6);
1031 		mltmask.sin6_family = AF_INET6;
1032 		mltmask.sin6_addr = in6mask32;
1033 #define	MLTMASK_LEN  4	/* mltmask's masklen (=32bit=4octet) */
1034 
1035 		/*
1036 		 * join link-local all-nodes address
1037 		 */
1038 		bzero(&mltaddr, sizeof(mltaddr));
1039 		mltaddr.sin6_len = sizeof(struct sockaddr_in6);
1040 		mltaddr.sin6_family = AF_INET6;
1041 		mltaddr.sin6_addr = in6addr_linklocal_allnodes;
1042 		if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) !=
1043 		    0)
1044 			goto cleanup; /* XXX: should not fail */
1045 
1046 		/*
1047 		 * XXX: do we really need this automatic routes?
1048 		 * We should probably reconsider this stuff.  Most applications
1049 		 * actually do not need the routes, since they usually specify
1050 		 * the outgoing interface.
1051 		 */
1052 		rt = rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL);
1053 		if (rt) {
1054 			/* XXX: only works in !SCOPEDROUTING case. */
1055 			if (memcmp(&mltaddr.sin6_addr,
1056 			    &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
1057 			    MLTMASK_LEN)) {
1058 				RTFREE_LOCKED(rt);
1059 				rt = NULL;
1060 			}
1061 		}
1062 		if (!rt) {
1063 			error = rtrequest(RTM_ADD, (struct sockaddr *)&mltaddr,
1064 			    (struct sockaddr *)&ia->ia_addr,
1065 			    (struct sockaddr *)&mltmask, RTF_UP,
1066 			    (struct rtentry **)0);
1067 			if (error)
1068 				goto cleanup;
1069 		} else {
1070 			RTFREE_LOCKED(rt);
1071 		}
1072 
1073 		imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
1074 		if (!imm) {
1075 			nd6log((LOG_WARNING,
1076 			    "in6_update_ifa: addmulti failed for "
1077 			    "%s on %s (errno=%d)\n",
1078 			    ip6_sprintf(ip6buf, &mltaddr.sin6_addr),
1079 			    if_name(ifp), error));
1080 			goto cleanup;
1081 		}
1082 		LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1083 
1084 		/*
1085 		 * join node information group address
1086 		 */
1087 		delay = 0;
1088 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1089 			/*
1090 			 * The spec doesn't say anything about delay for this
1091 			 * group, but the same logic should apply.
1092 			 */
1093 			delay = arc4random() %
1094 			    (MAX_RTR_SOLICITATION_DELAY * hz);
1095 		}
1096 		if (in6_nigroup(ifp, NULL, -1, &mltaddr.sin6_addr) == 0) {
1097 			imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error,
1098 			    delay); /* XXX jinmei */
1099 			if (!imm) {
1100 				nd6log((LOG_WARNING, "in6_update_ifa: "
1101 				    "addmulti failed for %s on %s "
1102 				    "(errno=%d)\n",
1103 				    ip6_sprintf(ip6buf, &mltaddr.sin6_addr),
1104 				    if_name(ifp), error));
1105 				/* XXX not very fatal, go on... */
1106 			} else {
1107 				LIST_INSERT_HEAD(&ia->ia6_memberships,
1108 				    imm, i6mm_chain);
1109 			}
1110 		}
1111 
1112 		/*
1113 		 * join interface-local all-nodes address.
1114 		 * (ff01::1%ifN, and ff01::%ifN/32)
1115 		 */
1116 		mltaddr.sin6_addr = in6addr_nodelocal_allnodes;
1117 		if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL))
1118 		    != 0)
1119 			goto cleanup; /* XXX: should not fail */
1120 		/* XXX: again, do we really need the route? */
1121 		rt = rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL);
1122 		if (rt) {
1123 			if (memcmp(&mltaddr.sin6_addr,
1124 			    &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
1125 			    MLTMASK_LEN)) {
1126 				RTFREE_LOCKED(rt);
1127 				rt = NULL;
1128 			}
1129 		}
1130 		if (!rt) {
1131 			error = rtrequest(RTM_ADD, (struct sockaddr *)&mltaddr,
1132 			    (struct sockaddr *)&ia->ia_addr,
1133 			    (struct sockaddr *)&mltmask, RTF_UP,
1134 			    (struct rtentry **)0);
1135 			if (error)
1136 				goto cleanup;
1137 		} else
1138 			RTFREE_LOCKED(rt);
1139 
1140 		imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
1141 		if (!imm) {
1142 			nd6log((LOG_WARNING, "in6_update_ifa: "
1143 			    "addmulti failed for %s on %s "
1144 			    "(errno=%d)\n",
1145 			    ip6_sprintf(ip6buf, &mltaddr.sin6_addr),
1146 			    if_name(ifp), error));
1147 			goto cleanup;
1148 		}
1149 		LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1150 #undef	MLTMASK_LEN
1151 	}
1152 
1153 	/*
1154 	 * Perform DAD, if needed.
1155 	 * XXX It may be of use, if we can administratively
1156 	 * disable DAD.
1157 	 */
1158 	if (in6if_do_dad(ifp) && ((ifra->ifra_flags & IN6_IFF_NODAD) == 0) &&
1159 	    (ia->ia6_flags & IN6_IFF_TENTATIVE))
1160 	{
1161 		int mindelay, maxdelay;
1162 
1163 		delay = 0;
1164 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1165 			/*
1166 			 * We need to impose a delay before sending an NS
1167 			 * for DAD.  Check if we also needed a delay for the
1168 			 * corresponding MLD message.  If we did, the delay
1169 			 * should be larger than the MLD delay (this could be
1170 			 * relaxed a bit, but this simple logic is at least
1171 			 * safe).
1172 			 * XXX: Break data hiding guidelines and look at
1173 			 * state for the solicited multicast group.
1174 			 */
1175 			mindelay = 0;
1176 			if (in6m_sol != NULL &&
1177 			    in6m_sol->in6m_state == MLD_REPORTING_MEMBER) {
1178 				mindelay = in6m_sol->in6m_timer;
1179 			}
1180 			maxdelay = MAX_RTR_SOLICITATION_DELAY * hz;
1181 			if (maxdelay - mindelay == 0)
1182 				delay = 0;
1183 			else {
1184 				delay =
1185 				    (arc4random() % (maxdelay - mindelay)) +
1186 				    mindelay;
1187 			}
1188 		}
1189 		nd6_dad_start((struct ifaddr *)ia, delay);
1190 	}
1191 
1192 	KASSERT(hostIsNew, ("in6_update_ifa: !hostIsNew"));
1193 	ifa_free(&ia->ia_ifa);
1194 	return (error);
1195 
1196   unlink:
1197 	/*
1198 	 * XXX: if a change of an existing address failed, keep the entry
1199 	 * anyway.
1200 	 */
1201 	if (hostIsNew) {
1202 		in6_unlink_ifa(ia, ifp);
1203 		ifa_free(&ia->ia_ifa);
1204 	}
1205 	return (error);
1206 
1207   cleanup:
1208 	KASSERT(hostIsNew, ("in6_update_ifa: cleanup: !hostIsNew"));
1209 	ifa_free(&ia->ia_ifa);
1210 	in6_purgeaddr(&ia->ia_ifa);
1211 	return error;
1212 }
1213 
1214 void
1215 in6_purgeaddr(struct ifaddr *ifa)
1216 {
1217 	struct ifnet *ifp = ifa->ifa_ifp;
1218 	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1219 	struct in6_multi_mship *imm;
1220 	struct sockaddr_in6 mltaddr, mltmask;
1221 	struct rtentry rt0;
1222 	struct sockaddr_dl gateway;
1223 	struct sockaddr_in6 mask, addr;
1224 	int plen, error;
1225 	struct rtentry *rt;
1226 	struct ifaddr *ifa0, *nifa;
1227 
1228 	/*
1229 	 * find another IPv6 address as the gateway for the
1230 	 * link-local and node-local all-nodes multicast
1231 	 * address routes
1232 	 */
1233 	IF_ADDR_LOCK(ifp);
1234 	TAILQ_FOREACH_SAFE(ifa0, &ifp->if_addrhead, ifa_link, nifa) {
1235 		if ((ifa0->ifa_addr->sa_family != AF_INET6) ||
1236 		    memcmp(&satosin6(ifa0->ifa_addr)->sin6_addr,
1237 			   &ia->ia_addr.sin6_addr,
1238 			   sizeof(struct in6_addr)) == 0)
1239 			continue;
1240 		else
1241 			break;
1242 	}
1243 	if (ifa0 != NULL)
1244 		ifa_ref(ifa0);
1245 	IF_ADDR_UNLOCK(ifp);
1246 
1247 	/*
1248 	 * Remove the loopback route to the interface address.
