xref: /freebsd/sys/netinet6/in6.c (revision 5ab1c5846ff41be24b1f6beb0317bf8258cd4409)
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 sockaddr_dl gateway;
172 	struct sockaddr_in6 mask, addr;
173 	struct rtentry rt;
174 	int fibnum;
175 
176 	/*
177 	 * initialize for rtmsg generation
178 	 */
179 	bzero(&gateway, sizeof(gateway));
180 	gateway.sdl_len = sizeof(gateway);
181 	gateway.sdl_family = AF_LINK;
182 
183 	bzero(&rt, sizeof(rt));
184 	rt.rt_gateway = (struct sockaddr *)&gateway;
185 	memcpy(&mask, &ia->ia_prefixmask, sizeof(ia->ia_prefixmask));
186 	memcpy(&addr, &ia->ia_addr, sizeof(ia->ia_addr));
187 	rt_mask(&rt) = (struct sockaddr *)&mask;
188 	rt_key(&rt) = (struct sockaddr *)&addr;
189 	rt.rt_flags = RTF_HOST | RTF_STATIC;
190 	if (cmd == RTM_ADD)
191 		rt.rt_flags |= RTF_UP;
192 	fibnum = V_rt_add_addr_allfibs ? RT_ALL_FIBS : ia62ifa(ia)->ifa_ifp->if_fib;
193 	/* Announce arrival of local address to this FIB. */
194 	rt_newaddrmsg_fib(cmd, &ia->ia_ifa, 0, &rt, fibnum);
195 }
196 
197 int
198 in6_mask2len(struct in6_addr *mask, u_char *lim0)
199 {
200 	int x = 0, y;
201 	u_char *lim = lim0, *p;
202 
203 	/* ignore the scope_id part */
204 	if (lim0 == NULL || lim0 - (u_char *)mask > sizeof(*mask))
205 		lim = (u_char *)mask + sizeof(*mask);
206 	for (p = (u_char *)mask; p < lim; x++, p++) {
207 		if (*p != 0xff)
208 			break;
209 	}
210 	y = 0;
211 	if (p < lim) {
212 		for (y = 0; y < 8; y++) {
213 			if ((*p & (0x80 >> y)) == 0)
214 				break;
215 		}
216 	}
217 
218 	/*
219 	 * when the limit pointer is given, do a stricter check on the
220 	 * remaining bits.
221 	 */
222 	if (p < lim) {
223 		if (y != 0 && (*p & (0x00ff >> y)) != 0)
224 			return (-1);
225 		for (p = p + 1; p < lim; p++)
226 			if (*p != 0)
227 				return (-1);
228 	}
229 
230 	return x * 8 + y;
231 }
232 
233 #ifdef COMPAT_FREEBSD32
234 struct in6_ndifreq32 {
235 	char ifname[IFNAMSIZ];
236 	uint32_t ifindex;
237 };
238 #define	SIOCGDEFIFACE32_IN6	_IOWR('i', 86, struct in6_ndifreq32)
239 #endif
240 
241 int
242 in6_control(struct socket *so, u_long cmd, caddr_t data,
243     struct ifnet *ifp, struct thread *td)
244 {
245 	struct	in6_ifreq *ifr = (struct in6_ifreq *)data;
246 	struct	in6_ifaddr *ia = NULL;
247 	struct	in6_aliasreq *ifra = (struct in6_aliasreq *)data;
248 	struct sockaddr_in6 *sa6;
249 	int carp_attached = 0;
250 	int error;
251 	u_long ocmd = cmd;
252 
253 	/*
254 	 * Compat to make pre-10.x ifconfig(8) operable.
255 	 */
256 	if (cmd == OSIOCAIFADDR_IN6)
257 		cmd = SIOCAIFADDR_IN6;
258 
259 	switch (cmd) {
260 	case SIOCGETSGCNT_IN6:
261 	case SIOCGETMIFCNT_IN6:
262 		/*
263 		 * XXX mrt_ioctl has a 3rd, unused, FIB argument in route.c.
264 		 * We cannot see how that would be needed, so do not adjust the
265 		 * KPI blindly; more likely should clean up the IPv4 variant.
266 		 */
267 		return (mrt6_ioctl ? mrt6_ioctl(cmd, data) : EOPNOTSUPP);
268 	}
269 
270 	switch (cmd) {
271 	case SIOCAADDRCTL_POLICY:
272 	case SIOCDADDRCTL_POLICY:
273 		if (td != NULL) {
274 			error = priv_check(td, PRIV_NETINET_ADDRCTRL6);
275 			if (error)
276 				return (error);
277 		}
278 		return (in6_src_ioctl(cmd, data));
279 	}
280 
281 	if (ifp == NULL)
282 		return (EOPNOTSUPP);
283 
284 	switch (cmd) {
285 	case SIOCSNDFLUSH_IN6:
286 	case SIOCSPFXFLUSH_IN6:
287 	case SIOCSRTRFLUSH_IN6:
288 	case SIOCSDEFIFACE_IN6:
289 	case SIOCSIFINFO_FLAGS:
290 	case SIOCSIFINFO_IN6:
291 		if (td != NULL) {
292 			error = priv_check(td, PRIV_NETINET_ND6);
293 			if (error)
294 				return (error);
295 		}
296 		/* FALLTHROUGH */
297 	case OSIOCGIFINFO_IN6:
298 	case SIOCGIFINFO_IN6:
299 	case SIOCGNBRINFO_IN6:
300 	case SIOCGDEFIFACE_IN6:
301 		return (nd6_ioctl(cmd, data, ifp));
302 
303 #ifdef COMPAT_FREEBSD32
304 	case SIOCGDEFIFACE32_IN6:
305 		{
306 			struct in6_ndifreq ndif;
307 			struct in6_ndifreq32 *ndif32;
308 
309 			error = nd6_ioctl(SIOCGDEFIFACE_IN6, (caddr_t)&ndif,
310 			    ifp);
311 			if (error)
312 				return (error);
313 			ndif32 = (struct in6_ndifreq32 *)data;
314 			ndif32->ifindex = ndif.ifindex;
315 			return (0);
316 		}
317 #endif
318 	}
319 
320 	switch (cmd) {
321 	case SIOCSIFPREFIX_IN6:
322 	case SIOCDIFPREFIX_IN6:
323 	case SIOCAIFPREFIX_IN6:
324 	case SIOCCIFPREFIX_IN6:
325 	case SIOCSGIFPREFIX_IN6:
326 	case SIOCGIFPREFIX_IN6:
327 		log(LOG_NOTICE,
328 		    "prefix ioctls are now invalidated. "
329 		    "please use ifconfig.\n");
330 		return (EOPNOTSUPP);
331 	}
332 
333 	switch (cmd) {
334 	case SIOCSSCOPE6:
335 		if (td != NULL) {
336 			error = priv_check(td, PRIV_NETINET_SCOPE6);
337 			if (error)
338 				return (error);
339 		}
340 		/* FALLTHROUGH */
341 	case SIOCGSCOPE6:
342 	case SIOCGSCOPE6DEF:
343 		return (scope6_ioctl(cmd, data, ifp));
344 	}
345 
346 	/*
347 	 * Find address for this interface, if it exists.
348 	 *
349 	 * In netinet code, we have checked ifra_addr in SIOCSIF*ADDR operation
350 	 * only, and used the first interface address as the target of other
351 	 * operations (without checking ifra_addr).  This was because netinet
352 	 * code/API assumed at most 1 interface address per interface.
353 	 * Since IPv6 allows a node to assign multiple addresses
354 	 * on a single interface, we almost always look and check the
355 	 * presence of ifra_addr, and reject invalid ones here.
356 	 * It also decreases duplicated code among SIOC*_IN6 operations.
357 	 */
358 	switch (cmd) {
359 	case SIOCAIFADDR_IN6:
360 	case SIOCSIFPHYADDR_IN6:
361 		sa6 = &ifra->ifra_addr;
362 		break;
363 	case SIOCSIFADDR_IN6:
364 	case SIOCGIFADDR_IN6:
365 	case SIOCSIFDSTADDR_IN6:
366 	case SIOCSIFNETMASK_IN6:
367 	case SIOCGIFDSTADDR_IN6:
368 	case SIOCGIFNETMASK_IN6:
369 	case SIOCDIFADDR_IN6:
370 	case SIOCGIFPSRCADDR_IN6:
371 	case SIOCGIFPDSTADDR_IN6:
372 	case SIOCGIFAFLAG_IN6:
373 	case SIOCSNDFLUSH_IN6:
374 	case SIOCSPFXFLUSH_IN6:
375 	case SIOCSRTRFLUSH_IN6:
376 	case SIOCGIFALIFETIME_IN6:
377 	case SIOCGIFSTAT_IN6:
378 	case SIOCGIFSTAT_ICMP6:
379 		sa6 = &ifr->ifr_addr;
380 		break;
381 	case SIOCSIFADDR:
382 	case SIOCSIFBRDADDR:
383 	case SIOCSIFDSTADDR:
384 	case SIOCSIFNETMASK:
385 		/*
386 		 * Although we should pass any non-INET6 ioctl requests
387 		 * down to driver, we filter some legacy INET requests.
388 		 * Drivers trust SIOCSIFADDR et al to come from an already
389 		 * privileged layer, and do not perform any credentials
390 		 * checks or input validation.
391 		 */
392 		return (EINVAL);
393 	default:
394 		sa6 = NULL;
395 		break;
396 	}
397 	if (sa6 && sa6->sin6_family == AF_INET6) {
398 		if (sa6->sin6_scope_id != 0)
399 			error = sa6_embedscope(sa6, 0);
400 		else
401 			error = in6_setscope(&sa6->sin6_addr, ifp, NULL);
402 		if (error != 0)
403 			return (error);
404 		if (td != NULL && (error = prison_check_ip6(td->td_ucred,
405 		    &sa6->sin6_addr)) != 0)
406 			return (error);
407 		ia = in6ifa_ifpwithaddr(ifp, &sa6->sin6_addr);
408 	} else
409 		ia = NULL;
410 
411 	switch (cmd) {
412 	case SIOCSIFADDR_IN6:
413 	case SIOCSIFDSTADDR_IN6:
414 	case SIOCSIFNETMASK_IN6:
415 		/*
416 		 * Since IPv6 allows a node to assign multiple addresses
417 		 * on a single interface, SIOCSIFxxx ioctls are deprecated.
418 		 */
419 		/* we decided to obsolete this command (20000704) */
420 		error = EINVAL;
421 		goto out;
422 
423 	case SIOCDIFADDR_IN6:
424 		/*
425 		 * for IPv4, we look for existing in_ifaddr here to allow
426 		 * "ifconfig if0 delete" to remove the first IPv4 address on
427 		 * the interface.  For IPv6, as the spec allows multiple
428 		 * interface address from the day one, we consider "remove the
429 		 * first one" semantics to be not preferable.
430 		 */
431 		if (ia == NULL) {
432 			error = EADDRNOTAVAIL;
433 			goto out;
434 		}
435 		/* FALLTHROUGH */
436 	case SIOCAIFADDR_IN6:
437 		/*
438 		 * We always require users to specify a valid IPv6 address for
439 		 * the corresponding operation.
440 		 */
441 		if (ifra->ifra_addr.sin6_family != AF_INET6 ||
442 		    ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6)) {
443 			error = EAFNOSUPPORT;
444 			goto out;
445 		}
446 
447 		if (td != NULL) {
448 			error = priv_check(td, (cmd == SIOCDIFADDR_IN6) ?
