xref: /freebsd/sys/netinet6/in6.c (revision 0263a1a1bcdcab7c436c9ad3e27260bff845f11b)
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 SIOCSIFALIFETIME_IN6:
364 	case SIOCGIFSTAT_IN6:
365 	case SIOCGIFSTAT_ICMP6:
366 		sa6 = &ifr->ifr_addr;
367 		break;
368 	case SIOCSIFADDR:
369 	case SIOCSIFBRDADDR:
370 	case SIOCSIFDSTADDR:
371 	case SIOCSIFNETMASK:
372 		/*
373 		 * Although we should pass any non-INET6 ioctl requests
374 		 * down to driver, we filter some legacy INET requests.
375 		 * Drivers trust SIOCSIFADDR et al to come from an already
376 		 * privileged layer, and do not perform any credentials
377 		 * checks or input validation.
378 		 */
379 		return (EINVAL);
380 	default:
381 		sa6 = NULL;
382 		break;
383 	}
384 	if (sa6 && sa6->sin6_family == AF_INET6) {
385 		if (sa6->sin6_scope_id != 0)
386 			error = sa6_embedscope(sa6, 0);
387 		else
388 			error = in6_setscope(&sa6->sin6_addr, ifp, NULL);
389 		if (error != 0)
390 			return (error);
391 		if (td != NULL && (error = prison_check_ip6(td->td_ucred,
392 		    &sa6->sin6_addr)) != 0)
393 			return (error);
394 		ia = in6ifa_ifpwithaddr(ifp, &sa6->sin6_addr);
395 	} else
396 		ia = NULL;
397 
398 	switch (cmd) {
399 	case SIOCSIFADDR_IN6:
400 	case SIOCSIFDSTADDR_IN6:
401 	case SIOCSIFNETMASK_IN6:
402 		/*
403 		 * Since IPv6 allows a node to assign multiple addresses
404 		 * on a single interface, SIOCSIFxxx ioctls are deprecated.
405 		 */
406 		/* we decided to obsolete this command (20000704) */
407 		error = EINVAL;
408 		goto out;
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 			error = EADDRNOTAVAIL;
420 			goto out;
421 		}
422 		/* FALLTHROUGH */
423 	case SIOCAIFADDR_IN6:
424 		/*
425 		 * We always require users to specify a valid IPv6 address for
426 		 * the corresponding operation.
427 		 */
428 		if (ifra->ifra_addr.sin6_family != AF_INET6 ||
429 		    ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6)) {
430 			error = EAFNOSUPPORT;
431 			goto out;
432 		}
433 
434 		if (td != NULL) {
435 			error = priv_check(td, (cmd == SIOCDIFADDR_IN6) ?
436 			    PRIV_NET_DELIFADDR : PRIV_NET_ADDIFADDR);
437 			if (error)
438 				goto out;
439 		}
440 		/* FALLTHROUGH */
441 	case SIOCGIFSTAT_IN6:
442 	case SIOCGIFSTAT_ICMP6:
443 		if (ifp->if_afdata[AF_INET6] == NULL) {
444 			error = EPFNOSUPPORT;
445 			goto out;
446 		}
447 		break;
448 
449 	case SIOCGIFADDR_IN6:
450 		/* This interface is basically deprecated. use SIOCGIFCONF. */
451 		/* FALLTHROUGH */
452 	case SIOCGIFAFLAG_IN6:
453 	case SIOCGIFNETMASK_IN6:
454 	case SIOCGIFDSTADDR_IN6:
455 	case SIOCGIFALIFETIME_IN6:
456 		/* must think again about its semantics */
457 		if (ia == NULL) {
458 			error = EADDRNOTAVAIL;
459 			goto out;
460 		}
461 		break;
462 
463 	case SIOCSIFALIFETIME_IN6:
464 	    {
465 		struct in6_addrlifetime *lt;
466 
467 		if (td != NULL) {
468 			error = priv_check(td, PRIV_NETINET_ALIFETIME6);
469 			if (error)
470 				goto out;
471 		}
472 		if (ia == NULL) {
473 			error = EADDRNOTAVAIL;
474 			goto out;
475 		}
476 		/* sanity for overflow - beware unsigned */
477 		lt = &ifr->ifr_ifru.ifru_lifetime;
478 		if (lt->ia6t_vltime != ND6_INFINITE_LIFETIME &&
479 		    lt->ia6t_vltime + time_uptime < time_uptime) {
480 			error = EINVAL;
481 			goto out;
482 		}
483 		if (lt->ia6t_pltime != ND6_INFINITE_LIFETIME &&
484 		    lt->ia6t_pltime + time_uptime < time_uptime) {
485 			error = EINVAL;
486 			goto out;
487 		}
488 		break;
489 	    }
490 	}
491 
492 	switch (cmd) {
493 	case SIOCGIFADDR_IN6:
494 		ifr->ifr_addr = ia->ia_addr;
495 		if ((error = sa6_recoverscope(&ifr->ifr_addr)) != 0)
496 			goto out;
497 		break;
498 
499 	case SIOCGIFDSTADDR_IN6:
500 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
501 			error = EINVAL;
502 			goto out;
503 		}
504 		/*
505 		 * XXX: should we check if ifa_dstaddr is NULL and return
506 		 * an error?
507 		 */
508 		ifr->ifr_dstaddr = ia->ia_dstaddr;
509 		if ((error = sa6_recoverscope(&ifr->ifr_dstaddr)) != 0)
510 			goto out;
511 		break;
512 
513 	case SIOCGIFNETMASK_IN6:
514 		ifr->ifr_addr = ia->ia_prefixmask;
515 		break;
516 
517 	case SIOCGIFAFLAG_IN6:
518 		ifr->ifr_ifru.ifru_flags6 = ia->ia6_flags;
519 		break;
520 
521 	case SIOCGIFSTAT_IN6:
522 		COUNTER_ARRAY_COPY(((struct in6_ifextra *)
523 		    ifp->if_afdata[AF_INET6])->in6_ifstat,
524 		    &ifr->ifr_ifru.ifru_stat,
525 		    sizeof(struct in6_ifstat) / sizeof(uint64_t));
526 		break;
527 
528 	case SIOCGIFSTAT_ICMP6:
529 		COUNTER_ARRAY_COPY(((struct in6_ifextra *)
530 		    ifp->if_afdata[AF_INET6])->icmp6_ifstat,
531 		    &ifr->ifr_ifru.ifru_icmp6stat,
532 		    sizeof(struct icmp6_ifstat) / sizeof(uint64_t));
533 		break;
534 
535 	case SIOCGIFALIFETIME_IN6:
536 		ifr->ifr_ifru.ifru_lifetime = ia->ia6_lifetime;
537 		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
538 			time_t maxexpire;
539 			struct in6_addrlifetime *retlt =
540 			    &ifr->ifr_ifru.ifru_lifetime;
541 
542 			/*
543 			 * XXX: adjust expiration time assuming time_t is
544 			 * signed.
545 			 */
546 			maxexpire = (-1) &
547 			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
548 			if (ia->ia6_lifetime.ia6t_vltime <
549 			    maxexpire - ia->ia6_updatetime) {
550 				retlt->ia6t_expire = ia->ia6_updatetime +
551 				    ia->ia6_lifetime.ia6t_vltime;
552 			} else
553 				retlt->ia6t_expire = maxexpire;
554 		}
555 		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
556 			time_t maxexpire;
557 			struct in6_addrlifetime *retlt =
558 			    &ifr->ifr_ifru.ifru_lifetime;
559 
560 			/*
561 			 * XXX: adjust expiration time assuming time_t is
562 			 * signed.
563 			 */
564 			maxexpire = (-1) &
565 			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
566 			if (ia->ia6_lifetime.ia6t_pltime <
567 			    maxexpire - ia->ia6_updatetime) {
568 				retlt->ia6t_preferred = ia->ia6_updatetime +
569 				    ia->ia6_lifetime.ia6t_pltime;
570 			} else
571 				retlt->ia6t_preferred = maxexpire;
572 		}
573 		break;
574 
575 	case SIOCSIFALIFETIME_IN6:
576 		ia->ia6_lifetime = ifr->ifr_ifru.ifru_lifetime;
577 		/* for sanity */
578 		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
579 			ia->ia6_lifetime.ia6t_expire =
580 				time_uptime + ia->ia6_lifetime.ia6t_vltime;
581 		} else
582 			ia->ia6_lifetime.ia6t_expire = 0;
583 		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
584 			ia->ia6_lifetime.ia6t_preferred =
585 				time_uptime + ia->ia6_lifetime.ia6t_pltime;
586 		} else
587 			ia->ia6_lifetime.ia6t_preferred = 0;
588 		break;
589 
590 	case SIOCAIFADDR_IN6:
591 	{
592 		struct nd_prefixctl pr0;
593 		struct nd_prefix *pr;
594 
595 		/*
596 		 * first, make or update the interface address structure,
597 		 * and link it to the list.
598 		 */
599 		if ((error = in6_update_ifa(ifp, ifra, ia, 0)) != 0)
600 			goto out;
601 		if (ia != NULL)
602 			ifa_free(&ia->ia_ifa);
603 		if ((ia = in6ifa_ifpwithaddr(ifp, &ifra->ifra_addr.sin6_addr))
604 		    == NULL) {
605 			/*
606 			 * this can happen when the user specify the 0 valid
607 			 * lifetime.
608 			 */
609 			break;
610 		}
611 
612 		if (cmd == ocmd && ifra->ifra_vhid > 0) {
613 			if (carp_attach_p != NULL)
614 				error = (*carp_attach_p)(&ia->ia_ifa,
615 				    ifra->ifra_vhid);
616 			else
617 				error = EPROTONOSUPPORT;
618 			if (error)
619 				goto out;
620 			else
621 				carp_attached = 1;
622 		}
623 
624 		/*
625 		 * then, make the prefix on-link on the interface.
626 		 * XXX: we'd rather create the prefix before the address, but
627 		 * we need at least one address to install the corresponding
628 		 * interface route, so we configure the address first.
629 		 */
630 
631 		/*
632 		 * convert mask to prefix length (prefixmask has already
633 		 * been validated in in6_update_ifa().
634 		 */
635 		bzero(&pr0, sizeof(pr0));
636 		pr0.ndpr_ifp = ifp;
637 		pr0.ndpr_plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
638 		    NULL);
639 		if (pr0.ndpr_plen == 128) {
640 			/* we don't need to install a host route. */
641 			goto aifaddr_out;
642 		}
643 		pr0.ndpr_prefix = ifra->ifra_addr;
644 		/* apply the mask for safety. */
645 		IN6_MASK_ADDR(&pr0.ndpr_prefix.sin6_addr,
646 		    &ifra->ifra_prefixmask.sin6_addr);
647 
648 		/*
649 		 * XXX: since we don't have an API to set prefix (not address)
650 		 * lifetimes, we just use the same lifetimes as addresses.