1249 	 * The check for the current setting of "nd6_useloopback"
1250 	 * is not needed.
1251 	 */
1252 	if (ia->ia_flags & IFA_RTSELF) {
1253 		error = ifa_del_loopback_route((struct ifaddr *)ia,
1254 				       (struct sockaddr *)&ia->ia_addr);
1255 		if (error == 0)
1256 			ia->ia_flags &= ~IFA_RTSELF;
1257 	}
1258 
1259 	/* stop DAD processing */
1260 	nd6_dad_stop(ifa);
1261 
1262 	IF_AFDATA_LOCK(ifp);
1263 	lla_lookup(LLTABLE6(ifp), (LLE_DELETE | LLE_IFADDR),
1264 	    (struct sockaddr *)&ia->ia_addr);
1265 	IF_AFDATA_UNLOCK(ifp);
1266 
1267 	/*
1268 	 * initialize for rtmsg generation
1269 	 */
1270 	bzero(&gateway, sizeof(gateway));
1271 	gateway.sdl_len = sizeof(gateway);
1272 	gateway.sdl_family = AF_LINK;
1273 	gateway.sdl_nlen = 0;
1274 	gateway.sdl_alen = ifp->if_addrlen;
1275 	/* */
1276 	bzero(&rt0, sizeof(rt0));
1277 	rt0.rt_gateway = (struct sockaddr *)&gateway;
1278 	memcpy(&mask, &ia->ia_prefixmask, sizeof(ia->ia_prefixmask));
1279 	memcpy(&addr, &ia->ia_addr, sizeof(ia->ia_addr));
1280 	rt_mask(&rt0) = (struct sockaddr *)&mask;
1281 	rt_key(&rt0) = (struct sockaddr *)&addr;
1282 	rt0.rt_flags = RTF_HOST | RTF_STATIC;
1283 	rt_newaddrmsg(RTM_DELETE, ifa, 0, &rt0);
1284 
1285 	/*
1286 	 * leave from multicast groups we have joined for the interface
1287 	 */
1288 	while ((imm = ia->ia6_memberships.lh_first) != NULL) {
1289 		LIST_REMOVE(imm, i6mm_chain);
1290 		in6_leavegroup(imm);
1291 	}
1292 
1293 	/*
1294 	 * remove the link-local all-nodes address
1295 	 */
1296 	bzero(&mltmask, sizeof(mltmask));
1297 	mltmask.sin6_len = sizeof(struct sockaddr_in6);
1298 	mltmask.sin6_family = AF_INET6;
1299 	mltmask.sin6_addr = in6mask32;
1300 
1301 	bzero(&mltaddr, sizeof(mltaddr));
1302 	mltaddr.sin6_len = sizeof(struct sockaddr_in6);
1303 	mltaddr.sin6_family = AF_INET6;
1304 	mltaddr.sin6_addr = in6addr_linklocal_allnodes;
1305 
1306 	if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) !=
1307 	    0)
1308 		goto cleanup;
1309 
1310 	rt = rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL);
1311 	if (rt != NULL && rt->rt_gateway != NULL &&
1312 	    (memcmp(&satosin6(rt->rt_gateway)->sin6_addr,
1313 		    &ia->ia_addr.sin6_addr,
1314 		    sizeof(ia->ia_addr.sin6_addr)) == 0)) {
1315 		/*
1316 		 * if no more IPv6 address exists on this interface
1317 		 * then remove the multicast address route
1318 		 */
1319 		if (ifa0 == NULL) {
1320 			memcpy(&mltaddr.sin6_addr, &satosin6(rt_key(rt))->sin6_addr,
1321 			       sizeof(mltaddr.sin6_addr));
1322 			RTFREE_LOCKED(rt);
1323 			error = rtrequest(RTM_DELETE, (struct sockaddr *)&mltaddr,
1324 					  (struct sockaddr *)&ia->ia_addr,
1325 					  (struct sockaddr *)&mltmask, RTF_UP,
1326 					  (struct rtentry **)0);
1327 			if (error)
1328 				log(LOG_INFO, "in6_purgeaddr: link-local all-nodes"
1329 				    "multicast address deletion error\n");
1330 		} else {
1331 			/*
1332 			 * replace the gateway of the route
1333 			 */
1334 			struct sockaddr_in6 sa;
1335 
1336 			bzero(&sa, sizeof(sa));
1337 			sa.sin6_len = sizeof(struct sockaddr_in6);
1338 			sa.sin6_family = AF_INET6;
1339 			memcpy(&sa.sin6_addr, &satosin6(ifa0->ifa_addr)->sin6_addr,
1340 			       sizeof(sa.sin6_addr));
1341 			in6_setscope(&sa.sin6_addr, ifa0->ifa_ifp, NULL);
1342 			memcpy(rt->rt_gateway, &sa, sizeof(sa));
1343 			RTFREE_LOCKED(rt);
1344 		}
1345 	} else {
1346 		if (rt != NULL)
1347 			RTFREE_LOCKED(rt);
1348 	}
1349 
1350 	/*
1351 	 * remove the node-local all-nodes address
1352 	 */
1353 	mltaddr.sin6_addr = in6addr_nodelocal_allnodes;
1354 	if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) !=
1355 	    0)
1356 		goto cleanup;
1357 
1358 	rt = rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL);
1359 	if (rt != NULL && rt->rt_gateway != NULL &&
1360 	    (memcmp(&satosin6(rt->rt_gateway)->sin6_addr,
1361 		    &ia->ia_addr.sin6_addr,
1362 		    sizeof(ia->ia_addr.sin6_addr)) == 0)) {
1363 		/*
1364 		 * if no more IPv6 address exists on this interface
1365 		 * then remove the multicast address route
1366 		 */
1367 		if (ifa0 == NULL) {
1368 			memcpy(&mltaddr.sin6_addr, &satosin6(rt_key(rt))->sin6_addr,
1369 			       sizeof(mltaddr.sin6_addr));
1370 
1371 			RTFREE_LOCKED(rt);
1372 			error = rtrequest(RTM_DELETE, (struct sockaddr *)&mltaddr,
1373 					  (struct sockaddr *)&ia->ia_addr,
1374 					  (struct sockaddr *)&mltmask, RTF_UP,
1375 					  (struct rtentry **)0);
1376 
1377 			if (error)
1378 				log(LOG_INFO, "in6_purgeaddr: node-local all-nodes"
1379 				    "multicast address deletion error\n");
1380 		} else {
1381 			/*
1382 			 * replace the gateway of the route
1383 			 */
1384 			struct sockaddr_in6 sa;
1385 
1386 			bzero(&sa, sizeof(sa));
1387 			sa.sin6_len = sizeof(struct sockaddr_in6);
1388 			sa.sin6_family = AF_INET6;
1389 			memcpy(&sa.sin6_addr, &satosin6(ifa0->ifa_addr)->sin6_addr,
1390 			       sizeof(sa.sin6_addr));
1391 			in6_setscope(&sa.sin6_addr, ifa0->ifa_ifp, NULL);
1392 			memcpy(rt->rt_gateway, &sa, sizeof(sa));
1393 			RTFREE_LOCKED(rt);
1394 		}
1395 	} else {
1396 		if (rt != NULL)
1397 			RTFREE_LOCKED(rt);
1398 	}
1399 
1400 cleanup:
1401 
1402 	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1403 	if ((ia->ia_flags & IFA_ROUTE) && plen == 128) {
1404 		int error;
1405 		struct sockaddr *dstaddr;
1406 
1407 		/*
1408 		 * use the interface address if configuring an
1409 		 * interface address with a /128 prefix len
1410 		 */
1411 		if (ia->ia_dstaddr.sin6_family == AF_INET6)
1412 			dstaddr = (struct sockaddr *)&ia->ia_dstaddr;
1413 		else
1414 			dstaddr = (struct sockaddr *)&ia->ia_addr;
1415 
1416 		error = rtrequest(RTM_DELETE,
1417 		    (struct sockaddr *)dstaddr,
1418 		    (struct sockaddr *)&ia->ia_addr,
1419 		    (struct sockaddr *)&ia->ia_prefixmask,
1420 		    ia->ia_flags | RTF_HOST, NULL);
1421 		if (error != 0)
1422 			return;
1423 		ia->ia_flags &= ~IFA_ROUTE;
1424 	}
1425 	if (ifa0 != NULL)
1426 		ifa_free(ifa0);
1427 
1428 	in6_unlink_ifa(ia, ifp);
1429 }
1430 
1431 static void
1432 in6_unlink_ifa(struct in6_ifaddr *ia, struct ifnet *ifp)
1433 {
1434 	int	s = splnet();
1435 
1436 	IF_ADDR_LOCK(ifp);
1437 	TAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
1438 	IF_ADDR_UNLOCK(ifp);
1439 	ifa_free(&ia->ia_ifa);			/* if_addrhead */
1440 
1441 	/*
1442 	 * Defer the release of what might be the last reference to the
1443 	 * in6_ifaddr so that it can't be freed before the remainder of the
1444 	 * cleanup.