449 			    PRIV_NET_DELIFADDR : PRIV_NET_ADDIFADDR);
450 			if (error)
451 				goto out;
452 		}
453 		/* FALLTHROUGH */
454 	case SIOCGIFSTAT_IN6:
455 	case SIOCGIFSTAT_ICMP6:
456 		if (ifp->if_afdata[AF_INET6] == NULL) {
457 			error = EPFNOSUPPORT;
458 			goto out;
459 		}
460 		break;
461 
462 	case SIOCGIFADDR_IN6:
463 		/* This interface is basically deprecated. use SIOCGIFCONF. */
464 		/* FALLTHROUGH */
465 	case SIOCGIFAFLAG_IN6:
466 	case SIOCGIFNETMASK_IN6:
467 	case SIOCGIFDSTADDR_IN6:
468 	case SIOCGIFALIFETIME_IN6:
469 		/* must think again about its semantics */
470 		if (ia == NULL) {
471 			error = EADDRNOTAVAIL;
472 			goto out;
473 		}
474 		break;
475 	}
476 
477 	switch (cmd) {
478 	case SIOCGIFADDR_IN6:
479 		ifr->ifr_addr = ia->ia_addr;
480 		if ((error = sa6_recoverscope(&ifr->ifr_addr)) != 0)
481 			goto out;
482 		break;
483 
484 	case SIOCGIFDSTADDR_IN6:
485 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
486 			error = EINVAL;
487 			goto out;
488 		}
489 		ifr->ifr_dstaddr = ia->ia_dstaddr;
490 		if ((error = sa6_recoverscope(&ifr->ifr_dstaddr)) != 0)
491 			goto out;
492 		break;
493 
494 	case SIOCGIFNETMASK_IN6:
495 		ifr->ifr_addr = ia->ia_prefixmask;
496 		break;
497 
498 	case SIOCGIFAFLAG_IN6:
499 		ifr->ifr_ifru.ifru_flags6 = ia->ia6_flags;
500 		break;
501 
502 	case SIOCGIFSTAT_IN6:
503 		COUNTER_ARRAY_COPY(((struct in6_ifextra *)
504 		    ifp->if_afdata[AF_INET6])->in6_ifstat,
505 		    &ifr->ifr_ifru.ifru_stat,
506 		    sizeof(struct in6_ifstat) / sizeof(uint64_t));
507 		break;
508 
509 	case SIOCGIFSTAT_ICMP6:
510 		COUNTER_ARRAY_COPY(((struct in6_ifextra *)
511 		    ifp->if_afdata[AF_INET6])->icmp6_ifstat,
512 		    &ifr->ifr_ifru.ifru_icmp6stat,
513 		    sizeof(struct icmp6_ifstat) / sizeof(uint64_t));
514 		break;
515 
516 	case SIOCGIFALIFETIME_IN6:
517 		ifr->ifr_ifru.ifru_lifetime = ia->ia6_lifetime;
518 		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
519 			time_t maxexpire;
520 			struct in6_addrlifetime *retlt =
521 			    &ifr->ifr_ifru.ifru_lifetime;
522 
523 			/*
524 			 * XXX: adjust expiration time assuming time_t is
525 			 * signed.
526 			 */
527 			maxexpire = (-1) &
528 			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
529 			if (ia->ia6_lifetime.ia6t_vltime <
530 			    maxexpire - ia->ia6_updatetime) {
531 				retlt->ia6t_expire = ia->ia6_updatetime +
532 				    ia->ia6_lifetime.ia6t_vltime;
533 			} else
534 				retlt->ia6t_expire = maxexpire;
535 		}
536 		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
537 			time_t maxexpire;
538 			struct in6_addrlifetime *retlt =
539 			    &ifr->ifr_ifru.ifru_lifetime;
540 
541 			/*
542 			 * XXX: adjust expiration time assuming time_t is
543 			 * signed.
544 			 */
545 			maxexpire = (-1) &
546 			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
547 			if (ia->ia6_lifetime.ia6t_pltime <
548 			    maxexpire - ia->ia6_updatetime) {
549 				retlt->ia6t_preferred = ia->ia6_updatetime +
550 				    ia->ia6_lifetime.ia6t_pltime;
551 			} else
552 				retlt->ia6t_preferred = maxexpire;
553 		}
554 		break;
555 
556 	case SIOCAIFADDR_IN6:
557 	{
558 		struct nd_prefixctl pr0;
559 		struct nd_prefix *pr;
560 
561 		/*
562 		 * first, make or update the interface address structure,
563 		 * and link it to the list.
564 		 */
565 		if ((error = in6_update_ifa(ifp, ifra, ia, 0)) != 0)
566 			goto out;
567 		if (ia != NULL) {
568 			if (ia->ia_ifa.ifa_carp)
569 				(*carp_detach_p)(&ia->ia_ifa, true);
570 			ifa_free(&ia->ia_ifa);
571 		}
572 		if ((ia = in6ifa_ifpwithaddr(ifp, &ifra->ifra_addr.sin6_addr))
573 		    == NULL) {
574 			/*
575 			 * this can happen when the user specify the 0 valid
576 			 * lifetime.
577 			 */
578 			break;
579 		}
580 
581 		if (cmd == ocmd && ifra->ifra_vhid > 0) {
582 			if (carp_attach_p != NULL)
583 				error = (*carp_attach_p)(&ia->ia_ifa,
584 				    ifra->ifra_vhid);
585 			else
586 				error = EPROTONOSUPPORT;
587 			if (error)
588 				goto out;
589 			else
590 				carp_attached = 1;
591 		}
592 
593 		/*
594 		 * then, make the prefix on-link on the interface.
595 		 * XXX: we'd rather create the prefix before the address, but
596 		 * we need at least one address to install the corresponding
597 		 * interface route, so we configure the address first.
598 		 */
599 
600 		/*
601 		 * convert mask to prefix length (prefixmask has already
602 		 * been validated in in6_update_ifa().
603 		 */
604 		bzero(&pr0, sizeof(pr0));
605 		pr0.ndpr_ifp = ifp;
606 		pr0.ndpr_plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
607 		    NULL);
608 		if (pr0.ndpr_plen == 128) {
609 			/* we don't need to install a host route. */
610 			goto aifaddr_out;
611 		}
612 		pr0.ndpr_prefix = ifra->ifra_addr;
613 		/* apply the mask for safety. */
614 		IN6_MASK_ADDR(&pr0.ndpr_prefix.sin6_addr,
615 		    &ifra->ifra_prefixmask.sin6_addr);
616 
617 		/*
618 		 * XXX: since we don't have an API to set prefix (not address)
619 		 * lifetimes, we just use the same lifetimes as addresses.
620 		 * The (temporarily) installed lifetimes can be overridden by
621 		 * later advertised RAs (when accept_rtadv is non 0), which is
622 		 * an intended behavior.
623 		 */
624 		pr0.ndpr_raf_onlink = 1; /* should be configurable? */
625 		pr0.ndpr_raf_auto =
626 		    ((ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0);
627 		pr0.ndpr_vltime = ifra->ifra_lifetime.ia6t_vltime;
628 		pr0.ndpr_pltime = ifra->ifra_lifetime.ia6t_pltime;
629 
630 		/* add the prefix if not yet. */
631 		if ((pr = nd6_prefix_lookup(&pr0)) == NULL) {
632 			/*
633 			 * nd6_prelist_add will install the corresponding
634 			 * interface route.
635 			 */
636 			if ((error = nd6_prelist_add(&pr0, NULL, &pr)) != 0) {
637 				if (carp_attached)
638 					(*carp_detach_p)(&ia->ia_ifa, false);
639 				goto out;
640 			}
641 		}
642 
643 		/* relate the address to the prefix */
644 		if (ia->ia6_ndpr == NULL) {
645 			ia->ia6_ndpr = pr;
646 			pr->ndpr_addrcnt++;
647 
648 			/*
649 			 * If this is the first autoconf address from the
650 			 * prefix, create a temporary address as well
651 			 * (when required).
652 			 */
653 			if ((ia->ia6_flags & IN6_IFF_AUTOCONF) &&
654 			    V_ip6_use_tempaddr && pr->ndpr_addrcnt == 1) {
655 				int e;
656 				if ((e = in6_tmpifadd(ia, 1, 0)) != 0) {
657 					log(LOG_NOTICE, "in6_control: failed "
658 					    "to create a temporary address, "
659 					    "errno=%d\n", e);
660 				}
661 			}
662 		}
663 		nd6_prefix_rele(pr);
664 
665 		/*
666 		 * this might affect the status of autoconfigured addresses,
667 		 * that is, this address might make other addresses detached.
668 		 */
669 		pfxlist_onlink_check();
670 
671 aifaddr_out:
672 		/*
673 		 * Try to clear the flag when a new IPv6 address is added
674 		 * onto an IFDISABLED interface and it succeeds.
675 		 */
676 		if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) {
677 			struct in6_ndireq nd;
678 
679 			memset(&nd, 0, sizeof(nd));
680 			nd.ndi.flags = ND_IFINFO(ifp)->flags;
681 			nd.ndi.flags &= ~ND6_IFF_IFDISABLED;
682 			if (nd6_ioctl(SIOCSIFINFO_FLAGS, (caddr_t)&nd, ifp) < 0)
683 				log(LOG_NOTICE, "SIOCAIFADDR_IN6: "
684 				    "SIOCSIFINFO_FLAGS for -ifdisabled "
685 				    "failed.");
686 			/*
687 			 * Ignore failure of clearing the flag intentionally.
688 			 * The failure means address duplication was detected.
689 			 */
690 		}
691 		break;
692 	}
693 
694 	case SIOCDIFADDR_IN6:
695 	{
696 		struct nd_prefix *pr;
697 
698 		/*
699 		 * If the address being deleted is the only one that owns
700 		 * the corresponding prefix, expire the prefix as well.
701 		 * XXX: theoretically, we don't have to worry about such
702 		 * relationship, since we separate the address management
703 		 * and the prefix management.  We do this, however, to provide
704 		 * as much backward compatibility as possible in terms of
705 		 * the ioctl operation.
706 		 * Note that in6_purgeaddr() will decrement ndpr_addrcnt.
707 		 */
708 		pr = ia->ia6_ndpr;
709 		in6_purgeaddr(&ia->ia_ifa);
710 		if (pr != NULL && pr->ndpr_addrcnt == 0) {
711 			ND6_WLOCK();
712 			nd6_prefix_unlink(pr, NULL);
713 			ND6_WUNLOCK();
714 			nd6_prefix_del(pr);
715 		}
716 		EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, &ia->ia_ifa,
717 		    IFADDR_EVENT_DEL);
718 		break;
719 	}
720 
721 	default:
722 		if (ifp->if_ioctl == NULL) {
723 			error = EOPNOTSUPP;
724 			goto out;
725 		}
726 		error = (*ifp->if_ioctl)(ifp, cmd, data);
727 		goto out;
728 	}
729 
730 	error = 0;
731 out:
732 	if (ia != NULL)
733 		ifa_free(&ia->ia_ifa);
734 	return (error);
735 }
736 
737 
738 static struct in6_multi_mship *
739 in6_joingroup_legacy(struct ifnet *ifp, const struct in6_addr *mcaddr,
740     int *errorp, int delay)
741 {
742 	struct in6_multi_mship *imm;
743 	int error;
744 
745 	imm = malloc(sizeof(*imm), M_IP6MADDR, M_NOWAIT);
746 	if (imm == NULL) {
747 		*errorp = ENOBUFS;
748 		return (NULL);
749 	}
750 
751 	delay = (delay * PR_FASTHZ) / hz;
752 
753 	error = in6_joingroup(ifp, mcaddr, NULL, &imm->i6mm_maddr, delay);
754 	if (error) {
755 		*errorp = error;
756 		free(imm, M_IP6MADDR);
757 		return (NULL);
758 	}
759 
760 	return (imm);
761 }
762 /*
763  * Join necessary multicast groups.  Factored out from in6_update_ifa().
