xref: /freebsd/sys/netinet6/in6.c (revision 079171874c9bf263b69e3af10784ad2bcd1fe699)
1 /*-
2  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. Neither the name of the project nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	$KAME: in6.c,v 1.259 2002/01/21 11:37:50 keiichi Exp $
30  */
31 
32 /*-
33  * Copyright (c) 1982, 1986, 1991, 1993
34  *	The Regents of the University of California.  All rights reserved.
35  *
36  * Redistribution and use in source and binary forms, with or without
37  * modification, are permitted provided that the following conditions
38  * are met:
39  * 1. Redistributions of source code must retain the above copyright
40  *    notice, this list of conditions and the following disclaimer.
41  * 2. Redistributions in binary form must reproduce the above copyright
42  *    notice, this list of conditions and the following disclaimer in the
43  *    documentation and/or other materials provided with the distribution.
44  * 4. Neither the name of the University nor the names of its contributors
45  *    may be used to endorse or promote products derived from this software
46  *    without specific prior written permission.
47  *
48  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58  * SUCH DAMAGE.
59  *
60  *	@(#)in.c	8.2 (Berkeley) 11/15/93
61  */
62 
63 #include <sys/cdefs.h>
64 __FBSDID("$FreeBSD$");
65 
66 #include "opt_compat.h"
67 #include "opt_inet.h"
68 #include "opt_inet6.h"
69 
70 #include <sys/param.h>
71 #include <sys/eventhandler.h>
72 #include <sys/errno.h>
73 #include <sys/jail.h>
74 #include <sys/malloc.h>
75 #include <sys/socket.h>
76 #include <sys/socketvar.h>
77 #include <sys/sockio.h>
78 #include <sys/systm.h>
79 #include <sys/priv.h>
80 #include <sys/proc.h>
81 #include <sys/time.h>
82 #include <sys/kernel.h>
83 #include <sys/lock.h>
84 #include <sys/rmlock.h>
85 #include <sys/syslog.h>
86 
87 #include <net/if.h>
88 #include <net/if_var.h>
89 #include <net/if_types.h>
90 #include <net/route.h>
91 #include <net/if_dl.h>
92 #include <net/vnet.h>
93 
94 #include <netinet/in.h>
95 #include <netinet/in_var.h>
96 #include <net/if_llatbl.h>
97 #include <netinet/if_ether.h>
98 #include <netinet/in_systm.h>
99 #include <netinet/ip.h>
100 #include <netinet/in_pcb.h>
101 #include <netinet/ip_carp.h>
102 
103 #include <netinet/ip6.h>
104 #include <netinet6/ip6_var.h>
105 #include <netinet6/nd6.h>
106 #include <netinet6/mld6_var.h>
107 #include <netinet6/ip6_mroute.h>
108 #include <netinet6/in6_ifattach.h>
109 #include <netinet6/scope6_var.h>
110 #include <netinet6/in6_fib.h>
111 #include <netinet6/in6_pcb.h>
112 
113 VNET_DECLARE(int, icmp6_nodeinfo_oldmcprefix);
114 #define V_icmp6_nodeinfo_oldmcprefix	VNET(icmp6_nodeinfo_oldmcprefix)
115 
116 /*
117  * Definitions of some costant IP6 addresses.
118  */
119 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
120 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
121 const struct in6_addr in6addr_nodelocal_allnodes =
122 	IN6ADDR_NODELOCAL_ALLNODES_INIT;
123 const struct in6_addr in6addr_linklocal_allnodes =
124 	IN6ADDR_LINKLOCAL_ALLNODES_INIT;
125 const struct in6_addr in6addr_linklocal_allrouters =
126 	IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
127 const struct in6_addr in6addr_linklocal_allv2routers =
128 	IN6ADDR_LINKLOCAL_ALLV2ROUTERS_INIT;
129 
130 const struct in6_addr in6mask0 = IN6MASK0;
131 const struct in6_addr in6mask32 = IN6MASK32;
132 const struct in6_addr in6mask64 = IN6MASK64;
133 const struct in6_addr in6mask96 = IN6MASK96;
134 const struct in6_addr in6mask128 = IN6MASK128;
135 
136 const struct sockaddr_in6 sa6_any =
137 	{ sizeof(sa6_any), AF_INET6, 0, 0, IN6ADDR_ANY_INIT, 0 };
138 
139 static int in6_notify_ifa(struct ifnet *, struct in6_ifaddr *,
140 	struct in6_aliasreq *, int);
141 static void in6_unlink_ifa(struct in6_ifaddr *, struct ifnet *);
142 
143 static int in6_validate_ifra(struct ifnet *, struct in6_aliasreq *,
144     struct in6_ifaddr *, int);
145 static struct in6_ifaddr *in6_alloc_ifa(struct ifnet *,
146     struct in6_aliasreq *, int flags);
147 static int in6_update_ifa_internal(struct ifnet *, struct in6_aliasreq *,
148     struct in6_ifaddr *, int, int);
149 static int in6_broadcast_ifa(struct ifnet *, struct in6_aliasreq *,
150     struct in6_ifaddr *, int);
151 
152 #define ifa2ia6(ifa)	((struct in6_ifaddr *)(ifa))
153 #define ia62ifa(ia6)	(&((ia6)->ia_ifa))
154 
155 
156 void
157 in6_newaddrmsg(struct in6_ifaddr *ia, int cmd)
158 {
159 	struct sockaddr_dl gateway;
160 	struct sockaddr_in6 mask, addr;
161 	struct rtentry rt;
162 
163 	/*
164 	 * initialize for rtmsg generation
165 	 */
166 	bzero(&gateway, sizeof(gateway));
167 	gateway.sdl_len = sizeof(gateway);
168 	gateway.sdl_family = AF_LINK;
169 
170 	bzero(&rt, sizeof(rt));
171 	rt.rt_gateway = (struct sockaddr *)&gateway;
172 	memcpy(&mask, &ia->ia_prefixmask, sizeof(ia->ia_prefixmask));
173 	memcpy(&addr, &ia->ia_addr, sizeof(ia->ia_addr));
174 	rt_mask(&rt) = (struct sockaddr *)&mask;
175 	rt_key(&rt) = (struct sockaddr *)&addr;
176 	rt.rt_flags = RTF_HOST | RTF_STATIC;
177 	if (cmd == RTM_ADD)
178 		rt.rt_flags |= RTF_UP;
179 	/* Announce arrival of local address to all FIBs. */
180 	rt_newaddrmsg(cmd, &ia->ia_ifa, 0, &rt);
181 }
182 
183 int
184 in6_mask2len(struct in6_addr *mask, u_char *lim0)
185 {
186 	int x = 0, y;
187 	u_char *lim = lim0, *p;
188 
189 	/* ignore the scope_id part */
190 	if (lim0 == NULL || lim0 - (u_char *)mask > sizeof(*mask))
191 		lim = (u_char *)mask + sizeof(*mask);
192 	for (p = (u_char *)mask; p < lim; x++, p++) {
193 		if (*p != 0xff)
194 			break;
195 	}
196 	y = 0;
197 	if (p < lim) {
198 		for (y = 0; y < 8; y++) {
199 			if ((*p & (0x80 >> y)) == 0)
200 				break;
201 		}
202 	}
203 
204 	/*
205 	 * when the limit pointer is given, do a stricter check on the
206 	 * remaining bits.