651 		 * The (temporarily) installed lifetimes can be overridden by
652 		 * later advertised RAs (when accept_rtadv is non 0), which is
653 		 * an intended behavior.
654 		 */
655 		pr0.ndpr_raf_onlink = 1; /* should be configurable? */
656 		pr0.ndpr_raf_auto =
657 		    ((ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0);
658 		pr0.ndpr_vltime = ifra->ifra_lifetime.ia6t_vltime;
659 		pr0.ndpr_pltime = ifra->ifra_lifetime.ia6t_pltime;
660 
661 		/* add the prefix if not yet. */
662 		if ((pr = nd6_prefix_lookup(&pr0)) == NULL) {
663 			/*
664 			 * nd6_prelist_add will install the corresponding
665 			 * interface route.
666 			 */
667 			if ((error = nd6_prelist_add(&pr0, NULL, &pr)) != 0) {
668 				if (carp_attached)
669 					(*carp_detach_p)(&ia->ia_ifa);
670 				goto out;
671 			}
672 			if (pr == NULL) {
673 				if (carp_attached)
674 					(*carp_detach_p)(&ia->ia_ifa);
675 				log(LOG_ERR, "nd6_prelist_add succeeded but "
676 				    "no prefix\n");
677 				error = EINVAL;
678 				goto out;
679 			}
680 		}
681 
682 		/* relate the address to the prefix */
683 		if (ia->ia6_ndpr == NULL) {
684 			ia->ia6_ndpr = pr;
685 			pr->ndpr_refcnt++;
686 
687 			/*
688 			 * If this is the first autoconf address from the
689 			 * prefix, create a temporary address as well
690 			 * (when required).
691 			 */
692 			if ((ia->ia6_flags & IN6_IFF_AUTOCONF) &&
693 			    V_ip6_use_tempaddr && pr->ndpr_refcnt == 1) {
694 				int e;
695 				if ((e = in6_tmpifadd(ia, 1, 0)) != 0) {
696 					log(LOG_NOTICE, "in6_control: failed "
697 					    "to create a temporary address, "
698 					    "errno=%d\n", e);
699 				}
700 			}
701 		}
702 
703 		/*
704 		 * this might affect the status of autoconfigured addresses,
705 		 * that is, this address might make other addresses detached.
706 		 */
707 		pfxlist_onlink_check();
708 aifaddr_out:
709 		if (error != 0 || ia == NULL)
710 			break;
711 		/*
712 		 * Try to clear the flag when a new IPv6 address is added
713 		 * onto an IFDISABLED interface and it succeeds.
714 		 */
715 		if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) {
716 			struct in6_ndireq nd;
717 
718 			memset(&nd, 0, sizeof(nd));
719 			nd.ndi.flags = ND_IFINFO(ifp)->flags;
720 			nd.ndi.flags &= ~ND6_IFF_IFDISABLED;
721 			if (nd6_ioctl(SIOCSIFINFO_FLAGS, (caddr_t)&nd, ifp) < 0)
722 				log(LOG_NOTICE, "SIOCAIFADDR_IN6: "
723 				    "SIOCSIFINFO_FLAGS for -ifdisabled "
724 				    "failed.");
725 			/*
726 			 * Ignore failure of clearing the flag intentionally.
727 			 * The failure means address duplication was detected.
728 			 */
729 		}
730 		EVENTHANDLER_INVOKE(ifaddr_event, ifp);
731 		break;
732 	}
733 
734 	case SIOCDIFADDR_IN6:
735 	{
736 		struct nd_prefix *pr;
737 
738 		/*
739 		 * If the address being deleted is the only one that owns
740 		 * the corresponding prefix, expire the prefix as well.
741 		 * XXX: theoretically, we don't have to worry about such
742 		 * relationship, since we separate the address management
743 		 * and the prefix management.  We do this, however, to provide
744 		 * as much backward compatibility as possible in terms of
745 		 * the ioctl operation.
746 		 * Note that in6_purgeaddr() will decrement ndpr_refcnt.
747 		 */
748 		pr = ia->ia6_ndpr;
749 		in6_purgeaddr(&ia->ia_ifa);
750 		if (pr && pr->ndpr_refcnt == 0)
751 			prelist_remove(pr);
752 		EVENTHANDLER_INVOKE(ifaddr_event, ifp);
753 		break;
754 	}
755 
756 	default:
757 		if (ifp->if_ioctl == NULL) {
758 			error = EOPNOTSUPP;
759 			goto out;
760 		}
761 		error = (*ifp->if_ioctl)(ifp, cmd, data);
762 		goto out;
763 	}
764 
765 	error = 0;
766 out:
767 	if (ia != NULL)
768 		ifa_free(&ia->ia_ifa);
769 	return (error);
770 }
771 
772 
773 /*
774  * Join necessary multicast groups.  Factored out from in6_update_ifa().
775  * This entire work should only be done once, for the default FIB.
776  */
777 static int
778 in6_update_ifa_join_mc(struct ifnet *ifp, struct in6_aliasreq *ifra,
779     struct in6_ifaddr *ia, int flags, struct in6_multi **in6m_sol)
780 {
781 	char ip6buf[INET6_ADDRSTRLEN];
782 	struct in6_addr mltaddr;
783 	struct in6_multi_mship *imm;
784 	int delay, error;
785 
786 	KASSERT(in6m_sol != NULL, ("%s: in6m_sol is NULL", __func__));
787 
788 	/* Join solicited multicast addr for new host id. */
789 	bzero(&mltaddr, sizeof(struct in6_addr));
790 	mltaddr.s6_addr32[0] = IPV6_ADDR_INT32_MLL;
791 	mltaddr.s6_addr32[2] = htonl(1);
792 	mltaddr.s6_addr32[3] = ifra->ifra_addr.sin6_addr.s6_addr32[3];
793 	mltaddr.s6_addr8[12] = 0xff;
794 	if ((error = in6_setscope(&mltaddr, ifp, NULL)) != 0) {
795 		/* XXX: should not happen */
796 		log(LOG_ERR, "%s: in6_setscope failed\n", __func__);
797 		goto cleanup;
798 	}
799 	delay = error = 0;
800 	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
801 		/*
802 		 * We need a random delay for DAD on the address being
803 		 * configured.  It also means delaying transmission of the
804 		 * corresponding MLD report to avoid report collision.
805 		 * [RFC 4861, Section 6.3.7]
806 		 */
807 		delay = arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz);
808 	}
809 	imm = in6_joingroup(ifp, &mltaddr, &error, delay);
810 	if (imm == NULL) {
811 		nd6log((LOG_WARNING, "%s: in6_joingroup failed for %s on %s "
812 		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf, &mltaddr),
813 		    if_name(ifp), error));
814 		goto cleanup;
815 	}
816 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
817 	*in6m_sol = imm->i6mm_maddr;
818 
819 	/*
820 	 * Join link-local all-nodes address.
821 	 */
822 	mltaddr = in6addr_linklocal_allnodes;
823 	if ((error = in6_setscope(&mltaddr, ifp, NULL)) != 0)
824 		goto cleanup; /* XXX: should not fail */
825 
826 	imm = in6_joingroup(ifp, &mltaddr, &error, 0);
827 	if (imm == NULL) {
828 		nd6log((LOG_WARNING, "%s: in6_joingroup failed for %s on %s "
829 		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf, &mltaddr),
830 		    if_name(ifp), error));
831 		goto cleanup;
832 	}
833 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
834 
835 	/*
836 	 * Join node information group address.
837 	 */
838 	delay = 0;
839 	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
840 		/*
841 		 * The spec does not say anything about delay for this group,
842 		 * but the same logic should apply.
843 		 */
844 		delay = arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz);
845 	}
846 	if (in6_nigroup(ifp, NULL, -1, &mltaddr) == 0) {
847 		/* XXX jinmei */
848 		imm = in6_joingroup(ifp, &mltaddr, &error, delay);
849 		if (imm == NULL)
850 			nd6log((LOG_WARNING,
851 			    "%s: in6_joingroup failed for %s on %s "
852 			    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
853 			    &mltaddr), if_name(ifp), error));
854 			/* XXX not very fatal, go on... */
855 		else
856 			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
857 	}
858 	if (V_icmp6_nodeinfo_oldmcprefix &&
859 	    in6_nigroup_oldmcprefix(ifp, NULL, -1, &mltaddr) == 0) {
860 		imm = in6_joingroup(ifp, &mltaddr, &error, delay);
861 		if (imm == NULL)
862 			nd6log((LOG_WARNING,
863 			    "%s: in6_joingroup failed for %s on %s "
864 			    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
865 			    &mltaddr), if_name(ifp), error));
866 			/* XXX not very fatal, go on... */
867 		else
868 			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
869 	}
870 
871 	/*
872 	 * Join interface-local all-nodes address.
873 	 * (ff01::1%ifN, and ff01::%ifN/32)
874 	 */
875 	mltaddr = in6addr_nodelocal_allnodes;
876 	if ((error = in6_setscope(&mltaddr, ifp, NULL)) != 0)
877 		goto cleanup; /* XXX: should not fail */
878 
879 	imm = in6_joingroup(ifp, &mltaddr, &error, 0);
880 	if (imm == NULL) {
881 		nd6log((LOG_WARNING, "%s: in6_joingroup failed for %s on %s "
882 		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
883 		    &mltaddr), if_name(ifp), error));
884 		goto cleanup;
885 	}
886 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
887 
888 cleanup:
889 	return (error);
890 }
891 
892 /*
893  * Update parameters of an IPv6 interface address.
894  * If necessary, a new entry is created and linked into address chains.
895  * This function is separated from in6_control().
896  */
897 int
898 in6_update_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
899     struct in6_ifaddr *ia, int flags)
900 {
901 	int error, hostIsNew = 0;
902 
903 	if ((error = in6_validate_ifra(ifp, ifra, ia, flags)) != 0)
904 		return (error);
905 
906 	if (ia == NULL) {
907 		hostIsNew = 1;
908 		if ((ia = in6_alloc_ifa(ifp, ifra, flags)) == NULL)
909 			return (ENOBUFS);
910 	}
911 
912 	error = in6_update_ifa_internal(ifp, ifra, ia, hostIsNew, flags);
913 	if (error != 0) {
914 		if (hostIsNew != 0) {
915 			in6_unlink_ifa(ia, ifp);
916 			ifa_free(&ia->ia_ifa);
917 		}
918 		return (error);
919 	}
920 
921 	if (hostIsNew)
922 		error = in6_broadcast_ifa(ifp, ifra, ia, flags);
923 
924 	return (error);
925 }
926 
927 /*
928  * Fill in basic IPv6 address request info.
