xref: /freebsd/sys/netinet6/in6.c (revision 604d0dd551fa7d275cdf69911c9df4595a42f49f)
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/if_dl.h>
95 #include <net/vnet.h>
96 
97 #include <netinet/in.h>
98 #include <netinet/in_var.h>
99 #include <net/if_llatbl.h>
100 #include <netinet/if_ether.h>
101 #include <netinet/in_systm.h>
102 #include <netinet/ip.h>
103 #include <netinet/in_pcb.h>
104 #include <netinet/ip_carp.h>
105 
106 #include <netinet/ip6.h>
107 #include <netinet6/ip6_var.h>
108 #include <netinet6/nd6.h>
109 #include <netinet6/mld6_var.h>
110 #include <netinet6/ip6_mroute.h>
111 #include <netinet6/in6_ifattach.h>
112 #include <netinet6/scope6_var.h>
113 #include <netinet6/in6_fib.h>
114 #include <netinet6/in6_pcb.h>
115 
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 
168 void
169 in6_newaddrmsg(struct in6_ifaddr *ia, int cmd)
170 {
171 	struct rt_addrinfo info;
172 	struct ifaddr *ifa;
173 	struct sockaddr_dl gateway;
174 	int fibnum;
175 
176 	ifa = &ia->ia_ifa;
177 
178 	/*
179 	 * Prepare info data for the host route.
180 	 * This code mimics one from ifa_maintain_loopback_route().
181 	 */
182 	bzero(&info, sizeof(struct rt_addrinfo));
183 	info.rti_flags = ifa->ifa_flags | RTF_HOST | RTF_STATIC | RTF_PINNED;
184 	info.rti_info[RTAX_DST] = ifa->ifa_addr;
185 	info.rti_info[RTAX_GATEWAY] = (struct sockaddr *)&gateway;
186 	link_init_sdl(ifa->ifa_ifp, (struct sockaddr *)&gateway, ifa->ifa_ifp->if_type);
187 	if (cmd != RTM_DELETE)
188 		info.rti_ifp = V_loif;
189 
190 
191 	fibnum = V_rt_add_addr_allfibs ? RT_ALL_FIBS : ia62ifa(ia)->ifa_ifp->if_fib;
192 
193 	if (cmd == RTM_ADD) {
194 		rt_addrmsg(cmd, &ia->ia_ifa, fibnum);
195 		rt_routemsg_info(cmd, &info, fibnum);
196 	} else if (cmd == RTM_DELETE) {
197 		rt_routemsg_info(cmd, &info, fibnum);
198 		rt_addrmsg(cmd, &ia->ia_ifa, fibnum);
199 	}
200 }
201 
202 int
203 in6_mask2len(struct in6_addr *mask, u_char *lim0)
204 {
205 	int x = 0, y;
206 	u_char *lim = lim0, *p;
207 
208 	/* ignore the scope_id part */
209 	if (lim0 == NULL || lim0 - (u_char *)mask > sizeof(*mask))
210 		lim = (u_char *)mask + sizeof(*mask);
211 	for (p = (u_char *)mask; p < lim; x++, p++) {
212 		if (*p != 0xff)
213 			break;
214 	}
215 	y = 0;
216 	if (p < lim) {
217 		for (y = 0; y < 8; y++) {
218 			if ((*p & (0x80 >> y)) == 0)
219 				break;
220 		}
221 	}
222 
223 	/*
224 	 * when the limit pointer is given, do a stricter check on the
225 	 * remaining bits.
226 	 */
227 	if (p < lim) {
228 		if (y != 0 && (*p & (0x00ff >> y)) != 0)
229 			return (-1);
230 		for (p = p + 1; p < lim; p++)
231 			if (*p != 0)
232 				return (-1);
233 	}
234 
235 	return x * 8 + y;
236 }
237 
238 #ifdef COMPAT_FREEBSD32
239 struct in6_ndifreq32 {
240 	char ifname[IFNAMSIZ];
241 	uint32_t ifindex;
242 };
243 #define	SIOCGDEFIFACE32_IN6	_IOWR('i', 86, struct in6_ndifreq32)
244 #endif
245 
246 int
247 in6_control(struct socket *so, u_long cmd, caddr_t data,
248     struct ifnet *ifp, struct thread *td)
249 {
250 	struct	in6_ifreq *ifr = (struct in6_ifreq *)data;
251 	struct	in6_ifaddr *ia = NULL;
252 	struct	in6_aliasreq *ifra = (struct in6_aliasreq *)data;
253 	struct sockaddr_in6 *sa6;
254 	int carp_attached = 0;
255 	int error;
256 	u_long ocmd = cmd;
257 
258 	/*
259 	 * Compat to make pre-10.x ifconfig(8) operable.
260 	 */
261 	if (cmd == OSIOCAIFADDR_IN6)
262 		cmd = SIOCAIFADDR_IN6;
263 
264 	switch (cmd) {
265 	case SIOCGETSGCNT_IN6:
266 	case SIOCGETMIFCNT_IN6:
267 		/*
268 		 * XXX mrt_ioctl has a 3rd, unused, FIB argument in route.c.
269 		 * We cannot see how that would be needed, so do not adjust the
270 		 * KPI blindly; more likely should clean up the IPv4 variant.
271 		 */
272 		return (mrt6_ioctl ? mrt6_ioctl(cmd, data) : EOPNOTSUPP);
273 	}
274 
275 	switch (cmd) {
276 	case SIOCAADDRCTL_POLICY:
277 	case SIOCDADDRCTL_POLICY:
278 		if (td != NULL) {
279 			error = priv_check(td, PRIV_NETINET_ADDRCTRL6);
280 			if (error)
281 				return (error);
282 		}
283 		return (in6_src_ioctl(cmd, data));
284 	}
285 
286 	if (ifp == NULL)
287 		return (EOPNOTSUPP);
288 
289 	switch (cmd) {
290 	case SIOCSNDFLUSH_IN6:
291 	case SIOCSPFXFLUSH_IN6:
292 	case SIOCSRTRFLUSH_IN6:
293 	case SIOCSDEFIFACE_IN6:
294 	case SIOCSIFINFO_FLAGS:
295 	case SIOCSIFINFO_IN6:
296 		if (td != NULL) {
297 			error = priv_check(td, PRIV_NETINET_ND6);
298 			if (error)
299 				return (error);
300 		}
301 		/* FALLTHROUGH */
302 	case OSIOCGIFINFO_IN6:
303 	case SIOCGIFINFO_IN6:
304 	case SIOCGNBRINFO_IN6:
305 	case SIOCGDEFIFACE_IN6:
306 		return (nd6_ioctl(cmd, data, ifp));
307 
308 #ifdef COMPAT_FREEBSD32
309 	case SIOCGDEFIFACE32_IN6:
310 		{
311 			struct in6_ndifreq ndif;
312 			struct in6_ndifreq32 *ndif32;
313 
314 			error = nd6_ioctl(SIOCGDEFIFACE_IN6, (caddr_t)&ndif,
315 			    ifp);
316 			if (error)
317 				return (error);
318 			ndif32 = (struct in6_ndifreq32 *)data;
319 			ndif32->ifindex = ndif.ifindex;
320 			return (0);
321 		}
322 #endif
323 	}
324 
325 	switch (cmd) {
326 	case SIOCSIFPREFIX_IN6:
327 	case SIOCDIFPREFIX_IN6:
328 	case SIOCAIFPREFIX_IN6:
329 	case SIOCCIFPREFIX_IN6:
330 	case SIOCSGIFPREFIX_IN6:
331 	case SIOCGIFPREFIX_IN6:
332 		log(LOG_NOTICE,
333 		    "prefix ioctls are now invalidated. "
334 		    "please use ifconfig.\n");
335 		return (EOPNOTSUPP);
336 	}
337 
338 	switch (cmd) {
339 	case SIOCSSCOPE6:
340 		if (td != NULL) {
341 			error = priv_check(td, PRIV_NETINET_SCOPE6);
342 			if (error)
343 				return (error);
344 		}
345 		/* FALLTHROUGH */
346 	case SIOCGSCOPE6:
347 	case SIOCGSCOPE6DEF:
348 		return (scope6_ioctl(cmd, data, ifp));
349 	}
350 
351 	/*
352 	 * Find address for this interface, if it exists.
353 	 *
354 	 * In netinet code, we have checked ifra_addr in SIOCSIF*ADDR operation
355 	 * only, and used the first interface address as the target of other
356 	 * operations (without checking ifra_addr).  This was because netinet
357 	 * code/API assumed at most 1 interface address per interface.
358 	 * Since IPv6 allows a node to assign multiple addresses
359 	 * on a single interface, we almost always look and check the
360 	 * presence of ifra_addr, and reject invalid ones here.
361 	 * It also decreases duplicated code among SIOC*_IN6 operations.
362 	 */
363 	switch (cmd) {
364 	case SIOCAIFADDR_IN6:
365 	case SIOCSIFPHYADDR_IN6:
366 		sa6 = &ifra->ifra_addr;
367 		break;
368 	case SIOCSIFADDR_IN6:
369 	case SIOCGIFADDR_IN6:
370 	case SIOCSIFDSTADDR_IN6:
371 	case SIOCSIFNETMASK_IN6:
372 	case SIOCGIFDSTADDR_IN6:
373 	case SIOCGIFNETMASK_IN6:
374 	case SIOCDIFADDR_IN6:
375 	case SIOCGIFPSRCADDR_IN6:
376 	case SIOCGIFPDSTADDR_IN6:
377 	case SIOCGIFAFLAG_IN6:
378 	case SIOCSNDFLUSH_IN6:
379 	case SIOCSPFXFLUSH_IN6:
380 	case SIOCSRTRFLUSH_IN6:
381 	case SIOCGIFALIFETIME_IN6:
382 	case SIOCGIFSTAT_IN6:
383 	case SIOCGIFSTAT_ICMP6:
384 		sa6 = &ifr->ifr_addr;
385 		break;
386 	case SIOCSIFADDR:
387 	case SIOCSIFBRDADDR:
388 	case SIOCSIFDSTADDR:
389 	case SIOCSIFNETMASK:
390 		/*
391 		 * Although we should pass any non-INET6 ioctl requests
392 		 * down to driver, we filter some legacy INET requests.