764  * This entire work should only be done once, for the default FIB.
765  */
766 static int
767 in6_update_ifa_join_mc(struct ifnet *ifp, struct in6_aliasreq *ifra,
768     struct in6_ifaddr *ia, int flags, struct in6_multi **in6m_sol)
769 {
770 	char ip6buf[INET6_ADDRSTRLEN];
771 	struct in6_addr mltaddr;
772 	struct in6_multi_mship *imm;
773 	int delay, error;
774 
775 	KASSERT(in6m_sol != NULL, ("%s: in6m_sol is NULL", __func__));
776 
777 	/* Join solicited multicast addr for new host id. */
778 	bzero(&mltaddr, sizeof(struct in6_addr));
779 	mltaddr.s6_addr32[0] = IPV6_ADDR_INT32_MLL;
780 	mltaddr.s6_addr32[2] = htonl(1);
781 	mltaddr.s6_addr32[3] = ifra->ifra_addr.sin6_addr.s6_addr32[3];
782 	mltaddr.s6_addr8[12] = 0xff;
783 	if ((error = in6_setscope(&mltaddr, ifp, NULL)) != 0) {
784 		/* XXX: should not happen */
785 		log(LOG_ERR, "%s: in6_setscope failed\n", __func__);
786 		goto cleanup;
787 	}
788 	delay = error = 0;
789 	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
790 		/*
791 		 * We need a random delay for DAD on the address being
792 		 * configured.  It also means delaying transmission of the
793 		 * corresponding MLD report to avoid report collision.
794 		 * [RFC 4861, Section 6.3.7]
795 		 */
796 		delay = arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz);
797 	}
798 	imm = in6_joingroup_legacy(ifp, &mltaddr, &error, delay);
799 	if (imm == NULL) {
800 		nd6log((LOG_WARNING, "%s: in6_joingroup failed for %s on %s "
801 		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf, &mltaddr),
802 		    if_name(ifp), error));
803 		goto cleanup;
804 	}
805 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
806 	*in6m_sol = imm->i6mm_maddr;
807 
808 	/*
809 	 * Join link-local all-nodes address.
810 	 */
811 	mltaddr = in6addr_linklocal_allnodes;
812 	if ((error = in6_setscope(&mltaddr, ifp, NULL)) != 0)
813 		goto cleanup; /* XXX: should not fail */
814 
815 	imm = in6_joingroup_legacy(ifp, &mltaddr, &error, 0);
816 	if (imm == NULL) {
817 		nd6log((LOG_WARNING, "%s: in6_joingroup failed for %s on %s "
818 		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf, &mltaddr),
819 		    if_name(ifp), error));
820 		goto cleanup;
821 	}
822 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
823 
824 	/*
825 	 * Join node information group address.
826 	 */
827 	delay = 0;
828 	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
829 		/*
830 		 * The spec does not say anything about delay for this group,
831 		 * but the same logic should apply.
832 		 */
833 		delay = arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz);
834 	}
835 	if (in6_nigroup(ifp, NULL, -1, &mltaddr) == 0) {
836 		/* XXX jinmei */
837 		imm = in6_joingroup_legacy(ifp, &mltaddr, &error, delay);
838 		if (imm == NULL)
839 			nd6log((LOG_WARNING,
840 			    "%s: in6_joingroup failed for %s on %s "
841 			    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
842 			    &mltaddr), if_name(ifp), error));
843 			/* XXX not very fatal, go on... */
844 		else
845 			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
846 	}
847 	if (V_icmp6_nodeinfo_oldmcprefix &&
848 	    in6_nigroup_oldmcprefix(ifp, NULL, -1, &mltaddr) == 0) {
849 		imm = in6_joingroup_legacy(ifp, &mltaddr, &error, delay);
850 		if (imm == NULL)
851 			nd6log((LOG_WARNING,
852 			    "%s: in6_joingroup failed for %s on %s "
853 			    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
854 			    &mltaddr), if_name(ifp), error));
855 			/* XXX not very fatal, go on... */
856 		else
857 			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
858 	}
859 
860 	/*
861 	 * Join interface-local all-nodes address.
862 	 * (ff01::1%ifN, and ff01::%ifN/32)
863 	 */
864 	mltaddr = in6addr_nodelocal_allnodes;
865 	if ((error = in6_setscope(&mltaddr, ifp, NULL)) != 0)
866 		goto cleanup; /* XXX: should not fail */
867 
868 	imm = in6_joingroup_legacy(ifp, &mltaddr, &error, 0);
869 	if (imm == NULL) {
870 		nd6log((LOG_WARNING, "%s: in6_joingroup failed for %s on %s "
871 		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
872 		    &mltaddr), if_name(ifp), error));
873 		goto cleanup;
874 	}
875 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
876 
877 cleanup:
878 	return (error);
879 }
880 
881 /*
882  * Update parameters of an IPv6 interface address.
883  * If necessary, a new entry is created and linked into address chains.
884  * This function is separated from in6_control().
885  */
886 int
887 in6_update_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
888     struct in6_ifaddr *ia, int flags)
889 {
890 	int error, hostIsNew = 0;
891 
892 	if ((error = in6_validate_ifra(ifp, ifra, ia, flags)) != 0)
893 		return (error);
894 
895 	if (ia == NULL) {
896 		hostIsNew = 1;
897 		if ((ia = in6_alloc_ifa(ifp, ifra, flags)) == NULL)
898 			return (ENOBUFS);
899 	}
900 
901 	error = in6_update_ifa_internal(ifp, ifra, ia, hostIsNew, flags);
902 	if (error != 0) {
903 		if (hostIsNew != 0) {
904 			in6_unlink_ifa(ia, ifp);
905 			ifa_free(&ia->ia_ifa);
906 		}
907 		return (error);
908 	}
909 
910 	if (hostIsNew)
911 		error = in6_broadcast_ifa(ifp, ifra, ia, flags);
912 
913 	return (error);
914 }
915 
916 /*
917  * Fill in basic IPv6 address request info.
918  */
919 void
920 in6_prepare_ifra(struct in6_aliasreq *ifra, const struct in6_addr *addr,
921     const struct in6_addr *mask)
922 {
923 
924 	memset(ifra, 0, sizeof(struct in6_aliasreq));
925 
926 	ifra->ifra_addr.sin6_family = AF_INET6;
927 	ifra->ifra_addr.sin6_len = sizeof(struct sockaddr_in6);
928 	if (addr != NULL)
929 		ifra->ifra_addr.sin6_addr = *addr;
930 
931 	ifra->ifra_prefixmask.sin6_family = AF_INET6;
932 	ifra->ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
933 	if (mask != NULL)
934 		ifra->ifra_prefixmask.sin6_addr = *mask;
935 }
936 
937 static int
938 in6_validate_ifra(struct ifnet *ifp, struct in6_aliasreq *ifra,
939     struct in6_ifaddr *ia, int flags)
940 {
941 	int plen = -1;
942 	struct sockaddr_in6 dst6;
943 	struct in6_addrlifetime *lt;
944 	char ip6buf[INET6_ADDRSTRLEN];
945 
946 	/* Validate parameters */
947 	if (ifp == NULL || ifra == NULL) /* this maybe redundant */
948 		return (EINVAL);
949 
950 	/*
951 	 * The destination address for a p2p link must have a family
952 	 * of AF_UNSPEC or AF_INET6.
953 	 */
954 	if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
955 	    ifra->ifra_dstaddr.sin6_family != AF_INET6 &&
956 	    ifra->ifra_dstaddr.sin6_family != AF_UNSPEC)
957 		return (EAFNOSUPPORT);
958 
959 	/*
960 	 * Validate address
961 	 */
962 	if (ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6) ||
963 	    ifra->ifra_addr.sin6_family != AF_INET6)
964 		return (EINVAL);
965 
966 	/*
967 	 * validate ifra_prefixmask.  don't check sin6_family, netmask
968 	 * does not carry fields other than sin6_len.
969 	 */
970 	if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6))
971 		return (EINVAL);
972 	/*
973 	 * Because the IPv6 address architecture is classless, we require
974 	 * users to specify a (non 0) prefix length (mask) for a new address.
975 	 * We also require the prefix (when specified) mask is valid, and thus
976 	 * reject a non-consecutive mask.
977 	 */
978 	if (ia == NULL && ifra->ifra_prefixmask.sin6_len == 0)
979 		return (EINVAL);
980 	if (ifra->ifra_prefixmask.sin6_len != 0) {
981 		plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
982 		    (u_char *)&ifra->ifra_prefixmask +
983 		    ifra->ifra_prefixmask.sin6_len);
984 		if (plen <= 0)
985 			return (EINVAL);
986 	} else {
987 		/*
988 		 * In this case, ia must not be NULL.  We just use its prefix
989 		 * length.
990 		 */
991 		plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
992 	}
993 	/*
994 	 * If the destination address on a p2p interface is specified,
995 	 * and the address is a scoped one, validate/set the scope
996 	 * zone identifier.
997 	 */
998 	dst6 = ifra->ifra_dstaddr;
999 	if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) != 0 &&
1000 	    (dst6.sin6_family == AF_INET6)) {
1001 		struct in6_addr in6_tmp;
1002 		u_int32_t zoneid;
1003 
1004 		in6_tmp = dst6.sin6_addr;
1005 		if (in6_setscope(&in6_tmp, ifp, &zoneid))
1006 			return (EINVAL); /* XXX: should be impossible */
1007 
1008 		if (dst6.sin6_scope_id != 0) {
1009 			if (dst6.sin6_scope_id != zoneid)
1010 				return (EINVAL);
1011 		} else		/* user omit to specify the ID. */
1012 			dst6.sin6_scope_id = zoneid;
1013 
1014 		/* convert into the internal form */
1015 		if (sa6_embedscope(&dst6, 0))
1016 			return (EINVAL); /* XXX: should be impossible */
1017 	}
1018 	/* Modify original ifra_dstaddr to reflect changes */
1019 	ifra->ifra_dstaddr = dst6;
1020 
1021 	/*
1022 	 * The destination address can be specified only for a p2p or a
1023 	 * loopback interface.  If specified, the corresponding prefix length
1024 	 * must be 128.
1025 	 */
1026 	if (ifra->ifra_dstaddr.sin6_family == AF_INET6) {
1027 		if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) == 0) {
1028 			/* XXX: noisy message */
1029 			nd6log((LOG_INFO, "in6_update_ifa: a destination can "
1030 			    "be specified for a p2p or a loopback IF only\n"));
1031 			return (EINVAL);
1032 		}
1033 		if (plen != 128) {
1034 			nd6log((LOG_INFO, "in6_update_ifa: prefixlen should "
1035 			    "be 128 when dstaddr is specified\n"));
1036 			return (EINVAL);
1037 		}
1038 	}
1039 	/* lifetime consistency check */
1040 	lt = &ifra->ifra_lifetime;
1041 	if (lt->ia6t_pltime > lt->ia6t_vltime)
1042 		return (EINVAL);
1043 	if (lt->ia6t_vltime == 0) {
1044 		/*
1045 		 * the following log might be noisy, but this is a typical
1046 		 * configuration mistake or a tool's bug.
1047 		 */
1048 		nd6log((LOG_INFO,
1049 		    "in6_update_ifa: valid lifetime is 0 for %s\n",
1050 		    ip6_sprintf(ip6buf, &ifra->ifra_addr.sin6_addr)));
1051 
1052 		if (ia == NULL)
1053 			return (0); /* there's nothing to do */
1054 	}
1055 
1056 	/* Check prefix mask */
1057 	if (ia != NULL && ifra->ifra_prefixmask.sin6_len != 0) {
1058 		/*
1059 		 * We prohibit changing the prefix length of an existing
1060 		 * address, because
1061 		 * + such an operation should be rare in IPv6, and
1062 		 * + the operation would confuse prefix management.
