xref: /netbsd-src/sys/netinet6/in6.c (revision 946379e7b37692fc43f68eb0d1c10daa0a7f3b6c)
1 /*	$NetBSD: in6.c,v 1.194 2015/12/12 23:34:25 christos Exp $	*/
2 /*	$KAME: in6.c,v 1.198 2001/07/18 09:12:38 itojun 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 <sys/cdefs.h>
65 __KERNEL_RCSID(0, "$NetBSD: in6.c,v 1.194 2015/12/12 23:34:25 christos Exp $");
66 
67 #ifdef _KERNEL_OPT
68 #include "opt_inet.h"
69 #include "opt_compat_netbsd.h"
70 #endif
71 
72 #include <sys/param.h>
73 #include <sys/ioctl.h>
74 #include <sys/errno.h>
75 #include <sys/malloc.h>
76 #include <sys/socket.h>
77 #include <sys/socketvar.h>
78 #include <sys/sockio.h>
79 #include <sys/systm.h>
80 #include <sys/proc.h>
81 #include <sys/time.h>
82 #include <sys/kernel.h>
83 #include <sys/syslog.h>
84 #include <sys/kauth.h>
85 #include <sys/cprng.h>
86 #include <sys/kmem.h>
87 
88 #include <net/if.h>
89 #include <net/if_types.h>
90 #include <net/if_llatbl.h>
91 #include <net/if_ether.h>
92 #include <net/if_dl.h>
93 #include <net/pfil.h>
94 #include <net/route.h>
95 
96 #include <netinet/in.h>
97 #include <netinet/in_var.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 
107 #include <net/net_osdep.h>
108 
109 #ifdef COMPAT_50
110 #include <compat/netinet6/in6_var.h>
111 #endif
112 
113 MALLOC_DEFINE(M_IP6OPT, "ip6_options", "IPv6 options");
114 
115 /* enable backward compatibility code for obsoleted ioctls */
116 #define COMPAT_IN6IFIOCTL
117 
118 #ifdef	IN6_DEBUG
119 #define	IN6_DPRINTF(__fmt, ...)	printf(__fmt, __VA_ARGS__)
120 #else
121 #define	IN6_DPRINTF(__fmt, ...)	do { } while (/*CONSTCOND*/0)
122 #endif /* IN6_DEBUG */
123 
124 /*
125  * Definitions of some constant IP6 addresses.
126  */
127 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
128 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
129 const struct in6_addr in6addr_nodelocal_allnodes =
130 	IN6ADDR_NODELOCAL_ALLNODES_INIT;
131 const struct in6_addr in6addr_linklocal_allnodes =
132 	IN6ADDR_LINKLOCAL_ALLNODES_INIT;
133 const struct in6_addr in6addr_linklocal_allrouters =
134 	IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
135 
136 const struct in6_addr in6mask0 = IN6MASK0;
137 const struct in6_addr in6mask32 = IN6MASK32;
138 const struct in6_addr in6mask64 = IN6MASK64;
139 const struct in6_addr in6mask96 = IN6MASK96;
140 const struct in6_addr in6mask128 = IN6MASK128;
141 
142 const struct sockaddr_in6 sa6_any = {sizeof(sa6_any), AF_INET6,
143 				     0, 0, IN6ADDR_ANY_INIT, 0};
144 
145 static int in6_lifaddr_ioctl(struct socket *, u_long, void *,
146 	struct ifnet *);
147 static int in6_ifinit(struct ifnet *, struct in6_ifaddr *,
148 	const struct sockaddr_in6 *, int);
149 static void in6_unlink_ifa(struct in6_ifaddr *, struct ifnet *);
150 
151 /*
152  * Add ownaddr as loopback rtentry.  We previously add the route only if
153  * necessary (ex. on a p2p link).  However, since we now manage addresses
154  * separately from prefixes, we should always add the route.  We can't
155  * rely on the cloning mechanism from the corresponding interface route
156  * any more.
157  */
158 void
159 in6_ifaddlocal(struct ifaddr *ifa)
160 {
161 
162 	if (IN6_ARE_ADDR_EQUAL(IFA_IN6(ifa), &in6addr_any) ||
163 	    (ifa->ifa_ifp->if_flags & IFF_POINTOPOINT &&
164 	    IN6_ARE_ADDR_EQUAL(IFA_IN6(ifa), IFA_DSTIN6(ifa))))
165 	{
166 		rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
167 		return;
168 	}
169 
170 	rt_ifa_addlocal(ifa);
171 }
172 
173 /*
174  * Remove loopback rtentry of ownaddr generated by in6_ifaddlocal(),
175  * if it exists.
176  */
177 void
178 in6_ifremlocal(struct ifaddr *ifa)
179 {
180 	struct in6_ifaddr *ia;
181 	struct ifaddr *alt_ifa = NULL;
182 	int ia_count = 0;
183 
184 	/*
185 	 * Some of BSD variants do not remove cloned routes
186 	 * from an interface direct route, when removing the direct route
187 	 * (see comments in net/net_osdep.h).  Even for variants that do remove
188 	 * cloned routes, they could fail to remove the cloned routes when
189 	 * we handle multple addresses that share a common prefix.
190 	 * So, we should remove the route corresponding to the deleted address.
191 	 */
192 
193 	/*
194 	 * Delete the entry only if exactly one ifaddr matches the
195 	 * address, ifa->ifa_addr.
196 	 *
197 	 * If more than one ifaddr matches, replace the ifaddr in
198 	 * the routing table, rt_ifa, with a different ifaddr than
199 	 * the one we are purging, ifa.  It is important to do
200 	 * this, or else the routing table can accumulate dangling
201 	 * pointers rt->rt_ifa->ifa_ifp to destroyed interfaces,
202 	 * which will lead to crashes, later.  (More than one ifaddr
203 	 * can match if we assign the same address to multiple---probably
204 	 * p2p---interfaces.)
205 	 *
206 	 * XXX An old comment at this place said, "we should avoid
207 	 * XXX such a configuration [i.e., interfaces with the same
208 	 * XXX addressed assigned --ed.] in IPv6...".  I do not
209 	 * XXX agree, especially now that I have fixed the dangling
210 	 * XXX ifp-pointers bug.
211 	 */
212 	for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
213 		if (!IN6_ARE_ADDR_EQUAL(IFA_IN6(ifa), &ia->ia_addr.sin6_addr))
214 			continue;
215 		if (ia->ia_ifp != ifa->ifa_ifp)
216 			alt_ifa = &ia->ia_ifa;
217 		if (++ia_count > 1 && alt_ifa != NULL)
218 			break;
219 	}
220 
221 	if (ia_count == 0)
222 		return;
223 
224 	rt_ifa_remlocal(ifa, ia_count == 1 ? NULL : alt_ifa);
225 }
226 
227 int
228 in6_mask2len(struct in6_addr *mask, u_char *lim0)
229 {
230 	int x = 0, y;
231 	u_char *lim = lim0, *p;
232 
233 	/* ignore the scope_id part */
234 	if (lim0 == NULL || lim0 - (u_char *)mask > sizeof(*mask))
235 		lim = (u_char *)mask + sizeof(*mask);
236 	for (p = (u_char *)mask; p < lim; x++, p++) {
237 		if (*p != 0xff)
238 			break;
239 	}
240 	y = 0;
241 	if (p < lim) {
242 		for (y = 0; y < NBBY; y++) {
243 			if ((*p & (0x80 >> y)) == 0)
244 				break;
245 		}
246 	}
247 
248 	/*
249 	 * when the limit pointer is given, do a stricter check on the
250 	 * remaining bits.
251 	 */
252 	if (p < lim) {
253 		if (y != 0 && (*p & (0x00ff >> y)) != 0)
254 			return -1;
255 		for (p = p + 1; p < lim; p++)
256 			if (*p != 0)
257 				return -1;
258 	}
259 
260 	return x * NBBY + y;
261 }
262 
263 #define ifa2ia6(ifa)	((struct in6_ifaddr *)(ifa))
264 #define ia62ifa(ia6)	(&((ia6)->ia_ifa))
265 
266 static int
267 in6_control1(struct socket *so, u_long cmd, void *data, struct ifnet *ifp)
268 {
269 	struct	in6_ifreq *ifr = (struct in6_ifreq *)data;
270 	struct	in6_ifaddr *ia = NULL;
271 	struct	in6_aliasreq *ifra = (struct in6_aliasreq *)data;
272 	struct sockaddr_in6 *sa6;
273 	int error;
274 
275 	switch (cmd) {
276 	case SIOCAADDRCTL_POLICY:
277 	case SIOCDADDRCTL_POLICY:
278 		/* Privileged. */
279 		return in6_src_ioctl(cmd, data);
280 	/*
281 	 * XXX: Fix me, once we fix SIOCSIFADDR, SIOCIFDSTADDR, etc.
282 	 */
283 	case SIOCSIFADDR:
284 	case SIOCSIFDSTADDR:
285 	case SIOCSIFBRDADDR:
286 	case SIOCSIFNETMASK:
287 		return EOPNOTSUPP;
288 	case SIOCGETSGCNT_IN6:
289 	case SIOCGETMIFCNT_IN6:
290 		return mrt6_ioctl(cmd, data);
291 	case SIOCGIFADDRPREF:
292 	case SIOCSIFADDRPREF:
293 		if (ifp == NULL)
294 			return EINVAL;
295 		return ifaddrpref_ioctl(so, cmd, data, ifp);
296 	}
297 
298 	if (ifp == NULL)
299 		return EOPNOTSUPP;
300 
301 	switch (cmd) {
302 	case SIOCSNDFLUSH_IN6:
303 	case SIOCSPFXFLUSH_IN6:
304 	case SIOCSRTRFLUSH_IN6:
305 	case SIOCSDEFIFACE_IN6:
306 	case SIOCSIFINFO_FLAGS:
307 	case SIOCSIFINFO_IN6:
308 		/* Privileged. */
309 		/* FALLTHROUGH */
310 	case OSIOCGIFINFO_IN6:
311 	case SIOCGIFINFO_IN6:
312 	case SIOCGDRLST_IN6:
313 	case SIOCGPRLST_IN6:
314 	case SIOCGNBRINFO_IN6:
315 	case SIOCGDEFIFACE_IN6:
316 		return nd6_ioctl(cmd, data, ifp);
317 	}
318 
319 	switch (cmd) {
320 	case SIOCSIFPREFIX_IN6:
321 	case SIOCDIFPREFIX_IN6:
322 	case SIOCAIFPREFIX_IN6:
323 	case SIOCCIFPREFIX_IN6:
324 	case SIOCSGIFPREFIX_IN6:
325 	case SIOCGIFPREFIX_IN6:
326 		log(LOG_NOTICE,
327 		    "prefix ioctls are now invalidated. "
328 		    "please use ifconfig.\n");
329 		return EOPNOTSUPP;
330 	}
331 
332 	switch (cmd) {
333 	case SIOCALIFADDR:
334 	case SIOCDLIFADDR:
335 		/* Privileged. */
336 		/* FALLTHROUGH */
337 	case SIOCGLIFADDR:
338 		return in6_lifaddr_ioctl(so, cmd, data, ifp);
339 	}
340 
341 	/*
342 	 * Find address for this interface, if it exists.
343 	 *
344 	 * In netinet code, we have checked ifra_addr in SIOCSIF*ADDR operation
345 	 * only, and used the first interface address as the target of other
346 	 * operations (without checking ifra_addr).  This was because netinet
347 	 * code/API assumed at most 1 interface address per interface.
348 	 * Since IPv6 allows a node to assign multiple addresses
349 	 * on a single interface, we almost always look and check the
350 	 * presence of ifra_addr, and reject invalid ones here.
351 	 * It also decreases duplicated code among SIOC*_IN6 operations.