393 		 * Drivers trust SIOCSIFADDR et al to come from an already
394 		 * privileged layer, and do not perform any credentials
395 		 * checks or input validation.
396 		 */
397 		return (EINVAL);
398 	default:
399 		sa6 = NULL;
400 		break;
401 	}
402 	if (sa6 && sa6->sin6_family == AF_INET6) {
403 		if (sa6->sin6_scope_id != 0)
404 			error = sa6_embedscope(sa6, 0);
405 		else
406 			error = in6_setscope(&sa6->sin6_addr, ifp, NULL);
407 		if (error != 0)
408 			return (error);
409 		if (td != NULL && (error = prison_check_ip6(td->td_ucred,
410 		    &sa6->sin6_addr)) != 0)
411 			return (error);
412 		ia = in6ifa_ifpwithaddr(ifp, &sa6->sin6_addr);
413 	} else
414 		ia = NULL;
415 
416 	switch (cmd) {
417 	case SIOCSIFADDR_IN6:
418 	case SIOCSIFDSTADDR_IN6:
419 	case SIOCSIFNETMASK_IN6:
420 		/*
421 		 * Since IPv6 allows a node to assign multiple addresses
422 		 * on a single interface, SIOCSIFxxx ioctls are deprecated.
423 		 */
424 		/* we decided to obsolete this command (20000704) */
425 		error = EINVAL;
426 		goto out;
427 
428 	case SIOCDIFADDR_IN6:
429 		/*
430 		 * for IPv4, we look for existing in_ifaddr here to allow
431 		 * "ifconfig if0 delete" to remove the first IPv4 address on
432 		 * the interface.  For IPv6, as the spec allows multiple
433 		 * interface address from the day one, we consider "remove the
434 		 * first one" semantics to be not preferable.
435 		 */
436 		if (ia == NULL) {
437 			error = EADDRNOTAVAIL;
438 			goto out;
439 		}
440 		/* FALLTHROUGH */
441 	case SIOCAIFADDR_IN6:
442 		/*
443 		 * We always require users to specify a valid IPv6 address for
444 		 * the corresponding operation.
445 		 */
446 		if (ifra->ifra_addr.sin6_family != AF_INET6 ||
447 		    ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6)) {
448 			error = EAFNOSUPPORT;
449 			goto out;
450 		}
451 
452 		if (td != NULL) {
453 			error = priv_check(td, (cmd == SIOCDIFADDR_IN6) ?
454 			    PRIV_NET_DELIFADDR : PRIV_NET_ADDIFADDR);
455 			if (error)
456 				goto out;
457 		}
458 		/* FALLTHROUGH */
459 	case SIOCGIFSTAT_IN6:
460 	case SIOCGIFSTAT_ICMP6:
461 		if (ifp->if_afdata[AF_INET6] == NULL) {
462 			error = EPFNOSUPPORT;
463 			goto out;
464 		}
465 		break;
466 
467 	case SIOCGIFADDR_IN6:
468 		/* This interface is basically deprecated. use SIOCGIFCONF. */
469 		/* FALLTHROUGH */
470 	case SIOCGIFAFLAG_IN6:
471 	case SIOCGIFNETMASK_IN6:
472 	case SIOCGIFDSTADDR_IN6:
473 	case SIOCGIFALIFETIME_IN6:
474 		/* must think again about its semantics */
475 		if (ia == NULL) {
476 			error = EADDRNOTAVAIL;
477 			goto out;
478 		}
479 		break;
480 	}
481 
482 	switch (cmd) {
483 	case SIOCGIFADDR_IN6:
484 		ifr->ifr_addr = ia->ia_addr;
485 		if ((error = sa6_recoverscope(&ifr->ifr_addr)) != 0)
486 			goto out;
487 		break;
488 
489 	case SIOCGIFDSTADDR_IN6:
490 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
491 			error = EINVAL;
492 			goto out;
493 		}
494 		ifr->ifr_dstaddr = ia->ia_dstaddr;
495 		if ((error = sa6_recoverscope(&ifr->ifr_dstaddr)) != 0)
496 			goto out;
497 		break;
498 
499 	case SIOCGIFNETMASK_IN6:
500 		ifr->ifr_addr = ia->ia_prefixmask;
501 		break;
502 
503 	case SIOCGIFAFLAG_IN6:
504 		ifr->ifr_ifru.ifru_flags6 = ia->ia6_flags;
505 		break;
506 
507 	case SIOCGIFSTAT_IN6:
508 		COUNTER_ARRAY_COPY(((struct in6_ifextra *)
509 		    ifp->if_afdata[AF_INET6])->in6_ifstat,
510 		    &ifr->ifr_ifru.ifru_stat,
511 		    sizeof(struct in6_ifstat) / sizeof(uint64_t));
512 		break;
513 
514 	case SIOCGIFSTAT_ICMP6:
515 		COUNTER_ARRAY_COPY(((struct in6_ifextra *)
516 		    ifp->if_afdata[AF_INET6])->icmp6_ifstat,
517 		    &ifr->ifr_ifru.ifru_icmp6stat,
518 		    sizeof(struct icmp6_ifstat) / sizeof(uint64_t));
519 		break;
520 
521 	case SIOCGIFALIFETIME_IN6:
522 		ifr->ifr_ifru.ifru_lifetime = ia->ia6_lifetime;
523 		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
524 			time_t maxexpire;
525 			struct in6_addrlifetime *retlt =
526 			    &ifr->ifr_ifru.ifru_lifetime;
527 
528 			/*
529 			 * XXX: adjust expiration time assuming time_t is
530 			 * signed.
531 			 */
532 			maxexpire = (-1) &
533 			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
534 			if (ia->ia6_lifetime.ia6t_vltime <
535 			    maxexpire - ia->ia6_updatetime) {
536 				retlt->ia6t_expire = ia->ia6_updatetime +
537 				    ia->ia6_lifetime.ia6t_vltime;
538 			} else
539 				retlt->ia6t_expire = maxexpire;
540 		}
541 		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
542 			time_t maxexpire;
543 			struct in6_addrlifetime *retlt =
544 			    &ifr->ifr_ifru.ifru_lifetime;
545 
546 			/*
547 			 * XXX: adjust expiration time assuming time_t is
548 			 * signed.
549 			 */
550 			maxexpire = (-1) &
551 			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
552 			if (ia->ia6_lifetime.ia6t_pltime <
553 			    maxexpire - ia->ia6_updatetime) {
554 				retlt->ia6t_preferred = ia->ia6_updatetime +
555 				    ia->ia6_lifetime.ia6t_pltime;
556 			} else
557 				retlt->ia6t_preferred = maxexpire;
558 		}
559 		break;
560 
561 	case SIOCAIFADDR_IN6:
562 	{
563 		struct nd_prefixctl pr0;
564 		struct nd_prefix *pr;
565 
566 		/*
567 		 * first, make or update the interface address structure,
568 		 * and link it to the list.
569 		 */
570 		if ((error = in6_update_ifa(ifp, ifra, ia, 0)) != 0)
571 			goto out;
572 		if (ia != NULL) {
573 			if (ia->ia_ifa.ifa_carp)
574 				(*carp_detach_p)(&ia->ia_ifa, true);
575 			ifa_free(&ia->ia_ifa);
576 		}
577 		if ((ia = in6ifa_ifpwithaddr(ifp, &ifra->ifra_addr.sin6_addr))
578 		    == NULL) {
579 			/*
580 			 * this can happen when the user specify the 0 valid
581 			 * lifetime.
582 			 */
583 			break;
584 		}
585 
586 		if (cmd == ocmd && ifra->ifra_vhid > 0) {
587 			if (carp_attach_p != NULL)
588 				error = (*carp_attach_p)(&ia->ia_ifa,
589 				    ifra->ifra_vhid);
590 			else
591 				error = EPROTONOSUPPORT;
592 			if (error)
593 				goto out;
594 			else
595 				carp_attached = 1;
596 		}
597 
598 		/*
599 		 * then, make the prefix on-link on the interface.
600 		 * XXX: we'd rather create the prefix before the address, but
601 		 * we need at least one address to install the corresponding
602 		 * interface route, so we configure the address first.
603 		 */
604 
605 		/*
606 		 * convert mask to prefix length (prefixmask has already
607 		 * been validated in in6_update_ifa().
608 		 */
609 		bzero(&pr0, sizeof(pr0));
610 		pr0.ndpr_ifp = ifp;
611 		pr0.ndpr_plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
612 		    NULL);
613 		if (pr0.ndpr_plen == 128) {
614 			/* we don't need to install a host route. */
615 			goto aifaddr_out;
616 		}
617 		pr0.ndpr_prefix = ifra->ifra_addr;
618 		/* apply the mask for safety. */
619 		IN6_MASK_ADDR(&pr0.ndpr_prefix.sin6_addr,
620 		    &ifra->ifra_prefixmask.sin6_addr);
621 
622 		/*
623 		 * XXX: since we don't have an API to set prefix (not address)
624 		 * lifetimes, we just use the same lifetimes as addresses.