929  */
930 void
931 in6_prepare_ifra(struct in6_aliasreq *ifra, const struct in6_addr *addr,
932     const struct in6_addr *mask)
933 {
934 
935 	memset(ifra, 0, sizeof(struct in6_aliasreq));
936 
937 	ifra->ifra_addr.sin6_family = AF_INET6;
938 	ifra->ifra_addr.sin6_len = sizeof(struct sockaddr_in6);
939 	if (addr != NULL)
940 		ifra->ifra_addr.sin6_addr = *addr;
941 
942 	ifra->ifra_prefixmask.sin6_family = AF_INET6;
943 	ifra->ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
944 	if (mask != NULL)
945 		ifra->ifra_prefixmask.sin6_addr = *mask;
946 }
947 
948 static int
949 in6_validate_ifra(struct ifnet *ifp, struct in6_aliasreq *ifra,
950     struct in6_ifaddr *ia, int flags)
951 {
952 	int plen = -1;
953 	struct sockaddr_in6 dst6;
954 	struct in6_addrlifetime *lt;
955 	char ip6buf[INET6_ADDRSTRLEN];
956 
957 	/* Validate parameters */
958 	if (ifp == NULL || ifra == NULL) /* this maybe redundant */
959 		return (EINVAL);
960 
961 	/*
962 	 * The destination address for a p2p link must have a family
963 	 * of AF_UNSPEC or AF_INET6.
964 	 */
965 	if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
966 	    ifra->ifra_dstaddr.sin6_family != AF_INET6 &&
967 	    ifra->ifra_dstaddr.sin6_family != AF_UNSPEC)
968 		return (EAFNOSUPPORT);
969 
970 	/*
971 	 * Validate address
972 	 */
973 	if (ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6) ||
974 	    ifra->ifra_addr.sin6_family != AF_INET6)
975 		return (EINVAL);
976 
977 	/*
978 	 * validate ifra_prefixmask.  don't check sin6_family, netmask
979 	 * does not carry fields other than sin6_len.
980 	 */
981 	if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6))
982 		return (EINVAL);
983 	/*
984 	 * Because the IPv6 address architecture is classless, we require
985 	 * users to specify a (non 0) prefix length (mask) for a new address.
986 	 * We also require the prefix (when specified) mask is valid, and thus
987 	 * reject a non-consecutive mask.
988 	 */
989 	if (ia == NULL && ifra->ifra_prefixmask.sin6_len == 0)
990 		return (EINVAL);
991 	if (ifra->ifra_prefixmask.sin6_len != 0) {
992 		plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
993 		    (u_char *)&ifra->ifra_prefixmask +
994 		    ifra->ifra_prefixmask.sin6_len);
995 		if (plen <= 0)
996 			return (EINVAL);
997 	} else {
998 		/*
999 		 * In this case, ia must not be NULL.  We just use its prefix
1000 		 * length.
1001 		 */
1002 		plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
1003 	}
1004 	/*
1005 	 * If the destination address on a p2p interface is specified,
1006 	 * and the address is a scoped one, validate/set the scope
1007 	 * zone identifier.
1008 	 */
1009 	dst6 = ifra->ifra_dstaddr;
1010 	if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) != 0 &&
1011 	    (dst6.sin6_family == AF_INET6)) {
1012 		struct in6_addr in6_tmp;
1013 		u_int32_t zoneid;
1014 
1015 		in6_tmp = dst6.sin6_addr;
1016 		if (in6_setscope(&in6_tmp, ifp, &zoneid))
1017 			return (EINVAL); /* XXX: should be impossible */
1018 
1019 		if (dst6.sin6_scope_id != 0) {
1020 			if (dst6.sin6_scope_id != zoneid)
1021 				return (EINVAL);
1022 		} else		/* user omit to specify the ID. */
1023 			dst6.sin6_scope_id = zoneid;
1024 
1025 		/* convert into the internal form */
1026 		if (sa6_embedscope(&dst6, 0))
1027 			return (EINVAL); /* XXX: should be impossible */
1028 	}
1029 	/* Modify original ifra_dstaddr to reflect changes */
1030 	ifra->ifra_dstaddr = dst6;
1031 
1032 	/*
1033 	 * The destination address can be specified only for a p2p or a
1034 	 * loopback interface.  If specified, the corresponding prefix length
1035 	 * must be 128.
1036 	 */
1037 	if (ifra->ifra_dstaddr.sin6_family == AF_INET6) {
1038 		if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) == 0) {
1039 			/* XXX: noisy message */
1040 			nd6log((LOG_INFO, "in6_update_ifa: a destination can "
1041 			    "be specified for a p2p or a loopback IF only\n"));
1042 			return (EINVAL);
1043 		}
1044 		if (plen != 128) {
1045 			nd6log((LOG_INFO, "in6_update_ifa: prefixlen should "
1046 			    "be 128 when dstaddr is specified\n"));
1047 			return (EINVAL);
1048 		}
1049 	}
1050 	/* lifetime consistency check */
1051 	lt = &ifra->ifra_lifetime;
1052 	if (lt->ia6t_pltime > lt->ia6t_vltime)
1053 		return (EINVAL);
1054 	if (lt->ia6t_vltime == 0) {
1055 		/*
1056 		 * the following log might be noisy, but this is a typical
1057 		 * configuration mistake or a tool's bug.
1058 		 */
1059 		nd6log((LOG_INFO,
1060 		    "in6_update_ifa: valid lifetime is 0 for %s\n",
1061 		    ip6_sprintf(ip6buf, &ifra->ifra_addr.sin6_addr)));
1062 
1063 		if (ia == NULL)
1064 			return (0); /* there's nothing to do */
1065 	}
1066 
1067 	/* Check prefix mask */
1068 	if (ia != NULL && ifra->ifra_prefixmask.sin6_len != 0) {
1069 		/*
1070 		 * We prohibit changing the prefix length of an existing
1071 		 * address, because
1072 		 * + such an operation should be rare in IPv6, and
1073 		 * + the operation would confuse prefix management.
1074 		 */
1075 		if (ia->ia_prefixmask.sin6_len != 0 &&
1076 		    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) != plen) {
1077 			nd6log((LOG_INFO, "in6_validate_ifa: the prefix length "
1078 			    "of an existing %s address should not be changed\n",
1079 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1080 
1081 			return (EINVAL);
1082 		}
1083 	}
1084 
1085 	return (0);
1086 }
1087 
1088 
1089 /*
1090  * Allocate a new ifaddr and link it into chains.
1091  */
1092 static struct in6_ifaddr *
1093 in6_alloc_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra, int flags)
1094 {
1095 	struct in6_ifaddr *ia;
1096 
1097 	/*
1098 	 * When in6_alloc_ifa() is called in a process of a received
1099 	 * RA, it is called under an interrupt context.  So, we should
1100 	 * call malloc with M_NOWAIT.
1101 	 */
1102 	ia = (struct in6_ifaddr *)ifa_alloc(sizeof(*ia), M_NOWAIT);
1103 	if (ia == NULL)
1104 		return (NULL);
1105 	LIST_INIT(&ia->ia6_memberships);
1106 	/* Initialize the address and masks, and put time stamp */
1107 	ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
1108 	ia->ia_addr.sin6_family = AF_INET6;
1109 	ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
1110 	/* XXX: Can we assign ,sin6_addr and skip the rest? */
1111 	ia->ia_addr = ifra->ifra_addr;
1112 	ia->ia6_createtime = time_uptime;
1113 	if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
1114 		/*
1115 		 * Some functions expect that ifa_dstaddr is not
1116 		 * NULL for p2p interfaces.
1117 		 */
1118 		ia->ia_ifa.ifa_dstaddr =
1119 		    (struct sockaddr *)&ia->ia_dstaddr;
1120 	} else {
1121 		ia->ia_ifa.ifa_dstaddr = NULL;
1122 	}
1123 
1124 	/* set prefix mask if any */
1125 	ia->ia_ifa.ifa_netmask = (struct sockaddr *)&ia->ia_prefixmask;
1126 	if (ifra->ifra_prefixmask.sin6_len != 0) {
1127 		ia->ia_prefixmask.sin6_family = AF_INET6;
1128 		ia->ia_prefixmask.sin6_len = ifra->ifra_prefixmask.sin6_len;
1129 		ia->ia_prefixmask.sin6_addr = ifra->ifra_prefixmask.sin6_addr;
1130 	}
1131 
1132 	ia->ia_ifp = ifp;
1133 	ifa_ref(&ia->ia_ifa);			/* if_addrhead */
1134 	IF_ADDR_WLOCK(ifp);
1135 	TAILQ_INSERT_TAIL(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
1136 	IF_ADDR_WUNLOCK(ifp);
1137 
1138 	ifa_ref(&ia->ia_ifa);			/* in6_ifaddrhead */
1139 	IN6_IFADDR_WLOCK();
1140 	TAILQ_INSERT_TAIL(&V_in6_ifaddrhead, ia, ia_link);
1141 	LIST_INSERT_HEAD(IN6ADDR_HASH(&ia->ia_addr.sin6_addr), ia, ia6_hash);
1142 	IN6_IFADDR_WUNLOCK();
1143 
1144 	return (ia);
1145 }
1146 
1147 /*
1148  * Update/configure interface address parameters:
1149  *
1150  * 1) Update lifetime
1151  * 2) Update interface metric ad flags
1152  * 3) Notify other subsystems
1153  */
1154 static int
1155 in6_update_ifa_internal(struct ifnet *ifp, struct in6_aliasreq *ifra,
1156     struct in6_ifaddr *ia, int hostIsNew, int flags)
1157 {
1158 	int error;
1159 
1160 	/* update timestamp */
1161 	ia->ia6_updatetime = time_uptime;
1162 
1163 	/*
1164 	 * Set lifetimes.  We do not refer to ia6t_expire and ia6t_preferred
1165 	 * to see if the address is deprecated or invalidated, but initialize
1166 	 * these members for applications.