1445 	 */
1446 	IN6_IFADDR_WLOCK();
1447 	TAILQ_REMOVE(&V_in6_ifaddrhead, ia, ia_link);
1448 	IN6_IFADDR_WUNLOCK();
1449 
1450 	/*
1451 	 * Release the reference to the base prefix.  There should be a
1452 	 * positive reference.
1453 	 */
1454 	if (ia->ia6_ndpr == NULL) {
1455 		nd6log((LOG_NOTICE,
1456 		    "in6_unlink_ifa: autoconf'ed address "
1457 		    "%p has no prefix\n", ia));
1458 	} else {
1459 		ia->ia6_ndpr->ndpr_refcnt--;
1460 		ia->ia6_ndpr = NULL;
1461 	}
1462 
1463 	/*
1464 	 * Also, if the address being removed is autoconf'ed, call
1465 	 * pfxlist_onlink_check() since the release might affect the status of
1466 	 * other (detached) addresses.
1467 	 */
1468 	if ((ia->ia6_flags & IN6_IFF_AUTOCONF)) {
1469 		pfxlist_onlink_check();
1470 	}
1471 	ifa_free(&ia->ia_ifa);			/* in6_ifaddrhead */
1472 	splx(s);
1473 }
1474 
1475 void
1476 in6_purgeif(struct ifnet *ifp)
1477 {
1478 	struct ifaddr *ifa, *nifa;
1479 
1480 	TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, nifa) {
1481 		if (ifa->ifa_addr->sa_family != AF_INET6)
1482 			continue;
1483 		in6_purgeaddr(ifa);
1484 	}
1485 
1486 	in6_ifdetach(ifp);
1487 }
1488 
1489 /*
1490  * SIOC[GAD]LIFADDR.
1491  *	SIOCGLIFADDR: get first address. (?)
1492  *	SIOCGLIFADDR with IFLR_PREFIX:
1493  *		get first address that matches the specified prefix.
1494  *	SIOCALIFADDR: add the specified address.
1495  *	SIOCALIFADDR with IFLR_PREFIX:
1496  *		add the specified prefix, filling hostid part from
1497  *		the first link-local address.  prefixlen must be <= 64.
1498  *	SIOCDLIFADDR: delete the specified address.
1499  *	SIOCDLIFADDR with IFLR_PREFIX:
1500  *		delete the first address that matches the specified prefix.
1501  * return values:
1502  *	EINVAL on invalid parameters
1503  *	EADDRNOTAVAIL on prefix match failed/specified address not found
1504  *	other values may be returned from in6_ioctl()
1505  *
1506  * NOTE: SIOCALIFADDR(with IFLR_PREFIX set) allows prefixlen less than 64.
1507  * this is to accomodate address naming scheme other than RFC2374,
1508  * in the future.
1509  * RFC2373 defines interface id to be 64bit, but it allows non-RFC2374
1510  * address encoding scheme. (see figure on page 8)
1511  */
1512 static int
1513 in6_lifaddr_ioctl(struct socket *so, u_long cmd, caddr_t data,
1514     struct ifnet *ifp, struct thread *td)
1515 {
1516 	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
1517 	struct ifaddr *ifa;
1518 	struct sockaddr *sa;
1519 
1520 	/* sanity checks */
1521 	if (!data || !ifp) {
1522 		panic("invalid argument to in6_lifaddr_ioctl");
1523 		/* NOTREACHED */
1524 	}
1525 
1526 	switch (cmd) {
1527 	case SIOCGLIFADDR:
1528 		/* address must be specified on GET with IFLR_PREFIX */
1529 		if ((iflr->flags & IFLR_PREFIX) == 0)
1530 			break;
1531 		/* FALLTHROUGH */
1532 	case SIOCALIFADDR:
1533 	case SIOCDLIFADDR:
1534 		/* address must be specified on ADD and DELETE */
1535 		sa = (struct sockaddr *)&iflr->addr;
1536 		if (sa->sa_family != AF_INET6)
1537 			return EINVAL;
1538 		if (sa->sa_len != sizeof(struct sockaddr_in6))
1539 			return EINVAL;
1540 		/* XXX need improvement */
1541 		sa = (struct sockaddr *)&iflr->dstaddr;
1542 		if (sa->sa_family && sa->sa_family != AF_INET6)
1543 			return EINVAL;
1544 		if (sa->sa_len && sa->sa_len != sizeof(struct sockaddr_in6))
1545 			return EINVAL;
1546 		break;
1547 	default: /* shouldn't happen */
1548 #if 0
1549 		panic("invalid cmd to in6_lifaddr_ioctl");
1550 		/* NOTREACHED */
1551 #else
1552 		return EOPNOTSUPP;
1553 #endif
1554 	}
1555 	if (sizeof(struct in6_addr) * 8 < iflr->prefixlen)
1556 		return EINVAL;
1557 
1558 	switch (cmd) {
1559 	case SIOCALIFADDR:
1560 	    {
1561 		struct in6_aliasreq ifra;
1562 		struct in6_addr *hostid = NULL;
1563 		int prefixlen;
1564 
1565 		ifa = NULL;
1566 		if ((iflr->flags & IFLR_PREFIX) != 0) {
1567 			struct sockaddr_in6 *sin6;
1568 
1569 			/*
1570 			 * hostid is to fill in the hostid part of the
1571 			 * address.  hostid points to the first link-local
1572 			 * address attached to the interface.
1573 			 */
1574 			ifa = (struct ifaddr *)in6ifa_ifpforlinklocal(ifp, 0);
1575 			if (!ifa)
1576 				return EADDRNOTAVAIL;
1577 			hostid = IFA_IN6(ifa);
1578 
1579 			/* prefixlen must be <= 64. */
1580 			if (64 < iflr->prefixlen)
1581 				return EINVAL;
1582 			prefixlen = iflr->prefixlen;
1583 
1584 			/* hostid part must be zero. */
1585 			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1586 			if (sin6->sin6_addr.s6_addr32[2] != 0 ||
1587 			    sin6->sin6_addr.s6_addr32[3] != 0) {
1588 				return EINVAL;
1589 			}
1590 		} else
1591 			prefixlen = iflr->prefixlen;
1592 
1593 		/* copy args to in6_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
1594 		bzero(&ifra, sizeof(ifra));
1595 		bcopy(iflr->iflr_name, ifra.ifra_name, sizeof(ifra.ifra_name));
1596 
1597 		bcopy(&iflr->addr, &ifra.ifra_addr,
1598 		    ((struct sockaddr *)&iflr->addr)->sa_len);
1599 		if (hostid) {
1600 			/* fill in hostid part */
1601 			ifra.ifra_addr.sin6_addr.s6_addr32[2] =
1602 			    hostid->s6_addr32[2];
1603 			ifra.ifra_addr.sin6_addr.s6_addr32[3] =
1604 			    hostid->s6_addr32[3];
1605 		}
1606 
1607 		if (((struct sockaddr *)&iflr->dstaddr)->sa_family) { /* XXX */
1608 			bcopy(&iflr->dstaddr, &ifra.ifra_dstaddr,
1609 			    ((struct sockaddr *)&iflr->dstaddr)->sa_len);
1610 			if (hostid) {
1611 				ifra.ifra_dstaddr.sin6_addr.s6_addr32[2] =
1612 				    hostid->s6_addr32[2];
1613 				ifra.ifra_dstaddr.sin6_addr.s6_addr32[3] =
1614 				    hostid->s6_addr32[3];
1615 			}
1616 		}
1617 		if (ifa != NULL)
1618 			ifa_free(ifa);
1619 
1620 		ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
1621 		in6_prefixlen2mask(&ifra.ifra_prefixmask.sin6_addr, prefixlen);
1622 
1623 		ifra.ifra_flags = iflr->flags & ~IFLR_PREFIX;
1624 		return in6_control(so, SIOCAIFADDR_IN6, (caddr_t)&ifra, ifp, td);
1625 	    }
1626 	case SIOCGLIFADDR:
1627 	case SIOCDLIFADDR:
1628 	    {
1629 		struct in6_ifaddr *ia;
1630 		struct in6_addr mask, candidate, match;
1631 		struct sockaddr_in6 *sin6;
1632 		int cmp;
1633 
1634 		bzero(&mask, sizeof(mask));
1635 		if (iflr->flags & IFLR_PREFIX) {
1636 			/* lookup a prefix rather than address. */
1637 			in6_prefixlen2mask(&mask, iflr->prefixlen);
1638 
1639 			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1640 			bcopy(&sin6->sin6_addr, &match, sizeof(match));
1641 			match.s6_addr32[0] &= mask.s6_addr32[0];
1642 			match.s6_addr32[1] &= mask.s6_addr32[1];
1643 			match.s6_addr32[2] &= mask.s6_addr32[2];
1644 			match.s6_addr32[3] &= mask.s6_addr32[3];
1645 
1646 			/* if you set extra bits, that's wrong */
1647 			if (bcmp(&match, &sin6->sin6_addr, sizeof(match)))
1648 				return EINVAL;
1649 
1650 			cmp = 1;
1651 		} else {
1652 			if (cmd == SIOCGLIFADDR) {
1653 				/* on getting an address, take the 1st match */
1654 				cmp = 0;	/* XXX */
1655 			} else {
1656 				/* on deleting an address, do exact match */
1657 				in6_prefixlen2mask(&mask, 128);
1658 				sin6 = (struct sockaddr_in6 *)&iflr->addr;
1659 				bcopy(&sin6->sin6_addr, &match, sizeof(match));
1660 
1661 				cmp = 1;
1662 			}
1663 		}
1664 
1665 		IF_ADDR_LOCK(ifp);
1666 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1667 			if (ifa->ifa_addr->sa_family != AF_INET6)
1668 				continue;
1669 			if (!cmp)
1670 				break;
1671 
1672 			/*
1673 			 * XXX: this is adhoc, but is necessary to allow
1674 			 * a user to specify fe80::/64 (not /10) for a
1675 			 * link-local address.
