xref: /freebsd/sys/netinet6/in6.c (revision e17f5b1d)
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 /*
119  * struct in6_ifreq and struct ifreq must be type punnable for common members
120  * of ifr_ifru to allow accessors to be shared.
121  */
122 _Static_assert(offsetof(struct in6_ifreq, ifr_ifru) ==
123     offsetof(struct ifreq, ifr_ifru),
124     "struct in6_ifreq and struct ifreq are not type punnable");
125 
126 VNET_DECLARE(int, icmp6_nodeinfo_oldmcprefix);
127 #define V_icmp6_nodeinfo_oldmcprefix	VNET(icmp6_nodeinfo_oldmcprefix)
128 
129 /*
130  * Definitions of some costant IP6 addresses.
131  */
132 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
133 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
134 const struct in6_addr in6addr_nodelocal_allnodes =
135 	IN6ADDR_NODELOCAL_ALLNODES_INIT;
136 const struct in6_addr in6addr_linklocal_allnodes =
137 	IN6ADDR_LINKLOCAL_ALLNODES_INIT;
138 const struct in6_addr in6addr_linklocal_allrouters =
139 	IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
140 const struct in6_addr in6addr_linklocal_allv2routers =
141 	IN6ADDR_LINKLOCAL_ALLV2ROUTERS_INIT;
142 
143 const struct in6_addr in6mask0 = IN6MASK0;
144 const struct in6_addr in6mask32 = IN6MASK32;
145 const struct in6_addr in6mask64 = IN6MASK64;
146 const struct in6_addr in6mask96 = IN6MASK96;
147 const struct in6_addr in6mask128 = IN6MASK128;
148 
149 const struct sockaddr_in6 sa6_any =
150 	{ sizeof(sa6_any), AF_INET6, 0, 0, IN6ADDR_ANY_INIT, 0 };
151 
152 static int in6_notify_ifa(struct ifnet *, struct in6_ifaddr *,
153 	struct in6_aliasreq *, int);
154 static void in6_unlink_ifa(struct in6_ifaddr *, struct ifnet *);
155 
156 static int in6_validate_ifra(struct ifnet *, struct in6_aliasreq *,
157     struct in6_ifaddr *, int);
158 static struct in6_ifaddr *in6_alloc_ifa(struct ifnet *,
159     struct in6_aliasreq *, int flags);
160 static int in6_update_ifa_internal(struct ifnet *, struct in6_aliasreq *,
161     struct in6_ifaddr *, int, int);
162 static int in6_broadcast_ifa(struct ifnet *, struct in6_aliasreq *,
163     struct in6_ifaddr *, int);
164 
165 #define ifa2ia6(ifa)	((struct in6_ifaddr *)(ifa))
166 #define ia62ifa(ia6)	(&((ia6)->ia_ifa))
167 
168 
169 void
170 in6_newaddrmsg(struct in6_ifaddr *ia, int cmd)
171 {
172 	struct rt_addrinfo info;
173 	struct ifaddr *ifa;
174 	struct sockaddr_dl gateway;
175 	int fibnum;
176 
177 	ifa = &ia->ia_ifa;
178 
179 	/*
180 	 * Prepare info data for the host route.
181 	 * This code mimics one from ifa_maintain_loopback_route().
182 	 */
183 	bzero(&info, sizeof(struct rt_addrinfo));
184 	info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC | RTF_PINNED;
185 	info.rti_info[RTAX_DST] = ifa->ifa_addr;
186 	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&gateway;
187 	link_init_sdl(ifa->ifa_ifp, (struct sockaddr *)&gateway, ifa->ifa_ifp->if_type);
188 	if (cmd != RTM_DELETE)
189 		info.rti_ifp = V_loif;
190 
191 
192 	fibnum = V_rt_add_addr_allfibs ? RT_ALL_FIBS : ia62ifa(ia)->ifa_ifp->if_fib;
193 
194 	if (cmd == RTM_ADD) {
195 		rt_addrmsg(cmd, &ia->ia_ifa, fibnum);
196 		rt_routemsg_info(cmd, &info, fibnum);
197 	} else if (cmd == RTM_DELETE) {
198 		rt_routemsg_info(cmd, &info, fibnum);
199 		rt_addrmsg(cmd, &ia->ia_ifa, fibnum);
200 	}
201 }
202 
203 int
204 in6_mask2len(struct in6_addr *mask, u_char *lim0)
205 {
206 	int x = 0, y;
207 	u_char *lim = lim0, *p;
208 
209 	/* ignore the scope_id part */
210 	if (lim0 == NULL || lim0 - (u_char *)mask > sizeof(*mask))
211 		lim = (u_char *)mask + sizeof(*mask);
212 	for (p = (u_char *)mask; p < lim; x++, p++) {
213 		if (*p != 0xff)
214 			break;
215 	}
216 	y = 0;
217 	if (p < lim) {
218 		for (y = 0; y < 8; y++) {
219 			if ((*p & (0x80 >> y)) == 0)
220 				break;
221 		}
222 	}
223 
224 	/*
225 	 * when the limit pointer is given, do a stricter check on the
226 	 * remaining bits.
227 	 */
228 	if (p < lim) {
229 		if (y != 0 && (*p & (0x00ff >> y)) != 0)
230 			return (-1);
231 		for (p = p + 1; p < lim; p++)
232 			if (*p != 0)
233 				return (-1);
234 	}
235 
236 	return x * 8 + y;
237 }
238 
239 #ifdef COMPAT_FREEBSD32
240 struct in6_ndifreq32 {
241 	char ifname[IFNAMSIZ];
242 	uint32_t ifindex;
243 };
244 #define	SIOCGDEFIFACE32_IN6	_IOWR('i', 86, struct in6_ndifreq32)
245 #endif
246 
247 int
248 in6_control(struct socket *so, u_long cmd, caddr_t data,
249     struct ifnet *ifp, struct thread *td)
250 {
251 	struct	in6_ifreq *ifr = (struct in6_ifreq *)data;
252 	struct	in6_ifaddr *ia = NULL;
253 	struct	in6_aliasreq *ifra = (struct in6_aliasreq *)data;
254 	struct sockaddr_in6 *sa6;
255 	int carp_attached = 0;
256 	int error;
257 	u_long ocmd = cmd;
258 
259 	/*
260 	 * Compat to make pre-10.x ifconfig(8) operable.
261 	 */
262 	if (cmd == OSIOCAIFADDR_IN6)
263 		cmd = SIOCAIFADDR_IN6;
264 
265 	switch (cmd) {
266 	case SIOCGETSGCNT_IN6:
267 	case SIOCGETMIFCNT_IN6:
268 		/*
269 		 * XXX mrt_ioctl has a 3rd, unused, FIB argument in route.c.
270 		 * We cannot see how that would be needed, so do not adjust the
271 		 * KPI blindly; more likely should clean up the IPv4 variant.
272 		 */
273 		return (mrt6_ioctl ? mrt6_ioctl(cmd, data) : EOPNOTSUPP);
274 	}
275 
276 	switch (cmd) {
277 	case SIOCAADDRCTL_POLICY:
278 	case SIOCDADDRCTL_POLICY:
279 		if (td != NULL) {
280 			error = priv_check(td, PRIV_NETINET_ADDRCTRL6);
281 			if (error)
282 				return (error);
283 		}
284 		return (in6_src_ioctl(cmd, data));
285 	}
286 
287 	if (ifp == NULL)
288 		return (EOPNOTSUPP);
289 
290 	switch (cmd) {
291 	case SIOCSNDFLUSH_IN6:
292 	case SIOCSPFXFLUSH_IN6:
293 	case SIOCSRTRFLUSH_IN6:
294 	case SIOCSDEFIFACE_IN6:
295 	case SIOCSIFINFO_FLAGS:
296 	case SIOCSIFINFO_IN6:
297 		if (td != NULL) {
298 			error = priv_check(td, PRIV_NETINET_ND6);
299 			if (error)
300 				return (error);
301 		}
302 		/* FALLTHROUGH */
303 	case OSIOCGIFINFO_IN6:
304 	case SIOCGIFINFO_IN6:
305 	case SIOCGNBRINFO_IN6:
306 	case SIOCGDEFIFACE_IN6:
307 		return (nd6_ioctl(cmd, data, ifp));
308 
309 #ifdef COMPAT_FREEBSD32
310 	case SIOCGDEFIFACE32_IN6:
311 		{
312 			struct in6_ndifreq ndif;
313 			struct in6_ndifreq32 *ndif32;
314 
315 			error = nd6_ioctl(SIOCGDEFIFACE_IN6, (caddr_t)&ndif,
316 			    ifp);
317 			if (error)
318 				return (error);
319 			ndif32 = (struct in6_ndifreq32 *)data;
320 			ndif32->ifindex = ndif.ifindex;
321 			return (0);
322 		}
323 #endif
324 	}
325 
326 	switch (cmd) {
327 	case SIOCSIFPREFIX_IN6:
328 	case SIOCDIFPREFIX_IN6:
329 	case SIOCAIFPREFIX_IN6:
330 	case SIOCCIFPREFIX_IN6:
331 	case SIOCSGIFPREFIX_IN6:
332 	case SIOCGIFPREFIX_IN6:
333 		log(LOG_NOTICE,
334 		    "prefix ioctls are now invalidated. "
335 		    "please use ifconfig.\n");
336 		return (EOPNOTSUPP);
337 	}
338 
339 	switch (cmd) {
340 	case SIOCSSCOPE6:
341 		if (td != NULL) {
342 			error = priv_check(td, PRIV_NETINET_SCOPE6);
343 			if (error)
344 				return (error);
345 		}
346 		/* FALLTHROUGH */
347 	case SIOCGSCOPE6:
348 	case SIOCGSCOPE6DEF:
349 		return (scope6_ioctl(cmd, data, ifp));
350 	}
351 
352 	/*
353 	 * Find address for this interface, if it exists.
354 	 *
355 	 * In netinet code, we have checked ifra_addr in SIOCSIF*ADDR operation
356 	 * only, and used the first interface address as the target of other
357 	 * operations (without checking ifra_addr).  This was because netinet
358 	 * code/API assumed at most 1 interface address per interface.
359 	 * Since IPv6 allows a node to assign multiple addresses
360 	 * on a single interface, we almost always look and check the
361 	 * presence of ifra_addr, and reject invalid ones here.
362 	 * It also decreases duplicated code among SIOC*_IN6 operations.
363 	 */
364 	switch (cmd) {
365 	case SIOCAIFADDR_IN6:
366 	case SIOCSIFPHYADDR_IN6:
367 		sa6 = &ifra->ifra_addr;
368 		break;
369 	case SIOCSIFADDR_IN6:
370 	case SIOCGIFADDR_IN6:
371 	case SIOCSIFDSTADDR_IN6:
372 	case SIOCSIFNETMASK_IN6:
373 	case SIOCGIFDSTADDR_IN6:
374 	case SIOCGIFNETMASK_IN6:
375 	case SIOCDIFADDR_IN6:
376 	case SIOCGIFPSRCADDR_IN6:
377 	case SIOCGIFPDSTADDR_IN6:
378 	case SIOCGIFAFLAG_IN6:
379 	case SIOCSNDFLUSH_IN6:
380 	case SIOCSPFXFLUSH_IN6:
381 	case SIOCSRTRFLUSH_IN6:
382 	case SIOCGIFALIFETIME_IN6:
383 	case SIOCGIFSTAT_IN6:
384 	case SIOCGIFSTAT_ICMP6:
385 		sa6 = &ifr->ifr_addr;
386 		break;
387 	case SIOCSIFADDR:
388 	case SIOCSIFBRDADDR:
389 	case SIOCSIFDSTADDR:
390 	case SIOCSIFNETMASK:
391 		/*
392 		 * Although we should pass any non-INET6 ioctl requests
393 		 * down to driver, we filter some legacy INET requests.
394 		 * Drivers trust SIOCSIFADDR et al to come from an already
395 		 * privileged layer, and do not perform any credentials
396 		 * checks or input validation.