1063 		 */
1064 		if (ia->ia_prefixmask.sin6_len != 0 &&
1065 		    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) != plen) {
1066 			nd6log((LOG_INFO, "in6_validate_ifa: the prefix length "
1067 			    "of an existing %s address should not be changed\n",
1068 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1069 
1070 			return (EINVAL);
1071 		}
1072 	}
1073 
1074 	return (0);
1075 }
1076 
1077 
1078 /*
1079  * Allocate a new ifaddr and link it into chains.
1080  */
1081 static struct in6_ifaddr *
1082 in6_alloc_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra, int flags)
1083 {
1084 	struct in6_ifaddr *ia;
1085 
1086 	/*
1087 	 * When in6_alloc_ifa() is called in a process of a received
1088 	 * RA, it is called under an interrupt context.  So, we should
1089 	 * call malloc with M_NOWAIT.
1090 	 */
1091 	ia = (struct in6_ifaddr *)ifa_alloc(sizeof(*ia), M_NOWAIT);
1092 	if (ia == NULL)
1093 		return (NULL);
1094 	LIST_INIT(&ia->ia6_memberships);
1095 	/* Initialize the address and masks, and put time stamp */
1096 	ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
1097 	ia->ia_addr.sin6_family = AF_INET6;
1098 	ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
1099 	/* XXX: Can we assign ,sin6_addr and skip the rest? */
1100 	ia->ia_addr = ifra->ifra_addr;
1101 	ia->ia6_createtime = time_uptime;
1102 	if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
1103 		/*
1104 		 * Some functions expect that ifa_dstaddr is not
1105 		 * NULL for p2p interfaces.
1106 		 */
1107 		ia->ia_ifa.ifa_dstaddr =
1108 		    (struct sockaddr *)&ia->ia_dstaddr;
1109 	} else {
1110 		ia->ia_ifa.ifa_dstaddr = NULL;
1111 	}
1112 
1113 	/* set prefix mask if any */
1114 	ia->ia_ifa.ifa_netmask = (struct sockaddr *)&ia->ia_prefixmask;
1115 	if (ifra->ifra_prefixmask.sin6_len != 0) {
1116 		ia->ia_prefixmask.sin6_family = AF_INET6;
1117 		ia->ia_prefixmask.sin6_len = ifra->ifra_prefixmask.sin6_len;
1118 		ia->ia_prefixmask.sin6_addr = ifra->ifra_prefixmask.sin6_addr;
1119 	}
1120 
1121 	ia->ia_ifp = ifp;
1122 	ifa_ref(&ia->ia_ifa);			/* if_addrhead */
1123 	IF_ADDR_WLOCK(ifp);
1124 	CK_STAILQ_INSERT_TAIL(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
1125 	IF_ADDR_WUNLOCK(ifp);
1126 
1127 	ifa_ref(&ia->ia_ifa);			/* in6_ifaddrhead */
1128 	IN6_IFADDR_WLOCK();
1129 	CK_STAILQ_INSERT_TAIL(&V_in6_ifaddrhead, ia, ia_link);
1130 	CK_LIST_INSERT_HEAD(IN6ADDR_HASH(&ia->ia_addr.sin6_addr), ia, ia6_hash);
1131 	IN6_IFADDR_WUNLOCK();
1132 
1133 	return (ia);
1134 }
1135 
1136 /*
1137  * Update/configure interface address parameters:
1138  *
1139  * 1) Update lifetime
1140  * 2) Update interface metric ad flags
1141  * 3) Notify other subsystems
1142  */
1143 static int
1144 in6_update_ifa_internal(struct ifnet *ifp, struct in6_aliasreq *ifra,
1145     struct in6_ifaddr *ia, int hostIsNew, int flags)
1146 {
1147 	int error;
1148 
1149 	/* update timestamp */
1150 	ia->ia6_updatetime = time_uptime;
1151 
1152 	/*
1153 	 * Set lifetimes.  We do not refer to ia6t_expire and ia6t_preferred
1154 	 * to see if the address is deprecated or invalidated, but initialize
1155 	 * these members for applications.
1156 	 */
1157 	ia->ia6_lifetime = ifra->ifra_lifetime;
1158 	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1159 		ia->ia6_lifetime.ia6t_expire =
1160 		    time_uptime + ia->ia6_lifetime.ia6t_vltime;
1161 	} else
1162 		ia->ia6_lifetime.ia6t_expire = 0;
1163 	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1164 		ia->ia6_lifetime.ia6t_preferred =
1165 		    time_uptime + ia->ia6_lifetime.ia6t_pltime;
1166 	} else
1167 		ia->ia6_lifetime.ia6t_preferred = 0;
1168 
1169 	/*
1170 	 * backward compatibility - if IN6_IFF_DEPRECATED is set from the
1171 	 * userland, make it deprecated.
1172 	 */
1173 	if ((ifra->ifra_flags & IN6_IFF_DEPRECATED) != 0) {
1174 		ia->ia6_lifetime.ia6t_pltime = 0;
1175 		ia->ia6_lifetime.ia6t_preferred = time_uptime;
1176 	}
1177 
1178 	/*
1179 	 * configure address flags.
1180 	 */
1181 	ia->ia6_flags = ifra->ifra_flags;
1182 
1183 	/*
1184 	 * Make the address tentative before joining multicast addresses,
1185 	 * so that corresponding MLD responses would not have a tentative
1186 	 * source address.
1187 	 */
1188 	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/* safety */
1189 
1190 	/*
1191 	 * DAD should be performed for an new address or addresses on
1192 	 * an interface with ND6_IFF_IFDISABLED.
1193 	 */
1194 	if (in6if_do_dad(ifp) &&
1195 	    (hostIsNew || (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)))
1196 		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1197 
1198 	/* notify other subsystems */
1199 	error = in6_notify_ifa(ifp, ia, ifra, hostIsNew);
1200 
1201 	return (error);
1202 }
1203 
1204 /*
1205  * Do link-level ifa job:
1206  * 1) Add lle entry for added address
1207  * 2) Notifies routing socket users about new address
1208  * 3) join appropriate multicast group
1209  * 4) start DAD if enabled
1210  */
1211 static int
1212 in6_broadcast_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
1213     struct in6_ifaddr *ia, int flags)
1214 {
1215 	struct in6_multi *in6m_sol;
1216 	int error = 0;
1217 
1218 	/* Add local address to lltable, if necessary (ex. on p2p link). */
1219 	if ((error = nd6_add_ifa_lle(ia)) != 0) {
1220 		in6_purgeaddr(&ia->ia_ifa);
1221 		ifa_free(&ia->ia_ifa);
1222 		return (error);
1223 	}
1224 
1225 	/* Join necessary multicast groups. */
1226 	in6m_sol = NULL;
1227 	if ((ifp->if_flags & IFF_MULTICAST) != 0) {
1228 		error = in6_update_ifa_join_mc(ifp, ifra, ia, flags, &in6m_sol);
1229 		if (error != 0) {
1230 			in6_purgeaddr(&ia->ia_ifa);
1231 			ifa_free(&ia->ia_ifa);
1232 			return (error);
1233 		}
1234 	}
1235 
1236 	/* Perform DAD, if the address is TENTATIVE. */
1237 	if ((ia->ia6_flags & IN6_IFF_TENTATIVE)) {
1238 		int delay, mindelay, maxdelay;
1239 
1240 		delay = 0;
1241 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1242 			/*
1243 			 * We need to impose a delay before sending an NS
1244 			 * for DAD.  Check if we also needed a delay for the
1245 			 * corresponding MLD message.  If we did, the delay
1246 			 * should be larger than the MLD delay (this could be
1247 			 * relaxed a bit, but this simple logic is at least
1248 			 * safe).
1249 			 * XXX: Break data hiding guidelines and look at
1250 			 * state for the solicited multicast group.
1251 			 */
1252 			mindelay = 0;
1253 			if (in6m_sol != NULL &&
1254 			    in6m_sol->in6m_state == MLD_REPORTING_MEMBER) {
1255 				mindelay = in6m_sol->in6m_timer;
1256 			}
1257 			maxdelay = MAX_RTR_SOLICITATION_DELAY * hz;
1258 			if (maxdelay - mindelay == 0)
1259 				delay = 0;
1260 			else {
1261 				delay =
1262 				    (arc4random() % (maxdelay - mindelay)) +
1263 				    mindelay;
1264 			}
1265 		}
1266 		nd6_dad_start((struct ifaddr *)ia, delay);
1267 	}
1268 
1269 	in6_newaddrmsg(ia, RTM_ADD);
1270 	ifa_free(&ia->ia_ifa);
1271 	return (error);
1272 }
1273 
1274 void
1275 in6_purgeaddr(struct ifaddr *ifa)
1276 {
1277 	struct ifnet *ifp = ifa->ifa_ifp;
1278 	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1279 	struct in6_multi_mship *imm;
1280 	int plen, error;
1281 
1282 	if (ifa->ifa_carp)
1283 		(*carp_detach_p)(ifa, false);
1284 
1285 	/*
1286 	 * Remove the loopback route to the interface address.
1287 	 * The check for the current setting of "nd6_useloopback"
1288 	 * is not needed.
1289 	 */
1290 	if (ia->ia_flags & IFA_RTSELF) {
1291 		error = ifa_del_loopback_route((struct ifaddr *)ia,
1292 		    (struct sockaddr *)&ia->ia_addr);
1293 		if (error == 0)
1294 			ia->ia_flags &= ~IFA_RTSELF;
1295 	}
1296 
1297 	/* stop DAD processing */
1298 	nd6_dad_stop(ifa);
1299 
1300 	/* Leave multicast groups. */
1301 	while ((imm = LIST_FIRST(&ia->ia6_memberships)) != NULL) {
1302 		LIST_REMOVE(imm, i6mm_chain);
1303 		if (imm->i6mm_maddr != NULL)
1304 			in6_leavegroup(imm->i6mm_maddr, NULL);
1305 		free(imm, M_IP6MADDR);
1306 	}
1307 	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1308 	if ((ia->ia_flags & IFA_ROUTE) && plen == 128) {
1309 		error = rtinit(&(ia->ia_ifa), RTM_DELETE, ia->ia_flags |
1310 		    (ia->ia_dstaddr.sin6_family == AF_INET6 ? RTF_HOST : 0));
1311 		if (error != 0)
1312 			log(LOG_INFO, "%s: err=%d, destination address delete "
1313 			    "failed\n", __func__, error);
1314 		ia->ia_flags &= ~IFA_ROUTE;
1315 	}
1316 
1317 	in6_newaddrmsg(ia, RTM_DELETE);
1318 	in6_unlink_ifa(ia, ifp);
1319 }
1320 
1321 static void
1322 in6_unlink_ifa(struct in6_ifaddr *ia, struct ifnet *ifp)
1323 {
1324 	char ip6buf[INET6_ADDRSTRLEN];
1325 	int remove_lle;
1326 
1327 	IF_ADDR_WLOCK(ifp);
1328 	CK_STAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifaddr, ifa_link);
1329 	IF_ADDR_WUNLOCK(ifp);
1330 	ifa_free(&ia->ia_ifa);			/* if_addrhead */
1331 
1332 	/*
1333 	 * Defer the release of what might be the last reference to the
1334 	 * in6_ifaddr so that it can't be freed before the remainder of the
1335 	 * cleanup.