352 	 */
353 	switch (cmd) {
354 	case SIOCAIFADDR_IN6:
355 #ifdef OSIOCAIFADDR_IN6
356 	case OSIOCAIFADDR_IN6:
357 #endif
358 #ifdef OSIOCSIFPHYADDR_IN6
359 	case OSIOCSIFPHYADDR_IN6:
360 #endif
361 	case SIOCSIFPHYADDR_IN6:
362 		sa6 = &ifra->ifra_addr;
363 		break;
364 	case SIOCSIFADDR_IN6:
365 	case SIOCGIFADDR_IN6:
366 	case SIOCSIFDSTADDR_IN6:
367 	case SIOCSIFNETMASK_IN6:
368 	case SIOCGIFDSTADDR_IN6:
369 	case SIOCGIFNETMASK_IN6:
370 	case SIOCDIFADDR_IN6:
371 	case SIOCGIFPSRCADDR_IN6:
372 	case SIOCGIFPDSTADDR_IN6:
373 	case SIOCGIFAFLAG_IN6:
374 	case SIOCSNDFLUSH_IN6:
375 	case SIOCSPFXFLUSH_IN6:
376 	case SIOCSRTRFLUSH_IN6:
377 	case SIOCGIFALIFETIME_IN6:
378 #ifdef OSIOCGIFALIFETIME_IN6
379 	case OSIOCGIFALIFETIME_IN6:
380 #endif
381 	case SIOCGIFSTAT_IN6:
382 	case SIOCGIFSTAT_ICMP6:
383 		sa6 = &ifr->ifr_addr;
384 		break;
385 	default:
386 		sa6 = NULL;
387 		break;
388 	}
389 	if (sa6 && sa6->sin6_family == AF_INET6) {
390 		if (sa6->sin6_scope_id != 0)
391 			error = sa6_embedscope(sa6, 0);
392 		else
393 			error = in6_setscope(&sa6->sin6_addr, ifp, NULL);
394 		if (error != 0)
395 			return error;
396 		ia = in6ifa_ifpwithaddr(ifp, &sa6->sin6_addr);
397 	} else
398 		ia = NULL;
399 
400 	switch (cmd) {
401 	case SIOCSIFADDR_IN6:
402 	case SIOCSIFDSTADDR_IN6:
403 	case SIOCSIFNETMASK_IN6:
404 		/*
405 		 * Since IPv6 allows a node to assign multiple addresses
406 		 * on a single interface, SIOCSIFxxx ioctls are deprecated.
407 		 */
408 		return EINVAL;
409 
410 	case SIOCDIFADDR_IN6:
411 		/*
412 		 * for IPv4, we look for existing in_ifaddr here to allow
413 		 * "ifconfig if0 delete" to remove the first IPv4 address on
414 		 * the interface.  For IPv6, as the spec allows multiple
415 		 * interface address from the day one, we consider "remove the
416 		 * first one" semantics to be not preferable.
417 		 */
418 		if (ia == NULL)
419 			return EADDRNOTAVAIL;
420 		/* FALLTHROUGH */
421 #ifdef OSIOCAIFADDR_IN6
422 	case OSIOCAIFADDR_IN6:
423 #endif
424 	case SIOCAIFADDR_IN6:
425 		/*
426 		 * We always require users to specify a valid IPv6 address for
427 		 * the corresponding operation.
428 		 */
429 		if (ifra->ifra_addr.sin6_family != AF_INET6 ||
430 		    ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6))
431 			return EAFNOSUPPORT;
432 		/* Privileged. */
433 
434 		break;
435 
436 	case SIOCGIFADDR_IN6:
437 		/* This interface is basically deprecated. use SIOCGIFCONF. */
438 		/* FALLTHROUGH */
439 	case SIOCGIFAFLAG_IN6:
440 	case SIOCGIFNETMASK_IN6:
441 	case SIOCGIFDSTADDR_IN6:
442 	case SIOCGIFALIFETIME_IN6:
443 #ifdef OSIOCGIFALIFETIME_IN6
444 	case OSIOCGIFALIFETIME_IN6:
445 #endif
446 		/* must think again about its semantics */
447 		if (ia == NULL)
448 			return EADDRNOTAVAIL;
449 		break;
450 	}
451 
452 	switch (cmd) {
453 
454 	case SIOCGIFADDR_IN6:
455 		ifr->ifr_addr = ia->ia_addr;
456 		if ((error = sa6_recoverscope(&ifr->ifr_addr)) != 0)
457 			return error;
458 		break;
459 
460 	case SIOCGIFDSTADDR_IN6:
461 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
462 			return EINVAL;
463 		/*
464 		 * XXX: should we check if ifa_dstaddr is NULL and return
465 		 * an error?
466 		 */
467 		ifr->ifr_dstaddr = ia->ia_dstaddr;
468 		if ((error = sa6_recoverscope(&ifr->ifr_dstaddr)) != 0)
469 			return error;
470 		break;
471 
472 	case SIOCGIFNETMASK_IN6:
473 		ifr->ifr_addr = ia->ia_prefixmask;
474 		break;
475 
476 	case SIOCGIFAFLAG_IN6:
477 		ifr->ifr_ifru.ifru_flags6 = ia->ia6_flags;
478 		break;
479 
480 	case SIOCGIFSTAT_IN6:
481 		if (ifp == NULL)
482 			return EINVAL;
483 		memset(&ifr->ifr_ifru.ifru_stat, 0,
484 		    sizeof(ifr->ifr_ifru.ifru_stat));
485 		ifr->ifr_ifru.ifru_stat =
486 		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->in6_ifstat;
487 		break;
488 
489 	case SIOCGIFSTAT_ICMP6:
490 		if (ifp == NULL)
491 			return EINVAL;
492 		memset(&ifr->ifr_ifru.ifru_icmp6stat, 0,
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 #ifdef OSIOCGIFALIFETIME_IN6
499 	case OSIOCGIFALIFETIME_IN6:
500 #endif
501 	case SIOCGIFALIFETIME_IN6:
502 		ifr->ifr_ifru.ifru_lifetime = ia->ia6_lifetime;
503 		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
504 			time_t maxexpire;
505 			struct in6_addrlifetime *retlt =
506 			    &ifr->ifr_ifru.ifru_lifetime;
507 
508 			/*
509 			 * XXX: adjust expiration time assuming time_t is
510 			 * signed.
511 			 */
512 			maxexpire = ((time_t)~0) &
513 			    ~((time_t)1 << ((sizeof(maxexpire) * NBBY) - 1));
514 			if (ia->ia6_lifetime.ia6t_vltime <
515 			    maxexpire - ia->ia6_updatetime) {
516 				retlt->ia6t_expire = ia->ia6_updatetime +
517 				    ia->ia6_lifetime.ia6t_vltime;
518 				retlt->ia6t_expire = retlt->ia6t_expire ?
519 				    time_mono_to_wall(retlt->ia6t_expire) :
520 				    0;
521 			} else
522 				retlt->ia6t_expire = maxexpire;
523 		}
524 		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
525 			time_t maxexpire;
526 			struct in6_addrlifetime *retlt =
527 			    &ifr->ifr_ifru.ifru_lifetime;
528 
529 			/*
530 			 * XXX: adjust expiration time assuming time_t is
531 			 * signed.
532 			 */
533 			maxexpire = ((time_t)~0) &
534 			    ~((time_t)1 << ((sizeof(maxexpire) * NBBY) - 1));
535 			if (ia->ia6_lifetime.ia6t_pltime <
536 			    maxexpire - ia->ia6_updatetime) {
537 				retlt->ia6t_preferred = ia->ia6_updatetime +
538 				    ia->ia6_lifetime.ia6t_pltime;
539 				retlt->ia6t_preferred = retlt->ia6t_preferred ?
540 				    time_mono_to_wall(retlt->ia6t_preferred) :
541 				    0;
542 			} else
543 				retlt->ia6t_preferred = maxexpire;
544 		}
545 #ifdef OSIOCFIFALIFETIME_IN6
546 		if (cmd == OSIOCFIFALIFETIME_IN6)
547 			in6_addrlifetime_to_in6_addrlifetime50(
548 			    &ifr->ifru.ifru_lifetime);
549 #endif
550 		break;
551 
552 #ifdef OSIOCAIFADDR_IN6
553 	case OSIOCAIFADDR_IN6:
554 		in6_aliasreq50_to_in6_aliasreq(ifra);
555 		/*FALLTHROUGH*/
556 #endif
557 	case SIOCAIFADDR_IN6:
558 	{
559 		int i;
560 		struct nd_prefixctl prc0;
561 		struct nd_prefix *pr;
562 		struct in6_addrlifetime *lt;
563 
564 		/* reject read-only flags */
565 		if ((ifra->ifra_flags & IN6_IFF_DUPLICATED) != 0 ||
566 		    (ifra->ifra_flags & IN6_IFF_DETACHED) != 0 ||
567 		    (ifra->ifra_flags & IN6_IFF_TENTATIVE) != 0 ||
568 		    (ifra->ifra_flags & IN6_IFF_NODAD) != 0 ||
569 		    (ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0) {
570 			return EINVAL;
571 		}
572 		/*
573 		 * ia6t_expire and ia6t_preferred won't be used for now,
574 		 * so just in case.
575 		 */
576 		lt = &ifra->ifra_lifetime;
577 		if (lt->ia6t_expire != 0)
578 			lt->ia6t_expire = time_wall_to_mono(lt->ia6t_expire);
579 		if (lt->ia6t_preferred != 0)
580 			lt->ia6t_preferred =
581 			    time_wall_to_mono(lt->ia6t_preferred);
582 		/*
583 		 * first, make or update the interface address structure,
584 		 * and link it to the list.
585 		 */
586 		if ((error = in6_update_ifa(ifp, ifra, ia, 0)) != 0)
587 			return error;
588 		if ((ia = in6ifa_ifpwithaddr(ifp, &ifra->ifra_addr.sin6_addr))
589 		    == NULL) {
590 		    	/*
591 			 * this can happen when the user specify the 0 valid
592 			 * lifetime.
593 			 */
594 			break;
595 		}
596 
597 		/*
598 		 * then, make the prefix on-link on the interface.
599 		 * XXX: we'd rather create the prefix before the address, but
600 		 * we need at least one address to install the corresponding
601 		 * interface route, so we configure the address first.
602 		 */
603 
604 		/*
605 		 * convert mask to prefix length (prefixmask has already
606 		 * been validated in in6_update_ifa().
607 		 */
608 		memset(&prc0, 0, sizeof(prc0));
609 		prc0.ndprc_ifp = ifp;
610 		prc0.ndprc_plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
611 		    NULL);
612 		if (prc0.ndprc_plen == 128) {
613 			break;	/* we don't need to install a host route. */
614 		}
615 		prc0.ndprc_prefix = ifra->ifra_addr;
616 		/* apply the mask for safety. */
617 		for (i = 0; i < 4; i++) {
618 			prc0.ndprc_prefix.sin6_addr.s6_addr32[i] &=
619 			    ifra->ifra_prefixmask.sin6_addr.s6_addr32[i];
620 		}
621 		/*
622 		 * XXX: since we don't have an API to set prefix (not address)
623 		 * lifetimes, we just use the same lifetimes as addresses.
624 		 * The (temporarily) installed lifetimes can be overridden by
625 		 * later advertised RAs (when accept_rtadv is non 0), which is
626 		 * an intended behavior.
627 		 */
628 		prc0.ndprc_raf_onlink = 1; /* should be configurable? */
629 		prc0.ndprc_raf_auto =
630 		    ((ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0);
631 		prc0.ndprc_vltime = ifra->ifra_lifetime.ia6t_vltime;
632 		prc0.ndprc_pltime = ifra->ifra_lifetime.ia6t_pltime;
633 
634 		/* add the prefix if not yet. */
635 		if ((pr = nd6_prefix_lookup(&prc0)) == NULL) {
636 			/*
637 			 * nd6_prelist_add will install the corresponding
638 			 * interface route.
639 			 */
640 			if ((error = nd6_prelist_add(&prc0, NULL, &pr)) != 0)
641 				return error;
642 			if (pr == NULL) {
643 				log(LOG_ERR, "nd6_prelist_add succeeded but "
644 				    "no prefix\n");
645 				return EINVAL; /* XXX panic here? */
646 			}
647 		}
648 
649 		/* relate the address to the prefix */
650 		if (ia->ia6_ndpr == NULL) {
651 			ia->ia6_ndpr = pr;
652 			pr->ndpr_refcnt++;
653 
654 			/*
655 			 * If this is the first autoconf address from the
656 			 * prefix, create a temporary address as well
657 			 * (when required).
658 			 */
659 			if ((ia->ia6_flags & IN6_IFF_AUTOCONF) &&
660 			    ip6_use_tempaddr && pr->ndpr_refcnt == 1) {
661 				int e;
662 				if ((e = in6_tmpifadd(ia, 1, 0)) != 0) {
663 					log(LOG_NOTICE, "in6_control: failed "
664 					    "to create a temporary address, "
665 					    "errno=%d\n", e);
666 				}
667 			}
668 		}
669 
670 		/*
671 		 * this might affect the status of autoconfigured addresses,
672 		 * that is, this address might make other addresses detached.
673 		 */
674 		pfxlist_onlink_check();
675 
676 		(void)pfil_run_hooks(if_pfil, (struct mbuf **)SIOCAIFADDR_IN6,
677 		    ifp, PFIL_IFADDR);
678 		break;
679 	}
680 
681 	case SIOCDIFADDR_IN6:
682 	{
683 		struct nd_prefix *pr;
684 
685 		/*
686 		 * If the address being deleted is the only one that owns
687 		 * the corresponding prefix, expire the prefix as well.
688 		 * XXX: theoretically, we don't have to worry about such
689 		 * relationship, since we separate the address management
690 		 * and the prefix management.  We do this, however, to provide
691 		 * as much backward compatibility as possible in terms of
692 		 * the ioctl operation.
693 		 * Note that in6_purgeaddr() will decrement ndpr_refcnt.