207 	 */
208 	if (p < lim) {
209 		if (y != 0 && (*p & (0x00ff >> y)) != 0)
210 			return (-1);
211 		for (p = p + 1; p < lim; p++)
212 			if (*p != 0)
213 				return (-1);
214 	}
215 
216 	return x * 8 + y;
217 }
218 
219 #ifdef COMPAT_FREEBSD32
220 struct in6_ndifreq32 {
221 	char ifname[IFNAMSIZ];
222 	uint32_t ifindex;
223 };
224 #define	SIOCGDEFIFACE32_IN6	_IOWR('i', 86, struct in6_ndifreq32)
225 #endif
226 
227 int
228 in6_control(struct socket *so, u_long cmd, caddr_t data,
229     struct ifnet *ifp, struct thread *td)
230 {
231 	struct	in6_ifreq *ifr = (struct in6_ifreq *)data;
232 	struct	in6_ifaddr *ia = NULL;
233 	struct	in6_aliasreq *ifra = (struct in6_aliasreq *)data;
234 	struct sockaddr_in6 *sa6;
235 	int carp_attached = 0;
236 	int error;
237 	u_long ocmd = cmd;
238 
239 	/*
240 	 * Compat to make pre-10.x ifconfig(8) operable.
241 	 */
242 	if (cmd == OSIOCAIFADDR_IN6)
243 		cmd = SIOCAIFADDR_IN6;
244 
245 	switch (cmd) {
246 	case SIOCGETSGCNT_IN6:
247 	case SIOCGETMIFCNT_IN6:
248 		/*
249 		 * XXX mrt_ioctl has a 3rd, unused, FIB argument in route.c.
250 		 * We cannot see how that would be needed, so do not adjust the
251 		 * KPI blindly; more likely should clean up the IPv4 variant.
252 		 */
253 		return (mrt6_ioctl ? mrt6_ioctl(cmd, data) : EOPNOTSUPP);
254 	}
255 
256 	switch (cmd) {
257 	case SIOCAADDRCTL_POLICY:
258 	case SIOCDADDRCTL_POLICY:
259 		if (td != NULL) {
260 			error = priv_check(td, PRIV_NETINET_ADDRCTRL6);
261 			if (error)
262 				return (error);
263 		}
264 		return (in6_src_ioctl(cmd, data));
265 	}
266 
267 	if (ifp == NULL)
268 		return (EOPNOTSUPP);
269 
270 	switch (cmd) {
271 	case SIOCSNDFLUSH_IN6:
272 	case SIOCSPFXFLUSH_IN6:
273 	case SIOCSRTRFLUSH_IN6:
274 	case SIOCSDEFIFACE_IN6:
275 	case SIOCSIFINFO_FLAGS:
276 	case SIOCSIFINFO_IN6:
277 		if (td != NULL) {
278 			error = priv_check(td, PRIV_NETINET_ND6);
279 			if (error)
280 				return (error);
281 		}
282 		/* FALLTHROUGH */
283 	case OSIOCGIFINFO_IN6:
284 	case SIOCGIFINFO_IN6:
285 	case SIOCGNBRINFO_IN6:
286 	case SIOCGDEFIFACE_IN6:
287 		return (nd6_ioctl(cmd, data, ifp));
288 
289 #ifdef COMPAT_FREEBSD32
290 	case SIOCGDEFIFACE32_IN6:
291 		{
292 			struct in6_ndifreq ndif;
293 			struct in6_ndifreq32 *ndif32;
294 
295 			error = nd6_ioctl(SIOCGDEFIFACE_IN6, (caddr_t)&ndif,
296 			    ifp);
297 			if (error)
298 				return (error);
299 			ndif32 = (struct in6_ndifreq32 *)data;
300 			ndif32->ifindex = ndif.ifindex;
301 			return (0);
302 		}
303 #endif
304 	}
305 
306 	switch (cmd) {
307 	case SIOCSIFPREFIX_IN6:
308 	case SIOCDIFPREFIX_IN6:
309 	case SIOCAIFPREFIX_IN6:
310 	case SIOCCIFPREFIX_IN6:
311 	case SIOCSGIFPREFIX_IN6:
312 	case SIOCGIFPREFIX_IN6:
313 		log(LOG_NOTICE,
314 		    "prefix ioctls are now invalidated. "
315 		    "please use ifconfig.\n");
316 		return (EOPNOTSUPP);
317 	}
318 
319 	switch (cmd) {
320 	case SIOCSSCOPE6:
321 		if (td != NULL) {
322 			error = priv_check(td, PRIV_NETINET_SCOPE6);
323 			if (error)
324 				return (error);
325 		}
326 		/* FALLTHROUGH */
327 	case SIOCGSCOPE6:
328 	case SIOCGSCOPE6DEF:
329 		return (scope6_ioctl(cmd, data, ifp));
330 	}
331 
332 	/*
333 	 * Find address for this interface, if it exists.
334 	 *
335 	 * In netinet code, we have checked ifra_addr in SIOCSIF*ADDR operation
336 	 * only, and used the first interface address as the target of other
337 	 * operations (without checking ifra_addr).  This was because netinet
338 	 * code/API assumed at most 1 interface address per interface.
339 	 * Since IPv6 allows a node to assign multiple addresses
340 	 * on a single interface, we almost always look and check the
341 	 * presence of ifra_addr, and reject invalid ones here.
342 	 * It also decreases duplicated code among SIOC*_IN6 operations.
343 	 */
344 	switch (cmd) {
345 	case SIOCAIFADDR_IN6:
346 	case SIOCSIFPHYADDR_IN6:
347 		sa6 = &ifra->ifra_addr;
348 		break;
349 	case SIOCSIFADDR_IN6:
350 	case SIOCGIFADDR_IN6:
351 	case SIOCSIFDSTADDR_IN6:
352 	case SIOCSIFNETMASK_IN6:
353 	case SIOCGIFDSTADDR_IN6:
354 	case SIOCGIFNETMASK_IN6:
355 	case SIOCDIFADDR_IN6:
356 	case SIOCGIFPSRCADDR_IN6:
357 	case SIOCGIFPDSTADDR_IN6:
358 	case SIOCGIFAFLAG_IN6:
359 	case SIOCSNDFLUSH_IN6:
360 	case SIOCSPFXFLUSH_IN6:
361 	case SIOCSRTRFLUSH_IN6:
362 	case SIOCGIFALIFETIME_IN6:
363 	case SIOCSIFALIFETIME_IN6:
364 	case SIOCGIFSTAT_IN6:
365 	case SIOCGIFSTAT_ICMP6:
366 		sa6 = &ifr->ifr_addr;
367 		break;
368 	case SIOCSIFADDR:
369 	case SIOCSIFBRDADDR:
370 	case SIOCSIFDSTADDR:
371 	case SIOCSIFNETMASK:
372 		/*
373 		 * Although we should pass any non-INET6 ioctl requests
374 		 * down to driver, we filter some legacy INET requests.
375 		 * Drivers trust SIOCSIFADDR et al to come from an already
376 		 * privileged layer, and do not perform any credentials
377 		 * checks or input validation.
378 		 */
379 		return (EINVAL);
380 	default:
381 		sa6 = NULL;
382 		break;
383 	}
384 	if (sa6 && sa6->sin6_family == AF_INET6) {
385 		if (sa6->sin6_scope_id != 0)
386 			error = sa6_embedscope(sa6, 0);
387 		else
388 			error = in6_setscope(&sa6->sin6_addr, ifp, NULL);
389 		if (error != 0)
390 			return (error);
391 		if (td != NULL && (error = prison_check_ip6(td->td_ucred,
392 		    &sa6->sin6_addr)) != 0)
393 			return (error);
394 		ia = in6ifa_ifpwithaddr(ifp, &sa6->sin6_addr);
395 	} else
396 		ia = NULL;
397 
398 	switch (cmd) {
399 	case SIOCSIFADDR_IN6:
400 	case SIOCSIFDSTADDR_IN6:
401 	case SIOCSIFNETMASK_IN6:
402 		/*
403 		 * Since IPv6 allows a node to assign multiple addresses
404 		 * on a single interface, SIOCSIFxxx ioctls are deprecated.
