xref: /freebsd/sys/netinet6/in6.c (revision ee7231c92936b17c592ad9d4dd92a560ff320191)
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/route_ctl.h>
95 #include <net/route/nhop.h>
96 #include <net/if_dl.h>
97 #include <net/vnet.h>
98 
99 #include <netinet/in.h>
100 #include <netinet/in_var.h>
101 #include <net/if_llatbl.h>
102 #include <netinet/if_ether.h>
103 #include <netinet/in_systm.h>
104 #include <netinet/ip.h>
105 #include <netinet/in_pcb.h>
106 #include <netinet/ip_carp.h>
107 
108 #include <netinet/ip6.h>
109 #include <netinet6/ip6_var.h>
110 #include <netinet6/nd6.h>
111 #include <netinet6/mld6_var.h>
112 #include <netinet6/ip6_mroute.h>
113 #include <netinet6/in6_ifattach.h>
114 #include <netinet6/scope6_var.h>
115 #include <netinet6/in6_fib.h>
116 #include <netinet6/in6_pcb.h>
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 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 	fibnum = ia62ifa(ia)->ifa_ifp->if_fib;
191 
192 	if (cmd == RTM_ADD) {
193 		rt_addrmsg(cmd, &ia->ia_ifa, fibnum);
194 		rt_routemsg_info(cmd, &info, fibnum);
195 	} else if (cmd == RTM_DELETE) {
196 		rt_routemsg_info(cmd, &info, fibnum);
197 		rt_addrmsg(cmd, &ia->ia_ifa, fibnum);
198 	}
199 }
200 
201 int
202 in6_mask2len(struct in6_addr *mask, u_char *lim0)
203 {
204 	int x = 0, y;
205 	u_char *lim = lim0, *p;
206 
207 	/* ignore the scope_id part */
208 	if (lim0 == NULL || lim0 - (u_char *)mask > sizeof(*mask))
209 		lim = (u_char *)mask + sizeof(*mask);
210 	for (p = (u_char *)mask; p < lim; x++, p++) {
211 		if (*p != 0xff)
212 			break;
213 	}
214 	y = 0;
215 	if (p < lim) {
216 		for (y = 0; y < 8; y++) {
217 			if ((*p & (0x80 >> y)) == 0)
218 				break;
219 		}
220 	}
221 
222 	/*
223 	 * when the limit pointer is given, do a stricter check on the
224 	 * remaining bits.
225 	 */
226 	if (p < lim) {
227 		if (y != 0 && (*p & (0x00ff >> y)) != 0)
228 			return (-1);
229 		for (p = p + 1; p < lim; p++)
230 			if (*p != 0)
231 				return (-1);
232 	}
233 
234 	return x * 8 + y;
235 }
236 
237 #ifdef COMPAT_FREEBSD32
238 struct in6_ndifreq32 {
239 	char ifname[IFNAMSIZ];
240 	uint32_t ifindex;
241 };
242 #define	SIOCGDEFIFACE32_IN6	_IOWR('i', 86, struct in6_ndifreq32)
243 #endif
244 
245 int
246 in6_control(struct socket *so, u_long cmd, caddr_t data,
247     struct ifnet *ifp, struct thread *td)
248 {
249 	struct	in6_ifreq *ifr = (struct in6_ifreq *)data;
250 	struct	in6_ifaddr *ia = NULL;
251 	struct	in6_aliasreq *ifra = (struct in6_aliasreq *)data;
252 	struct sockaddr_in6 *sa6;
253 	int carp_attached = 0;
254 	int error;
255 	u_long ocmd = cmd;
256 
257 	/*
258 	 * Compat to make pre-10.x ifconfig(8) operable.
259 	 */
260 	if (cmd == OSIOCAIFADDR_IN6)
261 		cmd = SIOCAIFADDR_IN6;
262 
263 	switch (cmd) {
264 	case SIOCGETSGCNT_IN6:
265 	case SIOCGETMIFCNT_IN6:
266 		/*
267 		 * XXX mrt_ioctl has a 3rd, unused, FIB argument in route.c.
268 		 * We cannot see how that would be needed, so do not adjust the
269 		 * KPI blindly; more likely should clean up the IPv4 variant.
270 		 */
271 		return (mrt6_ioctl ? mrt6_ioctl(cmd, data) : EOPNOTSUPP);
272 	}
273 
274 	switch (cmd) {
275 	case SIOCAADDRCTL_POLICY:
276 	case SIOCDADDRCTL_POLICY:
277 		if (td != NULL) {
278 			error = priv_check(td, PRIV_NETINET_ADDRCTRL6);
279 			if (error)
280 				return (error);
281 		}
282 		return (in6_src_ioctl(cmd, data));
283 	}
284 
285 	if (ifp == NULL)
286 		return (EOPNOTSUPP);
287 
288 	switch (cmd) {
289 	case SIOCSNDFLUSH_IN6:
290 	case SIOCSPFXFLUSH_IN6:
291 	case SIOCSRTRFLUSH_IN6:
292 	case SIOCSDEFIFACE_IN6:
293 	case SIOCSIFINFO_FLAGS:
294 	case SIOCSIFINFO_IN6:
295 		if (td != NULL) {
296 			error = priv_check(td, PRIV_NETINET_ND6);
297 			if (error)
298 				return (error);
299 		}
300 		/* FALLTHROUGH */
301 	case OSIOCGIFINFO_IN6:
302 	case SIOCGIFINFO_IN6:
303 	case SIOCGNBRINFO_IN6:
304 	case SIOCGDEFIFACE_IN6:
305 		return (nd6_ioctl(cmd, data, ifp));
306 
307 #ifdef COMPAT_FREEBSD32
308 	case SIOCGDEFIFACE32_IN6:
309 		{
310 			struct in6_ndifreq ndif;
311 			struct in6_ndifreq32 *ndif32;
312 
313 			error = nd6_ioctl(SIOCGDEFIFACE_IN6, (caddr_t)&ndif,
314 			    ifp);
315 			if (error)
316 				return (error);
317 			ndif32 = (struct in6_ndifreq32 *)data;
318 			ndif32->ifindex = ndif.ifindex;
319 			return (0);
320 		}
321 #endif
322 	}
323 
324 	switch (cmd) {
325 	case SIOCSIFPREFIX_IN6:
326 	case SIOCDIFPREFIX_IN6:
327 	case SIOCAIFPREFIX_IN6:
328 	case SIOCCIFPREFIX_IN6:
329 	case SIOCSGIFPREFIX_IN6:
330 	case SIOCGIFPREFIX_IN6:
331 		log(LOG_NOTICE,
332 		    "prefix ioctls are now invalidated. "
333 		    "please use ifconfig.\n");
334 		return (EOPNOTSUPP);
335 	}
336 
337 	switch (cmd) {
338 	case SIOCSSCOPE6:
339 		if (td != NULL) {
340 			error = priv_check(td, PRIV_NETINET_SCOPE6);
341 			if (error)
342 				return (error);
343 		}
344 		/* FALLTHROUGH */
345 	case SIOCGSCOPE6:
346 	case SIOCGSCOPE6DEF:
347 		return (scope6_ioctl(cmd, data, ifp));
348 	}
349 
350 	/*
351 	 * Find address for this interface, if it exists.
352 	 *
353 	 * In netinet code, we have checked ifra_addr in SIOCSIF*ADDR operation
354 	 * only, and used the first interface address as the target of other
355 	 * operations (without checking ifra_addr).  This was because netinet
356 	 * code/API assumed at most 1 interface address per interface.
357 	 * Since IPv6 allows a node to assign multiple addresses
358 	 * on a single interface, we almost always look and check the
359 	 * presence of ifra_addr, and reject invalid ones here.
360 	 * It also decreases duplicated code among SIOC*_IN6 operations.
361 	 */
362 	switch (cmd) {
363 	case SIOCAIFADDR_IN6:
364 	case SIOCSIFPHYADDR_IN6:
365 		sa6 = &ifra->ifra_addr;
366 		break;
367 	case SIOCSIFADDR_IN6:
368 	case SIOCGIFADDR_IN6:
369 	case SIOCSIFDSTADDR_IN6:
370 	case SIOCSIFNETMASK_IN6:
371 	case SIOCGIFDSTADDR_IN6:
372 	case SIOCGIFNETMASK_IN6:
373 	case SIOCDIFADDR_IN6:
374 	case SIOCGIFPSRCADDR_IN6:
375 	case SIOCGIFPDSTADDR_IN6:
376 	case SIOCGIFAFLAG_IN6:
377 	case SIOCSNDFLUSH_IN6:
378 	case SIOCSPFXFLUSH_IN6:
379 	case SIOCSRTRFLUSH_IN6:
380 	case SIOCGIFALIFETIME_IN6:
381 	case SIOCGIFSTAT_IN6:
382 	case SIOCGIFSTAT_ICMP6:
383 		sa6 = &ifr->ifr_addr;
384 		break;
385 	case SIOCSIFADDR:
386 	case SIOCSIFBRDADDR:
387 	case SIOCSIFDSTADDR:
388 	case SIOCSIFNETMASK:
389 		/*
390 		 * Although we should pass any non-INET6 ioctl requests
391 		 * down to driver, we filter some legacy INET requests.
392 		 * Drivers trust SIOCSIFADDR et al to come from an already
393 		 * privileged layer, and do not perform any credentials
394 		 * checks or input validation.
395 		 */
396 		return (EINVAL);
397 	default:
398 		sa6 = NULL;
399 		break;
400 	}
401 	if (sa6 && sa6->sin6_family == AF_INET6) {
402 		if (sa6->sin6_scope_id != 0)
403 			error = sa6_embedscope(sa6, 0);
404 		else
405 			error = in6_setscope(&sa6->sin6_addr, ifp, NULL);
406 		if (error != 0)
407 			return (error);
408 		if (td != NULL && (error = prison_check_ip6(td->td_ucred,
409 		    &sa6->sin6_addr)) != 0)
410 			return (error);
411 		ia = in6ifa_ifpwithaddr(ifp, &sa6->sin6_addr);
412 	} else
413 		ia = NULL;
414 
415 	switch (cmd) {
416 	case SIOCSIFADDR_IN6:
417 	case SIOCSIFDSTADDR_IN6:
418 	case SIOCSIFNETMASK_IN6:
419 		/*
420 		 * Since IPv6 allows a node to assign multiple addresses
421 		 * on a single interface, SIOCSIFxxx ioctls are deprecated.
422 		 */
423 		/* we decided to obsolete this command (20000704) */
424 		error = EINVAL;
425 		goto out;
426 
427 	case SIOCDIFADDR_IN6:
428 		/*
429 		 * for IPv4, we look for existing in_ifaddr here to allow
430 		 * "ifconfig if0 delete" to remove the first IPv4 address on
431 		 * the interface.  For IPv6, as the spec allows multiple
432 		 * interface address from the day one, we consider "remove the
433 		 * first one" semantics to be not preferable.
434 		 */
435 		if (ia == NULL) {
436 			error = EADDRNOTAVAIL;
437 			goto out;
438 		}
439 		/* FALLTHROUGH */
440 	case SIOCAIFADDR_IN6:
441 		/*
442 		 * We always require users to specify a valid IPv6 address for
443 		 * the corresponding operation.
444 		 */
445 		if (ifra->ifra_addr.sin6_family != AF_INET6 ||
446 		    ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6)) {
447 			error = EAFNOSUPPORT;
448 			goto out;
449 		}
450 
451 		if (td != NULL) {
452 			error = priv_check(td, (cmd == SIOCDIFADDR_IN6) ?