694 		 */
695 		pr = ia->ia6_ndpr;
696 		in6_purgeaddr(&ia->ia_ifa);
697 		if (pr && pr->ndpr_refcnt == 0)
698 			prelist_remove(pr);
699 		(void)pfil_run_hooks(if_pfil, (struct mbuf **)SIOCDIFADDR_IN6,
700 		    ifp, PFIL_IFADDR);
701 		break;
702 	}
703 
704 	default:
705 		return ENOTTY;
706 	}
707 
708 	return 0;
709 }
710 
711 int
712 in6_control(struct socket *so, u_long cmd, void *data, struct ifnet *ifp)
713 {
714 	int error, s;
715 
716 	switch (cmd) {
717 	case SIOCSNDFLUSH_IN6:
718 	case SIOCSPFXFLUSH_IN6:
719 	case SIOCSRTRFLUSH_IN6:
720 	case SIOCSDEFIFACE_IN6:
721 	case SIOCSIFINFO_FLAGS:
722 	case SIOCSIFINFO_IN6:
723 
724 	case SIOCALIFADDR:
725 	case SIOCDLIFADDR:
726 
727 	case SIOCDIFADDR_IN6:
728 #ifdef OSIOCAIFADDR_IN6
729 	case OSIOCAIFADDR_IN6:
730 #endif
731 	case SIOCAIFADDR_IN6:
732 
733 	case SIOCAADDRCTL_POLICY:
734 	case SIOCDADDRCTL_POLICY:
735 
736 		if (kauth_authorize_network(curlwp->l_cred,
737 		    KAUTH_NETWORK_SOCKET,
738 		    KAUTH_REQ_NETWORK_SOCKET_SETPRIV,
739 		    so, NULL, NULL))
740 			return EPERM;
741 		break;
742 	}
743 
744 	s = splnet();
745 	error = in6_control1(so , cmd, data, ifp);
746 	splx(s);
747 	return error;
748 }
749 
750 /*
751  * Update parameters of an IPv6 interface address.
752  * If necessary, a new entry is created and linked into address chains.
753  * This function is separated from in6_control().
754  * XXX: should this be performed under splnet()?
755  */
756 static int
757 in6_update_ifa1(struct ifnet *ifp, struct in6_aliasreq *ifra,
758     struct in6_ifaddr *ia, int flags)
759 {
760 	int error = 0, hostIsNew = 0, plen = -1;
761 	struct in6_ifaddr *oia;
762 	struct sockaddr_in6 dst6;
763 	struct in6_addrlifetime *lt;
764 	struct in6_multi_mship *imm;
765 	struct in6_multi *in6m_sol;
766 	struct rtentry *rt;
767 	int dad_delay, was_tentative;
768 
769 	in6m_sol = NULL;
770 
771 	/* Validate parameters */
772 	if (ifp == NULL || ifra == NULL) /* this maybe redundant */
773 		return EINVAL;
774 
775 	/*
776 	 * The destination address for a p2p link must have a family
777 	 * of AF_UNSPEC or AF_INET6.
778 	 */
779 	if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
780 	    ifra->ifra_dstaddr.sin6_family != AF_INET6 &&
781 	    ifra->ifra_dstaddr.sin6_family != AF_UNSPEC)
782 		return EAFNOSUPPORT;
783 	/*
784 	 * validate ifra_prefixmask.  don't check sin6_family, netmask
785 	 * does not carry fields other than sin6_len.
786 	 */
787 	if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6))
788 		return EINVAL;
789 	/*
790 	 * Because the IPv6 address architecture is classless, we require
791 	 * users to specify a (non 0) prefix length (mask) for a new address.
792 	 * We also require the prefix (when specified) mask is valid, and thus
793 	 * reject a non-consecutive mask.
794 	 */
795 	if (ia == NULL && ifra->ifra_prefixmask.sin6_len == 0)
796 		return EINVAL;
797 	if (ifra->ifra_prefixmask.sin6_len != 0) {
798 		plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
799 		    (u_char *)&ifra->ifra_prefixmask +
800 		    ifra->ifra_prefixmask.sin6_len);
801 		if (plen <= 0)
802 			return EINVAL;
803 	} else {
804 		/*
805 		 * In this case, ia must not be NULL.  We just use its prefix
806 		 * length.
807 		 */
808 		plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
809 	}
810 	/*
811 	 * If the destination address on a p2p interface is specified,
812 	 * and the address is a scoped one, validate/set the scope
813 	 * zone identifier.
814 	 */
815 	dst6 = ifra->ifra_dstaddr;
816 	if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) != 0 &&
817 	    (dst6.sin6_family == AF_INET6)) {
818 		struct in6_addr in6_tmp;
819 		u_int32_t zoneid;
820 
821 		in6_tmp = dst6.sin6_addr;
822 		if (in6_setscope(&in6_tmp, ifp, &zoneid))
823 			return EINVAL; /* XXX: should be impossible */
824 
825 		if (dst6.sin6_scope_id != 0) {
826 			if (dst6.sin6_scope_id != zoneid)
827 				return EINVAL;
828 		} else		/* user omit to specify the ID. */
829 			dst6.sin6_scope_id = zoneid;
830 
831 		/* convert into the internal form */
832 		if (sa6_embedscope(&dst6, 0))
833 			return EINVAL; /* XXX: should be impossible */
834 	}
835 	/*
836 	 * The destination address can be specified only for a p2p or a
837 	 * loopback interface.  If specified, the corresponding prefix length
838 	 * must be 128.
839 	 */
840 	if (ifra->ifra_dstaddr.sin6_family == AF_INET6) {
841 #ifdef FORCE_P2PPLEN
842 		int i;
843 #endif
844 
845 		if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) == 0) {
846 			/* XXX: noisy message */
847 			nd6log((LOG_INFO, "in6_update_ifa: a destination can "
848 			    "be specified for a p2p or a loopback IF only\n"));
849 			return EINVAL;
850 		}
851 		if (plen != 128) {
852 			nd6log((LOG_INFO, "in6_update_ifa: prefixlen should "
853 			    "be 128 when dstaddr is specified\n"));
854 #ifdef FORCE_P2PPLEN
855 			/*
856 			 * To be compatible with old configurations,
857 			 * such as ifconfig gif0 inet6 2001::1 2001::2
858 			 * prefixlen 126, we override the specified
859 			 * prefixmask as if the prefix length was 128.
860 			 */
861 			ifra->ifra_prefixmask.sin6_len =
862 			    sizeof(struct sockaddr_in6);
863 			for (i = 0; i < 4; i++)
864 				ifra->ifra_prefixmask.sin6_addr.s6_addr32[i] =
865 				    0xffffffff;
866 			plen = 128;
867 #else
868 			return EINVAL;
869 #endif
870 		}
871 	}
872 	/* lifetime consistency check */
873 	lt = &ifra->ifra_lifetime;
874 	if (lt->ia6t_pltime > lt->ia6t_vltime)
875 		return EINVAL;
876 	if (lt->ia6t_vltime == 0) {
877 		/*
878 		 * the following log might be noisy, but this is a typical
879 		 * configuration mistake or a tool's bug.
880 		 */
881 		nd6log((LOG_INFO,
882 		    "in6_update_ifa: valid lifetime is 0 for %s\n",
883 		    ip6_sprintf(&ifra->ifra_addr.sin6_addr)));
884 
885 		if (ia == NULL)
886 			return 0; /* there's nothing to do */
887 	}
888 
889 	/*
890 	 * If this is a new address, allocate a new ifaddr and link it
891 	 * into chains.
892 	 */
893 	if (ia == NULL) {
894 		hostIsNew = 1;
895 		/*
896 		 * When in6_update_ifa() is called in a process of a received
897 		 * RA, it is called under an interrupt context.  So, we should
898 		 * call malloc with M_NOWAIT.
899 		 */
900 		ia = (struct in6_ifaddr *) malloc(sizeof(*ia), M_IFADDR,
901 		    M_NOWAIT);
902 		if (ia == NULL)
903 			return ENOBUFS;
904 		memset(ia, 0, sizeof(*ia));
905 		LIST_INIT(&ia->ia6_memberships);
906 		/* Initialize the address and masks, and put time stamp */
907 		ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
908 		ia->ia_addr.sin6_family = AF_INET6;
909 		ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
910 		ia->ia6_createtime = time_uptime;
911 		if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
912 			/*
913 			 * XXX: some functions expect that ifa_dstaddr is not
914 			 * NULL for p2p interfaces.
915 			 */
916 			ia->ia_ifa.ifa_dstaddr =
917 			    (struct sockaddr *)&ia->ia_dstaddr;
918 		} else {
919 			ia->ia_ifa.ifa_dstaddr = NULL;
920 		}
921 		ia->ia_ifa.ifa_netmask =
922 		    (struct sockaddr *)&ia->ia_prefixmask;
923 
924 		ia->ia_ifp = ifp;
925 		if ((oia = in6_ifaddr) != NULL) {
926 			for ( ; oia->ia_next; oia = oia->ia_next)
927 				continue;
928 			oia->ia_next = ia;
929 		} else
930 			in6_ifaddr = ia;
931 		/* gain a refcnt for the link from in6_ifaddr */
932 		ifaref(&ia->ia_ifa);
933 
934 		ifa_insert(ifp, &ia->ia_ifa);
935 	}
936 
937 	/* update timestamp */
938 	ia->ia6_updatetime = time_uptime;
939 
940 	/* set prefix mask */
941 	if (ifra->ifra_prefixmask.sin6_len) {
942 		/*
943 		 * We prohibit changing the prefix length of an existing
944 		 * address, because
945 		 * + such an operation should be rare in IPv6, and
946 		 * + the operation would confuse prefix management.
947 		 */
948 		if (ia->ia_prefixmask.sin6_len &&
949 		    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) != plen) {
950 			nd6log((LOG_INFO, "in6_update_ifa: the prefix length of an"
951 			    " existing (%s) address should not be changed\n",
952 			    ip6_sprintf(&ia->ia_addr.sin6_addr)));
953 			error = EINVAL;
954 			goto unlink;
955 		}
956 		ia->ia_prefixmask = ifra->ifra_prefixmask;
957 	}
958 
959 	/*
960 	 * If a new destination address is specified, scrub the old one and
961 	 * install the new destination.  Note that the interface must be
962 	 * p2p or loopback (see the check above.)
963 	 */
964 	if (dst6.sin6_family == AF_INET6 &&
965 	    !IN6_ARE_ADDR_EQUAL(&dst6.sin6_addr, &ia->ia_dstaddr.sin6_addr)) {
966 		if ((ia->ia_flags & IFA_ROUTE) != 0 &&
967 		    rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST) != 0) {
968 			nd6log((LOG_ERR, "in6_update_ifa: failed to remove "
969 			    "a route to the old destination: %s\n",
970 			    ip6_sprintf(&ia->ia_addr.sin6_addr)));
971 			/* proceed anyway... */
972 		} else
973 			ia->ia_flags &= ~IFA_ROUTE;
974 		ia->ia_dstaddr = dst6;
975 	}
976 
977 	/*
978 	 * Set lifetimes.  We do not refer to ia6t_expire and ia6t_preferred
979 	 * to see if the address is deprecated or invalidated, but initialize
980 	 * these members for applications.
981 	 */
982 	ia->ia6_lifetime = ifra->ifra_lifetime;
983 	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
984 		ia->ia6_lifetime.ia6t_expire =
985 		    time_uptime + ia->ia6_lifetime.ia6t_vltime;
986 	} else
987 		ia->ia6_lifetime.ia6t_expire = 0;
988 	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
989 		ia->ia6_lifetime.ia6t_preferred =
990 		    time_uptime + ia->ia6_lifetime.ia6t_pltime;
991 	} else
992 		ia->ia6_lifetime.ia6t_preferred = 0;
993 
994 	/*
995 	 * configure address flags.
996 	 * We need to preserve tentative state so DAD works if
997 	 * something adds the same address before DAD finishes.
998 	 */
999 	was_tentative = ia->ia6_flags & (IN6_IFF_TENTATIVE|IN6_IFF_DUPLICATED);
1000 	ia->ia6_flags = ifra->ifra_flags;
1001 
1002 	/*
1003 	 * Make the address tentative before joining multicast addresses,
1004 	 * so that corresponding MLD responses would not have a tentative
1005 	 * source address.
1006 	 */
1007 	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/* safety */
1008 	if (ifp->if_link_state == LINK_STATE_DOWN) {
1009 		ia->ia6_flags |= IN6_IFF_DETACHED;
1010 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1011 	} else if ((hostIsNew || was_tentative) && if_do_dad(ifp))
1012 		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1013 
1014 	/*
1015 	 * backward compatibility - if IN6_IFF_DEPRECATED is set from the
1016 	 * userland, make it deprecated.
