xref: /freebsd/sys/netinet6/in6.c (revision 0f7f3352c8bc463607912e2463d13e52d44a4cae)
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/eventhandler.h>
72 #include <sys/errno.h>
73 #include <sys/jail.h>
74 #include <sys/malloc.h>
75 #include <sys/socket.h>
76 #include <sys/socketvar.h>
77 #include <sys/sockio.h>
78 #include <sys/systm.h>
79 #include <sys/priv.h>
80 #include <sys/proc.h>
81 #include <sys/time.h>
82 #include <sys/kernel.h>
83 #include <sys/lock.h>
84 #include <sys/rmlock.h>
85 #include <sys/syslog.h>
86 
87 #include <net/if.h>
88 #include <net/if_var.h>
89 #include <net/if_types.h>
90 #include <net/route.h>
91 #include <net/if_dl.h>
92 #include <net/vnet.h>
93 
94 #include <netinet/in.h>
95 #include <netinet/in_var.h>
96 #include <net/if_llatbl.h>
97 #include <netinet/if_ether.h>
98 #include <netinet/in_systm.h>
99 #include <netinet/ip.h>
100 #include <netinet/in_pcb.h>
101 #include <netinet/ip_carp.h>
102 
103 #include <netinet/ip6.h>
104 #include <netinet6/ip6_var.h>
105 #include <netinet6/nd6.h>
106 #include <netinet6/mld6_var.h>
107 #include <netinet6/ip6_mroute.h>
108 #include <netinet6/in6_ifattach.h>
109 #include <netinet6/scope6_var.h>
110 #include <netinet6/in6_fib.h>
111 #include <netinet6/in6_pcb.h>
112 
113 VNET_DECLARE(int, icmp6_nodeinfo_oldmcprefix);
114 #define V_icmp6_nodeinfo_oldmcprefix	VNET(icmp6_nodeinfo_oldmcprefix)
115 
116 /*
117  * Definitions of some costant IP6 addresses.
118  */
119 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
120 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
121 const struct in6_addr in6addr_nodelocal_allnodes =
122 	IN6ADDR_NODELOCAL_ALLNODES_INIT;
123 const struct in6_addr in6addr_linklocal_allnodes =
124 	IN6ADDR_LINKLOCAL_ALLNODES_INIT;
125 const struct in6_addr in6addr_linklocal_allrouters =
126 	IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
127 const struct in6_addr in6addr_linklocal_allv2routers =
128 	IN6ADDR_LINKLOCAL_ALLV2ROUTERS_INIT;
129 
130 const struct in6_addr in6mask0 = IN6MASK0;
131 const struct in6_addr in6mask32 = IN6MASK32;
132 const struct in6_addr in6mask64 = IN6MASK64;
133 const struct in6_addr in6mask96 = IN6MASK96;
134 const struct in6_addr in6mask128 = IN6MASK128;
135 
136 const struct sockaddr_in6 sa6_any =
137 	{ sizeof(sa6_any), AF_INET6, 0, 0, IN6ADDR_ANY_INIT, 0 };
138 
139 static int in6_notify_ifa(struct ifnet *, struct in6_ifaddr *,
140 	struct in6_aliasreq *, int);
141 static void in6_unlink_ifa(struct in6_ifaddr *, struct ifnet *);
142 
143 static int in6_validate_ifra(struct ifnet *, struct in6_aliasreq *,
144     struct in6_ifaddr *, int);
145 static struct in6_ifaddr *in6_alloc_ifa(struct ifnet *,
146     struct in6_aliasreq *, int flags);
147 static int in6_update_ifa_internal(struct ifnet *, struct in6_aliasreq *,
148     struct in6_ifaddr *, int, int);
149 static int in6_broadcast_ifa(struct ifnet *, struct in6_aliasreq *,
150     struct in6_ifaddr *, int);
151 
152 #define ifa2ia6(ifa)	((struct in6_ifaddr *)(ifa))
153 #define ia62ifa(ia6)	(&((ia6)->ia_ifa))
154 
155 
156 void
157 in6_newaddrmsg(struct in6_ifaddr *ia, int cmd)
158 {
159 	struct sockaddr_dl gateway;
160 	struct sockaddr_in6 mask, addr;
161 	struct rtentry rt;
162 
163 	/*
164 	 * initialize for rtmsg generation
165 	 */
166 	bzero(&gateway, sizeof(gateway));
167 	gateway.sdl_len = sizeof(gateway);
168 	gateway.sdl_family = AF_LINK;
169 
170 	bzero(&rt, sizeof(rt));
171 	rt.rt_gateway = (struct sockaddr *)&gateway;
172 	memcpy(&mask, &ia->ia_prefixmask, sizeof(ia->ia_prefixmask));
173 	memcpy(&addr, &ia->ia_addr, sizeof(ia->ia_addr));
174 	rt_mask(&rt) = (struct sockaddr *)&mask;
175 	rt_key(&rt) = (struct sockaddr *)&addr;
176 	rt.rt_flags = RTF_HOST | RTF_STATIC;
177 	if (cmd == RTM_ADD)
178 		rt.rt_flags |= RTF_UP;
179 	/* Announce arrival of local address to all FIBs. */
180 	rt_newaddrmsg(cmd, &ia->ia_ifa, 0, &rt);
181 }
182 
183 int
184 in6_mask2len(struct in6_addr *mask, u_char *lim0)
185 {
186 	int x = 0, y;
187 	u_char *lim = lim0, *p;
188 
189 	/* ignore the scope_id part */
190 	if (lim0 == NULL || lim0 - (u_char *)mask > sizeof(*mask))
191 		lim = (u_char *)mask + sizeof(*mask);
192 	for (p = (u_char *)mask; p < lim; x++, p++) {
193 		if (*p != 0xff)
194 			break;
195 	}
196 	y = 0;
197 	if (p < lim) {
198 		for (y = 0; y < 8; y++) {
199 			if ((*p & (0x80 >> y)) == 0)
200 				break;
201 		}
202 	}
203 
204 	/*
205 	 * when the limit pointer is given, do a stricter check on the
206 	 * remaining bits.
207 	 */
208 	if (p < lim) {
209 		if (y != 0 && (*p & (0x00ff >> y)) != 0)
210 			return (-1);
211 		for (p = p + 1; p < lim; p++)
212 			if (*p != 0)
213 				return (-1);
214 	}
215 
216 	return x * 8 + y;
217 }
218 
219 #ifdef COMPAT_FREEBSD32
220 struct in6_ndifreq32 {
221 	char ifname[IFNAMSIZ];
222 	uint32_t ifindex;
223 };
224 #define	SIOCGDEFIFACE32_IN6	_IOWR('i', 86, struct in6_ndifreq32)
225 #endif
226 
227 int
228 in6_control(struct socket *so, u_long cmd, caddr_t data,
229     struct ifnet *ifp, struct thread *td)
230 {
231 	struct	in6_ifreq *ifr = (struct in6_ifreq *)data;
232 	struct	in6_ifaddr *ia = NULL;
233 	struct	in6_aliasreq *ifra = (struct in6_aliasreq *)data;
234 	struct sockaddr_in6 *sa6;
235 	int carp_attached = 0;
236 	int error;
237 	u_long ocmd = cmd;
238 
239 	/*
240 	 * Compat to make pre-10.x ifconfig(8) operable.
241 	 */
242 	if (cmd == OSIOCAIFADDR_IN6)
243 		cmd = SIOCAIFADDR_IN6;
244 
245 	switch (cmd) {
246 	case SIOCGETSGCNT_IN6:
247 	case SIOCGETMIFCNT_IN6:
248 		/*
249 		 * XXX mrt_ioctl has a 3rd, unused, FIB argument in route.c.
250 		 * We cannot see how that would be needed, so do not adjust the
251 		 * KPI blindly; more likely should clean up the IPv4 variant.
252 		 */
253 		return (mrt6_ioctl ? mrt6_ioctl(cmd, data) : EOPNOTSUPP);
254 	}
255 
256 	switch (cmd) {
257 	case SIOCAADDRCTL_POLICY:
258 	case SIOCDADDRCTL_POLICY:
259 		if (td != NULL) {
260 			error = priv_check(td, PRIV_NETINET_ADDRCTRL6);
261 			if (error)
262 				return (error);
263 		}
264 		return (in6_src_ioctl(cmd, data));
265 	}
266 
267 	if (ifp == NULL)
268 		return (EOPNOTSUPP);
269 
270 	switch (cmd) {
271 	case SIOCSNDFLUSH_IN6:
272 	case SIOCSPFXFLUSH_IN6:
273 	case SIOCSRTRFLUSH_IN6:
274 	case SIOCSDEFIFACE_IN6:
275 	case SIOCSIFINFO_FLAGS:
276 	case SIOCSIFINFO_IN6:
277 		if (td != NULL) {
278 			error = priv_check(td, PRIV_NETINET_ND6);
279 			if (error)
280 				return (error);
281 		}
282 		/* FALLTHROUGH */
283 	case OSIOCGIFINFO_IN6:
284 	case SIOCGIFINFO_IN6:
285 	case SIOCGNBRINFO_IN6:
286 	case SIOCGDEFIFACE_IN6:
287 		return (nd6_ioctl(cmd, data, ifp));
288 
289 #ifdef COMPAT_FREEBSD32
290 	case SIOCGDEFIFACE32_IN6:
291 		{
292 			struct in6_ndifreq ndif;
293 			struct in6_ndifreq32 *ndif32;
294 
295 			error = nd6_ioctl(SIOCGDEFIFACE_IN6, (caddr_t)&ndif,
296 			    ifp);
297 			if (error)
298 				return (error);
299 			ndif32 = (struct in6_ndifreq32 *)data;
300 			ndif32->ifindex = ndif.ifindex;
301 			return (0);
302 		}
303 #endif
304 	}
305 
306 	switch (cmd) {
307 	case SIOCSIFPREFIX_IN6:
308 	case SIOCDIFPREFIX_IN6:
309 	case SIOCAIFPREFIX_IN6:
310 	case SIOCCIFPREFIX_IN6:
311 	case SIOCSGIFPREFIX_IN6:
312 	case SIOCGIFPREFIX_IN6:
313 		log(LOG_NOTICE,
314 		    "prefix ioctls are now invalidated. "
315 		    "please use ifconfig.\n");
316 		return (EOPNOTSUPP);
317 	}
318 
319 	switch (cmd) {
320 	case SIOCSSCOPE6:
321 		if (td != NULL) {
322 			error = priv_check(td, PRIV_NETINET_SCOPE6);
323 			if (error)
324 				return (error);
325 		}
326 		/* FALLTHROUGH */
327 	case SIOCGSCOPE6:
328 	case SIOCGSCOPE6DEF:
329 		return (scope6_ioctl(cmd, data, ifp));
330 	}
331 
332 	/*
333 	 * Find address for this interface, if it exists.
334 	 *
335 	 * In netinet code, we have checked ifra_addr in SIOCSIF*ADDR operation
336 	 * only, and used the first interface address as the target of other
337 	 * operations (without checking ifra_addr).  This was because netinet
338 	 * code/API assumed at most 1 interface address per interface.
339 	 * Since IPv6 allows a node to assign multiple addresses
340 	 * on a single interface, we almost always look and check the
341 	 * presence of ifra_addr, and reject invalid ones here.
342 	 * It also decreases duplicated code among SIOC*_IN6 operations.