1167 	 */
1168 	ia->ia6_lifetime = ifra->ifra_lifetime;
1169 	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1170 		ia->ia6_lifetime.ia6t_expire =
1171 		    time_uptime + ia->ia6_lifetime.ia6t_vltime;
1172 	} else
1173 		ia->ia6_lifetime.ia6t_expire = 0;
1174 	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1175 		ia->ia6_lifetime.ia6t_preferred =
1176 		    time_uptime + ia->ia6_lifetime.ia6t_pltime;
1177 	} else
1178 		ia->ia6_lifetime.ia6t_preferred = 0;
1179 
1180 	/*
1181 	 * backward compatibility - if IN6_IFF_DEPRECATED is set from the
1182 	 * userland, make it deprecated.
1183 	 */
1184 	if ((ifra->ifra_flags & IN6_IFF_DEPRECATED) != 0) {
1185 		ia->ia6_lifetime.ia6t_pltime = 0;
1186 		ia->ia6_lifetime.ia6t_preferred = time_uptime;
1187 	}
1188 
1189 	/*
1190 	 * configure address flags.
1191 	 */
1192 	ia->ia6_flags = ifra->ifra_flags;
1193 
1194 	/*
1195 	 * Make the address tentative before joining multicast addresses,
1196 	 * so that corresponding MLD responses would not have a tentative
1197 	 * source address.
1198 	 */
1199 	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/* safety */
1200 
1201 	/*
1202 	 * DAD should be performed for an new address or addresses on
1203 	 * an interface with ND6_IFF_IFDISABLED.
1204 	 */
1205 	if (in6if_do_dad(ifp) &&
1206 	    (hostIsNew || (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)))
1207 		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1208 
1209 	/* notify other subsystems */
1210 	error = in6_notify_ifa(ifp, ia, ifra, hostIsNew);
1211 
1212 	return (error);
1213 }
1214 
1215 /*
1216  * Do link-level ifa job:
1217  * 1) Add lle entry for added address
1218  * 2) Notifies routing socket users about new address
1219  * 3) join appropriate multicast group
1220  * 4) start DAD if enabled
1221  */
1222 static int
1223 in6_broadcast_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
1224     struct in6_ifaddr *ia, int flags)
1225 {
1226 	struct in6_multi *in6m_sol;
1227 	int error = 0;
1228 
1229 	/* Add local address to lltable, if necessary (ex. on p2p link). */
1230 	if ((error = nd6_add_ifa_lle(ia)) != 0) {
1231 		in6_purgeaddr(&ia->ia_ifa);
1232 		ifa_free(&ia->ia_ifa);
1233 		return (error);
1234 	}
1235 
1236 	/* Join necessary multicast groups. */
1237 	in6m_sol = NULL;
1238 	if ((ifp->if_flags & IFF_MULTICAST) != 0) {
1239 		error = in6_update_ifa_join_mc(ifp, ifra, ia, flags, &in6m_sol);
1240 		if (error != 0) {
1241 			in6_purgeaddr(&ia->ia_ifa);
1242 			ifa_free(&ia->ia_ifa);
1243 			return (error);
1244 		}
1245 	}
1246 
1247 	/* Perform DAD, if the address is TENTATIVE. */
1248 	if ((ia->ia6_flags & IN6_IFF_TENTATIVE)) {
1249 		int delay, mindelay, maxdelay;
1250 
1251 		delay = 0;
1252 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1253 			/*
1254 			 * We need to impose a delay before sending an NS
1255 			 * for DAD.  Check if we also needed a delay for the
1256 			 * corresponding MLD message.  If we did, the delay
1257 			 * should be larger than the MLD delay (this could be
1258 			 * relaxed a bit, but this simple logic is at least
1259 			 * safe).
1260 			 * XXX: Break data hiding guidelines and look at
1261 			 * state for the solicited multicast group.
1262 			 */
1263 			mindelay = 0;
1264 			if (in6m_sol != NULL &&
1265 			    in6m_sol->in6m_state == MLD_REPORTING_MEMBER) {
1266 				mindelay = in6m_sol->in6m_timer;
1267 			}
1268 			maxdelay = MAX_RTR_SOLICITATION_DELAY * hz;
1269 			if (maxdelay - mindelay == 0)
1270 				delay = 0;
1271 			else {
1272 				delay =
1273 				    (arc4random() % (maxdelay - mindelay)) +
1274 				    mindelay;
1275 			}
1276 		}
1277 		nd6_dad_start((struct ifaddr *)ia, delay);
1278 	}
1279 
1280 	in6_newaddrmsg(ia, RTM_ADD);
1281 	ifa_free(&ia->ia_ifa);
1282 	return (error);
1283 }
1284 
1285 void
1286 in6_purgeaddr(struct ifaddr *ifa)
1287 {
1288 	struct ifnet *ifp = ifa->ifa_ifp;
1289 	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1290 	struct in6_multi_mship *imm;
1291 	int plen, error;
1292 
1293 	if (ifa->ifa_carp)
1294 		(*carp_detach_p)(ifa);
1295 
1296 	/*
1297 	 * Remove the loopback route to the interface address.
1298 	 * The check for the current setting of "nd6_useloopback"
1299 	 * is not needed.
1300 	 */
1301 	if (ia->ia_flags & IFA_RTSELF) {
1302 		error = ifa_del_loopback_route((struct ifaddr *)ia,
1303 		    (struct sockaddr *)&ia->ia_addr);
1304 		if (error == 0)
1305 			ia->ia_flags &= ~IFA_RTSELF;
1306 	}
1307 
1308 	/* stop DAD processing */
1309 	nd6_dad_stop(ifa);
1310 
1311 	/* Leave multicast groups. */
1312 	while ((imm = LIST_FIRST(&ia->ia6_memberships)) != NULL) {
1313 		LIST_REMOVE(imm, i6mm_chain);
1314 		in6_leavegroup(imm);
1315 	}
1316 	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1317 	if ((ia->ia_flags & IFA_ROUTE) && plen == 128) {
1318 		error = rtinit(&(ia->ia_ifa), RTM_DELETE, ia->ia_flags |
1319 		    (ia->ia_dstaddr.sin6_family == AF_INET6 ? RTF_HOST : 0));
1320 		if (error != 0)
1321 			log(LOG_INFO, "%s: err=%d, destination address delete "
1322 			    "failed\n", __func__, error);
1323 		ia->ia_flags &= ~IFA_ROUTE;
1324 	}
1325 
1326 	in6_newaddrmsg(ia, RTM_DELETE);
1327 	in6_unlink_ifa(ia, ifp);
1328 }
1329 
1330 static void
1331 in6_unlink_ifa(struct in6_ifaddr *ia, struct ifnet *ifp)
1332 {
1333 	char ip6buf[INET6_ADDRSTRLEN];
1334 	int remove_lle;
1335 
1336 	IF_ADDR_WLOCK(ifp);
1337 	TAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
1338 	IF_ADDR_WUNLOCK(ifp);
1339 	ifa_free(&ia->ia_ifa);			/* if_addrhead */
1340 
1341 	/*
1342 	 * Defer the release of what might be the last reference to the
1343 	 * in6_ifaddr so that it can't be freed before the remainder of the
1344 	 * cleanup.
1345 	 */
1346 	IN6_IFADDR_WLOCK();
1347 	TAILQ_REMOVE(&V_in6_ifaddrhead, ia, ia_link);
1348 	LIST_REMOVE(ia, ia6_hash);
1349 	IN6_IFADDR_WUNLOCK();
1350 
1351 	/*
1352 	 * Release the reference to the base prefix.  There should be a
1353 	 * positive reference.
1354 	 */
1355 	remove_lle = 0;
1356 	if (ia->ia6_ndpr == NULL) {
1357 		nd6log((LOG_NOTICE,
1358 		    "in6_unlink_ifa: autoconf'ed address "
1359 		    "%s has no prefix\n", ip6_sprintf(ip6buf, IA6_IN6(ia))));
1360 	} else {
1361 		ia->ia6_ndpr->ndpr_refcnt--;
1362 		/* Do not delete lles within prefix if refcont != 0 */
1363 		if (ia->ia6_ndpr->ndpr_refcnt == 0)
1364 			remove_lle = 1;
1365 		ia->ia6_ndpr = NULL;
1366 	}
1367 
1368 	nd6_rem_ifa_lle(ia, remove_lle);
1369 
1370 	/*
1371 	 * Also, if the address being removed is autoconf'ed, call
1372 	 * pfxlist_onlink_check() since the release might affect the status of
1373 	 * other (detached) addresses.
1374 	 */
1375 	if ((ia->ia6_flags & IN6_IFF_AUTOCONF)) {
1376 		pfxlist_onlink_check();
1377 	}
1378 	ifa_free(&ia->ia_ifa);			/* in6_ifaddrhead */
1379 }
1380 
1381 /*
1382  * Notifies other subsystems about address change/arrival:
1383  * 1) Notifies device handler on the first IPv6 address assignment
1384  * 2) Handle routing table changes for P2P links and route
1385  * 3) Handle routing table changes for address host route
1386  */
1387 static int
1388 in6_notify_ifa(struct ifnet *ifp, struct in6_ifaddr *ia,
1389     struct in6_aliasreq *ifra, int hostIsNew)
1390 {
1391 	int	error = 0, plen, ifacount = 0;
1392 	struct ifaddr *ifa;
1393 	struct sockaddr_in6 *pdst;
1394 	char ip6buf[INET6_ADDRSTRLEN];
1395 
1396 	/*
1397 	 * Give the interface a chance to initialize
1398 	 * if this is its first address,
1399 	 */
1400 	if (hostIsNew != 0) {
1401 		IF_ADDR_RLOCK(ifp);
1402 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1403 			if (ifa->ifa_addr->sa_family != AF_INET6)
1404 				continue;
1405 			ifacount++;
1406 		}
1407 		IF_ADDR_RUNLOCK(ifp);
1408 	}
1409 
1410 	if (ifacount <= 1 && ifp->if_ioctl) {
1411 		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
1412 		if (error)
1413 			return (error);
1414 	}
1415 
1416 	/*
1417 	 * If a new destination address is specified, scrub the old one and
1418 	 * install the new destination.  Note that the interface must be
1419 	 * p2p or loopback.
1420 	 */
1421 	pdst = &ifra->ifra_dstaddr;
1422 	if (pdst->sin6_family == AF_INET6 &&
1423 	    !IN6_ARE_ADDR_EQUAL(&pdst->sin6_addr, &ia->ia_dstaddr.sin6_addr)) {
1424 		if ((ia->ia_flags & IFA_ROUTE) != 0 &&
1425 		    (rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST) != 0)) {
1426 			nd6log((LOG_ERR, "in6_update_ifa_internal: failed to "
1427 			    "remove a route to the old destination: %s\n",
1428 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1429 			/* proceed anyway... */
1430 		} else
1431 			ia->ia_flags &= ~IFA_ROUTE;
1432 		ia->ia_dstaddr = *pdst;
1433 	}
1434 
1435 	/*
1436 	 * If a new destination address is specified for a point-to-point
1437 	 * interface, install a route to the destination as an interface
1438 	 * direct route.