397 		 */
398 		return (EINVAL);
399 	default:
400 		sa6 = NULL;
401 		break;
402 	}
403 	if (sa6 && sa6->sin6_family == AF_INET6) {
404 		if (sa6->sin6_scope_id != 0)
405 			error = sa6_embedscope(sa6, 0);
406 		else
407 			error = in6_setscope(&sa6->sin6_addr, ifp, NULL);
408 		if (error != 0)
409 			return (error);
410 		if (td != NULL && (error = prison_check_ip6(td->td_ucred,
411 		    &sa6->sin6_addr)) != 0)
412 			return (error);
413 		ia = in6ifa_ifpwithaddr(ifp, &sa6->sin6_addr);
414 	} else
415 		ia = NULL;
416 
417 	switch (cmd) {
418 	case SIOCSIFADDR_IN6:
419 	case SIOCSIFDSTADDR_IN6:
420 	case SIOCSIFNETMASK_IN6:
421 		/*
422 		 * Since IPv6 allows a node to assign multiple addresses
423 		 * on a single interface, SIOCSIFxxx ioctls are deprecated.
424 		 */
425 		/* we decided to obsolete this command (20000704) */
426 		error = EINVAL;
427 		goto out;
428 
429 	case SIOCDIFADDR_IN6:
430 		/*
431 		 * for IPv4, we look for existing in_ifaddr here to allow
432 		 * "ifconfig if0 delete" to remove the first IPv4 address on
433 		 * the interface.  For IPv6, as the spec allows multiple
434 		 * interface address from the day one, we consider "remove the
435 		 * first one" semantics to be not preferable.
436 		 */
437 		if (ia == NULL) {
438 			error = EADDRNOTAVAIL;
439 			goto out;
440 		}
441 		/* FALLTHROUGH */
442 	case SIOCAIFADDR_IN6:
443 		/*
444 		 * We always require users to specify a valid IPv6 address for
445 		 * the corresponding operation.
446 		 */
447 		if (ifra->ifra_addr.sin6_family != AF_INET6 ||
448 		    ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6)) {
449 			error = EAFNOSUPPORT;
450 			goto out;
451 		}
452 
453 		if (td != NULL) {
454 			error = priv_check(td, (cmd == SIOCDIFADDR_IN6) ?
455 			    PRIV_NET_DELIFADDR : PRIV_NET_ADDIFADDR);
456 			if (error)
457 				goto out;
458 		}
459 		/* FALLTHROUGH */
460 	case SIOCGIFSTAT_IN6:
461 	case SIOCGIFSTAT_ICMP6:
462 		if (ifp->if_afdata[AF_INET6] == NULL) {
463 			error = EPFNOSUPPORT;
464 			goto out;
465 		}
466 		break;
467 
468 	case SIOCGIFADDR_IN6:
469 		/* This interface is basically deprecated. use SIOCGIFCONF. */
470 		/* FALLTHROUGH */
471 	case SIOCGIFAFLAG_IN6:
472 	case SIOCGIFNETMASK_IN6:
473 	case SIOCGIFDSTADDR_IN6:
474 	case SIOCGIFALIFETIME_IN6:
475 		/* must think again about its semantics */
476 		if (ia == NULL) {
477 			error = EADDRNOTAVAIL;
478 			goto out;
479 		}
480 		break;
481 	}
482 
483 	switch (cmd) {
484 	case SIOCGIFADDR_IN6:
485 		ifr->ifr_addr = ia->ia_addr;
486 		if ((error = sa6_recoverscope(&ifr->ifr_addr)) != 0)
487 			goto out;
488 		break;
489 
490 	case SIOCGIFDSTADDR_IN6:
491 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
492 			error = EINVAL;
493 			goto out;
494 		}
495 		ifr->ifr_dstaddr = ia->ia_dstaddr;
496 		if ((error = sa6_recoverscope(&ifr->ifr_dstaddr)) != 0)
497 			goto out;
498 		break;
499 
500 	case SIOCGIFNETMASK_IN6:
501 		ifr->ifr_addr = ia->ia_prefixmask;
502 		break;
503 
504 	case SIOCGIFAFLAG_IN6:
505 		ifr->ifr_ifru.ifru_flags6 = ia->ia6_flags;
506 		break;
507 
508 	case SIOCGIFSTAT_IN6:
509 		COUNTER_ARRAY_COPY(((struct in6_ifextra *)
510 		    ifp->if_afdata[AF_INET6])->in6_ifstat,
511 		    &ifr->ifr_ifru.ifru_stat,
512 		    sizeof(struct in6_ifstat) / sizeof(uint64_t));
513 		break;
514 
515 	case SIOCGIFSTAT_ICMP6:
516 		COUNTER_ARRAY_COPY(((struct in6_ifextra *)
517 		    ifp->if_afdata[AF_INET6])->icmp6_ifstat,
518 		    &ifr->ifr_ifru.ifru_icmp6stat,
519 		    sizeof(struct icmp6_ifstat) / sizeof(uint64_t));
520 		break;
521 
522 	case SIOCGIFALIFETIME_IN6:
523 		ifr->ifr_ifru.ifru_lifetime = ia->ia6_lifetime;
524 		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
525 			time_t maxexpire;
526 			struct in6_addrlifetime *retlt =
527 			    &ifr->ifr_ifru.ifru_lifetime;
528 
529 			/*
530 			 * XXX: adjust expiration time assuming time_t is
531 			 * signed.
532 			 */
533 			maxexpire = (-1) &
534 			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
535 			if (ia->ia6_lifetime.ia6t_vltime <
536 			    maxexpire - ia->ia6_updatetime) {
537 				retlt->ia6t_expire = ia->ia6_updatetime +
538 				    ia->ia6_lifetime.ia6t_vltime;
539 			} else
540 				retlt->ia6t_expire = maxexpire;
541 		}
542 		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
543 			time_t maxexpire;
544 			struct in6_addrlifetime *retlt =
545 			    &ifr->ifr_ifru.ifru_lifetime;
546 
547 			/*
548 			 * XXX: adjust expiration time assuming time_t is
549 			 * signed.
550 			 */
551 			maxexpire = (-1) &
552 			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
553 			if (ia->ia6_lifetime.ia6t_pltime <
554 			    maxexpire - ia->ia6_updatetime) {
555 				retlt->ia6t_preferred = ia->ia6_updatetime +
556 				    ia->ia6_lifetime.ia6t_pltime;
557 			} else
558 				retlt->ia6t_preferred = maxexpire;
559 		}
560 		break;
561 
562 	case SIOCAIFADDR_IN6:
563 	{
564 		struct nd_prefixctl pr0;
565 		struct nd_prefix *pr;
566 
567 		/*
568 		 * first, make or update the interface address structure,
569 		 * and link it to the list.
570 		 */
571 		if ((error = in6_update_ifa(ifp, ifra, ia, 0)) != 0)
572 			goto out;
573 		if (ia != NULL) {
574 			if (ia->ia_ifa.ifa_carp)
575 				(*carp_detach_p)(&ia->ia_ifa, true);
576 			ifa_free(&ia->ia_ifa);
577 		}
578 		if ((ia = in6ifa_ifpwithaddr(ifp, &ifra->ifra_addr.sin6_addr))
579 		    == NULL) {
580 			/*
581 			 * this can happen when the user specify the 0 valid
582 			 * lifetime.
583 			 */
584 			break;
585 		}
586 
587 		if (cmd == ocmd && ifra->ifra_vhid > 0) {
588 			if (carp_attach_p != NULL)
589 				error = (*carp_attach_p)(&ia->ia_ifa,
590 				    ifra->ifra_vhid);
591 			else
592 				error = EPROTONOSUPPORT;
593 			if (error)
594 				goto out;
595 			else
596 				carp_attached = 1;
597 		}
598 
599 		/*
600 		 * then, make the prefix on-link on the interface.
601 		 * XXX: we'd rather create the prefix before the address, but
602 		 * we need at least one address to install the corresponding
603 		 * interface route, so we configure the address first.
604 		 */
605 
606 		/*
607 		 * convert mask to prefix length (prefixmask has already
608 		 * been validated in in6_update_ifa().
609 		 */
610 		bzero(&pr0, sizeof(pr0));
611 		pr0.ndpr_ifp = ifp;
612 		pr0.ndpr_plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
613 		    NULL);
614 		if (pr0.ndpr_plen == 128) {
615 			/* we don't need to install a host route. */
616 			goto aifaddr_out;
617 		}
618 		pr0.ndpr_prefix = ifra->ifra_addr;
619 		/* apply the mask for safety. */
620 		IN6_MASK_ADDR(&pr0.ndpr_prefix.sin6_addr,
621 		    &ifra->ifra_prefixmask.sin6_addr);
622 
623 		/*
624 		 * XXX: since we don't have an API to set prefix (not address)
625 		 * lifetimes, we just use the same lifetimes as addresses.
626 		 * The (temporarily) installed lifetimes can be overridden by
627 		 * later advertised RAs (when accept_rtadv is non 0), which is
628 		 * an intended behavior.
629 		 */
630 		pr0.ndpr_raf_onlink = 1; /* should be configurable? */
631 		pr0.ndpr_raf_auto =
632 		    ((ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0);
633 		pr0.ndpr_vltime = ifra->ifra_lifetime.ia6t_vltime;
634 		pr0.ndpr_pltime = ifra->ifra_lifetime.ia6t_pltime;
635 
636 		/* add the prefix if not yet. */
637 		if ((pr = nd6_prefix_lookup(&pr0)) == NULL) {
638 			/*
639 			 * nd6_prelist_add will install the corresponding
640 			 * interface route.
641 			 */
642 			if ((error = nd6_prelist_add(&pr0, NULL, &pr)) != 0) {
643 				if (carp_attached)
644 					(*carp_detach_p)(&ia->ia_ifa, false);
645 				goto out;
646 			}
647 		}
648 
649 		/* relate the address to the prefix */
650 		if (ia->ia6_ndpr == NULL) {
651 			ia->ia6_ndpr = pr;
652 			pr->ndpr_addrcnt++;
653 
654 			/*
655 			 * If this is the first autoconf address from the
656 			 * prefix, create a temporary address as well
657 			 * (when required).
658 			 */
659 			if ((ia->ia6_flags & IN6_IFF_AUTOCONF) &&
660 			    V_ip6_use_tempaddr && pr->ndpr_addrcnt == 1) {
661 				int e;
662 				if ((e = in6_tmpifadd(ia, 1, 0)) != 0) {
663 					log(LOG_NOTICE, "in6_control: failed "
664 					    "to create a temporary address, "
665 					    "errno=%d\n", e);
666 				}
667 			}
668 		}
669 		nd6_prefix_rele(pr);
670 
671 		/*
672 		 * this might affect the status of autoconfigured addresses,
673 		 * that is, this address might make other addresses detached.
674 		 */
675 		pfxlist_onlink_check();
676 
677 aifaddr_out:
678 		/*
679 		 * Try to clear the flag when a new IPv6 address is added
680 		 * onto an IFDISABLED interface and it succeeds.
681 		 */
682 		if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) {
683 			struct in6_ndireq nd;
684 
685 			memset(&nd, 0, sizeof(nd));
686 			nd.ndi.flags = ND_IFINFO(ifp)->flags;
687 			nd.ndi.flags &= ~ND6_IFF_IFDISABLED;
688 			if (nd6_ioctl(SIOCSIFINFO_FLAGS, (caddr_t)&nd, ifp) < 0)
689 				log(LOG_NOTICE, "SIOCAIFADDR_IN6: "
690 				    "SIOCSIFINFO_FLAGS for -ifdisabled "
691 				    "failed.");
692 			/*
693 			 * Ignore failure of clearing the flag intentionally.
694 			 * The failure means address duplication was detected.
695 			 */
696 		}
697 		break;
698 	}
699 
700 	case SIOCDIFADDR_IN6:
701 	{
702 		struct nd_prefix *pr;
703 
704 		/*
705 		 * If the address being deleted is the only one that owns
706 		 * the corresponding prefix, expire the prefix as well.