625 		 * The (temporarily) installed lifetimes can be overridden by
626 		 * later advertised RAs (when accept_rtadv is non 0), which is
627 		 * an intended behavior.
628 		 */
629 		pr0.ndpr_raf_onlink = 1; /* should be configurable? */
630 		pr0.ndpr_raf_auto =
631 		    ((ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0);
632 		pr0.ndpr_vltime = ifra->ifra_lifetime.ia6t_vltime;
633 		pr0.ndpr_pltime = ifra->ifra_lifetime.ia6t_pltime;
634 
635 		/* add the prefix if not yet. */
636 		if ((pr = nd6_prefix_lookup(&pr0)) == NULL) {
637 			/*
638 			 * nd6_prelist_add will install the corresponding
639 			 * interface route.
640 			 */
641 			if ((error = nd6_prelist_add(&pr0, NULL, &pr)) != 0) {
642 				if (carp_attached)
643 					(*carp_detach_p)(&ia->ia_ifa, false);
644 				goto out;
645 			}
646 		}
647 
648 		/* relate the address to the prefix */
649 		if (ia->ia6_ndpr == NULL) {
650 			ia->ia6_ndpr = pr;
651 			pr->ndpr_addrcnt++;
652 
653 			/*
654 			 * If this is the first autoconf address from the
655 			 * prefix, create a temporary address as well
656 			 * (when required).
657 			 */
658 			if ((ia->ia6_flags & IN6_IFF_AUTOCONF) &&
659 			    V_ip6_use_tempaddr && pr->ndpr_addrcnt == 1) {
660 				int e;
661 				if ((e = in6_tmpifadd(ia, 1, 0)) != 0) {
662 					log(LOG_NOTICE, "in6_control: failed "
663 					    "to create a temporary address, "
664 					    "errno=%d\n", e);
665 				}
666 			}
667 		}
668 		nd6_prefix_rele(pr);
669 
670 		/*
671 		 * this might affect the status of autoconfigured addresses,
672 		 * that is, this address might make other addresses detached.
673 		 */
674 		pfxlist_onlink_check();
675 
676 aifaddr_out:
677 		/*
678 		 * Try to clear the flag when a new IPv6 address is added
679 		 * onto an IFDISABLED interface and it succeeds.
680 		 */
681 		if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) {
682 			struct in6_ndireq nd;
683 
684 			memset(&nd, 0, sizeof(nd));
685 			nd.ndi.flags = ND_IFINFO(ifp)->flags;
686 			nd.ndi.flags &= ~ND6_IFF_IFDISABLED;
687 			if (nd6_ioctl(SIOCSIFINFO_FLAGS, (caddr_t)&nd, ifp) < 0)
688 				log(LOG_NOTICE, "SIOCAIFADDR_IN6: "
689 				    "SIOCSIFINFO_FLAGS for -ifdisabled "
690 				    "failed.");
691 			/*
692 			 * Ignore failure of clearing the flag intentionally.
693 			 * The failure means address duplication was detected.
694 			 */
695 		}
696 		break;
697 	}
698 
699 	case SIOCDIFADDR_IN6:
700 	{
701 		struct nd_prefix *pr;
702 
703 		/*
704 		 * If the address being deleted is the only one that owns
705 		 * the corresponding prefix, expire the prefix as well.
706 		 * XXX: theoretically, we don't have to worry about such
707 		 * relationship, since we separate the address management
708 		 * and the prefix management.  We do this, however, to provide
709 		 * as much backward compatibility as possible in terms of
710 		 * the ioctl operation.
711 		 * Note that in6_purgeaddr() will decrement ndpr_addrcnt.
712 		 */
713 		pr = ia->ia6_ndpr;
714 		in6_purgeaddr(&ia->ia_ifa);
715 		if (pr != NULL && pr->ndpr_addrcnt == 0) {
716 			ND6_WLOCK();
717 			nd6_prefix_unlink(pr, NULL);
718 			ND6_WUNLOCK();
719 			nd6_prefix_del(pr);
720 		}
721 		EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, &ia->ia_ifa,
722 		    IFADDR_EVENT_DEL);
723 		break;
724 	}
725 
726 	default:
727 		if (ifp->if_ioctl == NULL) {
728 			error = EOPNOTSUPP;
729 			goto out;
730 		}
731 		error = (*ifp->if_ioctl)(ifp, cmd, data);
732 		goto out;
733 	}
734 
735 	error = 0;
736 out:
737 	if (ia != NULL)
738 		ifa_free(&ia->ia_ifa);
739 	return (error);
740 }
741 
742 
743 static struct in6_multi_mship *
744 in6_joingroup_legacy(struct ifnet *ifp, const struct in6_addr *mcaddr,
745     int *errorp, int delay)
746 {
747 	struct in6_multi_mship *imm;
748 	int error;
749 
750 	imm = malloc(sizeof(*imm), M_IP6MADDR, M_NOWAIT);
751 	if (imm == NULL) {
752 		*errorp = ENOBUFS;
753 		return (NULL);
754 	}
755 
756 	delay = (delay * PR_FASTHZ) / hz;
757 
758 	error = in6_joingroup(ifp, mcaddr, NULL, &imm->i6mm_maddr, delay);
759 	if (error) {
760 		*errorp = error;
761 		free(imm, M_IP6MADDR);
762 		return (NULL);
763 	}
764 
765 	return (imm);
766 }
767 /*
768  * Join necessary multicast groups.  Factored out from in6_update_ifa().
769  * This entire work should only be done once, for the default FIB.
770  */
771 static int
772 in6_update_ifa_join_mc(struct ifnet *ifp, struct in6_aliasreq *ifra,
773     struct in6_ifaddr *ia, int flags, struct in6_multi **in6m_sol)
774 {
775 	char ip6buf[INET6_ADDRSTRLEN];
776 	struct in6_addr mltaddr;
777 	struct in6_multi_mship *imm;
778 	int delay, error;
779 
780 	KASSERT(in6m_sol != NULL, ("%s: in6m_sol is NULL", __func__));
781 
782 	/* Join solicited multicast addr for new host id. */
783 	bzero(&mltaddr, sizeof(struct in6_addr));
784 	mltaddr.s6_addr32[0] = IPV6_ADDR_INT32_MLL;
785 	mltaddr.s6_addr32[2] = htonl(1);
786 	mltaddr.s6_addr32[3] = ifra->ifra_addr.sin6_addr.s6_addr32[3];
787 	mltaddr.s6_addr8[12] = 0xff;
788 	if ((error = in6_setscope(&mltaddr, ifp, NULL)) != 0) {
789 		/* XXX: should not happen */
790 		log(LOG_ERR, "%s: in6_setscope failed\n", __func__);
791 		goto cleanup;
792 	}
793 	delay = error = 0;
794 	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
795 		/*
796 		 * We need a random delay for DAD on the address being
797 		 * configured.  It also means delaying transmission of the
798 		 * corresponding MLD report to avoid report collision.
799 		 * [RFC 4861, Section 6.3.7]
800 		 */
801 		delay = arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz);
802 	}
803 	imm = in6_joingroup_legacy(ifp, &mltaddr, &error, delay);
804 	if (imm == NULL) {
805 		nd6log((LOG_WARNING, "%s: in6_joingroup failed for %s on %s "
806 		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf, &mltaddr),
807 		    if_name(ifp), error));
808 		goto cleanup;
809 	}
810 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
811 	*in6m_sol = imm->i6mm_maddr;
812 
813 	/*
814 	 * Join link-local all-nodes address.
815 	 */
816 	mltaddr = in6addr_linklocal_allnodes;
817 	if ((error = in6_setscope(&mltaddr, ifp, NULL)) != 0)
818 		goto cleanup; /* XXX: should not fail */
819 
820 	imm = in6_joingroup_legacy(ifp, &mltaddr, &error, 0);
821 	if (imm == NULL) {
822 		nd6log((LOG_WARNING, "%s: in6_joingroup failed for %s on %s "
823 		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf, &mltaddr),
824 		    if_name(ifp), error));
825 		goto cleanup;
826 	}
827 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
828 
829 	/*
830 	 * Join node information group address.
831 	 */
832 	delay = 0;
833 	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
834 		/*
835 		 * The spec does not say anything about delay for this group,
836 		 * but the same logic should apply.
837 		 */
838 		delay = arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz);
839 	}
840 	if (in6_nigroup(ifp, NULL, -1, &mltaddr) == 0) {
841 		/* XXX jinmei */
842 		imm = in6_joingroup_legacy(ifp, &mltaddr, &error, delay);
843 		if (imm == NULL)
844 			nd6log((LOG_WARNING,
845 			    "%s: in6_joingroup failed for %s on %s "
846 			    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
847 			    &mltaddr), if_name(ifp), error));
848 			/* XXX not very fatal, go on... */
849 		else
850 			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
851 	}
852 	if (V_icmp6_nodeinfo_oldmcprefix &&
853 	    in6_nigroup_oldmcprefix(ifp, NULL, -1, &mltaddr) == 0) {
854 		imm = in6_joingroup_legacy(ifp, &mltaddr, &error, delay);
855 		if (imm == NULL)
856 			nd6log((LOG_WARNING,
857 			    "%s: in6_joingroup failed for %s on %s "
858 			    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
859 			    &mltaddr), if_name(ifp), error));
860 			/* XXX not very fatal, go on... */
861 		else
862 			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
863 	}
864 
865 	/*
866 	 * Join interface-local all-nodes address.