405 		 */
406 		/* we decided to obsolete this command (20000704) */
407 		error = EINVAL;
408 		goto out;
409 
410 	case SIOCDIFADDR_IN6:
411 		/*
412 		 * for IPv4, we look for existing in_ifaddr here to allow
413 		 * "ifconfig if0 delete" to remove the first IPv4 address on
414 		 * the interface.  For IPv6, as the spec allows multiple
415 		 * interface address from the day one, we consider "remove the
416 		 * first one" semantics to be not preferable.
417 		 */
418 		if (ia == NULL) {
419 			error = EADDRNOTAVAIL;
420 			goto out;
421 		}
422 		/* FALLTHROUGH */
423 	case SIOCAIFADDR_IN6:
424 		/*
425 		 * We always require users to specify a valid IPv6 address for
426 		 * the corresponding operation.
427 		 */
428 		if (ifra->ifra_addr.sin6_family != AF_INET6 ||
429 		    ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6)) {
430 			error = EAFNOSUPPORT;
431 			goto out;
432 		}
433 
434 		if (td != NULL) {
435 			error = priv_check(td, (cmd == SIOCDIFADDR_IN6) ?
436 			    PRIV_NET_DELIFADDR : PRIV_NET_ADDIFADDR);
437 			if (error)
438 				goto out;
439 		}
440 		/* FALLTHROUGH */
441 	case SIOCGIFSTAT_IN6:
442 	case SIOCGIFSTAT_ICMP6:
443 		if (ifp->if_afdata[AF_INET6] == NULL) {
444 			error = EPFNOSUPPORT;
445 			goto out;
446 		}
447 		break;
448 
449 	case SIOCGIFADDR_IN6:
450 		/* This interface is basically deprecated. use SIOCGIFCONF. */
451 		/* FALLTHROUGH */
452 	case SIOCGIFAFLAG_IN6:
453 	case SIOCGIFNETMASK_IN6:
454 	case SIOCGIFDSTADDR_IN6:
455 	case SIOCGIFALIFETIME_IN6:
456 		/* must think again about its semantics */
457 		if (ia == NULL) {
458 			error = EADDRNOTAVAIL;
459 			goto out;
460 		}
461 		break;
462 
463 	case SIOCSIFALIFETIME_IN6:
464 	    {
465 		struct in6_addrlifetime *lt;
466 
467 		if (td != NULL) {
468 			error = priv_check(td, PRIV_NETINET_ALIFETIME6);
469 			if (error)
470 				goto out;
471 		}
472 		if (ia == NULL) {
473 			error = EADDRNOTAVAIL;
474 			goto out;
475 		}
476 		/* sanity for overflow - beware unsigned */
477 		lt = &ifr->ifr_ifru.ifru_lifetime;
478 		if (lt->ia6t_vltime != ND6_INFINITE_LIFETIME &&
479 		    lt->ia6t_vltime + time_uptime < time_uptime) {
480 			error = EINVAL;
481 			goto out;
482 		}
483 		if (lt->ia6t_pltime != ND6_INFINITE_LIFETIME &&
484 		    lt->ia6t_pltime + time_uptime < time_uptime) {
485 			error = EINVAL;
486 			goto out;
487 		}
488 		break;
489 	    }
490 	}
491 
492 	switch (cmd) {
493 	case SIOCGIFADDR_IN6:
494 		ifr->ifr_addr = ia->ia_addr;
495 		if ((error = sa6_recoverscope(&ifr->ifr_addr)) != 0)
496 			goto out;
497 		break;
498 
499 	case SIOCGIFDSTADDR_IN6:
500 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
501 			error = EINVAL;
502 			goto out;
503 		}
504 		/*
505 		 * XXX: should we check if ifa_dstaddr is NULL and return
506 		 * an error?
507 		 */
508 		ifr->ifr_dstaddr = ia->ia_dstaddr;
509 		if ((error = sa6_recoverscope(&ifr->ifr_dstaddr)) != 0)
510 			goto out;
511 		break;
512 
513 	case SIOCGIFNETMASK_IN6:
514 		ifr->ifr_addr = ia->ia_prefixmask;
515 		break;
516 
517 	case SIOCGIFAFLAG_IN6:
518 		ifr->ifr_ifru.ifru_flags6 = ia->ia6_flags;
519 		break;
520 
521 	case SIOCGIFSTAT_IN6:
522 		COUNTER_ARRAY_COPY(((struct in6_ifextra *)
523 		    ifp->if_afdata[AF_INET6])->in6_ifstat,
524 		    &ifr->ifr_ifru.ifru_stat,
525 		    sizeof(struct in6_ifstat) / sizeof(uint64_t));
526 		break;
527 
528 	case SIOCGIFSTAT_ICMP6:
529 		COUNTER_ARRAY_COPY(((struct in6_ifextra *)
530 		    ifp->if_afdata[AF_INET6])->icmp6_ifstat,
531 		    &ifr->ifr_ifru.ifru_icmp6stat,
532 		    sizeof(struct icmp6_ifstat) / sizeof(uint64_t));
533 		break;
534 
535 	case SIOCGIFALIFETIME_IN6:
536 		ifr->ifr_ifru.ifru_lifetime = ia->ia6_lifetime;
537 		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
538 			time_t maxexpire;
539 			struct in6_addrlifetime *retlt =
540 			    &ifr->ifr_ifru.ifru_lifetime;
541 
542 			/*
543 			 * XXX: adjust expiration time assuming time_t is
544 			 * signed.
545 			 */
546 			maxexpire = (-1) &
547 			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
548 			if (ia->ia6_lifetime.ia6t_vltime <
549 			    maxexpire - ia->ia6_updatetime) {
550 				retlt->ia6t_expire = ia->ia6_updatetime +
551 				    ia->ia6_lifetime.ia6t_vltime;
552 			} else
553 				retlt->ia6t_expire = maxexpire;
554 		}
555 		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
556 			time_t maxexpire;
557 			struct in6_addrlifetime *retlt =
558 			    &ifr->ifr_ifru.ifru_lifetime;
559 
560 			/*
561 			 * XXX: adjust expiration time assuming time_t is
562 			 * signed.
563 			 */
564 			maxexpire = (-1) &
565 			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
566 			if (ia->ia6_lifetime.ia6t_pltime <
567 			    maxexpire - ia->ia6_updatetime) {
568 				retlt->ia6t_preferred = ia->ia6_updatetime +
569 				    ia->ia6_lifetime.ia6t_pltime;
570 			} else
571 				retlt->ia6t_preferred = maxexpire;
572 		}
573 		break;
574 
575 	case SIOCSIFALIFETIME_IN6:
576 		ia->ia6_lifetime = ifr->ifr_ifru.ifru_lifetime;
577 		/* for sanity */
578 		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
579 			ia->ia6_lifetime.ia6t_expire =
580 				time_uptime + ia->ia6_lifetime.ia6t_vltime;
581 		} else
582 			ia->ia6_lifetime.ia6t_expire = 0;
583 		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
584 			ia->ia6_lifetime.ia6t_preferred =
585 				time_uptime + ia->ia6_lifetime.ia6t_pltime;
586 		} else
587 			ia->ia6_lifetime.ia6t_preferred = 0;
588 		break;
589 
590 	case SIOCAIFADDR_IN6:
591 	{
592 		struct nd_prefixctl pr0;
593 		struct nd_prefix *pr;
594 
595 		/*
596 		 * first, make or update the interface address structure,
597 		 * and link it to the list.