1676 			 */
1677 			bcopy(IFA_IN6(ifa), &candidate, sizeof(candidate));
1678 			in6_clearscope(&candidate);
1679 			candidate.s6_addr32[0] &= mask.s6_addr32[0];
1680 			candidate.s6_addr32[1] &= mask.s6_addr32[1];
1681 			candidate.s6_addr32[2] &= mask.s6_addr32[2];
1682 			candidate.s6_addr32[3] &= mask.s6_addr32[3];
1683 			if (IN6_ARE_ADDR_EQUAL(&candidate, &match))
1684 				break;
1685 		}
1686 		IF_ADDR_UNLOCK(ifp);
1687 		if (!ifa)
1688 			return EADDRNOTAVAIL;
1689 		ia = ifa2ia6(ifa);
1690 
1691 		if (cmd == SIOCGLIFADDR) {
1692 			int error;
1693 
1694 			/* fill in the if_laddrreq structure */
1695 			bcopy(&ia->ia_addr, &iflr->addr, ia->ia_addr.sin6_len);
1696 			error = sa6_recoverscope(
1697 			    (struct sockaddr_in6 *)&iflr->addr);
1698 			if (error != 0)
1699 				return (error);
1700 
1701 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1702 				bcopy(&ia->ia_dstaddr, &iflr->dstaddr,
1703 				    ia->ia_dstaddr.sin6_len);
1704 				error = sa6_recoverscope(
1705 				    (struct sockaddr_in6 *)&iflr->dstaddr);
1706 				if (error != 0)
1707 					return (error);
1708 			} else
1709 				bzero(&iflr->dstaddr, sizeof(iflr->dstaddr));
1710 
1711 			iflr->prefixlen =
1712 			    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
1713 
1714 			iflr->flags = ia->ia6_flags;	/* XXX */
1715 
1716 			return 0;
1717 		} else {
1718 			struct in6_aliasreq ifra;
1719 
1720 			/* fill in6_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
1721 			bzero(&ifra, sizeof(ifra));
1722 			bcopy(iflr->iflr_name, ifra.ifra_name,
1723 			    sizeof(ifra.ifra_name));
1724 
1725 			bcopy(&ia->ia_addr, &ifra.ifra_addr,
1726 			    ia->ia_addr.sin6_len);
1727 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1728 				bcopy(&ia->ia_dstaddr, &ifra.ifra_dstaddr,
1729 				    ia->ia_dstaddr.sin6_len);
1730 			} else {
1731 				bzero(&ifra.ifra_dstaddr,
1732 				    sizeof(ifra.ifra_dstaddr));
1733 			}
1734 			bcopy(&ia->ia_prefixmask, &ifra.ifra_dstaddr,
1735 			    ia->ia_prefixmask.sin6_len);
1736 
1737 			ifra.ifra_flags = ia->ia6_flags;
1738 			return in6_control(so, SIOCDIFADDR_IN6, (caddr_t)&ifra,
1739 			    ifp, td);
1740 		}
1741 	    }
1742 	}
1743 
1744 	return EOPNOTSUPP;	/* just for safety */
1745 }
1746 
1747 /*
1748  * Initialize an interface's intetnet6 address
1749  * and routing table entry.
1750  */
1751 static int
1752 in6_ifinit(struct ifnet *ifp, struct in6_ifaddr *ia,
1753     struct sockaddr_in6 *sin6, int newhost)
1754 {
1755 	int	error = 0, plen, ifacount = 0;
1756 	int	s = splimp();
1757 	struct ifaddr *ifa;
1758 
1759 	/*
1760 	 * Give the interface a chance to initialize
1761 	 * if this is its first address,
1762 	 * and to validate the address if necessary.
1763 	 */
1764 	IF_ADDR_LOCK(ifp);
1765 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1766 		if (ifa->ifa_addr->sa_family != AF_INET6)
1767 			continue;
1768 		ifacount++;
1769 	}
1770 	IF_ADDR_UNLOCK(ifp);
1771 
1772 	ia->ia_addr = *sin6;
1773 
1774 	if (ifacount <= 1 && ifp->if_ioctl) {
1775 		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
1776 		if (error) {
1777 			splx(s);
1778 			return (error);
1779 		}
1780 	}
1781 	splx(s);
1782 
1783 	ia->ia_ifa.ifa_metric = ifp->if_metric;
1784 
1785 	/* we could do in(6)_socktrim here, but just omit it at this moment. */
1786 
1787 	/*
1788 	 * Special case:
1789 	 * If a new destination address is specified for a point-to-point
1790 	 * interface, install a route to the destination as an interface
1791 	 * direct route.
1792 	 * XXX: the logic below rejects assigning multiple addresses on a p2p
1793 	 * interface that share the same destination.
1794 	 */
1795 	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1796 	if (!(ia->ia_flags & IFA_ROUTE) && plen == 128 &&
1797 	    ia->ia_dstaddr.sin6_family == AF_INET6) {
1798 		int rtflags = RTF_UP | RTF_HOST;
1799 
1800 		error = rtrequest(RTM_ADD,
1801 		    (struct sockaddr *)&ia->ia_dstaddr,
1802 		    (struct sockaddr *)&ia->ia_addr,
1803 		    (struct sockaddr *)&ia->ia_prefixmask,
1804 		    ia->ia_flags | rtflags, NULL);
1805 		if (error != 0)
1806 			return (error);
1807 		ia->ia_flags |= IFA_ROUTE;
1808 	}
1809 
1810 	/*
1811 	 * add a loopback route to self
1812 	 */
1813 	if (!(ia->ia_flags & IFA_ROUTE)
1814 	    && (V_nd6_useloopback
1815 		|| (ifp->if_flags & IFF_LOOPBACK))) {
1816 		error = ifa_add_loopback_route((struct ifaddr *)ia,
1817 				       (struct sockaddr *)&ia->ia_addr);
1818 		if (error == 0)
1819 			ia->ia_flags |= IFA_RTSELF;
1820 	}
1821 
1822 	/* Add ownaddr as loopback rtentry, if necessary (ex. on p2p link). */
1823 	if (newhost) {
1824 		struct llentry *ln;
1825 		struct rtentry rt;
1826 		struct sockaddr_dl gateway;
1827 		struct sockaddr_in6 mask, addr;
1828 
1829 		IF_AFDATA_LOCK(ifp);
1830 		ia->ia_ifa.ifa_rtrequest = NULL;
1831 
1832 		/* XXX QL
1833 		 * we need to report rt_newaddrmsg
1834 		 */
1835 		ln = lla_lookup(LLTABLE6(ifp), (LLE_CREATE | LLE_IFADDR | LLE_EXCLUSIVE),
1836 		    (struct sockaddr *)&ia->ia_addr);
1837 		IF_AFDATA_UNLOCK(ifp);
1838 		if (ln != NULL) {
1839 			ln->la_expire = 0;  /* for IPv6 this means permanent */
1840 			ln->ln_state = ND6_LLINFO_REACHABLE;
1841 			/*
1842 			 * initialize for rtmsg generation
1843 			 */
1844 			bzero(&gateway, sizeof(gateway));
1845 			gateway.sdl_len = sizeof(gateway);
1846 			gateway.sdl_family = AF_LINK;
1847 			gateway.sdl_nlen = 0;
1848 			gateway.sdl_alen = 6;
1849 			memcpy(gateway.sdl_data, &ln->ll_addr.mac_aligned, sizeof(ln->ll_addr));
1850 			/* */
1851 			LLE_WUNLOCK(ln);
1852 		}
1853 
1854 		bzero(&rt, sizeof(rt));
1855 		rt.rt_gateway = (struct sockaddr *)&gateway;
1856 		memcpy(&mask, &ia->ia_prefixmask, sizeof(ia->ia_prefixmask));
1857 		memcpy(&addr, &ia->ia_addr, sizeof(ia->ia_addr));
1858 		rt_mask(&rt) = (struct sockaddr *)&mask;
1859 		rt_key(&rt) = (struct sockaddr *)&addr;
1860 		rt.rt_flags = RTF_UP | RTF_HOST | RTF_STATIC;
1861 		rt_newaddrmsg(RTM_ADD, &ia->ia_ifa, 0, &rt);
1862 	}
1863 
1864 	return (error);
1865 }
1866 
1867 /*
1868  * Find an IPv6 interface link-local address specific to an interface.