1439 	 * XXX: the logic below rejects assigning multiple addresses on a p2p
1440 	 * interface that share the same destination.
1441 	 */
1442 	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1443 	if (!(ia->ia_flags & IFA_ROUTE) && plen == 128 &&
1444 	    ia->ia_dstaddr.sin6_family == AF_INET6) {
1445 		int rtflags = RTF_UP | RTF_HOST;
1446 		/*
1447 		 * Handle the case for ::1 .
1448 		 */
1449 		if (ifp->if_flags & IFF_LOOPBACK)
1450 			ia->ia_flags |= IFA_RTSELF;
1451 		error = rtinit(&ia->ia_ifa, RTM_ADD, ia->ia_flags | rtflags);
1452 		if (error)
1453 			return (error);
1454 		ia->ia_flags |= IFA_ROUTE;
1455 	}
1456 
1457 	/*
1458 	 * add a loopback route to self if not exists
1459 	 */
1460 	if (!(ia->ia_flags & IFA_RTSELF) && V_nd6_useloopback) {
1461 		error = ifa_add_loopback_route((struct ifaddr *)ia,
1462 		    (struct sockaddr *)&ia->ia_addr);
1463 		if (error == 0)
1464 			ia->ia_flags |= IFA_RTSELF;
1465 	}
1466 
1467 	return (error);
1468 }
1469 
1470 /*
1471  * Find an IPv6 interface link-local address specific to an interface.
1472  * ifaddr is returned referenced.
1473  */
1474 struct in6_ifaddr *
1475 in6ifa_ifpforlinklocal(struct ifnet *ifp, int ignoreflags)
1476 {
1477 	struct ifaddr *ifa;
1478 
1479 	IF_ADDR_RLOCK(ifp);
1480 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1481 		if (ifa->ifa_addr->sa_family != AF_INET6)
1482 			continue;
1483 		if (IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa))) {
1484 			if ((((struct in6_ifaddr *)ifa)->ia6_flags &
1485 			    ignoreflags) != 0)
1486 				continue;
1487 			ifa_ref(ifa);
1488 			break;
1489 		}
1490 	}
1491 	IF_ADDR_RUNLOCK(ifp);
1492 
1493 	return ((struct in6_ifaddr *)ifa);
1494 }
1495 
1496 
1497 /*
1498  * find the internet address corresponding to a given address.
1499  * ifaddr is returned referenced.
1500  */
1501 struct in6_ifaddr *
1502 in6ifa_ifwithaddr(const struct in6_addr *addr, uint32_t zoneid)
1503 {
1504 	struct rm_priotracker in6_ifa_tracker;
1505 	struct in6_ifaddr *ia;
1506 
1507 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1508 	LIST_FOREACH(ia, IN6ADDR_HASH(addr), ia6_hash) {
1509 		if (IN6_ARE_ADDR_EQUAL(IA6_IN6(ia), addr)) {
1510 			if (zoneid != 0 &&
1511 			    zoneid != ia->ia_addr.sin6_scope_id)
1512 				continue;
1513 			ifa_ref(&ia->ia_ifa);
1514 			break;
1515 		}
1516 	}
1517 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1518 	return (ia);
1519 }
1520 
1521 /*
1522  * find the internet address corresponding to a given interface and address.
1523  * ifaddr is returned referenced.
1524  */
1525 struct in6_ifaddr *
1526 in6ifa_ifpwithaddr(struct ifnet *ifp, const struct in6_addr *addr)
1527 {
1528 	struct ifaddr *ifa;
1529 
1530 	IF_ADDR_RLOCK(ifp);
1531 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1532 		if (ifa->ifa_addr->sa_family != AF_INET6)
1533 			continue;
1534 		if (IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa))) {
1535 			ifa_ref(ifa);
1536 			break;
1537 		}
1538 	}
1539 	IF_ADDR_RUNLOCK(ifp);
1540 
1541 	return ((struct in6_ifaddr *)ifa);
1542 }
1543 
1544 /*
1545  * Find a link-local scoped address on ifp and return it if any.
1546  */
1547 struct in6_ifaddr *
1548 in6ifa_llaonifp(struct ifnet *ifp)
1549 {
1550 	struct sockaddr_in6 *sin6;
1551 	struct ifaddr *ifa;
1552 
1553 	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)
1554 		return (NULL);
1555 	IF_ADDR_RLOCK(ifp);
1556 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1557 		if (ifa->ifa_addr->sa_family != AF_INET6)
1558 			continue;
1559 		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
1560 		if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr) ||
1561 		    IN6_IS_ADDR_MC_INTFACELOCAL(&sin6->sin6_addr) ||
1562 		    IN6_IS_ADDR_MC_NODELOCAL(&sin6->sin6_addr))
1563 			break;
1564 	}
1565 	IF_ADDR_RUNLOCK(ifp);
1566 
1567 	return ((struct in6_ifaddr *)ifa);
1568 }
1569 
1570 /*
1571  * Convert IP6 address to printable (loggable) representation. Caller
1572  * has to make sure that ip6buf is at least INET6_ADDRSTRLEN long.
1573  */
1574 static char digits[] = "0123456789abcdef";
1575 char *
1576 ip6_sprintf(char *ip6buf, const struct in6_addr *addr)
1577 {
1578 	int i, cnt = 0, maxcnt = 0, idx = 0, index = 0;
1579 	char *cp;
1580 	const u_int16_t *a = (const u_int16_t *)addr;
1581 	const u_int8_t *d;
1582 	int dcolon = 0, zero = 0;
1583 
1584 	cp = ip6buf;
1585 
1586 	for (i = 0; i < 8; i++) {
1587 		if (*(a + i) == 0) {
1588 			cnt++;
1589 			if (cnt == 1)
1590 				idx = i;
1591 		}
1592 		else if (maxcnt < cnt) {
1593 			maxcnt = cnt;
1594 			index = idx;
1595 			cnt = 0;
1596 		}
1597 	}
1598 	if (maxcnt < cnt) {
1599 		maxcnt = cnt;
1600 		index = idx;
1601 	}
1602 
1603 	for (i = 0; i < 8; i++) {
1604 		if (dcolon == 1) {
1605 			if (*a == 0) {
1606 				if (i == 7)
1607 					*cp++ = ':';
1608 				a++;
1609 				continue;
1610 			} else
1611 				dcolon = 2;
1612 		}
1613 		if (*a == 0) {
1614 			if (dcolon == 0 && *(a + 1) == 0 && i == index) {
1615 				if (i == 0)
1616 					*cp++ = ':';
1617 				*cp++ = ':';
1618 				dcolon = 1;
1619 			} else {
1620 				*cp++ = '0';
1621 				*cp++ = ':';
1622 			}
1623 			a++;
1624 			continue;
1625 		}
1626 		d = (const u_char *)a;
1627 		/* Try to eliminate leading zeros in printout like in :0001. */
1628 		zero = 1;
1629 		*cp = digits[*d >> 4];
1630 		if (*cp != '0') {
1631 			zero = 0;
1632 			cp++;
1633 		}
1634 		*cp = digits[*d++ & 0xf];
1635 		if (zero == 0 || (*cp != '0')) {
1636 			zero = 0;
1637 			cp++;
1638 		}
1639 		*cp = digits[*d >> 4];
1640 		if (zero == 0 || (*cp != '0')) {
1641 			zero = 0;
1642 			cp++;
1643 		}
1644 		*cp++ = digits[*d & 0xf];
1645 		*cp++ = ':';
1646 		a++;
1647 	}
1648 	*--cp = '\0';
1649 	return (ip6buf);
1650 }
1651 
1652 int
1653 in6_localaddr(struct in6_addr *in6)
1654 {
1655 	struct rm_priotracker in6_ifa_tracker;
1656 	struct in6_ifaddr *ia;
1657 
1658 	if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
1659 		return 1;
1660 
1661 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1662 	TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
1663 		if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
1664 		    &ia->ia_prefixmask.sin6_addr)) {
1665 			IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1666 			return 1;
1667 		}
1668 	}
1669 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1670 
1671 	return (0);
1672 }
1673 
1674 /*
1675  * Return 1 if an internet address is for the local host and configured
1676  * on one of its interfaces.
1677  */
1678 int
1679 in6_localip(struct in6_addr *in6)
1680 {
1681 	struct rm_priotracker in6_ifa_tracker;
1682 	struct in6_ifaddr *ia;
1683 
1684 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1685 	LIST_FOREACH(ia, IN6ADDR_HASH(in6), ia6_hash) {
1686 		if (IN6_ARE_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr)) {
1687 			IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1688 			return (1);
1689 		}
1690 	}
1691 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1692 	return (0);
1693 }
1694 
1695 /*
1696  * Return 1 if an internet address is configured on an interface.
1697  */
1698 int
1699 in6_ifhasaddr(struct ifnet *ifp, struct in6_addr *addr)
1700 {
1701 	struct in6_addr in6;
1702 	struct ifaddr *ifa;
1703 	struct in6_ifaddr *ia6;
1704 
1705 	in6 = *addr;
1706 	if (in6_clearscope(&in6))
1707 		return (0);
1708 	in6_setscope(&in6, ifp, NULL);
1709 
1710 	IF_ADDR_RLOCK(ifp);
1711 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1712 		if (ifa->ifa_addr->sa_family != AF_INET6)
1713 			continue;
1714 		ia6 = (struct in6_ifaddr *)ifa;
1715 		if (IN6_ARE_ADDR_EQUAL(&ia6->ia_addr.sin6_addr, &in6)) {
1716 			IF_ADDR_RUNLOCK(ifp);
1717 			return (1);
1718 		}
1719 	}
1720 	IF_ADDR_RUNLOCK(ifp);
1721 
1722 	return (0);
1723 }
1724 
1725 int
1726 in6_is_addr_deprecated(struct sockaddr_in6 *sa6)
1727 {
1728 	struct rm_priotracker in6_ifa_tracker;
1729 	struct in6_ifaddr *ia;
1730 
1731 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1732 	LIST_FOREACH(ia, IN6ADDR_HASH(&sa6->sin6_addr), ia6_hash) {
1733 		if (IN6_ARE_ADDR_EQUAL(IA6_IN6(ia), &sa6->sin6_addr)) {
1734 			if (ia->ia6_flags & IN6_IFF_DEPRECATED) {
1735 				IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1736 				return (1); /* true */
1737 			}
1738 			break;
1739 		}
1740 	}
1741 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1742 
1743 	return (0);		/* false */
1744 }
1745 
1746 /*
1747  * return length of part which dst and src are equal
1748  * hard coding...