598 		 */
599 		if ((error = in6_update_ifa(ifp, ifra, ia, 0)) != 0)
600 			goto out;
601 		if (ia != NULL)
602 			ifa_free(&ia->ia_ifa);
603 		if ((ia = in6ifa_ifpwithaddr(ifp, &ifra->ifra_addr.sin6_addr))
604 		    == NULL) {
605 			/*
606 			 * this can happen when the user specify the 0 valid
607 			 * lifetime.
608 			 */
609 			break;
610 		}
611 
612 		if (cmd == ocmd && ifra->ifra_vhid > 0) {
613 			if (carp_attach_p != NULL)
614 				error = (*carp_attach_p)(&ia->ia_ifa,
615 				    ifra->ifra_vhid);
616 			else
617 				error = EPROTONOSUPPORT;
618 			if (error)
619 				goto out;
620 			else
621 				carp_attached = 1;
622 		}
623 
624 		/*
625 		 * then, make the prefix on-link on the interface.
626 		 * XXX: we'd rather create the prefix before the address, but
627 		 * we need at least one address to install the corresponding
628 		 * interface route, so we configure the address first.
629 		 */
630 
631 		/*
632 		 * convert mask to prefix length (prefixmask has already
633 		 * been validated in in6_update_ifa().
634 		 */
635 		bzero(&pr0, sizeof(pr0));
636 		pr0.ndpr_ifp = ifp;
637 		pr0.ndpr_plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
638 		    NULL);
639 		if (pr0.ndpr_plen == 128) {
640 			/* we don't need to install a host route. */
641 			goto aifaddr_out;
642 		}
643 		pr0.ndpr_prefix = ifra->ifra_addr;
644 		/* apply the mask for safety. */
645 		IN6_MASK_ADDR(&pr0.ndpr_prefix.sin6_addr,
646 		    &ifra->ifra_prefixmask.sin6_addr);
647 
648 		/*
649 		 * XXX: since we don't have an API to set prefix (not address)
650 		 * lifetimes, we just use the same lifetimes as addresses.
651 		 * The (temporarily) installed lifetimes can be overridden by
652 		 * later advertised RAs (when accept_rtadv is non 0), which is
653 		 * an intended behavior.
654 		 */
655 		pr0.ndpr_raf_onlink = 1; /* should be configurable? */
656 		pr0.ndpr_raf_auto =
657 		    ((ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0);
658 		pr0.ndpr_vltime = ifra->ifra_lifetime.ia6t_vltime;
659 		pr0.ndpr_pltime = ifra->ifra_lifetime.ia6t_pltime;
660 
661 		/* add the prefix if not yet. */
662 		if ((pr = nd6_prefix_lookup(&pr0)) == NULL) {
663 			/*
664 			 * nd6_prelist_add will install the corresponding
665 			 * interface route.
666 			 */
667 			if ((error = nd6_prelist_add(&pr0, NULL, &pr)) != 0) {
668 				if (carp_attached)
669 					(*carp_detach_p)(&ia->ia_ifa);
670 				goto out;
671 			}
672 		}
673 
674 		/* relate the address to the prefix */
675 		if (ia->ia6_ndpr == NULL) {
676 			ia->ia6_ndpr = pr;
677 			pr->ndpr_refcnt++;
678 
679 			/*
680 			 * If this is the first autoconf address from the
681 			 * prefix, create a temporary address as well
682 			 * (when required).
683 			 */
684 			if ((ia->ia6_flags & IN6_IFF_AUTOCONF) &&
685 			    V_ip6_use_tempaddr && pr->ndpr_refcnt == 1) {
686 				int e;
687 				if ((e = in6_tmpifadd(ia, 1, 0)) != 0) {
688 					log(LOG_NOTICE, "in6_control: failed "
689 					    "to create a temporary address, "
690 					    "errno=%d\n", e);
691 				}
692 			}
693 		}
694 
695 		/*
696 		 * this might affect the status of autoconfigured addresses,
697 		 * that is, this address might make other addresses detached.
698 		 */
699 		pfxlist_onlink_check();
700 aifaddr_out:
701 		if (error != 0 || ia == NULL)
702 			break;
703 		/*
704 		 * Try to clear the flag when a new IPv6 address is added
705 		 * onto an IFDISABLED interface and it succeeds.
706 		 */
707 		if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) {
708 			struct in6_ndireq nd;
709 
710 			memset(&nd, 0, sizeof(nd));
711 			nd.ndi.flags = ND_IFINFO(ifp)->flags;
712 			nd.ndi.flags &= ~ND6_IFF_IFDISABLED;
713 			if (nd6_ioctl(SIOCSIFINFO_FLAGS, (caddr_t)&nd, ifp) < 0)
714 				log(LOG_NOTICE, "SIOCAIFADDR_IN6: "
715 				    "SIOCSIFINFO_FLAGS for -ifdisabled "
716 				    "failed.");
717 			/*
718 			 * Ignore failure of clearing the flag intentionally.
719 			 * The failure means address duplication was detected.
720 			 */
721 		}
722 		EVENTHANDLER_INVOKE(ifaddr_event, ifp);
723 		break;
724 	}
725 
726 	case SIOCDIFADDR_IN6:
727 	{
728 		struct nd_prefix *pr;
729 
730 		/*
731 		 * If the address being deleted is the only one that owns
732 		 * the corresponding prefix, expire the prefix as well.
733 		 * XXX: theoretically, we don't have to worry about such
734 		 * relationship, since we separate the address management
735 		 * and the prefix management.  We do this, however, to provide
736 		 * as much backward compatibility as possible in terms of
737 		 * the ioctl operation.
738 		 * Note that in6_purgeaddr() will decrement ndpr_refcnt.
739 		 */
740 		pr = ia->ia6_ndpr;
741 		in6_purgeaddr(&ia->ia_ifa);
742 		if (pr && pr->ndpr_refcnt == 0)
743 			prelist_remove(pr);
744 		EVENTHANDLER_INVOKE(ifaddr_event, ifp);
745 		break;
746 	}
747 
748 	default:
749 		if (ifp->if_ioctl == NULL) {
750 			error = EOPNOTSUPP;
751 			goto out;
752 		}
753 		error = (*ifp->if_ioctl)(ifp, cmd, data);
754 		goto out;
755 	}
756 
757 	error = 0;
758 out:
759 	if (ia != NULL)
760 		ifa_free(&ia->ia_ifa);
761 	return (error);
762 }
763 
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(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(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(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(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(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 /*
1082  * Allocate a new ifaddr and link it into chains.
1083  */
1084 static struct in6_ifaddr *
1085 in6_alloc_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra, int flags)
1086 {
1087 	struct in6_ifaddr *ia;
1088 
1089 	/*
1090 	 * When in6_alloc_ifa() is called in a process of a received
1091 	 * RA, it is called under an interrupt context.  So, we should
1092 	 * call malloc with M_NOWAIT.
1093 	 */
1094 	ia = (struct in6_ifaddr *)ifa_alloc(sizeof(*ia), M_NOWAIT);
1095 	if (ia == NULL)
1096 		return (NULL);
1097 	LIST_INIT(&ia->ia6_memberships);
1098 	/* Initialize the address and masks, and put time stamp */
1099 	ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
1100 	ia->ia_addr.sin6_family = AF_INET6;
1101 	ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
1102 	/* XXX: Can we assign ,sin6_addr and skip the rest? */
1103 	ia->ia_addr = ifra->ifra_addr;
1104 	ia->ia6_createtime = time_uptime;
1105 	if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
1106 		/*
1107 		 * Some functions expect that ifa_dstaddr is not
1108 		 * NULL for p2p interfaces.
