xref: /freebsd/sys/netinet6/in6.c (revision eac7052fdebb90caf2f653e06187bdbca837b9c7)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the project nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	$KAME: in6.c,v 1.259 2002/01/21 11:37:50 keiichi Exp $
32  */
33 
34 /*-
35  * Copyright (c) 1982, 1986, 1991, 1993
36  *	The Regents of the University of California.  All rights reserved.
37  *
38  * Redistribution and use in source and binary forms, with or without
39  * modification, are permitted provided that the following conditions
40  * are met:
41  * 1. Redistributions of source code must retain the above copyright
42  *    notice, this list of conditions and the following disclaimer.
43  * 2. Redistributions in binary form must reproduce the above copyright
44  *    notice, this list of conditions and the following disclaimer in the
45  *    documentation and/or other materials provided with the distribution.
46  * 3. Neither the name of the University nor the names of its contributors
47  *    may be used to endorse or promote products derived from this software
48  *    without specific prior written permission.
49  *
50  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
51  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
52  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
53  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
54  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
55  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
56  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
57  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
58  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
59  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
60  * SUCH DAMAGE.
61  *
62  *	@(#)in.c	8.2 (Berkeley) 11/15/93
63  */
64 
65 #include <sys/cdefs.h>
66 __FBSDID("$FreeBSD$");
67 
68 #include "opt_inet.h"
69 #include "opt_inet6.h"
70 
71 #include <sys/param.h>
72 #include <sys/eventhandler.h>
73 #include <sys/errno.h>
74 #include <sys/jail.h>
75 #include <sys/malloc.h>
76 #include <sys/socket.h>
77 #include <sys/socketvar.h>
78 #include <sys/sockio.h>
79 #include <sys/systm.h>
80 #include <sys/priv.h>
81 #include <sys/proc.h>
82 #include <sys/protosw.h>
83 #include <sys/time.h>
84 #include <sys/kernel.h>
85 #include <sys/lock.h>
86 #include <sys/rmlock.h>
87 #include <sys/sysctl.h>
88 #include <sys/syslog.h>
89 
90 #include <net/if.h>
91 #include <net/if_var.h>
92 #include <net/if_types.h>
93 #include <net/route.h>
94 #include <net/route/nhop.h>
95 #include <net/if_dl.h>
96 #include <net/vnet.h>
97 
98 #include <netinet/in.h>
99 #include <netinet/in_var.h>
100 #include <net/if_llatbl.h>
101 #include <netinet/if_ether.h>
102 #include <netinet/in_systm.h>
103 #include <netinet/ip.h>
104 #include <netinet/in_pcb.h>
105 #include <netinet/ip_carp.h>
106 
107 #include <netinet/ip6.h>
108 #include <netinet6/ip6_var.h>
109 #include <netinet6/nd6.h>
110 #include <netinet6/mld6_var.h>
111 #include <netinet6/ip6_mroute.h>
112 #include <netinet6/in6_ifattach.h>
113 #include <netinet6/scope6_var.h>
114 #include <netinet6/in6_fib.h>
115 #include <netinet6/in6_pcb.h>
116 
117 /*
118  * struct in6_ifreq and struct ifreq must be type punnable for common members
119  * of ifr_ifru to allow accessors to be shared.
120  */
121 _Static_assert(offsetof(struct in6_ifreq, ifr_ifru) ==
122     offsetof(struct ifreq, ifr_ifru),
123     "struct in6_ifreq and struct ifreq are not type punnable");
124 
125 VNET_DECLARE(int, icmp6_nodeinfo_oldmcprefix);
126 #define V_icmp6_nodeinfo_oldmcprefix	VNET(icmp6_nodeinfo_oldmcprefix)
127 
128 /*
129  * Definitions of some costant IP6 addresses.
130  */
131 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
132 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
133 const struct in6_addr in6addr_nodelocal_allnodes =
134 	IN6ADDR_NODELOCAL_ALLNODES_INIT;
135 const struct in6_addr in6addr_linklocal_allnodes =
136 	IN6ADDR_LINKLOCAL_ALLNODES_INIT;
137 const struct in6_addr in6addr_linklocal_allrouters =
138 	IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
139 const struct in6_addr in6addr_linklocal_allv2routers =
140 	IN6ADDR_LINKLOCAL_ALLV2ROUTERS_INIT;
141 
142 const struct in6_addr in6mask0 = IN6MASK0;
143 const struct in6_addr in6mask32 = IN6MASK32;
144 const struct in6_addr in6mask64 = IN6MASK64;
145 const struct in6_addr in6mask96 = IN6MASK96;
146 const struct in6_addr in6mask128 = IN6MASK128;
147 
148 const struct sockaddr_in6 sa6_any =
149 	{ sizeof(sa6_any), AF_INET6, 0, 0, IN6ADDR_ANY_INIT, 0 };
150 
151 static int in6_notify_ifa(struct ifnet *, struct in6_ifaddr *,
152 	struct in6_aliasreq *, int);
153 static void in6_unlink_ifa(struct in6_ifaddr *, struct ifnet *);
154 
155 static int in6_validate_ifra(struct ifnet *, struct in6_aliasreq *,
156     struct in6_ifaddr *, int);
157 static struct in6_ifaddr *in6_alloc_ifa(struct ifnet *,
158     struct in6_aliasreq *, int flags);
159 static int in6_update_ifa_internal(struct ifnet *, struct in6_aliasreq *,
160     struct in6_ifaddr *, int, int);
161 static int in6_broadcast_ifa(struct ifnet *, struct in6_aliasreq *,
162     struct in6_ifaddr *, int);
163 
164 #define ifa2ia6(ifa)	((struct in6_ifaddr *)(ifa))
165 #define ia62ifa(ia6)	(&((ia6)->ia_ifa))
166 
167 void
168 in6_newaddrmsg(struct in6_ifaddr *ia, int cmd)
169 {
170 	struct rt_addrinfo info;
171 	struct ifaddr *ifa;
172 	struct sockaddr_dl gateway;
173 	int fibnum;
174 
175 	ifa = &ia->ia_ifa;
176 
177 	/*
178 	 * Prepare info data for the host route.
179 	 * This code mimics one from ifa_maintain_loopback_route().
180 	 */
181 	bzero(&info, sizeof(struct rt_addrinfo));
182 	info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC | RTF_PINNED;
183 	info.rti_info[RTAX_DST] = ifa->ifa_addr;
184 	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&gateway;
185 	link_init_sdl(ifa->ifa_ifp, (struct sockaddr *)&gateway, ifa->ifa_ifp->if_type);
186 	if (cmd != RTM_DELETE)
187 		info.rti_ifp = V_loif;
188 
189 	fibnum = V_rt_add_addr_allfibs ? RT_ALL_FIBS : ia62ifa(ia)->ifa_ifp->if_fib;
190 
191 	if (cmd == RTM_ADD) {
192 		rt_addrmsg(cmd, &ia->ia_ifa, fibnum);
193 		rt_routemsg_info(cmd, &info, fibnum);
194 	} else if (cmd == RTM_DELETE) {
195 		rt_routemsg_info(cmd, &info, fibnum);
196 		rt_addrmsg(cmd, &ia->ia_ifa, fibnum);
197 	}
198 }
199 
200 int
201 in6_mask2len(struct in6_addr *mask, u_char *lim0)
202 {
203 	int x = 0, y;
204 	u_char *lim = lim0, *p;
205 
206 	/* ignore the scope_id part */
207 	if (lim0 == NULL || lim0 - (u_char *)mask > sizeof(*mask))
208 		lim = (u_char *)mask + sizeof(*mask);
209 	for (p = (u_char *)mask; p < lim; x++, p++) {
210 		if (*p != 0xff)
211 			break;
212 	}
213 	y = 0;
214 	if (p < lim) {
215 		for (y = 0; y < 8; y++) {
216 			if ((*p & (0x80 >> y)) == 0)
217 				break;
218 		}
219 	}
220 
221 	/*
222 	 * when the limit pointer is given, do a stricter check on the
223 	 * remaining bits.
224 	 */
225 	if (p < lim) {
226 		if (y != 0 && (*p & (0x00ff >> y)) != 0)
227 			return (-1);
228 		for (p = p + 1; p < lim; p++)
229 			if (*p != 0)
230 				return (-1);
231 	}
232 
233 	return x * 8 + y;
234 }
235 
236 #ifdef COMPAT_FREEBSD32
237 struct in6_ndifreq32 {
238 	char ifname[IFNAMSIZ];
239 	uint32_t ifindex;
240 };
241 #define	SIOCGDEFIFACE32_IN6	_IOWR('i', 86, struct in6_ndifreq32)
242 #endif
243 
244 int
245 in6_control(struct socket *so, u_long cmd, caddr_t data,
246     struct ifnet *ifp, struct thread *td)
247 {
248 	struct	in6_ifreq *ifr = (struct in6_ifreq *)data;
249 	struct	in6_ifaddr *ia = NULL;
250 	struct	in6_aliasreq *ifra = (struct in6_aliasreq *)data;
251 	struct sockaddr_in6 *sa6;
252 	int carp_attached = 0;
253 	int error;
254 	u_long ocmd = cmd;
255 
256 	/*
257 	 * Compat to make pre-10.x ifconfig(8) operable.
258 	 */
259 	if (cmd == OSIOCAIFADDR_IN6)
260 		cmd = SIOCAIFADDR_IN6;
261 
262 	switch (cmd) {
263 	case SIOCGETSGCNT_IN6:
264 	case SIOCGETMIFCNT_IN6:
265 		/*
266 		 * XXX mrt_ioctl has a 3rd, unused, FIB argument in route.c.
267 		 * We cannot see how that would be needed, so do not adjust the
268 		 * KPI blindly; more likely should clean up the IPv4 variant.
269 		 */
270 		return (mrt6_ioctl ? mrt6_ioctl(cmd, data) : EOPNOTSUPP);
271 	}
272 
273 	switch (cmd) {
274 	case SIOCAADDRCTL_POLICY:
275 	case SIOCDADDRCTL_POLICY:
276 		if (td != NULL) {
277 			error = priv_check(td, PRIV_NETINET_ADDRCTRL6);
278 			if (error)
279 				return (error);
280 		}
281 		return (in6_src_ioctl(cmd, data));
282 	}
283 
284 	if (ifp == NULL)
285 		return (EOPNOTSUPP);
286 
287 	switch (cmd) {
288 	case SIOCSNDFLUSH_IN6:
289 	case SIOCSPFXFLUSH_IN6:
290 	case SIOCSRTRFLUSH_IN6:
291 	case SIOCSDEFIFACE_IN6:
292 	case SIOCSIFINFO_FLAGS:
293 	case SIOCSIFINFO_IN6:
294 		if (td != NULL) {
295 			error = priv_check(td, PRIV_NETINET_ND6);
296 			if (error)
297 				return (error);
298 		}
299 		/* FALLTHROUGH */
300 	case OSIOCGIFINFO_IN6:
301 	case SIOCGIFINFO_IN6:
302 	case SIOCGNBRINFO_IN6:
303 	case SIOCGDEFIFACE_IN6:
304 		return (nd6_ioctl(cmd, data, ifp));
305 
306 #ifdef COMPAT_FREEBSD32
307 	case SIOCGDEFIFACE32_IN6:
308 		{
309 			struct in6_ndifreq ndif;
310 			struct in6_ndifreq32 *ndif32;
311 
312 			error = nd6_ioctl(SIOCGDEFIFACE_IN6, (caddr_t)&ndif,
313 			    ifp);
314 			if (error)
315 				return (error);
316 			ndif32 = (struct in6_ndifreq32 *)data;
317 			ndif32->ifindex = ndif.ifindex;
318 			return (0);
319 		}
320 #endif
321 	}
322 
323 	switch (cmd) {
324 	case SIOCSIFPREFIX_IN6:
325 	case SIOCDIFPREFIX_IN6:
326 	case SIOCAIFPREFIX_IN6:
327 	case SIOCCIFPREFIX_IN6:
328 	case SIOCSGIFPREFIX_IN6:
329 	case SIOCGIFPREFIX_IN6:
330 		log(LOG_NOTICE,
331 		    "prefix ioctls are now invalidated. "
332 		    "please use ifconfig.\n");
333 		return (EOPNOTSUPP);
334 	}
335 
336 	switch (cmd) {
337 	case SIOCSSCOPE6:
338 		if (td != NULL) {
339 			error = priv_check(td, PRIV_NETINET_SCOPE6);
340 			if (error)
341 				return (error);
342 		}
343 		/* FALLTHROUGH */
344 	case SIOCGSCOPE6:
345 	case SIOCGSCOPE6DEF:
346 		return (scope6_ioctl(cmd, data, ifp));
347 	}
348 
349 	/*
350 	 * Find address for this interface, if it exists.