1749  */
1750 int
1751 in6_matchlen(struct in6_addr *src, struct in6_addr *dst)
1752 {
1753 	int match = 0;
1754 	u_char *s = (u_char *)src, *d = (u_char *)dst;
1755 	u_char *lim = s + 16, r;
1756 
1757 	while (s < lim)
1758 		if ((r = (*d++ ^ *s++)) != 0) {
1759 			while (r < 128) {
1760 				match++;
1761 				r <<= 1;
1762 			}
1763 			break;
1764 		} else
1765 			match += 8;
1766 	return match;
1767 }
1768 
1769 /* XXX: to be scope conscious */
1770 int
1771 in6_are_prefix_equal(struct in6_addr *p1, struct in6_addr *p2, int len)
1772 {
1773 	int bytelen, bitlen;
1774 
1775 	/* sanity check */
1776 	if (0 > len || len > 128) {
1777 		log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n",
1778 		    len);
1779 		return (0);
1780 	}
1781 
1782 	bytelen = len / 8;
1783 	bitlen = len % 8;
1784 
1785 	if (bcmp(&p1->s6_addr, &p2->s6_addr, bytelen))
1786 		return (0);
1787 	if (bitlen != 0 &&
1788 	    p1->s6_addr[bytelen] >> (8 - bitlen) !=
1789 	    p2->s6_addr[bytelen] >> (8 - bitlen))
1790 		return (0);
1791 
1792 	return (1);
1793 }
1794 
1795 void
1796 in6_prefixlen2mask(struct in6_addr *maskp, int len)
1797 {
1798 	u_char maskarray[8] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
1799 	int bytelen, bitlen, i;
1800 
1801 	/* sanity check */
1802 	if (0 > len || len > 128) {
1803 		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
1804 		    len);
1805 		return;
1806 	}
1807 
1808 	bzero(maskp, sizeof(*maskp));
1809 	bytelen = len / 8;
1810 	bitlen = len % 8;
1811 	for (i = 0; i < bytelen; i++)
1812 		maskp->s6_addr[i] = 0xff;
1813 	if (bitlen)
1814 		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
1815 }
1816 
1817 /*
1818  * return the best address out of the same scope. if no address was
1819  * found, return the first valid address from designated IF.
1820  */
1821 struct in6_ifaddr *
1822 in6_ifawithifp(struct ifnet *ifp, struct in6_addr *dst)
1823 {
1824 	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
1825 	struct ifaddr *ifa;
1826 	struct in6_ifaddr *besta = 0;
1827 	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
1828 
1829 	dep[0] = dep[1] = NULL;
1830 
1831 	/*
1832 	 * We first look for addresses in the same scope.
1833 	 * If there is one, return it.
1834 	 * If two or more, return one which matches the dst longest.
1835 	 * If none, return one of global addresses assigned other ifs.
1836 	 */
1837 	IF_ADDR_RLOCK(ifp);
1838 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1839 		if (ifa->ifa_addr->sa_family != AF_INET6)
1840 			continue;
1841 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
1842 			continue; /* XXX: is there any case to allow anycast? */
1843 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
1844 			continue; /* don't use this interface */
1845 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
1846 			continue;
1847 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
1848 			if (V_ip6_use_deprecated)
1849 				dep[0] = (struct in6_ifaddr *)ifa;
1850 			continue;
1851 		}
1852 
1853 		if (dst_scope == in6_addrscope(IFA_IN6(ifa))) {
1854 			/*
1855 			 * call in6_matchlen() as few as possible
1856 			 */
1857 			if (besta) {
1858 				if (blen == -1)
1859 					blen = in6_matchlen(&besta->ia_addr.sin6_addr, dst);
1860 				tlen = in6_matchlen(IFA_IN6(ifa), dst);
1861 				if (tlen > blen) {
1862 					blen = tlen;
1863 					besta = (struct in6_ifaddr *)ifa;
1864 				}
1865 			} else
1866 				besta = (struct in6_ifaddr *)ifa;
1867 		}
1868 	}
1869 	if (besta) {
1870 		ifa_ref(&besta->ia_ifa);
1871 		IF_ADDR_RUNLOCK(ifp);
1872 		return (besta);
1873 	}
1874 
1875 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1876 		if (ifa->ifa_addr->sa_family != AF_INET6)
1877 			continue;
1878 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
1879 			continue; /* XXX: is there any case to allow anycast? */
1880 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
1881 			continue; /* don't use this interface */
1882 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
1883 			continue;
1884 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
1885 			if (V_ip6_use_deprecated)
1886 				dep[1] = (struct in6_ifaddr *)ifa;
1887 			continue;
1888 		}
1889 
1890 		if (ifa != NULL)
1891 			ifa_ref(ifa);
1892 		IF_ADDR_RUNLOCK(ifp);
1893 		return (struct in6_ifaddr *)ifa;
1894 	}
1895 
1896 	/* use the last-resort values, that are, deprecated addresses */
1897 	if (dep[0]) {
1898 		ifa_ref((struct ifaddr *)dep[0]);
1899 		IF_ADDR_RUNLOCK(ifp);
1900 		return dep[0];
1901 	}
1902 	if (dep[1]) {
1903 		ifa_ref((struct ifaddr *)dep[1]);
1904 		IF_ADDR_RUNLOCK(ifp);
1905 		return dep[1];
1906 	}
1907 
1908 	IF_ADDR_RUNLOCK(ifp);
1909 	return NULL;
1910 }
1911 
1912 /*
1913  * perform DAD when interface becomes IFF_UP.
1914  */
1915 void
1916 in6_if_up(struct ifnet *ifp)
1917 {
1918 	struct ifaddr *ifa;
1919 	struct in6_ifaddr *ia;
1920 
1921 	IF_ADDR_RLOCK(ifp);
1922 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1923 		if (ifa->ifa_addr->sa_family != AF_INET6)
1924 			continue;
1925 		ia = (struct in6_ifaddr *)ifa;
1926 		if (ia->ia6_flags & IN6_IFF_TENTATIVE) {
1927 			/*
1928 			 * The TENTATIVE flag was likely set by hand
1929 			 * beforehand, implicitly indicating the need for DAD.
1930 			 * We may be able to skip the random delay in this
1931 			 * case, but we impose delays just in case.
1932 			 */
1933 			nd6_dad_start(ifa,
1934 			    arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz));
1935 		}
1936 	}
1937 	IF_ADDR_RUNLOCK(ifp);
1938 
1939 	/*
1940 	 * special cases, like 6to4, are handled in in6_ifattach
1941 	 */
1942 	in6_ifattach(ifp, NULL);
1943 }
1944 
1945 int
1946 in6if_do_dad(struct ifnet *ifp)
1947 {
1948 	if ((ifp->if_flags & IFF_LOOPBACK) != 0)
1949 		return (0);
1950 
1951 	if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) ||
1952 	    (ND_IFINFO(ifp)->flags & ND6_IFF_NO_DAD))
1953 		return (0);
1954 
1955 	/*
1956 	 * Our DAD routine requires the interface up and running.
1957 	 * However, some interfaces can be up before the RUNNING
1958 	 * status.  Additionaly, users may try to assign addresses
1959 	 * before the interface becomes up (or running).
1960 	 * This function returns EAGAIN in that case.
1961 	 * The caller should mark "tentative" on the address instead of
1962 	 * performing DAD immediately.
1963 	 */
1964 	if (!((ifp->if_flags & IFF_UP) &&
1965 	    (ifp->if_drv_flags & IFF_DRV_RUNNING)))
1966 		return (EAGAIN);
1967 
1968 	return (1);
1969 }
1970 
1971 /*
1972  * Calculate max IPv6 MTU through all the interfaces and store it
1973  * to in6_maxmtu.
1974  */
1975 void
1976 in6_setmaxmtu(void)
1977 {
1978 	unsigned long maxmtu = 0;
1979 	struct ifnet *ifp;
1980 
1981 	IFNET_RLOCK_NOSLEEP();
1982 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1983 		/* this function can be called during ifnet initialization */
1984 		if (!ifp->if_afdata[AF_INET6])
1985 			continue;
1986 		if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
1987 		    IN6_LINKMTU(ifp) > maxmtu)
1988 			maxmtu = IN6_LINKMTU(ifp);
1989 	}
1990 	IFNET_RUNLOCK_NOSLEEP();
1991 	if (maxmtu)	/* update only when maxmtu is positive */
1992 		V_in6_maxmtu = maxmtu;
1993 }
1994 
1995 /*
1996  * Provide the length of interface identifiers to be used for the link attached
1997  * to the given interface.  The length should be defined in "IPv6 over
1998  * xxx-link" document.  Note that address architecture might also define
1999  * the length for a particular set of address prefixes, regardless of the
2000  * link type.  As clarified in rfc2462bis, those two definitions should be
2001  * consistent, and those really are as of August 2004.
2002  */
2003 int
2004 in6_if2idlen(struct ifnet *ifp)
2005 {
2006 	switch (ifp->if_type) {
2007 	case IFT_ETHER:		/* RFC2464 */
2008 	case IFT_PROPVIRTUAL:	/* XXX: no RFC. treat it as ether */
2009 	case IFT_L2VLAN:	/* ditto */
2010 	case IFT_IEEE80211:	/* ditto */
2011 	case IFT_BRIDGE:	/* bridge(4) only does Ethernet-like links */
2012 	case IFT_INFINIBAND:
2013 		return (64);
2014 	case IFT_FDDI:		/* RFC2467 */
2015 		return (64);
2016 	case IFT_ISO88025:	/* RFC2470 (IPv6 over Token Ring) */
2017 		return (64);
2018 	case IFT_PPP:		/* RFC2472 */
2019 		return (64);
2020 	case IFT_ARCNET:	/* RFC2497 */
2021 		return (64);
2022 	case IFT_FRELAY:	/* RFC2590 */
2023 		return (64);
2024 	case IFT_IEEE1394:	/* RFC3146 */
2025 		return (64);
2026 	case IFT_GIF:
2027 		return (64);	/* draft-ietf-v6ops-mech-v2-07 */
2028 	case IFT_LOOP:
2029 		return (64);	/* XXX: is this really correct? */
2030 	default:
2031 		/*
2032 		 * Unknown link type:
2033 		 * It might be controversial to use the today's common constant
2034 		 * of 64 for these cases unconditionally.  For full compliance,
2035 		 * we should return an error in this case.  On the other hand,
2036 		 * if we simply miss the standard for the link type or a new
2037 		 * standard is defined for a new link type, the IFID length
2038 		 * is very likely to be the common constant.  As a compromise,
2039 		 * we always use the constant, but make an explicit notice
2040 		 * indicating the "unknown" case.