707 		 * XXX: theoretically, we don't have to worry about such
708 		 * relationship, since we separate the address management
709 		 * and the prefix management.  We do this, however, to provide
710 		 * as much backward compatibility as possible in terms of
711 		 * the ioctl operation.
712 		 * Note that in6_purgeaddr() will decrement ndpr_addrcnt.
713 		 */
714 		pr = ia->ia6_ndpr;
715 		in6_purgeaddr(&ia->ia_ifa);
716 		if (pr != NULL && pr->ndpr_addrcnt == 0) {
717 			ND6_WLOCK();
718 			nd6_prefix_unlink(pr, NULL);
719 			ND6_WUNLOCK();
720 			nd6_prefix_del(pr);
721 		}
722 		EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, &ia->ia_ifa,
723 		    IFADDR_EVENT_DEL);
724 		break;
725 	}
726 
727 	default:
728 		if (ifp->if_ioctl == NULL) {
729 			error = EOPNOTSUPP;
730 			goto out;
731 		}
732 		error = (*ifp->if_ioctl)(ifp, cmd, data);
733 		goto out;
734 	}
735 
736 	error = 0;
737 out:
738 	if (ia != NULL)
739 		ifa_free(&ia->ia_ifa);
740 	return (error);
741 }
742 
743 
744 static struct in6_multi_mship *
745 in6_joingroup_legacy(struct ifnet *ifp, const struct in6_addr *mcaddr,
746     int *errorp, int delay)
747 {
748 	struct in6_multi_mship *imm;
749 	int error;
750 
751 	imm = malloc(sizeof(*imm), M_IP6MADDR, M_NOWAIT);
752 	if (imm == NULL) {
753 		*errorp = ENOBUFS;
754 		return (NULL);
755 	}
756 
757 	delay = (delay * PR_FASTHZ) / hz;
758 
759 	error = in6_joingroup(ifp, mcaddr, NULL, &imm->i6mm_maddr, delay);
760 	if (error) {
761 		*errorp = error;
762 		free(imm, M_IP6MADDR);
763 		return (NULL);
764 	}
765 
766 	return (imm);
767 }
768 /*
769  * Join necessary multicast groups.  Factored out from in6_update_ifa().
770  * This entire work should only be done once, for the default FIB.
771  */
772 static int
773 in6_update_ifa_join_mc(struct ifnet *ifp, struct in6_aliasreq *ifra,
774     struct in6_ifaddr *ia, int flags, struct in6_multi **in6m_sol)
775 {
776 	char ip6buf[INET6_ADDRSTRLEN];
777 	struct in6_addr mltaddr;
778 	struct in6_multi_mship *imm;
779 	int delay, error;
780 
781 	KASSERT(in6m_sol != NULL, ("%s: in6m_sol is NULL", __func__));
782 
783 	/* Join solicited multicast addr for new host id. */
784 	bzero(&mltaddr, sizeof(struct in6_addr));
785 	mltaddr.s6_addr32[0] = IPV6_ADDR_INT32_MLL;
786 	mltaddr.s6_addr32[2] = htonl(1);
787 	mltaddr.s6_addr32[3] = ifra->ifra_addr.sin6_addr.s6_addr32[3];
788 	mltaddr.s6_addr8[12] = 0xff;
789 	if ((error = in6_setscope(&mltaddr, ifp, NULL)) != 0) {
790 		/* XXX: should not happen */
791 		log(LOG_ERR, "%s: in6_setscope failed\n", __func__);
792 		goto cleanup;
793 	}
794 	delay = error = 0;
795 	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
796 		/*
797 		 * We need a random delay for DAD on the address being
798 		 * configured.  It also means delaying transmission of the
799 		 * corresponding MLD report to avoid report collision.
800 		 * [RFC 4861, Section 6.3.7]
801 		 */
802 		delay = arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz);
803 	}
804 	imm = in6_joingroup_legacy(ifp, &mltaddr, &error, delay);
805 	if (imm == NULL) {
806 		nd6log((LOG_WARNING, "%s: in6_joingroup failed for %s on %s "
807 		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf, &mltaddr),
808 		    if_name(ifp), error));
809 		goto cleanup;
810 	}
811 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
812 	*in6m_sol = imm->i6mm_maddr;
813 
814 	/*
815 	 * Join link-local all-nodes address.
816 	 */
817 	mltaddr = in6addr_linklocal_allnodes;
818 	if ((error = in6_setscope(&mltaddr, ifp, NULL)) != 0)
819 		goto cleanup; /* XXX: should not fail */
820 
821 	imm = in6_joingroup_legacy(ifp, &mltaddr, &error, 0);
822 	if (imm == NULL) {
823 		nd6log((LOG_WARNING, "%s: in6_joingroup failed for %s on %s "
824 		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf, &mltaddr),
825 		    if_name(ifp), error));
826 		goto cleanup;
827 	}
828 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
829 
830 	/*
831 	 * Join node information group address.
832 	 */
833 	delay = 0;
834 	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
835 		/*
836 		 * The spec does not say anything about delay for this group,
837 		 * but the same logic should apply.
838 		 */
839 		delay = arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz);
840 	}
841 	if (in6_nigroup(ifp, NULL, -1, &mltaddr) == 0) {
842 		/* XXX jinmei */
843 		imm = in6_joingroup_legacy(ifp, &mltaddr, &error, delay);
844 		if (imm == NULL)
845 			nd6log((LOG_WARNING,
846 			    "%s: in6_joingroup failed for %s on %s "
847 			    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
848 			    &mltaddr), if_name(ifp), error));
849 			/* XXX not very fatal, go on... */
850 		else
851 			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
852 	}
853 	if (V_icmp6_nodeinfo_oldmcprefix &&
854 	    in6_nigroup_oldmcprefix(ifp, NULL, -1, &mltaddr) == 0) {
855 		imm = in6_joingroup_legacy(ifp, &mltaddr, &error, delay);
856 		if (imm == NULL)
857 			nd6log((LOG_WARNING,
858 			    "%s: in6_joingroup failed for %s on %s "
859 			    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
860 			    &mltaddr), if_name(ifp), error));
861 			/* XXX not very fatal, go on... */
862 		else
863 			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
864 	}
865 
866 	/*
867 	 * Join interface-local all-nodes address.
868 	 * (ff01::1%ifN, and ff01::%ifN/32)
869 	 */
870 	mltaddr = in6addr_nodelocal_allnodes;
871 	if ((error = in6_setscope(&mltaddr, ifp, NULL)) != 0)
872 		goto cleanup; /* XXX: should not fail */
873 
874 	imm = in6_joingroup_legacy(ifp, &mltaddr, &error, 0);
875 	if (imm == NULL) {
876 		nd6log((LOG_WARNING, "%s: in6_joingroup failed for %s on %s "
877 		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
878 		    &mltaddr), if_name(ifp), error));
879 		goto cleanup;
880 	}
881 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
882 
883 cleanup:
884 	return (error);
885 }
886 
887 /*
888  * Update parameters of an IPv6 interface address.
889  * If necessary, a new entry is created and linked into address chains.
890  * This function is separated from in6_control().
891  */
892 int
893 in6_update_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
894     struct in6_ifaddr *ia, int flags)
895 {
896 	int error, hostIsNew = 0;
897 
898 	if ((error = in6_validate_ifra(ifp, ifra, ia, flags)) != 0)
899 		return (error);
900 
901 	if (ia == NULL) {
902 		hostIsNew = 1;
903 		if ((ia = in6_alloc_ifa(ifp, ifra, flags)) == NULL)
904 			return (ENOBUFS);
905 	}
906 
907 	error = in6_update_ifa_internal(ifp, ifra, ia, hostIsNew, flags);
908 	if (error != 0) {
909 		if (hostIsNew != 0) {
910 			in6_unlink_ifa(ia, ifp);
911 			ifa_free(&ia->ia_ifa);
912 		}
913 		return (error);
914 	}
915 
916 	if (hostIsNew)
917 		error = in6_broadcast_ifa(ifp, ifra, ia, flags);
918 
919 	return (error);
920 }
921 
922 /*
923  * Fill in basic IPv6 address request info.
924  */
925 void
926 in6_prepare_ifra(struct in6_aliasreq *ifra, const struct in6_addr *addr,
927     const struct in6_addr *mask)
928 {
929 
930 	memset(ifra, 0, sizeof(struct in6_aliasreq));
931 
932 	ifra->ifra_addr.sin6_family = AF_INET6;
933 	ifra->ifra_addr.sin6_len = sizeof(struct sockaddr_in6);
934 	if (addr != NULL)
935 		ifra->ifra_addr.sin6_addr = *addr;
936 
937 	ifra->ifra_prefixmask.sin6_family = AF_INET6;
938 	ifra->ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
939 	if (mask != NULL)
940 		ifra->ifra_prefixmask.sin6_addr = *mask;
941 }
942 
943 static int
944 in6_validate_ifra(struct ifnet *ifp, struct in6_aliasreq *ifra,
945     struct in6_ifaddr *ia, int flags)
946 {
947 	int plen = -1;
948 	struct sockaddr_in6 dst6;
949 	struct in6_addrlifetime *lt;
950 	char ip6buf[INET6_ADDRSTRLEN];
951 
952 	/* Validate parameters */
953 	if (ifp == NULL || ifra == NULL) /* this maybe redundant */
954 		return (EINVAL);
955 
956 	/*
957 	 * The destination address for a p2p link must have a family
958 	 * of AF_UNSPEC or AF_INET6.
959 	 */
960 	if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
961 	    ifra->ifra_dstaddr.sin6_family != AF_INET6 &&
962 	    ifra->ifra_dstaddr.sin6_family != AF_UNSPEC)
963 		return (EAFNOSUPPORT);
964 
965 	/*
966 	 * Validate address
967 	 */
968 	if (ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6) ||
969 	    ifra->ifra_addr.sin6_family != AF_INET6)
970 		return (EINVAL);
971 
972 	/*
973 	 * validate ifra_prefixmask.  don't check sin6_family, netmask
974 	 * does not carry fields other than sin6_len.
975 	 */
976 	if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6))
977 		return (EINVAL);
978 	/*
979 	 * Because the IPv6 address architecture is classless, we require
980 	 * users to specify a (non 0) prefix length (mask) for a new address.
981 	 * We also require the prefix (when specified) mask is valid, and thus
982 	 * reject a non-consecutive mask.
983 	 */
984 	if (ia == NULL && ifra->ifra_prefixmask.sin6_len == 0)
985 		return (EINVAL);
986 	if (ifra->ifra_prefixmask.sin6_len != 0) {
987 		plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
988 		    (u_char *)&ifra->ifra_prefixmask +
989 		    ifra->ifra_prefixmask.sin6_len);
990 		if (plen <= 0)
991 			return (EINVAL);
992 	} else {
993 		/*
994 		 * In this case, ia must not be NULL.  We just use its prefix
995 		 * length.
996 		 */
997 		plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
998 	}
999 	/*
1000 	 * If the destination address on a p2p interface is specified,
1001 	 * and the address is a scoped one, validate/set the scope
1002 	 * zone identifier.
1003 	 */
1004 	dst6 = ifra->ifra_dstaddr;
1005 	if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) != 0 &&
1006 	    (dst6.sin6_family == AF_INET6)) {
1007 		struct in6_addr in6_tmp;
1008 		u_int32_t zoneid;
1009 
1010 		in6_tmp = dst6.sin6_addr;
1011 		if (in6_setscope(&in6_tmp, ifp, &zoneid))
1012 			return (EINVAL); /* XXX: should be impossible */
1013 
1014 		if (dst6.sin6_scope_id != 0) {
1015 			if (dst6.sin6_scope_id != zoneid)
1016 				return (EINVAL);
1017 		} else		/* user omit to specify the ID. */
1018 			dst6.sin6_scope_id = zoneid;
1019 
1020 		/* convert into the internal form */
1021 		if (sa6_embedscope(&dst6, 0))
1022 			return (EINVAL); /* XXX: should be impossible */
1023 	}
1024 	/* Modify original ifra_dstaddr to reflect changes */
1025 	ifra->ifra_dstaddr = dst6;
1026 
1027 	/*
1028 	 * The destination address can be specified only for a p2p or a
1029 	 * loopback interface.  If specified, the corresponding prefix length
1030 	 * must be 128.