1869  * ifaddr is returned referenced.
1870  */
1871 struct in6_ifaddr *
1872 in6ifa_ifpforlinklocal(struct ifnet *ifp, int ignoreflags)
1873 {
1874 	struct ifaddr *ifa;
1875 
1876 	IF_ADDR_LOCK(ifp);
1877 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1878 		if (ifa->ifa_addr->sa_family != AF_INET6)
1879 			continue;
1880 		if (IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa))) {
1881 			if ((((struct in6_ifaddr *)ifa)->ia6_flags &
1882 			     ignoreflags) != 0)
1883 				continue;
1884 			ifa_ref(ifa);
1885 			break;
1886 		}
1887 	}
1888 	IF_ADDR_UNLOCK(ifp);
1889 
1890 	return ((struct in6_ifaddr *)ifa);
1891 }
1892 
1893 
1894 /*
1895  * find the internet address corresponding to a given interface and address.
1896  * ifaddr is returned referenced.
1897  */
1898 struct in6_ifaddr *
1899 in6ifa_ifpwithaddr(struct ifnet *ifp, struct in6_addr *addr)
1900 {
1901 	struct ifaddr *ifa;
1902 
1903 	IF_ADDR_LOCK(ifp);
1904 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1905 		if (ifa->ifa_addr->sa_family != AF_INET6)
1906 			continue;
1907 		if (IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa))) {
1908 			ifa_ref(ifa);
1909 			break;
1910 		}
1911 	}
1912 	IF_ADDR_UNLOCK(ifp);
1913 
1914 	return ((struct in6_ifaddr *)ifa);
1915 }
1916 
1917 /*
1918  * Convert IP6 address to printable (loggable) representation. Caller
1919  * has to make sure that ip6buf is at least INET6_ADDRSTRLEN long.
1920  */
1921 static char digits[] = "0123456789abcdef";
1922 char *
1923 ip6_sprintf(char *ip6buf, const struct in6_addr *addr)
1924 {
1925 	int i, cnt = 0, maxcnt = 0, idx = 0, index = 0;
1926 	char *cp;
1927 	const u_int16_t *a = (const u_int16_t *)addr;
1928 	const u_int8_t *d;
1929 	int dcolon = 0, zero = 0;
1930 
1931 	cp = ip6buf;
1932 
1933 	for (i = 0; i < 8; i++) {
1934 		if (*(a + i) == 0) {
1935 			cnt++;
1936 			if (cnt == 1)
1937 				idx = i;
1938 		}
1939 		else if (maxcnt < cnt) {
1940 			maxcnt = cnt;
1941 			index = idx;
1942 			cnt = 0;
1943 		}
1944 	}
1945 	if (maxcnt < cnt) {
1946 		maxcnt = cnt;
1947 		index = idx;
1948 	}
1949 
1950 	for (i = 0; i < 8; i++) {
1951 		if (dcolon == 1) {
1952 			if (*a == 0) {
1953 				if (i == 7)
1954 					*cp++ = ':';
1955 				a++;
1956 				continue;
1957 			} else
1958 				dcolon = 2;
1959 		}
1960 		if (*a == 0) {
1961 			if (dcolon == 0 && *(a + 1) == 0 && i == index) {
1962 				if (i == 0)
1963 					*cp++ = ':';
1964 				*cp++ = ':';
1965 				dcolon = 1;
1966 			} else {
1967 				*cp++ = '0';
1968 				*cp++ = ':';
1969 			}
1970 			a++;
1971 			continue;
1972 		}
1973 		d = (const u_char *)a;
1974 		/* Try to eliminate leading zeros in printout like in :0001. */
1975 		zero = 1;
1976 		*cp = digits[*d >> 4];
1977 		if (*cp != '0') {
1978 			zero = 0;
1979 			cp++;
1980 		}
1981 		*cp = digits[*d++ & 0xf];
1982 		if (zero == 0 || (*cp != '0')) {
1983 			zero = 0;
1984 			cp++;
1985 		}
1986 		*cp = digits[*d >> 4];
1987 		if (zero == 0 || (*cp != '0')) {
1988 			zero = 0;
1989 			cp++;
1990 		}
1991 		*cp++ = digits[*d & 0xf];
1992 		*cp++ = ':';
1993 		a++;
1994 	}
1995 	*--cp = '\0';
1996 	return (ip6buf);
1997 }
1998 
1999 int
2000 in6_localaddr(struct in6_addr *in6)
2001 {
2002 	struct in6_ifaddr *ia;
2003 
2004 	if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
2005 		return 1;
2006 
2007 	IN6_IFADDR_RLOCK();
2008 	TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
2009 		if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
2010 		    &ia->ia_prefixmask.sin6_addr)) {
2011 			IN6_IFADDR_RUNLOCK();
2012 			return 1;
2013 		}
2014 	}
2015 	IN6_IFADDR_RUNLOCK();
2016 
2017 	return (0);
2018 }
2019 
2020 /*
2021  * Return 1 if an internet address is for the local host and configured
2022  * on one of its interfaces.
2023  */
2024 int
2025 in6_localip(struct in6_addr *in6)
2026 {
2027 	struct in6_ifaddr *ia;
2028 
2029 	IN6_IFADDR_RLOCK();
2030 	TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
2031 		if (IN6_ARE_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr)) {
2032 			IN6_IFADDR_RUNLOCK();
2033 			return (1);
2034 		}
2035 	}
2036 	IN6_IFADDR_RUNLOCK();
2037 	return (0);
2038 }
2039 
2040 
2041 int
2042 in6_is_addr_deprecated(struct sockaddr_in6 *sa6)
2043 {
2044 	struct in6_ifaddr *ia;
2045 
2046 	IN6_IFADDR_RLOCK();
2047 	TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
2048 		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
2049 				       &sa6->sin6_addr) &&
2050 		    (ia->ia6_flags & IN6_IFF_DEPRECATED) != 0) {
2051 			IN6_IFADDR_RUNLOCK();
2052 			return (1); /* true */
2053 		}
2054 
2055 		/* XXX: do we still have to go thru the rest of the list? */
2056 	}
2057 	IN6_IFADDR_RUNLOCK();
2058 
2059 	return (0);		/* false */
2060 }
2061 
2062 /*
2063  * return length of part which dst and src are equal
2064  * hard coding...
2065  */
2066 int
2067 in6_matchlen(struct in6_addr *src, struct in6_addr *dst)
2068 {
2069 	int match = 0;
2070 	u_char *s = (u_char *)src, *d = (u_char *)dst;
2071 	u_char *lim = s + 16, r;
2072 
2073 	while (s < lim)
2074 		if ((r = (*d++ ^ *s++)) != 0) {
2075 			while (r < 128) {
2076 				match++;
2077 				r <<= 1;
2078 			}
2079 			break;
2080 		} else
2081 			match += 8;
2082 	return match;
2083 }
2084 
2085 /* XXX: to be scope conscious */
2086 int
2087 in6_are_prefix_equal(struct in6_addr *p1, struct in6_addr *p2, int len)
2088 {
2089 	int bytelen, bitlen;
2090 
2091 	/* sanity check */
2092 	if (0 > len || len > 128) {
2093 		log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n",
2094 		    len);
2095 		return (0);
2096 	}
2097 
2098 	bytelen = len / 8;
2099 	bitlen = len % 8;
2100 
2101 	if (bcmp(&p1->s6_addr, &p2->s6_addr, bytelen))
2102 		return (0);
2103 	if (bitlen != 0 &&
2104 	    p1->s6_addr[bytelen] >> (8 - bitlen) !=
2105 	    p2->s6_addr[bytelen] >> (8 - bitlen))
2106 		return (0);
2107 
2108 	return (1);
2109 }
2110 
2111 void
2112 in6_prefixlen2mask(struct in6_addr *maskp, int len)
2113 {
2114 	u_char maskarray[8] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
2115 	int bytelen, bitlen, i;
2116 
2117 	/* sanity check */
2118 	if (0 > len || len > 128) {
2119 		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
2120 		    len);
2121 		return;
2122 	}
2123 
2124 	bzero(maskp, sizeof(*maskp));
2125 	bytelen = len / 8;
2126 	bitlen = len % 8;
2127 	for (i = 0; i < bytelen; i++)
2128 		maskp->s6_addr[i] = 0xff;
2129 	if (bitlen)
2130 		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
2131 }
2132 
2133 /*
2134  * return the best address out of the same scope. if no address was
2135  * found, return the first valid address from designated IF.