453 			    PRIV_NET_DELIFADDR : PRIV_NET_ADDIFADDR);
454 			if (error)
455 				goto out;
456 		}
457 		/* FALLTHROUGH */
458 	case SIOCGIFSTAT_IN6:
459 	case SIOCGIFSTAT_ICMP6:
460 		if (ifp->if_afdata[AF_INET6] == NULL) {
461 			error = EPFNOSUPPORT;
462 			goto out;
463 		}
464 		break;
465 
466 	case SIOCGIFADDR_IN6:
467 		/* This interface is basically deprecated. use SIOCGIFCONF. */
468 		/* FALLTHROUGH */
469 	case SIOCGIFAFLAG_IN6:
470 	case SIOCGIFNETMASK_IN6:
471 	case SIOCGIFDSTADDR_IN6:
472 	case SIOCGIFALIFETIME_IN6:
473 		/* must think again about its semantics */
474 		if (ia == NULL) {
475 			error = EADDRNOTAVAIL;
476 			goto out;
477 		}
478 		break;
479 	}
480 
481 	switch (cmd) {
482 	case SIOCGIFADDR_IN6:
483 		ifr->ifr_addr = ia->ia_addr;
484 		if ((error = sa6_recoverscope(&ifr->ifr_addr)) != 0)
485 			goto out;
486 		break;
487 
488 	case SIOCGIFDSTADDR_IN6:
489 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
490 			error = EINVAL;
491 			goto out;
492 		}
493 		ifr->ifr_dstaddr = ia->ia_dstaddr;
494 		if ((error = sa6_recoverscope(&ifr->ifr_dstaddr)) != 0)
495 			goto out;
496 		break;
497 
498 	case SIOCGIFNETMASK_IN6:
499 		ifr->ifr_addr = ia->ia_prefixmask;
500 		break;
501 
502 	case SIOCGIFAFLAG_IN6:
503 		ifr->ifr_ifru.ifru_flags6 = ia->ia6_flags;
504 		break;
505 
506 	case SIOCGIFSTAT_IN6:
507 		COUNTER_ARRAY_COPY(((struct in6_ifextra *)
508 		    ifp->if_afdata[AF_INET6])->in6_ifstat,
509 		    &ifr->ifr_ifru.ifru_stat,
510 		    sizeof(struct in6_ifstat) / sizeof(uint64_t));
511 		break;
512 
513 	case SIOCGIFSTAT_ICMP6:
514 		COUNTER_ARRAY_COPY(((struct in6_ifextra *)
515 		    ifp->if_afdata[AF_INET6])->icmp6_ifstat,
516 		    &ifr->ifr_ifru.ifru_icmp6stat,
517 		    sizeof(struct icmp6_ifstat) / sizeof(uint64_t));
518 		break;
519 
520 	case SIOCGIFALIFETIME_IN6:
521 		ifr->ifr_ifru.ifru_lifetime = ia->ia6_lifetime;
522 		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
523 			time_t maxexpire;
524 			struct in6_addrlifetime *retlt =
525 			    &ifr->ifr_ifru.ifru_lifetime;
526 
527 			/*
528 			 * XXX: adjust expiration time assuming time_t is
529 			 * signed.
530 			 */
531 			maxexpire = (-1) &
532 			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
533 			if (ia->ia6_lifetime.ia6t_vltime <
534 			    maxexpire - ia->ia6_updatetime) {
535 				retlt->ia6t_expire = ia->ia6_updatetime +
536 				    ia->ia6_lifetime.ia6t_vltime;
537 			} else
538 				retlt->ia6t_expire = maxexpire;
539 		}
540 		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
541 			time_t maxexpire;
542 			struct in6_addrlifetime *retlt =
543 			    &ifr->ifr_ifru.ifru_lifetime;
544 
545 			/*
546 			 * XXX: adjust expiration time assuming time_t is
547 			 * signed.
548 			 */
549 			maxexpire = (-1) &
550 			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
551 			if (ia->ia6_lifetime.ia6t_pltime <
552 			    maxexpire - ia->ia6_updatetime) {
553 				retlt->ia6t_preferred = ia->ia6_updatetime +
554 				    ia->ia6_lifetime.ia6t_pltime;
555 			} else
556 				retlt->ia6t_preferred = maxexpire;
557 		}
558 		break;
559 
560 	case SIOCAIFADDR_IN6:
561 	{
562 		struct nd_prefixctl pr0;
563 		struct nd_prefix *pr;
564 
565 		/*
566 		 * first, make or update the interface address structure,
567 		 * and link it to the list.
568 		 */
569 		if ((error = in6_update_ifa(ifp, ifra, ia, 0)) != 0)
570 			goto out;
571 		if (ia != NULL) {
572 			if (ia->ia_ifa.ifa_carp)
573 				(*carp_detach_p)(&ia->ia_ifa, true);
574 			ifa_free(&ia->ia_ifa);
575 		}
576 		if ((ia = in6ifa_ifpwithaddr(ifp, &ifra->ifra_addr.sin6_addr))
577 		    == NULL) {
578 			/*
579 			 * this can happen when the user specify the 0 valid
580 			 * lifetime.
581 			 */
582 			break;
583 		}
584 
585 		if (cmd == ocmd && ifra->ifra_vhid > 0) {
586 			if (carp_attach_p != NULL)
587 				error = (*carp_attach_p)(&ia->ia_ifa,
588 				    ifra->ifra_vhid);
589 			else
590 				error = EPROTONOSUPPORT;
591 			if (error)
592 				goto out;
593 			else
594 				carp_attached = 1;
595 		}
596 
597 		/*
598 		 * then, make the prefix on-link on the interface.
599 		 * XXX: we'd rather create the prefix before the address, but
600 		 * we need at least one address to install the corresponding
601 		 * interface route, so we configure the address first.
602 		 */
603 
604 		/*
605 		 * convert mask to prefix length (prefixmask has already
606 		 * been validated in in6_update_ifa().
607 		 */
608 		bzero(&pr0, sizeof(pr0));
609 		pr0.ndpr_ifp = ifp;
610 		pr0.ndpr_plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
611 		    NULL);
612 		if (pr0.ndpr_plen == 128) {
613 			/* we don't need to install a host route. */
614 			goto aifaddr_out;
615 		}
616 		pr0.ndpr_prefix = ifra->ifra_addr;
617 		/* apply the mask for safety. */
618 		IN6_MASK_ADDR(&pr0.ndpr_prefix.sin6_addr,
619 		    &ifra->ifra_prefixmask.sin6_addr);
620 
621 		/*
622 		 * XXX: since we don't have an API to set prefix (not address)
623 		 * lifetimes, we just use the same lifetimes as addresses.
624 		 * The (temporarily) installed lifetimes can be overridden by
625 		 * later advertised RAs (when accept_rtadv is non 0), which is
626 		 * an intended behavior.
627 		 */
628 		pr0.ndpr_raf_onlink = 1; /* should be configurable? */
629 		pr0.ndpr_raf_auto =
630 		    ((ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0);
631 		pr0.ndpr_vltime = ifra->ifra_lifetime.ia6t_vltime;
632 		pr0.ndpr_pltime = ifra->ifra_lifetime.ia6t_pltime;
633 
634 		/* add the prefix if not yet. */
635 		if ((pr = nd6_prefix_lookup(&pr0)) == NULL) {
636 			/*
637 			 * nd6_prelist_add will install the corresponding
638 			 * interface route.
639 			 */
640 			if ((error = nd6_prelist_add(&pr0, NULL, &pr)) != 0) {
641 				if (carp_attached)
642 					(*carp_detach_p)(&ia->ia_ifa, false);
643 				goto out;
644 			}
645 		}
646 
647 		/* relate the address to the prefix */
648 		if (ia->ia6_ndpr == NULL) {
649 			ia->ia6_ndpr = pr;
650 			pr->ndpr_addrcnt++;
651 
652 			/*
653 			 * If this is the first autoconf address from the
654 			 * prefix, create a temporary address as well
655 			 * (when required).
656 			 */
657 			if ((ia->ia6_flags & IN6_IFF_AUTOCONF) &&
658 			    V_ip6_use_tempaddr && pr->ndpr_addrcnt == 1) {
659 				int e;
660 				if ((e = in6_tmpifadd(ia, 1, 0)) != 0) {
661 					log(LOG_NOTICE, "in6_control: failed "
662 					    "to create a temporary address, "
663 					    "errno=%d\n", e);
664 				}
665 			}
666 		}
667 		nd6_prefix_rele(pr);
668 
669 		/*
670 		 * this might affect the status of autoconfigured addresses,
671 		 * that is, this address might make other addresses detached.
672 		 */
673 		pfxlist_onlink_check();
674 
675 aifaddr_out:
676 		/*
677 		 * Try to clear the flag when a new IPv6 address is added
678 		 * onto an IFDISABLED interface and it succeeds.
679 		 */
680 		if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) {
681 			struct in6_ndireq nd;
682 
683 			memset(&nd, 0, sizeof(nd));
684 			nd.ndi.flags = ND_IFINFO(ifp)->flags;
685 			nd.ndi.flags &= ~ND6_IFF_IFDISABLED;
686 			if (nd6_ioctl(SIOCSIFINFO_FLAGS, (caddr_t)&nd, ifp) < 0)
687 				log(LOG_NOTICE, "SIOCAIFADDR_IN6: "
688 				    "SIOCSIFINFO_FLAGS for -ifdisabled "
689 				    "failed.");
690 			/*
691 			 * Ignore failure of clearing the flag intentionally.
692 			 * The failure means address duplication was detected.
693 			 */
694 		}
695 		break;
696 	}
697 
698 	case SIOCDIFADDR_IN6:
699 	{
700 		struct nd_prefix *pr;
701 
702 		/*
703 		 * If the address being deleted is the only one that owns
704 		 * the corresponding prefix, expire the prefix as well.
705 		 * XXX: theoretically, we don't have to worry about such
706 		 * relationship, since we separate the address management
707 		 * and the prefix management.  We do this, however, to provide
708 		 * as much backward compatibility as possible in terms of
709 		 * the ioctl operation.
710 		 * Note that in6_purgeaddr() will decrement ndpr_addrcnt.
711 		 */
712 		pr = ia->ia6_ndpr;
713 		in6_purgeaddr(&ia->ia_ifa);
714 		if (pr != NULL && pr->ndpr_addrcnt == 0) {
715 			ND6_WLOCK();
716 			nd6_prefix_unlink(pr, NULL);
717 			ND6_WUNLOCK();
718 			nd6_prefix_del(pr);
719 		}
720 		EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, &ia->ia_ifa,
721 		    IFADDR_EVENT_DEL);
722 		break;
723 	}
724 
725 	default:
726 		if (ifp->if_ioctl == NULL) {
727 			error = EOPNOTSUPP;
728 			goto out;
729 		}
730 		error = (*ifp->if_ioctl)(ifp, cmd, data);
731 		goto out;
732 	}
733 
734 	error = 0;
735 out:
736 	if (ia != NULL)
737 		ifa_free(&ia->ia_ifa);
738 	return (error);
739 }
740 
741 static struct in6_multi_mship *
742 in6_joingroup_legacy(struct ifnet *ifp, const struct in6_addr *mcaddr,
743     int *errorp, int delay)
744 {
745 	struct in6_multi_mship *imm;
746 	int error;
747 
748 	imm = malloc(sizeof(*imm), M_IP6MADDR, M_NOWAIT);
749 	if (imm == NULL) {
750 		*errorp = ENOBUFS;
751 		return (NULL);
752 	}
753 
754 	delay = (delay * PR_FASTHZ) / hz;
755 
756 	error = in6_joingroup(ifp, mcaddr, NULL, &imm->i6mm_maddr, delay);
757 	if (error) {
758 		*errorp = error;
759 		free(imm, M_IP6MADDR);
760 		return (NULL);
761 	}
762 
763 	return (imm);
764 }
765 /*
766  * Join necessary multicast groups.  Factored out from in6_update_ifa().
767  * This entire work should only be done once, for the default FIB.