1336 	 */
1337 	IN6_IFADDR_WLOCK();
1338 	CK_STAILQ_REMOVE(&V_in6_ifaddrhead, ia, in6_ifaddr, ia_link);
1339 	CK_LIST_REMOVE(ia, ia6_hash);
1340 	IN6_IFADDR_WUNLOCK();
1341 
1342 	/*
1343 	 * Release the reference to the base prefix.  There should be a
1344 	 * positive reference.
1345 	 */
1346 	remove_lle = 0;
1347 	if (ia->ia6_ndpr == NULL) {
1348 		nd6log((LOG_NOTICE,
1349 		    "in6_unlink_ifa: autoconf'ed address "
1350 		    "%s has no prefix\n", ip6_sprintf(ip6buf, IA6_IN6(ia))));
1351 	} else {
1352 		ia->ia6_ndpr->ndpr_addrcnt--;
1353 		/* Do not delete lles within prefix if refcont != 0 */
1354 		if (ia->ia6_ndpr->ndpr_addrcnt == 0)
1355 			remove_lle = 1;
1356 		ia->ia6_ndpr = NULL;
1357 	}
1358 
1359 	nd6_rem_ifa_lle(ia, remove_lle);
1360 
1361 	/*
1362 	 * Also, if the address being removed is autoconf'ed, call
1363 	 * pfxlist_onlink_check() since the release might affect the status of
1364 	 * other (detached) addresses.
1365 	 */
1366 	if ((ia->ia6_flags & IN6_IFF_AUTOCONF)) {
1367 		pfxlist_onlink_check();
1368 	}
1369 	ifa_free(&ia->ia_ifa);			/* in6_ifaddrhead */
1370 }
1371 
1372 /*
1373  * Notifies other subsystems about address change/arrival:
1374  * 1) Notifies device handler on the first IPv6 address assignment
1375  * 2) Handle routing table changes for P2P links and route
1376  * 3) Handle routing table changes for address host route
1377  */
1378 static int
1379 in6_notify_ifa(struct ifnet *ifp, struct in6_ifaddr *ia,
1380     struct in6_aliasreq *ifra, int hostIsNew)
1381 {
1382 	int	error = 0, plen, ifacount = 0;
1383 	struct ifaddr *ifa;
1384 	struct sockaddr_in6 *pdst;
1385 	char ip6buf[INET6_ADDRSTRLEN];
1386 
1387 	/*
1388 	 * Give the interface a chance to initialize
1389 	 * if this is its first address,
1390 	 */
1391 	if (hostIsNew != 0) {
1392 		struct epoch_tracker et;
1393 
1394 		NET_EPOCH_ENTER(et);
1395 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1396 			if (ifa->ifa_addr->sa_family != AF_INET6)
1397 				continue;
1398 			ifacount++;
1399 		}
1400 		NET_EPOCH_EXIT(et);
1401 	}
1402 
1403 	if (ifacount <= 1 && ifp->if_ioctl) {
1404 		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
1405 		if (error)
1406 			goto done;
1407 	}
1408 
1409 	/*
1410 	 * If a new destination address is specified, scrub the old one and
1411 	 * install the new destination.  Note that the interface must be
1412 	 * p2p or loopback.
1413 	 */
1414 	pdst = &ifra->ifra_dstaddr;
1415 	if (pdst->sin6_family == AF_INET6 &&
1416 	    !IN6_ARE_ADDR_EQUAL(&pdst->sin6_addr, &ia->ia_dstaddr.sin6_addr)) {
1417 		if ((ia->ia_flags & IFA_ROUTE) != 0 &&
1418 		    (rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST) != 0)) {
1419 			nd6log((LOG_ERR, "in6_update_ifa_internal: failed to "
1420 			    "remove a route to the old destination: %s\n",
1421 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1422 			/* proceed anyway... */
1423 		} else
1424 			ia->ia_flags &= ~IFA_ROUTE;
1425 		ia->ia_dstaddr = *pdst;
1426 	}
1427 
1428 	/*
1429 	 * If a new destination address is specified for a point-to-point
1430 	 * interface, install a route to the destination as an interface
1431 	 * direct route.
1432 	 * XXX: the logic below rejects assigning multiple addresses on a p2p
1433 	 * interface that share the same destination.
1434 	 */
1435 	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1436 	if (!(ia->ia_flags & IFA_ROUTE) && plen == 128 &&
1437 	    ia->ia_dstaddr.sin6_family == AF_INET6) {
1438 		int rtflags = RTF_UP | RTF_HOST;
1439 		/*
1440 		 * Handle the case for ::1 .
1441 		 */
1442 		if (ifp->if_flags & IFF_LOOPBACK)
1443 			ia->ia_flags |= IFA_RTSELF;
1444 		error = rtinit(&ia->ia_ifa, RTM_ADD, ia->ia_flags | rtflags);
1445 		if (error)
1446 			goto done;
1447 		ia->ia_flags |= IFA_ROUTE;
1448 	}
1449 
1450 	/*
1451 	 * add a loopback route to self if not exists
1452 	 */
1453 	if (!(ia->ia_flags & IFA_RTSELF) && V_nd6_useloopback) {
1454 		error = ifa_add_loopback_route((struct ifaddr *)ia,
1455 		    (struct sockaddr *)&ia->ia_addr);
1456 		if (error == 0)
1457 			ia->ia_flags |= IFA_RTSELF;
1458 	}
1459 done:
1460 	WITNESS_WARN(WARN_GIANTOK | WARN_SLEEPOK, NULL,
1461 	    "Invoking IPv6 network device address event may sleep");
1462 
1463 	ifa_ref(&ia->ia_ifa);
1464 	EVENTHANDLER_INVOKE(ifaddr_event_ext, ifp, &ia->ia_ifa,
1465 	    IFADDR_EVENT_ADD);
1466 	ifa_free(&ia->ia_ifa);
1467 
1468 	return (error);
1469 }
1470 
1471 /*
1472  * Find an IPv6 interface link-local address specific to an interface.
1473  * ifaddr is returned referenced.
1474  */
1475 struct in6_ifaddr *
1476 in6ifa_ifpforlinklocal(struct ifnet *ifp, int ignoreflags)
1477 {
1478 	struct ifaddr *ifa;
1479 
1480 	NET_EPOCH_ASSERT();
1481 
1482 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1483 		if (ifa->ifa_addr->sa_family != AF_INET6)
1484 			continue;
1485 		if (IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa))) {
1486 			if ((((struct in6_ifaddr *)ifa)->ia6_flags &
1487 			    ignoreflags) != 0)
1488 				continue;
1489 			ifa_ref(ifa);
1490 			break;
1491 		}
1492 	}
1493 
1494 	return ((struct in6_ifaddr *)ifa);
1495 }
1496 
1497 
1498 /*
1499  * find the interface address corresponding to a given IPv6 address.
1500  * ifaddr is returned referenced.
1501  */
1502 struct in6_ifaddr *
1503 in6ifa_ifwithaddr(const struct in6_addr *addr, uint32_t zoneid)
1504 {
1505 	struct rm_priotracker in6_ifa_tracker;
1506 	struct in6_ifaddr *ia;
1507 
1508 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1509 	CK_LIST_FOREACH(ia, IN6ADDR_HASH(addr), ia6_hash) {
1510 		if (IN6_ARE_ADDR_EQUAL(IA6_IN6(ia), addr)) {
1511 			if (zoneid != 0 &&
1512 			    zoneid != ia->ia_addr.sin6_scope_id)
1513 				continue;
1514 			ifa_ref(&ia->ia_ifa);
1515 			break;
1516 		}
1517 	}
1518 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1519 	return (ia);
1520 }
1521 
1522 /*
1523  * find the internet address corresponding to a given interface and address.
1524  * ifaddr is returned referenced.
1525  */
1526 struct in6_ifaddr *
1527 in6ifa_ifpwithaddr(struct ifnet *ifp, const struct in6_addr *addr)
1528 {
1529 	struct epoch_tracker et;
1530 	struct ifaddr *ifa;
1531 
1532 	NET_EPOCH_ENTER(et);
1533 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1534 		if (ifa->ifa_addr->sa_family != AF_INET6)
1535 			continue;
1536 		if (IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa))) {
1537 			ifa_ref(ifa);
1538 			break;
1539 		}
1540 	}
1541 	NET_EPOCH_EXIT(et);
1542 
1543 	return ((struct in6_ifaddr *)ifa);
1544 }
1545 
1546 /*
1547  * Find a link-local scoped address on ifp and return it if any.
1548  */
1549 struct in6_ifaddr *
1550 in6ifa_llaonifp(struct ifnet *ifp)
1551 {
1552 	struct epoch_tracker et;
1553 	struct sockaddr_in6 *sin6;
1554 	struct ifaddr *ifa;
1555 
1556 
1557 	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)
1558 		return (NULL);
1559 	NET_EPOCH_ENTER(et);
1560 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1561 		if (ifa->ifa_addr->sa_family != AF_INET6)
1562 			continue;
1563 		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
1564 		if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr) ||
1565 		    IN6_IS_ADDR_MC_INTFACELOCAL(&sin6->sin6_addr) ||
1566 		    IN6_IS_ADDR_MC_NODELOCAL(&sin6->sin6_addr))
1567 			break;
1568 	}
1569 	NET_EPOCH_EXIT(et);
1570 
1571 	return ((struct in6_ifaddr *)ifa);
1572 }
1573 
1574 /*
1575  * Convert IP6 address to printable (loggable) representation. Caller
1576  * has to make sure that ip6buf is at least INET6_ADDRSTRLEN long.
1577  */
1578 static char digits[] = "0123456789abcdef";
1579 char *
1580 ip6_sprintf(char *ip6buf, const struct in6_addr *addr)
1581 {
1582 	int i, cnt = 0, maxcnt = 0, idx = 0, index = 0;
1583 	char *cp;
1584 	const u_int16_t *a = (const u_int16_t *)addr;
1585 	const u_int8_t *d;
1586 	int dcolon = 0, zero = 0;
1587 
1588 	cp = ip6buf;
1589 
1590 	for (i = 0; i < 8; i++) {
1591 		if (*(a + i) == 0) {
1592 			cnt++;
1593 			if (cnt == 1)
1594 				idx = i;
1595 		}
1596 		else if (maxcnt < cnt) {
1597 			maxcnt = cnt;
1598 			index = idx;
1599 			cnt = 0;
1600 		}
1601 	}
1602 	if (maxcnt < cnt) {
1603 		maxcnt = cnt;
1604 		index = idx;
1605 	}
1606 
1607 	for (i = 0; i < 8; i++) {
1608 		if (dcolon == 1) {
1609 			if (*a == 0) {
1610 				if (i == 7)
1611 					*cp++ = ':';
1612 				a++;
1613 				continue;
1614 			} else
1615 				dcolon = 2;
1616 		}
1617 		if (*a == 0) {
1618 			if (dcolon == 0 && *(a + 1) == 0 && i == index) {
1619 				if (i == 0)
1620 					*cp++ = ':';
1621 				*cp++ = ':';
1622 				dcolon = 1;
1623 			} else {
1624 				*cp++ = '0';
1625 				*cp++ = ':';
1626 			}
1627 			a++;
1628 			continue;
1629 		}
1630 		d = (const u_char *)a;
1631 		/* Try to eliminate leading zeros in printout like in :0001. */
1632 		zero = 1;
1633 		*cp = digits[*d >> 4];
1634 		if (*cp != '0') {
1635 			zero = 0;
1636 			cp++;
1637 		}
1638 		*cp = digits[*d++ & 0xf];
1639 		if (zero == 0 || (*cp != '0')) {
1640 			zero = 0;
1641 			cp++;
1642 		}
1643 		*cp = digits[*d >> 4];
1644 		if (zero == 0 || (*cp != '0')) {
1645 			zero = 0;
1646 			cp++;
1647 		}
1648 		*cp++ = digits[*d & 0xf];
1649 		*cp++ = ':';
1650 		a++;
1651 	}
1652 	*--cp = '\0';
1653 	return (ip6buf);
1654 }
1655 
1656 int
1657 in6_localaddr(struct in6_addr *in6)
1658 {
1659 	struct rm_priotracker in6_ifa_tracker;
1660 	struct in6_ifaddr *ia;
1661 
1662 	if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
1663 		return 1;
1664 
1665 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1666 	CK_STAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
1667 		if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
1668 		    &ia->ia_prefixmask.sin6_addr)) {
1669 			IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1670 			return 1;
1671 		}
1672 	}
1673 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1674 
1675 	return (0);
1676 }
1677 
1678 /*
1679  * Return 1 if an internet address is for the local host and configured
1680  * on one of its interfaces.