1031 	 */
1032 	if (ifra->ifra_dstaddr.sin6_family == AF_INET6) {
1033 		if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) == 0) {
1034 			/* XXX: noisy message */
1035 			nd6log((LOG_INFO, "in6_update_ifa: a destination can "
1036 			    "be specified for a p2p or a loopback IF only\n"));
1037 			return (EINVAL);
1038 		}
1039 		if (plen != 128) {
1040 			nd6log((LOG_INFO, "in6_update_ifa: prefixlen should "
1041 			    "be 128 when dstaddr is specified\n"));
1042 			return (EINVAL);
1043 		}
1044 	}
1045 	/* lifetime consistency check */
1046 	lt = &ifra->ifra_lifetime;
1047 	if (lt->ia6t_pltime > lt->ia6t_vltime)
1048 		return (EINVAL);
1049 	if (lt->ia6t_vltime == 0) {
1050 		/*
1051 		 * the following log might be noisy, but this is a typical
1052 		 * configuration mistake or a tool's bug.
1053 		 */
1054 		nd6log((LOG_INFO,
1055 		    "in6_update_ifa: valid lifetime is 0 for %s\n",
1056 		    ip6_sprintf(ip6buf, &ifra->ifra_addr.sin6_addr)));
1057 
1058 		if (ia == NULL)
1059 			return (0); /* there's nothing to do */
1060 	}
1061 
1062 	/* Check prefix mask */
1063 	if (ia != NULL && ifra->ifra_prefixmask.sin6_len != 0) {
1064 		/*
1065 		 * We prohibit changing the prefix length of an existing
1066 		 * address, because
1067 		 * + such an operation should be rare in IPv6, and
1068 		 * + the operation would confuse prefix management.
1069 		 */
1070 		if (ia->ia_prefixmask.sin6_len != 0 &&
1071 		    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) != plen) {
1072 			nd6log((LOG_INFO, "in6_validate_ifa: the prefix length "
1073 			    "of an existing %s address should not be changed\n",
1074 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1075 
1076 			return (EINVAL);
1077 		}
1078 	}
1079 
1080 	return (0);
1081 }
1082 
1083 
1084 /*
1085  * Allocate a new ifaddr and link it into chains.
1086  */
1087 static struct in6_ifaddr *
1088 in6_alloc_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra, int flags)
1089 {
1090 	struct in6_ifaddr *ia;
1091 
1092 	/*
1093 	 * When in6_alloc_ifa() is called in a process of a received
1094 	 * RA, it is called under an interrupt context.  So, we should
1095 	 * call malloc with M_NOWAIT.
1096 	 */
1097 	ia = (struct in6_ifaddr *)ifa_alloc(sizeof(*ia), M_NOWAIT);
1098 	if (ia == NULL)
1099 		return (NULL);
1100 	LIST_INIT(&ia->ia6_memberships);
1101 	/* Initialize the address and masks, and put time stamp */
1102 	ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
1103 	ia->ia_addr.sin6_family = AF_INET6;
1104 	ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
1105 	/* XXX: Can we assign ,sin6_addr and skip the rest? */
1106 	ia->ia_addr = ifra->ifra_addr;
1107 	ia->ia6_createtime = time_uptime;
1108 	if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
1109 		/*
1110 		 * Some functions expect that ifa_dstaddr is not
1111 		 * NULL for p2p interfaces.
1112 		 */
1113 		ia->ia_ifa.ifa_dstaddr =
1114 		    (struct sockaddr *)&ia->ia_dstaddr;
1115 	} else {
1116 		ia->ia_ifa.ifa_dstaddr = NULL;
1117 	}
1118 
1119 	/* set prefix mask if any */
1120 	ia->ia_ifa.ifa_netmask = (struct sockaddr *)&ia->ia_prefixmask;
1121 	if (ifra->ifra_prefixmask.sin6_len != 0) {
1122 		ia->ia_prefixmask.sin6_family = AF_INET6;
1123 		ia->ia_prefixmask.sin6_len = ifra->ifra_prefixmask.sin6_len;
1124 		ia->ia_prefixmask.sin6_addr = ifra->ifra_prefixmask.sin6_addr;
1125 	}
1126 
1127 	ia->ia_ifp = ifp;
1128 	ifa_ref(&ia->ia_ifa);			/* if_addrhead */
1129 	IF_ADDR_WLOCK(ifp);
1130 	CK_STAILQ_INSERT_TAIL(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
1131 	IF_ADDR_WUNLOCK(ifp);
1132 
1133 	ifa_ref(&ia->ia_ifa);			/* in6_ifaddrhead */
1134 	IN6_IFADDR_WLOCK();
1135 	CK_STAILQ_INSERT_TAIL(&V_in6_ifaddrhead, ia, ia_link);
1136 	CK_LIST_INSERT_HEAD(IN6ADDR_HASH(&ia->ia_addr.sin6_addr), ia, ia6_hash);
1137 	IN6_IFADDR_WUNLOCK();
1138 
1139 	return (ia);
1140 }
1141 
1142 /*
1143  * Update/configure interface address parameters:
1144  *
1145  * 1) Update lifetime
1146  * 2) Update interface metric ad flags
1147  * 3) Notify other subsystems
1148  */
1149 static int
1150 in6_update_ifa_internal(struct ifnet *ifp, struct in6_aliasreq *ifra,
1151     struct in6_ifaddr *ia, int hostIsNew, int flags)
1152 {
1153 	int error;
1154 
1155 	/* update timestamp */
1156 	ia->ia6_updatetime = time_uptime;
1157 
1158 	/*
1159 	 * Set lifetimes.  We do not refer to ia6t_expire and ia6t_preferred
1160 	 * to see if the address is deprecated or invalidated, but initialize
1161 	 * these members for applications.
1162 	 */
1163 	ia->ia6_lifetime = ifra->ifra_lifetime;
1164 	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1165 		ia->ia6_lifetime.ia6t_expire =
1166 		    time_uptime + ia->ia6_lifetime.ia6t_vltime;
1167 	} else
1168 		ia->ia6_lifetime.ia6t_expire = 0;
1169 	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1170 		ia->ia6_lifetime.ia6t_preferred =
1171 		    time_uptime + ia->ia6_lifetime.ia6t_pltime;
1172 	} else
1173 		ia->ia6_lifetime.ia6t_preferred = 0;
1174 
1175 	/*
1176 	 * backward compatibility - if IN6_IFF_DEPRECATED is set from the
1177 	 * userland, make it deprecated.
1178 	 */
1179 	if ((ifra->ifra_flags & IN6_IFF_DEPRECATED) != 0) {
1180 		ia->ia6_lifetime.ia6t_pltime = 0;
1181 		ia->ia6_lifetime.ia6t_preferred = time_uptime;
1182 	}
1183 
1184 	/*
1185 	 * configure address flags.
1186 	 */
1187 	ia->ia6_flags = ifra->ifra_flags;
1188 
1189 	/*
1190 	 * Make the address tentative before joining multicast addresses,
1191 	 * so that corresponding MLD responses would not have a tentative
1192 	 * source address.
1193 	 */
1194 	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/* safety */
1195 
1196 	/*
1197 	 * DAD should be performed for an new address or addresses on
1198 	 * an interface with ND6_IFF_IFDISABLED.
1199 	 */
1200 	if (in6if_do_dad(ifp) &&
1201 	    (hostIsNew || (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)))
1202 		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1203 
1204 	/* notify other subsystems */
1205 	error = in6_notify_ifa(ifp, ia, ifra, hostIsNew);
1206 
1207 	return (error);
1208 }
1209 
1210 /*
1211  * Do link-level ifa job:
1212  * 1) Add lle entry for added address
1213  * 2) Notifies routing socket users about new address
1214  * 3) join appropriate multicast group
1215  * 4) start DAD if enabled
1216  */
1217 static int
1218 in6_broadcast_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
1219     struct in6_ifaddr *ia, int flags)
1220 {
1221 	struct in6_multi *in6m_sol;
1222 	int error = 0;
1223 
1224 	/* Add local address to lltable, if necessary (ex. on p2p link). */
1225 	if ((error = nd6_add_ifa_lle(ia)) != 0) {
1226 		in6_purgeaddr(&ia->ia_ifa);
1227 		ifa_free(&ia->ia_ifa);
1228 		return (error);
1229 	}
1230 
1231 	/* Join necessary multicast groups. */
1232 	in6m_sol = NULL;
1233 	if ((ifp->if_flags & IFF_MULTICAST) != 0) {
1234 		error = in6_update_ifa_join_mc(ifp, ifra, ia, flags, &in6m_sol);
1235 		if (error != 0) {
1236 			in6_purgeaddr(&ia->ia_ifa);
1237 			ifa_free(&ia->ia_ifa);
1238 			return (error);
1239 		}
1240 	}
1241 
1242 	/* Perform DAD, if the address is TENTATIVE. */
1243 	if ((ia->ia6_flags & IN6_IFF_TENTATIVE)) {
1244 		int delay, mindelay, maxdelay;
1245 
1246 		delay = 0;
1247 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1248 			/*
1249 			 * We need to impose a delay before sending an NS
1250 			 * for DAD.  Check if we also needed a delay for the
1251 			 * corresponding MLD message.  If we did, the delay
1252 			 * should be larger than the MLD delay (this could be
1253 			 * relaxed a bit, but this simple logic is at least
1254 			 * safe).
1255 			 * XXX: Break data hiding guidelines and look at
1256 			 * state for the solicited multicast group.
1257 			 */
1258 			mindelay = 0;
1259 			if (in6m_sol != NULL &&
1260 			    in6m_sol->in6m_state == MLD_REPORTING_MEMBER) {
1261 				mindelay = in6m_sol->in6m_timer;
1262 			}
1263 			maxdelay = MAX_RTR_SOLICITATION_DELAY * hz;
1264 			if (maxdelay - mindelay == 0)
1265 				delay = 0;
1266 			else {
1267 				delay =
1268 				    (arc4random() % (maxdelay - mindelay)) +
1269 				    mindelay;
1270 			}
1271 		}
1272 		nd6_dad_start((struct ifaddr *)ia, delay);
1273 	}
1274 
1275 	in6_newaddrmsg(ia, RTM_ADD);
1276 	ifa_free(&ia->ia_ifa);
1277 	return (error);
1278 }
1279 
1280 void
1281 in6_purgeaddr(struct ifaddr *ifa)
1282 {
1283 	struct ifnet *ifp = ifa->ifa_ifp;
1284 	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1285 	struct in6_multi_mship *imm;
1286 	int plen, error;
1287 
1288 	if (ifa->ifa_carp)
1289 		(*carp_detach_p)(ifa, false);
1290 
1291 	/*
1292 	 * Remove the loopback route to the interface address.
1293 	 * The check for the current setting of "nd6_useloopback"
1294 	 * is not needed.