343 	 */
344 	switch (cmd) {
345 	case SIOCAIFADDR_IN6:
346 	case SIOCSIFPHYADDR_IN6:
347 		sa6 = &ifra->ifra_addr;
348 		break;
349 	case SIOCSIFADDR_IN6:
350 	case SIOCGIFADDR_IN6:
351 	case SIOCSIFDSTADDR_IN6:
352 	case SIOCSIFNETMASK_IN6:
353 	case SIOCGIFDSTADDR_IN6:
354 	case SIOCGIFNETMASK_IN6:
355 	case SIOCDIFADDR_IN6:
356 	case SIOCGIFPSRCADDR_IN6:
357 	case SIOCGIFPDSTADDR_IN6:
358 	case SIOCGIFAFLAG_IN6:
359 	case SIOCSNDFLUSH_IN6:
360 	case SIOCSPFXFLUSH_IN6:
361 	case SIOCSRTRFLUSH_IN6:
362 	case SIOCGIFALIFETIME_IN6:
363 	case SIOCGIFSTAT_IN6:
364 	case SIOCGIFSTAT_ICMP6:
365 		sa6 = &ifr->ifr_addr;
366 		break;
367 	case SIOCSIFADDR:
368 	case SIOCSIFBRDADDR:
369 	case SIOCSIFDSTADDR:
370 	case SIOCSIFNETMASK:
371 		/*
372 		 * Although we should pass any non-INET6 ioctl requests
373 		 * down to driver, we filter some legacy INET requests.
374 		 * Drivers trust SIOCSIFADDR et al to come from an already
375 		 * privileged layer, and do not perform any credentials
376 		 * checks or input validation.
377 		 */
378 		return (EINVAL);
379 	default:
380 		sa6 = NULL;
381 		break;
382 	}
383 	if (sa6 && sa6->sin6_family == AF_INET6) {
384 		if (sa6->sin6_scope_id != 0)
385 			error = sa6_embedscope(sa6, 0);
386 		else
387 			error = in6_setscope(&sa6->sin6_addr, ifp, NULL);
388 		if (error != 0)
389 			return (error);
390 		if (td != NULL && (error = prison_check_ip6(td->td_ucred,
391 		    &sa6->sin6_addr)) != 0)
392 			return (error);
393 		ia = in6ifa_ifpwithaddr(ifp, &sa6->sin6_addr);
394 	} else
395 		ia = NULL;
396 
397 	switch (cmd) {
398 	case SIOCSIFADDR_IN6:
399 	case SIOCSIFDSTADDR_IN6:
400 	case SIOCSIFNETMASK_IN6:
401 		/*
402 		 * Since IPv6 allows a node to assign multiple addresses
403 		 * on a single interface, SIOCSIFxxx ioctls are deprecated.
404 		 */
405 		/* we decided to obsolete this command (20000704) */
406 		error = EINVAL;
407 		goto out;
408 
409 	case SIOCDIFADDR_IN6:
410 		/*
411 		 * for IPv4, we look for existing in_ifaddr here to allow
412 		 * "ifconfig if0 delete" to remove the first IPv4 address on
413 		 * the interface.  For IPv6, as the spec allows multiple
414 		 * interface address from the day one, we consider "remove the
415 		 * first one" semantics to be not preferable.
416 		 */
417 		if (ia == NULL) {
418 			error = EADDRNOTAVAIL;
419 			goto out;
420 		}
421 		/* FALLTHROUGH */
422 	case SIOCAIFADDR_IN6:
423 		/*
424 		 * We always require users to specify a valid IPv6 address for
425 		 * the corresponding operation.
426 		 */
427 		if (ifra->ifra_addr.sin6_family != AF_INET6 ||
428 		    ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6)) {
429 			error = EAFNOSUPPORT;
430 			goto out;
431 		}
432 
433 		if (td != NULL) {
434 			error = priv_check(td, (cmd == SIOCDIFADDR_IN6) ?
435 			    PRIV_NET_DELIFADDR : PRIV_NET_ADDIFADDR);
436 			if (error)
437 				goto out;
438 		}
439 		/* FALLTHROUGH */
440 	case SIOCGIFSTAT_IN6:
441 	case SIOCGIFSTAT_ICMP6:
442 		if (ifp->if_afdata[AF_INET6] == NULL) {
443 			error = EPFNOSUPPORT;
444 			goto out;
445 		}
446 		break;
447 
448 	case SIOCGIFADDR_IN6:
449 		/* This interface is basically deprecated. use SIOCGIFCONF. */
450 		/* FALLTHROUGH */
451 	case SIOCGIFAFLAG_IN6:
452 	case SIOCGIFNETMASK_IN6:
453 	case SIOCGIFDSTADDR_IN6:
454 	case SIOCGIFALIFETIME_IN6:
455 		/* must think again about its semantics */
456 		if (ia == NULL) {
457 			error = EADDRNOTAVAIL;
458 			goto out;
459 		}
460 		break;
461 	}
462 
463 	switch (cmd) {
464 	case SIOCGIFADDR_IN6:
465 		ifr->ifr_addr = ia->ia_addr;
466 		if ((error = sa6_recoverscope(&ifr->ifr_addr)) != 0)
467 			goto out;
468 		break;
469 
470 	case SIOCGIFDSTADDR_IN6:
471 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
472 			error = EINVAL;
473 			goto out;
474 		}
475 		/*
476 		 * XXX: should we check if ifa_dstaddr is NULL and return
477 		 * an error?
478 		 */
479 		ifr->ifr_dstaddr = ia->ia_dstaddr;
480 		if ((error = sa6_recoverscope(&ifr->ifr_dstaddr)) != 0)
481 			goto out;
482 		break;
483 
484 	case SIOCGIFNETMASK_IN6:
485 		ifr->ifr_addr = ia->ia_prefixmask;
486 		break;
487 
488 	case SIOCGIFAFLAG_IN6:
489 		ifr->ifr_ifru.ifru_flags6 = ia->ia6_flags;
490 		break;
491 
492 	case SIOCGIFSTAT_IN6:
493 		COUNTER_ARRAY_COPY(((struct in6_ifextra *)
494 		    ifp->if_afdata[AF_INET6])->in6_ifstat,
495 		    &ifr->ifr_ifru.ifru_stat,
496 		    sizeof(struct in6_ifstat) / sizeof(uint64_t));
497 		break;
498 
499 	case SIOCGIFSTAT_ICMP6:
500 		COUNTER_ARRAY_COPY(((struct in6_ifextra *)
501 		    ifp->if_afdata[AF_INET6])->icmp6_ifstat,
502 		    &ifr->ifr_ifru.ifru_icmp6stat,
503 		    sizeof(struct icmp6_ifstat) / sizeof(uint64_t));
504 		break;
505 
506 	case SIOCGIFALIFETIME_IN6:
507 		ifr->ifr_ifru.ifru_lifetime = ia->ia6_lifetime;
508 		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
509 			time_t maxexpire;
510 			struct in6_addrlifetime *retlt =
511 			    &ifr->ifr_ifru.ifru_lifetime;
512 
513 			/*
514 			 * XXX: adjust expiration time assuming time_t is
515 			 * signed.
516 			 */
517 			maxexpire = (-1) &
518 			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
519 			if (ia->ia6_lifetime.ia6t_vltime <
520 			    maxexpire - ia->ia6_updatetime) {
521 				retlt->ia6t_expire = ia->ia6_updatetime +
522 				    ia->ia6_lifetime.ia6t_vltime;
523 			} else
524 				retlt->ia6t_expire = maxexpire;
525 		}
526 		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
527 			time_t maxexpire;
528 			struct in6_addrlifetime *retlt =
529 			    &ifr->ifr_ifru.ifru_lifetime;
530 
531 			/*
532 			 * XXX: adjust expiration time assuming time_t is
533 			 * signed.
534 			 */
535 			maxexpire = (-1) &
536 			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
537 			if (ia->ia6_lifetime.ia6t_pltime <
538 			    maxexpire - ia->ia6_updatetime) {
539 				retlt->ia6t_preferred = ia->ia6_updatetime +
540 				    ia->ia6_lifetime.ia6t_pltime;
541 			} else
542 				retlt->ia6t_preferred = maxexpire;
543 		}
544 		break;
545 
546 	case SIOCAIFADDR_IN6:
547 	{
548 		struct nd_prefixctl pr0;
549 		struct nd_prefix *pr;
550 
551 		/*
552 		 * first, make or update the interface address structure,
553 		 * and link it to the list.
554 		 */
555 		if ((error = in6_update_ifa(ifp, ifra, ia, 0)) != 0)
556 			goto out;
557 		if (ia != NULL)
558 			ifa_free(&ia->ia_ifa);
559 		if ((ia = in6ifa_ifpwithaddr(ifp, &ifra->ifra_addr.sin6_addr))
560 		    == NULL) {
561 			/*
562 			 * this can happen when the user specify the 0 valid
563 			 * lifetime.
564 			 */
565 			break;
566 		}
567 
568 		if (cmd == ocmd && ifra->ifra_vhid > 0) {
569 			if (carp_attach_p != NULL)
570 				error = (*carp_attach_p)(&ia->ia_ifa,
571 				    ifra->ifra_vhid);
572 			else
573 				error = EPROTONOSUPPORT;
574 			if (error)
575 				goto out;
576 			else
577 				carp_attached = 1;
578 		}
579 
580 		/*
581 		 * then, make the prefix on-link on the interface.
582 		 * XXX: we'd rather create the prefix before the address, but
583 		 * we need at least one address to install the corresponding
584 		 * interface route, so we configure the address first.
585 		 */
586 
587 		/*
588 		 * convert mask to prefix length (prefixmask has already
589 		 * been validated in in6_update_ifa().
590 		 */
591 		bzero(&pr0, sizeof(pr0));
592 		pr0.ndpr_ifp = ifp;
593 		pr0.ndpr_plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
594 		    NULL);
595 		if (pr0.ndpr_plen == 128) {
596 			/* we don't need to install a host route. */
597 			goto aifaddr_out;
598 		}
599 		pr0.ndpr_prefix = ifra->ifra_addr;
600 		/* apply the mask for safety. */
601 		IN6_MASK_ADDR(&pr0.ndpr_prefix.sin6_addr,
602 		    &ifra->ifra_prefixmask.sin6_addr);
603 
604 		/*
605 		 * XXX: since we don't have an API to set prefix (not address)
606 		 * lifetimes, we just use the same lifetimes as addresses.
607 		 * The (temporarily) installed lifetimes can be overridden by
608 		 * later advertised RAs (when accept_rtadv is non 0), which is
609 		 * an intended behavior.
610 		 */
611 		pr0.ndpr_raf_onlink = 1; /* should be configurable? */
612 		pr0.ndpr_raf_auto =
613 		    ((ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0);
614 		pr0.ndpr_vltime = ifra->ifra_lifetime.ia6t_vltime;
615 		pr0.ndpr_pltime = ifra->ifra_lifetime.ia6t_pltime;
616 
617 		/* add the prefix if not yet. */
618 		if ((pr = nd6_prefix_lookup(&pr0)) == NULL) {
619 			/*
620 			 * nd6_prelist_add will install the corresponding
621 			 * interface route.
622 			 */
623 			if ((error = nd6_prelist_add(&pr0, NULL, &pr)) != 0) {
624 				if (carp_attached)
625 					(*carp_detach_p)(&ia->ia_ifa);
626 				goto out;
627 			}
628 		}
629 
630 		/* relate the address to the prefix */
631 		if (ia->ia6_ndpr == NULL) {
632 			ia->ia6_ndpr = pr;
633 			pr->ndpr_addrcnt++;
634 
635 			/*
636 			 * If this is the first autoconf address from the
637 			 * prefix, create a temporary address as well
638 			 * (when required).
639 			 */
640 			if ((ia->ia6_flags & IN6_IFF_AUTOCONF) &&
641 			    V_ip6_use_tempaddr && pr->ndpr_addrcnt == 1) {
642 				int e;
643 				if ((e = in6_tmpifadd(ia, 1, 0)) != 0) {
644 					log(LOG_NOTICE, "in6_control: failed "
645 					    "to create a temporary address, "
646 					    "errno=%d\n", e);
647 				}
648 			}
649 		}
650 		nd6_prefix_rele(pr);
651 
652 		/*
653 		 * this might affect the status of autoconfigured addresses,
654 		 * that is, this address might make other addresses detached.