1017 	 */
1018 	if ((ifra->ifra_flags & IN6_IFF_DEPRECATED) != 0) {
1019 		ia->ia6_lifetime.ia6t_pltime = 0;
1020 		ia->ia6_lifetime.ia6t_preferred = time_uptime;
1021 	}
1022 
1023 	/* reset the interface and routing table appropriately. */
1024 	if ((error = in6_ifinit(ifp, ia, &ifra->ifra_addr, hostIsNew)) != 0)
1025 		goto unlink;
1026 	/*
1027 	 * We are done if we have simply modified an existing address.
1028 	 */
1029 	if (!hostIsNew)
1030 		return error;
1031 
1032 	/*
1033 	 * Beyond this point, we should call in6_purgeaddr upon an error,
1034 	 * not just go to unlink.
1035 	 */
1036 
1037 	/* join necessary multicast groups */
1038 	if ((ifp->if_flags & IFF_MULTICAST) != 0) {
1039 		struct sockaddr_in6 mltaddr, mltmask;
1040 		struct in6_addr llsol;
1041 
1042 		/* join solicited multicast addr for new host id */
1043 		memset(&llsol, 0, sizeof(struct in6_addr));
1044 		llsol.s6_addr16[0] = htons(0xff02);
1045 		llsol.s6_addr32[1] = 0;
1046 		llsol.s6_addr32[2] = htonl(1);
1047 		llsol.s6_addr32[3] = ifra->ifra_addr.sin6_addr.s6_addr32[3];
1048 		llsol.s6_addr8[12] = 0xff;
1049 		if ((error = in6_setscope(&llsol, ifp, NULL)) != 0) {
1050 			/* XXX: should not happen */
1051 			log(LOG_ERR, "in6_update_ifa: "
1052 			    "in6_setscope failed\n");
1053 			goto cleanup;
1054 		}
1055 		dad_delay = 0;
1056 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1057 			/*
1058 			 * We need a random delay for DAD on the address
1059 			 * being configured.  It also means delaying
1060 			 * transmission of the corresponding MLD report to
1061 			 * avoid report collision.
1062 			 * [draft-ietf-ipv6-rfc2462bis-02.txt]
1063 			 */
1064 			dad_delay = cprng_fast32() %
1065 			    (MAX_RTR_SOLICITATION_DELAY * hz);
1066 		}
1067 
1068 #define	MLTMASK_LEN  4	/* mltmask's masklen (=32bit=4octet) */
1069 		/* join solicited multicast addr for new host id */
1070 		imm = in6_joingroup(ifp, &llsol, &error, dad_delay);
1071 		if (!imm) {
1072 			nd6log((LOG_ERR,
1073 			    "in6_update_ifa: addmulti "
1074 			    "failed for %s on %s (errno=%d)\n",
1075 			    ip6_sprintf(&llsol), if_name(ifp), error));
1076 			goto cleanup;
1077 		}
1078 		LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1079 		in6m_sol = imm->i6mm_maddr;
1080 
1081 		sockaddr_in6_init(&mltmask, &in6mask32, 0, 0, 0);
1082 
1083 		/*
1084 		 * join link-local all-nodes address
1085 		 */
1086 		sockaddr_in6_init(&mltaddr, &in6addr_linklocal_allnodes,
1087 		    0, 0, 0);
1088 		if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) != 0)
1089 			goto cleanup; /* XXX: should not fail */
1090 
1091 		/*
1092 		 * XXX: do we really need this automatic routes?
1093 		 * We should probably reconsider this stuff.  Most applications
1094 		 * actually do not need the routes, since they usually specify
1095 		 * the outgoing interface.
1096 		 */
1097 		rt = rtalloc1((struct sockaddr *)&mltaddr, 0);
1098 		if (rt) {
1099 			if (memcmp(&mltaddr.sin6_addr,
1100 			    &satocsin6(rt_getkey(rt))->sin6_addr,
1101 			    MLTMASK_LEN)) {
1102 				rtfree(rt);
1103 				rt = NULL;
1104 			} else if (rt->rt_ifp != ifp) {
1105 				IN6_DPRINTF("%s: rt_ifp %p -> %p (%s) "
1106 				    "network %04x:%04x::/32 = %04x:%04x::/32\n",
1107 				    __func__, rt->rt_ifp, ifp, ifp->if_xname,
1108 				    ntohs(mltaddr.sin6_addr.s6_addr16[0]),
1109 				    ntohs(mltaddr.sin6_addr.s6_addr16[1]),
1110 				    satocsin6(rt_getkey(rt))->sin6_addr.s6_addr16[0],
1111 				    satocsin6(rt_getkey(rt))->sin6_addr.s6_addr16[1]);
1112 				rt_replace_ifa(rt, &ia->ia_ifa);
1113 				rt->rt_ifp = ifp;
1114 			}
1115 		}
1116 		if (!rt) {
1117 			struct rt_addrinfo info;
1118 
1119 			memset(&info, 0, sizeof(info));
1120 			info.rti_info[RTAX_DST] = (struct sockaddr *)&mltaddr;
1121 			info.rti_info[RTAX_GATEWAY] =
1122 			    (struct sockaddr *)&ia->ia_addr;
1123 			info.rti_info[RTAX_NETMASK] =
1124 			    (struct sockaddr *)&mltmask;
1125 			info.rti_info[RTAX_IFA] =
1126 			    (struct sockaddr *)&ia->ia_addr;
1127 			/* XXX: we need RTF_CLONING to fake nd6_rtrequest */
1128 			info.rti_flags = RTF_UP | RTF_CLONING;
1129 			error = rtrequest1(RTM_ADD, &info, NULL);
1130 			if (error)
1131 				goto cleanup;
1132 		} else {
1133 			rtfree(rt);
1134 		}
1135 		imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
1136 		if (!imm) {
1137 			nd6log((LOG_WARNING,
1138 			    "in6_update_ifa: addmulti failed for "
1139 			    "%s on %s (errno=%d)\n",
1140 			    ip6_sprintf(&mltaddr.sin6_addr),
1141 			    if_name(ifp), error));
1142 			goto cleanup;
1143 		}
1144 		LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1145 
1146 		/*
1147 		 * join node information group address
1148 		 */
1149 		dad_delay = 0;
1150 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1151 			/*
1152 			 * The spec doesn't say anything about delay for this
1153 			 * group, but the same logic should apply.
1154 			 */
1155 			dad_delay = cprng_fast32() %
1156 			    (MAX_RTR_SOLICITATION_DELAY * hz);
1157 		}
1158 		if (in6_nigroup(ifp, hostname, hostnamelen, &mltaddr) != 0)
1159 			;
1160 		else if ((imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error,
1161 		          dad_delay)) == NULL) { /* XXX jinmei */
1162 			nd6log((LOG_WARNING, "in6_update_ifa: "
1163 			    "addmulti failed for %s on %s (errno=%d)\n",
1164 			    ip6_sprintf(&mltaddr.sin6_addr),
1165 			    if_name(ifp), error));
1166 			/* XXX not very fatal, go on... */
1167 		} else {
1168 			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1169 		}
1170 
1171 
1172 		/*
1173 		 * join interface-local all-nodes address.
1174 		 * (ff01::1%ifN, and ff01::%ifN/32)
1175 		 */
1176 		mltaddr.sin6_addr = in6addr_nodelocal_allnodes;
1177 		if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) != 0)
1178 			goto cleanup; /* XXX: should not fail */
1179 
1180 		/* XXX: again, do we really need the route? */
1181 		rt = rtalloc1((struct sockaddr *)&mltaddr, 0);
1182 		if (rt) {
1183 			/* 32bit came from "mltmask" */
1184 			if (memcmp(&mltaddr.sin6_addr,
1185 			    &satocsin6(rt_getkey(rt))->sin6_addr,
1186 			    32 / NBBY)) {
1187 				rtfree(rt);
1188 				rt = NULL;
1189 			} else if (rt->rt_ifp != ifp) {
1190 				IN6_DPRINTF("%s: rt_ifp %p -> %p (%s) "
1191 				    "network %04x:%04x::/32 = %04x:%04x::/32\n",
1192 				    __func__, rt->rt_ifp, ifp, ifp->if_xname,
1193 				    ntohs(mltaddr.sin6_addr.s6_addr16[0]),
1194 				    ntohs(mltaddr.sin6_addr.s6_addr16[1]),
1195 				    satocsin6(rt_getkey(rt))->sin6_addr.s6_addr16[0],
1196 				    satocsin6(rt_getkey(rt))->sin6_addr.s6_addr16[1]);
1197 				rt_replace_ifa(rt, &ia->ia_ifa);
1198 				rt->rt_ifp = ifp;
1199 			}
1200 		}
1201 		if (!rt) {
1202 			struct rt_addrinfo info;
1203 
1204 			memset(&info, 0, sizeof(info));
1205 			info.rti_info[RTAX_DST] = (struct sockaddr *)&mltaddr;
1206 			info.rti_info[RTAX_GATEWAY] =
1207 			    (struct sockaddr *)&ia->ia_addr;
1208 			info.rti_info[RTAX_NETMASK] =
1209 			    (struct sockaddr *)&mltmask;
1210 			info.rti_info[RTAX_IFA] =
1211 			    (struct sockaddr *)&ia->ia_addr;
1212 			info.rti_flags = RTF_UP | RTF_CLONING;
1213 			error = rtrequest1(RTM_ADD, &info, NULL);
1214 			if (error)
1215 				goto cleanup;
1216 #undef	MLTMASK_LEN
1217 		} else {
1218 			rtfree(rt);
1219 		}
1220 		imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
1221 		if (!imm) {
1222 			nd6log((LOG_WARNING, "in6_update_ifa: "
1223 			    "addmulti failed for %s on %s (errno=%d)\n",
1224 			    ip6_sprintf(&mltaddr.sin6_addr),
1225 			    if_name(ifp), error));
1226 			goto cleanup;
1227 		} else {
1228 			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1229 		}
1230 	}
1231 
1232 	/*
1233 	 * Perform DAD, if needed.
1234 	 * XXX It may be of use, if we can administratively
1235 	 * disable DAD.
1236 	 */
1237 	if (hostIsNew && if_do_dad(ifp) &&
1238 	    ((ifra->ifra_flags & IN6_IFF_NODAD) == 0) &&
1239 	    (ia->ia6_flags & IN6_IFF_TENTATIVE))
1240 	{
1241 		int mindelay, maxdelay;
1242 
1243 		dad_delay = 0;
1244 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1245 			/*
1246 			 * We need to impose a delay before sending an NS
1247 			 * for DAD.  Check if we also needed a delay for the
1248 			 * corresponding MLD message.  If we did, the delay
1249 			 * should be larger than the MLD delay (this could be
1250 			 * relaxed a bit, but this simple logic is at least
1251 			 * safe).
1252 			 */
1253 			mindelay = 0;
1254 			if (in6m_sol != NULL &&
1255 			    in6m_sol->in6m_state == MLD_REPORTPENDING) {
1256 				mindelay = in6m_sol->in6m_timer;
1257 			}
1258 			maxdelay = MAX_RTR_SOLICITATION_DELAY * hz;
1259 			if (maxdelay - mindelay == 0)
1260 				dad_delay = 0;
1261 			else {
1262 				dad_delay =
1263 				    (cprng_fast32() % (maxdelay - mindelay)) +
1264 				    mindelay;
1265 			}
1266 		}
1267 		/* +1 ensures callout is always used */
1268 		nd6_dad_start(&ia->ia_ifa, dad_delay + 1);
1269 	}
1270 
1271 	return error;
1272 
1273   unlink:
1274 	/*
1275 	 * XXX: if a change of an existing address failed, keep the entry
1276 	 * anyway.
1277 	 */
1278 	if (hostIsNew)
1279 		in6_unlink_ifa(ia, ifp);
1280 	return error;
1281 
1282   cleanup:
1283 	in6_purgeaddr(&ia->ia_ifa);
1284 	return error;
1285 }
1286 
1287 int
1288 in6_update_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
1289     struct in6_ifaddr *ia, int flags)
1290 {
1291 	int rc, s;
1292 
1293 	s = splnet();
1294 	rc = in6_update_ifa1(ifp, ifra, ia, flags);
1295 	splx(s);
1296 	return rc;
1297 }
1298 
1299 void
1300 in6_purgeaddr(struct ifaddr *ifa)
1301 {
1302 	struct ifnet *ifp = ifa->ifa_ifp;
1303 	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1304 	struct in6_multi_mship *imm;
1305 
1306 	/* stop DAD processing */
1307 	nd6_dad_stop(ifa);
1308 
1309 	/*
1310 	 * delete route to the destination of the address being purged.
1311 	 * The interface must be p2p or loopback in this case.