1295 	 */
1296 	if (ia->ia_flags & IFA_RTSELF) {
1297 		error = ifa_del_loopback_route((struct ifaddr *)ia,
1298 		    (struct sockaddr *)&ia->ia_addr);
1299 		if (error == 0)
1300 			ia->ia_flags &= ~IFA_RTSELF;
1301 	}
1302 
1303 	/* stop DAD processing */
1304 	nd6_dad_stop(ifa);
1305 
1306 	/* Leave multicast groups. */
1307 	while ((imm = LIST_FIRST(&ia->ia6_memberships)) != NULL) {
1308 		LIST_REMOVE(imm, i6mm_chain);
1309 		if (imm->i6mm_maddr != NULL)
1310 			in6_leavegroup(imm->i6mm_maddr, NULL);
1311 		free(imm, M_IP6MADDR);
1312 	}
1313 	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1314 	if ((ia->ia_flags & IFA_ROUTE) && plen == 128) {
1315 		error = rtinit(&(ia->ia_ifa), RTM_DELETE, ia->ia_flags |
1316 		    (ia->ia_dstaddr.sin6_family == AF_INET6 ? RTF_HOST : 0));
1317 		if (error != 0)
1318 			log(LOG_INFO, "%s: err=%d, destination address delete "
1319 			    "failed\n", __func__, error);
1320 		ia->ia_flags &= ~IFA_ROUTE;
1321 	}
1322 
1323 	in6_newaddrmsg(ia, RTM_DELETE);
1324 	in6_unlink_ifa(ia, ifp);
1325 }
1326 
1327 static void
1328 in6_unlink_ifa(struct in6_ifaddr *ia, struct ifnet *ifp)
1329 {
1330 	char ip6buf[INET6_ADDRSTRLEN];
1331 	int remove_lle;
1332 
1333 	IF_ADDR_WLOCK(ifp);
1334 	CK_STAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifaddr, ifa_link);
1335 	IF_ADDR_WUNLOCK(ifp);
1336 	ifa_free(&ia->ia_ifa);			/* if_addrhead */
1337 
1338 	/*
1339 	 * Defer the release of what might be the last reference to the
1340 	 * in6_ifaddr so that it can't be freed before the remainder of the
1341 	 * cleanup.
1342 	 */
1343 	IN6_IFADDR_WLOCK();
1344 	CK_STAILQ_REMOVE(&V_in6_ifaddrhead, ia, in6_ifaddr, ia_link);
1345 	CK_LIST_REMOVE(ia, ia6_hash);
1346 	IN6_IFADDR_WUNLOCK();
1347 
1348 	/*
1349 	 * Release the reference to the base prefix.  There should be a
1350 	 * positive reference.
1351 	 */
1352 	remove_lle = 0;
1353 	if (ia->ia6_ndpr == NULL) {
1354 		nd6log((LOG_NOTICE,
1355 		    "in6_unlink_ifa: autoconf'ed address "
1356 		    "%s has no prefix\n", ip6_sprintf(ip6buf, IA6_IN6(ia))));
1357 	} else {
1358 		ia->ia6_ndpr->ndpr_addrcnt--;
1359 		/* Do not delete lles within prefix if refcont != 0 */
1360 		if (ia->ia6_ndpr->ndpr_addrcnt == 0)
1361 			remove_lle = 1;
1362 		ia->ia6_ndpr = NULL;
1363 	}
1364 
1365 	nd6_rem_ifa_lle(ia, remove_lle);
1366 
1367 	/*
1368 	 * Also, if the address being removed is autoconf'ed, call
1369 	 * pfxlist_onlink_check() since the release might affect the status of
1370 	 * other (detached) addresses.
1371 	 */
1372 	if ((ia->ia6_flags & IN6_IFF_AUTOCONF)) {
1373 		pfxlist_onlink_check();
1374 	}
1375 	ifa_free(&ia->ia_ifa);			/* in6_ifaddrhead */
1376 }
1377 
1378 /*
1379  * Notifies other subsystems about address change/arrival:
1380  * 1) Notifies device handler on the first IPv6 address assignment
1381  * 2) Handle routing table changes for P2P links and route
1382  * 3) Handle routing table changes for address host route
1383  */
1384 static int
1385 in6_notify_ifa(struct ifnet *ifp, struct in6_ifaddr *ia,
1386     struct in6_aliasreq *ifra, int hostIsNew)
1387 {
1388 	int	error = 0, plen, ifacount = 0;
1389 	struct ifaddr *ifa;
1390 	struct sockaddr_in6 *pdst;
1391 	char ip6buf[INET6_ADDRSTRLEN];
1392 
1393 	/*
1394 	 * Give the interface a chance to initialize
1395 	 * if this is its first address,
1396 	 */
1397 	if (hostIsNew != 0) {
1398 		struct epoch_tracker et;
1399 
1400 		NET_EPOCH_ENTER(et);
1401 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1402 			if (ifa->ifa_addr->sa_family != AF_INET6)
1403 				continue;
1404 			ifacount++;
1405 		}
1406 		NET_EPOCH_EXIT(et);
1407 	}
1408 
1409 	if (ifacount <= 1 && ifp->if_ioctl) {
1410 		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
1411 		if (error)
1412 			goto done;
1413 	}
1414 
1415 	/*
1416 	 * If a new destination address is specified, scrub the old one and
1417 	 * install the new destination.  Note that the interface must be
1418 	 * p2p or loopback.
1419 	 */
1420 	pdst = &ifra->ifra_dstaddr;
1421 	if (pdst->sin6_family == AF_INET6 &&
1422 	    !IN6_ARE_ADDR_EQUAL(&pdst->sin6_addr, &ia->ia_dstaddr.sin6_addr)) {
1423 		if ((ia->ia_flags & IFA_ROUTE) != 0 &&
1424 		    (rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST) != 0)) {
1425 			nd6log((LOG_ERR, "in6_update_ifa_internal: failed to "
1426 			    "remove a route to the old destination: %s\n",
1427 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1428 			/* proceed anyway... */
1429 		} else
1430 			ia->ia_flags &= ~IFA_ROUTE;
1431 		ia->ia_dstaddr = *pdst;
1432 	}
1433 
1434 	/*
1435 	 * If a new destination address is specified for a point-to-point
1436 	 * interface, install a route to the destination as an interface
1437 	 * direct route.
1438 	 * XXX: the logic below rejects assigning multiple addresses on a p2p
1439 	 * interface that share the same destination.
1440 	 */
1441 	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1442 	if (!(ia->ia_flags & IFA_ROUTE) && plen == 128 &&
1443 	    ia->ia_dstaddr.sin6_family == AF_INET6) {
1444 		int rtflags = RTF_UP | RTF_HOST;
1445 		/*
1446 		 * Handle the case for ::1 .
1447 		 */
1448 		if (ifp->if_flags & IFF_LOOPBACK)
1449 			ia->ia_flags |= IFA_RTSELF;
1450 		error = rtinit(&ia->ia_ifa, RTM_ADD, ia->ia_flags | rtflags);
1451 		if (error)
1452 			goto done;
1453 		ia->ia_flags |= IFA_ROUTE;
1454 	}
1455 
1456 	/*
1457 	 * add a loopback route to self if not exists
1458 	 */
1459 	if (!(ia->ia_flags & IFA_RTSELF) && V_nd6_useloopback) {
1460 		error = ifa_add_loopback_route((struct ifaddr *)ia,
1461 		    (struct sockaddr *)&ia->ia_addr);
1462 		if (error == 0)
1463 			ia->ia_flags |= IFA_RTSELF;
1464 	}
1465 done:
1466 	WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
1467 	    "Invoking IPv6 network device address event may sleep");
1468 
1469 	ifa_ref(&ia->ia_ifa);
1470 	EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, &ia->ia_ifa,
1471 	    IFADDR_EVENT_ADD);
1472 	ifa_free(&ia->ia_ifa);
1473 
1474 	return (error);
1475 }
1476 
1477 /*
1478  * Find an IPv6 interface link-local address specific to an interface.
1479  * ifaddr is returned referenced.
1480  */
1481 struct in6_ifaddr *
1482 in6ifa_ifpforlinklocal(struct ifnet *ifp, int ignoreflags)
1483 {
1484 	struct ifaddr *ifa;
1485 
1486 	NET_EPOCH_ASSERT();
1487 
1488 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1489 		if (ifa->ifa_addr->sa_family != AF_INET6)
1490 			continue;
1491 		if (IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa))) {
1492 			if ((((struct in6_ifaddr *)ifa)->ia6_flags &
1493 			    ignoreflags) != 0)
1494 				continue;
1495 			ifa_ref(ifa);
1496 			break;
1497 		}
1498 	}
1499 
1500 	return ((struct in6_ifaddr *)ifa);
1501 }
1502 
1503 
1504 /*
1505  * find the interface address corresponding to a given IPv6 address.
1506  * ifaddr is returned referenced.
1507  */
1508 struct in6_ifaddr *
1509 in6ifa_ifwithaddr(const struct in6_addr *addr, uint32_t zoneid)
1510 {
1511 	struct rm_priotracker in6_ifa_tracker;
1512 	struct in6_ifaddr *ia;
1513 
1514 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1515 	CK_LIST_FOREACH(ia, IN6ADDR_HASH(addr), ia6_hash) {
1516 		if (IN6_ARE_ADDR_EQUAL(IA6_IN6(ia), addr)) {
1517 			if (zoneid != 0 &&
1518 			    zoneid != ia->ia_addr.sin6_scope_id)
1519 				continue;
1520 			ifa_ref(&ia->ia_ifa);
1521 			break;
1522 		}
1523 	}
1524 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1525 	return (ia);
1526 }
1527 
1528 /*
1529  * find the internet address corresponding to a given interface and address.
1530  * ifaddr is returned referenced.
1531  */
1532 struct in6_ifaddr *
1533 in6ifa_ifpwithaddr(struct ifnet *ifp, const struct in6_addr *addr)
1534 {
1535 	struct epoch_tracker et;
1536 	struct ifaddr *ifa;
1537 
1538 	NET_EPOCH_ENTER(et);
1539 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1540 		if (ifa->ifa_addr->sa_family != AF_INET6)
1541 			continue;
1542 		if (IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa))) {
1543 			ifa_ref(ifa);
1544 			break;
1545 		}
1546 	}
1547 	NET_EPOCH_EXIT(et);
1548 
1549 	return ((struct in6_ifaddr *)ifa);
1550 }
1551 
1552 /*
1553  * Find a link-local scoped address on ifp and return it if any.
1554  */
1555 struct in6_ifaddr *
1556 in6ifa_llaonifp(struct ifnet *ifp)
1557 {
1558 	struct epoch_tracker et;
1559 	struct sockaddr_in6 *sin6;
1560 	struct ifaddr *ifa;
1561 
1562 
1563 	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)
1564 		return (NULL);
1565 	NET_EPOCH_ENTER(et);
1566 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1567 		if (ifa->ifa_addr->sa_family != AF_INET6)
1568 			continue;
1569 		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
1570 		if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr) ||
1571 		    IN6_IS_ADDR_MC_INTFACELOCAL(&sin6->sin6_addr) ||
1572 		    IN6_IS_ADDR_MC_NODELOCAL(&sin6->sin6_addr))
1573 			break;
1574 	}
1575 	NET_EPOCH_EXIT(et);
1576 
1577 	return ((struct in6_ifaddr *)ifa);
1578 }
1579 
1580 /*
1581  * Convert IP6 address to printable (loggable) representation. Caller
1582  * has to make sure that ip6buf is at least INET6_ADDRSTRLEN long.