1109 		 */
1110 		ia->ia_ifa.ifa_dstaddr =
1111 		    (struct sockaddr *)&ia->ia_dstaddr;
1112 	} else {
1113 		ia->ia_ifa.ifa_dstaddr = NULL;
1114 	}
1115 
1116 	/* set prefix mask if any */
1117 	ia->ia_ifa.ifa_netmask = (struct sockaddr *)&ia->ia_prefixmask;
1118 	if (ifra->ifra_prefixmask.sin6_len != 0) {
1119 		ia->ia_prefixmask.sin6_family = AF_INET6;
1120 		ia->ia_prefixmask.sin6_len = ifra->ifra_prefixmask.sin6_len;
1121 		ia->ia_prefixmask.sin6_addr = ifra->ifra_prefixmask.sin6_addr;
1122 	}
1123 
1124 	ia->ia_ifp = ifp;
1125 	ifa_ref(&ia->ia_ifa);			/* if_addrhead */
1126 	IF_ADDR_WLOCK(ifp);
1127 	TAILQ_INSERT_TAIL(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
1128 	IF_ADDR_WUNLOCK(ifp);
1129 
1130 	ifa_ref(&ia->ia_ifa);			/* in6_ifaddrhead */
1131 	IN6_IFADDR_WLOCK();
1132 	TAILQ_INSERT_TAIL(&V_in6_ifaddrhead, ia, ia_link);
1133 	LIST_INSERT_HEAD(IN6ADDR_HASH(&ia->ia_addr.sin6_addr), ia, ia6_hash);
1134 	IN6_IFADDR_WUNLOCK();
1135 
1136 	return (ia);
1137 }
1138 
1139 /*
1140  * Update/configure interface address parameters:
1141  *
1142  * 1) Update lifetime
1143  * 2) Update interface metric ad flags
1144  * 3) Notify other subsystems
1145  */
1146 static int
1147 in6_update_ifa_internal(struct ifnet *ifp, struct in6_aliasreq *ifra,
1148     struct in6_ifaddr *ia, int hostIsNew, int flags)
1149 {
1150 	int error;
1151 
1152 	/* update timestamp */
1153 	ia->ia6_updatetime = time_uptime;
1154 
1155 	/*
1156 	 * Set lifetimes.  We do not refer to ia6t_expire and ia6t_preferred
1157 	 * to see if the address is deprecated or invalidated, but initialize
1158 	 * these members for applications.
1159 	 */
1160 	ia->ia6_lifetime = ifra->ifra_lifetime;
1161 	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1162 		ia->ia6_lifetime.ia6t_expire =
1163 		    time_uptime + ia->ia6_lifetime.ia6t_vltime;
1164 	} else
1165 		ia->ia6_lifetime.ia6t_expire = 0;
1166 	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1167 		ia->ia6_lifetime.ia6t_preferred =
1168 		    time_uptime + ia->ia6_lifetime.ia6t_pltime;
1169 	} else
1170 		ia->ia6_lifetime.ia6t_preferred = 0;
1171 
1172 	/*
1173 	 * backward compatibility - if IN6_IFF_DEPRECATED is set from the
1174 	 * userland, make it deprecated.
1175 	 */
1176 	if ((ifra->ifra_flags & IN6_IFF_DEPRECATED) != 0) {
1177 		ia->ia6_lifetime.ia6t_pltime = 0;
1178 		ia->ia6_lifetime.ia6t_preferred = time_uptime;
1179 	}
1180 
1181 	/*
1182 	 * configure address flags.
1183 	 */
1184 	ia->ia6_flags = ifra->ifra_flags;
1185 
1186 	/*
1187 	 * Make the address tentative before joining multicast addresses,
1188 	 * so that corresponding MLD responses would not have a tentative
1189 	 * source address.
1190 	 */
1191 	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/* safety */
1192 
1193 	/*
1194 	 * DAD should be performed for an new address or addresses on
1195 	 * an interface with ND6_IFF_IFDISABLED.
1196 	 */
1197 	if (in6if_do_dad(ifp) &&
1198 	    (hostIsNew || (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)))
1199 		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1200 
1201 	/* notify other subsystems */
1202 	error = in6_notify_ifa(ifp, ia, ifra, hostIsNew);
1203 
1204 	return (error);
1205 }
1206 
1207 /*
1208  * Do link-level ifa job:
1209  * 1) Add lle entry for added address
1210  * 2) Notifies routing socket users about new address
1211  * 3) join appropriate multicast group
1212  * 4) start DAD if enabled
1213  */
1214 static int
1215 in6_broadcast_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
1216     struct in6_ifaddr *ia, int flags)
1217 {
1218 	struct in6_multi *in6m_sol;
1219 	int error = 0;
1220 
1221 	/* Add local address to lltable, if necessary (ex. on p2p link). */
1222 	if ((error = nd6_add_ifa_lle(ia)) != 0) {
1223 		in6_purgeaddr(&ia->ia_ifa);
1224 		ifa_free(&ia->ia_ifa);
1225 		return (error);
1226 	}
1227 
1228 	/* Join necessary multicast groups. */
1229 	in6m_sol = NULL;
1230 	if ((ifp->if_flags & IFF_MULTICAST) != 0) {
1231 		error = in6_update_ifa_join_mc(ifp, ifra, ia, flags, &in6m_sol);
1232 		if (error != 0) {
1233 			in6_purgeaddr(&ia->ia_ifa);
1234 			ifa_free(&ia->ia_ifa);
1235 			return (error);
1236 		}
1237 	}
1238 
1239 	/* Perform DAD, if the address is TENTATIVE. */
1240 	if ((ia->ia6_flags & IN6_IFF_TENTATIVE)) {
1241 		int delay, mindelay, maxdelay;
1242 
1243 		delay = 0;
1244 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1245 			/*
1246 			 * We need to impose a delay before sending an NS
1247 			 * for DAD.  Check if we also needed a delay for the
1248 			 * corresponding MLD message.  If we did, the delay
1249 			 * should be larger than the MLD delay (this could be
1250 			 * relaxed a bit, but this simple logic is at least
1251 			 * safe).
1252 			 * XXX: Break data hiding guidelines and look at
1253 			 * state for the solicited multicast group.
1254 			 */
1255 			mindelay = 0;
1256 			if (in6m_sol != NULL &&
1257 			    in6m_sol->in6m_state == MLD_REPORTING_MEMBER) {
1258 				mindelay = in6m_sol->in6m_timer;
1259 			}
1260 			maxdelay = MAX_RTR_SOLICITATION_DELAY * hz;
1261 			if (maxdelay - mindelay == 0)
1262 				delay = 0;
1263 			else {
1264 				delay =
1265 				    (arc4random() % (maxdelay - mindelay)) +
1266 				    mindelay;
1267 			}
1268 		}
1269 		nd6_dad_start((struct ifaddr *)ia, delay);
1270 	}
1271 
1272 	in6_newaddrmsg(ia, RTM_ADD);
1273 	ifa_free(&ia->ia_ifa);
1274 	return (error);
1275 }
1276 
1277 void
1278 in6_purgeaddr(struct ifaddr *ifa)
1279 {
1280 	struct ifnet *ifp = ifa->ifa_ifp;
1281 	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1282 	struct in6_multi_mship *imm;
1283 	int plen, error;
1284 
1285 	if (ifa->ifa_carp)
1286 		(*carp_detach_p)(ifa);
1287 
1288 	/*
1289 	 * Remove the loopback route to the interface address.