655 		 */
656 		pfxlist_onlink_check();
657 
658 aifaddr_out:
659 		/*
660 		 * Try to clear the flag when a new IPv6 address is added
661 		 * onto an IFDISABLED interface and it succeeds.
662 		 */
663 		if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) {
664 			struct in6_ndireq nd;
665 
666 			memset(&nd, 0, sizeof(nd));
667 			nd.ndi.flags = ND_IFINFO(ifp)->flags;
668 			nd.ndi.flags &= ~ND6_IFF_IFDISABLED;
669 			if (nd6_ioctl(SIOCSIFINFO_FLAGS, (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 intentionally.
675 			 * The failure means address duplication was detected.
676 			 */
677 		}
678 		EVENTHANDLER_INVOKE(ifaddr_event, ifp);
679 		break;
680 	}
681 
682 	case SIOCDIFADDR_IN6:
683 	{
684 		struct nd_prefix *pr;
685 
686 		/*
687 		 * If the address being deleted is the only one that owns
688 		 * the corresponding prefix, expire the prefix as well.
689 		 * XXX: theoretically, we don't have to worry about such
690 		 * relationship, since we separate the address management
691 		 * and the prefix management.  We do this, however, to provide
692 		 * as much backward compatibility as possible in terms of
693 		 * the ioctl operation.
694 		 * Note that in6_purgeaddr() will decrement ndpr_addrcnt.
695 		 */
696 		pr = ia->ia6_ndpr;
697 		in6_purgeaddr(&ia->ia_ifa);
698 		if (pr != NULL && pr->ndpr_addrcnt == 0) {
699 			ND6_WLOCK();
700 			nd6_prefix_unlink(pr, NULL);
701 			ND6_WUNLOCK();
702 			nd6_prefix_del(pr);
703 		}
704 		EVENTHANDLER_INVOKE(ifaddr_event, ifp);
705 		break;
706 	}
707 
708 	default:
709 		if (ifp->if_ioctl == NULL) {
710 			error = EOPNOTSUPP;
711 			goto out;
712 		}
713 		error = (*ifp->if_ioctl)(ifp, cmd, data);
714 		goto out;
715 	}
716 
717 	error = 0;
718 out:
719 	if (ia != NULL)
720 		ifa_free(&ia->ia_ifa);
721 	return (error);
722 }
723 
724 
725 /*
726  * Join necessary multicast groups.  Factored out from in6_update_ifa().
727  * This entire work should only be done once, for the default FIB.
728  */
729 static int
730 in6_update_ifa_join_mc(struct ifnet *ifp, struct in6_aliasreq *ifra,
731     struct in6_ifaddr *ia, int flags, struct in6_multi **in6m_sol)
732 {
733 	char ip6buf[INET6_ADDRSTRLEN];
734 	struct in6_addr mltaddr;
735 	struct in6_multi_mship *imm;
736 	int delay, error;
737 
738 	KASSERT(in6m_sol != NULL, ("%s: in6m_sol is NULL", __func__));
739 
740 	/* Join solicited multicast addr for new host id. */
741 	bzero(&mltaddr, sizeof(struct in6_addr));
742 	mltaddr.s6_addr32[0] = IPV6_ADDR_INT32_MLL;
743 	mltaddr.s6_addr32[2] = htonl(1);
744 	mltaddr.s6_addr32[3] = ifra->ifra_addr.sin6_addr.s6_addr32[3];
745 	mltaddr.s6_addr8[12] = 0xff;
746 	if ((error = in6_setscope(&mltaddr, ifp, NULL)) != 0) {
747 		/* XXX: should not happen */
748 		log(LOG_ERR, "%s: in6_setscope failed\n", __func__);
749 		goto cleanup;
750 	}
751 	delay = error = 0;
752 	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
753 		/*
754 		 * We need a random delay for DAD on the address being
755 		 * configured.  It also means delaying transmission of the
756 		 * corresponding MLD report to avoid report collision.
757 		 * [RFC 4861, Section 6.3.7]
758 		 */
759 		delay = arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz);
760 	}
761 	imm = in6_joingroup(ifp, &mltaddr, &error, delay);
762 	if (imm == NULL) {
763 		nd6log((LOG_WARNING, "%s: in6_joingroup failed for %s on %s "
764 		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf, &mltaddr),
765 		    if_name(ifp), error));
766 		goto cleanup;
767 	}
768 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
769 	*in6m_sol = imm->i6mm_maddr;
770 
771 	/*
772 	 * Join link-local all-nodes address.
773 	 */
774 	mltaddr = in6addr_linklocal_allnodes;
775 	if ((error = in6_setscope(&mltaddr, ifp, NULL)) != 0)
776 		goto cleanup; /* XXX: should not fail */
777 
778 	imm = in6_joingroup(ifp, &mltaddr, &error, 0);
779 	if (imm == NULL) {
780 		nd6log((LOG_WARNING, "%s: in6_joingroup failed for %s on %s "
781 		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf, &mltaddr),
782 		    if_name(ifp), error));
783 		goto cleanup;
784 	}
785 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
786 
787 	/*
788 	 * Join node information group address.
789 	 */
790 	delay = 0;
791 	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
792 		/*
793 		 * The spec does not say anything about delay for this group,
794 		 * but the same logic should apply.
795 		 */
796 		delay = arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz);
797 	}
798 	if (in6_nigroup(ifp, NULL, -1, &mltaddr) == 0) {
799 		/* XXX jinmei */
800 		imm = in6_joingroup(ifp, &mltaddr, &error, delay);
801 		if (imm == NULL)
802 			nd6log((LOG_WARNING,
803 			    "%s: in6_joingroup failed for %s on %s "
804 			    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
805 			    &mltaddr), if_name(ifp), error));
806 			/* XXX not very fatal, go on... */
807 		else
808 			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
809 	}
810 	if (V_icmp6_nodeinfo_oldmcprefix &&
811 	    in6_nigroup_oldmcprefix(ifp, NULL, -1, &mltaddr) == 0) {
812 		imm = in6_joingroup(ifp, &mltaddr, &error, delay);
813 		if (imm == NULL)
814 			nd6log((LOG_WARNING,
815 			    "%s: in6_joingroup failed for %s on %s "
816 			    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
817 			    &mltaddr), if_name(ifp), error));
818 			/* XXX not very fatal, go on... */
819 		else
820 			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
821 	}
822 
823 	/*
824 	 * Join interface-local all-nodes address.
825 	 * (ff01::1%ifN, and ff01::%ifN/32)
826 	 */
827 	mltaddr = in6addr_nodelocal_allnodes;
828 	if ((error = in6_setscope(&mltaddr, ifp, NULL)) != 0)
829 		goto cleanup; /* XXX: should not fail */
830 
831 	imm = in6_joingroup(ifp, &mltaddr, &error, 0);
832 	if (imm == NULL) {
833 		nd6log((LOG_WARNING, "%s: in6_joingroup failed for %s on %s "
834 		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
835 		    &mltaddr), if_name(ifp), error));
836 		goto cleanup;
837 	}
838 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
839 
840 cleanup:
841 	return (error);
842 }
843 
844 /*
845  * Update parameters of an IPv6 interface address.
846  * If necessary, a new entry is created and linked into address chains.
847  * This function is separated from in6_control().
848  */
849 int
850 in6_update_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
851     struct in6_ifaddr *ia, int flags)
852 {
853 	int error, hostIsNew = 0;
854 
855 	if ((error = in6_validate_ifra(ifp, ifra, ia, flags)) != 0)
856 		return (error);
857 
858 	if (ia == NULL) {
859 		hostIsNew = 1;
860 		if ((ia = in6_alloc_ifa(ifp, ifra, flags)) == NULL)
861 			return (ENOBUFS);
862 	}
863 
864 	error = in6_update_ifa_internal(ifp, ifra, ia, hostIsNew, flags);
865 	if (error != 0) {
866 		if (hostIsNew != 0) {
867 			in6_unlink_ifa(ia, ifp);
868 			ifa_free(&ia->ia_ifa);
869 		}
870 		return (error);
871 	}
872 
873 	if (hostIsNew)
874 		error = in6_broadcast_ifa(ifp, ifra, ia, flags);
875 
876 	return (error);
877 }
878 
879 /*
880  * Fill in basic IPv6 address request info.
881  */
882 void
883 in6_prepare_ifra(struct in6_aliasreq *ifra, const struct in6_addr *addr,
884     const struct in6_addr *mask)
885 {
886 
887 	memset(ifra, 0, sizeof(struct in6_aliasreq));
888 
889 	ifra->ifra_addr.sin6_family = AF_INET6;
890 	ifra->ifra_addr.sin6_len = sizeof(struct sockaddr_in6);
891 	if (addr != NULL)
892 		ifra->ifra_addr.sin6_addr = *addr;
893 
894 	ifra->ifra_prefixmask.sin6_family = AF_INET6;
895 	ifra->ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
896 	if (mask != NULL)
897 		ifra->ifra_prefixmask.sin6_addr = *mask;
898 }
899 
900 static int
901 in6_validate_ifra(struct ifnet *ifp, struct in6_aliasreq *ifra,
902     struct in6_ifaddr *ia, int flags)
903 {
904 	int plen = -1;
905 	struct sockaddr_in6 dst6;
906 	struct in6_addrlifetime *lt;
907 	char ip6buf[INET6_ADDRSTRLEN];
908 
909 	/* Validate parameters */
910 	if (ifp == NULL || ifra == NULL) /* this maybe redundant */
911 		return (EINVAL);
912 
913 	/*
914 	 * The destination address for a p2p link must have a family
915 	 * of AF_UNSPEC or AF_INET6.
916 	 */
917 	if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
918 	    ifra->ifra_dstaddr.sin6_family != AF_INET6 &&
919 	    ifra->ifra_dstaddr.sin6_family != AF_UNSPEC)
920 		return (EAFNOSUPPORT);
921 
922 	/*
923 	 * Validate address
924 	 */
925 	if (ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6) ||
926 	    ifra->ifra_addr.sin6_family != AF_INET6)
927 		return (EINVAL);
928 
929 	/*
930 	 * validate ifra_prefixmask.  don't check sin6_family, netmask
931 	 * does not carry fields other than sin6_len.
932 	 */
933 	if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6))
934 		return (EINVAL);
935 	/*
936 	 * Because the IPv6 address architecture is classless, we require
937 	 * users to specify a (non 0) prefix length (mask) for a new address.
938 	 * We also require the prefix (when specified) mask is valid, and thus
939 	 * reject a non-consecutive mask.
940 	 */
941 	if (ia == NULL && ifra->ifra_prefixmask.sin6_len == 0)
942 		return (EINVAL);
943 	if (ifra->ifra_prefixmask.sin6_len != 0) {
944 		plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
945 		    (u_char *)&ifra->ifra_prefixmask +
946 		    ifra->ifra_prefixmask.sin6_len);
947 		if (plen <= 0)
948 			return (EINVAL);
949 	} else {
950 		/*
951 		 * In this case, ia must not be NULL.  We just use its prefix
952 		 * length.
953 		 */
954 		plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
955 	}
956 	/*
957 	 * If the destination address on a p2p interface is specified,
958 	 * and the address is a scoped one, validate/set the scope
959 	 * zone identifier.
960 	 */
961 	dst6 = ifra->ifra_dstaddr;
962 	if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) != 0 &&
963 	    (dst6.sin6_family == AF_INET6)) {
964 		struct in6_addr in6_tmp;
965 		u_int32_t zoneid;
966 
967 		in6_tmp = dst6.sin6_addr;
968 		if (in6_setscope(&in6_tmp, ifp, &zoneid))
969 			return (EINVAL); /* XXX: should be impossible */
970 
971 		if (dst6.sin6_scope_id != 0) {
972 			if (dst6.sin6_scope_id != zoneid)
973 				return (EINVAL);
974 		} else		/* user omit to specify the ID. */
975 			dst6.sin6_scope_id = zoneid;
976 
977 		/* convert into the internal form */
978 		if (sa6_embedscope(&dst6, 0))
979 			return (EINVAL); /* XXX: should be impossible */
980 	}
981 	/* Modify original ifra_dstaddr to reflect changes */
982 	ifra->ifra_dstaddr = dst6;
983 
984 	/*
985 	 * The destination address can be specified only for a p2p or a
986 	 * loopback interface.  If specified, the corresponding prefix length
987 	 * must be 128.