768  */
769 static int
770 in6_update_ifa_join_mc(struct ifnet *ifp, struct in6_aliasreq *ifra,
771     struct in6_ifaddr *ia, int flags, struct in6_multi **in6m_sol)
772 {
773 	char ip6buf[INET6_ADDRSTRLEN];
774 	struct in6_addr mltaddr;
775 	struct in6_multi_mship *imm;
776 	int delay, error;
777 
778 	KASSERT(in6m_sol != NULL, ("%s: in6m_sol is NULL", __func__));
779 
780 	/* Join solicited multicast addr for new host id. */
781 	bzero(&mltaddr, sizeof(struct in6_addr));
782 	mltaddr.s6_addr32[0] = IPV6_ADDR_INT32_MLL;
783 	mltaddr.s6_addr32[2] = htonl(1);
784 	mltaddr.s6_addr32[3] = ifra->ifra_addr.sin6_addr.s6_addr32[3];
785 	mltaddr.s6_addr8[12] = 0xff;
786 	if ((error = in6_setscope(&mltaddr, ifp, NULL)) != 0) {
787 		/* XXX: should not happen */
788 		log(LOG_ERR, "%s: in6_setscope failed\n", __func__);
789 		goto cleanup;
790 	}
791 	delay = error = 0;
792 	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
793 		/*
794 		 * We need a random delay for DAD on the address being
795 		 * configured.  It also means delaying transmission of the
796 		 * corresponding MLD report to avoid report collision.
797 		 * [RFC 4861, Section 6.3.7]
798 		 */
799 		delay = arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz);
800 	}
801 	imm = in6_joingroup_legacy(ifp, &mltaddr, &error, delay);
802 	if (imm == NULL) {
803 		nd6log((LOG_WARNING, "%s: in6_joingroup failed for %s on %s "
804 		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf, &mltaddr),
805 		    if_name(ifp), error));
806 		goto cleanup;
807 	}
808 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
809 	*in6m_sol = imm->i6mm_maddr;
810 
811 	/*
812 	 * Join link-local all-nodes address.
813 	 */
814 	mltaddr = in6addr_linklocal_allnodes;
815 	if ((error = in6_setscope(&mltaddr, ifp, NULL)) != 0)
816 		goto cleanup; /* XXX: should not fail */
817 
818 	imm = in6_joingroup_legacy(ifp, &mltaddr, &error, 0);
819 	if (imm == NULL) {
820 		nd6log((LOG_WARNING, "%s: in6_joingroup failed for %s on %s "
821 		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf, &mltaddr),
822 		    if_name(ifp), error));
823 		goto cleanup;
824 	}
825 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
826 
827 	/*
828 	 * Join node information group address.
829 	 */
830 	delay = 0;
831 	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
832 		/*
833 		 * The spec does not say anything about delay for this group,
834 		 * but the same logic should apply.
835 		 */
836 		delay = arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz);
837 	}
838 	if (in6_nigroup(ifp, NULL, -1, &mltaddr) == 0) {
839 		/* XXX jinmei */
840 		imm = in6_joingroup_legacy(ifp, &mltaddr, &error, delay);
841 		if (imm == NULL)
842 			nd6log((LOG_WARNING,
843 			    "%s: in6_joingroup failed for %s on %s "
844 			    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
845 			    &mltaddr), if_name(ifp), error));
846 			/* XXX not very fatal, go on... */
847 		else
848 			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
849 	}
850 	if (V_icmp6_nodeinfo_oldmcprefix &&
851 	    in6_nigroup_oldmcprefix(ifp, NULL, -1, &mltaddr) == 0) {
852 		imm = in6_joingroup_legacy(ifp, &mltaddr, &error, delay);
853 		if (imm == NULL)
854 			nd6log((LOG_WARNING,
855 			    "%s: in6_joingroup failed for %s on %s "
856 			    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
857 			    &mltaddr), if_name(ifp), error));
858 			/* XXX not very fatal, go on... */
859 		else
860 			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
861 	}
862 
863 	/*
864 	 * Join interface-local all-nodes address.
865 	 * (ff01::1%ifN, and ff01::%ifN/32)
866 	 */
867 	mltaddr = in6addr_nodelocal_allnodes;
868 	if ((error = in6_setscope(&mltaddr, ifp, NULL)) != 0)
869 		goto cleanup; /* XXX: should not fail */
870 
871 	imm = in6_joingroup_legacy(ifp, &mltaddr, &error, 0);
872 	if (imm == NULL) {
873 		nd6log((LOG_WARNING, "%s: in6_joingroup failed for %s on %s "
874 		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
875 		    &mltaddr), if_name(ifp), error));
876 		goto cleanup;
877 	}
878 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
879 
880 cleanup:
881 	return (error);
882 }
883 
884 /*
885  * Update parameters of an IPv6 interface address.
886  * If necessary, a new entry is created and linked into address chains.
887  * This function is separated from in6_control().
888  */
889 int
890 in6_update_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
891     struct in6_ifaddr *ia, int flags)
892 {
893 	int error, hostIsNew = 0;
894 
895 	if ((error = in6_validate_ifra(ifp, ifra, ia, flags)) != 0)
896 		return (error);
897 
898 	if (ia == NULL) {
899 		hostIsNew = 1;
900 		if ((ia = in6_alloc_ifa(ifp, ifra, flags)) == NULL)
901 			return (ENOBUFS);
902 	}
903 
904 	error = in6_update_ifa_internal(ifp, ifra, ia, hostIsNew, flags);
905 	if (error != 0) {
906 		if (hostIsNew != 0) {
907 			in6_unlink_ifa(ia, ifp);
908 			ifa_free(&ia->ia_ifa);
909 		}
910 		return (error);
911 	}
912 
913 	if (hostIsNew)
914 		error = in6_broadcast_ifa(ifp, ifra, ia, flags);
915 
916 	return (error);
917 }
918 
919 /*
920  * Fill in basic IPv6 address request info.
921  */
922 void
923 in6_prepare_ifra(struct in6_aliasreq *ifra, const struct in6_addr *addr,
924     const struct in6_addr *mask)
925 {
926 
927 	memset(ifra, 0, sizeof(struct in6_aliasreq));
928 
929 	ifra->ifra_addr.sin6_family = AF_INET6;
930 	ifra->ifra_addr.sin6_len = sizeof(struct sockaddr_in6);
931 	if (addr != NULL)
932 		ifra->ifra_addr.sin6_addr = *addr;
933 
934 	ifra->ifra_prefixmask.sin6_family = AF_INET6;
935 	ifra->ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
936 	if (mask != NULL)
937 		ifra->ifra_prefixmask.sin6_addr = *mask;
938 }
939 
940 static int
941 in6_validate_ifra(struct ifnet *ifp, struct in6_aliasreq *ifra,
942     struct in6_ifaddr *ia, int flags)
943 {
944 	int plen = -1;
945 	struct sockaddr_in6 dst6;
946 	struct in6_addrlifetime *lt;
947 	char ip6buf[INET6_ADDRSTRLEN];
948 
949 	/* Validate parameters */
950 	if (ifp == NULL || ifra == NULL) /* this maybe redundant */
951 		return (EINVAL);
952 
953 	/*
954 	 * The destination address for a p2p link must have a family
955 	 * of AF_UNSPEC or AF_INET6.
956 	 */
957 	if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
958 	    ifra->ifra_dstaddr.sin6_family != AF_INET6 &&
959 	    ifra->ifra_dstaddr.sin6_family != AF_UNSPEC)
960 		return (EAFNOSUPPORT);
961 
962 	/*
963 	 * Validate address
964 	 */
965 	if (ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6) ||
966 	    ifra->ifra_addr.sin6_family != AF_INET6)
967 		return (EINVAL);
968 
969 	/*
970 	 * validate ifra_prefixmask.  don't check sin6_family, netmask
971 	 * does not carry fields other than sin6_len.
972 	 */
973 	if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6))
974 		return (EINVAL);
975 	/*
976 	 * Because the IPv6 address architecture is classless, we require
977 	 * users to specify a (non 0) prefix length (mask) for a new address.
978 	 * We also require the prefix (when specified) mask is valid, and thus
979 	 * reject a non-consecutive mask.
980 	 */
981 	if (ia == NULL && ifra->ifra_prefixmask.sin6_len == 0)
982 		return (EINVAL);
983 	if (ifra->ifra_prefixmask.sin6_len != 0) {
984 		plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
985 		    (u_char *)&ifra->ifra_prefixmask +
986 		    ifra->ifra_prefixmask.sin6_len);
987 		if (plen <= 0)
988 			return (EINVAL);
989 	} else {
990 		/*
991 		 * In this case, ia must not be NULL.  We just use its prefix
992 		 * length.
993 		 */
994 		plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
995 	}
996 	/*
997 	 * If the destination address on a p2p interface is specified,
998 	 * and the address is a scoped one, validate/set the scope
999 	 * zone identifier.
1000 	 */
1001 	dst6 = ifra->ifra_dstaddr;
1002 	if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) != 0 &&
1003 	    (dst6.sin6_family == AF_INET6)) {
1004 		struct in6_addr in6_tmp;
1005 		u_int32_t zoneid;
1006 
1007 		in6_tmp = dst6.sin6_addr;
1008 		if (in6_setscope(&in6_tmp, ifp, &zoneid))
1009 			return (EINVAL); /* XXX: should be impossible */
1010 
1011 		if (dst6.sin6_scope_id != 0) {
1012 			if (dst6.sin6_scope_id != zoneid)
1013 				return (EINVAL);
1014 		} else		/* user omit to specify the ID. */
1015 			dst6.sin6_scope_id = zoneid;
1016 
1017 		/* convert into the internal form */
1018 		if (sa6_embedscope(&dst6, 0))
1019 			return (EINVAL); /* XXX: should be impossible */
1020 	}
1021 	/* Modify original ifra_dstaddr to reflect changes */
1022 	ifra->ifra_dstaddr = dst6;
1023 
1024 	/*
1025 	 * The destination address can be specified only for a p2p or a
1026 	 * loopback interface.  If specified, the corresponding prefix length
1027 	 * must be 128.
1028 	 */
1029 	if (ifra->ifra_dstaddr.sin6_family == AF_INET6) {
1030 		if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) == 0) {
1031 			/* XXX: noisy message */
1032 			nd6log((LOG_INFO, "in6_update_ifa: a destination can "
1033 			    "be specified for a p2p or a loopback IF only\n"));
1034 			return (EINVAL);
1035 		}
1036 		if (plen != 128) {
1037 			nd6log((LOG_INFO, "in6_update_ifa: prefixlen should "
1038 			    "be 128 when dstaddr is specified\n"));
1039 			return (EINVAL);
1040 		}
1041 	}
1042 	/* lifetime consistency check */
1043 	lt = &ifra->ifra_lifetime;
1044 	if (lt->ia6t_pltime > lt->ia6t_vltime)
1045 		return (EINVAL);
1046 	if (lt->ia6t_vltime == 0) {
1047 		/*
1048 		 * the following log might be noisy, but this is a typical
1049 		 * configuration mistake or a tool's bug.
1050 		 */
1051 		nd6log((LOG_INFO,
1052 		    "in6_update_ifa: valid lifetime is 0 for %s\n",
1053 		    ip6_sprintf(ip6buf, &ifra->ifra_addr.sin6_addr)));
1054 
1055 		if (ia == NULL)
1056 			return (0); /* there's nothing to do */
1057 	}
1058 
1059 	/* Check prefix mask */
1060 	if (ia != NULL && ifra->ifra_prefixmask.sin6_len != 0) {
1061 		/*
1062 		 * We prohibit changing the prefix length of an existing
1063 		 * address, because
1064 		 * + such an operation should be rare in IPv6, and
1065 		 * + the operation would confuse prefix management.
1066 		 */
1067 		if (ia->ia_prefixmask.sin6_len != 0 &&
1068 		    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) != plen) {
1069 			nd6log((LOG_INFO, "in6_validate_ifa: the prefix length "
1070 			    "of an existing %s address should not be changed\n",
1071 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1072 
1073 			return (EINVAL);
1074 		}
1075 	}
1076 
1077 	return (0);
1078 }
1079 
1080 /*
1081  * Allocate a new ifaddr and link it into chains.