867 	 * (ff01::1%ifN, and ff01::%ifN/32)
868 	 */
869 	mltaddr = in6addr_nodelocal_allnodes;
870 	if ((error = in6_setscope(&mltaddr, ifp, NULL)) != 0)
871 		goto cleanup; /* XXX: should not fail */
872 
873 	imm = in6_joingroup_legacy(ifp, &mltaddr, &error, 0);
874 	if (imm == NULL) {
875 		nd6log((LOG_WARNING, "%s: in6_joingroup failed for %s on %s "
876 		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
877 		    &mltaddr), if_name(ifp), error));
878 		goto cleanup;
879 	}
880 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
881 
882 cleanup:
883 	return (error);
884 }
885 
886 /*
887  * Update parameters of an IPv6 interface address.
888  * If necessary, a new entry is created and linked into address chains.
889  * This function is separated from in6_control().
890  */
891 int
892 in6_update_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
893     struct in6_ifaddr *ia, int flags)
894 {
895 	int error, hostIsNew = 0;
896 
897 	if ((error = in6_validate_ifra(ifp, ifra, ia, flags)) != 0)
898 		return (error);
899 
900 	if (ia == NULL) {
901 		hostIsNew = 1;
902 		if ((ia = in6_alloc_ifa(ifp, ifra, flags)) == NULL)
903 			return (ENOBUFS);
904 	}
905 
906 	error = in6_update_ifa_internal(ifp, ifra, ia, hostIsNew, flags);
907 	if (error != 0) {
908 		if (hostIsNew != 0) {
909 			in6_unlink_ifa(ia, ifp);
910 			ifa_free(&ia->ia_ifa);
911 		}
912 		return (error);
913 	}
914 
915 	if (hostIsNew)
916 		error = in6_broadcast_ifa(ifp, ifra, ia, flags);
917 
918 	return (error);
919 }
920 
921 /*
922  * Fill in basic IPv6 address request info.
923  */
924 void
925 in6_prepare_ifra(struct in6_aliasreq *ifra, const struct in6_addr *addr,
926     const struct in6_addr *mask)
927 {
928 
929 	memset(ifra, 0, sizeof(struct in6_aliasreq));
930 
931 	ifra->ifra_addr.sin6_family = AF_INET6;
932 	ifra->ifra_addr.sin6_len = sizeof(struct sockaddr_in6);
933 	if (addr != NULL)
934 		ifra->ifra_addr.sin6_addr = *addr;
935 
936 	ifra->ifra_prefixmask.sin6_family = AF_INET6;
937 	ifra->ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
938 	if (mask != NULL)
939 		ifra->ifra_prefixmask.sin6_addr = *mask;
940 }
941 
942 static int
943 in6_validate_ifra(struct ifnet *ifp, struct in6_aliasreq *ifra,
944     struct in6_ifaddr *ia, int flags)
945 {
946 	int plen = -1;
947 	struct sockaddr_in6 dst6;
948 	struct in6_addrlifetime *lt;
949 	char ip6buf[INET6_ADDRSTRLEN];
950 
951 	/* Validate parameters */
952 	if (ifp == NULL || ifra == NULL) /* this maybe redundant */
953 		return (EINVAL);
954 
955 	/*
956 	 * The destination address for a p2p link must have a family
957 	 * of AF_UNSPEC or AF_INET6.
958 	 */
959 	if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
960 	    ifra->ifra_dstaddr.sin6_family != AF_INET6 &&
961 	    ifra->ifra_dstaddr.sin6_family != AF_UNSPEC)
962 		return (EAFNOSUPPORT);
963 
964 	/*
965 	 * Validate address
966 	 */
967 	if (ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6) ||
968 	    ifra->ifra_addr.sin6_family != AF_INET6)
969 		return (EINVAL);
970 
971 	/*
972 	 * validate ifra_prefixmask.  don't check sin6_family, netmask
973 	 * does not carry fields other than sin6_len.
974 	 */
975 	if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6))
976 		return (EINVAL);
977 	/*
978 	 * Because the IPv6 address architecture is classless, we require
979 	 * users to specify a (non 0) prefix length (mask) for a new address.
980 	 * We also require the prefix (when specified) mask is valid, and thus
981 	 * reject a non-consecutive mask.
982 	 */
983 	if (ia == NULL && ifra->ifra_prefixmask.sin6_len == 0)
984 		return (EINVAL);
985 	if (ifra->ifra_prefixmask.sin6_len != 0) {
986 		plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
987 		    (u_char *)&ifra->ifra_prefixmask +
988 		    ifra->ifra_prefixmask.sin6_len);
989 		if (plen <= 0)
990 			return (EINVAL);
991 	} else {
992 		/*
993 		 * In this case, ia must not be NULL.  We just use its prefix
994 		 * length.
995 		 */
996 		plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
997 	}
998 	/*
999 	 * If the destination address on a p2p interface is specified,
1000 	 * and the address is a scoped one, validate/set the scope
1001 	 * zone identifier.
1002 	 */
1003 	dst6 = ifra->ifra_dstaddr;
1004 	if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) != 0 &&
1005 	    (dst6.sin6_family == AF_INET6)) {
1006 		struct in6_addr in6_tmp;
1007 		u_int32_t zoneid;
1008 
1009 		in6_tmp = dst6.sin6_addr;
1010 		if (in6_setscope(&in6_tmp, ifp, &zoneid))
1011 			return (EINVAL); /* XXX: should be impossible */
1012 
1013 		if (dst6.sin6_scope_id != 0) {
1014 			if (dst6.sin6_scope_id != zoneid)
1015 				return (EINVAL);
1016 		} else		/* user omit to specify the ID. */
1017 			dst6.sin6_scope_id = zoneid;
1018 
1019 		/* convert into the internal form */
1020 		if (sa6_embedscope(&dst6, 0))
1021 			return (EINVAL); /* XXX: should be impossible */
1022 	}
1023 	/* Modify original ifra_dstaddr to reflect changes */
1024 	ifra->ifra_dstaddr = dst6;
1025 
1026 	/*
1027 	 * The destination address can be specified only for a p2p or a
1028 	 * loopback interface.  If specified, the corresponding prefix length
1029 	 * must be 128.
1030 	 */
1031 	if (ifra->ifra_dstaddr.sin6_family == AF_INET6) {
1032 		if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) == 0) {
1033 			/* XXX: noisy message */
1034 			nd6log((LOG_INFO, "in6_update_ifa: a destination can "
1035 			    "be specified for a p2p or a loopback IF only\n"));
1036 			return (EINVAL);
1037 		}
1038 		if (plen != 128) {
1039 			nd6log((LOG_INFO, "in6_update_ifa: prefixlen should "
1040 			    "be 128 when dstaddr is specified\n"));
1041 			return (EINVAL);
1042 		}
1043 	}
1044 	/* lifetime consistency check */
1045 	lt = &ifra->ifra_lifetime;
1046 	if (lt->ia6t_pltime > lt->ia6t_vltime)
1047 		return (EINVAL);
1048 	if (lt->ia6t_vltime == 0) {
1049 		/*
1050 		 * the following log might be noisy, but this is a typical
1051 		 * configuration mistake or a tool's bug.
1052 		 */
1053 		nd6log((LOG_INFO,
1054 		    "in6_update_ifa: valid lifetime is 0 for %s\n",
1055 		    ip6_sprintf(ip6buf, &ifra->ifra_addr.sin6_addr)));
1056 
1057 		if (ia == NULL)
1058 			return (0); /* there's nothing to do */
1059 	}
1060 
1061 	/* Check prefix mask */
1062 	if (ia != NULL && ifra->ifra_prefixmask.sin6_len != 0) {
1063 		/*
1064 		 * We prohibit changing the prefix length of an existing
1065 		 * address, because
1066 		 * + such an operation should be rare in IPv6, and
1067 		 * + the operation would confuse prefix management.
1068 		 */
1069 		if (ia->ia_prefixmask.sin6_len != 0 &&
1070 		    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) != plen) {
1071 			nd6log((LOG_INFO, "in6_validate_ifa: the prefix length "
1072 			    "of an existing %s address should not be changed\n",
1073 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1074 
1075 			return (EINVAL);
1076 		}
1077 	}
1078 
1079 	return (0);
1080 }
1081 
1082 
1083 /*
1084  * Allocate a new ifaddr and link it into chains.
1085  */
1086 static struct in6_ifaddr *
1087 in6_alloc_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra, int flags)
1088 {
1089 	struct in6_ifaddr *ia;
1090 
1091 	/*
1092 	 * When in6_alloc_ifa() is called in a process of a received
1093 	 * RA, it is called under an interrupt context.  So, we should
1094 	 * call malloc with M_NOWAIT.
1095 	 */
1096 	ia = (struct in6_ifaddr *)ifa_alloc(sizeof(*ia), M_NOWAIT);
1097 	if (ia == NULL)
1098 		return (NULL);
1099 	LIST_INIT(&ia->ia6_memberships);
1100 	/* Initialize the address and masks, and put time stamp */
1101 	ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
1102 	ia->ia_addr.sin6_family = AF_INET6;
1103 	ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
1104 	/* XXX: Can we assign ,sin6_addr and skip the rest? */
1105 	ia->ia_addr = ifra->ifra_addr;
1106 	ia->ia6_createtime = time_uptime;
1107 	if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
1108 		/*
1109 		 * Some functions expect that ifa_dstaddr is not
1110 		 * NULL for p2p interfaces.