988 	 */
989 	if (ifra->ifra_dstaddr.sin6_family == AF_INET6) {
990 		if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) == 0) {
991 			/* XXX: noisy message */
992 			nd6log((LOG_INFO, "in6_update_ifa: a destination can "
993 			    "be specified for a p2p or a loopback IF only\n"));
994 			return (EINVAL);
995 		}
996 		if (plen != 128) {
997 			nd6log((LOG_INFO, "in6_update_ifa: prefixlen should "
998 			    "be 128 when dstaddr is specified\n"));
999 			return (EINVAL);
1000 		}
1001 	}
1002 	/* lifetime consistency check */
1003 	lt = &ifra->ifra_lifetime;
1004 	if (lt->ia6t_pltime > lt->ia6t_vltime)
1005 		return (EINVAL);
1006 	if (lt->ia6t_vltime == 0) {
1007 		/*
1008 		 * the following log might be noisy, but this is a typical
1009 		 * configuration mistake or a tool's bug.
1010 		 */
1011 		nd6log((LOG_INFO,
1012 		    "in6_update_ifa: valid lifetime is 0 for %s\n",
1013 		    ip6_sprintf(ip6buf, &ifra->ifra_addr.sin6_addr)));
1014 
1015 		if (ia == NULL)
1016 			return (0); /* there's nothing to do */
1017 	}
1018 
1019 	/* Check prefix mask */
1020 	if (ia != NULL && ifra->ifra_prefixmask.sin6_len != 0) {
1021 		/*
1022 		 * We prohibit changing the prefix length of an existing
1023 		 * address, because
1024 		 * + such an operation should be rare in IPv6, and
1025 		 * + the operation would confuse prefix management.
1026 		 */
1027 		if (ia->ia_prefixmask.sin6_len != 0 &&
1028 		    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) != plen) {
1029 			nd6log((LOG_INFO, "in6_validate_ifa: the prefix length "
1030 			    "of an existing %s address should not be changed\n",
1031 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1032 
1033 			return (EINVAL);
1034 		}
1035 	}
1036 
1037 	return (0);
1038 }
1039 
1040 
1041 /*
1042  * Allocate a new ifaddr and link it into chains.
1043  */
1044 static struct in6_ifaddr *
1045 in6_alloc_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra, int flags)
1046 {
1047 	struct in6_ifaddr *ia;
1048 
1049 	/*
1050 	 * When in6_alloc_ifa() is called in a process of a received
1051 	 * RA, it is called under an interrupt context.  So, we should
1052 	 * call malloc with M_NOWAIT.
1053 	 */
1054 	ia = (struct in6_ifaddr *)ifa_alloc(sizeof(*ia), M_NOWAIT);
1055 	if (ia == NULL)
1056 		return (NULL);
1057 	LIST_INIT(&ia->ia6_memberships);
1058 	/* Initialize the address and masks, and put time stamp */
1059 	ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
1060 	ia->ia_addr.sin6_family = AF_INET6;
1061 	ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
1062 	/* XXX: Can we assign ,sin6_addr and skip the rest? */
1063 	ia->ia_addr = ifra->ifra_addr;
1064 	ia->ia6_createtime = time_uptime;
1065 	if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
1066 		/*
1067 		 * Some functions expect that ifa_dstaddr is not
1068 		 * NULL for p2p interfaces.
1069 		 */
1070 		ia->ia_ifa.ifa_dstaddr =
1071 		    (struct sockaddr *)&ia->ia_dstaddr;
1072 	} else {
1073 		ia->ia_ifa.ifa_dstaddr = NULL;
1074 	}
1075 
1076 	/* set prefix mask if any */
1077 	ia->ia_ifa.ifa_netmask = (struct sockaddr *)&ia->ia_prefixmask;
1078 	if (ifra->ifra_prefixmask.sin6_len != 0) {
1079 		ia->ia_prefixmask.sin6_family = AF_INET6;
1080 		ia->ia_prefixmask.sin6_len = ifra->ifra_prefixmask.sin6_len;
1081 		ia->ia_prefixmask.sin6_addr = ifra->ifra_prefixmask.sin6_addr;
1082 	}
1083 
1084 	ia->ia_ifp = ifp;
1085 	ifa_ref(&ia->ia_ifa);			/* if_addrhead */
1086 	IF_ADDR_WLOCK(ifp);
1087 	TAILQ_INSERT_TAIL(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
1088 	IF_ADDR_WUNLOCK(ifp);
1089 
1090 	ifa_ref(&ia->ia_ifa);			/* in6_ifaddrhead */
1091 	IN6_IFADDR_WLOCK();
1092 	TAILQ_INSERT_TAIL(&V_in6_ifaddrhead, ia, ia_link);
1093 	LIST_INSERT_HEAD(IN6ADDR_HASH(&ia->ia_addr.sin6_addr), ia, ia6_hash);
1094 	IN6_IFADDR_WUNLOCK();
1095 
1096 	return (ia);
1097 }
1098 
1099 /*
1100  * Update/configure interface address parameters:
1101  *
1102  * 1) Update lifetime
1103  * 2) Update interface metric ad flags
1104  * 3) Notify other subsystems
1105  */
1106 static int
1107 in6_update_ifa_internal(struct ifnet *ifp, struct in6_aliasreq *ifra,
1108     struct in6_ifaddr *ia, int hostIsNew, int flags)
1109 {
1110 	int error;
1111 
1112 	/* update timestamp */
1113 	ia->ia6_updatetime = time_uptime;
1114 
1115 	/*
1116 	 * Set lifetimes.  We do not refer to ia6t_expire and ia6t_preferred
1117 	 * to see if the address is deprecated or invalidated, but initialize
1118 	 * these members for applications.
1119 	 */
1120 	ia->ia6_lifetime = ifra->ifra_lifetime;
1121 	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1122 		ia->ia6_lifetime.ia6t_expire =
1123 		    time_uptime + ia->ia6_lifetime.ia6t_vltime;
1124 	} else
1125 		ia->ia6_lifetime.ia6t_expire = 0;
1126 	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1127 		ia->ia6_lifetime.ia6t_preferred =
1128 		    time_uptime + ia->ia6_lifetime.ia6t_pltime;
1129 	} else
1130 		ia->ia6_lifetime.ia6t_preferred = 0;
1131 
1132 	/*
1133 	 * backward compatibility - if IN6_IFF_DEPRECATED is set from the
1134 	 * userland, make it deprecated.
1135 	 */
1136 	if ((ifra->ifra_flags & IN6_IFF_DEPRECATED) != 0) {
1137 		ia->ia6_lifetime.ia6t_pltime = 0;
1138 		ia->ia6_lifetime.ia6t_preferred = time_uptime;
1139 	}
1140 
1141 	/*
1142 	 * configure address flags.
1143 	 */
1144 	ia->ia6_flags = ifra->ifra_flags;
1145 
1146 	/*
1147 	 * Make the address tentative before joining multicast addresses,
1148 	 * so that corresponding MLD responses would not have a tentative
1149 	 * source address.
1150 	 */
1151 	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/* safety */
1152 
1153 	/*
1154 	 * DAD should be performed for an new address or addresses on
1155 	 * an interface with ND6_IFF_IFDISABLED.
1156 	 */
1157 	if (in6if_do_dad(ifp) &&
1158 	    (hostIsNew || (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)))
1159 		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1160 
1161 	/* notify other subsystems */
1162 	error = in6_notify_ifa(ifp, ia, ifra, hostIsNew);
1163 
1164 	return (error);
1165 }
1166 
1167 /*
1168  * Do link-level ifa job:
1169  * 1) Add lle entry for added address
1170  * 2) Notifies routing socket users about new address
1171  * 3) join appropriate multicast group
1172  * 4) start DAD if enabled
1173  */
1174 static int
1175 in6_broadcast_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
1176     struct in6_ifaddr *ia, int flags)
1177 {
1178 	struct in6_multi *in6m_sol;
1179 	int error = 0;
1180 
1181 	/* Add local address to lltable, if necessary (ex. on p2p link). */
1182 	if ((error = nd6_add_ifa_lle(ia)) != 0) {
1183 		in6_purgeaddr(&ia->ia_ifa);
1184 		ifa_free(&ia->ia_ifa);
1185 		return (error);
1186 	}
1187 
1188 	/* Join necessary multicast groups. */
1189 	in6m_sol = NULL;
1190 	if ((ifp->if_flags & IFF_MULTICAST) != 0) {
1191 		error = in6_update_ifa_join_mc(ifp, ifra, ia, flags, &in6m_sol);
1192 		if (error != 0) {
1193 			in6_purgeaddr(&ia->ia_ifa);
1194 			ifa_free(&ia->ia_ifa);
1195 			return (error);
1196 		}
1197 	}
1198 
1199 	/* Perform DAD, if the address is TENTATIVE. */
1200 	if ((ia->ia6_flags & IN6_IFF_TENTATIVE)) {
1201 		int delay, mindelay, maxdelay;
1202 
1203 		delay = 0;
1204 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1205 			/*
1206 			 * We need to impose a delay before sending an NS
1207 			 * for DAD.  Check if we also needed a delay for the
1208 			 * corresponding MLD message.  If we did, the delay
1209 			 * should be larger than the MLD delay (this could be
1210 			 * relaxed a bit, but this simple logic is at least
1211 			 * safe).
1212 			 * XXX: Break data hiding guidelines and look at
1213 			 * state for the solicited multicast group.
1214 			 */
1215 			mindelay = 0;
1216 			if (in6m_sol != NULL &&
1217 			    in6m_sol->in6m_state == MLD_REPORTING_MEMBER) {
1218 				mindelay = in6m_sol->in6m_timer;
1219 			}
1220 			maxdelay = MAX_RTR_SOLICITATION_DELAY * hz;
1221 			if (maxdelay - mindelay == 0)
1222 				delay = 0;
1223 			else {
1224 				delay =
1225 				    (arc4random() % (maxdelay - mindelay)) +
1226 				    mindelay;
1227 			}
1228 		}
1229 		nd6_dad_start((struct ifaddr *)ia, delay);
1230 	}
1231 
1232 	in6_newaddrmsg(ia, RTM_ADD);
1233 	ifa_free(&ia->ia_ifa);
1234 	return (error);
1235 }
1236 
1237 void
1238 in6_purgeaddr(struct ifaddr *ifa)
1239 {
1240 	struct ifnet *ifp = ifa->ifa_ifp;
1241 	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1242 	struct in6_multi_mship *imm;
1243 	int plen, error;
1244 
1245 	if (ifa->ifa_carp)
1246 		(*carp_detach_p)(ifa);
1247 
1248 	/*
1249 	 * Remove the loopback route to the interface address.
1250 	 * The check for the current setting of "nd6_useloopback"
1251 	 * is not needed.