1082  */
1083 static struct in6_ifaddr *
1084 in6_alloc_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra, int flags)
1085 {
1086 	struct in6_ifaddr *ia;
1087 
1088 	/*
1089 	 * When in6_alloc_ifa() is called in a process of a received
1090 	 * RA, it is called under an interrupt context.  So, we should
1091 	 * call malloc with M_NOWAIT.
1092 	 */
1093 	ia = (struct in6_ifaddr *)ifa_alloc(sizeof(*ia), M_NOWAIT);
1094 	if (ia == NULL)
1095 		return (NULL);
1096 	LIST_INIT(&ia->ia6_memberships);
1097 	/* Initialize the address and masks, and put time stamp */
1098 	ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
1099 	ia->ia_addr.sin6_family = AF_INET6;
1100 	ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
1101 	/* XXX: Can we assign ,sin6_addr and skip the rest? */
1102 	ia->ia_addr = ifra->ifra_addr;
1103 	ia->ia6_createtime = time_uptime;
1104 	if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
1105 		/*
1106 		 * Some functions expect that ifa_dstaddr is not
1107 		 * NULL for p2p interfaces.
1108 		 */
1109 		ia->ia_ifa.ifa_dstaddr =
1110 		    (struct sockaddr *)&ia->ia_dstaddr;
1111 	} else {
1112 		ia->ia_ifa.ifa_dstaddr = NULL;
1113 	}
1114 
1115 	/* set prefix mask if any */
1116 	ia->ia_ifa.ifa_netmask = (struct sockaddr *)&ia->ia_prefixmask;
1117 	if (ifra->ifra_prefixmask.sin6_len != 0) {
1118 		ia->ia_prefixmask.sin6_family = AF_INET6;
1119 		ia->ia_prefixmask.sin6_len = ifra->ifra_prefixmask.sin6_len;
1120 		ia->ia_prefixmask.sin6_addr = ifra->ifra_prefixmask.sin6_addr;
1121 	}
1122 
1123 	ia->ia_ifp = ifp;
1124 	ifa_ref(&ia->ia_ifa);			/* if_addrhead */
1125 	IF_ADDR_WLOCK(ifp);
1126 	CK_STAILQ_INSERT_TAIL(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
1127 	IF_ADDR_WUNLOCK(ifp);
1128 
1129 	ifa_ref(&ia->ia_ifa);			/* in6_ifaddrhead */
1130 	IN6_IFADDR_WLOCK();
1131 	CK_STAILQ_INSERT_TAIL(&V_in6_ifaddrhead, ia, ia_link);
1132 	CK_LIST_INSERT_HEAD(IN6ADDR_HASH(&ia->ia_addr.sin6_addr), ia, ia6_hash);
1133 	IN6_IFADDR_WUNLOCK();
1134 
1135 	return (ia);
1136 }
1137 
1138 /*
1139  * Update/configure interface address parameters:
1140  *
1141  * 1) Update lifetime
1142  * 2) Update interface metric ad flags
1143  * 3) Notify other subsystems
1144  */
1145 static int
1146 in6_update_ifa_internal(struct ifnet *ifp, struct in6_aliasreq *ifra,
1147     struct in6_ifaddr *ia, int hostIsNew, int flags)
1148 {
1149 	int error;
1150 
1151 	/* update timestamp */
1152 	ia->ia6_updatetime = time_uptime;
1153 
1154 	/*
1155 	 * Set lifetimes.  We do not refer to ia6t_expire and ia6t_preferred
1156 	 * to see if the address is deprecated or invalidated, but initialize
1157 	 * these members for applications.
1158 	 */
1159 	ia->ia6_lifetime = ifra->ifra_lifetime;
1160 	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1161 		ia->ia6_lifetime.ia6t_expire =
1162 		    time_uptime + ia->ia6_lifetime.ia6t_vltime;
1163 	} else
1164 		ia->ia6_lifetime.ia6t_expire = 0;
1165 	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1166 		ia->ia6_lifetime.ia6t_preferred =
1167 		    time_uptime + ia->ia6_lifetime.ia6t_pltime;
1168 	} else
1169 		ia->ia6_lifetime.ia6t_preferred = 0;
1170 
1171 	/*
1172 	 * backward compatibility - if IN6_IFF_DEPRECATED is set from the
1173 	 * userland, make it deprecated.
1174 	 */
1175 	if ((ifra->ifra_flags & IN6_IFF_DEPRECATED) != 0) {
1176 		ia->ia6_lifetime.ia6t_pltime = 0;
1177 		ia->ia6_lifetime.ia6t_preferred = time_uptime;
1178 	}
1179 
1180 	/*
1181 	 * configure address flags.
1182 	 */
1183 	ia->ia6_flags = ifra->ifra_flags;
1184 
1185 	/*
1186 	 * Make the address tentative before joining multicast addresses,
1187 	 * so that corresponding MLD responses would not have a tentative
1188 	 * source address.
1189 	 */
1190 	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/* safety */
1191 
1192 	/*
1193 	 * DAD should be performed for an new address or addresses on
1194 	 * an interface with ND6_IFF_IFDISABLED.
1195 	 */
1196 	if (in6if_do_dad(ifp) &&
1197 	    (hostIsNew || (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)))
1198 		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1199 
1200 	/* notify other subsystems */
1201 	error = in6_notify_ifa(ifp, ia, ifra, hostIsNew);
1202 
1203 	return (error);
1204 }
1205 
1206 /*
1207  * Do link-level ifa job:
1208  * 1) Add lle entry for added address
1209  * 2) Notifies routing socket users about new address
1210  * 3) join appropriate multicast group
1211  * 4) start DAD if enabled
1212  */
1213 static int
1214 in6_broadcast_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
1215     struct in6_ifaddr *ia, int flags)
1216 {
1217 	struct in6_multi *in6m_sol;
1218 	int error = 0;
1219 
1220 	/* Add local address to lltable, if necessary (ex. on p2p link). */
1221 	if ((error = nd6_add_ifa_lle(ia)) != 0) {
1222 		in6_purgeaddr(&ia->ia_ifa);
1223 		ifa_free(&ia->ia_ifa);
1224 		return (error);
1225 	}
1226 
1227 	/* Join necessary multicast groups. */
1228 	in6m_sol = NULL;
1229 	if ((ifp->if_flags & IFF_MULTICAST) != 0) {
1230 		error = in6_update_ifa_join_mc(ifp, ifra, ia, flags, &in6m_sol);
1231 		if (error != 0) {
1232 			in6_purgeaddr(&ia->ia_ifa);
1233 			ifa_free(&ia->ia_ifa);
1234 			return (error);
1235 		}
1236 	}
1237 
1238 	/* Perform DAD, if the address is TENTATIVE. */
1239 	if ((ia->ia6_flags & IN6_IFF_TENTATIVE)) {
1240 		int delay, mindelay, maxdelay;
1241 
1242 		delay = 0;
1243 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1244 			/*
1245 			 * We need to impose a delay before sending an NS
1246 			 * for DAD.  Check if we also needed a delay for the
1247 			 * corresponding MLD message.  If we did, the delay
1248 			 * should be larger than the MLD delay (this could be
1249 			 * relaxed a bit, but this simple logic is at least
1250 			 * safe).
1251 			 * XXX: Break data hiding guidelines and look at
1252 			 * state for the solicited multicast group.
1253 			 */
1254 			mindelay = 0;
1255 			if (in6m_sol != NULL &&
1256 			    in6m_sol->in6m_state == MLD_REPORTING_MEMBER) {
1257 				mindelay = in6m_sol->in6m_timer;
1258 			}
1259 			maxdelay = MAX_RTR_SOLICITATION_DELAY * hz;
1260 			if (maxdelay - mindelay == 0)
1261 				delay = 0;
1262 			else {
1263 				delay =
1264 				    (arc4random() % (maxdelay - mindelay)) +
1265 				    mindelay;
1266 			}
1267 		}
1268 		nd6_dad_start((struct ifaddr *)ia, delay);
1269 	}
1270 
1271 	in6_newaddrmsg(ia, RTM_ADD);
1272 	ifa_free(&ia->ia_ifa);
1273 	return (error);
1274 }
1275 
1276 /*
1277  * Adds or deletes interface route for p2p ifa.
1278  * Returns 0 on success or errno.
1279  */
1280 static int
1281 in6_handle_dstaddr_rtrequest(int cmd, struct in6_ifaddr *ia)
1282 {
1283 	struct epoch_tracker et;
1284 	struct ifaddr *ifa = &ia->ia_ifa;
1285 	int error;
1286 
1287 	/* Prepare gateway */
1288 	struct sockaddr_dl_short sdl = {
1289 		.sdl_family = AF_LINK,
1290 		.sdl_len = sizeof(struct sockaddr_dl_short),
1291 		.sdl_type = ifa->ifa_ifp->if_type,
1292 		.sdl_index = ifa->ifa_ifp->if_index,
1293 	};
1294 
1295 	struct sockaddr_in6 dst = {
1296 		.sin6_family = AF_INET6,
1297 		.sin6_len = sizeof(struct sockaddr_in6),
1298 		.sin6_addr = ia->ia_dstaddr.sin6_addr,
1299 	};
1300 
1301 	struct rt_addrinfo info = {
1302 		.rti_ifa = ifa,
1303 		.rti_flags = RTF_PINNED | RTF_HOST,
1304 		.rti_info = {
1305 			[RTAX_DST] = (struct sockaddr *)&dst,
1306 			[RTAX_GATEWAY] = (struct sockaddr *)&sdl,
1307 		},
1308 	};
1309 	/* Don't set additional per-gw filters on removal */
1310 
1311 	NET_EPOCH_ENTER(et);
1312 	error = rib_handle_ifaddr_info(ifa->ifa_ifp->if_fib, cmd, &info);
1313 	NET_EPOCH_EXIT(et);
1314 
1315 	return (error);
1316 }
1317 
1318 void
1319 in6_purgeaddr(struct ifaddr *ifa)
1320 {
1321 	struct ifnet *ifp = ifa->ifa_ifp;
1322 	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1323 	struct in6_multi_mship *imm;
1324 	int plen, error;
1325 
1326 	if (ifa->ifa_carp)
1327 		(*carp_detach_p)(ifa, false);
1328 
1329 	/*
1330 	 * Remove the loopback route to the interface address.
1331 	 * The check for the current setting of "nd6_useloopback"
1332 	 * is not needed.
1333 	 */
1334 	if (ia->ia_flags & IFA_RTSELF) {
1335 		error = ifa_del_loopback_route((struct ifaddr *)ia,
1336 		    (struct sockaddr *)&ia->ia_addr);
1337 		if (error == 0)
1338 			ia->ia_flags &= ~IFA_RTSELF;
1339 	}
1340 
1341 	/* stop DAD processing */
1342 	nd6_dad_stop(ifa);
1343 
1344 	/* Leave multicast groups. */
1345 	while ((imm = LIST_FIRST(&ia->ia6_memberships)) != NULL) {
1346 		LIST_REMOVE(imm, i6mm_chain);
1347 		if (imm->i6mm_maddr != NULL)
1348 			in6_leavegroup(imm->i6mm_maddr, NULL);
1349 		free(imm, M_IP6MADDR);
1350 	}
1351 	/* Check if we need to remove p2p route */
1352 	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1353 	if (ia->ia_dstaddr.sin6_family != AF_INET6)
1354 		plen = 0;
1355 	if ((ia->ia_flags & IFA_ROUTE) && plen == 128) {
1356 		error = in6_handle_dstaddr_rtrequest(RTM_DELETE, ia);
1357 		if (error != 0)
1358 			log(LOG_INFO, "%s: err=%d, destination address delete "
1359 			    "failed\n", __func__, error);
1360 		ia->ia_flags &= ~IFA_ROUTE;
1361 	}
1362 
1363 	in6_newaddrmsg(ia, RTM_DELETE);
1364 	in6_unlink_ifa(ia, ifp);
1365 }
1366 
1367 static void
1368 in6_unlink_ifa(struct in6_ifaddr *ia, struct ifnet *ifp)
1369 {
1370 	char ip6buf[INET6_ADDRSTRLEN];
1371 	int remove_lle;
1372 
1373 	IF_ADDR_WLOCK(ifp);
1374 	CK_STAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifaddr, ifa_link);
1375 	IF_ADDR_WUNLOCK(ifp);
1376 	ifa_free(&ia->ia_ifa);			/* if_addrhead */
1377 
1378 	/*
1379 	 * Defer the release of what might be the last reference to the
1380 	 * in6_ifaddr so that it can't be freed before the remainder of the
1381 	 * cleanup.