2041 		 */
2042 		printf("in6_if2idlen: unknown link type (%d)\n", ifp->if_type);
2043 		return (64);
2044 	}
2045 }
2046 
2047 #include <sys/sysctl.h>
2048 
2049 struct in6_llentry {
2050 	struct llentry		base;
2051 };
2052 
2053 #define	IN6_LLTBL_DEFAULT_HSIZE	32
2054 #define	IN6_LLTBL_HASH(k, h) \
2055 	(((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1))
2056 
2057 /*
2058  * Do actual deallocation of @lle.
2059  * Called by LLE_FREE_LOCKED when number of references
2060  * drops to zero.
2061  */
2062 static void
2063 in6_lltable_destroy_lle(struct llentry *lle)
2064 {
2065 
2066 	LLE_WUNLOCK(lle);
2067 	LLE_LOCK_DESTROY(lle);
2068 	LLE_REQ_DESTROY(lle);
2069 	free(lle, M_LLTABLE);
2070 }
2071 
2072 static struct llentry *
2073 in6_lltable_new(const struct in6_addr *addr6, u_int flags)
2074 {
2075 	struct in6_llentry *lle;
2076 
2077 	lle = malloc(sizeof(struct in6_llentry), M_LLTABLE, M_NOWAIT | M_ZERO);
2078 	if (lle == NULL)		/* NB: caller generates msg */
2079 		return NULL;
2080 
2081 	lle->base.r_l3addr.addr6 = *addr6;
2082 	lle->base.lle_refcnt = 1;
2083 	lle->base.lle_free = in6_lltable_destroy_lle;
2084 	LLE_LOCK_INIT(&lle->base);
2085 	LLE_REQ_INIT(&lle->base);
2086 	callout_init(&lle->base.lle_timer, 1);
2087 
2088 	return (&lle->base);
2089 }
2090 
2091 static int
2092 in6_lltable_match_prefix(const struct sockaddr *saddr,
2093     const struct sockaddr *smask, u_int flags, struct llentry *lle)
2094 {
2095 	const struct in6_addr *addr, *mask, *lle_addr;
2096 
2097 	addr = &((const struct sockaddr_in6 *)saddr)->sin6_addr;
2098 	mask = &((const struct sockaddr_in6 *)smask)->sin6_addr;
2099 	lle_addr = &lle->r_l3addr.addr6;
2100 
2101 	if (IN6_ARE_MASKED_ADDR_EQUAL(lle_addr, addr, mask) == 0)
2102 		return (0);
2103 
2104 	if (lle->la_flags & LLE_IFADDR) {
2105 
2106 		/*
2107 		 * Delete LLE_IFADDR records IFF address & flag matches.
2108 		 * Note that addr is the interface address within prefix
2109 		 * being matched.
2110 		 */
2111 		if (IN6_ARE_ADDR_EQUAL(addr, lle_addr) &&
2112 		    (flags & LLE_STATIC) != 0)
2113 			return (1);
2114 		return (0);
2115 	}
2116 
2117 	/* flags & LLE_STATIC means deleting both dynamic and static entries */
2118 	if ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC))
2119 		return (1);
2120 
2121 	return (0);
2122 }
2123 
2124 static void
2125 in6_lltable_free_entry(struct lltable *llt, struct llentry *lle)
2126 {
2127 	struct ifnet *ifp;
2128 
2129 	LLE_WLOCK_ASSERT(lle);
2130 	KASSERT(llt != NULL, ("lltable is NULL"));
2131 
2132 	/* Unlink entry from table */
2133 	if ((lle->la_flags & LLE_LINKED) != 0) {
2134 
2135 		ifp = llt->llt_ifp;
2136 		IF_AFDATA_WLOCK_ASSERT(ifp);
2137 		lltable_unlink_entry(llt, lle);
2138 	}
2139 
2140 	if (callout_stop(&lle->lle_timer) > 0)
2141 		LLE_REMREF(lle);
2142 
2143 	llentry_free(lle);
2144 }
2145 
2146 static int
2147 in6_lltable_rtcheck(struct ifnet *ifp,
2148 		    u_int flags,
2149 		    const struct sockaddr *l3addr)
2150 {
2151 	const struct sockaddr_in6 *sin6;
2152 	struct nhop6_basic nh6;
2153 	struct in6_addr dst;
2154 	uint32_t scopeid;
2155 	int error;
2156 	char ip6buf[INET6_ADDRSTRLEN];
2157 
2158 	KASSERT(l3addr->sa_family == AF_INET6,
2159 	    ("sin_family %d", l3addr->sa_family));
2160 
2161 	/* Our local addresses are always only installed on the default FIB. */
2162 
2163 	sin6 = (const struct sockaddr_in6 *)l3addr;
2164 	in6_splitscope(&sin6->sin6_addr, &dst, &scopeid);
2165 	error = fib6_lookup_nh_basic(RT_DEFAULT_FIB, &dst, scopeid, 0, 0, &nh6);
2166 	if (error != 0 || (nh6.nh_flags & NHF_GATEWAY) || nh6.nh_ifp != ifp) {
2167 		struct ifaddr *ifa;
2168 		/*
2169 		 * Create an ND6 cache for an IPv6 neighbor
2170 		 * that is not covered by our own prefix.
2171 		 */
2172 		ifa = ifaof_ifpforaddr(l3addr, ifp);
2173 		if (ifa != NULL) {
2174 			ifa_free(ifa);
2175 			return 0;
2176 		}
2177 		log(LOG_INFO, "IPv6 address: \"%s\" is not on the network\n",
2178 		    ip6_sprintf(ip6buf, &sin6->sin6_addr));
2179 		return EINVAL;
2180 	}
2181 	return 0;
2182 }
2183 
2184 static inline uint32_t
2185 in6_lltable_hash_dst(const struct in6_addr *dst, uint32_t hsize)
2186 {
2187 
2188 	return (IN6_LLTBL_HASH(dst->s6_addr32[3], hsize));
2189 }
2190 
2191 static uint32_t
2192 in6_lltable_hash(const struct llentry *lle, uint32_t hsize)
2193 {
2194 
2195 	return (in6_lltable_hash_dst(&lle->r_l3addr.addr6, hsize));
2196 }
2197 
2198 static void
2199 in6_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
2200 {
2201 	struct sockaddr_in6 *sin6;
2202 
2203 	sin6 = (struct sockaddr_in6 *)sa;
2204 	bzero(sin6, sizeof(*sin6));
2205 	sin6->sin6_family = AF_INET6;
2206 	sin6->sin6_len = sizeof(*sin6);
2207 	sin6->sin6_addr = lle->r_l3addr.addr6;
2208 }
2209 
2210 static inline struct llentry *
2211 in6_lltable_find_dst(struct lltable *llt, const struct in6_addr *dst)
2212 {
2213 	struct llentry *lle;
2214 	struct llentries *lleh;
2215 	u_int hashidx;
2216 
2217 	hashidx = in6_lltable_hash_dst(dst, llt->llt_hsize);
2218 	lleh = &llt->lle_head[hashidx];
2219 	LIST_FOREACH(lle, lleh, lle_next) {
2220 		if (lle->la_flags & LLE_DELETED)
2221 			continue;
2222 		if (IN6_ARE_ADDR_EQUAL(&lle->r_l3addr.addr6, dst))
2223 			break;
2224 	}
2225 
2226 	return (lle);
2227 }
2228 
2229 static void
2230 in6_lltable_delete_entry(struct lltable *llt, struct llentry *lle)
2231 {
2232 
2233 	lle->la_flags |= LLE_DELETED;
2234 	EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_DELETED);
2235 #ifdef DIAGNOSTIC
2236 	log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle);
2237 #endif
2238 	llentry_free(lle);
2239 }
2240 
2241 static struct llentry *
2242 in6_lltable_alloc(struct lltable *llt, u_int flags,
2243 	const struct sockaddr *l3addr)
2244 {
2245 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2246 	struct ifnet *ifp = llt->llt_ifp;
2247 	struct llentry *lle;
2248 	char linkhdr[LLE_MAX_LINKHDR];
2249 	size_t linkhdrsize;
2250 	int lladdr_off;
2251 
2252 	KASSERT(l3addr->sa_family == AF_INET6,
2253 	    ("sin_family %d", l3addr->sa_family));
2254 
2255 	/*
2256 	 * A route that covers the given address must have
2257 	 * been installed 1st because we are doing a resolution,
2258 	 * verify this.