351 	 *
352 	 * In netinet code, we have checked ifra_addr in SIOCSIF*ADDR operation
353 	 * only, and used the first interface address as the target of other
354 	 * operations (without checking ifra_addr).  This was because netinet
355 	 * code/API assumed at most 1 interface address per interface.
356 	 * Since IPv6 allows a node to assign multiple addresses
357 	 * on a single interface, we almost always look and check the
358 	 * presence of ifra_addr, and reject invalid ones here.
359 	 * It also decreases duplicated code among SIOC*_IN6 operations.
360 	 */
361 	switch (cmd) {
362 	case SIOCAIFADDR_IN6:
363 	case SIOCSIFPHYADDR_IN6:
364 		sa6 = &ifra->ifra_addr;
365 		break;
366 	case SIOCSIFADDR_IN6:
367 	case SIOCGIFADDR_IN6:
368 	case SIOCSIFDSTADDR_IN6:
369 	case SIOCSIFNETMASK_IN6:
370 	case SIOCGIFDSTADDR_IN6:
371 	case SIOCGIFNETMASK_IN6:
372 	case SIOCDIFADDR_IN6:
373 	case SIOCGIFPSRCADDR_IN6:
374 	case SIOCGIFPDSTADDR_IN6:
375 	case SIOCGIFAFLAG_IN6:
376 	case SIOCSNDFLUSH_IN6:
377 	case SIOCSPFXFLUSH_IN6:
378 	case SIOCSRTRFLUSH_IN6:
379 	case SIOCGIFALIFETIME_IN6:
380 	case SIOCGIFSTAT_IN6:
381 	case SIOCGIFSTAT_ICMP6:
382 		sa6 = &ifr->ifr_addr;
383 		break;
384 	case SIOCSIFADDR:
385 	case SIOCSIFBRDADDR:
386 	case SIOCSIFDSTADDR:
387 	case SIOCSIFNETMASK:
388 		/*
389 		 * Although we should pass any non-INET6 ioctl requests
390 		 * down to driver, we filter some legacy INET requests.
391 		 * Drivers trust SIOCSIFADDR et al to come from an already
392 		 * privileged layer, and do not perform any credentials
393 		 * checks or input validation.
394 		 */
395 		return (EINVAL);
396 	default:
397 		sa6 = NULL;
398 		break;
399 	}
400 	if (sa6 && sa6->sin6_family == AF_INET6) {
401 		if (sa6->sin6_scope_id != 0)
402 			error = sa6_embedscope(sa6, 0);
403 		else
404 			error = in6_setscope(&sa6->sin6_addr, ifp, NULL);
405 		if (error != 0)
406 			return (error);
407 		if (td != NULL && (error = prison_check_ip6(td->td_ucred,
408 		    &sa6->sin6_addr)) != 0)
409 			return (error);
410 		ia = in6ifa_ifpwithaddr(ifp, &sa6->sin6_addr);
411 	} else
412 		ia = NULL;
413 
414 	switch (cmd) {
415 	case SIOCSIFADDR_IN6:
416 	case SIOCSIFDSTADDR_IN6:
417 	case SIOCSIFNETMASK_IN6:
418 		/*
419 		 * Since IPv6 allows a node to assign multiple addresses
420 		 * on a single interface, SIOCSIFxxx ioctls are deprecated.
421 		 */
422 		/* we decided to obsolete this command (20000704) */
423 		error = EINVAL;
424 		goto out;
425 
426 	case SIOCDIFADDR_IN6:
427 		/*
428 		 * for IPv4, we look for existing in_ifaddr here to allow
429 		 * "ifconfig if0 delete" to remove the first IPv4 address on
430 		 * the interface.  For IPv6, as the spec allows multiple
431 		 * interface address from the day one, we consider "remove the
432 		 * first one" semantics to be not preferable.
433 		 */
434 		if (ia == NULL) {
435 			error = EADDRNOTAVAIL;
436 			goto out;
437 		}
438 		/* FALLTHROUGH */
439 	case SIOCAIFADDR_IN6:
440 		/*
441 		 * We always require users to specify a valid IPv6 address for
442 		 * the corresponding operation.
443 		 */
444 		if (ifra->ifra_addr.sin6_family != AF_INET6 ||
445 		    ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6)) {
446 			error = EAFNOSUPPORT;
447 			goto out;
448 		}
449 
450 		if (td != NULL) {
451 			error = priv_check(td, (cmd == SIOCDIFADDR_IN6) ?
452 			    PRIV_NET_DELIFADDR : PRIV_NET_ADDIFADDR);
453 			if (error)
454 				goto out;
455 		}
456 		/* FALLTHROUGH */
457 	case SIOCGIFSTAT_IN6:
458 	case SIOCGIFSTAT_ICMP6:
459 		if (ifp->if_afdata[AF_INET6] == NULL) {
460 			error = EPFNOSUPPORT;
461 			goto out;
462 		}
463 		break;
464 
465 	case SIOCGIFADDR_IN6:
466 		/* This interface is basically deprecated. use SIOCGIFCONF. */
467 		/* FALLTHROUGH */
468 	case SIOCGIFAFLAG_IN6:
469 	case SIOCGIFNETMASK_IN6:
470 	case SIOCGIFDSTADDR_IN6:
471 	case SIOCGIFALIFETIME_IN6:
472 		/* must think again about its semantics */
473 		if (ia == NULL) {
474 			error = EADDRNOTAVAIL;
475 			goto out;
476 		}
477 		break;
478 	}
479 
480 	switch (cmd) {
481 	case SIOCGIFADDR_IN6:
482 		ifr->ifr_addr = ia->ia_addr;
483 		if ((error = sa6_recoverscope(&ifr->ifr_addr)) != 0)
484 			goto out;
485 		break;
486 
487 	case SIOCGIFDSTADDR_IN6:
488 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
489 			error = EINVAL;
490 			goto out;
491 		}
492 		ifr->ifr_dstaddr = ia->ia_dstaddr;
493 		if ((error = sa6_recoverscope(&ifr->ifr_dstaddr)) != 0)
494 			goto out;
495 		break;
496 
497 	case SIOCGIFNETMASK_IN6:
498 		ifr->ifr_addr = ia->ia_prefixmask;
499 		break;
500 
501 	case SIOCGIFAFLAG_IN6:
502 		ifr->ifr_ifru.ifru_flags6 = ia->ia6_flags;
503 		break;
504 
505 	case SIOCGIFSTAT_IN6:
506 		COUNTER_ARRAY_COPY(((struct in6_ifextra *)
507 		    ifp->if_afdata[AF_INET6])->in6_ifstat,
508 		    &ifr->ifr_ifru.ifru_stat,
509 		    sizeof(struct in6_ifstat) / sizeof(uint64_t));
510 		break;
511 
512 	case SIOCGIFSTAT_ICMP6:
513 		COUNTER_ARRAY_COPY(((struct in6_ifextra *)
514 		    ifp->if_afdata[AF_INET6])->icmp6_ifstat,
515 		    &ifr->ifr_ifru.ifru_icmp6stat,
516 		    sizeof(struct icmp6_ifstat) / sizeof(uint64_t));
517 		break;
518 
519 	case SIOCGIFALIFETIME_IN6:
520 		ifr->ifr_ifru.ifru_lifetime = ia->ia6_lifetime;
521 		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
522 			time_t maxexpire;
523 			struct in6_addrlifetime *retlt =
524 			    &ifr->ifr_ifru.ifru_lifetime;
525 
526 			/*
527 			 * XXX: adjust expiration time assuming time_t is
528 			 * signed.
529 			 */
530 			maxexpire = (-1) &
531 			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
532 			if (ia->ia6_lifetime.ia6t_vltime <
533 			    maxexpire - ia->ia6_updatetime) {
534 				retlt->ia6t_expire = ia->ia6_updatetime +
535 				    ia->ia6_lifetime.ia6t_vltime;
536 			} else
537 				retlt->ia6t_expire = maxexpire;
538 		}
539 		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
540 			time_t maxexpire;
541 			struct in6_addrlifetime *retlt =
542 			    &ifr->ifr_ifru.ifru_lifetime;
543 
544 			/*
545 			 * XXX: adjust expiration time assuming time_t is
546 			 * signed.
547 			 */
548 			maxexpire = (-1) &
549 			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
550 			if (ia->ia6_lifetime.ia6t_pltime <
551 			    maxexpire - ia->ia6_updatetime) {
552 				retlt->ia6t_preferred = ia->ia6_updatetime +
553 				    ia->ia6_lifetime.ia6t_pltime;
554 			} else
555 				retlt->ia6t_preferred = maxexpire;
556 		}
557 		break;
558 
559 	case SIOCAIFADDR_IN6:
560 	{
561 		struct nd_prefixctl pr0;
562 		struct nd_prefix *pr;
563 
564 		/*
565 		 * first, make or update the interface address structure,
566 		 * and link it to the list.
567 		 */
568 		if ((error = in6_update_ifa(ifp, ifra, ia, 0)) != 0)
569 			goto out;
570 		if (ia != NULL) {
571 			if (ia->ia_ifa.ifa_carp)
572 				(*carp_detach_p)(&ia->ia_ifa, true);
573 			ifa_free(&ia->ia_ifa);
574 		}
575 		if ((ia = in6ifa_ifpwithaddr(ifp, &ifra->ifra_addr.sin6_addr))
576 		    == NULL) {
577 			/*
578 			 * this can happen when the user specify the 0 valid
579 			 * lifetime.
580 			 */
581 			break;
582 		}
583 
584 		if (cmd == ocmd && ifra->ifra_vhid > 0) {
585 			if (carp_attach_p != NULL)
586 				error = (*carp_attach_p)(&ia->ia_ifa,
587 				    ifra->ifra_vhid);
588 			else
589 				error = EPROTONOSUPPORT;
590 			if (error)
591 				goto out;
592 			else
593 				carp_attached = 1;
594 		}
595 
596 		/*
597 		 * then, make the prefix on-link on the interface.
598 		 * XXX: we'd rather create the prefix before the address, but
599 		 * we need at least one address to install the corresponding
600 		 * interface route, so we configure the address first.
601 		 */
602 
603 		/*
604 		 * convert mask to prefix length (prefixmask has already
605 		 * been validated in in6_update_ifa().
606 		 */
607 		bzero(&pr0, sizeof(pr0));
608 		pr0.ndpr_ifp = ifp;
609 		pr0.ndpr_plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
610 		    NULL);
611 		if (pr0.ndpr_plen == 128) {
612 			/* we don't need to install a host route. */
613 			goto aifaddr_out;
614 		}
615 		pr0.ndpr_prefix = ifra->ifra_addr;
616 		/* apply the mask for safety. */
617 		IN6_MASK_ADDR(&pr0.ndpr_prefix.sin6_addr,
618 		    &ifra->ifra_prefixmask.sin6_addr);
619 
620 		/*
621 		 * XXX: since we don't have an API to set prefix (not address)
622 		 * lifetimes, we just use the same lifetimes as addresses.
623 		 * The (temporarily) installed lifetimes can be overridden by
624 		 * later advertised RAs (when accept_rtadv is non 0), which is
625 		 * an intended behavior.
626 		 */
627 		pr0.ndpr_raf_onlink = 1; /* should be configurable? */
628 		pr0.ndpr_raf_auto =
629 		    ((ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0);
630 		pr0.ndpr_vltime = ifra->ifra_lifetime.ia6t_vltime;
631 		pr0.ndpr_pltime = ifra->ifra_lifetime.ia6t_pltime;
632 
633 		/* add the prefix if not yet. */
634 		if ((pr = nd6_prefix_lookup(&pr0)) == NULL) {
635 			/*
636 			 * nd6_prelist_add will install the corresponding
637 			 * interface route.