1681  */
1682 int
1683 in6_localip(struct in6_addr *in6)
1684 {
1685 	struct rm_priotracker in6_ifa_tracker;
1686 	struct in6_ifaddr *ia;
1687 
1688 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1689 	CK_LIST_FOREACH(ia, IN6ADDR_HASH(in6), ia6_hash) {
1690 		if (IN6_ARE_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr)) {
1691 			IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1692 			return (1);
1693 		}
1694 	}
1695 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1696 	return (0);
1697 }
1698 
1699 /*
1700  * Return 1 if an internet address is configured on an interface.
1701  */
1702 int
1703 in6_ifhasaddr(struct ifnet *ifp, struct in6_addr *addr)
1704 {
1705 	struct in6_addr in6;
1706 	struct ifaddr *ifa;
1707 	struct in6_ifaddr *ia6;
1708 
1709 	NET_EPOCH_ASSERT();
1710 
1711 	in6 = *addr;
1712 	if (in6_clearscope(&in6))
1713 		return (0);
1714 	in6_setscope(&in6, ifp, NULL);
1715 
1716 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1717 		if (ifa->ifa_addr->sa_family != AF_INET6)
1718 			continue;
1719 		ia6 = (struct in6_ifaddr *)ifa;
1720 		if (IN6_ARE_ADDR_EQUAL(&ia6->ia_addr.sin6_addr, &in6))
1721 			return (1);
1722 	}
1723 
1724 	return (0);
1725 }
1726 
1727 int
1728 in6_is_addr_deprecated(struct sockaddr_in6 *sa6)
1729 {
1730 	struct rm_priotracker in6_ifa_tracker;
1731 	struct in6_ifaddr *ia;
1732 
1733 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1734 	CK_LIST_FOREACH(ia, IN6ADDR_HASH(&sa6->sin6_addr), ia6_hash) {
1735 		if (IN6_ARE_ADDR_EQUAL(IA6_IN6(ia), &sa6->sin6_addr)) {
1736 			if (ia->ia6_flags & IN6_IFF_DEPRECATED) {
1737 				IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1738 				return (1); /* true */
1739 			}
1740 			break;
1741 		}
1742 	}
1743 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1744 
1745 	return (0);		/* false */
1746 }
1747 
1748 /*
1749  * return length of part which dst and src are equal
1750  * hard coding...
1751  */
1752 int
1753 in6_matchlen(struct in6_addr *src, struct in6_addr *dst)
1754 {
1755 	int match = 0;
1756 	u_char *s = (u_char *)src, *d = (u_char *)dst;
1757 	u_char *lim = s + 16, r;
1758 
1759 	while (s < lim)
1760 		if ((r = (*d++ ^ *s++)) != 0) {
1761 			while (r < 128) {
1762 				match++;
1763 				r <<= 1;
1764 			}
1765 			break;
1766 		} else
1767 			match += 8;
1768 	return match;
1769 }
1770 
1771 /* XXX: to be scope conscious */
1772 int
1773 in6_are_prefix_equal(struct in6_addr *p1, struct in6_addr *p2, int len)
1774 {
1775 	int bytelen, bitlen;
1776 
1777 	/* sanity check */
1778 	if (0 > len || len > 128) {
1779 		log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n",
1780 		    len);
1781 		return (0);
1782 	}
1783 
1784 	bytelen = len / 8;
1785 	bitlen = len % 8;
1786 
1787 	if (bcmp(&p1->s6_addr, &p2->s6_addr, bytelen))
1788 		return (0);
1789 	if (bitlen != 0 &&
1790 	    p1->s6_addr[bytelen] >> (8 - bitlen) !=
1791 	    p2->s6_addr[bytelen] >> (8 - bitlen))
1792 		return (0);
1793 
1794 	return (1);
1795 }
1796 
1797 void
1798 in6_prefixlen2mask(struct in6_addr *maskp, int len)
1799 {
1800 	u_char maskarray[8] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
1801 	int bytelen, bitlen, i;
1802 
1803 	/* sanity check */
1804 	if (0 > len || len > 128) {
1805 		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
1806 		    len);
1807 		return;
1808 	}
1809 
1810 	bzero(maskp, sizeof(*maskp));
1811 	bytelen = len / 8;
1812 	bitlen = len % 8;
1813 	for (i = 0; i < bytelen; i++)
1814 		maskp->s6_addr[i] = 0xff;
1815 	if (bitlen)
1816 		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
1817 }
1818 
1819 /*
1820  * return the best address out of the same scope. if no address was
1821  * found, return the first valid address from designated IF.
1822  */
1823 struct in6_ifaddr *
1824 in6_ifawithifp(struct ifnet *ifp, struct in6_addr *dst)
1825 {
1826 	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
1827 	struct ifaddr *ifa;
1828 	struct in6_ifaddr *besta = NULL;
1829 	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
1830 
1831 	NET_EPOCH_ASSERT();
1832 
1833 	dep[0] = dep[1] = NULL;
1834 
1835 	/*
1836 	 * We first look for addresses in the same scope.
1837 	 * If there is one, return it.
1838 	 * If two or more, return one which matches the dst longest.
1839 	 * If none, return one of global addresses assigned other ifs.
1840 	 */
1841 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1842 		if (ifa->ifa_addr->sa_family != AF_INET6)
1843 			continue;
1844 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
1845 			continue; /* XXX: is there any case to allow anycast? */
1846 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
1847 			continue; /* don't use this interface */
1848 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
1849 			continue;
1850 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
1851 			if (V_ip6_use_deprecated)
1852 				dep[0] = (struct in6_ifaddr *)ifa;
1853 			continue;
1854 		}
1855 
1856 		if (dst_scope == in6_addrscope(IFA_IN6(ifa))) {
1857 			/*
1858 			 * call in6_matchlen() as few as possible
1859 			 */
1860 			if (besta) {
1861 				if (blen == -1)
1862 					blen = in6_matchlen(&besta->ia_addr.sin6_addr, dst);
1863 				tlen = in6_matchlen(IFA_IN6(ifa), dst);
1864 				if (tlen > blen) {
1865 					blen = tlen;
1866 					besta = (struct in6_ifaddr *)ifa;
1867 				}
1868 			} else
1869 				besta = (struct in6_ifaddr *)ifa;
1870 		}
1871 	}
1872 	if (besta) {
1873 		ifa_ref(&besta->ia_ifa);
1874 		return (besta);
1875 	}
1876 
1877 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1878 		if (ifa->ifa_addr->sa_family != AF_INET6)
1879 			continue;
1880 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
1881 			continue; /* XXX: is there any case to allow anycast? */
1882 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
1883 			continue; /* don't use this interface */
1884 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
1885 			continue;
1886 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
1887 			if (V_ip6_use_deprecated)
1888 				dep[1] = (struct in6_ifaddr *)ifa;
1889 			continue;
1890 		}
1891 
1892 		if (ifa != NULL)
1893 			ifa_ref(ifa);
1894 		return (struct in6_ifaddr *)ifa;
1895 	}
1896 
1897 	/* use the last-resort values, that are, deprecated addresses */
1898 	if (dep[0]) {
1899 		ifa_ref((struct ifaddr *)dep[0]);
1900 		return dep[0];
1901 	}
1902 	if (dep[1]) {
1903 		ifa_ref((struct ifaddr *)dep[1]);
1904 		return dep[1];
1905 	}
1906 
1907 	return NULL;
1908 }
1909 
1910 /*
1911  * perform DAD when interface becomes IFF_UP.
1912  */
1913 void
1914 in6_if_up(struct ifnet *ifp)
1915 {
1916 	struct epoch_tracker et;
1917 	struct ifaddr *ifa;
1918 	struct in6_ifaddr *ia;
1919 
1920 	NET_EPOCH_ENTER(et);
1921 	CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1922 		if (ifa->ifa_addr->sa_family != AF_INET6)
1923 			continue;
1924 		ia = (struct in6_ifaddr *)ifa;
1925 		if (ia->ia6_flags & IN6_IFF_TENTATIVE) {
1926 			/*
1927 			 * The TENTATIVE flag was likely set by hand
1928 			 * beforehand, implicitly indicating the need for DAD.
1929 			 * We may be able to skip the random delay in this
1930 			 * case, but we impose delays just in case.
1931 			 */
1932 			nd6_dad_start(ifa,
1933 			    arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz));
1934 		}
1935 	}
1936 	NET_EPOCH_EXIT(et);
1937 
1938 	/*
1939 	 * special cases, like 6to4, are handled in in6_ifattach
1940 	 */
1941 	in6_ifattach(ifp, NULL);
1942 }
1943 
1944 int
1945 in6if_do_dad(struct ifnet *ifp)
1946 {
1947 
1948 	if ((ifp->if_flags & IFF_LOOPBACK) != 0)
1949 		return (0);
1950 	if ((ifp->if_flags & IFF_MULTICAST) == 0)
1951 		return (0);
1952 	if ((ND_IFINFO(ifp)->flags &
1953 	    (ND6_IFF_IFDISABLED | ND6_IFF_NO_DAD)) != 0)
1954 		return (0);
1955 	return (1);
1956 }
1957 
1958 /*
1959  * Calculate max IPv6 MTU through all the interfaces and store it
1960  * to in6_maxmtu.
1961  */
1962 void
1963 in6_setmaxmtu(void)
1964 {
1965 	struct epoch_tracker et;
1966 	unsigned long maxmtu = 0;
1967 	struct ifnet *ifp;
1968 
1969 	NET_EPOCH_ENTER(et);
1970 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1971 		/* this function can be called during ifnet initialization */
1972 		if (!ifp->if_afdata[AF_INET6])
1973 			continue;
1974 		if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
1975 		    IN6_LINKMTU(ifp) > maxmtu)
1976 			maxmtu = IN6_LINKMTU(ifp);
1977 	}
1978 	NET_EPOCH_EXIT(et);
1979 	if (maxmtu)	/* update only when maxmtu is positive */
1980 		V_in6_maxmtu = maxmtu;
1981 }
1982 
1983 /*
1984  * Provide the length of interface identifiers to be used for the link attached
1985  * to the given interface.  The length should be defined in "IPv6 over
1986  * xxx-link" document.  Note that address architecture might also define
1987  * the length for a particular set of address prefixes, regardless of the
1988  * link type.  As clarified in rfc2462bis, those two definitions should be
1989  * consistent, and those really are as of August 2004.