1252 	 */
1253 	if (ia->ia_flags & IFA_RTSELF) {
1254 		error = ifa_del_loopback_route((struct ifaddr *)ia,
1255 		    (struct sockaddr *)&ia->ia_addr);
1256 		if (error == 0)
1257 			ia->ia_flags &= ~IFA_RTSELF;
1258 	}
1259 
1260 	/* stop DAD processing */
1261 	nd6_dad_stop(ifa);
1262 
1263 	/* Leave multicast groups. */
1264 	while ((imm = LIST_FIRST(&ia->ia6_memberships)) != NULL) {
1265 		LIST_REMOVE(imm, i6mm_chain);
1266 		in6_leavegroup(imm);
1267 	}
1268 	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1269 	if ((ia->ia_flags & IFA_ROUTE) && plen == 128) {
1270 		error = rtinit(&(ia->ia_ifa), RTM_DELETE, ia->ia_flags |
1271 		    (ia->ia_dstaddr.sin6_family == AF_INET6 ? RTF_HOST : 0));
1272 		if (error != 0)
1273 			log(LOG_INFO, "%s: err=%d, destination address delete "
1274 			    "failed\n", __func__, error);
1275 		ia->ia_flags &= ~IFA_ROUTE;
1276 	}
1277 
1278 	in6_newaddrmsg(ia, RTM_DELETE);
1279 	in6_unlink_ifa(ia, ifp);
1280 }
1281 
1282 static void
1283 in6_unlink_ifa(struct in6_ifaddr *ia, struct ifnet *ifp)
1284 {
1285 	char ip6buf[INET6_ADDRSTRLEN];
1286 	int remove_lle;
1287 
1288 	IF_ADDR_WLOCK(ifp);
1289 	TAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
1290 	IF_ADDR_WUNLOCK(ifp);
1291 	ifa_free(&ia->ia_ifa);			/* if_addrhead */
1292 
1293 	/*
1294 	 * Defer the release of what might be the last reference to the
1295 	 * in6_ifaddr so that it can't be freed before the remainder of the
1296 	 * cleanup.
1297 	 */
1298 	IN6_IFADDR_WLOCK();
1299 	TAILQ_REMOVE(&V_in6_ifaddrhead, ia, ia_link);
1300 	LIST_REMOVE(ia, ia6_hash);
1301 	IN6_IFADDR_WUNLOCK();
1302 
1303 	/*
1304 	 * Release the reference to the base prefix.  There should be a
1305 	 * positive reference.
1306 	 */
1307 	remove_lle = 0;
1308 	if (ia->ia6_ndpr == NULL) {
1309 		nd6log((LOG_NOTICE,
1310 		    "in6_unlink_ifa: autoconf'ed address "
1311 		    "%s has no prefix\n", ip6_sprintf(ip6buf, IA6_IN6(ia))));
1312 	} else {
1313 		ia->ia6_ndpr->ndpr_addrcnt--;
1314 		/* Do not delete lles within prefix if refcont != 0 */
1315 		if (ia->ia6_ndpr->ndpr_addrcnt == 0)
1316 			remove_lle = 1;
1317 		ia->ia6_ndpr = NULL;
1318 	}
1319 
1320 	nd6_rem_ifa_lle(ia, remove_lle);
1321 
1322 	/*
1323 	 * Also, if the address being removed is autoconf'ed, call
1324 	 * pfxlist_onlink_check() since the release might affect the status of
1325 	 * other (detached) addresses.
1326 	 */
1327 	if ((ia->ia6_flags & IN6_IFF_AUTOCONF)) {
1328 		pfxlist_onlink_check();
1329 	}
1330 	ifa_free(&ia->ia_ifa);			/* in6_ifaddrhead */
1331 }
1332 
1333 /*
1334  * Notifies other subsystems about address change/arrival:
1335  * 1) Notifies device handler on the first IPv6 address assignment
1336  * 2) Handle routing table changes for P2P links and route
1337  * 3) Handle routing table changes for address host route
1338  */
1339 static int
1340 in6_notify_ifa(struct ifnet *ifp, struct in6_ifaddr *ia,
1341     struct in6_aliasreq *ifra, int hostIsNew)
1342 {
1343 	int	error = 0, plen, ifacount = 0;
1344 	struct ifaddr *ifa;
1345 	struct sockaddr_in6 *pdst;
1346 	char ip6buf[INET6_ADDRSTRLEN];
1347 
1348 	/*
1349 	 * Give the interface a chance to initialize
1350 	 * if this is its first address,
1351 	 */
1352 	if (hostIsNew != 0) {
1353 		IF_ADDR_RLOCK(ifp);
1354 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1355 			if (ifa->ifa_addr->sa_family != AF_INET6)
1356 				continue;
1357 			ifacount++;
1358 		}
1359 		IF_ADDR_RUNLOCK(ifp);
1360 	}
1361 
1362 	if (ifacount <= 1 && ifp->if_ioctl) {
1363 		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
1364 		if (error)
1365 			return (error);
1366 	}
1367 
1368 	/*
1369 	 * If a new destination address is specified, scrub the old one and
1370 	 * install the new destination.  Note that the interface must be
1371 	 * p2p or loopback.
1372 	 */
1373 	pdst = &ifra->ifra_dstaddr;
1374 	if (pdst->sin6_family == AF_INET6 &&
1375 	    !IN6_ARE_ADDR_EQUAL(&pdst->sin6_addr, &ia->ia_dstaddr.sin6_addr)) {
1376 		if ((ia->ia_flags & IFA_ROUTE) != 0 &&
1377 		    (rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST) != 0)) {
1378 			nd6log((LOG_ERR, "in6_update_ifa_internal: failed to "
1379 			    "remove a route to the old destination: %s\n",
1380 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1381 			/* proceed anyway... */
1382 		} else
1383 			ia->ia_flags &= ~IFA_ROUTE;
1384 		ia->ia_dstaddr = *pdst;
1385 	}
1386 
1387 	/*
1388 	 * If a new destination address is specified for a point-to-point
1389 	 * interface, install a route to the destination as an interface
1390 	 * direct route.
1391 	 * XXX: the logic below rejects assigning multiple addresses on a p2p
1392 	 * interface that share the same destination.
1393 	 */
1394 	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1395 	if (!(ia->ia_flags & IFA_ROUTE) && plen == 128 &&
1396 	    ia->ia_dstaddr.sin6_family == AF_INET6) {
1397 		int rtflags = RTF_UP | RTF_HOST;
1398 		/*
1399 		 * Handle the case for ::1 .
1400 		 */
1401 		if (ifp->if_flags & IFF_LOOPBACK)
1402 			ia->ia_flags |= IFA_RTSELF;
1403 		error = rtinit(&ia->ia_ifa, RTM_ADD, ia->ia_flags | rtflags);
1404 		if (error)
1405 			return (error);
1406 		ia->ia_flags |= IFA_ROUTE;
1407 	}
1408 
1409 	/*
1410 	 * add a loopback route to self if not exists
1411 	 */
1412 	if (!(ia->ia_flags & IFA_RTSELF) && V_nd6_useloopback) {
1413 		error = ifa_add_loopback_route((struct ifaddr *)ia,
1414 		    (struct sockaddr *)&ia->ia_addr);
1415 		if (error == 0)
1416 			ia->ia_flags |= IFA_RTSELF;
1417 	}
1418 
1419 	return (error);
1420 }
1421 
1422 /*
1423  * Find an IPv6 interface link-local address specific to an interface.
1424  * ifaddr is returned referenced.
1425  */
1426 struct in6_ifaddr *
1427 in6ifa_ifpforlinklocal(struct ifnet *ifp, int ignoreflags)
1428 {
1429 	struct ifaddr *ifa;
1430 
1431 	IF_ADDR_RLOCK(ifp);
1432 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1433 		if (ifa->ifa_addr->sa_family != AF_INET6)
1434 			continue;
1435 		if (IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa))) {
1436 			if ((((struct in6_ifaddr *)ifa)->ia6_flags &
1437 			    ignoreflags) != 0)
1438 				continue;
1439 			ifa_ref(ifa);
1440 			break;
1441 		}
1442 	}
1443 	IF_ADDR_RUNLOCK(ifp);
1444 
1445 	return ((struct in6_ifaddr *)ifa);
1446 }
1447 
1448 
1449 /*
1450  * find the internet address corresponding to a given address.
1451  * ifaddr is returned referenced.
1452  */
1453 struct in6_ifaddr *
1454 in6ifa_ifwithaddr(const struct in6_addr *addr, uint32_t zoneid)
1455 {
1456 	struct rm_priotracker in6_ifa_tracker;
1457 	struct in6_ifaddr *ia;
1458 
1459 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1460 	LIST_FOREACH(ia, IN6ADDR_HASH(addr), ia6_hash) {
1461 		if (IN6_ARE_ADDR_EQUAL(IA6_IN6(ia), addr)) {
1462 			if (zoneid != 0 &&
1463 			    zoneid != ia->ia_addr.sin6_scope_id)
1464 				continue;
1465 			ifa_ref(&ia->ia_ifa);
1466 			break;
1467 		}
1468 	}
1469 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1470 	return (ia);
1471 }
1472 
1473 /*
1474  * find the internet address corresponding to a given interface and address.
1475  * ifaddr is returned referenced.
1476  */
1477 struct in6_ifaddr *
1478 in6ifa_ifpwithaddr(struct ifnet *ifp, const struct in6_addr *addr)
1479 {
1480 	struct ifaddr *ifa;
1481 
1482 	IF_ADDR_RLOCK(ifp);
1483 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1484 		if (ifa->ifa_addr->sa_family != AF_INET6)
1485 			continue;
1486 		if (IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa))) {
1487 			ifa_ref(ifa);
1488 			break;
1489 		}
1490 	}
1491 	IF_ADDR_RUNLOCK(ifp);
1492 
1493 	return ((struct in6_ifaddr *)ifa);
1494 }
1495 
1496 /*
1497  * Find a link-local scoped address on ifp and return it if any.
1498  */
1499 struct in6_ifaddr *
1500 in6ifa_llaonifp(struct ifnet *ifp)
1501 {
1502 	struct sockaddr_in6 *sin6;
1503 	struct ifaddr *ifa;
1504 
1505 	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)
1506 		return (NULL);
1507 	IF_ADDR_RLOCK(ifp);
1508 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1509 		if (ifa->ifa_addr->sa_family != AF_INET6)
1510 			continue;
1511 		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
1512 		if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr) ||
1513 		    IN6_IS_ADDR_MC_INTFACELOCAL(&sin6->sin6_addr) ||
1514 		    IN6_IS_ADDR_MC_NODELOCAL(&sin6->sin6_addr))
1515 			break;
1516 	}
1517 	IF_ADDR_RUNLOCK(ifp);
1518 
1519 	return ((struct in6_ifaddr *)ifa);
1520 }
1521 
1522 /*
1523  * Convert IP6 address to printable (loggable) representation. Caller
1524  * has to make sure that ip6buf is at least INET6_ADDRSTRLEN long.
1525  */
1526 static char digits[] = "0123456789abcdef";
1527 char *
1528 ip6_sprintf(char *ip6buf, const struct in6_addr *addr)
1529 {
1530 	int i, cnt = 0, maxcnt = 0, idx = 0, index = 0;
1531 	char *cp;
1532 	const u_int16_t *a = (const u_int16_t *)addr;
1533 	const u_int8_t *d;
1534 	int dcolon = 0, zero = 0;
1535 
1536 	cp = ip6buf;
1537 
1538 	for (i = 0; i < 8; i++) {
1539 		if (*(a + i) == 0) {
1540 			cnt++;
1541 			if (cnt == 1)
1542 				idx = i;
1543 		}
1544 		else if (maxcnt < cnt) {
1545 			maxcnt = cnt;
1546 			index = idx;
1547 			cnt = 0;
1548 		}
1549 	}
1550 	if (maxcnt < cnt) {
1551 		maxcnt = cnt;
1552 		index = idx;
1553 	}
1554 
1555 	for (i = 0; i < 8; i++) {
1556 		if (dcolon == 1) {
1557 			if (*a == 0) {
1558 				if (i == 7)
1559 					*cp++ = ':';
1560 				a++;
1561 				continue;
1562 			} else
1563 				dcolon = 2;
1564 		}
1565 		if (*a == 0) {
1566 			if (dcolon == 0 && *(a + 1) == 0 && i == index) {
1567 				if (i == 0)
1568 					*cp++ = ':';
1569 				*cp++ = ':';
1570 				dcolon = 1;
1571 			} else {
1572 				*cp++ = '0';
1573 				*cp++ = ':';
1574 			}
1575 			a++;
1576 			continue;
1577 		}
1578 		d = (const u_char *)a;
1579 		/* Try to eliminate leading zeros in printout like in :0001. */
1580 		zero = 1;
1581 		*cp = digits[*d >> 4];
1582 		if (*cp != '0') {
1583 			zero = 0;
1584 			cp++;
1585 		}
1586 		*cp = digits[*d++ & 0xf];
1587 		if (zero == 0 || (*cp != '0')) {
1588 			zero = 0;
1589 			cp++;
1590 		}
1591 		*cp = digits[*d >> 4];
1592 		if (zero == 0 || (*cp != '0')) {
1593 			zero = 0;
1594 			cp++;
1595 		}
1596 		*cp++ = digits[*d & 0xf];
1597 		*cp++ = ':';
1598 		a++;
1599 	}
1600 	*--cp = '\0';
1601 	return (ip6buf);
1602 }
1603 
1604 int
1605 in6_localaddr(struct in6_addr *in6)
1606 {
1607 	struct rm_priotracker in6_ifa_tracker;
1608 	struct in6_ifaddr *ia;
1609 
1610 	if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
1611 		return 1;
1612 
1613 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1614 	TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
1615 		if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
1616 		    &ia->ia_prefixmask.sin6_addr)) {
1617 			IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1618 			return 1;
1619 		}
1620 	}
1621 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1622 
1623 	return (0);
1624 }
1625 
1626 /*
1627  * Return 1 if an internet address is for the local host and configured
1628  * on one of its interfaces.