1312 	 */
1313 	if ((ia->ia_flags & IFA_ROUTE) != 0 && ia->ia_dstaddr.sin6_len != 0) {
1314 		int e;
1315 
1316 		if ((e = rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST))
1317 		    != 0) {
1318 			log(LOG_ERR, "in6_purgeaddr: failed to remove "
1319 			    "a route to the p2p destination: %s on %s, "
1320 			    "errno=%d\n",
1321 			    ip6_sprintf(&ia->ia_addr.sin6_addr), if_name(ifp),
1322 			    e);
1323 			/* proceed anyway... */
1324 		} else
1325 			ia->ia_flags &= ~IFA_ROUTE;
1326 	}
1327 
1328 	/* Remove ownaddr's loopback rtentry, if it exists. */
1329 	in6_ifremlocal(&(ia->ia_ifa));
1330 
1331 	/*
1332 	 * leave from multicast groups we have joined for the interface
1333 	 */
1334 	while ((imm = LIST_FIRST(&ia->ia6_memberships)) != NULL) {
1335 		LIST_REMOVE(imm, i6mm_chain);
1336 		in6_leavegroup(imm);
1337 	}
1338 
1339 	in6_unlink_ifa(ia, ifp);
1340 }
1341 
1342 static void
1343 in6_unlink_ifa(struct in6_ifaddr *ia, struct ifnet *ifp)
1344 {
1345 	struct in6_ifaddr *oia;
1346 	int	s = splnet();
1347 
1348 	ifa_remove(ifp, &ia->ia_ifa);
1349 
1350 	oia = ia;
1351 	if (oia == (ia = in6_ifaddr))
1352 		in6_ifaddr = ia->ia_next;
1353 	else {
1354 		while (ia->ia_next && (ia->ia_next != oia))
1355 			ia = ia->ia_next;
1356 		if (ia->ia_next)
1357 			ia->ia_next = oia->ia_next;
1358 		else {
1359 			/* search failed */
1360 			printf("Couldn't unlink in6_ifaddr from in6_ifaddr\n");
1361 		}
1362 	}
1363 
1364 	/*
1365 	 * XXX thorpej@NetBSD.org -- if the interface is going
1366 	 * XXX away, don't save the multicast entries, delete them!
1367 	 */
1368 	if (LIST_EMPTY(&oia->ia6_multiaddrs))
1369 		;
1370 	else if (oia->ia_ifa.ifa_ifp->if_output == if_nulloutput) {
1371 		struct in6_multi *in6m, *next;
1372 
1373 		for (in6m = LIST_FIRST(&oia->ia6_multiaddrs); in6m != NULL;
1374 		     in6m = next) {
1375 			next = LIST_NEXT(in6m, in6m_entry);
1376 			in6_delmulti(in6m);
1377 		}
1378 	} else
1379 		in6_savemkludge(oia);
1380 
1381 	/*
1382 	 * Release the reference to the base prefix.  There should be a
1383 	 * positive reference.
1384 	 */
1385 	if (oia->ia6_ndpr == NULL) {
1386 		nd6log((LOG_NOTICE, "in6_unlink_ifa: autoconf'ed address "
1387 		    "%p has no prefix\n", oia));
1388 	} else {
1389 		oia->ia6_ndpr->ndpr_refcnt--;
1390 		oia->ia6_ndpr = NULL;
1391 	}
1392 
1393 	/*
1394 	 * Also, if the address being removed is autoconf'ed, call
1395 	 * pfxlist_onlink_check() since the release might affect the status of
1396 	 * other (detached) addresses.
1397 	 */
1398 	if ((oia->ia6_flags & IN6_IFF_AUTOCONF) != 0)
1399 		pfxlist_onlink_check();
1400 
1401 	/*
1402 	 * release another refcnt for the link from in6_ifaddr.
1403 	 * Note that we should decrement the refcnt at least once for all *BSD.
1404 	 */
1405 	ifafree(&oia->ia_ifa);
1406 
1407 	splx(s);
1408 }
1409 
1410 void
1411 in6_purgeif(struct ifnet *ifp)
1412 {
1413 	if_purgeaddrs(ifp, AF_INET6, in6_purgeaddr);
1414 
1415 	in6_ifdetach(ifp);
1416 }
1417 
1418 /*
1419  * SIOC[GAD]LIFADDR.
1420  *	SIOCGLIFADDR: get first address. (?)
1421  *	SIOCGLIFADDR with IFLR_PREFIX:
1422  *		get first address that matches the specified prefix.
1423  *	SIOCALIFADDR: add the specified address.
1424  *	SIOCALIFADDR with IFLR_PREFIX:
1425  *		add the specified prefix, filling hostid part from
1426  *		the first link-local address.  prefixlen must be <= 64.
1427  *	SIOCDLIFADDR: delete the specified address.
1428  *	SIOCDLIFADDR with IFLR_PREFIX:
1429  *		delete the first address that matches the specified prefix.
1430  * return values:
1431  *	EINVAL on invalid parameters
1432  *	EADDRNOTAVAIL on prefix match failed/specified address not found
1433  *	other values may be returned from in6_ioctl()
1434  *
1435  * NOTE: SIOCALIFADDR(with IFLR_PREFIX set) allows prefixlen less than 64.
1436  * this is to accommodate address naming scheme other than RFC2374,
1437  * in the future.
1438  * RFC2373 defines interface id to be 64bit, but it allows non-RFC2374
1439  * address encoding scheme. (see figure on page 8)
1440  */
1441 static int
1442 in6_lifaddr_ioctl(struct socket *so, u_long cmd, void *data,
1443 	struct ifnet *ifp)
1444 {
1445 	struct in6_ifaddr *ia;
1446 	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
1447 	struct ifaddr *ifa;
1448 	struct sockaddr *sa;
1449 
1450 	/* sanity checks */
1451 	if (!data || !ifp) {
1452 		panic("invalid argument to in6_lifaddr_ioctl");
1453 		/* NOTREACHED */
1454 	}
1455 
1456 	switch (cmd) {
1457 	case SIOCGLIFADDR:
1458 		/* address must be specified on GET with IFLR_PREFIX */
1459 		if ((iflr->flags & IFLR_PREFIX) == 0)
1460 			break;
1461 		/* FALLTHROUGH */
1462 	case SIOCALIFADDR:
1463 	case SIOCDLIFADDR:
1464 		/* address must be specified on ADD and DELETE */
1465 		sa = (struct sockaddr *)&iflr->addr;
1466 		if (sa->sa_family != AF_INET6)
1467 			return EINVAL;
1468 		if (sa->sa_len != sizeof(struct sockaddr_in6))
1469 			return EINVAL;
1470 		/* XXX need improvement */
1471 		sa = (struct sockaddr *)&iflr->dstaddr;
1472 		if (sa->sa_family && sa->sa_family != AF_INET6)
1473 			return EINVAL;
1474 		if (sa->sa_len && sa->sa_len != sizeof(struct sockaddr_in6))
1475 			return EINVAL;
1476 		break;
1477 	default: /* shouldn't happen */
1478 #if 0
1479 		panic("invalid cmd to in6_lifaddr_ioctl");
1480 		/* NOTREACHED */
1481 #else
1482 		return EOPNOTSUPP;
1483 #endif
1484 	}
1485 	if (sizeof(struct in6_addr) * NBBY < iflr->prefixlen)
1486 		return EINVAL;
1487 
1488 	switch (cmd) {
1489 	case SIOCALIFADDR:
1490 	    {
1491 		struct in6_aliasreq ifra;
1492 		struct in6_addr *xhostid = NULL;
1493 		int prefixlen;
1494 
1495 		if ((iflr->flags & IFLR_PREFIX) != 0) {
1496 			struct sockaddr_in6 *sin6;
1497 
1498 			/*
1499 			 * xhostid is to fill in the hostid part of the
1500 			 * address.  xhostid points to the first link-local
1501 			 * address attached to the interface.
1502 			 */
1503 			ia = in6ifa_ifpforlinklocal(ifp, 0);
1504 			if (ia == NULL)
1505 				return EADDRNOTAVAIL;
1506 			xhostid = IFA_IN6(&ia->ia_ifa);
1507 
1508 		 	/* prefixlen must be <= 64. */
1509 			if (64 < iflr->prefixlen)
1510 				return EINVAL;
1511 			prefixlen = iflr->prefixlen;
1512 
1513 			/* hostid part must be zero. */
1514 			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1515 			if (sin6->sin6_addr.s6_addr32[2] != 0
1516 			 || sin6->sin6_addr.s6_addr32[3] != 0) {
1517 				return EINVAL;
1518 			}
1519 		} else
1520 			prefixlen = iflr->prefixlen;
1521 
1522 		/* copy args to in6_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
1523 		memset(&ifra, 0, sizeof(ifra));
1524 		memcpy(ifra.ifra_name, iflr->iflr_name, sizeof(ifra.ifra_name));
1525 
1526 		memcpy(&ifra.ifra_addr, &iflr->addr,
1527 		    ((struct sockaddr *)&iflr->addr)->sa_len);
1528 		if (xhostid) {
1529 			/* fill in hostid part */
1530 			ifra.ifra_addr.sin6_addr.s6_addr32[2] =
1531 			    xhostid->s6_addr32[2];
1532 			ifra.ifra_addr.sin6_addr.s6_addr32[3] =
1533 			    xhostid->s6_addr32[3];
1534 		}
1535 
1536 		if (((struct sockaddr *)&iflr->dstaddr)->sa_family) { /* XXX */
1537 			memcpy(&ifra.ifra_dstaddr, &iflr->dstaddr,
1538 			    ((struct sockaddr *)&iflr->dstaddr)->sa_len);
1539 			if (xhostid) {
1540 				ifra.ifra_dstaddr.sin6_addr.s6_addr32[2] =
1541 				    xhostid->s6_addr32[2];
1542 				ifra.ifra_dstaddr.sin6_addr.s6_addr32[3] =
1543 				    xhostid->s6_addr32[3];
1544 			}
1545 		}
1546 
1547 		ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
1548 		in6_prefixlen2mask(&ifra.ifra_prefixmask.sin6_addr, prefixlen);
1549 
1550 		ifra.ifra_lifetime.ia6t_vltime = ND6_INFINITE_LIFETIME;
1551 		ifra.ifra_lifetime.ia6t_pltime = ND6_INFINITE_LIFETIME;
1552 		ifra.ifra_flags = iflr->flags & ~IFLR_PREFIX;
1553 		return in6_control(so, SIOCAIFADDR_IN6, &ifra, ifp);
1554 	    }
1555 	case SIOCGLIFADDR:
1556 	case SIOCDLIFADDR:
1557 	    {
1558 		struct in6_addr mask, candidate, match;
1559 		struct sockaddr_in6 *sin6;
1560 		int cmp;
1561 
1562 		memset(&mask, 0, sizeof(mask));
1563 		if (iflr->flags & IFLR_PREFIX) {
1564 			/* lookup a prefix rather than address. */
1565 			in6_prefixlen2mask(&mask, iflr->prefixlen);
1566 
1567 			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1568 			memcpy(&match, &sin6->sin6_addr, sizeof(match));
1569 			match.s6_addr32[0] &= mask.s6_addr32[0];
1570 			match.s6_addr32[1] &= mask.s6_addr32[1];
1571 			match.s6_addr32[2] &= mask.s6_addr32[2];
1572 			match.s6_addr32[3] &= mask.s6_addr32[3];
1573 
1574 			/* if you set extra bits, that's wrong */
1575 			if (memcmp(&match, &sin6->sin6_addr, sizeof(match)))
1576 				return EINVAL;
1577 
1578 			cmp = 1;
1579 		} else {
1580 			if (cmd == SIOCGLIFADDR) {
1581 				/* on getting an address, take the 1st match */
1582 				cmp = 0;	/* XXX */
1583 			} else {
1584 				/* on deleting an address, do exact match */
1585 				in6_prefixlen2mask(&mask, 128);
1586 				sin6 = (struct sockaddr_in6 *)&iflr->addr;
1587 				memcpy(&match, &sin6->sin6_addr, sizeof(match));
1588 
1589 				cmp = 1;
1590 			}
1591 		}
1592 
1593 		IFADDR_FOREACH(ifa, ifp) {
1594 			if (ifa->ifa_addr->sa_family != AF_INET6)
1595 				continue;
1596 			if (!cmp)
1597 				break;
1598 
1599 			/*
1600 			 * XXX: this is adhoc, but is necessary to allow
1601 			 * a user to specify fe80::/64 (not /10) for a
1602 			 * link-local address.