1382 	 */
1383 	IN6_IFADDR_WLOCK();
1384 	CK_STAILQ_REMOVE(&V_in6_ifaddrhead, ia, in6_ifaddr, ia_link);
1385 	CK_LIST_REMOVE(ia, ia6_hash);
1386 	IN6_IFADDR_WUNLOCK();
1387 
1388 	/*
1389 	 * Release the reference to the base prefix.  There should be a
1390 	 * positive reference.
1391 	 */
1392 	remove_lle = 0;
1393 	if (ia->ia6_ndpr == NULL) {
1394 		nd6log((LOG_NOTICE,
1395 		    "in6_unlink_ifa: autoconf'ed address "
1396 		    "%s has no prefix\n", ip6_sprintf(ip6buf, IA6_IN6(ia))));
1397 	} else {
1398 		ia->ia6_ndpr->ndpr_addrcnt--;
1399 		/* Do not delete lles within prefix if refcont != 0 */
1400 		if (ia->ia6_ndpr->ndpr_addrcnt == 0)
1401 			remove_lle = 1;
1402 		ia->ia6_ndpr = NULL;
1403 	}
1404 
1405 	nd6_rem_ifa_lle(ia, remove_lle);
1406 
1407 	/*
1408 	 * Also, if the address being removed is autoconf'ed, call
1409 	 * pfxlist_onlink_check() since the release might affect the status of
1410 	 * other (detached) addresses.
1411 	 */
1412 	if ((ia->ia6_flags & IN6_IFF_AUTOCONF)) {
1413 		pfxlist_onlink_check();
1414 	}
1415 	ifa_free(&ia->ia_ifa);			/* in6_ifaddrhead */
1416 }
1417 
1418 /*
1419  * Notifies other subsystems about address change/arrival:
1420  * 1) Notifies device handler on the first IPv6 address assignment
1421  * 2) Handle routing table changes for P2P links and route
1422  * 3) Handle routing table changes for address host route
1423  */
1424 static int
1425 in6_notify_ifa(struct ifnet *ifp, struct in6_ifaddr *ia,
1426     struct in6_aliasreq *ifra, int hostIsNew)
1427 {
1428 	int	error = 0, plen, ifacount = 0;
1429 	struct ifaddr *ifa;
1430 	struct sockaddr_in6 *pdst;
1431 	char ip6buf[INET6_ADDRSTRLEN];
1432 
1433 	/*
1434 	 * Give the interface a chance to initialize
1435 	 * if this is its first address,
1436 	 */
1437 	if (hostIsNew != 0) {
1438 		struct epoch_tracker et;
1439 
1440 		NET_EPOCH_ENTER(et);
1441 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1442 			if (ifa->ifa_addr->sa_family != AF_INET6)
1443 				continue;
1444 			ifacount++;
1445 		}
1446 		NET_EPOCH_EXIT(et);
1447 	}
1448 
1449 	if (ifacount <= 1 && ifp->if_ioctl) {
1450 		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
1451 		if (error)
1452 			goto done;
1453 	}
1454 
1455 	/*
1456 	 * If a new destination address is specified, scrub the old one and
1457 	 * install the new destination.  Note that the interface must be
1458 	 * p2p or loopback.
1459 	 */
1460 	pdst = &ifra->ifra_dstaddr;
1461 	if (pdst->sin6_family == AF_INET6 &&
1462 	    !IN6_ARE_ADDR_EQUAL(&pdst->sin6_addr, &ia->ia_dstaddr.sin6_addr)) {
1463 		if ((ia->ia_flags & IFA_ROUTE) != 0 &&
1464 		    (in6_handle_dstaddr_rtrequest(RTM_DELETE, ia) != 0)) {
1465 			nd6log((LOG_ERR, "in6_update_ifa_internal: failed to "
1466 			    "remove a route to the old destination: %s\n",
1467 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1468 			/* proceed anyway... */
1469 		} else
1470 			ia->ia_flags &= ~IFA_ROUTE;
1471 		ia->ia_dstaddr = *pdst;
1472 	}
1473 
1474 	/*
1475 	 * If a new destination address is specified for a point-to-point
1476 	 * interface, install a route to the destination as an interface
1477 	 * direct route.
1478 	 * XXX: the logic below rejects assigning multiple addresses on a p2p
1479 	 * interface that share the same destination.
1480 	 */
1481 	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1482 	if (!(ia->ia_flags & IFA_ROUTE) && plen == 128 &&
1483 	    ia->ia_dstaddr.sin6_family == AF_INET6) {
1484 		/*
1485 		 * Handle the case for ::1 .
1486 		 */
1487 		if (ifp->if_flags & IFF_LOOPBACK)
1488 			ia->ia_flags |= IFA_RTSELF;
1489 		error = in6_handle_dstaddr_rtrequest(RTM_ADD, ia);
1490 		if (error)
1491 			goto done;
1492 		ia->ia_flags |= IFA_ROUTE;
1493 	}
1494 
1495 	/*
1496 	 * add a loopback route to self if not exists
1497 	 */
1498 	if (!(ia->ia_flags & IFA_RTSELF) && V_nd6_useloopback) {
1499 		error = ifa_add_loopback_route((struct ifaddr *)ia,
1500 		    (struct sockaddr *)&ia->ia_addr);
1501 		if (error == 0)
1502 			ia->ia_flags |= IFA_RTSELF;
1503 	}
1504 done:
1505 	WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
1506 	    "Invoking IPv6 network device address event may sleep");
1507 
1508 	ifa_ref(&ia->ia_ifa);
1509 	EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, &ia->ia_ifa,
1510 	    IFADDR_EVENT_ADD);
1511 	ifa_free(&ia->ia_ifa);
1512 
1513 	return (error);
1514 }
1515 
1516 /*
1517  * Find an IPv6 interface link-local address specific to an interface.
1518  * ifaddr is returned referenced.
1519  */
1520 struct in6_ifaddr *
1521 in6ifa_ifpforlinklocal(struct ifnet *ifp, int ignoreflags)
1522 {
1523 	struct ifaddr *ifa;
1524 
1525 	NET_EPOCH_ASSERT();
1526 
1527 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1528 		if (ifa->ifa_addr->sa_family != AF_INET6)
1529 			continue;
1530 		if (IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa))) {
1531 			if ((((struct in6_ifaddr *)ifa)->ia6_flags &
1532 			    ignoreflags) != 0)
1533 				continue;
1534 			ifa_ref(ifa);
1535 			break;
1536 		}
1537 	}
1538 
1539 	return ((struct in6_ifaddr *)ifa);
1540 }
1541 
1542 /*
1543  * find the interface address corresponding to a given IPv6 address.
1544  * ifaddr is returned referenced.
1545  */
1546 struct in6_ifaddr *
1547 in6ifa_ifwithaddr(const struct in6_addr *addr, uint32_t zoneid)
1548 {
1549 	struct rm_priotracker in6_ifa_tracker;
1550 	struct in6_ifaddr *ia;
1551 
1552 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1553 	CK_LIST_FOREACH(ia, IN6ADDR_HASH(addr), ia6_hash) {
1554 		if (IN6_ARE_ADDR_EQUAL(IA6_IN6(ia), addr)) {
1555 			if (zoneid != 0 &&
1556 			    zoneid != ia->ia_addr.sin6_scope_id)
1557 				continue;
1558 			ifa_ref(&ia->ia_ifa);
1559 			break;
1560 		}
1561 	}
1562 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1563 	return (ia);
1564 }
1565 
1566 /*
1567  * find the internet address corresponding to a given interface and address.
1568  * ifaddr is returned referenced.
1569  */
1570 struct in6_ifaddr *
1571 in6ifa_ifpwithaddr(struct ifnet *ifp, const struct in6_addr *addr)
1572 {
1573 	struct epoch_tracker et;
1574 	struct ifaddr *ifa;
1575 
1576 	NET_EPOCH_ENTER(et);
1577 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1578 		if (ifa->ifa_addr->sa_family != AF_INET6)
1579 			continue;
1580 		if (IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa))) {
1581 			ifa_ref(ifa);
1582 			break;
1583 		}
1584 	}
1585 	NET_EPOCH_EXIT(et);
1586 
1587 	return ((struct in6_ifaddr *)ifa);
1588 }
1589 
1590 /*
1591  * Find a link-local scoped address on ifp and return it if any.
1592  */
1593 struct in6_ifaddr *
1594 in6ifa_llaonifp(struct ifnet *ifp)
1595 {
1596 	struct epoch_tracker et;
1597 	struct sockaddr_in6 *sin6;
1598 	struct ifaddr *ifa;
1599 
1600 	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)
1601 		return (NULL);
1602 	NET_EPOCH_ENTER(et);
1603 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1604 		if (ifa->ifa_addr->sa_family != AF_INET6)
1605 			continue;
1606 		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
1607 		if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr) ||
1608 		    IN6_IS_ADDR_MC_INTFACELOCAL(&sin6->sin6_addr) ||
1609 		    IN6_IS_ADDR_MC_NODELOCAL(&sin6->sin6_addr))
1610 			break;
1611 	}
1612 	NET_EPOCH_EXIT(et);
1613 
1614 	return ((struct in6_ifaddr *)ifa);
1615 }
1616 
1617 /*
1618  * Convert IP6 address to printable (loggable) representation. Caller
1619  * has to make sure that ip6buf is at least INET6_ADDRSTRLEN long.
1620  */
1621 static char digits[] = "0123456789abcdef";
1622 char *
1623 ip6_sprintf(char *ip6buf, const struct in6_addr *addr)
1624 {
1625 	int i, cnt = 0, maxcnt = 0, idx = 0, index = 0;
1626 	char *cp;
1627 	const u_int16_t *a = (const u_int16_t *)addr;
1628 	const u_int8_t *d;
1629 	int dcolon = 0, zero = 0;
1630 
1631 	cp = ip6buf;
1632 
1633 	for (i = 0; i < 8; i++) {
1634 		if (*(a + i) == 0) {
1635 			cnt++;
1636 			if (cnt == 1)
1637 				idx = i;
1638 		}
1639 		else if (maxcnt < cnt) {
1640 			maxcnt = cnt;
1641 			index = idx;
1642 			cnt = 0;
1643 		}
1644 	}
1645 	if (maxcnt < cnt) {
1646 		maxcnt = cnt;
1647 		index = idx;
1648 	}
1649 
1650 	for (i = 0; i < 8; i++) {
1651 		if (dcolon == 1) {
1652 			if (*a == 0) {
1653 				if (i == 7)
1654 					*cp++ = ':';
1655 				a++;
1656 				continue;
1657 			} else
1658 				dcolon = 2;
1659 		}
1660 		if (*a == 0) {
1661 			if (dcolon == 0 && *(a + 1) == 0 && i == index) {
1662 				if (i == 0)
1663 					*cp++ = ':';
1664 				*cp++ = ':';
1665 				dcolon = 1;
1666 			} else {
1667 				*cp++ = '0';
1668 				*cp++ = ':';
1669 			}
1670 			a++;
1671 			continue;
1672 		}
1673 		d = (const u_char *)a;
1674 		/* Try to eliminate leading zeros in printout like in :0001. */
1675 		zero = 1;
1676 		*cp = digits[*d >> 4];
1677 		if (*cp != '0') {
1678 			zero = 0;
1679 			cp++;
1680 		}
1681 		*cp = digits[*d++ & 0xf];
1682 		if (zero == 0 || (*cp != '0')) {
1683 			zero = 0;
1684 			cp++;
1685 		}
1686 		*cp = digits[*d >> 4];
1687 		if (zero == 0 || (*cp != '0')) {
1688 			zero = 0;
1689 			cp++;
1690 		}
1691 		*cp++ = digits[*d & 0xf];
1692 		*cp++ = ':';
1693 		a++;
1694 	}
1695 	*--cp = '\0';
1696 	return (ip6buf);
1697 }
1698 
1699 int
1700 in6_localaddr(struct in6_addr *in6)
1701 {
1702 	struct rm_priotracker in6_ifa_tracker;
1703 	struct in6_ifaddr *ia;
1704 
1705 	if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
1706 		return 1;
1707 
1708 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1709 	CK_STAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
1710 		if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
1711 		    &ia->ia_prefixmask.sin6_addr)) {
1712 			IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1713 			return 1;
1714 		}
1715 	}
1716 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1717 
1718 	return (0);
1719 }
1720 
1721 /*
1722  * Return 1 if an internet address is for the local host and configured
1723  * on one of its interfaces.