1290 	 * The check for the current setting of "nd6_useloopback"
1291 	 * is not needed.
1292 	 */
1293 	if (ia->ia_flags & IFA_RTSELF) {
1294 		error = ifa_del_loopback_route((struct ifaddr *)ia,
1295 		    (struct sockaddr *)&ia->ia_addr);
1296 		if (error == 0)
1297 			ia->ia_flags &= ~IFA_RTSELF;
1298 	}
1299 
1300 	/* stop DAD processing */
1301 	nd6_dad_stop(ifa);
1302 
1303 	/* Leave multicast groups. */
1304 	while ((imm = LIST_FIRST(&ia->ia6_memberships)) != NULL) {
1305 		LIST_REMOVE(imm, i6mm_chain);
1306 		in6_leavegroup(imm);
1307 	}
1308 	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1309 	if ((ia->ia_flags & IFA_ROUTE) && plen == 128) {
1310 		error = rtinit(&(ia->ia_ifa), RTM_DELETE, ia->ia_flags |
1311 		    (ia->ia_dstaddr.sin6_family == AF_INET6 ? RTF_HOST : 0));
1312 		if (error != 0)
1313 			log(LOG_INFO, "%s: err=%d, destination address delete "
1314 			    "failed\n", __func__, error);
1315 		ia->ia_flags &= ~IFA_ROUTE;
1316 	}
1317 
1318 	in6_newaddrmsg(ia, RTM_DELETE);
1319 	in6_unlink_ifa(ia, ifp);
1320 }
1321 
1322 static void
1323 in6_unlink_ifa(struct in6_ifaddr *ia, struct ifnet *ifp)
1324 {
1325 	char ip6buf[INET6_ADDRSTRLEN];
1326 	int remove_lle;
1327 
1328 	IF_ADDR_WLOCK(ifp);
1329 	TAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
1330 	IF_ADDR_WUNLOCK(ifp);
1331 	ifa_free(&ia->ia_ifa);			/* if_addrhead */
1332 
1333 	/*
1334 	 * Defer the release of what might be the last reference to the
1335 	 * in6_ifaddr so that it can't be freed before the remainder of the
1336 	 * cleanup.
1337 	 */
1338 	IN6_IFADDR_WLOCK();
1339 	TAILQ_REMOVE(&V_in6_ifaddrhead, ia, ia_link);
1340 	LIST_REMOVE(ia, ia6_hash);
1341 	IN6_IFADDR_WUNLOCK();
1342 
1343 	/*
1344 	 * Release the reference to the base prefix.  There should be a
1345 	 * positive reference.
1346 	 */
1347 	remove_lle = 0;
1348 	if (ia->ia6_ndpr == NULL) {
1349 		nd6log((LOG_NOTICE,
1350 		    "in6_unlink_ifa: autoconf'ed address "
1351 		    "%s has no prefix\n", ip6_sprintf(ip6buf, IA6_IN6(ia))));
1352 	} else {
1353 		ia->ia6_ndpr->ndpr_refcnt--;
1354 		/* Do not delete lles within prefix if refcont != 0 */
1355 		if (ia->ia6_ndpr->ndpr_refcnt == 0)
1356 			remove_lle = 1;
1357 		ia->ia6_ndpr = NULL;
1358 	}
1359 
1360 	nd6_rem_ifa_lle(ia, remove_lle);
1361 
1362 	/*
1363 	 * Also, if the address being removed is autoconf'ed, call
1364 	 * pfxlist_onlink_check() since the release might affect the status of
1365 	 * other (detached) addresses.
1366 	 */
1367 	if ((ia->ia6_flags & IN6_IFF_AUTOCONF)) {
1368 		pfxlist_onlink_check();
1369 	}
1370 	ifa_free(&ia->ia_ifa);			/* in6_ifaddrhead */
1371 }
1372 
1373 /*
1374  * Notifies other subsystems about address change/arrival:
1375  * 1) Notifies device handler on the first IPv6 address assignment
1376  * 2) Handle routing table changes for P2P links and route
1377  * 3) Handle routing table changes for address host route
1378  */
1379 static int
1380 in6_notify_ifa(struct ifnet *ifp, struct in6_ifaddr *ia,
1381     struct in6_aliasreq *ifra, int hostIsNew)
1382 {
1383 	int	error = 0, plen, ifacount = 0;
1384 	struct ifaddr *ifa;
1385 	struct sockaddr_in6 *pdst;
1386 	char ip6buf[INET6_ADDRSTRLEN];
1387 
1388 	/*
1389 	 * Give the interface a chance to initialize
1390 	 * if this is its first address,
1391 	 */
1392 	if (hostIsNew != 0) {
1393 		IF_ADDR_RLOCK(ifp);
1394 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1395 			if (ifa->ifa_addr->sa_family != AF_INET6)
1396 				continue;
1397 			ifacount++;
1398 		}
1399 		IF_ADDR_RUNLOCK(ifp);
1400 	}
1401 
1402 	if (ifacount <= 1 && ifp->if_ioctl) {
1403 		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
1404 		if (error)
1405 			return (error);
1406 	}
1407 
1408 	/*
1409 	 * If a new destination address is specified, scrub the old one and
1410 	 * install the new destination.  Note that the interface must be
1411 	 * p2p or loopback.
1412 	 */
1413 	pdst = &ifra->ifra_dstaddr;
1414 	if (pdst->sin6_family == AF_INET6 &&
1415 	    !IN6_ARE_ADDR_EQUAL(&pdst->sin6_addr, &ia->ia_dstaddr.sin6_addr)) {
1416 		if ((ia->ia_flags & IFA_ROUTE) != 0 &&
1417 		    (rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST) != 0)) {
1418 			nd6log((LOG_ERR, "in6_update_ifa_internal: failed to "
1419 			    "remove a route to the old destination: %s\n",
1420 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1421 			/* proceed anyway... */
1422 		} else
1423 			ia->ia_flags &= ~IFA_ROUTE;
1424 		ia->ia_dstaddr = *pdst;
1425 	}
1426 
1427 	/*
1428 	 * If a new destination address is specified for a point-to-point
1429 	 * interface, install a route to the destination as an interface
1430 	 * direct route.
1431 	 * XXX: the logic below rejects assigning multiple addresses on a p2p
1432 	 * interface that share the same destination.
1433 	 */
1434 	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1435 	if (!(ia->ia_flags & IFA_ROUTE) && plen == 128 &&
1436 	    ia->ia_dstaddr.sin6_family == AF_INET6) {
1437 		int rtflags = RTF_UP | RTF_HOST;
1438 		/*
1439 		 * Handle the case for ::1 .
1440 		 */
1441 		if (ifp->if_flags & IFF_LOOPBACK)
1442 			ia->ia_flags |= IFA_RTSELF;
1443 		error = rtinit(&ia->ia_ifa, RTM_ADD, ia->ia_flags | rtflags);
1444 		if (error)
1445 			return (error);
1446 		ia->ia_flags |= IFA_ROUTE;
1447 	}
1448 
1449 	/*
1450 	 * add a loopback route to self if not exists
1451 	 */
1452 	if (!(ia->ia_flags & IFA_RTSELF) && V_nd6_useloopback) {
1453 		error = ifa_add_loopback_route((struct ifaddr *)ia,
1454 		    (struct sockaddr *)&ia->ia_addr);
1455 		if (error == 0)
1456 			ia->ia_flags |= IFA_RTSELF;
1457 	}
1458 
1459 	return (error);
1460 }
1461 
1462 /*
1463  * Find an IPv6 interface link-local address specific to an interface.