1603 			 */
1604 			memcpy(&candidate, IFA_IN6(ifa), sizeof(candidate));
1605 			in6_clearscope(&candidate);
1606 			candidate.s6_addr32[0] &= mask.s6_addr32[0];
1607 			candidate.s6_addr32[1] &= mask.s6_addr32[1];
1608 			candidate.s6_addr32[2] &= mask.s6_addr32[2];
1609 			candidate.s6_addr32[3] &= mask.s6_addr32[3];
1610 			if (IN6_ARE_ADDR_EQUAL(&candidate, &match))
1611 				break;
1612 		}
1613 		if (!ifa)
1614 			return EADDRNOTAVAIL;
1615 		ia = ifa2ia6(ifa);
1616 
1617 		if (cmd == SIOCGLIFADDR) {
1618 			int error;
1619 
1620 			/* fill in the if_laddrreq structure */
1621 			memcpy(&iflr->addr, &ia->ia_addr, ia->ia_addr.sin6_len);
1622 			error = sa6_recoverscope(
1623 			    (struct sockaddr_in6 *)&iflr->addr);
1624 			if (error != 0)
1625 				return error;
1626 
1627 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1628 				memcpy(&iflr->dstaddr, &ia->ia_dstaddr,
1629 				    ia->ia_dstaddr.sin6_len);
1630 				error = sa6_recoverscope(
1631 				    (struct sockaddr_in6 *)&iflr->dstaddr);
1632 				if (error != 0)
1633 					return error;
1634 			} else
1635 				memset(&iflr->dstaddr, 0, sizeof(iflr->dstaddr));
1636 
1637 			iflr->prefixlen =
1638 			    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
1639 
1640 			iflr->flags = ia->ia6_flags;	/* XXX */
1641 
1642 			return 0;
1643 		} else {
1644 			struct in6_aliasreq ifra;
1645 
1646 			/* fill in6_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
1647 			memset(&ifra, 0, sizeof(ifra));
1648 			memcpy(ifra.ifra_name, iflr->iflr_name,
1649 			    sizeof(ifra.ifra_name));
1650 
1651 			memcpy(&ifra.ifra_addr, &ia->ia_addr,
1652 			    ia->ia_addr.sin6_len);
1653 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1654 				memcpy(&ifra.ifra_dstaddr, &ia->ia_dstaddr,
1655 				    ia->ia_dstaddr.sin6_len);
1656 			} else {
1657 				memset(&ifra.ifra_dstaddr, 0,
1658 				    sizeof(ifra.ifra_dstaddr));
1659 			}
1660 			memcpy(&ifra.ifra_dstaddr, &ia->ia_prefixmask,
1661 			    ia->ia_prefixmask.sin6_len);
1662 
1663 			ifra.ifra_flags = ia->ia6_flags;
1664 			return in6_control(so, SIOCDIFADDR_IN6, &ifra, ifp);
1665 		}
1666 	    }
1667 	}
1668 
1669 	return EOPNOTSUPP;	/* just for safety */
1670 }
1671 
1672 /*
1673  * Initialize an interface's internet6 address
1674  * and routing table entry.
1675  */
1676 static int
1677 in6_ifinit(struct ifnet *ifp, struct in6_ifaddr *ia,
1678 	const struct sockaddr_in6 *sin6, int newhost)
1679 {
1680 	int	error = 0, plen, ifacount = 0;
1681 	int	s = splnet();
1682 	struct ifaddr *ifa;
1683 
1684 	/*
1685 	 * Give the interface a chance to initialize
1686 	 * if this is its first address,
1687 	 * and to validate the address if necessary.
1688 	 */
1689 	IFADDR_FOREACH(ifa, ifp) {
1690 		if (ifa->ifa_addr == NULL)
1691 			continue;	/* just for safety */
1692 		if (ifa->ifa_addr->sa_family != AF_INET6)
1693 			continue;
1694 		ifacount++;
1695 	}
1696 
1697 	ia->ia_addr = *sin6;
1698 
1699 	if (ifacount <= 1 &&
1700 	    (error = if_addr_init(ifp, &ia->ia_ifa, true)) != 0) {
1701 		splx(s);
1702 		return error;
1703 	}
1704 	splx(s);
1705 
1706 	ia->ia_ifa.ifa_metric = ifp->if_metric;
1707 
1708 	/* we could do in(6)_socktrim here, but just omit it at this moment. */
1709 
1710 	/*
1711 	 * Special case:
1712 	 * If the destination address is specified for a point-to-point
1713 	 * interface, install a route to the destination as an interface
1714 	 * direct route.
1715 	 */
1716 	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1717 	if (plen == 128 && ia->ia_dstaddr.sin6_family == AF_INET6) {
1718 		if ((error = rtinit(&ia->ia_ifa, RTM_ADD,
1719 				    RTF_UP | RTF_HOST)) != 0)
1720 			return error;
1721 		ia->ia_flags |= IFA_ROUTE;
1722 	}
1723 
1724 	/* Add ownaddr as loopback rtentry, if necessary (ex. on p2p link). */
1725 	if (newhost) {
1726 		/* set the rtrequest function to create llinfo */
1727 		if (ifp->if_flags & IFF_POINTOPOINT)
1728 			ia->ia_ifa.ifa_rtrequest = p2p_rtrequest;
1729 		else if ((ifp->if_flags & IFF_LOOPBACK) == 0)
1730 			ia->ia_ifa.ifa_rtrequest = nd6_rtrequest;
1731 		in6_ifaddlocal(&ia->ia_ifa);
1732 	} else {
1733 		/* Inform the routing socket of new flags/timings */
1734 		rt_newaddrmsg(RTM_NEWADDR, &ia->ia_ifa, 0, NULL);
1735 	}
1736 
1737 	if (ifp->if_flags & IFF_MULTICAST)
1738 		in6_restoremkludge(ia, ifp);
1739 
1740 	return error;
1741 }
1742 
1743 static struct ifaddr *
1744 bestifa(struct ifaddr *best_ifa, struct ifaddr *ifa)
1745 {
1746 	if (best_ifa == NULL || best_ifa->ifa_preference < ifa->ifa_preference)
1747 		return ifa;
1748 	return best_ifa;
1749 }
1750 
1751 /*
1752  * Find an IPv6 interface link-local address specific to an interface.
1753  */
1754 struct in6_ifaddr *
1755 in6ifa_ifpforlinklocal(const struct ifnet *ifp, const int ignoreflags)
1756 {
1757 	struct ifaddr *best_ifa = NULL, *ifa;
1758 
1759 	IFADDR_FOREACH(ifa, ifp) {
1760 		if (ifa->ifa_addr == NULL)
1761 			continue;	/* just for safety */
1762 		if (ifa->ifa_addr->sa_family != AF_INET6)
1763 			continue;
1764 		if (!IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa)))
1765 			continue;
1766 		if ((((struct in6_ifaddr *)ifa)->ia6_flags & ignoreflags) != 0)
1767 			continue;
1768 		best_ifa = bestifa(best_ifa, ifa);
1769 	}
1770 
1771 	return (struct in6_ifaddr *)best_ifa;
1772 }
1773 
1774 /*
1775  * find the internet address corresponding to a given address.
1776  * ifaddr is returned referenced.
1777  */
1778 struct in6_ifaddr *
1779 in6ifa_ifwithaddr(const struct in6_addr *addr, uint32_t zoneid)
1780 {
1781 	struct in6_ifaddr *ia;
1782 
1783 #ifdef __FreeBSD__
1784 	IN6_IFADDR_RLOCK();
1785 	LIST_FOREACH(ia, IN6ADDR_HASH(addr), ia6_hash) {
1786 #else
1787 	for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
1788 #endif
1789 		if (IN6_ARE_ADDR_EQUAL(IA6_IN6(ia), addr)) {
1790 			if (zoneid != 0 &&
1791 			    zoneid != ia->ia_addr.sin6_scope_id)
1792 				continue;
1793 			ifaref(&ia->ia_ifa);
1794 			break;
1795 		}
1796 	}
1797 #ifdef __FreeBSD__
1798 	IN6_IFADDR_RUNLOCK();
1799 #endif
1800 	return ia;
1801 }
1802 
1803 /*
1804  * find the internet address corresponding to a given interface and address.
1805  */
1806 struct in6_ifaddr *
1807 in6ifa_ifpwithaddr(const struct ifnet *ifp, const struct in6_addr *addr)
1808 {
1809 	struct ifaddr *best_ifa = NULL, *ifa;
1810 
1811 	IFADDR_FOREACH(ifa, ifp) {
1812 		if (ifa->ifa_addr == NULL)
1813 			continue;	/* just for safety */
1814 		if (ifa->ifa_addr->sa_family != AF_INET6)
1815 			continue;
1816 		if (!IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa)))
1817 			continue;
1818 		best_ifa = bestifa(best_ifa, ifa);
1819 	}
1820 
1821 	return (struct in6_ifaddr *)best_ifa;
1822 }
1823 
1824 static struct in6_ifaddr *
1825 bestia(struct in6_ifaddr *best_ia, struct in6_ifaddr *ia)
1826 {
1827 	if (best_ia == NULL ||
1828 	    best_ia->ia_ifa.ifa_preference < ia->ia_ifa.ifa_preference)
1829 		return ia;
1830 	return best_ia;
1831 }
1832 
1833 /*
1834  * Convert IP6 address to printable (loggable) representation.
1835  */
1836 char *
1837 ip6_sprintf(const struct in6_addr *addr)
1838 {
1839 	static int ip6round = 0;
1840 	static char ip6buf[8][INET6_ADDRSTRLEN];
1841 	char *cp = ip6buf[ip6round++ & 7];
1842 
1843 	in6_print(cp, INET6_ADDRSTRLEN, addr);
1844 	return cp;
1845 }
1846 
1847 /*
1848  * Determine if an address is on a local network.
1849  */
1850 int
1851 in6_localaddr(const struct in6_addr *in6)
1852 {
1853 	struct in6_ifaddr *ia;
1854 
1855 	if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
1856 		return 1;
1857 
1858 	for (ia = in6_ifaddr; ia; ia = ia->ia_next)
1859 		if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
1860 					      &ia->ia_prefixmask.sin6_addr))
1861 			return 1;
1862 
1863 	return 0;
1864 }
1865 
1866 int
1867 in6_is_addr_deprecated(struct sockaddr_in6 *sa6)
1868 {
1869 	struct in6_ifaddr *ia;
1870 
1871 	for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
1872 		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
1873 		    &sa6->sin6_addr) &&
1874 #ifdef SCOPEDROUTING
1875 		    ia->ia_addr.sin6_scope_id == sa6->sin6_scope_id &&
1876 #endif
1877 		    (ia->ia6_flags & IN6_IFF_DEPRECATED) != 0)
1878 			return 1; /* true */
1879 
1880 		/* XXX: do we still have to go thru the rest of the list? */
1881 	}
1882 
1883 	return 0;		/* false */
1884 }
1885 
1886 /*
1887  * return length of part which dst and src are equal
1888  * hard coding...
1889  */
1890 int
1891 in6_matchlen(struct in6_addr *src, struct in6_addr *dst)
1892 {
1893 	int match = 0;
1894 	u_char *s = (u_char *)src, *d = (u_char *)dst;
1895 	u_char *lim = s + 16, r;
1896 
1897 	while (s < lim)
1898 		if ((r = (*d++ ^ *s++)) != 0) {
1899 			while (r < 128) {
1900 				match++;
1901 				r <<= 1;
1902 			}
1903 			break;
1904 		} else
1905 			match += NBBY;
1906 	return match;
1907 }
1908 
1909 /* XXX: to be scope conscious */
1910 int
1911 in6_are_prefix_equal(struct in6_addr *p1, struct in6_addr *p2, int len)
1912 {
1913 	int bytelen, bitlen;
1914 
1915 	/* sanity check */
1916 	if (len < 0 || len > 128) {
1917 		log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n",
1918 		    len);
1919 		return 0;
1920 	}
1921 
1922 	bytelen = len / NBBY;
1923 	bitlen = len % NBBY;
1924 
1925 	if (memcmp(&p1->s6_addr, &p2->s6_addr, bytelen))
1926 		return 0;
1927 	if (bitlen != 0 &&
1928 	    p1->s6_addr[bytelen] >> (NBBY - bitlen) !=
1929 	    p2->s6_addr[bytelen] >> (NBBY - bitlen))
1930 		return 0;
1931 
1932 	return 1;
1933 }
1934 
1935 void
1936 in6_prefixlen2mask(struct in6_addr *maskp, int len)
1937 {
1938 	static const u_char maskarray[NBBY] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
1939 	int bytelen, bitlen, i;
1940 
1941 	/* sanity check */
1942 	if (len < 0 || len > 128) {
1943 		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
1944 		    len);
1945 		return;
1946 	}
1947 
1948 	memset(maskp, 0, sizeof(*maskp));
1949 	bytelen = len / NBBY;
1950 	bitlen = len % NBBY;
1951 	for (i = 0; i < bytelen; i++)
1952 		maskp->s6_addr[i] = 0xff;
1953 	if (bitlen)
1954 		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
1955 }
1956 
1957 /*
1958  * return the best address out of the same scope. if no address was
1959  * found, return the first valid address from designated IF.
1960  */
1961 struct in6_ifaddr *
1962 in6_ifawithifp(struct ifnet *ifp, struct in6_addr *dst)
1963 {
1964 	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
1965 	struct ifaddr *ifa;
1966 	struct in6_ifaddr *best_ia = NULL, *ia;
1967 	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
1968 
1969 	dep[0] = dep[1] = NULL;
1970 
1971 	/*
1972 	 * We first look for addresses in the same scope.