2136  */
2137 struct in6_ifaddr *
2138 in6_ifawithifp(struct ifnet *ifp, struct in6_addr *dst)
2139 {
2140 	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
2141 	struct ifaddr *ifa;
2142 	struct in6_ifaddr *besta = 0;
2143 	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
2144 
2145 	dep[0] = dep[1] = NULL;
2146 
2147 	/*
2148 	 * We first look for addresses in the same scope.
2149 	 * If there is one, return it.
2150 	 * If two or more, return one which matches the dst longest.
2151 	 * If none, return one of global addresses assigned other ifs.
2152 	 */
2153 	IF_ADDR_LOCK(ifp);
2154 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2155 		if (ifa->ifa_addr->sa_family != AF_INET6)
2156 			continue;
2157 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
2158 			continue; /* XXX: is there any case to allow anycast? */
2159 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
2160 			continue; /* don't use this interface */
2161 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
2162 			continue;
2163 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2164 			if (V_ip6_use_deprecated)
2165 				dep[0] = (struct in6_ifaddr *)ifa;
2166 			continue;
2167 		}
2168 
2169 		if (dst_scope == in6_addrscope(IFA_IN6(ifa))) {
2170 			/*
2171 			 * call in6_matchlen() as few as possible
2172 			 */
2173 			if (besta) {
2174 				if (blen == -1)
2175 					blen = in6_matchlen(&besta->ia_addr.sin6_addr, dst);
2176 				tlen = in6_matchlen(IFA_IN6(ifa), dst);
2177 				if (tlen > blen) {
2178 					blen = tlen;
2179 					besta = (struct in6_ifaddr *)ifa;
2180 				}
2181 			} else
2182 				besta = (struct in6_ifaddr *)ifa;
2183 		}
2184 	}
2185 	if (besta) {
2186 		ifa_ref(&besta->ia_ifa);
2187 		IF_ADDR_UNLOCK(ifp);
2188 		return (besta);
2189 	}
2190 	IF_ADDR_UNLOCK(ifp);
2191 
2192 	IN6_IFADDR_RLOCK();
2193 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2194 		if (ifa->ifa_addr->sa_family != AF_INET6)
2195 			continue;
2196 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
2197 			continue; /* XXX: is there any case to allow anycast? */
2198 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
2199 			continue; /* don't use this interface */
2200 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
2201 			continue;
2202 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2203 			if (V_ip6_use_deprecated)
2204 				dep[1] = (struct in6_ifaddr *)ifa;
2205 			continue;
2206 		}
2207 
2208 		if (ifa != NULL)
2209 			ifa_ref(ifa);
2210 		IN6_IFADDR_RUNLOCK();
2211 		return (struct in6_ifaddr *)ifa;
2212 	}
2213 	IN6_IFADDR_RUNLOCK();
2214 
2215 	/* use the last-resort values, that are, deprecated addresses */
2216 	if (dep[0])
2217 		return dep[0];
2218 	if (dep[1])
2219 		return dep[1];
2220 
2221 	return NULL;
2222 }
2223 
2224 /*
2225  * perform DAD when interface becomes IFF_UP.
2226  */
2227 void
2228 in6_if_up(struct ifnet *ifp)
2229 {
2230 	struct ifaddr *ifa;
2231 	struct in6_ifaddr *ia;
2232 
2233 	IF_ADDR_LOCK(ifp);
2234 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2235 		if (ifa->ifa_addr->sa_family != AF_INET6)
2236 			continue;
2237 		ia = (struct in6_ifaddr *)ifa;
2238 		if (ia->ia6_flags & IN6_IFF_TENTATIVE) {
2239 			/*
2240 			 * The TENTATIVE flag was likely set by hand
2241 			 * beforehand, implicitly indicating the need for DAD.
2242 			 * We may be able to skip the random delay in this
2243 			 * case, but we impose delays just in case.
2244 			 */
2245 			nd6_dad_start(ifa,
2246 			    arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz));
2247 		}
2248 	}
2249 	IF_ADDR_UNLOCK(ifp);
2250 
2251 	/*
2252 	 * special cases, like 6to4, are handled in in6_ifattach
2253 	 */
2254 	in6_ifattach(ifp, NULL);
2255 }
2256 
2257 int
2258 in6if_do_dad(struct ifnet *ifp)
2259 {
2260 	if ((ifp->if_flags & IFF_LOOPBACK) != 0)
2261 		return (0);
2262 
2263 	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)
2264 		return (0);
2265 
2266 	switch (ifp->if_type) {
2267 #ifdef IFT_DUMMY
2268 	case IFT_DUMMY:
2269 #endif
2270 	case IFT_FAITH:
2271 		/*
2272 		 * These interfaces do not have the IFF_LOOPBACK flag,
2273 		 * but loop packets back.  We do not have to do DAD on such
2274 		 * interfaces.  We should even omit it, because loop-backed
2275 		 * NS would confuse the DAD procedure.
2276 		 */
2277 		return (0);
2278 	default:
2279 		/*
2280 		 * Our DAD routine requires the interface up and running.
2281 		 * However, some interfaces can be up before the RUNNING
2282 		 * status.  Additionaly, users may try to assign addresses
2283 		 * before the interface becomes up (or running).
2284 		 * We simply skip DAD in such a case as a work around.
2285 		 * XXX: we should rather mark "tentative" on such addresses,
2286 		 * and do DAD after the interface becomes ready.
2287 		 */
2288 		if (!((ifp->if_flags & IFF_UP) &&
2289 		    (ifp->if_drv_flags & IFF_DRV_RUNNING)))
2290 			return (0);
2291 
2292 		return (1);
2293 	}
2294 }
2295 
2296 /*
2297  * Calculate max IPv6 MTU through all the interfaces and store it
2298  * to in6_maxmtu.
2299  */
2300 void
2301 in6_setmaxmtu(void)
2302 {
2303 	unsigned long maxmtu = 0;
2304 	struct ifnet *ifp;
2305 
2306 	IFNET_RLOCK_NOSLEEP();
2307 	for (ifp = TAILQ_FIRST(&V_ifnet); ifp;
2308 	    ifp = TAILQ_NEXT(ifp, if_list)) {
2309 		/* this function can be called during ifnet initialization */
2310 		if (!ifp->if_afdata[AF_INET6])
2311 			continue;
2312 		if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
2313 		    IN6_LINKMTU(ifp) > maxmtu)
2314 			maxmtu = IN6_LINKMTU(ifp);
2315 	}
2316 	IFNET_RUNLOCK_NOSLEEP();
2317 	if (maxmtu)	     /* update only when maxmtu is positive */
2318 		V_in6_maxmtu = maxmtu;
2319 }
2320 
2321 /*
2322  * Provide the length of interface identifiers to be used for the link attached
2323  * to the given interface.  The length should be defined in "IPv6 over
2324  * xxx-link" document.  Note that address architecture might also define
2325  * the length for a particular set of address prefixes, regardless of the
2326  * link type.  As clarified in rfc2462bis, those two definitions should be
2327  * consistent, and those really are as of August 2004.