638 			 */
639 			if ((error = nd6_prelist_add(&pr0, NULL, &pr)) != 0) {
640 				if (carp_attached)
641 					(*carp_detach_p)(&ia->ia_ifa, false);
642 				goto out;
643 			}
644 		}
645 
646 		/* relate the address to the prefix */
647 		if (ia->ia6_ndpr == NULL) {
648 			ia->ia6_ndpr = pr;
649 			pr->ndpr_addrcnt++;
650 
651 			/*
652 			 * If this is the first autoconf address from the
653 			 * prefix, create a temporary address as well
654 			 * (when required).
655 			 */
656 			if ((ia->ia6_flags & IN6_IFF_AUTOCONF) &&
657 			    V_ip6_use_tempaddr && pr->ndpr_addrcnt == 1) {
658 				int e;
659 				if ((e = in6_tmpifadd(ia, 1, 0)) != 0) {
660 					log(LOG_NOTICE, "in6_control: failed "
661 					    "to create a temporary address, "
662 					    "errno=%d\n", e);
663 				}
664 			}
665 		}
666 		nd6_prefix_rele(pr);
667 
668 		/*
669 		 * this might affect the status of autoconfigured addresses,
670 		 * that is, this address might make other addresses detached.
671 		 */
672 		pfxlist_onlink_check();
673 
674 aifaddr_out:
675 		/*
676 		 * Try to clear the flag when a new IPv6 address is added
677 		 * onto an IFDISABLED interface and it succeeds.
678 		 */
679 		if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) {
680 			struct in6_ndireq nd;
681 
682 			memset(&nd, 0, sizeof(nd));
683 			nd.ndi.flags = ND_IFINFO(ifp)->flags;
684 			nd.ndi.flags &= ~ND6_IFF_IFDISABLED;
685 			if (nd6_ioctl(SIOCSIFINFO_FLAGS, (caddr_t)&nd, ifp) < 0)
686 				log(LOG_NOTICE, "SIOCAIFADDR_IN6: "
687 				    "SIOCSIFINFO_FLAGS for -ifdisabled "
688 				    "failed.");
689 			/*
690 			 * Ignore failure of clearing the flag intentionally.
691 			 * The failure means address duplication was detected.
692 			 */
693 		}
694 		break;
695 	}
696 
697 	case SIOCDIFADDR_IN6:
698 	{
699 		struct nd_prefix *pr;
700 
701 		/*
702 		 * If the address being deleted is the only one that owns
703 		 * the corresponding prefix, expire the prefix as well.
704 		 * XXX: theoretically, we don't have to worry about such
705 		 * relationship, since we separate the address management
706 		 * and the prefix management.  We do this, however, to provide
707 		 * as much backward compatibility as possible in terms of
708 		 * the ioctl operation.
709 		 * Note that in6_purgeaddr() will decrement ndpr_addrcnt.
710 		 */
711 		pr = ia->ia6_ndpr;
712 		in6_purgeaddr(&ia->ia_ifa);
713 		if (pr != NULL && pr->ndpr_addrcnt == 0) {
714 			ND6_WLOCK();
715 			nd6_prefix_unlink(pr, NULL);
716 			ND6_WUNLOCK();
717 			nd6_prefix_del(pr);
718 		}
719 		EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, &ia->ia_ifa,
720 		    IFADDR_EVENT_DEL);
721 		break;
722 	}
723 
724 	default:
725 		if (ifp->if_ioctl == NULL) {
726 			error = EOPNOTSUPP;
727 			goto out;
728 		}
729 		error = (*ifp->if_ioctl)(ifp, cmd, data);
730 		goto out;
731 	}
732 
733 	error = 0;
734 out:
735 	if (ia != NULL)
736 		ifa_free(&ia->ia_ifa);
737 	return (error);
738 }
739 
740 static struct in6_multi_mship *
741 in6_joingroup_legacy(struct ifnet *ifp, const struct in6_addr *mcaddr,
742     int *errorp, int delay)
743 {
744 	struct in6_multi_mship *imm;
745 	int error;
746 
747 	imm = malloc(sizeof(*imm), M_IP6MADDR, M_NOWAIT);
748 	if (imm == NULL) {
749 		*errorp = ENOBUFS;
750 		return (NULL);
751 	}
752 
753 	delay = (delay * PR_FASTHZ) / hz;
754 
755 	error = in6_joingroup(ifp, mcaddr, NULL, &imm->i6mm_maddr, delay);
756 	if (error) {
757 		*errorp = error;
758 		free(imm, M_IP6MADDR);
759 		return (NULL);
760 	}
761 
762 	return (imm);
763 }
764 /*
765  * Join necessary multicast groups.  Factored out from in6_update_ifa().
766  * This entire work should only be done once, for the default FIB.
767  */
768 static int
769 in6_update_ifa_join_mc(struct ifnet *ifp, struct in6_aliasreq *ifra,
770     struct in6_ifaddr *ia, int flags, struct in6_multi **in6m_sol)
771 {
772 	char ip6buf[INET6_ADDRSTRLEN];
773 	struct in6_addr mltaddr;
774 	struct in6_multi_mship *imm;
775 	int delay, error;
776 
777 	KASSERT(in6m_sol != NULL, ("%s: in6m_sol is NULL", __func__));
778 
779 	/* Join solicited multicast addr for new host id. */
780 	bzero(&mltaddr, sizeof(struct in6_addr));
781 	mltaddr.s6_addr32[0] = IPV6_ADDR_INT32_MLL;
782 	mltaddr.s6_addr32[2] = htonl(1);
783 	mltaddr.s6_addr32[3] = ifra->ifra_addr.sin6_addr.s6_addr32[3];
784 	mltaddr.s6_addr8[12] = 0xff;
785 	if ((error = in6_setscope(&mltaddr, ifp, NULL)) != 0) {
786 		/* XXX: should not happen */
787 		log(LOG_ERR, "%s: in6_setscope failed\n", __func__);
788 		goto cleanup;
789 	}
790 	delay = error = 0;
791 	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
792 		/*
793 		 * We need a random delay for DAD on the address being
794 		 * configured.  It also means delaying transmission of the
795 		 * corresponding MLD report to avoid report collision.
796 		 * [RFC 4861, Section 6.3.7]
797 		 */
798 		delay = arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz);
799 	}
800 	imm = in6_joingroup_legacy(ifp, &mltaddr, &error, delay);
801 	if (imm == NULL) {
802 		nd6log((LOG_WARNING, "%s: in6_joingroup failed for %s on %s "
803 		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf, &mltaddr),
804 		    if_name(ifp), error));
805 		goto cleanup;
806 	}
807 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
808 	*in6m_sol = imm->i6mm_maddr;
809 
810 	/*
811 	 * Join link-local all-nodes address.
812 	 */
813 	mltaddr = in6addr_linklocal_allnodes;
814 	if ((error = in6_setscope(&mltaddr, ifp, NULL)) != 0)
815 		goto cleanup; /* XXX: should not fail */
816 
817 	imm = in6_joingroup_legacy(ifp, &mltaddr, &error, 0);
818 	if (imm == NULL) {
819 		nd6log((LOG_WARNING, "%s: in6_joingroup failed for %s on %s "
820 		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf, &mltaddr),
821 		    if_name(ifp), error));
822 		goto cleanup;
823 	}
824 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
825 
826 	/*
827 	 * Join node information group address.
828 	 */
829 	delay = 0;
830 	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
831 		/*
832 		 * The spec does not say anything about delay for this group,
833 		 * but the same logic should apply.
834 		 */
835 		delay = arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz);
836 	}
837 	if (in6_nigroup(ifp, NULL, -1, &mltaddr) == 0) {
838 		/* XXX jinmei */
839 		imm = in6_joingroup_legacy(ifp, &mltaddr, &error, delay);
840 		if (imm == NULL)
841 			nd6log((LOG_WARNING,
842 			    "%s: in6_joingroup failed for %s on %s "
843 			    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
844 			    &mltaddr), if_name(ifp), error));
845 			/* XXX not very fatal, go on... */
846 		else
847 			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
848 	}
849 	if (V_icmp6_nodeinfo_oldmcprefix &&
850 	    in6_nigroup_oldmcprefix(ifp, NULL, -1, &mltaddr) == 0) {
851 		imm = in6_joingroup_legacy(ifp, &mltaddr, &error, delay);
852 		if (imm == NULL)
853 			nd6log((LOG_WARNING,
854 			    "%s: in6_joingroup failed for %s on %s "
855 			    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
856 			    &mltaddr), if_name(ifp), error));
857 			/* XXX not very fatal, go on... */
858 		else
859 			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
860 	}
861 
862 	/*
863 	 * Join interface-local all-nodes address.
864 	 * (ff01::1%ifN, and ff01::%ifN/32)
865 	 */
866 	mltaddr = in6addr_nodelocal_allnodes;
867 	if ((error = in6_setscope(&mltaddr, ifp, NULL)) != 0)
868 		goto cleanup; /* XXX: should not fail */
869 
870 	imm = in6_joingroup_legacy(ifp, &mltaddr, &error, 0);
871 	if (imm == NULL) {
872 		nd6log((LOG_WARNING, "%s: in6_joingroup failed for %s on %s "
873 		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
874 		    &mltaddr), if_name(ifp), error));
875 		goto cleanup;
876 	}
877 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
878 
879 cleanup:
880 	return (error);
881 }
882 
883 /*
884  * Update parameters of an IPv6 interface address.
885  * If necessary, a new entry is created and linked into address chains.
886  * This function is separated from in6_control().
887  */
888 int
889 in6_update_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
890     struct in6_ifaddr *ia, int flags)
891 {
892 	int error, hostIsNew = 0;
893 
894 	if ((error = in6_validate_ifra(ifp, ifra, ia, flags)) != 0)
895 		return (error);
896 
897 	if (ia == NULL) {
898 		hostIsNew = 1;
899 		if ((ia = in6_alloc_ifa(ifp, ifra, flags)) == NULL)
900 			return (ENOBUFS);
901 	}
902 
903 	error = in6_update_ifa_internal(ifp, ifra, ia, hostIsNew, flags);
904 	if (error != 0) {
905 		if (hostIsNew != 0) {
906 			in6_unlink_ifa(ia, ifp);
907 			ifa_free(&ia->ia_ifa);
908 		}
909 		return (error);
910 	}
911 
912 	if (hostIsNew)
913 		error = in6_broadcast_ifa(ifp, ifra, ia, flags);
914 
915 	return (error);
916 }
917 
918 /*
919  * Fill in basic IPv6 address request info.
920  */
921 void
922 in6_prepare_ifra(struct in6_aliasreq *ifra, const struct in6_addr *addr,
923     const struct in6_addr *mask)
924 {
925 
926 	memset(ifra, 0, sizeof(struct in6_aliasreq));
927 
928 	ifra->ifra_addr.sin6_family = AF_INET6;
929 	ifra->ifra_addr.sin6_len = sizeof(struct sockaddr_in6);
930 	if (addr != NULL)
931 		ifra->ifra_addr.sin6_addr = *addr;
932 
933 	ifra->ifra_prefixmask.sin6_family = AF_INET6;
934 	ifra->ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
935 	if (mask != NULL)
936 		ifra->ifra_prefixmask.sin6_addr = *mask;
937 }
938 
939 static int
940 in6_validate_ifra(struct ifnet *ifp, struct in6_aliasreq *ifra,
941     struct in6_ifaddr *ia, int flags)
942 {
943 	int plen = -1;
944 	struct sockaddr_in6 dst6;
945 	struct in6_addrlifetime *lt;
946 	char ip6buf[INET6_ADDRSTRLEN];
947 
948 	/* Validate parameters */
949 	if (ifp == NULL || ifra == NULL) /* this maybe redundant */
950 		return (EINVAL);
951 
952 	/*
953 	 * The destination address for a p2p link must have a family
954 	 * of AF_UNSPEC or AF_INET6.
955 	 */
956 	if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
957 	    ifra->ifra_dstaddr.sin6_family != AF_INET6 &&
958 	    ifra->ifra_dstaddr.sin6_family != AF_UNSPEC)
959 		return (EAFNOSUPPORT);
960 
961 	/*
962 	 * Validate address
963 	 */
964 	if (ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6) ||
965 	    ifra->ifra_addr.sin6_family != AF_INET6)
966 		return (EINVAL);
967 
968 	/*
969 	 * validate ifra_prefixmask.  don't check sin6_family, netmask
970 	 * does not carry fields other than sin6_len.