1111 		 */
1112 		ia->ia_ifa.ifa_dstaddr =
1113 		    (struct sockaddr *)&ia->ia_dstaddr;
1114 	} else {
1115 		ia->ia_ifa.ifa_dstaddr = NULL;
1116 	}
1117 
1118 	/* set prefix mask if any */
1119 	ia->ia_ifa.ifa_netmask = (struct sockaddr *)&ia->ia_prefixmask;
1120 	if (ifra->ifra_prefixmask.sin6_len != 0) {
1121 		ia->ia_prefixmask.sin6_family = AF_INET6;
1122 		ia->ia_prefixmask.sin6_len = ifra->ifra_prefixmask.sin6_len;
1123 		ia->ia_prefixmask.sin6_addr = ifra->ifra_prefixmask.sin6_addr;
1124 	}
1125 
1126 	ia->ia_ifp = ifp;
1127 	ifa_ref(&ia->ia_ifa);			/* if_addrhead */
1128 	IF_ADDR_WLOCK(ifp);
1129 	CK_STAILQ_INSERT_TAIL(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
1130 	IF_ADDR_WUNLOCK(ifp);
1131 
1132 	ifa_ref(&ia->ia_ifa);			/* in6_ifaddrhead */
1133 	IN6_IFADDR_WLOCK();
1134 	CK_STAILQ_INSERT_TAIL(&V_in6_ifaddrhead, ia, ia_link);
1135 	CK_LIST_INSERT_HEAD(IN6ADDR_HASH(&ia->ia_addr.sin6_addr), ia, ia6_hash);
1136 	IN6_IFADDR_WUNLOCK();
1137 
1138 	return (ia);
1139 }
1140 
1141 /*
1142  * Update/configure interface address parameters:
1143  *
1144  * 1) Update lifetime
1145  * 2) Update interface metric ad flags
1146  * 3) Notify other subsystems
1147  */
1148 static int
1149 in6_update_ifa_internal(struct ifnet *ifp, struct in6_aliasreq *ifra,
1150     struct in6_ifaddr *ia, int hostIsNew, int flags)
1151 {
1152 	int error;
1153 
1154 	/* update timestamp */
1155 	ia->ia6_updatetime = time_uptime;
1156 
1157 	/*
1158 	 * Set lifetimes.  We do not refer to ia6t_expire and ia6t_preferred
1159 	 * to see if the address is deprecated or invalidated, but initialize
1160 	 * these members for applications.
1161 	 */
1162 	ia->ia6_lifetime = ifra->ifra_lifetime;
1163 	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1164 		ia->ia6_lifetime.ia6t_expire =
1165 		    time_uptime + ia->ia6_lifetime.ia6t_vltime;
1166 	} else
1167 		ia->ia6_lifetime.ia6t_expire = 0;
1168 	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1169 		ia->ia6_lifetime.ia6t_preferred =
1170 		    time_uptime + ia->ia6_lifetime.ia6t_pltime;
1171 	} else
1172 		ia->ia6_lifetime.ia6t_preferred = 0;
1173 
1174 	/*
1175 	 * backward compatibility - if IN6_IFF_DEPRECATED is set from the
1176 	 * userland, make it deprecated.
1177 	 */
1178 	if ((ifra->ifra_flags & IN6_IFF_DEPRECATED) != 0) {
1179 		ia->ia6_lifetime.ia6t_pltime = 0;
1180 		ia->ia6_lifetime.ia6t_preferred = time_uptime;
1181 	}
1182 
1183 	/*
1184 	 * configure address flags.
1185 	 */
1186 	ia->ia6_flags = ifra->ifra_flags;
1187 
1188 	/*
1189 	 * Make the address tentative before joining multicast addresses,
1190 	 * so that corresponding MLD responses would not have a tentative
1191 	 * source address.
1192 	 */
1193 	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/* safety */
1194 
1195 	/*
1196 	 * DAD should be performed for an new address or addresses on
1197 	 * an interface with ND6_IFF_IFDISABLED.
1198 	 */
1199 	if (in6if_do_dad(ifp) &&
1200 	    (hostIsNew || (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)))
1201 		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1202 
1203 	/* notify other subsystems */
1204 	error = in6_notify_ifa(ifp, ia, ifra, hostIsNew);
1205 
1206 	return (error);
1207 }
1208 
1209 /*
1210  * Do link-level ifa job:
1211  * 1) Add lle entry for added address
1212  * 2) Notifies routing socket users about new address
1213  * 3) join appropriate multicast group
1214  * 4) start DAD if enabled
1215  */
1216 static int
1217 in6_broadcast_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
1218     struct in6_ifaddr *ia, int flags)
1219 {
1220 	struct in6_multi *in6m_sol;
1221 	int error = 0;
1222 
1223 	/* Add local address to lltable, if necessary (ex. on p2p link). */
1224 	if ((error = nd6_add_ifa_lle(ia)) != 0) {
1225 		in6_purgeaddr(&ia->ia_ifa);
1226 		ifa_free(&ia->ia_ifa);
1227 		return (error);
1228 	}
1229 
1230 	/* Join necessary multicast groups. */
1231 	in6m_sol = NULL;
1232 	if ((ifp->if_flags & IFF_MULTICAST) != 0) {
1233 		error = in6_update_ifa_join_mc(ifp, ifra, ia, flags, &in6m_sol);
1234 		if (error != 0) {
1235 			in6_purgeaddr(&ia->ia_ifa);
1236 			ifa_free(&ia->ia_ifa);
1237 			return (error);
1238 		}
1239 	}
1240 
1241 	/* Perform DAD, if the address is TENTATIVE. */
1242 	if ((ia->ia6_flags & IN6_IFF_TENTATIVE)) {
1243 		int delay, mindelay, maxdelay;
1244 
1245 		delay = 0;
1246 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1247 			/*
1248 			 * We need to impose a delay before sending an NS
1249 			 * for DAD.  Check if we also needed a delay for the
1250 			 * corresponding MLD message.  If we did, the delay
1251 			 * should be larger than the MLD delay (this could be
1252 			 * relaxed a bit, but this simple logic is at least
1253 			 * safe).
1254 			 * XXX: Break data hiding guidelines and look at
1255 			 * state for the solicited multicast group.
1256 			 */
1257 			mindelay = 0;
1258 			if (in6m_sol != NULL &&
1259 			    in6m_sol->in6m_state == MLD_REPORTING_MEMBER) {
1260 				mindelay = in6m_sol->in6m_timer;
1261 			}
1262 			maxdelay = MAX_RTR_SOLICITATION_DELAY * hz;
1263 			if (maxdelay - mindelay == 0)
1264 				delay = 0;
1265 			else {
1266 				delay =
1267 				    (arc4random() % (maxdelay - mindelay)) +
1268 				    mindelay;
1269 			}
1270 		}
1271 		nd6_dad_start((struct ifaddr *)ia, delay);
1272 	}
1273 
1274 	in6_newaddrmsg(ia, RTM_ADD);
1275 	ifa_free(&ia->ia_ifa);
1276 	return (error);
1277 }
1278 
1279 void
1280 in6_purgeaddr(struct ifaddr *ifa)
1281 {
1282 	struct ifnet *ifp = ifa->ifa_ifp;
1283 	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1284 	struct in6_multi_mship *imm;
1285 	int plen, error;
1286 
1287 	if (ifa->ifa_carp)
1288 		(*carp_detach_p)(ifa, false);
1289 
1290 	/*
1291 	 * Remove the loopback route to the interface address.
1292 	 * The check for the current setting of "nd6_useloopback"
1293 	 * is not needed.
1294 	 */
1295 	if (ia->ia_flags & IFA_RTSELF) {
1296 		error = ifa_del_loopback_route((struct ifaddr *)ia,
1297 		    (struct sockaddr *)&ia->ia_addr);
1298 		if (error == 0)
1299 			ia->ia_flags &= ~IFA_RTSELF;
1300 	}
1301 
1302 	/* stop DAD processing */
1303 	nd6_dad_stop(ifa);
1304 
1305 	/* Leave multicast groups. */
1306 	while ((imm = LIST_FIRST(&ia->ia6_memberships)) != NULL) {
1307 		LIST_REMOVE(imm, i6mm_chain);
1308 		if (imm->i6mm_maddr != NULL)
1309 			in6_leavegroup(imm->i6mm_maddr, NULL);
1310 		free(imm, M_IP6MADDR);
1311 	}
1312 	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1313 	if ((ia->ia_flags & IFA_ROUTE) && plen == 128) {
1314 		error = rtinit(&(ia->ia_ifa), RTM_DELETE, ia->ia_flags |
1315 		    (ia->ia_dstaddr.sin6_family == AF_INET6 ? RTF_HOST : 0));
1316 		if (error != 0)
1317 			log(LOG_INFO, "%s: err=%d, destination address delete "
1318 			    "failed\n", __func__, error);
1319 		ia->ia_flags &= ~IFA_ROUTE;
1320 	}
1321 
1322 	in6_newaddrmsg(ia, RTM_DELETE);
1323 	in6_unlink_ifa(ia, ifp);
1324 }
1325 
1326 static void
1327 in6_unlink_ifa(struct in6_ifaddr *ia, struct ifnet *ifp)
1328 {
1329 	char ip6buf[INET6_ADDRSTRLEN];
1330 	int remove_lle;
1331 
1332 	IF_ADDR_WLOCK(ifp);
1333 	CK_STAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifaddr, ifa_link);
1334 	IF_ADDR_WUNLOCK(ifp);
1335 	ifa_free(&ia->ia_ifa);			/* if_addrhead */
1336 
1337 	/*
1338 	 * Defer the release of what might be the last reference to the
1339 	 * in6_ifaddr so that it can't be freed before the remainder of the
1340 	 * cleanup.
1341 	 */
1342 	IN6_IFADDR_WLOCK();
1343 	CK_STAILQ_REMOVE(&V_in6_ifaddrhead, ia, in6_ifaddr, ia_link);
1344 	CK_LIST_REMOVE(ia, ia6_hash);
1345 	IN6_IFADDR_WUNLOCK();
1346 
1347 	/*
1348 	 * Release the reference to the base prefix.  There should be a
1349 	 * positive reference.