1724  */
1725 int
1726 in6_localip(struct in6_addr *in6)
1727 {
1728 	struct rm_priotracker in6_ifa_tracker;
1729 	struct in6_ifaddr *ia;
1730 
1731 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1732 	CK_LIST_FOREACH(ia, IN6ADDR_HASH(in6), ia6_hash) {
1733 		if (IN6_ARE_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr)) {
1734 			IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1735 			return (1);
1736 		}
1737 	}
1738 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1739 	return (0);
1740 }
1741 
1742 /*
1743  * Return 1 if an internet address is configured on an interface.
1744  */
1745 int
1746 in6_ifhasaddr(struct ifnet *ifp, struct in6_addr *addr)
1747 {
1748 	struct in6_addr in6;
1749 	struct ifaddr *ifa;
1750 	struct in6_ifaddr *ia6;
1751 
1752 	NET_EPOCH_ASSERT();
1753 
1754 	in6 = *addr;
1755 	if (in6_clearscope(&in6))
1756 		return (0);
1757 	in6_setscope(&in6, ifp, NULL);
1758 
1759 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1760 		if (ifa->ifa_addr->sa_family != AF_INET6)
1761 			continue;
1762 		ia6 = (struct in6_ifaddr *)ifa;
1763 		if (IN6_ARE_ADDR_EQUAL(&ia6->ia_addr.sin6_addr, &in6))
1764 			return (1);
1765 	}
1766 
1767 	return (0);
1768 }
1769 
1770 int
1771 in6_is_addr_deprecated(struct sockaddr_in6 *sa6)
1772 {
1773 	struct rm_priotracker in6_ifa_tracker;
1774 	struct in6_ifaddr *ia;
1775 
1776 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1777 	CK_LIST_FOREACH(ia, IN6ADDR_HASH(&sa6->sin6_addr), ia6_hash) {
1778 		if (IN6_ARE_ADDR_EQUAL(IA6_IN6(ia), &sa6->sin6_addr)) {
1779 			if (ia->ia6_flags & IN6_IFF_DEPRECATED) {
1780 				IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1781 				return (1); /* true */
1782 			}
1783 			break;
1784 		}
1785 	}
1786 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1787 
1788 	return (0);		/* false */
1789 }
1790 
1791 /*
1792  * return length of part which dst and src are equal
1793  * hard coding...
1794  */
1795 int
1796 in6_matchlen(struct in6_addr *src, struct in6_addr *dst)
1797 {
1798 	int match = 0;
1799 	u_char *s = (u_char *)src, *d = (u_char *)dst;
1800 	u_char *lim = s + 16, r;
1801 
1802 	while (s < lim)
1803 		if ((r = (*d++ ^ *s++)) != 0) {
1804 			while (r < 128) {
1805 				match++;
1806 				r <<= 1;
1807 			}
1808 			break;
1809 		} else
1810 			match += 8;
1811 	return match;
1812 }
1813 
1814 /* XXX: to be scope conscious */
1815 int
1816 in6_are_prefix_equal(struct in6_addr *p1, struct in6_addr *p2, int len)
1817 {
1818 	int bytelen, bitlen;
1819 
1820 	/* sanity check */
1821 	if (0 > len || len > 128) {
1822 		log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n",
1823 		    len);
1824 		return (0);
1825 	}
1826 
1827 	bytelen = len / 8;
1828 	bitlen = len % 8;
1829 
1830 	if (bcmp(&p1->s6_addr, &p2->s6_addr, bytelen))
1831 		return (0);
1832 	if (bitlen != 0 &&
1833 	    p1->s6_addr[bytelen] >> (8 - bitlen) !=
1834 	    p2->s6_addr[bytelen] >> (8 - bitlen))
1835 		return (0);
1836 
1837 	return (1);
1838 }
1839 
1840 void
1841 in6_prefixlen2mask(struct in6_addr *maskp, int len)
1842 {
1843 	u_char maskarray[8] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
1844 	int bytelen, bitlen, i;
1845 
1846 	/* sanity check */
1847 	if (0 > len || len > 128) {
1848 		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
1849 		    len);
1850 		return;
1851 	}
1852 
1853 	bzero(maskp, sizeof(*maskp));
1854 	bytelen = len / 8;
1855 	bitlen = len % 8;
1856 	for (i = 0; i < bytelen; i++)
1857 		maskp->s6_addr[i] = 0xff;
1858 	if (bitlen)
1859 		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
1860 }
1861 
1862 /*
1863  * return the best address out of the same scope. if no address was
1864  * found, return the first valid address from designated IF.
1865  */
1866 struct in6_ifaddr *
1867 in6_ifawithifp(struct ifnet *ifp, struct in6_addr *dst)
1868 {
1869 	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
1870 	struct ifaddr *ifa;
1871 	struct in6_ifaddr *besta = NULL;
1872 	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
1873 
1874 	NET_EPOCH_ASSERT();
1875 
1876 	dep[0] = dep[1] = NULL;
1877 
1878 	/*
1879 	 * We first look for addresses in the same scope.
1880 	 * If there is one, return it.
1881 	 * If two or more, return one which matches the dst longest.
1882 	 * If none, return one of global addresses assigned other ifs.
1883 	 */
1884 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1885 		if (ifa->ifa_addr->sa_family != AF_INET6)
1886 			continue;
1887 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
1888 			continue; /* XXX: is there any case to allow anycast? */
1889 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
1890 			continue; /* don't use this interface */
1891 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
1892 			continue;
1893 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
1894 			if (V_ip6_use_deprecated)
1895 				dep[0] = (struct in6_ifaddr *)ifa;
1896 			continue;
1897 		}
1898 
1899 		if (dst_scope == in6_addrscope(IFA_IN6(ifa))) {
1900 			/*
1901 			 * call in6_matchlen() as few as possible
1902 			 */
1903 			if (besta) {
1904 				if (blen == -1)
1905 					blen = in6_matchlen(&besta->ia_addr.sin6_addr, dst);
1906 				tlen = in6_matchlen(IFA_IN6(ifa), dst);
1907 				if (tlen > blen) {
1908 					blen = tlen;
1909 					besta = (struct in6_ifaddr *)ifa;
1910 				}
1911 			} else
1912 				besta = (struct in6_ifaddr *)ifa;
1913 		}
1914 	}
1915 	if (besta) {
1916 		ifa_ref(&besta->ia_ifa);
1917 		return (besta);
1918 	}
1919 
1920 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1921 		if (ifa->ifa_addr->sa_family != AF_INET6)
1922 			continue;
1923 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
1924 			continue; /* XXX: is there any case to allow anycast? */
1925 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
1926 			continue; /* don't use this interface */
1927 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
1928 			continue;
1929 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
1930 			if (V_ip6_use_deprecated)
1931 				dep[1] = (struct in6_ifaddr *)ifa;
1932 			continue;
1933 		}
1934 
1935 		if (ifa != NULL)
1936 			ifa_ref(ifa);
1937 		return (struct in6_ifaddr *)ifa;
1938 	}
1939 
1940 	/* use the last-resort values, that are, deprecated addresses */
1941 	if (dep[0]) {
1942 		ifa_ref((struct ifaddr *)dep[0]);
1943 		return dep[0];
1944 	}
1945 	if (dep[1]) {
1946 		ifa_ref((struct ifaddr *)dep[1]);
1947 		return dep[1];
1948 	}
1949 
1950 	return NULL;
1951 }
1952 
1953 /*
1954  * perform DAD when interface becomes IFF_UP.
1955  */
1956 void
1957 in6_if_up(struct ifnet *ifp)
1958 {
1959 	struct epoch_tracker et;
1960 	struct ifaddr *ifa;
1961 	struct in6_ifaddr *ia;
1962 
1963 	NET_EPOCH_ENTER(et);
1964 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1965 		if (ifa->ifa_addr->sa_family != AF_INET6)
1966 			continue;
1967 		ia = (struct in6_ifaddr *)ifa;
1968 		if (ia->ia6_flags & IN6_IFF_TENTATIVE) {
1969 			/*
1970 			 * The TENTATIVE flag was likely set by hand
1971 			 * beforehand, implicitly indicating the need for DAD.
1972 			 * We may be able to skip the random delay in this
1973 			 * case, but we impose delays just in case.
1974 			 */
1975 			nd6_dad_start(ifa,
1976 			    arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz));
1977 		}
1978 	}
1979 	NET_EPOCH_EXIT(et);
1980 
1981 	/*
1982 	 * special cases, like 6to4, are handled in in6_ifattach
1983 	 */
1984 	in6_ifattach(ifp, NULL);
1985 }
1986 
1987 int
1988 in6if_do_dad(struct ifnet *ifp)
1989 {
1990 
1991 	if ((ifp->if_flags & IFF_LOOPBACK) != 0)
1992 		return (0);
1993 	if ((ifp->if_flags & IFF_MULTICAST) == 0)
1994 		return (0);
1995 	if ((ND_IFINFO(ifp)->flags &
1996 	    (ND6_IFF_IFDISABLED | ND6_IFF_NO_DAD)) != 0)
1997 		return (0);
1998 	return (1);
1999 }
2000 
2001 /*
2002  * Calculate max IPv6 MTU through all the interfaces and store it
2003  * to in6_maxmtu.
2004  */
2005 void
2006 in6_setmaxmtu(void)
2007 {
2008 	struct epoch_tracker et;
2009 	unsigned long maxmtu = 0;
2010 	struct ifnet *ifp;
2011 
2012 	NET_EPOCH_ENTER(et);
2013 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
2014 		/* this function can be called during ifnet initialization */
2015 		if (!ifp->if_afdata[AF_INET6])
2016 			continue;
2017 		if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
2018 		    IN6_LINKMTU(ifp) > maxmtu)
2019 			maxmtu = IN6_LINKMTU(ifp);
2020 	}
2021 	NET_EPOCH_EXIT(et);
2022 	if (maxmtu)	/* update only when maxmtu is positive */
2023 		V_in6_maxmtu = maxmtu;
2024 }
2025 
2026 /*
2027  * Provide the length of interface identifiers to be used for the link attached
2028  * to the given interface.  The length should be defined in "IPv6 over
2029  * xxx-link" document.  Note that address architecture might also define
2030  * the length for a particular set of address prefixes, regardless of the
2031  * link type.  As clarified in rfc2462bis, those two definitions should be
2032  * consistent, and those really are as of August 2004.