1973 	 * If there is one, return it.
1974 	 * If two or more, return one which matches the dst longest.
1975 	 * If none, return one of global addresses assigned other ifs.
1976 	 */
1977 	IFADDR_FOREACH(ifa, ifp) {
1978 		if (ifa->ifa_addr->sa_family != AF_INET6)
1979 			continue;
1980 		ia = (struct in6_ifaddr *)ifa;
1981 		if (ia->ia6_flags & IN6_IFF_ANYCAST)
1982 			continue; /* XXX: is there any case to allow anycast? */
1983 		if (ia->ia6_flags & IN6_IFF_NOTREADY)
1984 			continue; /* don't use this interface */
1985 		if (ia->ia6_flags & IN6_IFF_DETACHED)
1986 			continue;
1987 		if (ia->ia6_flags & IN6_IFF_DEPRECATED) {
1988 			if (ip6_use_deprecated)
1989 				dep[0] = ia;
1990 			continue;
1991 		}
1992 
1993 		if (dst_scope != in6_addrscope(IFA_IN6(ifa)))
1994 			continue;
1995 		/*
1996 		 * call in6_matchlen() as few as possible
1997 		 */
1998 		if (best_ia == NULL) {
1999 			best_ia = ia;
2000 			continue;
2001 		}
2002 		if (blen == -1)
2003 			blen = in6_matchlen(&best_ia->ia_addr.sin6_addr, dst);
2004 		tlen = in6_matchlen(IFA_IN6(ifa), dst);
2005 		if (tlen > blen) {
2006 			blen = tlen;
2007 			best_ia = ia;
2008 		} else if (tlen == blen)
2009 			best_ia = bestia(best_ia, ia);
2010 	}
2011 	if (best_ia != NULL)
2012 		return best_ia;
2013 
2014 	IFADDR_FOREACH(ifa, ifp) {
2015 		if (ifa->ifa_addr->sa_family != AF_INET6)
2016 			continue;
2017 		ia = (struct in6_ifaddr *)ifa;
2018 		if (ia->ia6_flags & IN6_IFF_ANYCAST)
2019 			continue; /* XXX: is there any case to allow anycast? */
2020 		if (ia->ia6_flags & IN6_IFF_NOTREADY)
2021 			continue; /* don't use this interface */
2022 		if (ia->ia6_flags & IN6_IFF_DETACHED)
2023 			continue;
2024 		if (ia->ia6_flags & IN6_IFF_DEPRECATED) {
2025 			if (ip6_use_deprecated)
2026 				dep[1] = (struct in6_ifaddr *)ifa;
2027 			continue;
2028 		}
2029 
2030 		best_ia = bestia(best_ia, ia);
2031 	}
2032 	if (best_ia != NULL)
2033 		return best_ia;
2034 
2035 	/* use the last-resort values, that are, deprecated addresses */
2036 	if (dep[0])
2037 		return dep[0];
2038 	if (dep[1])
2039 		return dep[1];
2040 
2041 	return NULL;
2042 }
2043 
2044 /*
2045  * perform DAD when interface becomes IFF_UP.
2046  */
2047 void
2048 in6_if_link_up(struct ifnet *ifp)
2049 {
2050 	struct ifaddr *ifa;
2051 	struct in6_ifaddr *ia;
2052 
2053 	/* Ensure it's sane to run DAD */
2054 	if (ifp->if_link_state == LINK_STATE_DOWN)
2055 		return;
2056 	if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) != (IFF_UP|IFF_RUNNING))
2057 		return;
2058 
2059 	IFADDR_FOREACH(ifa, ifp) {
2060 		if (ifa->ifa_addr->sa_family != AF_INET6)
2061 			continue;
2062 		ia = (struct in6_ifaddr *)ifa;
2063 
2064 		/* If detached then mark as tentative */
2065 		if (ia->ia6_flags & IN6_IFF_DETACHED) {
2066 			ia->ia6_flags &= ~IN6_IFF_DETACHED;
2067 			if (if_do_dad(ifp)) {
2068 				ia->ia6_flags |= IN6_IFF_TENTATIVE;
2069 				nd6log((LOG_ERR, "in6_if_up: "
2070 				    "%s marked tentative\n",
2071 				    ip6_sprintf(&ia->ia_addr.sin6_addr)));
2072 			} else if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0)
2073 				rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
2074 		}
2075 
2076 		if (ia->ia6_flags & IN6_IFF_TENTATIVE) {
2077 			int rand_delay;
2078 
2079 			/* Clear the duplicated flag as we're starting DAD. */
2080 			ia->ia6_flags &= ~IN6_IFF_DUPLICATED;
2081 
2082 			/*
2083 			 * The TENTATIVE flag was likely set by hand
2084 			 * beforehand, implicitly indicating the need for DAD.
2085 			 * We may be able to skip the random delay in this
2086 			 * case, but we impose delays just in case.
2087 			 */
2088 			rand_delay = cprng_fast32() %
2089 			    (MAX_RTR_SOLICITATION_DELAY * hz);
2090 			/* +1 ensures callout is always used */
2091 			nd6_dad_start(ifa, rand_delay + 1);
2092 		}
2093 	}
2094 
2095 	/* Restore any detached prefixes */
2096 	pfxlist_onlink_check();
2097 }
2098 
2099 void
2100 in6_if_up(struct ifnet *ifp)
2101 {
2102 
2103 	/*
2104 	 * special cases, like 6to4, are handled in in6_ifattach
2105 	 */
2106 	in6_ifattach(ifp, NULL);
2107 
2108 	/* interface may not support link state, so bring it up also */
2109 	in6_if_link_up(ifp);
2110 }
2111 
2112 /*
2113  * Mark all addresses as detached.
2114  */
2115 void
2116 in6_if_link_down(struct ifnet *ifp)
2117 {
2118 	struct ifaddr *ifa;
2119 	struct in6_ifaddr *ia;
2120 
2121 	/* Any prefixes on this interface should be detached as well */
2122 	pfxlist_onlink_check();
2123 
2124 	IFADDR_FOREACH(ifa, ifp) {
2125 		if (ifa->ifa_addr->sa_family != AF_INET6)
2126 			continue;
2127 		ia = (struct in6_ifaddr *)ifa;
2128 
2129 		/* Stop DAD processing */
2130 		nd6_dad_stop(ifa);
2131 
2132 		/*
2133 		 * Mark the address as detached.
2134 		 * This satisfies RFC4862 Section 5.3, but we should apply
2135 		 * this logic to all addresses to be a good citizen and
2136 		 * avoid potential duplicated addresses.
2137 		 * When the interface comes up again, detached addresses
2138 		 * are marked tentative and DAD commences.
2139 		 */
2140 		if (!(ia->ia6_flags & IN6_IFF_DETACHED)) {
2141 			nd6log((LOG_DEBUG, "in6_if_down: "
2142 			    "%s marked detached\n",
2143 			    ip6_sprintf(&ia->ia_addr.sin6_addr)));
2144 			ia->ia6_flags |= IN6_IFF_DETACHED;
2145 			ia->ia6_flags &=
2146 			    ~(IN6_IFF_TENTATIVE | IN6_IFF_DUPLICATED);
2147 			rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
2148 		}
2149 	}
2150 }
2151 
2152 void
2153 in6_if_down(struct ifnet *ifp)
2154 {
2155 
2156 	in6_if_link_down(ifp);
2157 }
2158 
2159 void
2160 in6_if_link_state_change(struct ifnet *ifp, int link_state)
2161 {
2162 
2163 	switch (link_state) {
2164 	case LINK_STATE_DOWN:
2165 		in6_if_link_down(ifp);
2166 		break;
2167 	case LINK_STATE_UP:
2168 		in6_if_link_up(ifp);
2169 		break;
2170 	}
2171 }
2172 
2173 /*
2174  * Calculate max IPv6 MTU through all the interfaces and store it
2175  * to in6_maxmtu.
2176  */
2177 void
2178 in6_setmaxmtu(void)
2179 {
2180 	unsigned long maxmtu = 0;
2181 	struct ifnet *ifp;
2182 
2183 	IFNET_FOREACH(ifp) {
2184 		/* this function can be called during ifnet initialization */
2185 		if (!ifp->if_afdata[AF_INET6])
2186 			continue;
2187 		if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
2188 		    IN6_LINKMTU(ifp) > maxmtu)
2189 			maxmtu = IN6_LINKMTU(ifp);
2190 	}
2191 	if (maxmtu)	     /* update only when maxmtu is positive */
2192 		in6_maxmtu = maxmtu;
2193 }
2194 
2195 /*
2196  * Provide the length of interface identifiers to be used for the link attached
2197  * to the given interface.  The length should be defined in "IPv6 over
2198  * xxx-link" document.  Note that address architecture might also define
2199  * the length for a particular set of address prefixes, regardless of the
2200  * link type.  As clarified in rfc2462bis, those two definitions should be
2201  * consistent, and those really are as of August 2004.
2202  */
2203 int
2204 in6_if2idlen(struct ifnet *ifp)
2205 {
2206 	switch (ifp->if_type) {
2207 	case IFT_ETHER:		/* RFC2464 */
2208 	case IFT_PROPVIRTUAL:	/* XXX: no RFC. treat it as ether */
2209 	case IFT_L2VLAN:	/* ditto */
2210 	case IFT_IEEE80211:	/* ditto */
2211 	case IFT_FDDI:		/* RFC2467 */
2212 	case IFT_ISO88025:	/* RFC2470 (IPv6 over Token Ring) */
2213 	case IFT_PPP:		/* RFC2472 */
2214 	case IFT_ARCNET:	/* RFC2497 */
2215 	case IFT_FRELAY:	/* RFC2590 */
2216 	case IFT_IEEE1394:	/* RFC3146 */
2217 	case IFT_GIF:		/* draft-ietf-v6ops-mech-v2-07 */
2218 	case IFT_LOOP:		/* XXX: is this really correct? */
2219 		return 64;
2220 	default:
2221 		/*
2222 		 * Unknown link type:
2223 		 * It might be controversial to use the today's common constant
2224 		 * of 64 for these cases unconditionally.  For full compliance,
2225 		 * we should return an error in this case.  On the other hand,
2226 		 * if we simply miss the standard for the link type or a new
2227 		 * standard is defined for a new link type, the IFID length
2228 		 * is very likely to be the common constant.  As a compromise,
2229 		 * we always use the constant, but make an explicit notice
2230 		 * indicating the "unknown" case.
2231 		 */
2232 		printf("in6_if2idlen: unknown link type (%d)\n", ifp->if_type);
2233 		return 64;
2234 	}
2235 }
2236 
2237 struct in6_llentry {
2238 	struct llentry		base;
2239 };
2240 
2241 #define	IN6_LLTBL_DEFAULT_HSIZE	32
2242 #define	IN6_LLTBL_HASH(k, h) \
2243 	(((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1))
2244 
2245 /*
2246  * Do actual deallocation of @lle.
2247  * Called by LLE_FREE_LOCKED when number of references
2248  * drops to zero.
2249  */
2250 static void
2251 in6_lltable_destroy_lle(struct llentry *lle)
2252 {
2253 
2254 	LLE_WUNLOCK(lle);
2255 	LLE_LOCK_DESTROY(lle);
2256 	kmem_intr_free(lle, sizeof(struct in6_llentry));
2257 }
2258 
2259 static struct llentry *
2260 in6_lltable_new(const struct in6_addr *addr6, u_int flags)
2261 {
2262 	struct in6_llentry *lle;
2263 
2264 	lle = kmem_intr_zalloc(sizeof(struct in6_llentry), KM_NOSLEEP);
2265 	if (lle == NULL)		/* NB: caller generates msg */
2266 		return NULL;
2267 
2268 	lle->base.r_l3addr.addr6 = *addr6;
2269 	lle->base.lle_refcnt = 1;
2270 	lle->base.lle_free = in6_lltable_destroy_lle;
2271 	LLE_LOCK_INIT(&lle->base);
2272 	callout_init(&lle->base.lle_timer, CALLOUT_MPSAFE);
2273 
2274 	return &lle->base;
2275 }
2276 
2277 static int
2278 in6_lltable_match_prefix(const struct sockaddr *prefix,
2279     const struct sockaddr *mask, u_int flags, struct llentry *lle)
2280 {
2281 	const struct sockaddr_in6 *pfx = (const struct sockaddr_in6 *)prefix;
2282 	const struct sockaddr_in6 *msk = (const struct sockaddr_in6 *)mask;
2283 
2284 	if (IN6_ARE_MASKED_ADDR_EQUAL(&lle->r_l3addr.addr6,
2285 	    &pfx->sin6_addr, &msk->sin6_addr) &&
2286 	    ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC)))
2287 		return 1;
2288 
2289 	return 0;
2290 }
2291 
2292 static void
2293 in6_lltable_free_entry(struct lltable *llt, struct llentry *lle)
2294 {
2295 	struct ifnet *ifp __diagused;
2296 
2297 	LLE_WLOCK_ASSERT(lle);
2298 	KASSERT(llt != NULL);
2299 
2300 	/* Unlink entry from table */
2301 	if ((lle->la_flags & LLE_LINKED) != 0) {
2302 
2303 		ifp = llt->llt_ifp;
2304 		IF_AFDATA_WLOCK_ASSERT(ifp);
2305 		lltable_unlink_entry(llt, lle);
2306 	}
2307 
2308 	KASSERT(mutex_owned(softnet_lock));
2309 	callout_halt(&lle->lle_timer, softnet_lock);
2310 	LLE_REMREF(lle);
2311 
2312 	llentry_free(lle);
2313 }
2314 
2315 static int
2316 in6_lltable_rtcheck(struct ifnet *ifp,
2317 		    u_int flags,
2318 		    const struct sockaddr *l3addr)
2319 {
2320 	struct rtentry *rt;
2321 
2322 	KASSERTMSG(l3addr->sa_family == AF_INET6,
2323 	    "sin_family %d", l3addr->sa_family);
2324 
2325 	rt = rtalloc1(l3addr, 0);
2326 	if (rt == NULL || (rt->rt_flags & RTF_GATEWAY) || rt->rt_ifp != ifp) {
2327 		struct ifaddr *ifa;
2328 		/*
2329 		 * Create an ND6 cache for an IPv6 neighbor
2330 		 * that is not covered by our own prefix.