1990  */
1991 int
1992 in6_if2idlen(struct ifnet *ifp)
1993 {
1994 	switch (ifp->if_type) {
1995 	case IFT_ETHER:		/* RFC2464 */
1996 	case IFT_PROPVIRTUAL:	/* XXX: no RFC. treat it as ether */
1997 	case IFT_L2VLAN:	/* ditto */
1998 	case IFT_BRIDGE:	/* bridge(4) only does Ethernet-like links */
1999 	case IFT_INFINIBAND:
2000 		return (64);
2001 	case IFT_PPP:		/* RFC2472 */
2002 		return (64);
2003 	case IFT_FRELAY:	/* RFC2590 */
2004 		return (64);
2005 	case IFT_IEEE1394:	/* RFC3146 */
2006 		return (64);
2007 	case IFT_GIF:
2008 		return (64);	/* draft-ietf-v6ops-mech-v2-07 */
2009 	case IFT_LOOP:
2010 		return (64);	/* XXX: is this really correct? */
2011 	default:
2012 		/*
2013 		 * Unknown link type:
2014 		 * It might be controversial to use the today's common constant
2015 		 * of 64 for these cases unconditionally.  For full compliance,
2016 		 * we should return an error in this case.  On the other hand,
2017 		 * if we simply miss the standard for the link type or a new
2018 		 * standard is defined for a new link type, the IFID length
2019 		 * is very likely to be the common constant.  As a compromise,
2020 		 * we always use the constant, but make an explicit notice
2021 		 * indicating the "unknown" case.
2022 		 */
2023 		printf("in6_if2idlen: unknown link type (%d)\n", ifp->if_type);
2024 		return (64);
2025 	}
2026 }
2027 
2028 struct in6_llentry {
2029 	struct llentry		base;
2030 };
2031 
2032 #define	IN6_LLTBL_DEFAULT_HSIZE	32
2033 #define	IN6_LLTBL_HASH(k, h) \
2034 	(((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1))
2035 
2036 /*
2037  * Do actual deallocation of @lle.
2038  */
2039 static void
2040 in6_lltable_destroy_lle_unlocked(epoch_context_t ctx)
2041 {
2042 	struct llentry *lle;
2043 
2044 	lle = __containerof(ctx, struct llentry, lle_epoch_ctx);
2045 	LLE_LOCK_DESTROY(lle);
2046 	LLE_REQ_DESTROY(lle);
2047 	free(lle, M_LLTABLE);
2048 }
2049 
2050 /*
2051  * Called by LLE_FREE_LOCKED when number of references
2052  * drops to zero.
2053  */
2054 static void
2055 in6_lltable_destroy_lle(struct llentry *lle)
2056 {
2057 
2058 	LLE_WUNLOCK(lle);
2059 	epoch_call(net_epoch_preempt,  &lle->lle_epoch_ctx, in6_lltable_destroy_lle_unlocked);
2060 }
2061 
2062 static struct llentry *
2063 in6_lltable_new(const struct in6_addr *addr6, u_int flags)
2064 {
2065 	struct in6_llentry *lle;
2066 
2067 	lle = malloc(sizeof(struct in6_llentry), M_LLTABLE, M_NOWAIT | M_ZERO);
2068 	if (lle == NULL)		/* NB: caller generates msg */
2069 		return NULL;
2070 
2071 	lle->base.r_l3addr.addr6 = *addr6;
2072 	lle->base.lle_refcnt = 1;
2073 	lle->base.lle_free = in6_lltable_destroy_lle;
2074 	LLE_LOCK_INIT(&lle->base);
2075 	LLE_REQ_INIT(&lle->base);
2076 	callout_init(&lle->base.lle_timer, 1);
2077 
2078 	return (&lle->base);
2079 }
2080 
2081 static int
2082 in6_lltable_match_prefix(const struct sockaddr *saddr,
2083     const struct sockaddr *smask, u_int flags, struct llentry *lle)
2084 {
2085 	const struct in6_addr *addr, *mask, *lle_addr;
2086 
2087 	addr = &((const struct sockaddr_in6 *)saddr)->sin6_addr;
2088 	mask = &((const struct sockaddr_in6 *)smask)->sin6_addr;
2089 	lle_addr = &lle->r_l3addr.addr6;
2090 
2091 	if (IN6_ARE_MASKED_ADDR_EQUAL(lle_addr, addr, mask) == 0)
2092 		return (0);
2093 
2094 	if (lle->la_flags & LLE_IFADDR) {
2095 
2096 		/*
2097 		 * Delete LLE_IFADDR records IFF address & flag matches.
2098 		 * Note that addr is the interface address within prefix
2099 		 * being matched.
2100 		 */
2101 		if (IN6_ARE_ADDR_EQUAL(addr, lle_addr) &&
2102 		    (flags & LLE_STATIC) != 0)
2103 			return (1);
2104 		return (0);
2105 	}
2106 
2107 	/* flags & LLE_STATIC means deleting both dynamic and static entries */
2108 	if ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC))
2109 		return (1);
2110 
2111 	return (0);
2112 }
2113 
2114 static void
2115 in6_lltable_free_entry(struct lltable *llt, struct llentry *lle)
2116 {
2117 	struct ifnet *ifp;
2118 
2119 	LLE_WLOCK_ASSERT(lle);
2120 	KASSERT(llt != NULL, ("lltable is NULL"));
2121 
2122 	/* Unlink entry from table */
2123 	if ((lle->la_flags & LLE_LINKED) != 0) {
2124 
2125 		ifp = llt->llt_ifp;
2126 		IF_AFDATA_WLOCK_ASSERT(ifp);
2127 		lltable_unlink_entry(llt, lle);
2128 	}
2129 
2130 	llentry_free(lle);
2131 }
2132 
2133 static int
2134 in6_lltable_rtcheck(struct ifnet *ifp,
2135 		    u_int flags,
2136 		    const struct sockaddr *l3addr)
2137 {
2138 	const struct sockaddr_in6 *sin6;
2139 	struct nhop6_basic nh6;
2140 	struct in6_addr dst;
2141 	uint32_t scopeid;
2142 	int error;
2143 	char ip6buf[INET6_ADDRSTRLEN];
2144 	int fibnum;
2145 
2146 	NET_EPOCH_ASSERT();
2147 	KASSERT(l3addr->sa_family == AF_INET6,
2148 	    ("sin_family %d", l3addr->sa_family));
2149 
2150 	sin6 = (const struct sockaddr_in6 *)l3addr;
2151 	in6_splitscope(&sin6->sin6_addr, &dst, &scopeid);
2152 	fibnum = V_rt_add_addr_allfibs ? RT_DEFAULT_FIB : ifp->if_fib;
2153 	error = fib6_lookup_nh_basic(fibnum, &dst, scopeid, 0, 0, &nh6);
2154 	if (error != 0 || (nh6.nh_flags & NHF_GATEWAY) || nh6.nh_ifp != ifp) {
2155 		struct ifaddr *ifa;
2156 		/*
2157 		 * Create an ND6 cache for an IPv6 neighbor
2158 		 * that is not covered by our own prefix.
2159 		 */
2160 		ifa = ifaof_ifpforaddr(l3addr, ifp);
2161 		if (ifa != NULL) {
2162 			return 0;
2163 		}
2164 		log(LOG_INFO, "IPv6 address: \"%s\" is not on the network\n",
2165 		    ip6_sprintf(ip6buf, &sin6->sin6_addr));
2166 		return EINVAL;
2167 	}
2168 	return 0;
2169 }
2170 
2171 /*
2172  * Called by the datapath to indicate that the entry was used.
2173  */
2174 static void
2175 in6_lltable_mark_used(struct llentry *lle)
2176 {
2177 
2178 	LLE_REQ_LOCK(lle);
2179 	lle->r_skip_req = 0;
2180 
2181 	/*
2182 	 * Set the hit time so the callback function
2183 	 * can determine the remaining time before
2184 	 * transiting to the DELAY state.
2185 	 */
2186 	lle->lle_hittime = time_uptime;
2187 	LLE_REQ_UNLOCK(lle);
2188 }
2189 
2190 static inline uint32_t
2191 in6_lltable_hash_dst(const struct in6_addr *dst, uint32_t hsize)
2192 {
2193 
2194 	return (IN6_LLTBL_HASH(dst->s6_addr32[3], hsize));
2195 }
2196 
2197 static uint32_t
2198 in6_lltable_hash(const struct llentry *lle, uint32_t hsize)
2199 {
2200 
2201 	return (in6_lltable_hash_dst(&lle->r_l3addr.addr6, hsize));
2202 }
2203 
2204 static void
2205 in6_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
2206 {
2207 	struct sockaddr_in6 *sin6;
2208 
2209 	sin6 = (struct sockaddr_in6 *)sa;
2210 	bzero(sin6, sizeof(*sin6));
2211 	sin6->sin6_family = AF_INET6;
2212 	sin6->sin6_len = sizeof(*sin6);
2213 	sin6->sin6_addr = lle->r_l3addr.addr6;
2214 }
2215 
2216 static inline struct llentry *
2217 in6_lltable_find_dst(struct lltable *llt, const struct in6_addr *dst)
2218 {
2219 	struct llentry *lle;
2220 	struct llentries *lleh;
2221 	u_int hashidx;
2222 
2223 	hashidx = in6_lltable_hash_dst(dst, llt->llt_hsize);
2224 	lleh = &llt->lle_head[hashidx];
2225 	CK_LIST_FOREACH(lle, lleh, lle_next) {
2226 		if (lle->la_flags & LLE_DELETED)
2227 			continue;
2228 		if (IN6_ARE_ADDR_EQUAL(&lle->r_l3addr.addr6, dst))
2229 			break;
2230 	}
2231 
2232 	return (lle);
2233 }
2234 
2235 static void
2236 in6_lltable_delete_entry(struct lltable *llt, struct llentry *lle)
2237 {
2238 
2239 	lle->la_flags |= LLE_DELETED;
2240 	EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_DELETED);
2241 #ifdef DIAGNOSTIC
2242 	log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle);
2243 #endif
2244 	llentry_free(lle);
2245 }
2246 
2247 static struct llentry *
2248 in6_lltable_alloc(struct lltable *llt, u_int flags,
2249 	const struct sockaddr *l3addr)
2250 {
2251 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2252 	struct ifnet *ifp = llt->llt_ifp;
2253 	struct llentry *lle;
2254 	char linkhdr[LLE_MAX_LINKHDR];
2255 	size_t linkhdrsize;
2256 	int lladdr_off;
2257 
2258 	KASSERT(l3addr->sa_family == AF_INET6,
2259 	    ("sin_family %d", l3addr->sa_family));
2260 
2261 	/*
2262 	 * A route that covers the given address must have
2263 	 * been installed 1st because we are doing a resolution,
2264 	 * verify this.
2265 	 */
2266 	if (!(flags & LLE_IFADDR) &&
2267 	    in6_lltable_rtcheck(ifp, flags, l3addr) != 0)
2268 		return (NULL);
2269 
2270 	lle = in6_lltable_new(&sin6->sin6_addr, flags);
2271 	if (lle == NULL) {
2272 		log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
2273 		return (NULL);
2274 	}
2275 	lle->la_flags = flags;
2276 	if ((flags & LLE_IFADDR) == LLE_IFADDR) {
2277 		linkhdrsize = LLE_MAX_LINKHDR;
2278 		if (lltable_calc_llheader(ifp, AF_INET6, IF_LLADDR(ifp),
2279 		    linkhdr, &linkhdrsize, &lladdr_off) != 0) {
2280 			epoch_call(net_epoch_preempt,  &lle->lle_epoch_ctx, in6_lltable_destroy_lle_unlocked);
2281 			return (NULL);
2282 		}
2283 		lltable_set_entry_addr(ifp, lle, linkhdr, linkhdrsize,
2284 		    lladdr_off);
2285 		lle->la_flags |= LLE_STATIC;
2286 	}
2287 
2288 	if ((lle->la_flags & LLE_STATIC) != 0)
2289 		lle->ln_state = ND6_LLINFO_REACHABLE;
2290 
2291 	return (lle);
2292 }
2293 
2294 static struct llentry *
2295 in6_lltable_lookup(struct lltable *llt, u_int flags,
2296 	const struct sockaddr *l3addr)
2297 {
2298 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2299 	struct llentry *lle;
2300 
2301 	IF_AFDATA_LOCK_ASSERT(llt->llt_ifp);
2302 	KASSERT(l3addr->sa_family == AF_INET6,
2303 	    ("sin_family %d", l3addr->sa_family));
2304 	KASSERT((flags & (LLE_UNLOCKED | LLE_EXCLUSIVE)) !=
2305 	    (LLE_UNLOCKED | LLE_EXCLUSIVE),
2306 	    ("wrong lle request flags: %#x", flags));
2307 
2308 	lle = in6_lltable_find_dst(llt, &sin6->sin6_addr);
2309 	if (lle == NULL)
2310 		return (NULL);
2311 	if (flags & LLE_UNLOCKED)
2312 		return (lle);
2313 
2314 	if (flags & LLE_EXCLUSIVE)
2315 		LLE_WLOCK(lle);
2316 	else
2317 		LLE_RLOCK(lle);
2318 
2319 	/*
2320 	 * If the afdata lock is not held, the LLE may have been unlinked while
2321 	 * we were blocked on the LLE lock.  Check for this case.