1629  */
1630 int
1631 in6_localip(struct in6_addr *in6)
1632 {
1633 	struct rm_priotracker in6_ifa_tracker;
1634 	struct in6_ifaddr *ia;
1635 
1636 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1637 	LIST_FOREACH(ia, IN6ADDR_HASH(in6), ia6_hash) {
1638 		if (IN6_ARE_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr)) {
1639 			IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1640 			return (1);
1641 		}
1642 	}
1643 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1644 	return (0);
1645 }
1646 
1647 /*
1648  * Return 1 if an internet address is configured on an interface.
1649  */
1650 int
1651 in6_ifhasaddr(struct ifnet *ifp, struct in6_addr *addr)
1652 {
1653 	struct in6_addr in6;
1654 	struct ifaddr *ifa;
1655 	struct in6_ifaddr *ia6;
1656 
1657 	in6 = *addr;
1658 	if (in6_clearscope(&in6))
1659 		return (0);
1660 	in6_setscope(&in6, ifp, NULL);
1661 
1662 	IF_ADDR_RLOCK(ifp);
1663 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1664 		if (ifa->ifa_addr->sa_family != AF_INET6)
1665 			continue;
1666 		ia6 = (struct in6_ifaddr *)ifa;
1667 		if (IN6_ARE_ADDR_EQUAL(&ia6->ia_addr.sin6_addr, &in6)) {
1668 			IF_ADDR_RUNLOCK(ifp);
1669 			return (1);
1670 		}
1671 	}
1672 	IF_ADDR_RUNLOCK(ifp);
1673 
1674 	return (0);
1675 }
1676 
1677 int
1678 in6_is_addr_deprecated(struct sockaddr_in6 *sa6)
1679 {
1680 	struct rm_priotracker in6_ifa_tracker;
1681 	struct in6_ifaddr *ia;
1682 
1683 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1684 	LIST_FOREACH(ia, IN6ADDR_HASH(&sa6->sin6_addr), ia6_hash) {
1685 		if (IN6_ARE_ADDR_EQUAL(IA6_IN6(ia), &sa6->sin6_addr)) {
1686 			if (ia->ia6_flags & IN6_IFF_DEPRECATED) {
1687 				IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1688 				return (1); /* true */
1689 			}
1690 			break;
1691 		}
1692 	}
1693 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1694 
1695 	return (0);		/* false */
1696 }
1697 
1698 /*
1699  * return length of part which dst and src are equal
1700  * hard coding...
1701  */
1702 int
1703 in6_matchlen(struct in6_addr *src, struct in6_addr *dst)
1704 {
1705 	int match = 0;
1706 	u_char *s = (u_char *)src, *d = (u_char *)dst;
1707 	u_char *lim = s + 16, r;
1708 
1709 	while (s < lim)
1710 		if ((r = (*d++ ^ *s++)) != 0) {
1711 			while (r < 128) {
1712 				match++;
1713 				r <<= 1;
1714 			}
1715 			break;
1716 		} else
1717 			match += 8;
1718 	return match;
1719 }
1720 
1721 /* XXX: to be scope conscious */
1722 int
1723 in6_are_prefix_equal(struct in6_addr *p1, struct in6_addr *p2, int len)
1724 {
1725 	int bytelen, bitlen;
1726 
1727 	/* sanity check */
1728 	if (0 > len || len > 128) {
1729 		log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n",
1730 		    len);
1731 		return (0);
1732 	}
1733 
1734 	bytelen = len / 8;
1735 	bitlen = len % 8;
1736 
1737 	if (bcmp(&p1->s6_addr, &p2->s6_addr, bytelen))
1738 		return (0);
1739 	if (bitlen != 0 &&
1740 	    p1->s6_addr[bytelen] >> (8 - bitlen) !=
1741 	    p2->s6_addr[bytelen] >> (8 - bitlen))
1742 		return (0);
1743 
1744 	return (1);
1745 }
1746 
1747 void
1748 in6_prefixlen2mask(struct in6_addr *maskp, int len)
1749 {
1750 	u_char maskarray[8] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
1751 	int bytelen, bitlen, i;
1752 
1753 	/* sanity check */
1754 	if (0 > len || len > 128) {
1755 		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
1756 		    len);
1757 		return;
1758 	}
1759 
1760 	bzero(maskp, sizeof(*maskp));
1761 	bytelen = len / 8;
1762 	bitlen = len % 8;
1763 	for (i = 0; i < bytelen; i++)
1764 		maskp->s6_addr[i] = 0xff;
1765 	if (bitlen)
1766 		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
1767 }
1768 
1769 /*
1770  * return the best address out of the same scope. if no address was
1771  * found, return the first valid address from designated IF.
1772  */
1773 struct in6_ifaddr *
1774 in6_ifawithifp(struct ifnet *ifp, struct in6_addr *dst)
1775 {
1776 	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
1777 	struct ifaddr *ifa;
1778 	struct in6_ifaddr *besta = NULL;
1779 	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
1780 
1781 	dep[0] = dep[1] = NULL;
1782 
1783 	/*
1784 	 * We first look for addresses in the same scope.
1785 	 * If there is one, return it.
1786 	 * If two or more, return one which matches the dst longest.
1787 	 * If none, return one of global addresses assigned other ifs.
1788 	 */
1789 	IF_ADDR_RLOCK(ifp);
1790 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1791 		if (ifa->ifa_addr->sa_family != AF_INET6)
1792 			continue;
1793 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
1794 			continue; /* XXX: is there any case to allow anycast? */
1795 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
1796 			continue; /* don't use this interface */
1797 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
1798 			continue;
1799 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
1800 			if (V_ip6_use_deprecated)
1801 				dep[0] = (struct in6_ifaddr *)ifa;
1802 			continue;
1803 		}
1804 
1805 		if (dst_scope == in6_addrscope(IFA_IN6(ifa))) {
1806 			/*
1807 			 * call in6_matchlen() as few as possible
1808 			 */
1809 			if (besta) {
1810 				if (blen == -1)
1811 					blen = in6_matchlen(&besta->ia_addr.sin6_addr, dst);
1812 				tlen = in6_matchlen(IFA_IN6(ifa), dst);
1813 				if (tlen > blen) {
1814 					blen = tlen;
1815 					besta = (struct in6_ifaddr *)ifa;
1816 				}
1817 			} else
1818 				besta = (struct in6_ifaddr *)ifa;
1819 		}
1820 	}
1821 	if (besta) {
1822 		ifa_ref(&besta->ia_ifa);
1823 		IF_ADDR_RUNLOCK(ifp);
1824 		return (besta);
1825 	}
1826 
1827 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1828 		if (ifa->ifa_addr->sa_family != AF_INET6)
1829 			continue;
1830 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
1831 			continue; /* XXX: is there any case to allow anycast? */
1832 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
1833 			continue; /* don't use this interface */
1834 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
1835 			continue;
1836 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
1837 			if (V_ip6_use_deprecated)
1838 				dep[1] = (struct in6_ifaddr *)ifa;
1839 			continue;
1840 		}
1841 
1842 		if (ifa != NULL)
1843 			ifa_ref(ifa);
1844 		IF_ADDR_RUNLOCK(ifp);
1845 		return (struct in6_ifaddr *)ifa;
1846 	}
1847 
1848 	/* use the last-resort values, that are, deprecated addresses */
1849 	if (dep[0]) {
1850 		ifa_ref((struct ifaddr *)dep[0]);
1851 		IF_ADDR_RUNLOCK(ifp);
1852 		return dep[0];
1853 	}
1854 	if (dep[1]) {
1855 		ifa_ref((struct ifaddr *)dep[1]);
1856 		IF_ADDR_RUNLOCK(ifp);
1857 		return dep[1];
1858 	}
1859 
1860 	IF_ADDR_RUNLOCK(ifp);
1861 	return NULL;
1862 }
1863 
1864 /*
1865  * perform DAD when interface becomes IFF_UP.
1866  */
1867 void
1868 in6_if_up(struct ifnet *ifp)
1869 {
1870 	struct ifaddr *ifa;
1871 	struct in6_ifaddr *ia;
1872 
1873 	IF_ADDR_RLOCK(ifp);
1874 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1875 		if (ifa->ifa_addr->sa_family != AF_INET6)
1876 			continue;
1877 		ia = (struct in6_ifaddr *)ifa;
1878 		if (ia->ia6_flags & IN6_IFF_TENTATIVE) {
1879 			/*
1880 			 * The TENTATIVE flag was likely set by hand
1881 			 * beforehand, implicitly indicating the need for DAD.
1882 			 * We may be able to skip the random delay in this
1883 			 * case, but we impose delays just in case.
1884 			 */
1885 			nd6_dad_start(ifa,
1886 			    arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz));
1887 		}
1888 	}
1889 	IF_ADDR_RUNLOCK(ifp);
1890 
1891 	/*
1892 	 * special cases, like 6to4, are handled in in6_ifattach
1893 	 */
1894 	in6_ifattach(ifp, NULL);
1895 }
1896 
1897 int
1898 in6if_do_dad(struct ifnet *ifp)
1899 {
1900 	if ((ifp->if_flags & IFF_LOOPBACK) != 0)
1901 		return (0);
1902 
1903 	if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) ||
1904 	    (ND_IFINFO(ifp)->flags & ND6_IFF_NO_DAD))
1905 		return (0);
1906 
1907 	/*
1908 	 * Our DAD routine requires the interface up and running.
1909 	 * However, some interfaces can be up before the RUNNING
1910 	 * status.  Additionally, users may try to assign addresses
1911 	 * before the interface becomes up (or running).
1912 	 * This function returns EAGAIN in that case.
1913 	 * The caller should mark "tentative" on the address instead of
1914 	 * performing DAD immediately.
1915 	 */
1916 	if (!((ifp->if_flags & IFF_UP) &&
1917 	    (ifp->if_drv_flags & IFF_DRV_RUNNING)))
1918 		return (EAGAIN);
1919 
1920 	return (1);
1921 }
1922 
1923 /*
1924  * Calculate max IPv6 MTU through all the interfaces and store it
1925  * to in6_maxmtu.