1350 	 */
1351 	remove_lle = 0;
1352 	if (ia->ia6_ndpr == NULL) {
1353 		nd6log((LOG_NOTICE,
1354 		    "in6_unlink_ifa: autoconf'ed address "
1355 		    "%s has no prefix\n", ip6_sprintf(ip6buf, IA6_IN6(ia))));
1356 	} else {
1357 		ia->ia6_ndpr->ndpr_addrcnt--;
1358 		/* Do not delete lles within prefix if refcont != 0 */
1359 		if (ia->ia6_ndpr->ndpr_addrcnt == 0)
1360 			remove_lle = 1;
1361 		ia->ia6_ndpr = NULL;
1362 	}
1363 
1364 	nd6_rem_ifa_lle(ia, remove_lle);
1365 
1366 	/*
1367 	 * Also, if the address being removed is autoconf'ed, call
1368 	 * pfxlist_onlink_check() since the release might affect the status of
1369 	 * other (detached) addresses.
1370 	 */
1371 	if ((ia->ia6_flags & IN6_IFF_AUTOCONF)) {
1372 		pfxlist_onlink_check();
1373 	}
1374 	ifa_free(&ia->ia_ifa);			/* in6_ifaddrhead */
1375 }
1376 
1377 /*
1378  * Notifies other subsystems about address change/arrival:
1379  * 1) Notifies device handler on the first IPv6 address assignment
1380  * 2) Handle routing table changes for P2P links and route
1381  * 3) Handle routing table changes for address host route
1382  */
1383 static int
1384 in6_notify_ifa(struct ifnet *ifp, struct in6_ifaddr *ia,
1385     struct in6_aliasreq *ifra, int hostIsNew)
1386 {
1387 	int	error = 0, plen, ifacount = 0;
1388 	struct ifaddr *ifa;
1389 	struct sockaddr_in6 *pdst;
1390 	char ip6buf[INET6_ADDRSTRLEN];
1391 
1392 	/*
1393 	 * Give the interface a chance to initialize
1394 	 * if this is its first address,
1395 	 */
1396 	if (hostIsNew != 0) {
1397 		struct epoch_tracker et;
1398 
1399 		NET_EPOCH_ENTER(et);
1400 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1401 			if (ifa->ifa_addr->sa_family != AF_INET6)
1402 				continue;
1403 			ifacount++;
1404 		}
1405 		NET_EPOCH_EXIT(et);
1406 	}
1407 
1408 	if (ifacount <= 1 && ifp->if_ioctl) {
1409 		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
1410 		if (error)
1411 			goto done;
1412 	}
1413 
1414 	/*
1415 	 * If a new destination address is specified, scrub the old one and
1416 	 * install the new destination.  Note that the interface must be
1417 	 * p2p or loopback.
1418 	 */
1419 	pdst = &ifra->ifra_dstaddr;
1420 	if (pdst->sin6_family == AF_INET6 &&
1421 	    !IN6_ARE_ADDR_EQUAL(&pdst->sin6_addr, &ia->ia_dstaddr.sin6_addr)) {
1422 		if ((ia->ia_flags & IFA_ROUTE) != 0 &&
1423 		    (rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST) != 0)) {
1424 			nd6log((LOG_ERR, "in6_update_ifa_internal: failed to "
1425 			    "remove a route to the old destination: %s\n",
1426 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1427 			/* proceed anyway... */
1428 		} else
1429 			ia->ia_flags &= ~IFA_ROUTE;
1430 		ia->ia_dstaddr = *pdst;
1431 	}
1432 
1433 	/*
1434 	 * If a new destination address is specified for a point-to-point
1435 	 * interface, install a route to the destination as an interface
1436 	 * direct route.
1437 	 * XXX: the logic below rejects assigning multiple addresses on a p2p
1438 	 * interface that share the same destination.
1439 	 */
1440 	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1441 	if (!(ia->ia_flags & IFA_ROUTE) && plen == 128 &&
1442 	    ia->ia_dstaddr.sin6_family == AF_INET6) {
1443 		int rtflags = RTF_UP | RTF_HOST;
1444 		/*
1445 		 * Handle the case for ::1 .
1446 		 */
1447 		if (ifp->if_flags & IFF_LOOPBACK)
1448 			ia->ia_flags |= IFA_RTSELF;
1449 		error = rtinit(&ia->ia_ifa, RTM_ADD, ia->ia_flags | rtflags);
1450 		if (error)
1451 			goto done;
1452 		ia->ia_flags |= IFA_ROUTE;
1453 	}
1454 
1455 	/*
1456 	 * add a loopback route to self if not exists
1457 	 */
1458 	if (!(ia->ia_flags & IFA_RTSELF) && V_nd6_useloopback) {
1459 		error = ifa_add_loopback_route((struct ifaddr *)ia,
1460 		    (struct sockaddr *)&ia->ia_addr);
1461 		if (error == 0)
1462 			ia->ia_flags |= IFA_RTSELF;
1463 	}
1464 done:
1465 	WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
1466 	    "Invoking IPv6 network device address event may sleep");
1467 
1468 	ifa_ref(&ia->ia_ifa);
1469 	EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, &ia->ia_ifa,
1470 	    IFADDR_EVENT_ADD);
1471 	ifa_free(&ia->ia_ifa);
1472 
1473 	return (error);
1474 }
1475 
1476 /*
1477  * Find an IPv6 interface link-local address specific to an interface.
1478  * ifaddr is returned referenced.
1479  */
1480 struct in6_ifaddr *
1481 in6ifa_ifpforlinklocal(struct ifnet *ifp, int ignoreflags)
1482 {
1483 	struct ifaddr *ifa;
1484 
1485 	NET_EPOCH_ASSERT();
1486 
1487 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1488 		if (ifa->ifa_addr->sa_family != AF_INET6)
1489 			continue;
1490 		if (IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa))) {
1491 			if ((((struct in6_ifaddr *)ifa)->ia6_flags &
1492 			    ignoreflags) != 0)
1493 				continue;
1494 			ifa_ref(ifa);
1495 			break;
1496 		}
1497 	}
1498 
1499 	return ((struct in6_ifaddr *)ifa);
1500 }
1501 
1502 
1503 /*
1504  * find the interface address corresponding to a given IPv6 address.
1505  * ifaddr is returned referenced.
1506  */
1507 struct in6_ifaddr *
1508 in6ifa_ifwithaddr(const struct in6_addr *addr, uint32_t zoneid)
1509 {
1510 	struct rm_priotracker in6_ifa_tracker;
1511 	struct in6_ifaddr *ia;
1512 
1513 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1514 	CK_LIST_FOREACH(ia, IN6ADDR_HASH(addr), ia6_hash) {
1515 		if (IN6_ARE_ADDR_EQUAL(IA6_IN6(ia), addr)) {
1516 			if (zoneid != 0 &&
1517 			    zoneid != ia->ia_addr.sin6_scope_id)
1518 				continue;
1519 			ifa_ref(&ia->ia_ifa);
1520 			break;
1521 		}
1522 	}
1523 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1524 	return (ia);
1525 }
1526 
1527 /*
1528  * find the internet address corresponding to a given interface and address.
1529  * ifaddr is returned referenced.
1530  */
1531 struct in6_ifaddr *
1532 in6ifa_ifpwithaddr(struct ifnet *ifp, const struct in6_addr *addr)
1533 {
1534 	struct epoch_tracker et;
1535 	struct ifaddr *ifa;
1536 
1537 	NET_EPOCH_ENTER(et);
1538 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1539 		if (ifa->ifa_addr->sa_family != AF_INET6)
1540 			continue;
1541 		if (IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa))) {
1542 			ifa_ref(ifa);
1543 			break;
1544 		}
1545 	}
1546 	NET_EPOCH_EXIT(et);
1547 
1548 	return ((struct in6_ifaddr *)ifa);
1549 }
1550 
1551 /*
1552  * Find a link-local scoped address on ifp and return it if any.
1553  */
1554 struct in6_ifaddr *
1555 in6ifa_llaonifp(struct ifnet *ifp)
1556 {
1557 	struct epoch_tracker et;
1558 	struct sockaddr_in6 *sin6;
1559 	struct ifaddr *ifa;
1560 
1561 
1562 	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)
1563 		return (NULL);
1564 	NET_EPOCH_ENTER(et);
1565 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1566 		if (ifa->ifa_addr->sa_family != AF_INET6)
1567 			continue;
1568 		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
1569 		if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr) ||
1570 		    IN6_IS_ADDR_MC_INTFACELOCAL(&sin6->sin6_addr) ||
1571 		    IN6_IS_ADDR_MC_NODELOCAL(&sin6->sin6_addr))
1572 			break;
1573 	}
1574 	NET_EPOCH_EXIT(et);
1575 
1576 	return ((struct in6_ifaddr *)ifa);
1577 }
1578 
1579 /*
1580  * Convert IP6 address to printable (loggable) representation. Caller
1581  * has to make sure that ip6buf is at least INET6_ADDRSTRLEN long.