2328  */
2329 int
2330 in6_if2idlen(struct ifnet *ifp)
2331 {
2332 	switch (ifp->if_type) {
2333 	case IFT_ETHER:		/* RFC2464 */
2334 #ifdef IFT_PROPVIRTUAL
2335 	case IFT_PROPVIRTUAL:	/* XXX: no RFC. treat it as ether */
2336 #endif
2337 #ifdef IFT_L2VLAN
2338 	case IFT_L2VLAN:	/* ditto */
2339 #endif
2340 #ifdef IFT_IEEE80211
2341 	case IFT_IEEE80211:	/* ditto */
2342 #endif
2343 #ifdef IFT_MIP
2344 	case IFT_MIP:	/* ditto */
2345 #endif
2346 	case IFT_INFINIBAND:
2347 		return (64);
2348 	case IFT_FDDI:		/* RFC2467 */
2349 		return (64);
2350 	case IFT_ISO88025:	/* RFC2470 (IPv6 over Token Ring) */
2351 		return (64);
2352 	case IFT_PPP:		/* RFC2472 */
2353 		return (64);
2354 	case IFT_ARCNET:	/* RFC2497 */
2355 		return (64);
2356 	case IFT_FRELAY:	/* RFC2590 */
2357 		return (64);
2358 	case IFT_IEEE1394:	/* RFC3146 */
2359 		return (64);
2360 	case IFT_GIF:
2361 		return (64);	/* draft-ietf-v6ops-mech-v2-07 */
2362 	case IFT_LOOP:
2363 		return (64);	/* XXX: is this really correct? */
2364 	default:
2365 		/*
2366 		 * Unknown link type:
2367 		 * It might be controversial to use the today's common constant
2368 		 * of 64 for these cases unconditionally.  For full compliance,
2369 		 * we should return an error in this case.  On the other hand,
2370 		 * if we simply miss the standard for the link type or a new
2371 		 * standard is defined for a new link type, the IFID length
2372 		 * is very likely to be the common constant.  As a compromise,
2373 		 * we always use the constant, but make an explicit notice
2374 		 * indicating the "unknown" case.
2375 		 */
2376 		printf("in6_if2idlen: unknown link type (%d)\n", ifp->if_type);
2377 		return (64);
2378 	}
2379 }
2380 
2381 #include <sys/sysctl.h>
2382 
2383 struct in6_llentry {
2384 	struct llentry		base;
2385 	struct sockaddr_in6	l3_addr6;
2386 };
2387 
2388 static struct llentry *
2389 in6_lltable_new(const struct sockaddr *l3addr, u_int flags)
2390 {
2391 	struct in6_llentry *lle;
2392 
2393 	lle = malloc(sizeof(struct in6_llentry), M_LLTABLE,
2394 	    M_DONTWAIT | M_ZERO);
2395 	if (lle == NULL)		/* NB: caller generates msg */
2396 		return NULL;
2397 
2398 	lle->l3_addr6 = *(const struct sockaddr_in6 *)l3addr;
2399 	lle->base.lle_refcnt = 1;
2400 	LLE_LOCK_INIT(&lle->base);
2401 	callout_init_rw(&lle->base.ln_timer_ch, &lle->base.lle_lock,
2402 	    CALLOUT_RETURNUNLOCKED);
2403 
2404 	return &lle->base;
2405 }
2406 
2407 /*
2408  * Deletes an address from the address table.
2409  * This function is called by the timer functions
2410  * such as arptimer() and nd6_llinfo_timer(), and
2411  * the caller does the locking.
2412  */
2413 static void
2414 in6_lltable_free(struct lltable *llt, struct llentry *lle)
2415 {
2416 	LLE_WUNLOCK(lle);
2417 	LLE_LOCK_DESTROY(lle);
2418 	free(lle, M_LLTABLE);
2419 }
2420 
2421 static void
2422 in6_lltable_prefix_free(struct lltable *llt,
2423 			const struct sockaddr *prefix,
2424 			const struct sockaddr *mask,
2425 			u_int flags)
2426 {
2427 	const struct sockaddr_in6 *pfx = (const struct sockaddr_in6 *)prefix;
2428 	const struct sockaddr_in6 *msk = (const struct sockaddr_in6 *)mask;
2429 	struct llentry *lle, *next;
2430 	register int i;
2431 
2432 	/*
2433 	 * (flags & LLE_STATIC) means deleting all entries
2434 	 * including static ND6 entries
2435 	 */
2436 	for (i=0; i < LLTBL_HASHTBL_SIZE; i++) {
2437 		LIST_FOREACH_SAFE(lle, &llt->lle_head[i], lle_next, next) {
2438 			if (IN6_ARE_MASKED_ADDR_EQUAL(
2439 				    &((struct sockaddr_in6 *)L3_ADDR(lle))->sin6_addr,
2440 				    &pfx->sin6_addr,
2441 				    &msk->sin6_addr) &&
2442 			    ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC))) {
2443 				int canceled;
2444 
2445 				canceled = callout_drain(&lle->la_timer);
2446 				LLE_WLOCK(lle);
2447 				if (canceled)
2448 					LLE_REMREF(lle);
2449 				llentry_free(lle);
2450 			}
2451 		}
2452 	}
2453 }
2454 
2455 static int
2456 in6_lltable_rtcheck(struct ifnet *ifp,
2457 		    u_int flags,
2458 		    const struct sockaddr *l3addr)
2459 {
2460 	struct rtentry *rt;
2461 	char ip6buf[INET6_ADDRSTRLEN];
2462 
2463 	KASSERT(l3addr->sa_family == AF_INET6,
2464 	    ("sin_family %d", l3addr->sa_family));
2465 
2466 	/* XXX rtalloc1 should take a const param */
2467 	rt = rtalloc1(__DECONST(struct sockaddr *, l3addr), 0, 0);
2468 	if (rt == NULL || (rt->rt_flags & RTF_GATEWAY) || rt->rt_ifp != ifp) {
2469 		struct ifaddr *ifa;
2470 		/*
2471 		 * Create an ND6 cache for an IPv6 neighbor
2472 		 * that is not covered by our own prefix.
2473 		 */
2474 		/* XXX ifaof_ifpforaddr should take a const param */
2475 		ifa = ifaof_ifpforaddr(__DECONST(struct sockaddr *, l3addr), ifp);
2476 		if (ifa != NULL) {
2477 			ifa_free(ifa);
2478 			if (rt != NULL)
2479 				RTFREE_LOCKED(rt);
2480 			return 0;
2481 		}
2482 		log(LOG_INFO, "IPv6 address: \"%s\" is not on the network\n",
2483 		    ip6_sprintf(ip6buf, &((const struct sockaddr_in6 *)l3addr)->sin6_addr));
2484 		if (rt != NULL)
2485 			RTFREE_LOCKED(rt);
2486 		return EINVAL;
2487 	}
2488 	RTFREE_LOCKED(rt);
2489 	return 0;
2490 }
2491 
2492 static struct llentry *
2493 in6_lltable_lookup(struct lltable *llt, u_int flags,
2494 	const struct sockaddr *l3addr)
2495 {
2496 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2497 	struct ifnet *ifp = llt->llt_ifp;
2498 	struct llentry *lle;
2499 	struct llentries *lleh;
2500 	u_int hashkey;
2501 
2502 	IF_AFDATA_LOCK_ASSERT(ifp);
2503 	KASSERT(l3addr->sa_family == AF_INET6,
2504 	    ("sin_family %d", l3addr->sa_family));
2505 
2506 	hashkey = sin6->sin6_addr.s6_addr32[3];
2507 	lleh = &llt->lle_head[LLATBL_HASH(hashkey, LLTBL_HASHMASK)];
2508 	LIST_FOREACH(lle, lleh, lle_next) {
2509 		struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)L3_ADDR(lle);
2510 		if (lle->la_flags & LLE_DELETED)
2511 			continue;
2512 		if (bcmp(&sa6->sin6_addr, &sin6->sin6_addr,
2513 			 sizeof(struct in6_addr)) == 0)
2514 			break;
2515 	}
2516 
2517 	if (lle == NULL) {
2518 		if (!(flags & LLE_CREATE))
2519 			return (NULL);
2520 		/*
2521 		 * A route that covers the given address must have
2522 		 * been installed 1st because we are doing a resolution,
2523 		 * verify this.