1926  */
1927 void
1928 in6_setmaxmtu(void)
1929 {
1930 	unsigned long maxmtu = 0;
1931 	struct ifnet *ifp;
1932 
1933 	IFNET_RLOCK_NOSLEEP();
1934 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1935 		/* this function can be called during ifnet initialization */
1936 		if (!ifp->if_afdata[AF_INET6])
1937 			continue;
1938 		if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
1939 		    IN6_LINKMTU(ifp) > maxmtu)
1940 			maxmtu = IN6_LINKMTU(ifp);
1941 	}
1942 	IFNET_RUNLOCK_NOSLEEP();
1943 	if (maxmtu)	/* update only when maxmtu is positive */
1944 		V_in6_maxmtu = maxmtu;
1945 }
1946 
1947 /*
1948  * Provide the length of interface identifiers to be used for the link attached
1949  * to the given interface.  The length should be defined in "IPv6 over
1950  * xxx-link" document.  Note that address architecture might also define
1951  * the length for a particular set of address prefixes, regardless of the
1952  * link type.  As clarified in rfc2462bis, those two definitions should be
1953  * consistent, and those really are as of August 2004.
1954  */
1955 int
1956 in6_if2idlen(struct ifnet *ifp)
1957 {
1958 	switch (ifp->if_type) {
1959 	case IFT_ETHER:		/* RFC2464 */
1960 	case IFT_PROPVIRTUAL:	/* XXX: no RFC. treat it as ether */
1961 	case IFT_L2VLAN:	/* ditto */
1962 	case IFT_IEEE80211:	/* ditto */
1963 	case IFT_BRIDGE:	/* bridge(4) only does Ethernet-like links */
1964 	case IFT_INFINIBAND:
1965 		return (64);
1966 	case IFT_FDDI:		/* RFC2467 */
1967 		return (64);
1968 	case IFT_ISO88025:	/* RFC2470 (IPv6 over Token Ring) */
1969 		return (64);
1970 	case IFT_PPP:		/* RFC2472 */
1971 		return (64);
1972 	case IFT_ARCNET:	/* RFC2497 */
1973 		return (64);
1974 	case IFT_FRELAY:	/* RFC2590 */
1975 		return (64);
1976 	case IFT_IEEE1394:	/* RFC3146 */
1977 		return (64);
1978 	case IFT_GIF:
1979 		return (64);	/* draft-ietf-v6ops-mech-v2-07 */
1980 	case IFT_LOOP:
1981 		return (64);	/* XXX: is this really correct? */
1982 	default:
1983 		/*
1984 		 * Unknown link type:
1985 		 * It might be controversial to use the today's common constant
1986 		 * of 64 for these cases unconditionally.  For full compliance,
1987 		 * we should return an error in this case.  On the other hand,
1988 		 * if we simply miss the standard for the link type or a new
1989 		 * standard is defined for a new link type, the IFID length
1990 		 * is very likely to be the common constant.  As a compromise,
1991 		 * we always use the constant, but make an explicit notice
1992 		 * indicating the "unknown" case.
1993 		 */
1994 		printf("in6_if2idlen: unknown link type (%d)\n", ifp->if_type);
1995 		return (64);
1996 	}
1997 }
1998 
1999 #include <sys/sysctl.h>
2000 
2001 struct in6_llentry {
2002 	struct llentry		base;
2003 };
2004 
2005 #define	IN6_LLTBL_DEFAULT_HSIZE	32
2006 #define	IN6_LLTBL_HASH(k, h) \
2007 	(((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1))
2008 
2009 /*
2010  * Do actual deallocation of @lle.
2011  */
2012 static void
2013 in6_lltable_destroy_lle_unlocked(struct llentry *lle)
2014 {
2015 
2016 	LLE_LOCK_DESTROY(lle);
2017 	LLE_REQ_DESTROY(lle);
2018 	free(lle, M_LLTABLE);
2019 }
2020 
2021 /*
2022  * Called by LLE_FREE_LOCKED when number of references
2023  * drops to zero.
2024  */
2025 static void
2026 in6_lltable_destroy_lle(struct llentry *lle)
2027 {
2028 
2029 	LLE_WUNLOCK(lle);
2030 	in6_lltable_destroy_lle_unlocked(lle);
2031 }
2032 
2033 static struct llentry *
2034 in6_lltable_new(const struct in6_addr *addr6, u_int flags)
2035 {
2036 	struct in6_llentry *lle;
2037 
2038 	lle = malloc(sizeof(struct in6_llentry), M_LLTABLE, M_NOWAIT | M_ZERO);
2039 	if (lle == NULL)		/* NB: caller generates msg */
2040 		return NULL;
2041 
2042 	lle->base.r_l3addr.addr6 = *addr6;
2043 	lle->base.lle_refcnt = 1;
2044 	lle->base.lle_free = in6_lltable_destroy_lle;
2045 	LLE_LOCK_INIT(&lle->base);
2046 	LLE_REQ_INIT(&lle->base);
2047 	callout_init(&lle->base.lle_timer, 1);
2048 
2049 	return (&lle->base);
2050 }
2051 
2052 static int
2053 in6_lltable_match_prefix(const struct sockaddr *saddr,
2054     const struct sockaddr *smask, u_int flags, struct llentry *lle)
2055 {
2056 	const struct in6_addr *addr, *mask, *lle_addr;
2057 
2058 	addr = &((const struct sockaddr_in6 *)saddr)->sin6_addr;
2059 	mask = &((const struct sockaddr_in6 *)smask)->sin6_addr;
2060 	lle_addr = &lle->r_l3addr.addr6;
2061 
2062 	if (IN6_ARE_MASKED_ADDR_EQUAL(lle_addr, addr, mask) == 0)
2063 		return (0);
2064 
2065 	if (lle->la_flags & LLE_IFADDR) {
2066 
2067 		/*
2068 		 * Delete LLE_IFADDR records IFF address & flag matches.
2069 		 * Note that addr is the interface address within prefix
2070 		 * being matched.
2071 		 */
2072 		if (IN6_ARE_ADDR_EQUAL(addr, lle_addr) &&
2073 		    (flags & LLE_STATIC) != 0)
2074 			return (1);
2075 		return (0);
2076 	}
2077 
2078 	/* flags & LLE_STATIC means deleting both dynamic and static entries */
2079 	if ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC))
2080 		return (1);
2081 
2082 	return (0);
2083 }
2084 
2085 static void
2086 in6_lltable_free_entry(struct lltable *llt, struct llentry *lle)
2087 {
2088 	struct ifnet *ifp;
2089 
2090 	LLE_WLOCK_ASSERT(lle);
2091 	KASSERT(llt != NULL, ("lltable is NULL"));
2092 
2093 	/* Unlink entry from table */
2094 	if ((lle->la_flags & LLE_LINKED) != 0) {
2095 
2096 		ifp = llt->llt_ifp;
2097 		IF_AFDATA_WLOCK_ASSERT(ifp);
2098 		lltable_unlink_entry(llt, lle);
2099 	}
2100 
2101 	if (callout_stop(&lle->lle_timer) > 0)
2102 		LLE_REMREF(lle);
2103 
2104 	llentry_free(lle);
2105 }
2106 
2107 static int
2108 in6_lltable_rtcheck(struct ifnet *ifp,
2109 		    u_int flags,
2110 		    const struct sockaddr *l3addr)
2111 {
2112 	const struct sockaddr_in6 *sin6;
2113 	struct nhop6_basic nh6;
2114 	struct in6_addr dst;
2115 	uint32_t scopeid;
2116 	int error;
2117 	char ip6buf[INET6_ADDRSTRLEN];
2118 
2119 	KASSERT(l3addr->sa_family == AF_INET6,
2120 	    ("sin_family %d", l3addr->sa_family));
2121 
2122 	/* Our local addresses are always only installed on the default FIB. */
2123 
2124 	sin6 = (const struct sockaddr_in6 *)l3addr;
2125 	in6_splitscope(&sin6->sin6_addr, &dst, &scopeid);
2126 	error = fib6_lookup_nh_basic(RT_DEFAULT_FIB, &dst, scopeid, 0, 0, &nh6);
2127 	if (error != 0 || (nh6.nh_flags & NHF_GATEWAY) || nh6.nh_ifp != ifp) {
2128 		struct ifaddr *ifa;
2129 		/*
2130 		 * Create an ND6 cache for an IPv6 neighbor
2131 		 * that is not covered by our own prefix.
2132 		 */
2133 		ifa = ifaof_ifpforaddr(l3addr, ifp);
2134 		if (ifa != NULL) {
2135 			ifa_free(ifa);
2136 			return 0;
2137 		}
2138 		log(LOG_INFO, "IPv6 address: \"%s\" is not on the network\n",
2139 		    ip6_sprintf(ip6buf, &sin6->sin6_addr));
2140 		return EINVAL;
2141 	}
2142 	return 0;
2143 }
2144 
2145 static inline uint32_t
2146 in6_lltable_hash_dst(const struct in6_addr *dst, uint32_t hsize)
2147 {
2148 
2149 	return (IN6_LLTBL_HASH(dst->s6_addr32[3], hsize));
2150 }
2151 
2152 static uint32_t
2153 in6_lltable_hash(const struct llentry *lle, uint32_t hsize)
2154 {
2155 
2156 	return (in6_lltable_hash_dst(&lle->r_l3addr.addr6, hsize));
2157 }
2158 
2159 static void
2160 in6_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
2161 {
2162 	struct sockaddr_in6 *sin6;
2163 
2164 	sin6 = (struct sockaddr_in6 *)sa;
2165 	bzero(sin6, sizeof(*sin6));
2166 	sin6->sin6_family = AF_INET6;
2167 	sin6->sin6_len = sizeof(*sin6);
2168 	sin6->sin6_addr = lle->r_l3addr.addr6;
2169 }
2170 
2171 static inline struct llentry *
2172 in6_lltable_find_dst(struct lltable *llt, const struct in6_addr *dst)
2173 {
2174 	struct llentry *lle;
2175 	struct llentries *lleh;
2176 	u_int hashidx;
2177 
2178 	hashidx = in6_lltable_hash_dst(dst, llt->llt_hsize);
2179 	lleh = &llt->lle_head[hashidx];
2180 	LIST_FOREACH(lle, lleh, lle_next) {
2181 		if (lle->la_flags & LLE_DELETED)
2182 			continue;
2183 		if (IN6_ARE_ADDR_EQUAL(&lle->r_l3addr.addr6, dst))
2184 			break;
2185 	}
2186 
2187 	return (lle);
2188 }
2189 
2190 static void
2191 in6_lltable_delete_entry(struct lltable *llt, struct llentry *lle)
2192 {
2193 
2194 	lle->la_flags |= LLE_DELETED;
2195 	EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_DELETED);
2196 #ifdef DIAGNOSTIC
2197 	log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle);
2198 #endif
2199 	llentry_free(lle);
2200 }
2201 
2202 static struct llentry *
2203 in6_lltable_alloc(struct lltable *llt, u_int flags,
2204 	const struct sockaddr *l3addr)
2205 {
2206 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2207 	struct ifnet *ifp = llt->llt_ifp;
2208 	struct llentry *lle;
2209 	char linkhdr[LLE_MAX_LINKHDR];
2210 	size_t linkhdrsize;
2211 	int lladdr_off;
2212 
2213 	KASSERT(l3addr->sa_family == AF_INET6,
2214 	    ("sin_family %d", l3addr->sa_family));
2215 
2216 	/*
2217 	 * A route that covers the given address must have
2218 	 * been installed 1st because we are doing a resolution,
2219 	 * verify this.