1583  */
1584 static char digits[] = "0123456789abcdef";
1585 char *
1586 ip6_sprintf(char *ip6buf, const struct in6_addr *addr)
1587 {
1588 	int i, cnt = 0, maxcnt = 0, idx = 0, index = 0;
1589 	char *cp;
1590 	const u_int16_t *a = (const u_int16_t *)addr;
1591 	const u_int8_t *d;
1592 	int dcolon = 0, zero = 0;
1593 
1594 	cp = ip6buf;
1595 
1596 	for (i = 0; i < 8; i++) {
1597 		if (*(a + i) == 0) {
1598 			cnt++;
1599 			if (cnt == 1)
1600 				idx = i;
1601 		}
1602 		else if (maxcnt < cnt) {
1603 			maxcnt = cnt;
1604 			index = idx;
1605 			cnt = 0;
1606 		}
1607 	}
1608 	if (maxcnt < cnt) {
1609 		maxcnt = cnt;
1610 		index = idx;
1611 	}
1612 
1613 	for (i = 0; i < 8; i++) {
1614 		if (dcolon == 1) {
1615 			if (*a == 0) {
1616 				if (i == 7)
1617 					*cp++ = ':';
1618 				a++;
1619 				continue;
1620 			} else
1621 				dcolon = 2;
1622 		}
1623 		if (*a == 0) {
1624 			if (dcolon == 0 && *(a + 1) == 0 && i == index) {
1625 				if (i == 0)
1626 					*cp++ = ':';
1627 				*cp++ = ':';
1628 				dcolon = 1;
1629 			} else {
1630 				*cp++ = '0';
1631 				*cp++ = ':';
1632 			}
1633 			a++;
1634 			continue;
1635 		}
1636 		d = (const u_char *)a;
1637 		/* Try to eliminate leading zeros in printout like in :0001. */
1638 		zero = 1;
1639 		*cp = digits[*d >> 4];
1640 		if (*cp != '0') {
1641 			zero = 0;
1642 			cp++;
1643 		}
1644 		*cp = digits[*d++ & 0xf];
1645 		if (zero == 0 || (*cp != '0')) {
1646 			zero = 0;
1647 			cp++;
1648 		}
1649 		*cp = digits[*d >> 4];
1650 		if (zero == 0 || (*cp != '0')) {
1651 			zero = 0;
1652 			cp++;
1653 		}
1654 		*cp++ = digits[*d & 0xf];
1655 		*cp++ = ':';
1656 		a++;
1657 	}
1658 	*--cp = '\0';
1659 	return (ip6buf);
1660 }
1661 
1662 int
1663 in6_localaddr(struct in6_addr *in6)
1664 {
1665 	struct rm_priotracker in6_ifa_tracker;
1666 	struct in6_ifaddr *ia;
1667 
1668 	if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
1669 		return 1;
1670 
1671 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1672 	CK_STAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
1673 		if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
1674 		    &ia->ia_prefixmask.sin6_addr)) {
1675 			IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1676 			return 1;
1677 		}
1678 	}
1679 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1680 
1681 	return (0);
1682 }
1683 
1684 /*
1685  * Return 1 if an internet address is for the local host and configured
1686  * on one of its interfaces.
1687  */
1688 int
1689 in6_localip(struct in6_addr *in6)
1690 {
1691 	struct rm_priotracker in6_ifa_tracker;
1692 	struct in6_ifaddr *ia;
1693 
1694 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1695 	CK_LIST_FOREACH(ia, IN6ADDR_HASH(in6), ia6_hash) {
1696 		if (IN6_ARE_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr)) {
1697 			IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1698 			return (1);
1699 		}
1700 	}
1701 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1702 	return (0);
1703 }
1704 
1705 /*
1706  * Return 1 if an internet address is configured on an interface.
1707  */
1708 int
1709 in6_ifhasaddr(struct ifnet *ifp, struct in6_addr *addr)
1710 {
1711 	struct in6_addr in6;
1712 	struct ifaddr *ifa;
1713 	struct in6_ifaddr *ia6;
1714 
1715 	NET_EPOCH_ASSERT();
1716 
1717 	in6 = *addr;
1718 	if (in6_clearscope(&in6))
1719 		return (0);
1720 	in6_setscope(&in6, ifp, NULL);
1721 
1722 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1723 		if (ifa->ifa_addr->sa_family != AF_INET6)
1724 			continue;
1725 		ia6 = (struct in6_ifaddr *)ifa;
1726 		if (IN6_ARE_ADDR_EQUAL(&ia6->ia_addr.sin6_addr, &in6))
1727 			return (1);
1728 	}
1729 
1730 	return (0);
1731 }
1732 
1733 int
1734 in6_is_addr_deprecated(struct sockaddr_in6 *sa6)
1735 {
1736 	struct rm_priotracker in6_ifa_tracker;
1737 	struct in6_ifaddr *ia;
1738 
1739 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1740 	CK_LIST_FOREACH(ia, IN6ADDR_HASH(&sa6->sin6_addr), ia6_hash) {
1741 		if (IN6_ARE_ADDR_EQUAL(IA6_IN6(ia), &sa6->sin6_addr)) {
1742 			if (ia->ia6_flags & IN6_IFF_DEPRECATED) {
1743 				IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1744 				return (1); /* true */
1745 			}
1746 			break;
1747 		}
1748 	}
1749 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1750 
1751 	return (0);		/* false */
1752 }
1753 
1754 /*
1755  * return length of part which dst and src are equal
1756  * hard coding...
1757  */
1758 int
1759 in6_matchlen(struct in6_addr *src, struct in6_addr *dst)
1760 {
1761 	int match = 0;
1762 	u_char *s = (u_char *)src, *d = (u_char *)dst;
1763 	u_char *lim = s + 16, r;
1764 
1765 	while (s < lim)
1766 		if ((r = (*d++ ^ *s++)) != 0) {
1767 			while (r < 128) {
1768 				match++;
1769 				r <<= 1;
1770 			}
1771 			break;
1772 		} else
1773 			match += 8;
1774 	return match;
1775 }
1776 
1777 /* XXX: to be scope conscious */
1778 int
1779 in6_are_prefix_equal(struct in6_addr *p1, struct in6_addr *p2, int len)
1780 {
1781 	int bytelen, bitlen;
1782 
1783 	/* sanity check */
1784 	if (0 > len || len > 128) {
1785 		log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n",
1786 		    len);
1787 		return (0);
1788 	}
1789 
1790 	bytelen = len / 8;
1791 	bitlen = len % 8;
1792 
1793 	if (bcmp(&p1->s6_addr, &p2->s6_addr, bytelen))
1794 		return (0);
1795 	if (bitlen != 0 &&
1796 	    p1->s6_addr[bytelen] >> (8 - bitlen) !=
1797 	    p2->s6_addr[bytelen] >> (8 - bitlen))
1798 		return (0);
1799 
1800 	return (1);
1801 }
1802 
1803 void
1804 in6_prefixlen2mask(struct in6_addr *maskp, int len)
1805 {
1806 	u_char maskarray[8] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
1807 	int bytelen, bitlen, i;
1808 
1809 	/* sanity check */
1810 	if (0 > len || len > 128) {
1811 		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
1812 		    len);
1813 		return;
1814 	}
1815 
1816 	bzero(maskp, sizeof(*maskp));
1817 	bytelen = len / 8;
1818 	bitlen = len % 8;
1819 	for (i = 0; i < bytelen; i++)
1820 		maskp->s6_addr[i] = 0xff;
1821 	if (bitlen)
1822 		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
1823 }
1824 
1825 /*
1826  * return the best address out of the same scope. if no address was
1827  * found, return the first valid address from designated IF.
1828  */
1829 struct in6_ifaddr *
1830 in6_ifawithifp(struct ifnet *ifp, struct in6_addr *dst)
1831 {
1832 	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
1833 	struct ifaddr *ifa;
1834 	struct in6_ifaddr *besta = NULL;
1835 	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
1836 
1837 	NET_EPOCH_ASSERT();
1838 
1839 	dep[0] = dep[1] = NULL;
1840 
1841 	/*
1842 	 * We first look for addresses in the same scope.
1843 	 * If there is one, return it.
1844 	 * If two or more, return one which matches the dst longest.
1845 	 * If none, return one of global addresses assigned other ifs.
1846 	 */
1847 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1848 		if (ifa->ifa_addr->sa_family != AF_INET6)
1849 			continue;
1850 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
1851 			continue; /* XXX: is there any case to allow anycast? */
1852 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
1853 			continue; /* don't use this interface */
1854 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
1855 			continue;
1856 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
1857 			if (V_ip6_use_deprecated)
1858 				dep[0] = (struct in6_ifaddr *)ifa;
1859 			continue;
1860 		}
1861 
1862 		if (dst_scope == in6_addrscope(IFA_IN6(ifa))) {
1863 			/*
1864 			 * call in6_matchlen() as few as possible
1865 			 */
1866 			if (besta) {
1867 				if (blen == -1)
1868 					blen = in6_matchlen(&besta->ia_addr.sin6_addr, dst);
1869 				tlen = in6_matchlen(IFA_IN6(ifa), dst);
1870 				if (tlen > blen) {
1871 					blen = tlen;
1872 					besta = (struct in6_ifaddr *)ifa;
1873 				}
1874 			} else
1875 				besta = (struct in6_ifaddr *)ifa;
1876 		}
1877 	}
1878 	if (besta) {
1879 		ifa_ref(&besta->ia_ifa);
1880 		return (besta);
1881 	}
1882 
1883 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1884 		if (ifa->ifa_addr->sa_family != AF_INET6)
1885 			continue;
1886 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
1887 			continue; /* XXX: is there any case to allow anycast? */
1888 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
1889 			continue; /* don't use this interface */
1890 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
1891 			continue;
1892 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
1893 			if (V_ip6_use_deprecated)
1894 				dep[1] = (struct in6_ifaddr *)ifa;
1895 			continue;
1896 		}
1897 
1898 		if (ifa != NULL)
1899 			ifa_ref(ifa);
1900 		return (struct in6_ifaddr *)ifa;
1901 	}
1902 
1903 	/* use the last-resort values, that are, deprecated addresses */
1904 	if (dep[0]) {
1905 		ifa_ref((struct ifaddr *)dep[0]);
1906 		return dep[0];
1907 	}
1908 	if (dep[1]) {
1909 		ifa_ref((struct ifaddr *)dep[1]);
1910 		return dep[1];
1911 	}
1912 
1913 	return NULL;
1914 }
1915 
1916 /*
1917  * perform DAD when interface becomes IFF_UP.
1918  */
1919 void
1920 in6_if_up(struct ifnet *ifp)
1921 {
1922 	struct epoch_tracker et;
1923 	struct ifaddr *ifa;
1924 	struct in6_ifaddr *ia;
1925 
1926 	NET_EPOCH_ENTER(et);
1927 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1928 		if (ifa->ifa_addr->sa_family != AF_INET6)
1929 			continue;
1930 		ia = (struct in6_ifaddr *)ifa;
1931 		if (ia->ia6_flags & IN6_IFF_TENTATIVE) {
1932 			/*
1933 			 * The TENTATIVE flag was likely set by hand
1934 			 * beforehand, implicitly indicating the need for DAD.
1935 			 * We may be able to skip the random delay in this
1936 			 * case, but we impose delays just in case.
1937 			 */
1938 			nd6_dad_start(ifa,
1939 			    arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz));
1940 		}
1941 	}
1942 	NET_EPOCH_EXIT(et);
1943 
1944 	/*
1945 	 * special cases, like 6to4, are handled in in6_ifattach
1946 	 */
1947 	in6_ifattach(ifp, NULL);
1948 }
1949 
1950 int
1951 in6if_do_dad(struct ifnet *ifp)
1952 {
1953 
1954 	if ((ifp->if_flags & IFF_LOOPBACK) != 0)
1955 		return (0);
1956 	if ((ifp->if_flags & IFF_MULTICAST) == 0)
1957 		return (0);
1958 	if ((ND_IFINFO(ifp)->flags &
1959 	    (ND6_IFF_IFDISABLED | ND6_IFF_NO_DAD)) != 0)
1960 		return (0);
1961 	return (1);
1962 }
1963 
1964 /*
1965  * Calculate max IPv6 MTU through all the interfaces and store it
1966  * to in6_maxmtu.
1967  */
1968 void
1969 in6_setmaxmtu(void)
1970 {
1971 	struct epoch_tracker et;
1972 	unsigned long maxmtu = 0;
1973 	struct ifnet *ifp;
1974 
1975 	NET_EPOCH_ENTER(et);
1976 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1977 		/* this function can be called during ifnet initialization */
1978 		if (!ifp->if_afdata[AF_INET6])
1979 			continue;
1980 		if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
1981 		    IN6_LINKMTU(ifp) > maxmtu)
1982 			maxmtu = IN6_LINKMTU(ifp);
1983 	}
1984 	NET_EPOCH_EXIT(et);
1985 	if (maxmtu)	/* update only when maxmtu is positive */
1986 		V_in6_maxmtu = maxmtu;
1987 }
1988 
1989 /*
1990  * Provide the length of interface identifiers to be used for the link attached
1991  * to the given interface.  The length should be defined in "IPv6 over
1992  * xxx-link" document.  Note that address architecture might also define
1993  * the length for a particular set of address prefixes, regardless of the
1994  * link type.  As clarified in rfc2462bis, those two definitions should be
1995  * consistent, and those really are as of August 2004.