971 	 */
972 	if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6))
973 		return (EINVAL);
974 	/*
975 	 * Because the IPv6 address architecture is classless, we require
976 	 * users to specify a (non 0) prefix length (mask) for a new address.
977 	 * We also require the prefix (when specified) mask is valid, and thus
978 	 * reject a non-consecutive mask.
979 	 */
980 	if (ia == NULL && ifra->ifra_prefixmask.sin6_len == 0)
981 		return (EINVAL);
982 	if (ifra->ifra_prefixmask.sin6_len != 0) {
983 		plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
984 		    (u_char *)&ifra->ifra_prefixmask +
985 		    ifra->ifra_prefixmask.sin6_len);
986 		if (plen <= 0)
987 			return (EINVAL);
988 	} else {
989 		/*
990 		 * In this case, ia must not be NULL.  We just use its prefix
991 		 * length.
992 		 */
993 		plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
994 	}
995 	/*
996 	 * If the destination address on a p2p interface is specified,
997 	 * and the address is a scoped one, validate/set the scope
998 	 * zone identifier.
999 	 */
1000 	dst6 = ifra->ifra_dstaddr;
1001 	if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) != 0 &&
1002 	    (dst6.sin6_family == AF_INET6)) {
1003 		struct in6_addr in6_tmp;
1004 		u_int32_t zoneid;
1005 
1006 		in6_tmp = dst6.sin6_addr;
1007 		if (in6_setscope(&in6_tmp, ifp, &zoneid))
1008 			return (EINVAL); /* XXX: should be impossible */
1009 
1010 		if (dst6.sin6_scope_id != 0) {
1011 			if (dst6.sin6_scope_id != zoneid)
1012 				return (EINVAL);
1013 		} else		/* user omit to specify the ID. */
1014 			dst6.sin6_scope_id = zoneid;
1015 
1016 		/* convert into the internal form */
1017 		if (sa6_embedscope(&dst6, 0))
1018 			return (EINVAL); /* XXX: should be impossible */
1019 	}
1020 	/* Modify original ifra_dstaddr to reflect changes */
1021 	ifra->ifra_dstaddr = dst6;
1022 
1023 	/*
1024 	 * The destination address can be specified only for a p2p or a
1025 	 * loopback interface.  If specified, the corresponding prefix length
1026 	 * must be 128.
1027 	 */
1028 	if (ifra->ifra_dstaddr.sin6_family == AF_INET6) {
1029 		if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) == 0) {
1030 			/* XXX: noisy message */
1031 			nd6log((LOG_INFO, "in6_update_ifa: a destination can "
1032 			    "be specified for a p2p or a loopback IF only\n"));
1033 			return (EINVAL);
1034 		}
1035 		if (plen != 128) {
1036 			nd6log((LOG_INFO, "in6_update_ifa: prefixlen should "
1037 			    "be 128 when dstaddr is specified\n"));
1038 			return (EINVAL);
1039 		}
1040 	}
1041 	/* lifetime consistency check */
1042 	lt = &ifra->ifra_lifetime;
1043 	if (lt->ia6t_pltime > lt->ia6t_vltime)
1044 		return (EINVAL);
1045 	if (lt->ia6t_vltime == 0) {
1046 		/*
1047 		 * the following log might be noisy, but this is a typical
1048 		 * configuration mistake or a tool's bug.
1049 		 */
1050 		nd6log((LOG_INFO,
1051 		    "in6_update_ifa: valid lifetime is 0 for %s\n",
1052 		    ip6_sprintf(ip6buf, &ifra->ifra_addr.sin6_addr)));
1053 
1054 		if (ia == NULL)
1055 			return (0); /* there's nothing to do */
1056 	}
1057 
1058 	/* Check prefix mask */
1059 	if (ia != NULL && ifra->ifra_prefixmask.sin6_len != 0) {
1060 		/*
1061 		 * We prohibit changing the prefix length of an existing
1062 		 * address, because
1063 		 * + such an operation should be rare in IPv6, and
1064 		 * + the operation would confuse prefix management.
1065 		 */
1066 		if (ia->ia_prefixmask.sin6_len != 0 &&
1067 		    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) != plen) {
1068 			nd6log((LOG_INFO, "in6_validate_ifa: the prefix length "
1069 			    "of an existing %s address should not be changed\n",
1070 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1071 
1072 			return (EINVAL);
1073 		}
1074 	}
1075 
1076 	return (0);
1077 }
1078 
1079 /*
1080  * Allocate a new ifaddr and link it into chains.
1081  */
1082 static struct in6_ifaddr *
1083 in6_alloc_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra, int flags)
1084 {
1085 	struct in6_ifaddr *ia;
1086 
1087 	/*
1088 	 * When in6_alloc_ifa() is called in a process of a received
1089 	 * RA, it is called under an interrupt context.  So, we should
1090 	 * call malloc with M_NOWAIT.
1091 	 */
1092 	ia = (struct in6_ifaddr *)ifa_alloc(sizeof(*ia), M_NOWAIT);
1093 	if (ia == NULL)
1094 		return (NULL);
1095 	LIST_INIT(&ia->ia6_memberships);
1096 	/* Initialize the address and masks, and put time stamp */
1097 	ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
1098 	ia->ia_addr.sin6_family = AF_INET6;
1099 	ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
1100 	/* XXX: Can we assign ,sin6_addr and skip the rest? */
1101 	ia->ia_addr = ifra->ifra_addr;
1102 	ia->ia6_createtime = time_uptime;
1103 	if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
1104 		/*
1105 		 * Some functions expect that ifa_dstaddr is not
1106 		 * NULL for p2p interfaces.
1107 		 */
1108 		ia->ia_ifa.ifa_dstaddr =
1109 		    (struct sockaddr *)&ia->ia_dstaddr;
1110 	} else {
1111 		ia->ia_ifa.ifa_dstaddr = NULL;
1112 	}
1113 
1114 	/* set prefix mask if any */
1115 	ia->ia_ifa.ifa_netmask = (struct sockaddr *)&ia->ia_prefixmask;
1116 	if (ifra->ifra_prefixmask.sin6_len != 0) {
1117 		ia->ia_prefixmask.sin6_family = AF_INET6;
1118 		ia->ia_prefixmask.sin6_len = ifra->ifra_prefixmask.sin6_len;
1119 		ia->ia_prefixmask.sin6_addr = ifra->ifra_prefixmask.sin6_addr;
1120 	}
1121 
1122 	ia->ia_ifp = ifp;
1123 	ifa_ref(&ia->ia_ifa);			/* if_addrhead */
1124 	IF_ADDR_WLOCK(ifp);
1125 	CK_STAILQ_INSERT_TAIL(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
1126 	IF_ADDR_WUNLOCK(ifp);
1127 
1128 	ifa_ref(&ia->ia_ifa);			/* in6_ifaddrhead */
1129 	IN6_IFADDR_WLOCK();
1130 	CK_STAILQ_INSERT_TAIL(&V_in6_ifaddrhead, ia, ia_link);
1131 	CK_LIST_INSERT_HEAD(IN6ADDR_HASH(&ia->ia_addr.sin6_addr), ia, ia6_hash);
1132 	IN6_IFADDR_WUNLOCK();
1133 
1134 	return (ia);
1135 }
1136 
1137 /*
1138  * Update/configure interface address parameters:
1139  *
1140  * 1) Update lifetime
1141  * 2) Update interface metric ad flags
1142  * 3) Notify other subsystems
1143  */
1144 static int
1145 in6_update_ifa_internal(struct ifnet *ifp, struct in6_aliasreq *ifra,
1146     struct in6_ifaddr *ia, int hostIsNew, int flags)
1147 {
1148 	int error;
1149 
1150 	/* update timestamp */
1151 	ia->ia6_updatetime = time_uptime;
1152 
1153 	/*
1154 	 * Set lifetimes.  We do not refer to ia6t_expire and ia6t_preferred
1155 	 * to see if the address is deprecated or invalidated, but initialize
1156 	 * these members for applications.
1157 	 */
1158 	ia->ia6_lifetime = ifra->ifra_lifetime;
1159 	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1160 		ia->ia6_lifetime.ia6t_expire =
1161 		    time_uptime + ia->ia6_lifetime.ia6t_vltime;
1162 	} else
1163 		ia->ia6_lifetime.ia6t_expire = 0;
1164 	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1165 		ia->ia6_lifetime.ia6t_preferred =
1166 		    time_uptime + ia->ia6_lifetime.ia6t_pltime;
1167 	} else
1168 		ia->ia6_lifetime.ia6t_preferred = 0;
1169 
1170 	/*
1171 	 * backward compatibility - if IN6_IFF_DEPRECATED is set from the
1172 	 * userland, make it deprecated.
1173 	 */
1174 	if ((ifra->ifra_flags & IN6_IFF_DEPRECATED) != 0) {
1175 		ia->ia6_lifetime.ia6t_pltime = 0;
1176 		ia->ia6_lifetime.ia6t_preferred = time_uptime;
1177 	}
1178 
1179 	/*
1180 	 * configure address flags.
1181 	 */
1182 	ia->ia6_flags = ifra->ifra_flags;
1183 
1184 	/*
1185 	 * Make the address tentative before joining multicast addresses,
1186 	 * so that corresponding MLD responses would not have a tentative
1187 	 * source address.
1188 	 */
1189 	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/* safety */
1190 
1191 	/*
1192 	 * DAD should be performed for an new address or addresses on
1193 	 * an interface with ND6_IFF_IFDISABLED.
1194 	 */
1195 	if (in6if_do_dad(ifp) &&
1196 	    (hostIsNew || (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)))
1197 		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1198 
1199 	/* notify other subsystems */
1200 	error = in6_notify_ifa(ifp, ia, ifra, hostIsNew);
1201 
1202 	return (error);
1203 }
1204 
1205 /*
1206  * Do link-level ifa job:
1207  * 1) Add lle entry for added address
1208  * 2) Notifies routing socket users about new address
1209  * 3) join appropriate multicast group
1210  * 4) start DAD if enabled
1211  */
1212 static int
1213 in6_broadcast_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
1214     struct in6_ifaddr *ia, int flags)
1215 {
1216 	struct in6_multi *in6m_sol;
1217 	int error = 0;
1218 
1219 	/* Add local address to lltable, if necessary (ex. on p2p link). */
1220 	if ((error = nd6_add_ifa_lle(ia)) != 0) {
1221 		in6_purgeaddr(&ia->ia_ifa);
1222 		ifa_free(&ia->ia_ifa);
1223 		return (error);
1224 	}
1225 
1226 	/* Join necessary multicast groups. */
1227 	in6m_sol = NULL;
1228 	if ((ifp->if_flags & IFF_MULTICAST) != 0) {
1229 		error = in6_update_ifa_join_mc(ifp, ifra, ia, flags, &in6m_sol);
1230 		if (error != 0) {
1231 			in6_purgeaddr(&ia->ia_ifa);
1232 			ifa_free(&ia->ia_ifa);
1233 			return (error);
1234 		}
1235 	}
1236 
1237 	/* Perform DAD, if the address is TENTATIVE. */
1238 	if ((ia->ia6_flags & IN6_IFF_TENTATIVE)) {
1239 		int delay, mindelay, maxdelay;
1240 
1241 		delay = 0;
1242 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1243 			/*
1244 			 * We need to impose a delay before sending an NS
1245 			 * for DAD.  Check if we also needed a delay for the
1246 			 * corresponding MLD message.  If we did, the delay
1247 			 * should be larger than the MLD delay (this could be
1248 			 * relaxed a bit, but this simple logic is at least
1249 			 * safe).
1250 			 * XXX: Break data hiding guidelines and look at
1251 			 * state for the solicited multicast group.