2322 	 */
2323 	if (__predict_false((lle->la_flags & LLE_LINKED) == 0)) {
2324 		if (flags & LLE_EXCLUSIVE)
2325 			LLE_WUNLOCK(lle);
2326 		else
2327 			LLE_RUNLOCK(lle);
2328 		return (NULL);
2329 	}
2330 	return (lle);
2331 }
2332 
2333 static int
2334 in6_lltable_dump_entry(struct lltable *llt, struct llentry *lle,
2335     struct sysctl_req *wr)
2336 {
2337 	struct ifnet *ifp = llt->llt_ifp;
2338 	/* XXX stack use */
2339 	struct {
2340 		struct rt_msghdr	rtm;
2341 		struct sockaddr_in6	sin6;
2342 		/*
2343 		 * ndp.c assumes that sdl is word aligned
2344 		 */
2345 #ifdef __LP64__
2346 		uint32_t		pad;
2347 #endif
2348 		struct sockaddr_dl	sdl;
2349 	} ndpc;
2350 	struct sockaddr_dl *sdl;
2351 	int error;
2352 
2353 	bzero(&ndpc, sizeof(ndpc));
2354 	/* skip deleted entries */
2355 	if ((lle->la_flags & LLE_DELETED) == LLE_DELETED)
2356 		return (0);
2357 	/* Skip if jailed and not a valid IP of the prison. */
2358 	lltable_fill_sa_entry(lle, (struct sockaddr *)&ndpc.sin6);
2359 	if (prison_if(wr->td->td_ucred, (struct sockaddr *)&ndpc.sin6) != 0)
2360 		return (0);
2361 	/*
2362 	 * produce a msg made of:
2363 	 *  struct rt_msghdr;
2364 	 *  struct sockaddr_in6 (IPv6)
2365 	 *  struct sockaddr_dl;
2366 	 */
2367 	ndpc.rtm.rtm_msglen = sizeof(ndpc);
2368 	ndpc.rtm.rtm_version = RTM_VERSION;
2369 	ndpc.rtm.rtm_type = RTM_GET;
2370 	ndpc.rtm.rtm_flags = RTF_UP;
2371 	ndpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
2372 	if (V_deembed_scopeid)
2373 		sa6_recoverscope(&ndpc.sin6);
2374 
2375 	/* publish */
2376 	if (lle->la_flags & LLE_PUB)
2377 		ndpc.rtm.rtm_flags |= RTF_ANNOUNCE;
2378 
2379 	sdl = &ndpc.sdl;
2380 	sdl->sdl_family = AF_LINK;
2381 	sdl->sdl_len = sizeof(*sdl);
2382 	sdl->sdl_index = ifp->if_index;
2383 	sdl->sdl_type = ifp->if_type;
2384 	if ((lle->la_flags & LLE_VALID) == LLE_VALID) {
2385 		sdl->sdl_alen = ifp->if_addrlen;
2386 		bcopy(lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
2387 	} else {
2388 		sdl->sdl_alen = 0;
2389 		bzero(LLADDR(sdl), ifp->if_addrlen);
2390 	}
2391 	if (lle->la_expire != 0)
2392 		ndpc.rtm.rtm_rmx.rmx_expire = lle->la_expire +
2393 		    lle->lle_remtime / hz + time_second - time_uptime;
2394 	ndpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
2395 	if (lle->la_flags & LLE_STATIC)
2396 		ndpc.rtm.rtm_flags |= RTF_STATIC;
2397 	if (lle->la_flags & LLE_IFADDR)
2398 		ndpc.rtm.rtm_flags |= RTF_PINNED;
2399 	if (lle->ln_router != 0)
2400 		ndpc.rtm.rtm_flags |= RTF_GATEWAY;
2401 	ndpc.rtm.rtm_rmx.rmx_pksent = lle->la_asked;
2402 	/* Store state in rmx_weight value */
2403 	ndpc.rtm.rtm_rmx.rmx_state = lle->ln_state;
2404 	ndpc.rtm.rtm_index = ifp->if_index;
2405 	error = SYSCTL_OUT(wr, &ndpc, sizeof(ndpc));
2406 
2407 	return (error);
2408 }
2409 
2410 static struct lltable *
2411 in6_lltattach(struct ifnet *ifp)
2412 {
2413 	struct lltable *llt;
2414 
2415 	llt = lltable_allocate_htbl(IN6_LLTBL_DEFAULT_HSIZE);
2416 	llt->llt_af = AF_INET6;
2417 	llt->llt_ifp = ifp;
2418 
2419 	llt->llt_lookup = in6_lltable_lookup;
2420 	llt->llt_alloc_entry = in6_lltable_alloc;
2421 	llt->llt_delete_entry = in6_lltable_delete_entry;
2422 	llt->llt_dump_entry = in6_lltable_dump_entry;
2423 	llt->llt_hash = in6_lltable_hash;
2424 	llt->llt_fill_sa_entry = in6_lltable_fill_sa_entry;
2425 	llt->llt_free_entry = in6_lltable_free_entry;
2426 	llt->llt_match_prefix = in6_lltable_match_prefix;
2427 	llt->llt_mark_used = in6_lltable_mark_used;
2428  	lltable_link(llt);
2429 
2430 	return (llt);
2431 }
2432 
2433 void *
2434 in6_domifattach(struct ifnet *ifp)
2435 {
2436 	struct in6_ifextra *ext;
2437 
2438 	/* There are not IPv6-capable interfaces. */
2439 	switch (ifp->if_type) {
2440 	case IFT_PFLOG:
2441 	case IFT_PFSYNC:
2442 	case IFT_USB:
2443 		return (NULL);
2444 	}
2445 	ext = (struct in6_ifextra *)malloc(sizeof(*ext), M_IFADDR, M_WAITOK);
2446 	bzero(ext, sizeof(*ext));
2447 
2448 	ext->in6_ifstat = malloc(sizeof(counter_u64_t) *
2449 	    sizeof(struct in6_ifstat) / sizeof(uint64_t), M_IFADDR, M_WAITOK);
2450 	COUNTER_ARRAY_ALLOC(ext->in6_ifstat,
2451 	    sizeof(struct in6_ifstat) / sizeof(uint64_t), M_WAITOK);
2452 
2453 	ext->icmp6_ifstat = malloc(sizeof(counter_u64_t) *
2454 	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t), M_IFADDR,
2455 	    M_WAITOK);
2456 	COUNTER_ARRAY_ALLOC(ext->icmp6_ifstat,
2457 	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t), M_WAITOK);
2458 
2459 	ext->nd_ifinfo = nd6_ifattach(ifp);
2460 	ext->scope6_id = scope6_ifattach(ifp);
2461 	ext->lltable = in6_lltattach(ifp);
2462 
2463 	ext->mld_ifinfo = mld_domifattach(ifp);
2464 
2465 	return ext;
2466 }
2467 
2468 int
2469 in6_domifmtu(struct ifnet *ifp)
2470 {
2471 	if (ifp->if_afdata[AF_INET6] == NULL)
2472 		return ifp->if_mtu;
2473 
2474 	return (IN6_LINKMTU(ifp));
2475 }
2476 
2477 void
2478 in6_domifdetach(struct ifnet *ifp, void *aux)
2479 {
2480 	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
2481 
2482 	mld_domifdetach(ifp);
2483 	scope6_ifdetach(ext->scope6_id);
2484 	nd6_ifdetach(ifp, ext->nd_ifinfo);
2485 	lltable_free(ext->lltable);
2486 	COUNTER_ARRAY_FREE(ext->in6_ifstat,
2487 	    sizeof(struct in6_ifstat) / sizeof(uint64_t));
2488 	free(ext->in6_ifstat, M_IFADDR);
2489 	COUNTER_ARRAY_FREE(ext->icmp6_ifstat,
2490 	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t));
2491 	free(ext->icmp6_ifstat, M_IFADDR);
2492 	free(ext, M_IFADDR);
2493 }
2494 
2495 /*
2496  * Convert sockaddr_in6 to sockaddr_in.  Original sockaddr_in6 must be
2497  * v4 mapped addr or v4 compat addr
2498  */
2499 void
2500 in6_sin6_2_sin(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2501 {
2502 
2503 	bzero(sin, sizeof(*sin));
2504 	sin->sin_len = sizeof(struct sockaddr_in);
2505 	sin->sin_family = AF_INET;
2506 	sin->sin_port = sin6->sin6_port;
2507 	sin->sin_addr.s_addr = sin6->sin6_addr.s6_addr32[3];
2508 }
2509 
2510 /* Convert sockaddr_in to sockaddr_in6 in v4 mapped addr format. */
2511 void
2512 in6_sin_2_v4mapsin6(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2513 {
2514 	bzero(sin6, sizeof(*sin6));
2515 	sin6->sin6_len = sizeof(struct sockaddr_in6);
2516 	sin6->sin6_family = AF_INET6;
2517 	sin6->sin6_port = sin->sin_port;
2518 	sin6->sin6_addr.s6_addr32[0] = 0;
2519 	sin6->sin6_addr.s6_addr32[1] = 0;
2520 	sin6->sin6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
2521 	sin6->sin6_addr.s6_addr32[3] = sin->sin_addr.s_addr;
2522 }
2523 
2524 /* Convert sockaddr_in6 into sockaddr_in. */
2525 void
2526 in6_sin6_2_sin_in_sock(struct sockaddr *nam)
2527 {
2528 	struct sockaddr_in *sin_p;
2529 	struct sockaddr_in6 sin6;
2530 
2531 	/*
2532 	 * Save original sockaddr_in6 addr and convert it
2533 	 * to sockaddr_in.
2534 	 */
2535 	sin6 = *(struct sockaddr_in6 *)nam;
2536 	sin_p = (struct sockaddr_in *)nam;
2537 	in6_sin6_2_sin(sin_p, &sin6);
2538 }
2539 
2540 /* Convert sockaddr_in into sockaddr_in6 in v4 mapped addr format. */
2541 void
2542 in6_sin_2_v4mapsin6_in_sock(struct sockaddr **nam)
2543 {
2544 	struct sockaddr_in *sin_p;
2545 	struct sockaddr_in6 *sin6_p;
2546 
2547 	sin6_p = malloc(sizeof *sin6_p, M_SONAME, M_WAITOK);
2548 	sin_p = (struct sockaddr_in *)*nam;
2549 	in6_sin_2_v4mapsin6(sin_p, sin6_p);
2550 	free(*nam, M_SONAME);
2551 	*nam = (struct sockaddr *)sin6_p;
2552 }
2553