2524 		 */
2525 		if (!(flags & LLE_IFADDR) &&
2526 		    in6_lltable_rtcheck(ifp, flags, l3addr) != 0)
2527 			return NULL;
2528 
2529 		lle = in6_lltable_new(l3addr, flags);
2530 		if (lle == NULL) {
2531 			log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
2532 			return NULL;
2533 		}
2534 		lle->la_flags = flags & ~LLE_CREATE;
2535 		if ((flags & (LLE_CREATE | LLE_IFADDR)) == (LLE_CREATE | LLE_IFADDR)) {
2536 			bcopy(IF_LLADDR(ifp), &lle->ll_addr, ifp->if_addrlen);
2537 			lle->la_flags |= (LLE_VALID | LLE_STATIC);
2538 		}
2539 
2540 		lle->lle_tbl  = llt;
2541 		lle->lle_head = lleh;
2542 		LIST_INSERT_HEAD(lleh, lle, lle_next);
2543 	} else if (flags & LLE_DELETE) {
2544 		if (!(lle->la_flags & LLE_IFADDR) || (flags & LLE_IFADDR)) {
2545 			LLE_WLOCK(lle);
2546 			lle->la_flags = LLE_DELETED;
2547 			LLE_WUNLOCK(lle);
2548 #ifdef DIAGNOSTIC
2549 			log(LOG_INFO, "ifaddr cache = %p  is deleted\n", lle);
2550 #endif
2551 		}
2552 		lle = (void *)-1;
2553 	}
2554 	if (LLE_IS_VALID(lle)) {
2555 		if (flags & LLE_EXCLUSIVE)
2556 			LLE_WLOCK(lle);
2557 		else
2558 			LLE_RLOCK(lle);
2559 	}
2560 	return (lle);
2561 }
2562 
2563 static int
2564 in6_lltable_dump(struct lltable *llt, struct sysctl_req *wr)
2565 {
2566 	struct ifnet *ifp = llt->llt_ifp;
2567 	struct llentry *lle;
2568 	/* XXX stack use */
2569 	struct {
2570 		struct rt_msghdr	rtm;
2571 		struct sockaddr_in6	sin6;
2572 		/*
2573 		 * ndp.c assumes that sdl is word aligned
2574 		 */
2575 #ifdef __LP64__
2576 		uint32_t		pad;
2577 #endif
2578 		struct sockaddr_dl	sdl;
2579 	} ndpc;
2580 	int i, error;
2581 
2582 	if (ifp->if_flags & IFF_LOOPBACK)
2583 		return 0;
2584 
2585 	LLTABLE_LOCK_ASSERT();
2586 
2587 	error = 0;
2588 	for (i = 0; i < LLTBL_HASHTBL_SIZE; i++) {
2589 		LIST_FOREACH(lle, &llt->lle_head[i], lle_next) {
2590 			struct sockaddr_dl *sdl;
2591 
2592 			/* skip deleted or invalid entries */
2593 			if ((lle->la_flags & (LLE_DELETED|LLE_VALID)) != LLE_VALID)
2594 				continue;
2595 			/* Skip if jailed and not a valid IP of the prison. */
2596 			if (prison_if(wr->td->td_ucred, L3_ADDR(lle)) != 0)
2597 				continue;
2598 			/*
2599 			 * produce a msg made of:
2600 			 *  struct rt_msghdr;
2601 			 *  struct sockaddr_in6 (IPv6)
2602 			 *  struct sockaddr_dl;
2603 			 */
2604 			bzero(&ndpc, sizeof(ndpc));
2605 			ndpc.rtm.rtm_msglen = sizeof(ndpc);
2606 			ndpc.rtm.rtm_version = RTM_VERSION;
2607 			ndpc.rtm.rtm_type = RTM_GET;
2608 			ndpc.rtm.rtm_flags = RTF_UP;
2609 			ndpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
2610 			ndpc.sin6.sin6_family = AF_INET6;
2611 			ndpc.sin6.sin6_len = sizeof(ndpc.sin6);
2612 			bcopy(L3_ADDR(lle), &ndpc.sin6, L3_ADDR_LEN(lle));
2613 
2614 			/* publish */
2615 			if (lle->la_flags & LLE_PUB)
2616 				ndpc.rtm.rtm_flags |= RTF_ANNOUNCE;
2617 
2618 			sdl = &ndpc.sdl;
2619 			sdl->sdl_family = AF_LINK;
2620 			sdl->sdl_len = sizeof(*sdl);
2621 			sdl->sdl_alen = ifp->if_addrlen;
2622 			sdl->sdl_index = ifp->if_index;
2623 			sdl->sdl_type = ifp->if_type;
2624 			bcopy(&lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
2625 			ndpc.rtm.rtm_rmx.rmx_expire =
2626 			    lle->la_flags & LLE_STATIC ? 0 : lle->la_expire;
2627 			ndpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
2628 			if (lle->la_flags & LLE_STATIC)
2629 				ndpc.rtm.rtm_flags |= RTF_STATIC;
2630 			ndpc.rtm.rtm_index = ifp->if_index;
2631 			error = SYSCTL_OUT(wr, &ndpc, sizeof(ndpc));
2632 			if (error)
2633 				break;
2634 		}
2635 	}
2636 	return error;
2637 }
2638 
2639 void *
2640 in6_domifattach(struct ifnet *ifp)
2641 {
2642 	struct in6_ifextra *ext;
2643 
2644 	ext = (struct in6_ifextra *)malloc(sizeof(*ext), M_IFADDR, M_WAITOK);
2645 	bzero(ext, sizeof(*ext));
2646 
2647 	ext->in6_ifstat = (struct in6_ifstat *)malloc(sizeof(struct in6_ifstat),
2648 	    M_IFADDR, M_WAITOK);
2649 	bzero(ext->in6_ifstat, sizeof(*ext->in6_ifstat));
2650 
2651 	ext->icmp6_ifstat =
2652 	    (struct icmp6_ifstat *)malloc(sizeof(struct icmp6_ifstat),
2653 	    M_IFADDR, M_WAITOK);
2654 	bzero(ext->icmp6_ifstat, sizeof(*ext->icmp6_ifstat));
2655 
2656 	ext->nd_ifinfo = nd6_ifattach(ifp);
2657 	ext->scope6_id = scope6_ifattach(ifp);
2658 	ext->lltable = lltable_init(ifp, AF_INET6);
2659 	if (ext->lltable != NULL) {
2660 		ext->lltable->llt_free = in6_lltable_free;
2661 		ext->lltable->llt_prefix_free = in6_lltable_prefix_free;
2662 		ext->lltable->llt_lookup = in6_lltable_lookup;
2663 		ext->lltable->llt_dump = in6_lltable_dump;
2664 	}
2665 
2666 	ext->mld_ifinfo = mld_domifattach(ifp);
2667 
2668 	return ext;
2669 }
2670 
2671 void
2672 in6_domifdetach(struct ifnet *ifp, void *aux)
2673 {
2674 	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
2675 
2676 	mld_domifdetach(ifp);
2677 	scope6_ifdetach(ext->scope6_id);
2678 	nd6_ifdetach(ext->nd_ifinfo);
2679 	lltable_free(ext->lltable);
2680 	free(ext->in6_ifstat, M_IFADDR);
2681 	free(ext->icmp6_ifstat, M_IFADDR);
2682 	free(ext, M_IFADDR);
2683 }
2684 
2685 /*
2686  * Convert sockaddr_in6 to sockaddr_in.  Original sockaddr_in6 must be
2687  * v4 mapped addr or v4 compat addr
2688  */
2689 void
2690 in6_sin6_2_sin(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2691 {
2692 
2693 	bzero(sin, sizeof(*sin));
2694 	sin->sin_len = sizeof(struct sockaddr_in);
2695 	sin->sin_family = AF_INET;
2696 	sin->sin_port = sin6->sin6_port;
2697 	sin->sin_addr.s_addr = sin6->sin6_addr.s6_addr32[3];
2698 }
2699 
2700 /* Convert sockaddr_in to sockaddr_in6 in v4 mapped addr format. */
2701 void
2702 in6_sin_2_v4mapsin6(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2703 {
2704 	bzero(sin6, sizeof(*sin6));
2705 	sin6->sin6_len = sizeof(struct sockaddr_in6);
2706 	sin6->sin6_family = AF_INET6;
2707 	sin6->sin6_port = sin->sin_port;
2708 	sin6->sin6_addr.s6_addr32[0] = 0;
2709 	sin6->sin6_addr.s6_addr32[1] = 0;
2710 	sin6->sin6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
2711 	sin6->sin6_addr.s6_addr32[3] = sin->sin_addr.s_addr;
2712 }
2713 
2714 /* Convert sockaddr_in6 into sockaddr_in. */
2715 void
2716 in6_sin6_2_sin_in_sock(struct sockaddr *nam)
2717 {
2718 	struct sockaddr_in *sin_p;
2719 	struct sockaddr_in6 sin6;
2720 
2721 	/*
2722 	 * Save original sockaddr_in6 addr and convert it
2723 	 * to sockaddr_in.
2724 	 */
2725 	sin6 = *(struct sockaddr_in6 *)nam;
2726 	sin_p = (struct sockaddr_in *)nam;
2727 	in6_sin6_2_sin(sin_p, &sin6);
2728 }
2729 
2730 /* Convert sockaddr_in into sockaddr_in6 in v4 mapped addr format. */
2731 void
2732 in6_sin_2_v4mapsin6_in_sock(struct sockaddr **nam)
2733 {
2734 	struct sockaddr_in *sin_p;
2735 	struct sockaddr_in6 *sin6_p;
2736 
2737 	sin6_p = malloc(sizeof *sin6_p, M_SONAME,
2738 	       M_WAITOK);
2739 	sin_p = (struct sockaddr_in *)*nam;
2740 	in6_sin_2_v4mapsin6(sin_p, sin6_p);
2741 	free(*nam, M_SONAME);
2742 	*nam = (struct sockaddr *)sin6_p;
2743 }
2744