1252 			 */
1253 			mindelay = 0;
1254 			if (in6m_sol != NULL &&
1255 			    in6m_sol->in6m_state == MLD_REPORTING_MEMBER) {
1256 				mindelay = in6m_sol->in6m_timer;
1257 			}
1258 			maxdelay = MAX_RTR_SOLICITATION_DELAY * hz;
1259 			if (maxdelay - mindelay == 0)
1260 				delay = 0;
1261 			else {
1262 				delay =
1263 				    (arc4random() % (maxdelay - mindelay)) +
1264 				    mindelay;
1265 			}
1266 		}
1267 		nd6_dad_start((struct ifaddr *)ia, delay);
1268 	}
1269 
1270 	in6_newaddrmsg(ia, RTM_ADD);
1271 	ifa_free(&ia->ia_ifa);
1272 	return (error);
1273 }
1274 
1275 void
1276 in6_purgeaddr(struct ifaddr *ifa)
1277 {
1278 	struct ifnet *ifp = ifa->ifa_ifp;
1279 	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1280 	struct in6_multi_mship *imm;
1281 	int plen, error;
1282 
1283 	if (ifa->ifa_carp)
1284 		(*carp_detach_p)(ifa, false);
1285 
1286 	/*
1287 	 * Remove the loopback route to the interface address.
1288 	 * The check for the current setting of "nd6_useloopback"
1289 	 * is not needed.
1290 	 */
1291 	if (ia->ia_flags & IFA_RTSELF) {
1292 		error = ifa_del_loopback_route((struct ifaddr *)ia,
1293 		    (struct sockaddr *)&ia->ia_addr);
1294 		if (error == 0)
1295 			ia->ia_flags &= ~IFA_RTSELF;
1296 	}
1297 
1298 	/* stop DAD processing */
1299 	nd6_dad_stop(ifa);
1300 
1301 	/* Leave multicast groups. */
1302 	while ((imm = LIST_FIRST(&ia->ia6_memberships)) != NULL) {
1303 		LIST_REMOVE(imm, i6mm_chain);
1304 		if (imm->i6mm_maddr != NULL)
1305 			in6_leavegroup(imm->i6mm_maddr, NULL);
1306 		free(imm, M_IP6MADDR);
1307 	}
1308 	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1309 	if ((ia->ia_flags & IFA_ROUTE) && plen == 128) {
1310 		error = rtinit(&(ia->ia_ifa), RTM_DELETE, ia->ia_flags |
1311 		    (ia->ia_dstaddr.sin6_family == AF_INET6 ? RTF_HOST : 0));
1312 		if (error != 0)
1313 			log(LOG_INFO, "%s: err=%d, destination address delete "
1314 			    "failed\n", __func__, error);
1315 		ia->ia_flags &= ~IFA_ROUTE;
1316 	}
1317 
1318 	in6_newaddrmsg(ia, RTM_DELETE);
1319 	in6_unlink_ifa(ia, ifp);
1320 }
1321 
1322 static void
1323 in6_unlink_ifa(struct in6_ifaddr *ia, struct ifnet *ifp)
1324 {
1325 	char ip6buf[INET6_ADDRSTRLEN];
1326 	int remove_lle;
1327 
1328 	IF_ADDR_WLOCK(ifp);
1329 	CK_STAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifaddr, ifa_link);
1330 	IF_ADDR_WUNLOCK(ifp);
1331 	ifa_free(&ia->ia_ifa);			/* if_addrhead */
1332 
1333 	/*
1334 	 * Defer the release of what might be the last reference to the
1335 	 * in6_ifaddr so that it can't be freed before the remainder of the
1336 	 * cleanup.
1337 	 */
1338 	IN6_IFADDR_WLOCK();
1339 	CK_STAILQ_REMOVE(&V_in6_ifaddrhead, ia, in6_ifaddr, ia_link);
1340 	CK_LIST_REMOVE(ia, ia6_hash);
1341 	IN6_IFADDR_WUNLOCK();
1342 
1343 	/*
1344 	 * Release the reference to the base prefix.  There should be a
1345 	 * positive reference.
1346 	 */
1347 	remove_lle = 0;
1348 	if (ia->ia6_ndpr == NULL) {
1349 		nd6log((LOG_NOTICE,
1350 		    "in6_unlink_ifa: autoconf'ed address "
1351 		    "%s has no prefix\n", ip6_sprintf(ip6buf, IA6_IN6(ia))));
1352 	} else {
1353 		ia->ia6_ndpr->ndpr_addrcnt--;
1354 		/* Do not delete lles within prefix if refcont != 0 */
1355 		if (ia->ia6_ndpr->ndpr_addrcnt == 0)
1356 			remove_lle = 1;
1357 		ia->ia6_ndpr = NULL;
1358 	}
1359 
1360 	nd6_rem_ifa_lle(ia, remove_lle);
1361 
1362 	/*
1363 	 * Also, if the address being removed is autoconf'ed, call
1364 	 * pfxlist_onlink_check() since the release might affect the status of
1365 	 * other (detached) addresses.
1366 	 */
1367 	if ((ia->ia6_flags & IN6_IFF_AUTOCONF)) {
1368 		pfxlist_onlink_check();
1369 	}
1370 	ifa_free(&ia->ia_ifa);			/* in6_ifaddrhead */
1371 }
1372 
1373 /*
1374  * Notifies other subsystems about address change/arrival:
1375  * 1) Notifies device handler on the first IPv6 address assignment
1376  * 2) Handle routing table changes for P2P links and route
1377  * 3) Handle routing table changes for address host route
1378  */
1379 static int
1380 in6_notify_ifa(struct ifnet *ifp, struct in6_ifaddr *ia,
1381     struct in6_aliasreq *ifra, int hostIsNew)
1382 {
1383 	int	error = 0, plen, ifacount = 0;
1384 	struct ifaddr *ifa;
1385 	struct sockaddr_in6 *pdst;
1386 	char ip6buf[INET6_ADDRSTRLEN];
1387 
1388 	/*
1389 	 * Give the interface a chance to initialize
1390 	 * if this is its first address,
1391 	 */
1392 	if (hostIsNew != 0) {
1393 		struct epoch_tracker et;
1394 
1395 		NET_EPOCH_ENTER(et);
1396 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1397 			if (ifa->ifa_addr->sa_family != AF_INET6)
1398 				continue;
1399 			ifacount++;
1400 		}
1401 		NET_EPOCH_EXIT(et);
1402 	}
1403 
1404 	if (ifacount <= 1 && ifp->if_ioctl) {
1405 		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
1406 		if (error)
1407 			goto done;
1408 	}
1409 
1410 	/*
1411 	 * If a new destination address is specified, scrub the old one and
1412 	 * install the new destination.  Note that the interface must be
1413 	 * p2p or loopback.
1414 	 */
1415 	pdst = &ifra->ifra_dstaddr;
1416 	if (pdst->sin6_family == AF_INET6 &&
1417 	    !IN6_ARE_ADDR_EQUAL(&pdst->sin6_addr, &ia->ia_dstaddr.sin6_addr)) {
1418 		if ((ia->ia_flags & IFA_ROUTE) != 0 &&
1419 		    (rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST) != 0)) {
1420 			nd6log((LOG_ERR, "in6_update_ifa_internal: failed to "
1421 			    "remove a route to the old destination: %s\n",
1422 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1423 			/* proceed anyway... */
1424 		} else
1425 			ia->ia_flags &= ~IFA_ROUTE;
1426 		ia->ia_dstaddr = *pdst;
1427 	}
1428 
1429 	/*
1430 	 * If a new destination address is specified for a point-to-point
1431 	 * interface, install a route to the destination as an interface
1432 	 * direct route.
1433 	 * XXX: the logic below rejects assigning multiple addresses on a p2p
1434 	 * interface that share the same destination.
1435 	 */
1436 	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1437 	if (!(ia->ia_flags & IFA_ROUTE) && plen == 128 &&
1438 	    ia->ia_dstaddr.sin6_family == AF_INET6) {
1439 		int rtflags = RTF_UP | RTF_HOST;
1440 		/*
1441 		 * Handle the case for ::1 .
1442 		 */
1443 		if (ifp->if_flags & IFF_LOOPBACK)
1444 			ia->ia_flags |= IFA_RTSELF;
1445 		error = rtinit(&ia->ia_ifa, RTM_ADD, ia->ia_flags | rtflags);
1446 		if (error)
1447 			goto done;
1448 		ia->ia_flags |= IFA_ROUTE;
1449 	}
1450 
1451 	/*
1452 	 * add a loopback route to self if not exists
1453 	 */
1454 	if (!(ia->ia_flags & IFA_RTSELF) && V_nd6_useloopback) {
1455 		error = ifa_add_loopback_route((struct ifaddr *)ia,
1456 		    (struct sockaddr *)&ia->ia_addr);
1457 		if (error == 0)
1458 			ia->ia_flags |= IFA_RTSELF;
1459 	}
1460 done:
1461 	WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
1462 	    "Invoking IPv6 network device address event may sleep");
1463 
1464 	ifa_ref(&ia->ia_ifa);
1465 	EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, &ia->ia_ifa,
1466 	    IFADDR_EVENT_ADD);
1467 	ifa_free(&ia->ia_ifa);
1468 
1469 	return (error);
1470 }
1471 
1472 /*
1473  * Find an IPv6 interface link-local address specific to an interface.
1474  * ifaddr is returned referenced.
1475  */
1476 struct in6_ifaddr *
1477 in6ifa_ifpforlinklocal(struct ifnet *ifp, int ignoreflags)
1478 {
1479 	struct ifaddr *ifa;
1480 
1481 	NET_EPOCH_ASSERT();
1482 
1483 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1484 		if (ifa->ifa_addr->sa_family != AF_INET6)
1485 			continue;
1486 		if (IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa))) {
1487 			if ((((struct in6_ifaddr *)ifa)->ia6_flags &
1488 			    ignoreflags) != 0)
1489 				continue;
1490 			ifa_ref(ifa);
1491 			break;
1492 		}
1493 	}
1494 
1495 	return ((struct in6_ifaddr *)ifa);
1496 }
1497 
1498 /*
1499  * find the interface address corresponding to a given IPv6 address.
1500  * ifaddr is returned referenced.
1501  */
1502 struct in6_ifaddr *
1503 in6ifa_ifwithaddr(const struct in6_addr *addr, uint32_t zoneid)
1504 {
1505 	struct rm_priotracker in6_ifa_tracker;
1506 	struct in6_ifaddr *ia;
1507 
1508 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1509 	CK_LIST_FOREACH(ia, IN6ADDR_HASH(addr), ia6_hash) {
1510 		if (IN6_ARE_ADDR_EQUAL(IA6_IN6(ia), addr)) {
1511 			if (zoneid != 0 &&
1512 			    zoneid != ia->ia_addr.sin6_scope_id)
1513 				continue;
1514 			ifa_ref(&ia->ia_ifa);
1515 			break;
1516 		}
1517 	}
1518 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1519 	return (ia);
1520 }
1521 
1522 /*
1523  * find the internet address corresponding to a given interface and address.
1524  * ifaddr is returned referenced.
1525  */
1526 struct in6_ifaddr *
1527 in6ifa_ifpwithaddr(struct ifnet *ifp, const struct in6_addr *addr)
1528 {
1529 	struct epoch_tracker et;
1530 	struct ifaddr *ifa;
1531 
1532 	NET_EPOCH_ENTER(et);
1533 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1534 		if (ifa->ifa_addr->sa_family != AF_INET6)
1535 			continue;
1536 		if (IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa))) {
1537 			ifa_ref(ifa);
1538 			break;
1539 		}
1540 	}
1541 	NET_EPOCH_EXIT(et);
1542 
1543 	return ((struct in6_ifaddr *)ifa);
1544 }
1545 
1546 /*
1547  * Find a link-local scoped address on ifp and return it if any.
1548  */
1549 struct in6_ifaddr *
1550 in6ifa_llaonifp(struct ifnet *ifp)
1551 {
1552 	struct epoch_tracker et;
1553 	struct sockaddr_in6 *sin6;
1554 	struct ifaddr *ifa;
1555 
1556 	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)
1557 		return (NULL);
1558 	NET_EPOCH_ENTER(et);
1559 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1560 		if (ifa->ifa_addr->sa_family != AF_INET6)
1561 			continue;
1562 		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
1563 		if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr) ||
1564 		    IN6_IS_ADDR_MC_INTFACELOCAL(&sin6->sin6_addr) ||
1565 		    IN6_IS_ADDR_MC_NODELOCAL(&sin6->sin6_addr))
1566 			break;
1567 	}
1568 	NET_EPOCH_EXIT(et);
1569 
1570 	return ((struct in6_ifaddr *)ifa);
1571 }
1572 
1573 /*
1574  * Convert IP6 address to printable (loggable) representation. Caller
1575  * has to make sure that ip6buf is at least INET6_ADDRSTRLEN long.