2033  */
2034 int
2035 in6_if2idlen(struct ifnet *ifp)
2036 {
2037 	switch (ifp->if_type) {
2038 	case IFT_ETHER:		/* RFC2464 */
2039 	case IFT_PROPVIRTUAL:	/* XXX: no RFC. treat it as ether */
2040 	case IFT_L2VLAN:	/* ditto */
2041 	case IFT_BRIDGE:	/* bridge(4) only does Ethernet-like links */
2042 	case IFT_INFINIBAND:
2043 		return (64);
2044 	case IFT_PPP:		/* RFC2472 */
2045 		return (64);
2046 	case IFT_FRELAY:	/* RFC2590 */
2047 		return (64);
2048 	case IFT_IEEE1394:	/* RFC3146 */
2049 		return (64);
2050 	case IFT_GIF:
2051 		return (64);	/* draft-ietf-v6ops-mech-v2-07 */
2052 	case IFT_LOOP:
2053 		return (64);	/* XXX: is this really correct? */
2054 	default:
2055 		/*
2056 		 * Unknown link type:
2057 		 * It might be controversial to use the today's common constant
2058 		 * of 64 for these cases unconditionally.  For full compliance,
2059 		 * we should return an error in this case.  On the other hand,
2060 		 * if we simply miss the standard for the link type or a new
2061 		 * standard is defined for a new link type, the IFID length
2062 		 * is very likely to be the common constant.  As a compromise,
2063 		 * we always use the constant, but make an explicit notice
2064 		 * indicating the "unknown" case.
2065 		 */
2066 		printf("in6_if2idlen: unknown link type (%d)\n", ifp->if_type);
2067 		return (64);
2068 	}
2069 }
2070 
2071 struct in6_llentry {
2072 	struct llentry		base;
2073 };
2074 
2075 #define	IN6_LLTBL_DEFAULT_HSIZE	32
2076 #define	IN6_LLTBL_HASH(k, h) \
2077 	(((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1))
2078 
2079 /*
2080  * Do actual deallocation of @lle.
2081  */
2082 static void
2083 in6_lltable_destroy_lle_unlocked(epoch_context_t ctx)
2084 {
2085 	struct llentry *lle;
2086 
2087 	lle = __containerof(ctx, struct llentry, lle_epoch_ctx);
2088 	LLE_LOCK_DESTROY(lle);
2089 	LLE_REQ_DESTROY(lle);
2090 	free(lle, M_LLTABLE);
2091 }
2092 
2093 /*
2094  * Called by LLE_FREE_LOCKED when number of references
2095  * drops to zero.
2096  */
2097 static void
2098 in6_lltable_destroy_lle(struct llentry *lle)
2099 {
2100 
2101 	LLE_WUNLOCK(lle);
2102 	NET_EPOCH_CALL(in6_lltable_destroy_lle_unlocked, &lle->lle_epoch_ctx);
2103 }
2104 
2105 static struct llentry *
2106 in6_lltable_new(const struct in6_addr *addr6, u_int flags)
2107 {
2108 	struct in6_llentry *lle;
2109 
2110 	lle = malloc(sizeof(struct in6_llentry), M_LLTABLE, M_NOWAIT | M_ZERO);
2111 	if (lle == NULL)		/* NB: caller generates msg */
2112 		return NULL;
2113 
2114 	lle->base.r_l3addr.addr6 = *addr6;
2115 	lle->base.lle_refcnt = 1;
2116 	lle->base.lle_free = in6_lltable_destroy_lle;
2117 	LLE_LOCK_INIT(&lle->base);
2118 	LLE_REQ_INIT(&lle->base);
2119 	callout_init(&lle->base.lle_timer, 1);
2120 
2121 	return (&lle->base);
2122 }
2123 
2124 static int
2125 in6_lltable_match_prefix(const struct sockaddr *saddr,
2126     const struct sockaddr *smask, u_int flags, struct llentry *lle)
2127 {
2128 	const struct in6_addr *addr, *mask, *lle_addr;
2129 
2130 	addr = &((const struct sockaddr_in6 *)saddr)->sin6_addr;
2131 	mask = &((const struct sockaddr_in6 *)smask)->sin6_addr;
2132 	lle_addr = &lle->r_l3addr.addr6;
2133 
2134 	if (IN6_ARE_MASKED_ADDR_EQUAL(lle_addr, addr, mask) == 0)
2135 		return (0);
2136 
2137 	if (lle->la_flags & LLE_IFADDR) {
2138 		/*
2139 		 * Delete LLE_IFADDR records IFF address & flag matches.
2140 		 * Note that addr is the interface address within prefix
2141 		 * being matched.
2142 		 */
2143 		if (IN6_ARE_ADDR_EQUAL(addr, lle_addr) &&
2144 		    (flags & LLE_STATIC) != 0)
2145 			return (1);
2146 		return (0);
2147 	}
2148 
2149 	/* flags & LLE_STATIC means deleting both dynamic and static entries */
2150 	if ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC))
2151 		return (1);
2152 
2153 	return (0);
2154 }
2155 
2156 static void
2157 in6_lltable_free_entry(struct lltable *llt, struct llentry *lle)
2158 {
2159 	struct ifnet *ifp;
2160 
2161 	LLE_WLOCK_ASSERT(lle);
2162 	KASSERT(llt != NULL, ("lltable is NULL"));
2163 
2164 	/* Unlink entry from table */
2165 	if ((lle->la_flags & LLE_LINKED) != 0) {
2166 		ifp = llt->llt_ifp;
2167 		IF_AFDATA_WLOCK_ASSERT(ifp);
2168 		lltable_unlink_entry(llt, lle);
2169 	}
2170 
2171 	llentry_free(lle);
2172 }
2173 
2174 static int
2175 in6_lltable_rtcheck(struct ifnet *ifp,
2176 		    u_int flags,
2177 		    const struct sockaddr *l3addr)
2178 {
2179 	const struct sockaddr_in6 *sin6;
2180 	struct nhop_object *nh;
2181 	struct in6_addr dst;
2182 	uint32_t scopeid;
2183 	char ip6buf[INET6_ADDRSTRLEN];
2184 	int fibnum;
2185 
2186 	NET_EPOCH_ASSERT();
2187 	KASSERT(l3addr->sa_family == AF_INET6,
2188 	    ("sin_family %d", l3addr->sa_family));
2189 
2190 	sin6 = (const struct sockaddr_in6 *)l3addr;
2191 	in6_splitscope(&sin6->sin6_addr, &dst, &scopeid);
2192 	fibnum = V_rt_add_addr_allfibs ? RT_DEFAULT_FIB : ifp->if_fib;
2193 	nh = fib6_lookup(fibnum, &dst, scopeid, NHR_NONE, 0);
2194 	if (nh && ((nh->nh_flags & NHF_GATEWAY) || nh->nh_ifp != ifp)) {
2195 		struct ifaddr *ifa;
2196 		/*
2197 		 * Create an ND6 cache for an IPv6 neighbor
2198 		 * that is not covered by our own prefix.
2199 		 */
2200 		ifa = ifaof_ifpforaddr(l3addr, ifp);
2201 		if (ifa != NULL) {
2202 			return 0;
2203 		}
2204 		log(LOG_INFO, "IPv6 address: \"%s\" is not on the network\n",
2205 		    ip6_sprintf(ip6buf, &sin6->sin6_addr));
2206 		return EINVAL;
2207 	}
2208 	return 0;
2209 }
2210 
2211 /*
2212  * Called by the datapath to indicate that the entry was used.
2213  */
2214 static void
2215 in6_lltable_mark_used(struct llentry *lle)
2216 {
2217 
2218 	LLE_REQ_LOCK(lle);
2219 	lle->r_skip_req = 0;
2220 
2221 	/*
2222 	 * Set the hit time so the callback function
2223 	 * can determine the remaining time before
2224 	 * transiting to the DELAY state.
2225 	 */
2226 	lle->lle_hittime = time_uptime;
2227 	LLE_REQ_UNLOCK(lle);
2228 }
2229 
2230 static inline uint32_t
2231 in6_lltable_hash_dst(const struct in6_addr *dst, uint32_t hsize)
2232 {
2233 
2234 	return (IN6_LLTBL_HASH(dst->s6_addr32[3], hsize));
2235 }
2236 
2237 static uint32_t
2238 in6_lltable_hash(const struct llentry *lle, uint32_t hsize)
2239 {
2240 
2241 	return (in6_lltable_hash_dst(&lle->r_l3addr.addr6, hsize));
2242 }
2243 
2244 static void
2245 in6_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
2246 {
2247 	struct sockaddr_in6 *sin6;
2248 
2249 	sin6 = (struct sockaddr_in6 *)sa;
2250 	bzero(sin6, sizeof(*sin6));
2251 	sin6->sin6_family = AF_INET6;
2252 	sin6->sin6_len = sizeof(*sin6);
2253 	sin6->sin6_addr = lle->r_l3addr.addr6;
2254 }
2255 
2256 static inline struct llentry *
2257 in6_lltable_find_dst(struct lltable *llt, const struct in6_addr *dst)
2258 {
2259 	struct llentry *lle;
2260 	struct llentries *lleh;
2261 	u_int hashidx;
2262 
2263 	hashidx = in6_lltable_hash_dst(dst, llt->llt_hsize);
2264 	lleh = &llt->lle_head[hashidx];
2265 	CK_LIST_FOREACH(lle, lleh, lle_next) {
2266 		if (lle->la_flags & LLE_DELETED)
2267 			continue;
2268 		if (IN6_ARE_ADDR_EQUAL(&lle->r_l3addr.addr6, dst))
2269 			break;
2270 	}
2271 
2272 	return (lle);
2273 }
2274 
2275 static void
2276 in6_lltable_delete_entry(struct lltable *llt, struct llentry *lle)
2277 {
2278 
2279 	lle->la_flags |= LLE_DELETED;
2280 	EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_DELETED);
2281 #ifdef DIAGNOSTIC
2282 	log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle);
2283 #endif
2284 	llentry_free(lle);
2285 }
2286 
2287 static struct llentry *
2288 in6_lltable_alloc(struct lltable *llt, u_int flags,
2289 	const struct sockaddr *l3addr)
2290 {
2291 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2292 	struct ifnet *ifp = llt->llt_ifp;
2293 	struct llentry *lle;
2294 	char linkhdr[LLE_MAX_LINKHDR];
2295 	size_t linkhdrsize;
2296 	int lladdr_off;
2297 
2298 	KASSERT(l3addr->sa_family == AF_INET6,
2299 	    ("sin_family %d", l3addr->sa_family));
2300 
2301 	/*
2302 	 * A route that covers the given address must have
2303 	 * been installed 1st because we are doing a resolution,
2304 	 * verify this.