1464  * ifaddr is returned referenced.
1465  */
1466 struct in6_ifaddr *
1467 in6ifa_ifpforlinklocal(struct ifnet *ifp, int ignoreflags)
1468 {
1469 	struct ifaddr *ifa;
1470 
1471 	IF_ADDR_RLOCK(ifp);
1472 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1473 		if (ifa->ifa_addr->sa_family != AF_INET6)
1474 			continue;
1475 		if (IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa))) {
1476 			if ((((struct in6_ifaddr *)ifa)->ia6_flags &
1477 			    ignoreflags) != 0)
1478 				continue;
1479 			ifa_ref(ifa);
1480 			break;
1481 		}
1482 	}
1483 	IF_ADDR_RUNLOCK(ifp);
1484 
1485 	return ((struct in6_ifaddr *)ifa);
1486 }
1487 
1488 
1489 /*
1490  * find the internet address corresponding to a given address.
1491  * ifaddr is returned referenced.
1492  */
1493 struct in6_ifaddr *
1494 in6ifa_ifwithaddr(const struct in6_addr *addr, uint32_t zoneid)
1495 {
1496 	struct rm_priotracker in6_ifa_tracker;
1497 	struct in6_ifaddr *ia;
1498 
1499 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1500 	LIST_FOREACH(ia, IN6ADDR_HASH(addr), ia6_hash) {
1501 		if (IN6_ARE_ADDR_EQUAL(IA6_IN6(ia), addr)) {
1502 			if (zoneid != 0 &&
1503 			    zoneid != ia->ia_addr.sin6_scope_id)
1504 				continue;
1505 			ifa_ref(&ia->ia_ifa);
1506 			break;
1507 		}
1508 	}
1509 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1510 	return (ia);
1511 }
1512 
1513 /*
1514  * find the internet address corresponding to a given interface and address.
1515  * ifaddr is returned referenced.
1516  */
1517 struct in6_ifaddr *
1518 in6ifa_ifpwithaddr(struct ifnet *ifp, const struct in6_addr *addr)
1519 {
1520 	struct ifaddr *ifa;
1521 
1522 	IF_ADDR_RLOCK(ifp);
1523 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1524 		if (ifa->ifa_addr->sa_family != AF_INET6)
1525 			continue;
1526 		if (IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa))) {
1527 			ifa_ref(ifa);
1528 			break;
1529 		}
1530 	}
1531 	IF_ADDR_RUNLOCK(ifp);
1532 
1533 	return ((struct in6_ifaddr *)ifa);
1534 }
1535 
1536 /*
1537  * Find a link-local scoped address on ifp and return it if any.
1538  */
1539 struct in6_ifaddr *
1540 in6ifa_llaonifp(struct ifnet *ifp)
1541 {
1542 	struct sockaddr_in6 *sin6;
1543 	struct ifaddr *ifa;
1544 
1545 	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)
1546 		return (NULL);
1547 	IF_ADDR_RLOCK(ifp);
1548 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1549 		if (ifa->ifa_addr->sa_family != AF_INET6)
1550 			continue;
1551 		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
1552 		if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr) ||
1553 		    IN6_IS_ADDR_MC_INTFACELOCAL(&sin6->sin6_addr) ||
1554 		    IN6_IS_ADDR_MC_NODELOCAL(&sin6->sin6_addr))
1555 			break;
1556 	}
1557 	IF_ADDR_RUNLOCK(ifp);
1558 
1559 	return ((struct in6_ifaddr *)ifa);
1560 }
1561 
1562 /*
1563  * Convert IP6 address to printable (loggable) representation. Caller
1564  * has to make sure that ip6buf is at least INET6_ADDRSTRLEN long.
1565  */
1566 static char digits[] = "0123456789abcdef";
1567 char *
1568 ip6_sprintf(char *ip6buf, const struct in6_addr *addr)
1569 {
1570 	int i, cnt = 0, maxcnt = 0, idx = 0, index = 0;
1571 	char *cp;
1572 	const u_int16_t *a = (const u_int16_t *)addr;
1573 	const u_int8_t *d;
1574 	int dcolon = 0, zero = 0;
1575 
1576 	cp = ip6buf;
1577 
1578 	for (i = 0; i < 8; i++) {
1579 		if (*(a + i) == 0) {
1580 			cnt++;
1581 			if (cnt == 1)
1582 				idx = i;
1583 		}
1584 		else if (maxcnt < cnt) {
1585 			maxcnt = cnt;
1586 			index = idx;
1587 			cnt = 0;
1588 		}
1589 	}
1590 	if (maxcnt < cnt) {
1591 		maxcnt = cnt;
1592 		index = idx;
1593 	}
1594 
1595 	for (i = 0; i < 8; i++) {
1596 		if (dcolon == 1) {
1597 			if (*a == 0) {
1598 				if (i == 7)
1599 					*cp++ = ':';
1600 				a++;
1601 				continue;
1602 			} else
1603 				dcolon = 2;
1604 		}
1605 		if (*a == 0) {
1606 			if (dcolon == 0 && *(a + 1) == 0 && i == index) {
1607 				if (i == 0)
1608 					*cp++ = ':';
1609 				*cp++ = ':';
1610 				dcolon = 1;
1611 			} else {
1612 				*cp++ = '0';
1613 				*cp++ = ':';
1614 			}
1615 			a++;
1616 			continue;
1617 		}
1618 		d = (const u_char *)a;
1619 		/* Try to eliminate leading zeros in printout like in :0001. */
1620 		zero = 1;
1621 		*cp = digits[*d >> 4];
1622 		if (*cp != '0') {
1623 			zero = 0;
1624 			cp++;
1625 		}
1626 		*cp = digits[*d++ & 0xf];
1627 		if (zero == 0 || (*cp != '0')) {
1628 			zero = 0;
1629 			cp++;
1630 		}
1631 		*cp = digits[*d >> 4];
1632 		if (zero == 0 || (*cp != '0')) {
1633 			zero = 0;
1634 			cp++;
1635 		}
1636 		*cp++ = digits[*d & 0xf];
1637 		*cp++ = ':';
1638 		a++;
1639 	}
1640 	*--cp = '\0';
1641 	return (ip6buf);
1642 }
1643 
1644 int
1645 in6_localaddr(struct in6_addr *in6)
1646 {
1647 	struct rm_priotracker in6_ifa_tracker;
1648 	struct in6_ifaddr *ia;
1649 
1650 	if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
1651 		return 1;
1652 
1653 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1654 	TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
1655 		if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
1656 		    &ia->ia_prefixmask.sin6_addr)) {
1657 			IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1658 			return 1;
1659 		}
1660 	}
1661 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1662 
1663 	return (0);
1664 }
1665 
1666 /*
1667  * Return 1 if an internet address is for the local host and configured
1668  * on one of its interfaces.
1669  */
1670 int
1671 in6_localip(struct in6_addr *in6)
1672 {
1673 	struct rm_priotracker in6_ifa_tracker;
1674 	struct in6_ifaddr *ia;
1675 
1676 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1677 	LIST_FOREACH(ia, IN6ADDR_HASH(in6), ia6_hash) {
1678 		if (IN6_ARE_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr)) {
1679 			IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1680 			return (1);
1681 		}
1682 	}
1683 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1684 	return (0);
1685 }
1686 
1687 /*
1688  * Return 1 if an internet address is configured on an interface.