1576  */
1577 static char digits[] = "0123456789abcdef";
1578 char *
1579 ip6_sprintf(char *ip6buf, const struct in6_addr *addr)
1580 {
1581 	int i, cnt = 0, maxcnt = 0, idx = 0, index = 0;
1582 	char *cp;
1583 	const u_int16_t *a = (const u_int16_t *)addr;
1584 	const u_int8_t *d;
1585 	int dcolon = 0, zero = 0;
1586 
1587 	cp = ip6buf;
1588 
1589 	for (i = 0; i < 8; i++) {
1590 		if (*(a + i) == 0) {
1591 			cnt++;
1592 			if (cnt == 1)
1593 				idx = i;
1594 		}
1595 		else if (maxcnt < cnt) {
1596 			maxcnt = cnt;
1597 			index = idx;
1598 			cnt = 0;
1599 		}
1600 	}
1601 	if (maxcnt < cnt) {
1602 		maxcnt = cnt;
1603 		index = idx;
1604 	}
1605 
1606 	for (i = 0; i < 8; i++) {
1607 		if (dcolon == 1) {
1608 			if (*a == 0) {
1609 				if (i == 7)
1610 					*cp++ = ':';
1611 				a++;
1612 				continue;
1613 			} else
1614 				dcolon = 2;
1615 		}
1616 		if (*a == 0) {
1617 			if (dcolon == 0 && *(a + 1) == 0 && i == index) {
1618 				if (i == 0)
1619 					*cp++ = ':';
1620 				*cp++ = ':';
1621 				dcolon = 1;
1622 			} else {
1623 				*cp++ = '0';
1624 				*cp++ = ':';
1625 			}
1626 			a++;
1627 			continue;
1628 		}
1629 		d = (const u_char *)a;
1630 		/* Try to eliminate leading zeros in printout like in :0001. */
1631 		zero = 1;
1632 		*cp = digits[*d >> 4];
1633 		if (*cp != '0') {
1634 			zero = 0;
1635 			cp++;
1636 		}
1637 		*cp = digits[*d++ & 0xf];
1638 		if (zero == 0 || (*cp != '0')) {
1639 			zero = 0;
1640 			cp++;
1641 		}
1642 		*cp = digits[*d >> 4];
1643 		if (zero == 0 || (*cp != '0')) {
1644 			zero = 0;
1645 			cp++;
1646 		}
1647 		*cp++ = digits[*d & 0xf];
1648 		*cp++ = ':';
1649 		a++;
1650 	}
1651 	*--cp = '\0';
1652 	return (ip6buf);
1653 }
1654 
1655 int
1656 in6_localaddr(struct in6_addr *in6)
1657 {
1658 	struct rm_priotracker in6_ifa_tracker;
1659 	struct in6_ifaddr *ia;
1660 
1661 	if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
1662 		return 1;
1663 
1664 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1665 	CK_STAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
1666 		if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
1667 		    &ia->ia_prefixmask.sin6_addr)) {
1668 			IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1669 			return 1;
1670 		}
1671 	}
1672 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1673 
1674 	return (0);
1675 }
1676 
1677 /*
1678  * Return 1 if an internet address is for the local host and configured
1679  * on one of its interfaces.
1680  */
1681 int
1682 in6_localip(struct in6_addr *in6)
1683 {
1684 	struct rm_priotracker in6_ifa_tracker;
1685 	struct in6_ifaddr *ia;
1686 
1687 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1688 	CK_LIST_FOREACH(ia, IN6ADDR_HASH(in6), ia6_hash) {
1689 		if (IN6_ARE_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr)) {
1690 			IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1691 			return (1);
1692 		}
1693 	}
1694 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1695 	return (0);
1696 }
1697 
1698 /*
1699  * Return 1 if an internet address is configured on an interface.
1700  */
1701 int
1702 in6_ifhasaddr(struct ifnet *ifp, struct in6_addr *addr)
1703 {
1704 	struct in6_addr in6;
1705 	struct ifaddr *ifa;
1706 	struct in6_ifaddr *ia6;
1707 
1708 	NET_EPOCH_ASSERT();
1709 
1710 	in6 = *addr;
1711 	if (in6_clearscope(&in6))
1712 		return (0);
1713 	in6_setscope(&in6, ifp, NULL);
1714 
1715 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1716 		if (ifa->ifa_addr->sa_family != AF_INET6)
1717 			continue;
1718 		ia6 = (struct in6_ifaddr *)ifa;
1719 		if (IN6_ARE_ADDR_EQUAL(&ia6->ia_addr.sin6_addr, &in6))
1720 			return (1);
1721 	}
1722 
1723 	return (0);
1724 }
1725 
1726 int
1727 in6_is_addr_deprecated(struct sockaddr_in6 *sa6)
1728 {
1729 	struct rm_priotracker in6_ifa_tracker;
1730 	struct in6_ifaddr *ia;
1731 
1732 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1733 	CK_LIST_FOREACH(ia, IN6ADDR_HASH(&sa6->sin6_addr), ia6_hash) {
1734 		if (IN6_ARE_ADDR_EQUAL(IA6_IN6(ia), &sa6->sin6_addr)) {
1735 			if (ia->ia6_flags & IN6_IFF_DEPRECATED) {
1736 				IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1737 				return (1); /* true */
1738 			}
1739 			break;
1740 		}
1741 	}
1742 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1743 
1744 	return (0);		/* false */
1745 }
1746 
1747 /*
1748  * return length of part which dst and src are equal
1749  * hard coding...
1750  */
1751 int
1752 in6_matchlen(struct in6_addr *src, struct in6_addr *dst)
1753 {
1754 	int match = 0;
1755 	u_char *s = (u_char *)src, *d = (u_char *)dst;
1756 	u_char *lim = s + 16, r;
1757 
1758 	while (s < lim)
1759 		if ((r = (*d++ ^ *s++)) != 0) {
1760 			while (r < 128) {
1761 				match++;
1762 				r <<= 1;
1763 			}
1764 			break;
1765 		} else
1766 			match += 8;
1767 	return match;
1768 }
1769 
1770 /* XXX: to be scope conscious */
1771 int
1772 in6_are_prefix_equal(struct in6_addr *p1, struct in6_addr *p2, int len)
1773 {
1774 	int bytelen, bitlen;
1775 
1776 	/* sanity check */
1777 	if (0 > len || len > 128) {
1778 		log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n",
1779 		    len);
1780 		return (0);
1781 	}
1782 
1783 	bytelen = len / 8;
1784 	bitlen = len % 8;
1785 
1786 	if (bcmp(&p1->s6_addr, &p2->s6_addr, bytelen))
1787 		return (0);
1788 	if (bitlen != 0 &&
1789 	    p1->s6_addr[bytelen] >> (8 - bitlen) !=
1790 	    p2->s6_addr[bytelen] >> (8 - bitlen))
1791 		return (0);
1792 
1793 	return (1);
1794 }
1795 
1796 void
1797 in6_prefixlen2mask(struct in6_addr *maskp, int len)
1798 {
1799 	u_char maskarray[8] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
1800 	int bytelen, bitlen, i;
1801 
1802 	/* sanity check */
1803 	if (0 > len || len > 128) {
1804 		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
1805 		    len);
1806 		return;
1807 	}
1808 
1809 	bzero(maskp, sizeof(*maskp));
1810 	bytelen = len / 8;
1811 	bitlen = len % 8;
1812 	for (i = 0; i < bytelen; i++)
1813 		maskp->s6_addr[i] = 0xff;
1814 	if (bitlen)
1815 		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
1816 }
1817 
1818 /*
1819  * return the best address out of the same scope. if no address was
1820  * found, return the first valid address from designated IF.
1821  */
1822 struct in6_ifaddr *
1823 in6_ifawithifp(struct ifnet *ifp, struct in6_addr *dst)
1824 {
1825 	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
1826 	struct ifaddr *ifa;
1827 	struct in6_ifaddr *besta = NULL;
1828 	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
1829 
1830 	NET_EPOCH_ASSERT();
1831 
1832 	dep[0] = dep[1] = NULL;
1833 
1834 	/*
1835 	 * We first look for addresses in the same scope.
1836 	 * If there is one, return it.
1837 	 * If two or more, return one which matches the dst longest.
1838 	 * If none, return one of global addresses assigned other ifs.
1839 	 */
1840 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1841 		if (ifa->ifa_addr->sa_family != AF_INET6)
1842 			continue;
1843 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
1844 			continue; /* XXX: is there any case to allow anycast? */
1845 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
1846 			continue; /* don't use this interface */
1847 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
1848 			continue;
1849 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
1850 			if (V_ip6_use_deprecated)
1851 				dep[0] = (struct in6_ifaddr *)ifa;
1852 			continue;
1853 		}
1854 
1855 		if (dst_scope == in6_addrscope(IFA_IN6(ifa))) {
1856 			/*
1857 			 * call in6_matchlen() as few as possible
1858 			 */
1859 			if (besta) {
1860 				if (blen == -1)
1861 					blen = in6_matchlen(&besta->ia_addr.sin6_addr, dst);
1862 				tlen = in6_matchlen(IFA_IN6(ifa), dst);
1863 				if (tlen > blen) {
1864 					blen = tlen;
1865 					besta = (struct in6_ifaddr *)ifa;
1866 				}
1867 			} else
1868 				besta = (struct in6_ifaddr *)ifa;
1869 		}
1870 	}
1871 	if (besta) {
1872 		ifa_ref(&besta->ia_ifa);
1873 		return (besta);
1874 	}
1875 
1876 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1877 		if (ifa->ifa_addr->sa_family != AF_INET6)
1878 			continue;
1879 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
1880 			continue; /* XXX: is there any case to allow anycast? */
1881 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
1882 			continue; /* don't use this interface */
1883 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
1884 			continue;
1885 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
1886 			if (V_ip6_use_deprecated)
1887 				dep[1] = (struct in6_ifaddr *)ifa;
1888 			continue;
1889 		}
1890 
1891 		if (ifa != NULL)
1892 			ifa_ref(ifa);
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 		return dep[0];
1900 	}
1901 	if (dep[1]) {
1902 		ifa_ref((struct ifaddr *)dep[1]);
1903 		return dep[1];
1904 	}
1905 
1906 	return NULL;
1907 }
1908 
1909 /*
1910  * perform DAD when interface becomes IFF_UP.
1911  */
1912 void
1913 in6_if_up(struct ifnet *ifp)
1914 {
1915 	struct epoch_tracker et;
1916 	struct ifaddr *ifa;
1917 	struct in6_ifaddr *ia;
1918 
1919 	NET_EPOCH_ENTER(et);
1920 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1921 		if (ifa->ifa_addr->sa_family != AF_INET6)
1922 			continue;
1923 		ia = (struct in6_ifaddr *)ifa;
1924 		if (ia->ia6_flags & IN6_IFF_TENTATIVE) {
1925 			/*
1926 			 * The TENTATIVE flag was likely set by hand
1927 			 * beforehand, implicitly indicating the need for DAD.
1928 			 * We may be able to skip the random delay in this
1929 			 * case, but we impose delays just in case.
1930 			 */
1931 			nd6_dad_start(ifa,
1932 			    arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz));
1933 		}
1934 	}
1935 	NET_EPOCH_EXIT(et);
1936 
1937 	/*
1938 	 * special cases, like 6to4, are handled in in6_ifattach
1939 	 */
1940 	in6_ifattach(ifp, NULL);
1941 }
1942 
1943 int
1944 in6if_do_dad(struct ifnet *ifp)
1945 {
1946 
1947 	if ((ifp->if_flags & IFF_LOOPBACK) != 0)
1948 		return (0);
1949 	if ((ifp->if_flags & IFF_MULTICAST) == 0)
1950 		return (0);
1951 	if ((ND_IFINFO(ifp)->flags &
1952 	    (ND6_IFF_IFDISABLED | ND6_IFF_NO_DAD)) != 0)
1953 		return (0);
1954 	return (1);
1955 }
1956 
1957 /*
1958  * Calculate max IPv6 MTU through all the interfaces and store it
1959  * to in6_maxmtu.