2259 	 */
2260 	if (!(flags & LLE_IFADDR) &&
2261 	    in6_lltable_rtcheck(ifp, flags, l3addr) != 0)
2262 		return (NULL);
2263 
2264 	lle = in6_lltable_new(&sin6->sin6_addr, flags);
2265 	if (lle == NULL) {
2266 		log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
2267 		return (NULL);
2268 	}
2269 	lle->la_flags = flags;
2270 	if ((flags & LLE_IFADDR) == LLE_IFADDR) {
2271 		linkhdrsize = LLE_MAX_LINKHDR;
2272 		if (lltable_calc_llheader(ifp, AF_INET6, IF_LLADDR(ifp),
2273 		    linkhdr, &linkhdrsize, &lladdr_off) != 0)
2274 			return (NULL);
2275 		lltable_set_entry_addr(ifp, lle, linkhdr, linkhdrsize,
2276 		    lladdr_off);
2277 		lle->la_flags |= LLE_STATIC;
2278 	}
2279 
2280 	if ((lle->la_flags & LLE_STATIC) != 0)
2281 		lle->ln_state = ND6_LLINFO_REACHABLE;
2282 
2283 	return (lle);
2284 }
2285 
2286 static struct llentry *
2287 in6_lltable_lookup(struct lltable *llt, u_int flags,
2288 	const struct sockaddr *l3addr)
2289 {
2290 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2291 	struct llentry *lle;
2292 
2293 	IF_AFDATA_LOCK_ASSERT(llt->llt_ifp);
2294 	KASSERT(l3addr->sa_family == AF_INET6,
2295 	    ("sin_family %d", l3addr->sa_family));
2296 
2297 	lle = in6_lltable_find_dst(llt, &sin6->sin6_addr);
2298 
2299 	if (lle == NULL)
2300 		return (NULL);
2301 
2302 	KASSERT((flags & (LLE_UNLOCKED|LLE_EXCLUSIVE)) !=
2303 	    (LLE_UNLOCKED|LLE_EXCLUSIVE),("wrong lle request flags: 0x%X",
2304 	    flags));
2305 
2306 	if (flags & LLE_UNLOCKED)
2307 		return (lle);
2308 
2309 	if (flags & LLE_EXCLUSIVE)
2310 		LLE_WLOCK(lle);
2311 	else
2312 		LLE_RLOCK(lle);
2313 	return (lle);
2314 }
2315 
2316 static int
2317 in6_lltable_dump_entry(struct lltable *llt, struct llentry *lle,
2318     struct sysctl_req *wr)
2319 {
2320 	struct ifnet *ifp = llt->llt_ifp;
2321 	/* XXX stack use */
2322 	struct {
2323 		struct rt_msghdr	rtm;
2324 		struct sockaddr_in6	sin6;
2325 		/*
2326 		 * ndp.c assumes that sdl is word aligned
2327 		 */
2328 #ifdef __LP64__
2329 		uint32_t		pad;
2330 #endif
2331 		struct sockaddr_dl	sdl;
2332 	} ndpc;
2333 	struct sockaddr_dl *sdl;
2334 	int error;
2335 
2336 	bzero(&ndpc, sizeof(ndpc));
2337 			/* skip deleted entries */
2338 			if ((lle->la_flags & LLE_DELETED) == LLE_DELETED)
2339 				return (0);
2340 			/* Skip if jailed and not a valid IP of the prison. */
2341 			lltable_fill_sa_entry(lle,
2342 			    (struct sockaddr *)&ndpc.sin6);
2343 			if (prison_if(wr->td->td_ucred,
2344 			    (struct sockaddr *)&ndpc.sin6) != 0)
2345 				return (0);
2346 			/*
2347 			 * produce a msg made of:
2348 			 *  struct rt_msghdr;
2349 			 *  struct sockaddr_in6 (IPv6)
2350 			 *  struct sockaddr_dl;
2351 			 */
2352 			ndpc.rtm.rtm_msglen = sizeof(ndpc);
2353 			ndpc.rtm.rtm_version = RTM_VERSION;
2354 			ndpc.rtm.rtm_type = RTM_GET;
2355 			ndpc.rtm.rtm_flags = RTF_UP;
2356 			ndpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
2357 			if (V_deembed_scopeid)
2358 				sa6_recoverscope(&ndpc.sin6);
2359 
2360 			/* publish */
2361 			if (lle->la_flags & LLE_PUB)
2362 				ndpc.rtm.rtm_flags |= RTF_ANNOUNCE;
2363 
2364 			sdl = &ndpc.sdl;
2365 			sdl->sdl_family = AF_LINK;
2366 			sdl->sdl_len = sizeof(*sdl);
2367 			sdl->sdl_alen = ifp->if_addrlen;
2368 			sdl->sdl_index = ifp->if_index;
2369 			sdl->sdl_type = ifp->if_type;
2370 			bcopy(&lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
2371 			if (lle->la_expire != 0)
2372 				ndpc.rtm.rtm_rmx.rmx_expire = lle->la_expire +
2373 				    lle->lle_remtime / hz +
2374 				    time_second - time_uptime;
2375 			ndpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
2376 			if (lle->la_flags & LLE_STATIC)
2377 				ndpc.rtm.rtm_flags |= RTF_STATIC;
2378 			if (lle->la_flags & LLE_IFADDR)
2379 				ndpc.rtm.rtm_flags |= RTF_PINNED;
2380 			if (lle->ln_router != 0)
2381 				ndpc.rtm.rtm_flags |= RTF_GATEWAY;
2382 			ndpc.rtm.rtm_rmx.rmx_pksent = lle->la_asked;
2383 			/* Store state in rmx_weight value */
2384 			ndpc.rtm.rtm_rmx.rmx_state = lle->ln_state;
2385 			ndpc.rtm.rtm_index = ifp->if_index;
2386 			error = SYSCTL_OUT(wr, &ndpc, sizeof(ndpc));
2387 
2388 	return (error);
2389 }
2390 
2391 static struct lltable *
2392 in6_lltattach(struct ifnet *ifp)
2393 {
2394 	struct lltable *llt;
2395 
2396 	llt = lltable_allocate_htbl(IN6_LLTBL_DEFAULT_HSIZE);
2397 	llt->llt_af = AF_INET6;
2398 	llt->llt_ifp = ifp;
2399 
2400 	llt->llt_lookup = in6_lltable_lookup;
2401 	llt->llt_alloc_entry = in6_lltable_alloc;
2402 	llt->llt_delete_entry = in6_lltable_delete_entry;
2403 	llt->llt_dump_entry = in6_lltable_dump_entry;
2404 	llt->llt_hash = in6_lltable_hash;
2405 	llt->llt_fill_sa_entry = in6_lltable_fill_sa_entry;
2406 	llt->llt_free_entry = in6_lltable_free_entry;
2407 	llt->llt_match_prefix = in6_lltable_match_prefix;
2408  	lltable_link(llt);
2409 
2410 	return (llt);
2411 }
2412 
2413 void *
2414 in6_domifattach(struct ifnet *ifp)
2415 {
2416 	struct in6_ifextra *ext;
2417 
2418 	/* There are not IPv6-capable interfaces. */
2419 	switch (ifp->if_type) {
2420 	case IFT_PFLOG:
2421 	case IFT_PFSYNC:
2422 	case IFT_USB:
2423 		return (NULL);
2424 	}
2425 	ext = (struct in6_ifextra *)malloc(sizeof(*ext), M_IFADDR, M_WAITOK);
2426 	bzero(ext, sizeof(*ext));
2427 
2428 	ext->in6_ifstat = malloc(sizeof(counter_u64_t) *
2429 	    sizeof(struct in6_ifstat) / sizeof(uint64_t), M_IFADDR, M_WAITOK);
2430 	COUNTER_ARRAY_ALLOC(ext->in6_ifstat,
2431 	    sizeof(struct in6_ifstat) / sizeof(uint64_t), M_WAITOK);
2432 
2433 	ext->icmp6_ifstat = malloc(sizeof(counter_u64_t) *
2434 	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t), M_IFADDR,
2435 	    M_WAITOK);
2436 	COUNTER_ARRAY_ALLOC(ext->icmp6_ifstat,
2437 	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t), M_WAITOK);
2438 
2439 	ext->nd_ifinfo = nd6_ifattach(ifp);
2440 	ext->scope6_id = scope6_ifattach(ifp);
2441 	ext->lltable = in6_lltattach(ifp);
2442 
2443 	ext->mld_ifinfo = mld_domifattach(ifp);
2444 
2445 	return ext;
2446 }
2447 
2448 int
2449 in6_domifmtu(struct ifnet *ifp)
2450 {
2451 	if (ifp->if_afdata[AF_INET6] == NULL)
2452 		return ifp->if_mtu;
2453 
2454 	return (IN6_LINKMTU(ifp));
2455 }
2456 
2457 void
2458 in6_domifdetach(struct ifnet *ifp, void *aux)
2459 {
2460 	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
2461 
2462 	mld_domifdetach(ifp);
2463 	scope6_ifdetach(ext->scope6_id);
2464 	nd6_ifdetach(ext->nd_ifinfo);
2465 	lltable_free(ext->lltable);
2466 	COUNTER_ARRAY_FREE(ext->in6_ifstat,
2467 	    sizeof(struct in6_ifstat) / sizeof(uint64_t));
2468 	free(ext->in6_ifstat, M_IFADDR);
2469 	COUNTER_ARRAY_FREE(ext->icmp6_ifstat,
2470 	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t));
2471 	free(ext->icmp6_ifstat, M_IFADDR);
2472 	free(ext, M_IFADDR);
2473 }
2474 
2475 /*
2476  * Convert sockaddr_in6 to sockaddr_in.  Original sockaddr_in6 must be
2477  * v4 mapped addr or v4 compat addr
2478  */
2479 void
2480 in6_sin6_2_sin(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2481 {
2482 
2483 	bzero(sin, sizeof(*sin));
2484 	sin->sin_len = sizeof(struct sockaddr_in);
2485 	sin->sin_family = AF_INET;
2486 	sin->sin_port = sin6->sin6_port;
2487 	sin->sin_addr.s_addr = sin6->sin6_addr.s6_addr32[3];
2488 }
2489 
2490 /* Convert sockaddr_in to sockaddr_in6 in v4 mapped addr format. */
2491 void
2492 in6_sin_2_v4mapsin6(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2493 {
2494 	bzero(sin6, sizeof(*sin6));
2495 	sin6->sin6_len = sizeof(struct sockaddr_in6);
2496 	sin6->sin6_family = AF_INET6;
2497 	sin6->sin6_port = sin->sin_port;
2498 	sin6->sin6_addr.s6_addr32[0] = 0;
2499 	sin6->sin6_addr.s6_addr32[1] = 0;
2500 	sin6->sin6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
2501 	sin6->sin6_addr.s6_addr32[3] = sin->sin_addr.s_addr;
2502 }
2503 
2504 /* Convert sockaddr_in6 into sockaddr_in. */
2505 void
2506 in6_sin6_2_sin_in_sock(struct sockaddr *nam)
2507 {
2508 	struct sockaddr_in *sin_p;
2509 	struct sockaddr_in6 sin6;
2510 
2511 	/*
2512 	 * Save original sockaddr_in6 addr and convert it
2513 	 * to sockaddr_in.
2514 	 */
2515 	sin6 = *(struct sockaddr_in6 *)nam;
2516 	sin_p = (struct sockaddr_in *)nam;
2517 	in6_sin6_2_sin(sin_p, &sin6);
2518 }
2519 
2520 /* Convert sockaddr_in into sockaddr_in6 in v4 mapped addr format. */
2521 void
2522 in6_sin_2_v4mapsin6_in_sock(struct sockaddr **nam)
2523 {
2524 	struct sockaddr_in *sin_p;
2525 	struct sockaddr_in6 *sin6_p;
2526 
2527 	sin6_p = malloc(sizeof *sin6_p, M_SONAME, M_WAITOK);
2528 	sin_p = (struct sockaddr_in *)*nam;
2529 	in6_sin_2_v4mapsin6(sin_p, sin6_p);
2530 	free(*nam, M_SONAME);
2531 	*nam = (struct sockaddr *)sin6_p;
2532 }
2533