1996  */
1997 int
1998 in6_if2idlen(struct ifnet *ifp)
1999 {
2000 	switch (ifp->if_type) {
2001 	case IFT_ETHER:		/* RFC2464 */
2002 	case IFT_PROPVIRTUAL:	/* XXX: no RFC. treat it as ether */
2003 	case IFT_L2VLAN:	/* ditto */
2004 	case IFT_BRIDGE:	/* bridge(4) only does Ethernet-like links */
2005 	case IFT_INFINIBAND:
2006 		return (64);
2007 	case IFT_PPP:		/* RFC2472 */
2008 		return (64);
2009 	case IFT_FRELAY:	/* RFC2590 */
2010 		return (64);
2011 	case IFT_IEEE1394:	/* RFC3146 */
2012 		return (64);
2013 	case IFT_GIF:
2014 		return (64);	/* draft-ietf-v6ops-mech-v2-07 */
2015 	case IFT_LOOP:
2016 		return (64);	/* XXX: is this really correct? */
2017 	default:
2018 		/*
2019 		 * Unknown link type:
2020 		 * It might be controversial to use the today's common constant
2021 		 * of 64 for these cases unconditionally.  For full compliance,
2022 		 * we should return an error in this case.  On the other hand,
2023 		 * if we simply miss the standard for the link type or a new
2024 		 * standard is defined for a new link type, the IFID length
2025 		 * is very likely to be the common constant.  As a compromise,
2026 		 * we always use the constant, but make an explicit notice
2027 		 * indicating the "unknown" case.
2028 		 */
2029 		printf("in6_if2idlen: unknown link type (%d)\n", ifp->if_type);
2030 		return (64);
2031 	}
2032 }
2033 
2034 struct in6_llentry {
2035 	struct llentry		base;
2036 };
2037 
2038 #define	IN6_LLTBL_DEFAULT_HSIZE	32
2039 #define	IN6_LLTBL_HASH(k, h) \
2040 	(((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1))
2041 
2042 /*
2043  * Do actual deallocation of @lle.
2044  */
2045 static void
2046 in6_lltable_destroy_lle_unlocked(epoch_context_t ctx)
2047 {
2048 	struct llentry *lle;
2049 
2050 	lle = __containerof(ctx, struct llentry, lle_epoch_ctx);
2051 	LLE_LOCK_DESTROY(lle);
2052 	LLE_REQ_DESTROY(lle);
2053 	free(lle, M_LLTABLE);
2054 }
2055 
2056 /*
2057  * Called by LLE_FREE_LOCKED when number of references
2058  * drops to zero.
2059  */
2060 static void
2061 in6_lltable_destroy_lle(struct llentry *lle)
2062 {
2063 
2064 	LLE_WUNLOCK(lle);
2065 	NET_EPOCH_CALL(in6_lltable_destroy_lle_unlocked, &lle->lle_epoch_ctx);
2066 }
2067 
2068 static struct llentry *
2069 in6_lltable_new(const struct in6_addr *addr6, u_int flags)
2070 {
2071 	struct in6_llentry *lle;
2072 
2073 	lle = malloc(sizeof(struct in6_llentry), M_LLTABLE, M_NOWAIT | M_ZERO);
2074 	if (lle == NULL)		/* NB: caller generates msg */
2075 		return NULL;
2076 
2077 	lle->base.r_l3addr.addr6 = *addr6;
2078 	lle->base.lle_refcnt = 1;
2079 	lle->base.lle_free = in6_lltable_destroy_lle;
2080 	LLE_LOCK_INIT(&lle->base);
2081 	LLE_REQ_INIT(&lle->base);
2082 	callout_init(&lle->base.lle_timer, 1);
2083 
2084 	return (&lle->base);
2085 }
2086 
2087 static int
2088 in6_lltable_match_prefix(const struct sockaddr *saddr,
2089     const struct sockaddr *smask, u_int flags, struct llentry *lle)
2090 {
2091 	const struct in6_addr *addr, *mask, *lle_addr;
2092 
2093 	addr = &((const struct sockaddr_in6 *)saddr)->sin6_addr;
2094 	mask = &((const struct sockaddr_in6 *)smask)->sin6_addr;
2095 	lle_addr = &lle->r_l3addr.addr6;
2096 
2097 	if (IN6_ARE_MASKED_ADDR_EQUAL(lle_addr, addr, mask) == 0)
2098 		return (0);
2099 
2100 	if (lle->la_flags & LLE_IFADDR) {
2101 
2102 		/*
2103 		 * Delete LLE_IFADDR records IFF address & flag matches.
2104 		 * Note that addr is the interface address within prefix
2105 		 * being matched.
2106 		 */
2107 		if (IN6_ARE_ADDR_EQUAL(addr, lle_addr) &&
2108 		    (flags & LLE_STATIC) != 0)
2109 			return (1);
2110 		return (0);
2111 	}
2112 
2113 	/* flags & LLE_STATIC means deleting both dynamic and static entries */
2114 	if ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC))
2115 		return (1);
2116 
2117 	return (0);
2118 }
2119 
2120 static void
2121 in6_lltable_free_entry(struct lltable *llt, struct llentry *lle)
2122 {
2123 	struct ifnet *ifp;
2124 
2125 	LLE_WLOCK_ASSERT(lle);
2126 	KASSERT(llt != NULL, ("lltable is NULL"));
2127 
2128 	/* Unlink entry from table */
2129 	if ((lle->la_flags & LLE_LINKED) != 0) {
2130 
2131 		ifp = llt->llt_ifp;
2132 		IF_AFDATA_WLOCK_ASSERT(ifp);
2133 		lltable_unlink_entry(llt, lle);
2134 	}
2135 
2136 	llentry_free(lle);
2137 }
2138 
2139 static int
2140 in6_lltable_rtcheck(struct ifnet *ifp,
2141 		    u_int flags,
2142 		    const struct sockaddr *l3addr)
2143 {
2144 	const struct sockaddr_in6 *sin6;
2145 	struct nhop_object *nh;
2146 	struct in6_addr dst;
2147 	uint32_t scopeid;
2148 	char ip6buf[INET6_ADDRSTRLEN];
2149 	int fibnum;
2150 
2151 	NET_EPOCH_ASSERT();
2152 	KASSERT(l3addr->sa_family == AF_INET6,
2153 	    ("sin_family %d", l3addr->sa_family));
2154 
2155 	sin6 = (const struct sockaddr_in6 *)l3addr;
2156 	in6_splitscope(&sin6->sin6_addr, &dst, &scopeid);
2157 	fibnum = V_rt_add_addr_allfibs ? RT_DEFAULT_FIB : ifp->if_fib;
2158 	nh = fib6_lookup(fibnum, &dst, scopeid, NHR_NONE, 0);
2159 	if (nh && ((nh->nh_flags & NHF_GATEWAY) || nh->nh_ifp != ifp)) {
2160 		struct ifaddr *ifa;
2161 		/*
2162 		 * Create an ND6 cache for an IPv6 neighbor
2163 		 * that is not covered by our own prefix.
2164 		 */
2165 		ifa = ifaof_ifpforaddr(l3addr, ifp);
2166 		if (ifa != NULL) {
2167 			return 0;
2168 		}
2169 		log(LOG_INFO, "IPv6 address: \"%s\" is not on the network\n",
2170 		    ip6_sprintf(ip6buf, &sin6->sin6_addr));
2171 		return EINVAL;
2172 	}
2173 	return 0;
2174 }
2175 
2176 /*
2177  * Called by the datapath to indicate that the entry was used.
2178  */
2179 static void
2180 in6_lltable_mark_used(struct llentry *lle)
2181 {
2182 
2183 	LLE_REQ_LOCK(lle);
2184 	lle->r_skip_req = 0;
2185 
2186 	/*
2187 	 * Set the hit time so the callback function
2188 	 * can determine the remaining time before
2189 	 * transiting to the DELAY state.
2190 	 */
2191 	lle->lle_hittime = time_uptime;
2192 	LLE_REQ_UNLOCK(lle);
2193 }
2194 
2195 static inline uint32_t
2196 in6_lltable_hash_dst(const struct in6_addr *dst, uint32_t hsize)
2197 {
2198 
2199 	return (IN6_LLTBL_HASH(dst->s6_addr32[3], hsize));
2200 }
2201 
2202 static uint32_t
2203 in6_lltable_hash(const struct llentry *lle, uint32_t hsize)
2204 {
2205 
2206 	return (in6_lltable_hash_dst(&lle->r_l3addr.addr6, hsize));
2207 }
2208 
2209 static void
2210 in6_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
2211 {
2212 	struct sockaddr_in6 *sin6;
2213 
2214 	sin6 = (struct sockaddr_in6 *)sa;
2215 	bzero(sin6, sizeof(*sin6));
2216 	sin6->sin6_family = AF_INET6;
2217 	sin6->sin6_len = sizeof(*sin6);
2218 	sin6->sin6_addr = lle->r_l3addr.addr6;
2219 }
2220 
2221 static inline struct llentry *
2222 in6_lltable_find_dst(struct lltable *llt, const struct in6_addr *dst)
2223 {
2224 	struct llentry *lle;
2225 	struct llentries *lleh;
2226 	u_int hashidx;
2227 
2228 	hashidx = in6_lltable_hash_dst(dst, llt->llt_hsize);
2229 	lleh = &llt->lle_head[hashidx];
2230 	CK_LIST_FOREACH(lle, lleh, lle_next) {
2231 		if (lle->la_flags & LLE_DELETED)
2232 			continue;
2233 		if (IN6_ARE_ADDR_EQUAL(&lle->r_l3addr.addr6, dst))
2234 			break;
2235 	}
2236 
2237 	return (lle);
2238 }
2239 
2240 static void
2241 in6_lltable_delete_entry(struct lltable *llt, struct llentry *lle)
2242 {
2243 
2244 	lle->la_flags |= LLE_DELETED;
2245 	EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_DELETED);
2246 #ifdef DIAGNOSTIC
2247 	log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle);
2248 #endif
2249 	llentry_free(lle);
2250 }
2251 
2252 static struct llentry *
2253 in6_lltable_alloc(struct lltable *llt, u_int flags,
2254 	const struct sockaddr *l3addr)
2255 {
2256 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2257 	struct ifnet *ifp = llt->llt_ifp;
2258 	struct llentry *lle;
2259 	char linkhdr[LLE_MAX_LINKHDR];
2260 	size_t linkhdrsize;
2261 	int lladdr_off;
2262 
2263 	KASSERT(l3addr->sa_family == AF_INET6,
2264 	    ("sin_family %d", l3addr->sa_family));
2265 
2266 	/*
2267 	 * A route that covers the given address must have
2268 	 * been installed 1st because we are doing a resolution,
2269 	 * verify this.