1582  */
1583 static char digits[] = "0123456789abcdef";
1584 char *
1585 ip6_sprintf(char *ip6buf, const struct in6_addr *addr)
1586 {
1587 	int i, cnt = 0, maxcnt = 0, idx = 0, index = 0;
1588 	char *cp;
1589 	const u_int16_t *a = (const u_int16_t *)addr;
1590 	const u_int8_t *d;
1591 	int dcolon = 0, zero = 0;
1592 
1593 	cp = ip6buf;
1594 
1595 	for (i = 0; i < 8; i++) {
1596 		if (*(a + i) == 0) {
1597 			cnt++;
1598 			if (cnt == 1)
1599 				idx = i;
1600 		}
1601 		else if (maxcnt < cnt) {
1602 			maxcnt = cnt;
1603 			index = idx;
1604 			cnt = 0;
1605 		}
1606 	}
1607 	if (maxcnt < cnt) {
1608 		maxcnt = cnt;
1609 		index = idx;
1610 	}
1611 
1612 	for (i = 0; i < 8; i++) {
1613 		if (dcolon == 1) {
1614 			if (*a == 0) {
1615 				if (i == 7)
1616 					*cp++ = ':';
1617 				a++;
1618 				continue;
1619 			} else
1620 				dcolon = 2;
1621 		}
1622 		if (*a == 0) {
1623 			if (dcolon == 0 && *(a + 1) == 0 && i == index) {
1624 				if (i == 0)
1625 					*cp++ = ':';
1626 				*cp++ = ':';
1627 				dcolon = 1;
1628 			} else {
1629 				*cp++ = '0';
1630 				*cp++ = ':';
1631 			}
1632 			a++;
1633 			continue;
1634 		}
1635 		d = (const u_char *)a;
1636 		/* Try to eliminate leading zeros in printout like in :0001. */
1637 		zero = 1;
1638 		*cp = digits[*d >> 4];
1639 		if (*cp != '0') {
1640 			zero = 0;
1641 			cp++;
1642 		}
1643 		*cp = digits[*d++ & 0xf];
1644 		if (zero == 0 || (*cp != '0')) {
1645 			zero = 0;
1646 			cp++;
1647 		}
1648 		*cp = digits[*d >> 4];
1649 		if (zero == 0 || (*cp != '0')) {
1650 			zero = 0;
1651 			cp++;
1652 		}
1653 		*cp++ = digits[*d & 0xf];
1654 		*cp++ = ':';
1655 		a++;
1656 	}
1657 	*--cp = '\0';
1658 	return (ip6buf);
1659 }
1660 
1661 int
1662 in6_localaddr(struct in6_addr *in6)
1663 {
1664 	struct rm_priotracker in6_ifa_tracker;
1665 	struct in6_ifaddr *ia;
1666 
1667 	if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
1668 		return 1;
1669 
1670 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1671 	CK_STAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
1672 		if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
1673 		    &ia->ia_prefixmask.sin6_addr)) {
1674 			IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1675 			return 1;
1676 		}
1677 	}
1678 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1679 
1680 	return (0);
1681 }
1682 
1683 /*
1684  * Return 1 if an internet address is for the local host and configured
1685  * on one of its interfaces.
1686  */
1687 int
1688 in6_localip(struct in6_addr *in6)
1689 {
1690 	struct rm_priotracker in6_ifa_tracker;
1691 	struct in6_ifaddr *ia;
1692 
1693 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1694 	CK_LIST_FOREACH(ia, IN6ADDR_HASH(in6), ia6_hash) {
1695 		if (IN6_ARE_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr)) {
1696 			IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1697 			return (1);
1698 		}
1699 	}
1700 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1701 	return (0);
1702 }
1703 
1704 /*
1705  * Return 1 if an internet address is configured on an interface.
1706  */
1707 int
1708 in6_ifhasaddr(struct ifnet *ifp, struct in6_addr *addr)
1709 {
1710 	struct in6_addr in6;
1711 	struct ifaddr *ifa;
1712 	struct in6_ifaddr *ia6;
1713 
1714 	NET_EPOCH_ASSERT();
1715 
1716 	in6 = *addr;
1717 	if (in6_clearscope(&in6))
1718 		return (0);
1719 	in6_setscope(&in6, ifp, NULL);
1720 
1721 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1722 		if (ifa->ifa_addr->sa_family != AF_INET6)
1723 			continue;
1724 		ia6 = (struct in6_ifaddr *)ifa;
1725 		if (IN6_ARE_ADDR_EQUAL(&ia6->ia_addr.sin6_addr, &in6))
1726 			return (1);
1727 	}
1728 
1729 	return (0);
1730 }
1731 
1732 int
1733 in6_is_addr_deprecated(struct sockaddr_in6 *sa6)
1734 {
1735 	struct rm_priotracker in6_ifa_tracker;
1736 	struct in6_ifaddr *ia;
1737 
1738 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1739 	CK_LIST_FOREACH(ia, IN6ADDR_HASH(&sa6->sin6_addr), ia6_hash) {
1740 		if (IN6_ARE_ADDR_EQUAL(IA6_IN6(ia), &sa6->sin6_addr)) {
1741 			if (ia->ia6_flags & IN6_IFF_DEPRECATED) {
1742 				IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1743 				return (1); /* true */
1744 			}
1745 			break;
1746 		}
1747 	}
1748 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1749 
1750 	return (0);		/* false */
1751 }
1752 
1753 /*
1754  * return length of part which dst and src are equal
1755  * hard coding...
1756  */
1757 int
1758 in6_matchlen(struct in6_addr *src, struct in6_addr *dst)
1759 {
1760 	int match = 0;
1761 	u_char *s = (u_char *)src, *d = (u_char *)dst;
1762 	u_char *lim = s + 16, r;
1763 
1764 	while (s < lim)
1765 		if ((r = (*d++ ^ *s++)) != 0) {
1766 			while (r < 128) {
1767 				match++;
1768 				r <<= 1;
1769 			}
1770 			break;
1771 		} else
1772 			match += 8;
1773 	return match;
1774 }
1775 
1776 /* XXX: to be scope conscious */
1777 int
1778 in6_are_prefix_equal(struct in6_addr *p1, struct in6_addr *p2, int len)
1779 {
1780 	int bytelen, bitlen;
1781 
1782 	/* sanity check */
1783 	if (0 > len || len > 128) {
1784 		log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n",
1785 		    len);
1786 		return (0);
1787 	}
1788 
1789 	bytelen = len / 8;
1790 	bitlen = len % 8;
1791 
1792 	if (bcmp(&p1->s6_addr, &p2->s6_addr, bytelen))
1793 		return (0);
1794 	if (bitlen != 0 &&
1795 	    p1->s6_addr[bytelen] >> (8 - bitlen) !=
1796 	    p2->s6_addr[bytelen] >> (8 - bitlen))
1797 		return (0);
1798 
1799 	return (1);
1800 }
1801 
1802 void
1803 in6_prefixlen2mask(struct in6_addr *maskp, int len)
1804 {
1805 	u_char maskarray[8] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
1806 	int bytelen, bitlen, i;
1807 
1808 	/* sanity check */
1809 	if (0 > len || len > 128) {
1810 		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
1811 		    len);
1812 		return;
1813 	}
1814 
1815 	bzero(maskp, sizeof(*maskp));
1816 	bytelen = len / 8;
1817 	bitlen = len % 8;
1818 	for (i = 0; i < bytelen; i++)
1819 		maskp->s6_addr[i] = 0xff;
1820 	if (bitlen)
1821 		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
1822 }
1823 
1824 /*
1825  * return the best address out of the same scope. if no address was
1826  * found, return the first valid address from designated IF.
1827  */
1828 struct in6_ifaddr *
1829 in6_ifawithifp(struct ifnet *ifp, struct in6_addr *dst)
1830 {
1831 	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
1832 	struct ifaddr *ifa;
1833 	struct in6_ifaddr *besta = NULL;
1834 	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
1835 
1836 	NET_EPOCH_ASSERT();
1837 
1838 	dep[0] = dep[1] = NULL;
1839 
1840 	/*
1841 	 * We first look for addresses in the same scope.
1842 	 * If there is one, return it.
1843 	 * If two or more, return one which matches the dst longest.
1844 	 * If none, return one of global addresses assigned other ifs.
1845 	 */
1846 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1847 		if (ifa->ifa_addr->sa_family != AF_INET6)
1848 			continue;
1849 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
1850 			continue; /* XXX: is there any case to allow anycast? */
1851 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
1852 			continue; /* don't use this interface */
1853 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
1854 			continue;
1855 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
1856 			if (V_ip6_use_deprecated)
1857 				dep[0] = (struct in6_ifaddr *)ifa;
1858 			continue;
1859 		}
1860 
1861 		if (dst_scope == in6_addrscope(IFA_IN6(ifa))) {
1862 			/*
1863 			 * call in6_matchlen() as few as possible
1864 			 */
1865 			if (besta) {
1866 				if (blen == -1)
1867 					blen = in6_matchlen(&besta->ia_addr.sin6_addr, dst);
1868 				tlen = in6_matchlen(IFA_IN6(ifa), dst);
1869 				if (tlen > blen) {
1870 					blen = tlen;
1871 					besta = (struct in6_ifaddr *)ifa;
1872 				}
1873 			} else
1874 				besta = (struct in6_ifaddr *)ifa;
1875 		}
1876 	}
1877 	if (besta) {
1878 		ifa_ref(&besta->ia_ifa);
1879 		return (besta);
1880 	}
1881 
1882 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1883 		if (ifa->ifa_addr->sa_family != AF_INET6)
1884 			continue;
1885 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
1886 			continue; /* XXX: is there any case to allow anycast? */
1887 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
1888 			continue; /* don't use this interface */
1889 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
1890 			continue;
1891 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
1892 			if (V_ip6_use_deprecated)
1893 				dep[1] = (struct in6_ifaddr *)ifa;
1894 			continue;
1895 		}
1896 
1897 		if (ifa != NULL)
1898 			ifa_ref(ifa);
1899 		return (struct in6_ifaddr *)ifa;
1900 	}
1901 
1902 	/* use the last-resort values, that are, deprecated addresses */
1903 	if (dep[0]) {
1904 		ifa_ref((struct ifaddr *)dep[0]);
1905 		return dep[0];
1906 	}
1907 	if (dep[1]) {
1908 		ifa_ref((struct ifaddr *)dep[1]);
1909 		return dep[1];
1910 	}
1911 
1912 	return NULL;
1913 }
1914 
1915 /*
1916  * perform DAD when interface becomes IFF_UP.