2331 		 */
2332 		/* XXX ifaof_ifpforaddr should take a const param */
2333 		ifa = ifaof_ifpforaddr(l3addr, ifp);
2334 		if (ifa != NULL) {
2335 			ifafree(ifa);
2336 			if (rt != NULL)
2337 				rtfree(rt);
2338 			return 0;
2339 		}
2340 		log(LOG_INFO, "IPv6 address: \"%s\" is not on the network\n",
2341 		    ip6_sprintf(&((const struct sockaddr_in6 *)l3addr)->sin6_addr));
2342 		if (rt != NULL)
2343 			rtfree(rt);
2344 		return EINVAL;
2345 	}
2346 	rtfree(rt);
2347 	return 0;
2348 }
2349 
2350 static inline uint32_t
2351 in6_lltable_hash_dst(const struct in6_addr *dst, uint32_t hsize)
2352 {
2353 
2354 	return IN6_LLTBL_HASH(dst->s6_addr32[3], hsize);
2355 }
2356 
2357 static uint32_t
2358 in6_lltable_hash(const struct llentry *lle, uint32_t hsize)
2359 {
2360 
2361 	return in6_lltable_hash_dst(&lle->r_l3addr.addr6, hsize);
2362 }
2363 
2364 static void
2365 in6_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
2366 {
2367 	struct sockaddr_in6 *sin6;
2368 
2369 	sin6 = (struct sockaddr_in6 *)sa;
2370 	bzero(sin6, sizeof(*sin6));
2371 	sin6->sin6_family = AF_INET6;
2372 	sin6->sin6_len = sizeof(*sin6);
2373 	sin6->sin6_addr = lle->r_l3addr.addr6;
2374 }
2375 
2376 static inline struct llentry *
2377 in6_lltable_find_dst(struct lltable *llt, const struct in6_addr *dst)
2378 {
2379 	struct llentry *lle;
2380 	struct llentries *lleh;
2381 	u_int hashidx;
2382 
2383 	hashidx = in6_lltable_hash_dst(dst, llt->llt_hsize);
2384 	lleh = &llt->lle_head[hashidx];
2385 	LIST_FOREACH(lle, lleh, lle_next) {
2386 		if (lle->la_flags & LLE_DELETED)
2387 			continue;
2388 		if (IN6_ARE_ADDR_EQUAL(&lle->r_l3addr.addr6, dst))
2389 			break;
2390 	}
2391 
2392 	return lle;
2393 }
2394 
2395 static int
2396 in6_lltable_delete(struct lltable *llt, u_int flags,
2397 	const struct sockaddr *l3addr)
2398 {
2399 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2400 	struct llentry *lle;
2401 
2402 	IF_AFDATA_WLOCK_ASSERT(llt->llt_ifp);
2403 	KASSERTMSG(l3addr->sa_family == AF_INET6,
2404 	    "sin_family %d", l3addr->sa_family);
2405 
2406 	lle = in6_lltable_find_dst(llt, &sin6->sin6_addr);
2407 
2408 	if (lle == NULL)
2409 		return ENOENT;
2410 
2411 	if (!(lle->la_flags & LLE_IFADDR) || (flags & LLE_IFADDR)) {
2412 		LLE_WLOCK(lle);
2413 		lle->la_flags |= LLE_DELETED;
2414 #ifdef DIAGNOSTIC
2415 		log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle);
2416 #endif
2417 		if ((lle->la_flags & (LLE_STATIC | LLE_IFADDR)) == LLE_STATIC)
2418 			llentry_free(lle);
2419 		else
2420 			LLE_WUNLOCK(lle);
2421 	}
2422 
2423 	return 0;
2424 }
2425 
2426 static struct llentry *
2427 in6_lltable_create(struct lltable *llt, u_int flags,
2428 	const struct sockaddr *l3addr)
2429 {
2430 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2431 	struct ifnet *ifp = llt->llt_ifp;
2432 	struct llentry *lle;
2433 
2434 	IF_AFDATA_WLOCK_ASSERT(ifp);
2435 	KASSERTMSG(l3addr->sa_family == AF_INET6,
2436 	    "sin_family %d", l3addr->sa_family);
2437 
2438 	lle = in6_lltable_find_dst(llt, &sin6->sin6_addr);
2439 
2440 	if (lle != NULL) {
2441 		LLE_WLOCK(lle);
2442 		return lle;
2443 	}
2444 
2445 	/*
2446 	 * A route that covers the given address must have
2447 	 * been installed 1st because we are doing a resolution,
2448 	 * verify this.
2449 	 */
2450 	if (!(flags & LLE_IFADDR) &&
2451 	    in6_lltable_rtcheck(ifp, flags, l3addr) != 0)
2452 		return NULL;
2453 
2454 	lle = in6_lltable_new(&sin6->sin6_addr, flags);
2455 	if (lle == NULL) {
2456 		log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
2457 		return NULL;
2458 	}
2459 	lle->la_flags = flags;
2460 	if ((flags & LLE_IFADDR) == LLE_IFADDR) {
2461 		memcpy(&lle->ll_addr, CLLADDR(ifp->if_sadl), ifp->if_addrlen);
2462 		lle->la_flags |= (LLE_VALID | LLE_STATIC);
2463 	}
2464 
2465 	lltable_link_entry(llt, lle);
2466 	LLE_WLOCK(lle);
2467 
2468 	return lle;
2469 }
2470 
2471 static struct llentry *
2472 in6_lltable_lookup(struct lltable *llt, u_int flags,
2473 	const struct sockaddr *l3addr)
2474 {
2475 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2476 	struct llentry *lle;
2477 
2478 	IF_AFDATA_LOCK_ASSERT(llt->llt_ifp);
2479 	KASSERTMSG(l3addr->sa_family == AF_INET6,
2480 	    "sin_family %d", l3addr->sa_family);
2481 
2482 	lle = in6_lltable_find_dst(llt, &sin6->sin6_addr);
2483 
2484 	if (lle == NULL)
2485 		return NULL;
2486 
2487 	if (flags & LLE_EXCLUSIVE)
2488 		LLE_WLOCK(lle);
2489 	else
2490 		LLE_RLOCK(lle);
2491 	return lle;
2492 }
2493 
2494 static struct lltable *
2495 in6_lltattach(struct ifnet *ifp)
2496 {
2497 	struct lltable *llt;
2498 
2499 	llt = lltable_allocate_htbl(IN6_LLTBL_DEFAULT_HSIZE);
2500 	llt->llt_af = AF_INET6;
2501 	llt->llt_ifp = ifp;
2502 
2503 	llt->llt_lookup = in6_lltable_lookup;
2504 	llt->llt_create = in6_lltable_create;
2505 	llt->llt_delete = in6_lltable_delete;
2506 #if notyet
2507 	llt->llt_dump_entry = in6_lltable_dump_entry;
2508 #endif
2509 	llt->llt_hash = in6_lltable_hash;
2510 	llt->llt_fill_sa_entry = in6_lltable_fill_sa_entry;
2511 	llt->llt_free_entry = in6_lltable_free_entry;
2512 	llt->llt_match_prefix = in6_lltable_match_prefix;
2513 	lltable_link(llt);
2514 
2515 	return llt;
2516 }
2517 
2518 void *
2519 in6_domifattach(struct ifnet *ifp)
2520 {
2521 	struct in6_ifextra *ext;
2522 
2523 	ext = malloc(sizeof(*ext), M_IFADDR, M_WAITOK|M_ZERO);
2524 
2525 	ext->in6_ifstat = malloc(sizeof(struct in6_ifstat),
2526 	    M_IFADDR, M_WAITOK|M_ZERO);
2527 
2528 	ext->icmp6_ifstat = malloc(sizeof(struct icmp6_ifstat),
2529 	    M_IFADDR, M_WAITOK|M_ZERO);
2530 
2531 	ext->nd_ifinfo = nd6_ifattach(ifp);
2532 	ext->scope6_id = scope6_ifattach(ifp);
2533 	ext->nprefixes = 0;
2534 	ext->ndefrouters = 0;
2535 
2536 	ext->lltable = in6_lltattach(ifp);
2537 
2538 	return ext;
2539 }
2540 
2541 void
2542 in6_domifdetach(struct ifnet *ifp, void *aux)
2543 {
2544 	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
2545 
2546 	lltable_free(ext->lltable);
2547 	ext->lltable = NULL;
2548 	nd6_ifdetach(ifp, ext);
2549 	free(ext->in6_ifstat, M_IFADDR);
2550 	free(ext->icmp6_ifstat, M_IFADDR);
2551 	scope6_ifdetach(ext->scope6_id);
2552 	free(ext, M_IFADDR);
2553 }
2554 
2555 /*
2556  * Convert sockaddr_in6 to sockaddr_in.  Original sockaddr_in6 must be
2557  * v4 mapped addr or v4 compat addr
2558  */
2559 void
2560 in6_sin6_2_sin(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2561 {
2562 	memset(sin, 0, sizeof(*sin));
2563 	sin->sin_len = sizeof(struct sockaddr_in);
2564 	sin->sin_family = AF_INET;
2565 	sin->sin_port = sin6->sin6_port;
2566 	sin->sin_addr.s_addr = sin6->sin6_addr.s6_addr32[3];
2567 }
2568 
2569 /* Convert sockaddr_in to sockaddr_in6 in v4 mapped addr format. */
2570 void
2571 in6_sin_2_v4mapsin6(const struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2572 {
2573 	memset(sin6, 0, sizeof(*sin6));
2574 	sin6->sin6_len = sizeof(struct sockaddr_in6);
2575 	sin6->sin6_family = AF_INET6;
2576 	sin6->sin6_port = sin->sin_port;
2577 	sin6->sin6_addr.s6_addr32[0] = 0;
2578 	sin6->sin6_addr.s6_addr32[1] = 0;
2579 	sin6->sin6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
2580 	sin6->sin6_addr.s6_addr32[3] = sin->sin_addr.s_addr;
2581 }
2582 
2583 /* Convert sockaddr_in6 into sockaddr_in. */
2584 void
2585 in6_sin6_2_sin_in_sock(struct sockaddr *nam)
2586 {
2587 	struct sockaddr_in *sin_p;
2588 	struct sockaddr_in6 sin6;
2589 
2590 	/*
2591 	 * Save original sockaddr_in6 addr and convert it
2592 	 * to sockaddr_in.
2593 	 */
2594 	sin6 = *(struct sockaddr_in6 *)nam;
2595 	sin_p = (struct sockaddr_in *)nam;
2596 	in6_sin6_2_sin(sin_p, &sin6);
2597 }
2598 
2599 /* Convert sockaddr_in into sockaddr_in6 in v4 mapped addr format. */
2600 void
2601 in6_sin_2_v4mapsin6_in_sock(struct sockaddr **nam)
2602 {
2603 	struct sockaddr_in *sin_p;
2604 	struct sockaddr_in6 *sin6_p;
2605 
2606 	sin6_p = malloc(sizeof(*sin6_p), M_SONAME, M_WAITOK);
2607 	sin_p = (struct sockaddr_in *)*nam;
2608 	in6_sin_2_v4mapsin6(sin_p, sin6_p);
2609 	free(*nam, M_SONAME);
2610 	*nam = (struct sockaddr *)sin6_p;
2611 }
2612