1960  */
1961 void
1962 in6_setmaxmtu(void)
1963 {
1964 	struct epoch_tracker et;
1965 	unsigned long maxmtu = 0;
1966 	struct ifnet *ifp;
1967 
1968 	NET_EPOCH_ENTER(et);
1969 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1970 		/* this function can be called during ifnet initialization */
1971 		if (!ifp->if_afdata[AF_INET6])
1972 			continue;
1973 		if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
1974 		    IN6_LINKMTU(ifp) > maxmtu)
1975 			maxmtu = IN6_LINKMTU(ifp);
1976 	}
1977 	NET_EPOCH_EXIT(et);
1978 	if (maxmtu)	/* update only when maxmtu is positive */
1979 		V_in6_maxmtu = maxmtu;
1980 }
1981 
1982 /*
1983  * Provide the length of interface identifiers to be used for the link attached
1984  * to the given interface.  The length should be defined in "IPv6 over
1985  * xxx-link" document.  Note that address architecture might also define
1986  * the length for a particular set of address prefixes, regardless of the
1987  * link type.  As clarified in rfc2462bis, those two definitions should be
1988  * consistent, and those really are as of August 2004.
1989  */
1990 int
1991 in6_if2idlen(struct ifnet *ifp)
1992 {
1993 	switch (ifp->if_type) {
1994 	case IFT_ETHER:		/* RFC2464 */
1995 	case IFT_PROPVIRTUAL:	/* XXX: no RFC. treat it as ether */
1996 	case IFT_L2VLAN:	/* ditto */
1997 	case IFT_BRIDGE:	/* bridge(4) only does Ethernet-like links */
1998 	case IFT_INFINIBAND:
1999 		return (64);
2000 	case IFT_PPP:		/* RFC2472 */
2001 		return (64);
2002 	case IFT_FRELAY:	/* RFC2590 */
2003 		return (64);
2004 	case IFT_IEEE1394:	/* RFC3146 */
2005 		return (64);
2006 	case IFT_GIF:
2007 		return (64);	/* draft-ietf-v6ops-mech-v2-07 */
2008 	case IFT_LOOP:
2009 		return (64);	/* XXX: is this really correct? */
2010 	default:
2011 		/*
2012 		 * Unknown link type:
2013 		 * It might be controversial to use the today's common constant
2014 		 * of 64 for these cases unconditionally.  For full compliance,
2015 		 * we should return an error in this case.  On the other hand,
2016 		 * if we simply miss the standard for the link type or a new
2017 		 * standard is defined for a new link type, the IFID length
2018 		 * is very likely to be the common constant.  As a compromise,
2019 		 * we always use the constant, but make an explicit notice
2020 		 * indicating the "unknown" case.
2021 		 */
2022 		printf("in6_if2idlen: unknown link type (%d)\n", ifp->if_type);
2023 		return (64);
2024 	}
2025 }
2026 
2027 struct in6_llentry {
2028 	struct llentry		base;
2029 };
2030 
2031 #define	IN6_LLTBL_DEFAULT_HSIZE	32
2032 #define	IN6_LLTBL_HASH(k, h) \
2033 	(((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1))
2034 
2035 /*
2036  * Do actual deallocation of @lle.
2037  */
2038 static void
2039 in6_lltable_destroy_lle_unlocked(epoch_context_t ctx)
2040 {
2041 	struct llentry *lle;
2042 
2043 	lle = __containerof(ctx, struct llentry, lle_epoch_ctx);
2044 	LLE_LOCK_DESTROY(lle);
2045 	LLE_REQ_DESTROY(lle);
2046 	free(lle, M_LLTABLE);
2047 }
2048 
2049 /*
2050  * Called by LLE_FREE_LOCKED when number of references
2051  * drops to zero.
2052  */
2053 static void
2054 in6_lltable_destroy_lle(struct llentry *lle)
2055 {
2056 
2057 	LLE_WUNLOCK(lle);
2058 	NET_EPOCH_CALL(in6_lltable_destroy_lle_unlocked, &lle->lle_epoch_ctx);
2059 }
2060 
2061 static struct llentry *
2062 in6_lltable_new(const struct in6_addr *addr6, u_int flags)
2063 {
2064 	struct in6_llentry *lle;
2065 
2066 	lle = malloc(sizeof(struct in6_llentry), M_LLTABLE, M_NOWAIT | M_ZERO);
2067 	if (lle == NULL)		/* NB: caller generates msg */
2068 		return NULL;
2069 
2070 	lle->base.r_l3addr.addr6 = *addr6;
2071 	lle->base.lle_refcnt = 1;
2072 	lle->base.lle_free = in6_lltable_destroy_lle;
2073 	LLE_LOCK_INIT(&lle->base);
2074 	LLE_REQ_INIT(&lle->base);
2075 	callout_init(&lle->base.lle_timer, 1);
2076 
2077 	return (&lle->base);
2078 }
2079 
2080 static int
2081 in6_lltable_match_prefix(const struct sockaddr *saddr,
2082     const struct sockaddr *smask, u_int flags, struct llentry *lle)
2083 {
2084 	const struct in6_addr *addr, *mask, *lle_addr;
2085 
2086 	addr = &((const struct sockaddr_in6 *)saddr)->sin6_addr;
2087 	mask = &((const struct sockaddr_in6 *)smask)->sin6_addr;
2088 	lle_addr = &lle->r_l3addr.addr6;
2089 
2090 	if (IN6_ARE_MASKED_ADDR_EQUAL(lle_addr, addr, mask) == 0)
2091 		return (0);
2092 
2093 	if (lle->la_flags & LLE_IFADDR) {
2094 		/*
2095 		 * Delete LLE_IFADDR records IFF address & flag matches.
2096 		 * Note that addr is the interface address within prefix
2097 		 * being matched.
2098 		 */
2099 		if (IN6_ARE_ADDR_EQUAL(addr, lle_addr) &&
2100 		    (flags & LLE_STATIC) != 0)
2101 			return (1);
2102 		return (0);
2103 	}
2104 
2105 	/* flags & LLE_STATIC means deleting both dynamic and static entries */
2106 	if ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC))
2107 		return (1);
2108 
2109 	return (0);
2110 }
2111 
2112 static void
2113 in6_lltable_free_entry(struct lltable *llt, struct llentry *lle)
2114 {
2115 	struct ifnet *ifp;
2116 
2117 	LLE_WLOCK_ASSERT(lle);
2118 	KASSERT(llt != NULL, ("lltable is NULL"));
2119 
2120 	/* Unlink entry from table */
2121 	if ((lle->la_flags & LLE_LINKED) != 0) {
2122 		ifp = llt->llt_ifp;
2123 		IF_AFDATA_WLOCK_ASSERT(ifp);
2124 		lltable_unlink_entry(llt, lle);
2125 	}
2126 
2127 	llentry_free(lle);
2128 }
2129 
2130 static int
2131 in6_lltable_rtcheck(struct ifnet *ifp,
2132 		    u_int flags,
2133 		    const struct sockaddr *l3addr)
2134 {
2135 	const struct sockaddr_in6 *sin6;
2136 	struct nhop_object *nh;
2137 	struct in6_addr dst;
2138 	uint32_t scopeid;
2139 	char ip6buf[INET6_ADDRSTRLEN];
2140 	int fibnum;
2141 
2142 	NET_EPOCH_ASSERT();
2143 	KASSERT(l3addr->sa_family == AF_INET6,
2144 	    ("sin_family %d", l3addr->sa_family));
2145 
2146 	sin6 = (const struct sockaddr_in6 *)l3addr;
2147 	in6_splitscope(&sin6->sin6_addr, &dst, &scopeid);
2148 	fibnum = V_rt_add_addr_allfibs ? RT_DEFAULT_FIB : ifp->if_fib;
2149 	nh = fib6_lookup(fibnum, &dst, scopeid, NHR_NONE, 0);
2150 	if (nh && ((nh->nh_flags & NHF_GATEWAY) || nh->nh_ifp != ifp)) {
2151 		struct ifaddr *ifa;
2152 		/*
2153 		 * Create an ND6 cache for an IPv6 neighbor
2154 		 * that is not covered by our own prefix.
2155 		 */
2156 		ifa = ifaof_ifpforaddr(l3addr, ifp);
2157 		if (ifa != NULL) {
2158 			return 0;
2159 		}
2160 		log(LOG_INFO, "IPv6 address: \"%s\" is not on the network\n",
2161 		    ip6_sprintf(ip6buf, &sin6->sin6_addr));
2162 		return EINVAL;
2163 	}
2164 	return 0;
2165 }
2166 
2167 /*
2168  * Called by the datapath to indicate that the entry was used.
2169  */
2170 static void
2171 in6_lltable_mark_used(struct llentry *lle)
2172 {
2173 
2174 	LLE_REQ_LOCK(lle);
2175 	lle->r_skip_req = 0;
2176 
2177 	/*
2178 	 * Set the hit time so the callback function
2179 	 * can determine the remaining time before
2180 	 * transiting to the DELAY state.
2181 	 */
2182 	lle->lle_hittime = time_uptime;
2183 	LLE_REQ_UNLOCK(lle);
2184 }
2185 
2186 static inline uint32_t
2187 in6_lltable_hash_dst(const struct in6_addr *dst, uint32_t hsize)
2188 {
2189 
2190 	return (IN6_LLTBL_HASH(dst->s6_addr32[3], hsize));
2191 }
2192 
2193 static uint32_t
2194 in6_lltable_hash(const struct llentry *lle, uint32_t hsize)
2195 {
2196 
2197 	return (in6_lltable_hash_dst(&lle->r_l3addr.addr6, hsize));
2198 }
2199 
2200 static void
2201 in6_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
2202 {
2203 	struct sockaddr_in6 *sin6;
2204 
2205 	sin6 = (struct sockaddr_in6 *)sa;
2206 	bzero(sin6, sizeof(*sin6));
2207 	sin6->sin6_family = AF_INET6;
2208 	sin6->sin6_len = sizeof(*sin6);
2209 	sin6->sin6_addr = lle->r_l3addr.addr6;
2210 }
2211 
2212 static inline struct llentry *
2213 in6_lltable_find_dst(struct lltable *llt, const struct in6_addr *dst)
2214 {
2215 	struct llentry *lle;
2216 	struct llentries *lleh;
2217 	u_int hashidx;
2218 
2219 	hashidx = in6_lltable_hash_dst(dst, llt->llt_hsize);
2220 	lleh = &llt->lle_head[hashidx];
2221 	CK_LIST_FOREACH(lle, lleh, lle_next) {
2222 		if (lle->la_flags & LLE_DELETED)
2223 			continue;
2224 		if (IN6_ARE_ADDR_EQUAL(&lle->r_l3addr.addr6, dst))
2225 			break;
2226 	}
2227 
2228 	return (lle);
2229 }
2230 
2231 static void
2232 in6_lltable_delete_entry(struct lltable *llt, struct llentry *lle)
2233 {
2234 
2235 	lle->la_flags |= LLE_DELETED;
2236 	EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_DELETED);
2237 #ifdef DIAGNOSTIC
2238 	log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle);
2239 #endif
2240 	llentry_free(lle);
2241 }
2242 
2243 static struct llentry *
2244 in6_lltable_alloc(struct lltable *llt, u_int flags,
2245 	const struct sockaddr *l3addr)
2246 {
2247 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2248 	struct ifnet *ifp = llt->llt_ifp;
2249 	struct llentry *lle;
2250 	char linkhdr[LLE_MAX_LINKHDR];
2251 	size_t linkhdrsize;
2252 	int lladdr_off;
2253 
2254 	KASSERT(l3addr->sa_family == AF_INET6,
2255 	    ("sin_family %d", l3addr->sa_family));
2256 
2257 	/*
2258 	 * A route that covers the given address must have
2259 	 * been installed 1st because we are doing a resolution,
2260 	 * verify this.