1917  */
1918 void
1919 in6_if_up(struct ifnet *ifp)
1920 {
1921 	struct epoch_tracker et;
1922 	struct ifaddr *ifa;
1923 	struct in6_ifaddr *ia;
1924 
1925 	NET_EPOCH_ENTER(et);
1926 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1927 		if (ifa->ifa_addr->sa_family != AF_INET6)
1928 			continue;
1929 		ia = (struct in6_ifaddr *)ifa;
1930 		if (ia->ia6_flags & IN6_IFF_TENTATIVE) {
1931 			/*
1932 			 * The TENTATIVE flag was likely set by hand
1933 			 * beforehand, implicitly indicating the need for DAD.
1934 			 * We may be able to skip the random delay in this
1935 			 * case, but we impose delays just in case.
1936 			 */
1937 			nd6_dad_start(ifa,
1938 			    arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz));
1939 		}
1940 	}
1941 	NET_EPOCH_EXIT(et);
1942 
1943 	/*
1944 	 * special cases, like 6to4, are handled in in6_ifattach
1945 	 */
1946 	in6_ifattach(ifp, NULL);
1947 }
1948 
1949 int
1950 in6if_do_dad(struct ifnet *ifp)
1951 {
1952 
1953 	if ((ifp->if_flags & IFF_LOOPBACK) != 0)
1954 		return (0);
1955 	if ((ifp->if_flags & IFF_MULTICAST) == 0)
1956 		return (0);
1957 	if ((ND_IFINFO(ifp)->flags &
1958 	    (ND6_IFF_IFDISABLED | ND6_IFF_NO_DAD)) != 0)
1959 		return (0);
1960 	return (1);
1961 }
1962 
1963 /*
1964  * Calculate max IPv6 MTU through all the interfaces and store it
1965  * to in6_maxmtu.
1966  */
1967 void
1968 in6_setmaxmtu(void)
1969 {
1970 	struct epoch_tracker et;
1971 	unsigned long maxmtu = 0;
1972 	struct ifnet *ifp;
1973 
1974 	NET_EPOCH_ENTER(et);
1975 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1976 		/* this function can be called during ifnet initialization */
1977 		if (!ifp->if_afdata[AF_INET6])
1978 			continue;
1979 		if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
1980 		    IN6_LINKMTU(ifp) > maxmtu)
1981 			maxmtu = IN6_LINKMTU(ifp);
1982 	}
1983 	NET_EPOCH_EXIT(et);
1984 	if (maxmtu)	/* update only when maxmtu is positive */
1985 		V_in6_maxmtu = maxmtu;
1986 }
1987 
1988 /*
1989  * Provide the length of interface identifiers to be used for the link attached
1990  * to the given interface.  The length should be defined in "IPv6 over
1991  * xxx-link" document.  Note that address architecture might also define
1992  * the length for a particular set of address prefixes, regardless of the
1993  * link type.  As clarified in rfc2462bis, those two definitions should be
1994  * consistent, and those really are as of August 2004.
1995  */
1996 int
1997 in6_if2idlen(struct ifnet *ifp)
1998 {
1999 	switch (ifp->if_type) {
2000 	case IFT_ETHER:		/* RFC2464 */
2001 	case IFT_PROPVIRTUAL:	/* XXX: no RFC. treat it as ether */
2002 	case IFT_L2VLAN:	/* ditto */
2003 	case IFT_BRIDGE:	/* bridge(4) only does Ethernet-like links */
2004 	case IFT_INFINIBAND:
2005 		return (64);
2006 	case IFT_PPP:		/* RFC2472 */
2007 		return (64);
2008 	case IFT_FRELAY:	/* RFC2590 */
2009 		return (64);
2010 	case IFT_IEEE1394:	/* RFC3146 */
2011 		return (64);
2012 	case IFT_GIF:
2013 		return (64);	/* draft-ietf-v6ops-mech-v2-07 */
2014 	case IFT_LOOP:
2015 		return (64);	/* XXX: is this really correct? */
2016 	default:
2017 		/*
2018 		 * Unknown link type:
2019 		 * It might be controversial to use the today's common constant
2020 		 * of 64 for these cases unconditionally.  For full compliance,
2021 		 * we should return an error in this case.  On the other hand,
2022 		 * if we simply miss the standard for the link type or a new
2023 		 * standard is defined for a new link type, the IFID length
2024 		 * is very likely to be the common constant.  As a compromise,
2025 		 * we always use the constant, but make an explicit notice
2026 		 * indicating the "unknown" case.
2027 		 */
2028 		printf("in6_if2idlen: unknown link type (%d)\n", ifp->if_type);
2029 		return (64);
2030 	}
2031 }
2032 
2033 struct in6_llentry {
2034 	struct llentry		base;
2035 };
2036 
2037 #define	IN6_LLTBL_DEFAULT_HSIZE	32
2038 #define	IN6_LLTBL_HASH(k, h) \
2039 	(((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1))
2040 
2041 /*
2042  * Do actual deallocation of @lle.
2043  */
2044 static void
2045 in6_lltable_destroy_lle_unlocked(epoch_context_t ctx)
2046 {
2047 	struct llentry *lle;
2048 
2049 	lle = __containerof(ctx, struct llentry, lle_epoch_ctx);
2050 	LLE_LOCK_DESTROY(lle);
2051 	LLE_REQ_DESTROY(lle);
2052 	free(lle, M_LLTABLE);
2053 }
2054 
2055 /*
2056  * Called by LLE_FREE_LOCKED when number of references
2057  * drops to zero.
2058  */
2059 static void
2060 in6_lltable_destroy_lle(struct llentry *lle)
2061 {
2062 
2063 	LLE_WUNLOCK(lle);
2064 	NET_EPOCH_CALL(in6_lltable_destroy_lle_unlocked, &lle->lle_epoch_ctx);
2065 }
2066 
2067 static struct llentry *
2068 in6_lltable_new(const struct in6_addr *addr6, u_int flags)
2069 {
2070 	struct in6_llentry *lle;
2071 
2072 	lle = malloc(sizeof(struct in6_llentry), M_LLTABLE, M_NOWAIT | M_ZERO);
2073 	if (lle == NULL)		/* NB: caller generates msg */
2074 		return NULL;
2075 
2076 	lle->base.r_l3addr.addr6 = *addr6;
2077 	lle->base.lle_refcnt = 1;
2078 	lle->base.lle_free = in6_lltable_destroy_lle;
2079 	LLE_LOCK_INIT(&lle->base);
2080 	LLE_REQ_INIT(&lle->base);
2081 	callout_init(&lle->base.lle_timer, 1);
2082 
2083 	return (&lle->base);
2084 }
2085 
2086 static int
2087 in6_lltable_match_prefix(const struct sockaddr *saddr,
2088     const struct sockaddr *smask, u_int flags, struct llentry *lle)
2089 {
2090 	const struct in6_addr *addr, *mask, *lle_addr;
2091 
2092 	addr = &((const struct sockaddr_in6 *)saddr)->sin6_addr;
2093 	mask = &((const struct sockaddr_in6 *)smask)->sin6_addr;
2094 	lle_addr = &lle->r_l3addr.addr6;
2095 
2096 	if (IN6_ARE_MASKED_ADDR_EQUAL(lle_addr, addr, mask) == 0)
2097 		return (0);
2098 
2099 	if (lle->la_flags & LLE_IFADDR) {
2100 
2101 		/*
2102 		 * Delete LLE_IFADDR records IFF address & flag matches.
2103 		 * Note that addr is the interface address within prefix
2104 		 * being matched.
2105 		 */
2106 		if (IN6_ARE_ADDR_EQUAL(addr, lle_addr) &&
2107 		    (flags & LLE_STATIC) != 0)
2108 			return (1);
2109 		return (0);
2110 	}
2111 
2112 	/* flags & LLE_STATIC means deleting both dynamic and static entries */
2113 	if ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC))
2114 		return (1);
2115 
2116 	return (0);
2117 }
2118 
2119 static void
2120 in6_lltable_free_entry(struct lltable *llt, struct llentry *lle)
2121 {
2122 	struct ifnet *ifp;
2123 
2124 	LLE_WLOCK_ASSERT(lle);
2125 	KASSERT(llt != NULL, ("lltable is NULL"));
2126 
2127 	/* Unlink entry from table */
2128 	if ((lle->la_flags & LLE_LINKED) != 0) {
2129 
2130 		ifp = llt->llt_ifp;
2131 		IF_AFDATA_WLOCK_ASSERT(ifp);
2132 		lltable_unlink_entry(llt, lle);
2133 	}
2134 
2135 	llentry_free(lle);
2136 }
2137 
2138 static int
2139 in6_lltable_rtcheck(struct ifnet *ifp,
2140 		    u_int flags,
2141 		    const struct sockaddr *l3addr)
2142 {
2143 	const struct sockaddr_in6 *sin6;
2144 	struct nhop6_basic nh6;
2145 	struct in6_addr dst;
2146 	uint32_t scopeid;
2147 	int error;
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 	error = fib6_lookup_nh_basic(fibnum, &dst, scopeid, 0, 0, &nh6);
2159 	if (error != 0 || (nh6.nh_flags & NHF_GATEWAY) || nh6.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