2305 	 */
2306 	if (!(flags & LLE_IFADDR) &&
2307 	    in6_lltable_rtcheck(ifp, flags, l3addr) != 0)
2308 		return (NULL);
2309 
2310 	lle = in6_lltable_new(&sin6->sin6_addr, flags);
2311 	if (lle == NULL) {
2312 		log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
2313 		return (NULL);
2314 	}
2315 	lle->la_flags = flags;
2316 	if ((flags & LLE_IFADDR) == LLE_IFADDR) {
2317 		linkhdrsize = LLE_MAX_LINKHDR;
2318 		if (lltable_calc_llheader(ifp, AF_INET6, IF_LLADDR(ifp),
2319 		    linkhdr, &linkhdrsize, &lladdr_off) != 0) {
2320 			NET_EPOCH_CALL(in6_lltable_destroy_lle_unlocked, &lle->lle_epoch_ctx);
2321 			return (NULL);
2322 		}
2323 		lltable_set_entry_addr(ifp, lle, linkhdr, linkhdrsize,
2324 		    lladdr_off);
2325 		lle->la_flags |= LLE_STATIC;
2326 	}
2327 
2328 	if ((lle->la_flags & LLE_STATIC) != 0)
2329 		lle->ln_state = ND6_LLINFO_REACHABLE;
2330 
2331 	return (lle);
2332 }
2333 
2334 static struct llentry *
2335 in6_lltable_lookup(struct lltable *llt, u_int flags,
2336 	const struct sockaddr *l3addr)
2337 {
2338 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2339 	struct llentry *lle;
2340 
2341 	IF_AFDATA_LOCK_ASSERT(llt->llt_ifp);
2342 	KASSERT(l3addr->sa_family == AF_INET6,
2343 	    ("sin_family %d", l3addr->sa_family));
2344 	KASSERT((flags & (LLE_UNLOCKED | LLE_EXCLUSIVE)) !=
2345 	    (LLE_UNLOCKED | LLE_EXCLUSIVE),
2346 	    ("wrong lle request flags: %#x", flags));
2347 
2348 	lle = in6_lltable_find_dst(llt, &sin6->sin6_addr);
2349 	if (lle == NULL)
2350 		return (NULL);
2351 	if (flags & LLE_UNLOCKED)
2352 		return (lle);
2353 
2354 	if (flags & LLE_EXCLUSIVE)
2355 		LLE_WLOCK(lle);
2356 	else
2357 		LLE_RLOCK(lle);
2358 
2359 	/*
2360 	 * If the afdata lock is not held, the LLE may have been unlinked while
2361 	 * we were blocked on the LLE lock.  Check for this case.
2362 	 */
2363 	if (__predict_false((lle->la_flags & LLE_LINKED) == 0)) {
2364 		if (flags & LLE_EXCLUSIVE)
2365 			LLE_WUNLOCK(lle);
2366 		else
2367 			LLE_RUNLOCK(lle);
2368 		return (NULL);
2369 	}
2370 	return (lle);
2371 }
2372 
2373 static int
2374 in6_lltable_dump_entry(struct lltable *llt, struct llentry *lle,
2375     struct sysctl_req *wr)
2376 {
2377 	struct ifnet *ifp = llt->llt_ifp;
2378 	/* XXX stack use */
2379 	struct {
2380 		struct rt_msghdr	rtm;
2381 		struct sockaddr_in6	sin6;
2382 		/*
2383 		 * ndp.c assumes that sdl is word aligned
2384 		 */
2385 #ifdef __LP64__
2386 		uint32_t		pad;
2387 #endif
2388 		struct sockaddr_dl	sdl;
2389 	} ndpc;
2390 	struct sockaddr_dl *sdl;
2391 	int error;
2392 
2393 	bzero(&ndpc, sizeof(ndpc));
2394 	/* skip deleted entries */
2395 	if ((lle->la_flags & LLE_DELETED) == LLE_DELETED)
2396 		return (0);
2397 	/* Skip if jailed and not a valid IP of the prison. */
2398 	lltable_fill_sa_entry(lle, (struct sockaddr *)&ndpc.sin6);
2399 	if (prison_if(wr->td->td_ucred, (struct sockaddr *)&ndpc.sin6) != 0)
2400 		return (0);
2401 	/*
2402 	 * produce a msg made of:
2403 	 *  struct rt_msghdr;
2404 	 *  struct sockaddr_in6 (IPv6)
2405 	 *  struct sockaddr_dl;
2406 	 */
2407 	ndpc.rtm.rtm_msglen = sizeof(ndpc);
2408 	ndpc.rtm.rtm_version = RTM_VERSION;
2409 	ndpc.rtm.rtm_type = RTM_GET;
2410 	ndpc.rtm.rtm_flags = RTF_UP;
2411 	ndpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
2412 	sa6_recoverscope(&ndpc.sin6);
2413 
2414 	/* publish */
2415 	if (lle->la_flags & LLE_PUB)
2416 		ndpc.rtm.rtm_flags |= RTF_ANNOUNCE;
2417 
2418 	sdl = &ndpc.sdl;
2419 	sdl->sdl_family = AF_LINK;
2420 	sdl->sdl_len = sizeof(*sdl);
2421 	sdl->sdl_index = ifp->if_index;
2422 	sdl->sdl_type = ifp->if_type;
2423 	if ((lle->la_flags & LLE_VALID) == LLE_VALID) {
2424 		sdl->sdl_alen = ifp->if_addrlen;
2425 		bcopy(lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
2426 	} else {
2427 		sdl->sdl_alen = 0;
2428 		bzero(LLADDR(sdl), ifp->if_addrlen);
2429 	}
2430 	if (lle->la_expire != 0)
2431 		ndpc.rtm.rtm_rmx.rmx_expire = lle->la_expire +
2432 		    lle->lle_remtime / hz + time_second - time_uptime;
2433 	ndpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
2434 	if (lle->la_flags & LLE_STATIC)
2435 		ndpc.rtm.rtm_flags |= RTF_STATIC;
2436 	if (lle->la_flags & LLE_IFADDR)
2437 		ndpc.rtm.rtm_flags |= RTF_PINNED;
2438 	if (lle->ln_router != 0)
2439 		ndpc.rtm.rtm_flags |= RTF_GATEWAY;
2440 	ndpc.rtm.rtm_rmx.rmx_pksent = lle->la_asked;
2441 	/* Store state in rmx_weight value */
2442 	ndpc.rtm.rtm_rmx.rmx_state = lle->ln_state;
2443 	ndpc.rtm.rtm_index = ifp->if_index;
2444 	error = SYSCTL_OUT(wr, &ndpc, sizeof(ndpc));
2445 
2446 	return (error);
2447 }
2448 
2449 static struct lltable *
2450 in6_lltattach(struct ifnet *ifp)
2451 {
2452 	struct lltable *llt;
2453 
2454 	llt = lltable_allocate_htbl(IN6_LLTBL_DEFAULT_HSIZE);
2455 	llt->llt_af = AF_INET6;
2456 	llt->llt_ifp = ifp;
2457 
2458 	llt->llt_lookup = in6_lltable_lookup;
2459 	llt->llt_alloc_entry = in6_lltable_alloc;
2460 	llt->llt_delete_entry = in6_lltable_delete_entry;
2461 	llt->llt_dump_entry = in6_lltable_dump_entry;
2462 	llt->llt_hash = in6_lltable_hash;
2463 	llt->llt_fill_sa_entry = in6_lltable_fill_sa_entry;
2464 	llt->llt_free_entry = in6_lltable_free_entry;
2465 	llt->llt_match_prefix = in6_lltable_match_prefix;
2466 	llt->llt_mark_used = in6_lltable_mark_used;
2467  	lltable_link(llt);
2468 
2469 	return (llt);
2470 }
2471 
2472 void *
2473 in6_domifattach(struct ifnet *ifp)
2474 {
2475 	struct in6_ifextra *ext;
2476 
2477 	/* There are not IPv6-capable interfaces. */
2478 	switch (ifp->if_type) {
2479 	case IFT_PFLOG:
2480 	case IFT_PFSYNC:
2481 	case IFT_USB:
2482 		return (NULL);
2483 	}
2484 	ext = (struct in6_ifextra *)malloc(sizeof(*ext), M_IFADDR, M_WAITOK);
2485 	bzero(ext, sizeof(*ext));
2486 
2487 	ext->in6_ifstat = malloc(sizeof(counter_u64_t) *
2488 	    sizeof(struct in6_ifstat) / sizeof(uint64_t), M_IFADDR, M_WAITOK);
2489 	COUNTER_ARRAY_ALLOC(ext->in6_ifstat,
2490 	    sizeof(struct in6_ifstat) / sizeof(uint64_t), M_WAITOK);
2491 
2492 	ext->icmp6_ifstat = malloc(sizeof(counter_u64_t) *
2493 	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t), M_IFADDR,
2494 	    M_WAITOK);
2495 	COUNTER_ARRAY_ALLOC(ext->icmp6_ifstat,
2496 	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t), M_WAITOK);
2497 
2498 	ext->nd_ifinfo = nd6_ifattach(ifp);
2499 	ext->scope6_id = scope6_ifattach(ifp);
2500 	ext->lltable = in6_lltattach(ifp);
2501 
2502 	ext->mld_ifinfo = mld_domifattach(ifp);
2503 
2504 	return ext;
2505 }
2506 
2507 int
2508 in6_domifmtu(struct ifnet *ifp)
2509 {
2510 	if (ifp->if_afdata[AF_INET6] == NULL)
2511 		return ifp->if_mtu;
2512 
2513 	return (IN6_LINKMTU(ifp));
2514 }
2515 
2516 void
2517 in6_domifdetach(struct ifnet *ifp, void *aux)
2518 {
2519 	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
2520 
2521 	mld_domifdetach(ifp);
2522 	scope6_ifdetach(ext->scope6_id);
2523 	nd6_ifdetach(ifp, ext->nd_ifinfo);
2524 	lltable_free(ext->lltable);
2525 	COUNTER_ARRAY_FREE(ext->in6_ifstat,
2526 	    sizeof(struct in6_ifstat) / sizeof(uint64_t));
2527 	free(ext->in6_ifstat, M_IFADDR);
2528 	COUNTER_ARRAY_FREE(ext->icmp6_ifstat,
2529 	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t));
2530 	free(ext->icmp6_ifstat, M_IFADDR);
2531 	free(ext, M_IFADDR);
2532 }
2533 
2534 /*
2535  * Convert sockaddr_in6 to sockaddr_in.  Original sockaddr_in6 must be
2536  * v4 mapped addr or v4 compat addr
2537  */
2538 void
2539 in6_sin6_2_sin(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2540 {
2541 
2542 	bzero(sin, sizeof(*sin));
2543 	sin->sin_len = sizeof(struct sockaddr_in);
2544 	sin->sin_family = AF_INET;
2545 	sin->sin_port = sin6->sin6_port;
2546 	sin->sin_addr.s_addr = sin6->sin6_addr.s6_addr32[3];
2547 }
2548 
2549 /* Convert sockaddr_in to sockaddr_in6 in v4 mapped addr format. */
2550 void
2551 in6_sin_2_v4mapsin6(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2552 {
2553 	bzero(sin6, sizeof(*sin6));
2554 	sin6->sin6_len = sizeof(struct sockaddr_in6);
2555 	sin6->sin6_family = AF_INET6;
2556 	sin6->sin6_port = sin->sin_port;
2557 	sin6->sin6_addr.s6_addr32[0] = 0;
2558 	sin6->sin6_addr.s6_addr32[1] = 0;
2559 	sin6->sin6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
2560 	sin6->sin6_addr.s6_addr32[3] = sin->sin_addr.s_addr;
2561 }
2562 
2563 /* Convert sockaddr_in6 into sockaddr_in. */
2564 void
2565 in6_sin6_2_sin_in_sock(struct sockaddr *nam)
2566 {
2567 	struct sockaddr_in *sin_p;
2568 	struct sockaddr_in6 sin6;
2569 
2570 	/*
2571 	 * Save original sockaddr_in6 addr and convert it
2572 	 * to sockaddr_in.
2573 	 */
2574 	sin6 = *(struct sockaddr_in6 *)nam;
2575 	sin_p = (struct sockaddr_in *)nam;
2576 	in6_sin6_2_sin(sin_p, &sin6);
2577 }
2578 
2579 /* Convert sockaddr_in into sockaddr_in6 in v4 mapped addr format. */
2580 void
2581 in6_sin_2_v4mapsin6_in_sock(struct sockaddr **nam)
2582 {
2583 	struct sockaddr_in *sin_p;
2584 	struct sockaddr_in6 *sin6_p;
2585 
2586 	sin6_p = malloc(sizeof *sin6_p, M_SONAME, M_WAITOK);
2587 	sin_p = (struct sockaddr_in *)*nam;
2588 	in6_sin_2_v4mapsin6(sin_p, sin6_p);
2589 	free(*nam, M_SONAME);
2590 	*nam = (struct sockaddr *)sin6_p;
2591 }
2592