2261 	 */
2262 	if (!(flags & LLE_IFADDR) &&
2263 	    in6_lltable_rtcheck(ifp, flags, l3addr) != 0)
2264 		return (NULL);
2265 
2266 	lle = in6_lltable_new(&sin6->sin6_addr, flags);
2267 	if (lle == NULL) {
2268 		log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
2269 		return (NULL);
2270 	}
2271 	lle->la_flags = flags;
2272 	if ((flags & LLE_IFADDR) == LLE_IFADDR) {
2273 		linkhdrsize = LLE_MAX_LINKHDR;
2274 		if (lltable_calc_llheader(ifp, AF_INET6, IF_LLADDR(ifp),
2275 		    linkhdr, &linkhdrsize, &lladdr_off) != 0) {
2276 			NET_EPOCH_CALL(in6_lltable_destroy_lle_unlocked, &lle->lle_epoch_ctx);
2277 			return (NULL);
2278 		}
2279 		lltable_set_entry_addr(ifp, lle, linkhdr, linkhdrsize,
2280 		    lladdr_off);
2281 		lle->la_flags |= LLE_STATIC;
2282 	}
2283 
2284 	if ((lle->la_flags & LLE_STATIC) != 0)
2285 		lle->ln_state = ND6_LLINFO_REACHABLE;
2286 
2287 	return (lle);
2288 }
2289 
2290 static struct llentry *
2291 in6_lltable_lookup(struct lltable *llt, u_int flags,
2292 	const struct sockaddr *l3addr)
2293 {
2294 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2295 	struct llentry *lle;
2296 
2297 	IF_AFDATA_LOCK_ASSERT(llt->llt_ifp);
2298 	KASSERT(l3addr->sa_family == AF_INET6,
2299 	    ("sin_family %d", l3addr->sa_family));
2300 	KASSERT((flags & (LLE_UNLOCKED | LLE_EXCLUSIVE)) !=
2301 	    (LLE_UNLOCKED | LLE_EXCLUSIVE),
2302 	    ("wrong lle request flags: %#x", flags));
2303 
2304 	lle = in6_lltable_find_dst(llt, &sin6->sin6_addr);
2305 	if (lle == NULL)
2306 		return (NULL);
2307 	if (flags & LLE_UNLOCKED)
2308 		return (lle);
2309 
2310 	if (flags & LLE_EXCLUSIVE)
2311 		LLE_WLOCK(lle);
2312 	else
2313 		LLE_RLOCK(lle);
2314 
2315 	/*
2316 	 * If the afdata lock is not held, the LLE may have been unlinked while
2317 	 * we were blocked on the LLE lock.  Check for this case.
2318 	 */
2319 	if (__predict_false((lle->la_flags & LLE_LINKED) == 0)) {
2320 		if (flags & LLE_EXCLUSIVE)
2321 			LLE_WUNLOCK(lle);
2322 		else
2323 			LLE_RUNLOCK(lle);
2324 		return (NULL);
2325 	}
2326 	return (lle);
2327 }
2328 
2329 static int
2330 in6_lltable_dump_entry(struct lltable *llt, struct llentry *lle,
2331     struct sysctl_req *wr)
2332 {
2333 	struct ifnet *ifp = llt->llt_ifp;
2334 	/* XXX stack use */
2335 	struct {
2336 		struct rt_msghdr	rtm;
2337 		struct sockaddr_in6	sin6;
2338 		/*
2339 		 * ndp.c assumes that sdl is word aligned
2340 		 */
2341 #ifdef __LP64__
2342 		uint32_t		pad;
2343 #endif
2344 		struct sockaddr_dl	sdl;
2345 	} ndpc;
2346 	struct sockaddr_dl *sdl;
2347 	int error;
2348 
2349 	bzero(&ndpc, sizeof(ndpc));
2350 	/* skip deleted entries */
2351 	if ((lle->la_flags & LLE_DELETED) == LLE_DELETED)
2352 		return (0);
2353 	/* Skip if jailed and not a valid IP of the prison. */
2354 	lltable_fill_sa_entry(lle, (struct sockaddr *)&ndpc.sin6);
2355 	if (prison_if(wr->td->td_ucred, (struct sockaddr *)&ndpc.sin6) != 0)
2356 		return (0);
2357 	/*
2358 	 * produce a msg made of:
2359 	 *  struct rt_msghdr;
2360 	 *  struct sockaddr_in6 (IPv6)
2361 	 *  struct sockaddr_dl;
2362 	 */
2363 	ndpc.rtm.rtm_msglen = sizeof(ndpc);
2364 	ndpc.rtm.rtm_version = RTM_VERSION;
2365 	ndpc.rtm.rtm_type = RTM_GET;
2366 	ndpc.rtm.rtm_flags = RTF_UP;
2367 	ndpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
2368 	sa6_recoverscope(&ndpc.sin6);
2369 
2370 	/* publish */
2371 	if (lle->la_flags & LLE_PUB)
2372 		ndpc.rtm.rtm_flags |= RTF_ANNOUNCE;
2373 
2374 	sdl = &ndpc.sdl;
2375 	sdl->sdl_family = AF_LINK;
2376 	sdl->sdl_len = sizeof(*sdl);
2377 	sdl->sdl_index = ifp->if_index;
2378 	sdl->sdl_type = ifp->if_type;
2379 	if ((lle->la_flags & LLE_VALID) == LLE_VALID) {
2380 		sdl->sdl_alen = ifp->if_addrlen;
2381 		bcopy(lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
2382 	} else {
2383 		sdl->sdl_alen = 0;
2384 		bzero(LLADDR(sdl), ifp->if_addrlen);
2385 	}
2386 	if (lle->la_expire != 0)
2387 		ndpc.rtm.rtm_rmx.rmx_expire = lle->la_expire +
2388 		    lle->lle_remtime / hz + time_second - time_uptime;
2389 	ndpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
2390 	if (lle->la_flags & LLE_STATIC)
2391 		ndpc.rtm.rtm_flags |= RTF_STATIC;
2392 	if (lle->la_flags & LLE_IFADDR)
2393 		ndpc.rtm.rtm_flags |= RTF_PINNED;
2394 	if (lle->ln_router != 0)
2395 		ndpc.rtm.rtm_flags |= RTF_GATEWAY;
2396 	ndpc.rtm.rtm_rmx.rmx_pksent = lle->la_asked;
2397 	/* Store state in rmx_weight value */
2398 	ndpc.rtm.rtm_rmx.rmx_state = lle->ln_state;
2399 	ndpc.rtm.rtm_index = ifp->if_index;
2400 	error = SYSCTL_OUT(wr, &ndpc, sizeof(ndpc));
2401 
2402 	return (error);
2403 }
2404 
2405 static struct lltable *
2406 in6_lltattach(struct ifnet *ifp)
2407 {
2408 	struct lltable *llt;
2409 
2410 	llt = lltable_allocate_htbl(IN6_LLTBL_DEFAULT_HSIZE);
2411 	llt->llt_af = AF_INET6;
2412 	llt->llt_ifp = ifp;
2413 
2414 	llt->llt_lookup = in6_lltable_lookup;
2415 	llt->llt_alloc_entry = in6_lltable_alloc;
2416 	llt->llt_delete_entry = in6_lltable_delete_entry;
2417 	llt->llt_dump_entry = in6_lltable_dump_entry;
2418 	llt->llt_hash = in6_lltable_hash;
2419 	llt->llt_fill_sa_entry = in6_lltable_fill_sa_entry;
2420 	llt->llt_free_entry = in6_lltable_free_entry;
2421 	llt->llt_match_prefix = in6_lltable_match_prefix;
2422 	llt->llt_mark_used = in6_lltable_mark_used;
2423  	lltable_link(llt);
2424 
2425 	return (llt);
2426 }
2427 
2428 void *
2429 in6_domifattach(struct ifnet *ifp)
2430 {
2431 	struct in6_ifextra *ext;
2432 
2433 	/* There are not IPv6-capable interfaces. */
2434 	switch (ifp->if_type) {
2435 	case IFT_PFLOG:
2436 	case IFT_PFSYNC:
2437 	case IFT_USB:
2438 		return (NULL);
2439 	}
2440 	ext = (struct in6_ifextra *)malloc(sizeof(*ext), M_IFADDR, M_WAITOK);
2441 	bzero(ext, sizeof(*ext));
2442 
2443 	ext->in6_ifstat = malloc(sizeof(counter_u64_t) *
2444 	    sizeof(struct in6_ifstat) / sizeof(uint64_t), M_IFADDR, M_WAITOK);
2445 	COUNTER_ARRAY_ALLOC(ext->in6_ifstat,
2446 	    sizeof(struct in6_ifstat) / sizeof(uint64_t), M_WAITOK);
2447 
2448 	ext->icmp6_ifstat = malloc(sizeof(counter_u64_t) *
2449 	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t), M_IFADDR,
2450 	    M_WAITOK);
2451 	COUNTER_ARRAY_ALLOC(ext->icmp6_ifstat,
2452 	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t), M_WAITOK);
2453 
2454 	ext->nd_ifinfo = nd6_ifattach(ifp);
2455 	ext->scope6_id = scope6_ifattach(ifp);
2456 	ext->lltable = in6_lltattach(ifp);
2457 
2458 	ext->mld_ifinfo = mld_domifattach(ifp);
2459 
2460 	return ext;
2461 }
2462 
2463 int
2464 in6_domifmtu(struct ifnet *ifp)
2465 {
2466 	if (ifp->if_afdata[AF_INET6] == NULL)
2467 		return ifp->if_mtu;
2468 
2469 	return (IN6_LINKMTU(ifp));
2470 }
2471 
2472 void
2473 in6_domifdetach(struct ifnet *ifp, void *aux)
2474 {
2475 	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
2476 
2477 	mld_domifdetach(ifp);
2478 	scope6_ifdetach(ext->scope6_id);
2479 	nd6_ifdetach(ifp, ext->nd_ifinfo);
2480 	lltable_free(ext->lltable);
2481 	COUNTER_ARRAY_FREE(ext->in6_ifstat,
2482 	    sizeof(struct in6_ifstat) / sizeof(uint64_t));
2483 	free(ext->in6_ifstat, M_IFADDR);
2484 	COUNTER_ARRAY_FREE(ext->icmp6_ifstat,
2485 	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t));
2486 	free(ext->icmp6_ifstat, M_IFADDR);
2487 	free(ext, M_IFADDR);
2488 }
2489 
2490 /*
2491  * Convert sockaddr_in6 to sockaddr_in.  Original sockaddr_in6 must be
2492  * v4 mapped addr or v4 compat addr
2493  */
2494 void
2495 in6_sin6_2_sin(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2496 {
2497 
2498 	bzero(sin, sizeof(*sin));
2499 	sin->sin_len = sizeof(struct sockaddr_in);
2500 	sin->sin_family = AF_INET;
2501 	sin->sin_port = sin6->sin6_port;
2502 	sin->sin_addr.s_addr = sin6->sin6_addr.s6_addr32[3];
2503 }
2504 
2505 /* Convert sockaddr_in to sockaddr_in6 in v4 mapped addr format. */
2506 void
2507 in6_sin_2_v4mapsin6(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2508 {
2509 	bzero(sin6, sizeof(*sin6));
2510 	sin6->sin6_len = sizeof(struct sockaddr_in6);
2511 	sin6->sin6_family = AF_INET6;
2512 	sin6->sin6_port = sin->sin_port;
2513 	sin6->sin6_addr.s6_addr32[0] = 0;
2514 	sin6->sin6_addr.s6_addr32[1] = 0;
2515 	sin6->sin6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
2516 	sin6->sin6_addr.s6_addr32[3] = sin->sin_addr.s_addr;
2517 }
2518 
2519 /* Convert sockaddr_in6 into sockaddr_in. */
2520 void
2521 in6_sin6_2_sin_in_sock(struct sockaddr *nam)
2522 {
2523 	struct sockaddr_in *sin_p;
2524 	struct sockaddr_in6 sin6;
2525 
2526 	/*
2527 	 * Save original sockaddr_in6 addr and convert it
2528 	 * to sockaddr_in.
2529 	 */
2530 	sin6 = *(struct sockaddr_in6 *)nam;
2531 	sin_p = (struct sockaddr_in *)nam;
2532 	in6_sin6_2_sin(sin_p, &sin6);
2533 }
2534 
2535 /* Convert sockaddr_in into sockaddr_in6 in v4 mapped addr format. */
2536 void
2537 in6_sin_2_v4mapsin6_in_sock(struct sockaddr **nam)
2538 {
2539 	struct sockaddr_in *sin_p;
2540 	struct sockaddr_in6 *sin6_p;
2541 
2542 	sin6_p = malloc(sizeof *sin6_p, M_SONAME, M_WAITOK);
2543 	sin_p = (struct sockaddr_in *)*nam;
2544 	in6_sin_2_v4mapsin6(sin_p, sin6_p);
2545 	free(*nam, M_SONAME);
2546 	*nam = (struct sockaddr *)sin6_p;
2547 }
2548