2220 	 */
2221 	if (!(flags & LLE_IFADDR) &&
2222 	    in6_lltable_rtcheck(ifp, flags, l3addr) != 0)
2223 		return (NULL);
2224 
2225 	lle = in6_lltable_new(&sin6->sin6_addr, flags);
2226 	if (lle == NULL) {
2227 		log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
2228 		return (NULL);
2229 	}
2230 	lle->la_flags = flags;
2231 	if ((flags & LLE_IFADDR) == LLE_IFADDR) {
2232 		linkhdrsize = LLE_MAX_LINKHDR;
2233 		if (lltable_calc_llheader(ifp, AF_INET6, IF_LLADDR(ifp),
2234 		    linkhdr, &linkhdrsize, &lladdr_off) != 0) {
2235 			in6_lltable_destroy_lle_unlocked(lle);
2236 			return (NULL);
2237 		}
2238 		lltable_set_entry_addr(ifp, lle, linkhdr, linkhdrsize,
2239 		    lladdr_off);
2240 		lle->la_flags |= LLE_STATIC;
2241 	}
2242 
2243 	if ((lle->la_flags & LLE_STATIC) != 0)
2244 		lle->ln_state = ND6_LLINFO_REACHABLE;
2245 
2246 	return (lle);
2247 }
2248 
2249 static struct llentry *
2250 in6_lltable_lookup(struct lltable *llt, u_int flags,
2251 	const struct sockaddr *l3addr)
2252 {
2253 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2254 	struct llentry *lle;
2255 
2256 	IF_AFDATA_LOCK_ASSERT(llt->llt_ifp);
2257 	KASSERT(l3addr->sa_family == AF_INET6,
2258 	    ("sin_family %d", l3addr->sa_family));
2259 
2260 	lle = in6_lltable_find_dst(llt, &sin6->sin6_addr);
2261 
2262 	if (lle == NULL)
2263 		return (NULL);
2264 
2265 	KASSERT((flags & (LLE_UNLOCKED|LLE_EXCLUSIVE)) !=
2266 	    (LLE_UNLOCKED|LLE_EXCLUSIVE),("wrong lle request flags: 0x%X",
2267 	    flags));
2268 
2269 	if (flags & LLE_UNLOCKED)
2270 		return (lle);
2271 
2272 	if (flags & LLE_EXCLUSIVE)
2273 		LLE_WLOCK(lle);
2274 	else
2275 		LLE_RLOCK(lle);
2276 	return (lle);
2277 }
2278 
2279 static int
2280 in6_lltable_dump_entry(struct lltable *llt, struct llentry *lle,
2281     struct sysctl_req *wr)
2282 {
2283 	struct ifnet *ifp = llt->llt_ifp;
2284 	/* XXX stack use */
2285 	struct {
2286 		struct rt_msghdr	rtm;
2287 		struct sockaddr_in6	sin6;
2288 		/*
2289 		 * ndp.c assumes that sdl is word aligned
2290 		 */
2291 #ifdef __LP64__
2292 		uint32_t		pad;
2293 #endif
2294 		struct sockaddr_dl	sdl;
2295 	} ndpc;
2296 	struct sockaddr_dl *sdl;
2297 	int error;
2298 
2299 	bzero(&ndpc, sizeof(ndpc));
2300 			/* skip deleted entries */
2301 			if ((lle->la_flags & LLE_DELETED) == LLE_DELETED)
2302 				return (0);
2303 			/* Skip if jailed and not a valid IP of the prison. */
2304 			lltable_fill_sa_entry(lle,
2305 			    (struct sockaddr *)&ndpc.sin6);
2306 			if (prison_if(wr->td->td_ucred,
2307 			    (struct sockaddr *)&ndpc.sin6) != 0)
2308 				return (0);
2309 			/*
2310 			 * produce a msg made of:
2311 			 *  struct rt_msghdr;
2312 			 *  struct sockaddr_in6 (IPv6)
2313 			 *  struct sockaddr_dl;
2314 			 */
2315 			ndpc.rtm.rtm_msglen = sizeof(ndpc);
2316 			ndpc.rtm.rtm_version = RTM_VERSION;
2317 			ndpc.rtm.rtm_type = RTM_GET;
2318 			ndpc.rtm.rtm_flags = RTF_UP;
2319 			ndpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
2320 			if (V_deembed_scopeid)
2321 				sa6_recoverscope(&ndpc.sin6);
2322 
2323 			/* publish */
2324 			if (lle->la_flags & LLE_PUB)
2325 				ndpc.rtm.rtm_flags |= RTF_ANNOUNCE;
2326 
2327 			sdl = &ndpc.sdl;
2328 			sdl->sdl_family = AF_LINK;
2329 			sdl->sdl_len = sizeof(*sdl);
2330 			sdl->sdl_index = ifp->if_index;
2331 			sdl->sdl_type = ifp->if_type;
2332 			if ((lle->la_flags & LLE_VALID) == LLE_VALID) {
2333 				sdl->sdl_alen = ifp->if_addrlen;
2334 				bcopy(lle->ll_addr, LLADDR(sdl),
2335 				    ifp->if_addrlen);
2336 			} else {
2337 				sdl->sdl_alen = 0;
2338 				bzero(LLADDR(sdl), ifp->if_addrlen);
2339 			}
2340 			if (lle->la_expire != 0)
2341 				ndpc.rtm.rtm_rmx.rmx_expire = lle->la_expire +
2342 				    lle->lle_remtime / hz +
2343 				    time_second - time_uptime;
2344 			ndpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
2345 			if (lle->la_flags & LLE_STATIC)
2346 				ndpc.rtm.rtm_flags |= RTF_STATIC;
2347 			if (lle->la_flags & LLE_IFADDR)
2348 				ndpc.rtm.rtm_flags |= RTF_PINNED;
2349 			if (lle->ln_router != 0)
2350 				ndpc.rtm.rtm_flags |= RTF_GATEWAY;
2351 			ndpc.rtm.rtm_rmx.rmx_pksent = lle->la_asked;
2352 			/* Store state in rmx_weight value */
2353 			ndpc.rtm.rtm_rmx.rmx_state = lle->ln_state;
2354 			ndpc.rtm.rtm_index = ifp->if_index;
2355 			error = SYSCTL_OUT(wr, &ndpc, sizeof(ndpc));
2356 
2357 	return (error);
2358 }
2359 
2360 static struct lltable *
2361 in6_lltattach(struct ifnet *ifp)
2362 {
2363 	struct lltable *llt;
2364 
2365 	llt = lltable_allocate_htbl(IN6_LLTBL_DEFAULT_HSIZE);
2366 	llt->llt_af = AF_INET6;
2367 	llt->llt_ifp = ifp;
2368 
2369 	llt->llt_lookup = in6_lltable_lookup;
2370 	llt->llt_alloc_entry = in6_lltable_alloc;
2371 	llt->llt_delete_entry = in6_lltable_delete_entry;
2372 	llt->llt_dump_entry = in6_lltable_dump_entry;
2373 	llt->llt_hash = in6_lltable_hash;
2374 	llt->llt_fill_sa_entry = in6_lltable_fill_sa_entry;
2375 	llt->llt_free_entry = in6_lltable_free_entry;
2376 	llt->llt_match_prefix = in6_lltable_match_prefix;
2377  	lltable_link(llt);
2378 
2379 	return (llt);
2380 }
2381 
2382 void *
2383 in6_domifattach(struct ifnet *ifp)
2384 {
2385 	struct in6_ifextra *ext;
2386 
2387 	/* There are not IPv6-capable interfaces. */
2388 	switch (ifp->if_type) {
2389 	case IFT_PFLOG:
2390 	case IFT_PFSYNC:
2391 	case IFT_USB:
2392 		return (NULL);
2393 	}
2394 	ext = (struct in6_ifextra *)malloc(sizeof(*ext), M_IFADDR, M_WAITOK);
2395 	bzero(ext, sizeof(*ext));
2396 
2397 	ext->in6_ifstat = malloc(sizeof(counter_u64_t) *
2398 	    sizeof(struct in6_ifstat) / sizeof(uint64_t), M_IFADDR, M_WAITOK);
2399 	COUNTER_ARRAY_ALLOC(ext->in6_ifstat,
2400 	    sizeof(struct in6_ifstat) / sizeof(uint64_t), M_WAITOK);
2401 
2402 	ext->icmp6_ifstat = malloc(sizeof(counter_u64_t) *
2403 	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t), M_IFADDR,
2404 	    M_WAITOK);
2405 	COUNTER_ARRAY_ALLOC(ext->icmp6_ifstat,
2406 	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t), M_WAITOK);
2407 
2408 	ext->nd_ifinfo = nd6_ifattach(ifp);
2409 	ext->scope6_id = scope6_ifattach(ifp);
2410 	ext->lltable = in6_lltattach(ifp);
2411 
2412 	ext->mld_ifinfo = mld_domifattach(ifp);
2413 
2414 	return ext;
2415 }
2416 
2417 int
2418 in6_domifmtu(struct ifnet *ifp)
2419 {
2420 	if (ifp->if_afdata[AF_INET6] == NULL)
2421 		return ifp->if_mtu;
2422 
2423 	return (IN6_LINKMTU(ifp));
2424 }
2425 
2426 void
2427 in6_domifdetach(struct ifnet *ifp, void *aux)
2428 {
2429 	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
2430 
2431 	mld_domifdetach(ifp);
2432 	scope6_ifdetach(ext->scope6_id);
2433 	nd6_ifdetach(ifp, ext->nd_ifinfo);
2434 	lltable_free(ext->lltable);
2435 	COUNTER_ARRAY_FREE(ext->in6_ifstat,
2436 	    sizeof(struct in6_ifstat) / sizeof(uint64_t));
2437 	free(ext->in6_ifstat, M_IFADDR);
2438 	COUNTER_ARRAY_FREE(ext->icmp6_ifstat,
2439 	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t));
2440 	free(ext->icmp6_ifstat, M_IFADDR);
2441 	free(ext, M_IFADDR);
2442 }
2443 
2444 /*
2445  * Convert sockaddr_in6 to sockaddr_in.  Original sockaddr_in6 must be
2446  * v4 mapped addr or v4 compat addr
2447  */
2448 void
2449 in6_sin6_2_sin(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2450 {
2451 
2452 	bzero(sin, sizeof(*sin));
2453 	sin->sin_len = sizeof(struct sockaddr_in);
2454 	sin->sin_family = AF_INET;
2455 	sin->sin_port = sin6->sin6_port;
2456 	sin->sin_addr.s_addr = sin6->sin6_addr.s6_addr32[3];
2457 }
2458 
2459 /* Convert sockaddr_in to sockaddr_in6 in v4 mapped addr format. */
2460 void
2461 in6_sin_2_v4mapsin6(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2462 {
2463 	bzero(sin6, sizeof(*sin6));
2464 	sin6->sin6_len = sizeof(struct sockaddr_in6);
2465 	sin6->sin6_family = AF_INET6;
2466 	sin6->sin6_port = sin->sin_port;
2467 	sin6->sin6_addr.s6_addr32[0] = 0;
2468 	sin6->sin6_addr.s6_addr32[1] = 0;
2469 	sin6->sin6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
2470 	sin6->sin6_addr.s6_addr32[3] = sin->sin_addr.s_addr;
2471 }
2472 
2473 /* Convert sockaddr_in6 into sockaddr_in. */
2474 void
2475 in6_sin6_2_sin_in_sock(struct sockaddr *nam)
2476 {
2477 	struct sockaddr_in *sin_p;
2478 	struct sockaddr_in6 sin6;
2479 
2480 	/*
2481 	 * Save original sockaddr_in6 addr and convert it
2482 	 * to sockaddr_in.
2483 	 */
2484 	sin6 = *(struct sockaddr_in6 *)nam;
2485 	sin_p = (struct sockaddr_in *)nam;
2486 	in6_sin6_2_sin(sin_p, &sin6);
2487 }
2488 
2489 /* Convert sockaddr_in into sockaddr_in6 in v4 mapped addr format. */
2490 void
2491 in6_sin_2_v4mapsin6_in_sock(struct sockaddr **nam)
2492 {
2493 	struct sockaddr_in *sin_p;
2494 	struct sockaddr_in6 *sin6_p;
2495 
2496 	sin6_p = malloc(sizeof *sin6_p, M_SONAME, M_WAITOK);
2497 	sin_p = (struct sockaddr_in *)*nam;
2498 	in6_sin_2_v4mapsin6(sin_p, sin6_p);
2499 	free(*nam, M_SONAME);
2500 	*nam = (struct sockaddr *)sin6_p;
2501 }
2502