2270 	 */
2271 	if (!(flags & LLE_IFADDR) &&
2272 	    in6_lltable_rtcheck(ifp, flags, l3addr) != 0)
2273 		return (NULL);
2274 
2275 	lle = in6_lltable_new(&sin6->sin6_addr, flags);
2276 	if (lle == NULL) {
2277 		log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
2278 		return (NULL);
2279 	}
2280 	lle->la_flags = flags;
2281 	if ((flags & LLE_IFADDR) == LLE_IFADDR) {
2282 		linkhdrsize = LLE_MAX_LINKHDR;
2283 		if (lltable_calc_llheader(ifp, AF_INET6, IF_LLADDR(ifp),
2284 		    linkhdr, &linkhdrsize, &lladdr_off) != 0) {
2285 			NET_EPOCH_CALL(in6_lltable_destroy_lle_unlocked, &lle->lle_epoch_ctx);
2286 			return (NULL);
2287 		}
2288 		lltable_set_entry_addr(ifp, lle, linkhdr, linkhdrsize,
2289 		    lladdr_off);
2290 		lle->la_flags |= LLE_STATIC;
2291 	}
2292 
2293 	if ((lle->la_flags & LLE_STATIC) != 0)
2294 		lle->ln_state = ND6_LLINFO_REACHABLE;
2295 
2296 	return (lle);
2297 }
2298 
2299 static struct llentry *
2300 in6_lltable_lookup(struct lltable *llt, u_int flags,
2301 	const struct sockaddr *l3addr)
2302 {
2303 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2304 	struct llentry *lle;
2305 
2306 	IF_AFDATA_LOCK_ASSERT(llt->llt_ifp);
2307 	KASSERT(l3addr->sa_family == AF_INET6,
2308 	    ("sin_family %d", l3addr->sa_family));
2309 	KASSERT((flags & (LLE_UNLOCKED | LLE_EXCLUSIVE)) !=
2310 	    (LLE_UNLOCKED | LLE_EXCLUSIVE),
2311 	    ("wrong lle request flags: %#x", flags));
2312 
2313 	lle = in6_lltable_find_dst(llt, &sin6->sin6_addr);
2314 	if (lle == NULL)
2315 		return (NULL);
2316 	if (flags & LLE_UNLOCKED)
2317 		return (lle);
2318 
2319 	if (flags & LLE_EXCLUSIVE)
2320 		LLE_WLOCK(lle);
2321 	else
2322 		LLE_RLOCK(lle);
2323 
2324 	/*
2325 	 * If the afdata lock is not held, the LLE may have been unlinked while
2326 	 * we were blocked on the LLE lock.  Check for this case.
2327 	 */
2328 	if (__predict_false((lle->la_flags & LLE_LINKED) == 0)) {
2329 		if (flags & LLE_EXCLUSIVE)
2330 			LLE_WUNLOCK(lle);
2331 		else
2332 			LLE_RUNLOCK(lle);
2333 		return (NULL);
2334 	}
2335 	return (lle);
2336 }
2337 
2338 static int
2339 in6_lltable_dump_entry(struct lltable *llt, struct llentry *lle,
2340     struct sysctl_req *wr)
2341 {
2342 	struct ifnet *ifp = llt->llt_ifp;
2343 	/* XXX stack use */
2344 	struct {
2345 		struct rt_msghdr	rtm;
2346 		struct sockaddr_in6	sin6;
2347 		/*
2348 		 * ndp.c assumes that sdl is word aligned
2349 		 */
2350 #ifdef __LP64__
2351 		uint32_t		pad;
2352 #endif
2353 		struct sockaddr_dl	sdl;
2354 	} ndpc;
2355 	struct sockaddr_dl *sdl;
2356 	int error;
2357 
2358 	bzero(&ndpc, sizeof(ndpc));
2359 	/* skip deleted entries */
2360 	if ((lle->la_flags & LLE_DELETED) == LLE_DELETED)
2361 		return (0);
2362 	/* Skip if jailed and not a valid IP of the prison. */
2363 	lltable_fill_sa_entry(lle, (struct sockaddr *)&ndpc.sin6);
2364 	if (prison_if(wr->td->td_ucred, (struct sockaddr *)&ndpc.sin6) != 0)
2365 		return (0);
2366 	/*
2367 	 * produce a msg made of:
2368 	 *  struct rt_msghdr;
2369 	 *  struct sockaddr_in6 (IPv6)
2370 	 *  struct sockaddr_dl;
2371 	 */
2372 	ndpc.rtm.rtm_msglen = sizeof(ndpc);
2373 	ndpc.rtm.rtm_version = RTM_VERSION;
2374 	ndpc.rtm.rtm_type = RTM_GET;
2375 	ndpc.rtm.rtm_flags = RTF_UP;
2376 	ndpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
2377 	if (V_deembed_scopeid)
2378 		sa6_recoverscope(&ndpc.sin6);
2379 
2380 	/* publish */
2381 	if (lle->la_flags & LLE_PUB)
2382 		ndpc.rtm.rtm_flags |= RTF_ANNOUNCE;
2383 
2384 	sdl = &ndpc.sdl;
2385 	sdl->sdl_family = AF_LINK;
2386 	sdl->sdl_len = sizeof(*sdl);
2387 	sdl->sdl_index = ifp->if_index;
2388 	sdl->sdl_type = ifp->if_type;
2389 	if ((lle->la_flags & LLE_VALID) == LLE_VALID) {
2390 		sdl->sdl_alen = ifp->if_addrlen;
2391 		bcopy(lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
2392 	} else {
2393 		sdl->sdl_alen = 0;
2394 		bzero(LLADDR(sdl), ifp->if_addrlen);
2395 	}
2396 	if (lle->la_expire != 0)
2397 		ndpc.rtm.rtm_rmx.rmx_expire = lle->la_expire +
2398 		    lle->lle_remtime / hz + time_second - time_uptime;
2399 	ndpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
2400 	if (lle->la_flags & LLE_STATIC)
2401 		ndpc.rtm.rtm_flags |= RTF_STATIC;
2402 	if (lle->la_flags & LLE_IFADDR)
2403 		ndpc.rtm.rtm_flags |= RTF_PINNED;
2404 	if (lle->ln_router != 0)
2405 		ndpc.rtm.rtm_flags |= RTF_GATEWAY;
2406 	ndpc.rtm.rtm_rmx.rmx_pksent = lle->la_asked;
2407 	/* Store state in rmx_weight value */
2408 	ndpc.rtm.rtm_rmx.rmx_state = lle->ln_state;
2409 	ndpc.rtm.rtm_index = ifp->if_index;
2410 	error = SYSCTL_OUT(wr, &ndpc, sizeof(ndpc));
2411 
2412 	return (error);
2413 }
2414 
2415 static struct lltable *
2416 in6_lltattach(struct ifnet *ifp)
2417 {
2418 	struct lltable *llt;
2419 
2420 	llt = lltable_allocate_htbl(IN6_LLTBL_DEFAULT_HSIZE);
2421 	llt->llt_af = AF_INET6;
2422 	llt->llt_ifp = ifp;
2423 
2424 	llt->llt_lookup = in6_lltable_lookup;
2425 	llt->llt_alloc_entry = in6_lltable_alloc;
2426 	llt->llt_delete_entry = in6_lltable_delete_entry;
2427 	llt->llt_dump_entry = in6_lltable_dump_entry;
2428 	llt->llt_hash = in6_lltable_hash;
2429 	llt->llt_fill_sa_entry = in6_lltable_fill_sa_entry;
2430 	llt->llt_free_entry = in6_lltable_free_entry;
2431 	llt->llt_match_prefix = in6_lltable_match_prefix;
2432 	llt->llt_mark_used = in6_lltable_mark_used;
2433  	lltable_link(llt);
2434 
2435 	return (llt);
2436 }
2437 
2438 void *
2439 in6_domifattach(struct ifnet *ifp)
2440 {
2441 	struct in6_ifextra *ext;
2442 
2443 	/* There are not IPv6-capable interfaces. */
2444 	switch (ifp->if_type) {
2445 	case IFT_PFLOG:
2446 	case IFT_PFSYNC:
2447 	case IFT_USB:
2448 		return (NULL);
2449 	}
2450 	ext = (struct in6_ifextra *)malloc(sizeof(*ext), M_IFADDR, M_WAITOK);
2451 	bzero(ext, sizeof(*ext));
2452 
2453 	ext->in6_ifstat = malloc(sizeof(counter_u64_t) *
2454 	    sizeof(struct in6_ifstat) / sizeof(uint64_t), M_IFADDR, M_WAITOK);
2455 	COUNTER_ARRAY_ALLOC(ext->in6_ifstat,
2456 	    sizeof(struct in6_ifstat) / sizeof(uint64_t), M_WAITOK);
2457 
2458 	ext->icmp6_ifstat = malloc(sizeof(counter_u64_t) *
2459 	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t), M_IFADDR,
2460 	    M_WAITOK);
2461 	COUNTER_ARRAY_ALLOC(ext->icmp6_ifstat,
2462 	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t), M_WAITOK);
2463 
2464 	ext->nd_ifinfo = nd6_ifattach(ifp);
2465 	ext->scope6_id = scope6_ifattach(ifp);
2466 	ext->lltable = in6_lltattach(ifp);
2467 
2468 	ext->mld_ifinfo = mld_domifattach(ifp);
2469 
2470 	return ext;
2471 }
2472 
2473 int
2474 in6_domifmtu(struct ifnet *ifp)
2475 {
2476 	if (ifp->if_afdata[AF_INET6] == NULL)
2477 		return ifp->if_mtu;
2478 
2479 	return (IN6_LINKMTU(ifp));
2480 }
2481 
2482 void
2483 in6_domifdetach(struct ifnet *ifp, void *aux)
2484 {
2485 	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
2486 
2487 	mld_domifdetach(ifp);
2488 	scope6_ifdetach(ext->scope6_id);
2489 	nd6_ifdetach(ifp, ext->nd_ifinfo);
2490 	lltable_free(ext->lltable);
2491 	COUNTER_ARRAY_FREE(ext->in6_ifstat,
2492 	    sizeof(struct in6_ifstat) / sizeof(uint64_t));
2493 	free(ext->in6_ifstat, M_IFADDR);
2494 	COUNTER_ARRAY_FREE(ext->icmp6_ifstat,
2495 	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t));
2496 	free(ext->icmp6_ifstat, M_IFADDR);
2497 	free(ext, M_IFADDR);
2498 }
2499 
2500 /*
2501  * Convert sockaddr_in6 to sockaddr_in.  Original sockaddr_in6 must be
2502  * v4 mapped addr or v4 compat addr
2503  */
2504 void
2505 in6_sin6_2_sin(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2506 {
2507 
2508 	bzero(sin, sizeof(*sin));
2509 	sin->sin_len = sizeof(struct sockaddr_in);
2510 	sin->sin_family = AF_INET;
2511 	sin->sin_port = sin6->sin6_port;
2512 	sin->sin_addr.s_addr = sin6->sin6_addr.s6_addr32[3];
2513 }
2514 
2515 /* Convert sockaddr_in to sockaddr_in6 in v4 mapped addr format. */
2516 void
2517 in6_sin_2_v4mapsin6(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2518 {
2519 	bzero(sin6, sizeof(*sin6));
2520 	sin6->sin6_len = sizeof(struct sockaddr_in6);
2521 	sin6->sin6_family = AF_INET6;
2522 	sin6->sin6_port = sin->sin_port;
2523 	sin6->sin6_addr.s6_addr32[0] = 0;
2524 	sin6->sin6_addr.s6_addr32[1] = 0;
2525 	sin6->sin6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
2526 	sin6->sin6_addr.s6_addr32[3] = sin->sin_addr.s_addr;
2527 }
2528 
2529 /* Convert sockaddr_in6 into sockaddr_in. */
2530 void
2531 in6_sin6_2_sin_in_sock(struct sockaddr *nam)
2532 {
2533 	struct sockaddr_in *sin_p;
2534 	struct sockaddr_in6 sin6;
2535 
2536 	/*
2537 	 * Save original sockaddr_in6 addr and convert it
2538 	 * to sockaddr_in.
2539 	 */
2540 	sin6 = *(struct sockaddr_in6 *)nam;
2541 	sin_p = (struct sockaddr_in *)nam;
2542 	in6_sin6_2_sin(sin_p, &sin6);
2543 }
2544 
2545 /* Convert sockaddr_in into sockaddr_in6 in v4 mapped addr format. */
2546 void
2547 in6_sin_2_v4mapsin6_in_sock(struct sockaddr **nam)
2548 {
2549 	struct sockaddr_in *sin_p;
2550 	struct sockaddr_in6 *sin6_p;
2551 
2552 	sin6_p = malloc(sizeof *sin6_p, M_SONAME, M_WAITOK);
2553 	sin_p = (struct sockaddr_in *)*nam;
2554 	in6_sin_2_v4mapsin6(sin_p, sin6_p);
2555 	free(*nam, M_SONAME);
2556 	*nam = (struct sockaddr *)sin6_p;
2557 }
2558