1689  */
1690 int
1691 in6_ifhasaddr(struct ifnet *ifp, struct in6_addr *addr)
1692 {
1693 	struct in6_addr in6;
1694 	struct ifaddr *ifa;
1695 	struct in6_ifaddr *ia6;
1696 
1697 	in6 = *addr;
1698 	if (in6_clearscope(&in6))
1699 		return (0);
1700 	in6_setscope(&in6, ifp, NULL);
1701 
1702 	IF_ADDR_RLOCK(ifp);
1703 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1704 		if (ifa->ifa_addr->sa_family != AF_INET6)
1705 			continue;
1706 		ia6 = (struct in6_ifaddr *)ifa;
1707 		if (IN6_ARE_ADDR_EQUAL(&ia6->ia_addr.sin6_addr, &in6)) {
1708 			IF_ADDR_RUNLOCK(ifp);
1709 			return (1);
1710 		}
1711 	}
1712 	IF_ADDR_RUNLOCK(ifp);
1713 
1714 	return (0);
1715 }
1716 
1717 int
1718 in6_is_addr_deprecated(struct sockaddr_in6 *sa6)
1719 {
1720 	struct rm_priotracker in6_ifa_tracker;
1721 	struct in6_ifaddr *ia;
1722 
1723 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
1724 	LIST_FOREACH(ia, IN6ADDR_HASH(&sa6->sin6_addr), ia6_hash) {
1725 		if (IN6_ARE_ADDR_EQUAL(IA6_IN6(ia), &sa6->sin6_addr)) {
1726 			if (ia->ia6_flags & IN6_IFF_DEPRECATED) {
1727 				IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1728 				return (1); /* true */
1729 			}
1730 			break;
1731 		}
1732 	}
1733 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
1734 
1735 	return (0);		/* false */
1736 }
1737 
1738 /*
1739  * return length of part which dst and src are equal
1740  * hard coding...
1741  */
1742 int
1743 in6_matchlen(struct in6_addr *src, struct in6_addr *dst)
1744 {
1745 	int match = 0;
1746 	u_char *s = (u_char *)src, *d = (u_char *)dst;
1747 	u_char *lim = s + 16, r;
1748 
1749 	while (s < lim)
1750 		if ((r = (*d++ ^ *s++)) != 0) {
1751 			while (r < 128) {
1752 				match++;
1753 				r <<= 1;
1754 			}
1755 			break;
1756 		} else
1757 			match += 8;
1758 	return match;
1759 }
1760 
1761 /* XXX: to be scope conscious */
1762 int
1763 in6_are_prefix_equal(struct in6_addr *p1, struct in6_addr *p2, int len)
1764 {
1765 	int bytelen, bitlen;
1766 
1767 	/* sanity check */
1768 	if (0 > len || len > 128) {
1769 		log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n",
1770 		    len);
1771 		return (0);
1772 	}
1773 
1774 	bytelen = len / 8;
1775 	bitlen = len % 8;
1776 
1777 	if (bcmp(&p1->s6_addr, &p2->s6_addr, bytelen))
1778 		return (0);
1779 	if (bitlen != 0 &&
1780 	    p1->s6_addr[bytelen] >> (8 - bitlen) !=
1781 	    p2->s6_addr[bytelen] >> (8 - bitlen))
1782 		return (0);
1783 
1784 	return (1);
1785 }
1786 
1787 void
1788 in6_prefixlen2mask(struct in6_addr *maskp, int len)
1789 {
1790 	u_char maskarray[8] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
1791 	int bytelen, bitlen, i;
1792 
1793 	/* sanity check */
1794 	if (0 > len || len > 128) {
1795 		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
1796 		    len);
1797 		return;
1798 	}
1799 
1800 	bzero(maskp, sizeof(*maskp));
1801 	bytelen = len / 8;
1802 	bitlen = len % 8;
1803 	for (i = 0; i < bytelen; i++)
1804 		maskp->s6_addr[i] = 0xff;
1805 	if (bitlen)
1806 		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
1807 }
1808 
1809 /*
1810  * return the best address out of the same scope. if no address was
1811  * found, return the first valid address from designated IF.
1812  */
1813 struct in6_ifaddr *
1814 in6_ifawithifp(struct ifnet *ifp, struct in6_addr *dst)
1815 {
1816 	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
1817 	struct ifaddr *ifa;
1818 	struct in6_ifaddr *besta = NULL;
1819 	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
1820 
1821 	dep[0] = dep[1] = NULL;
1822 
1823 	/*
1824 	 * We first look for addresses in the same scope.
1825 	 * If there is one, return it.
1826 	 * If two or more, return one which matches the dst longest.
1827 	 * If none, return one of global addresses assigned other ifs.
1828 	 */
1829 	IF_ADDR_RLOCK(ifp);
1830 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1831 		if (ifa->ifa_addr->sa_family != AF_INET6)
1832 			continue;
1833 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
1834 			continue; /* XXX: is there any case to allow anycast? */
1835 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
1836 			continue; /* don't use this interface */
1837 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
1838 			continue;
1839 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
1840 			if (V_ip6_use_deprecated)
1841 				dep[0] = (struct in6_ifaddr *)ifa;
1842 			continue;
1843 		}
1844 
1845 		if (dst_scope == in6_addrscope(IFA_IN6(ifa))) {
1846 			/*
1847 			 * call in6_matchlen() as few as possible
1848 			 */
1849 			if (besta) {
1850 				if (blen == -1)
1851 					blen = in6_matchlen(&besta->ia_addr.sin6_addr, dst);
1852 				tlen = in6_matchlen(IFA_IN6(ifa), dst);
1853 				if (tlen > blen) {
1854 					blen = tlen;
1855 					besta = (struct in6_ifaddr *)ifa;
1856 				}
1857 			} else
1858 				besta = (struct in6_ifaddr *)ifa;
1859 		}
1860 	}
1861 	if (besta) {
1862 		ifa_ref(&besta->ia_ifa);
1863 		IF_ADDR_RUNLOCK(ifp);
1864 		return (besta);
1865 	}
1866 
1867 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1868 		if (ifa->ifa_addr->sa_family != AF_INET6)
1869 			continue;
1870 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
1871 			continue; /* XXX: is there any case to allow anycast? */
1872 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
1873 			continue; /* don't use this interface */
1874 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
1875 			continue;
1876 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
1877 			if (V_ip6_use_deprecated)
1878 				dep[1] = (struct in6_ifaddr *)ifa;
1879 			continue;
1880 		}
1881 
1882 		if (ifa != NULL)
1883 			ifa_ref(ifa);
1884 		IF_ADDR_RUNLOCK(ifp);
1885 		return (struct in6_ifaddr *)ifa;
1886 	}
1887 
1888 	/* use the last-resort values, that are, deprecated addresses */
1889 	if (dep[0]) {
1890 		ifa_ref((struct ifaddr *)dep[0]);
1891 		IF_ADDR_RUNLOCK(ifp);
1892 		return dep[0];
1893 	}
1894 	if (dep[1]) {
1895 		ifa_ref((struct ifaddr *)dep[1]);
1896 		IF_ADDR_RUNLOCK(ifp);
1897 		return dep[1];
1898 	}
1899 
1900 	IF_ADDR_RUNLOCK(ifp);
1901 	return NULL;
1902 }
1903 
1904 /*
1905  * perform DAD when interface becomes IFF_UP.
1906  */
1907 void
1908 in6_if_up(struct ifnet *ifp)
1909 {
1910 	struct ifaddr *ifa;
1911 	struct in6_ifaddr *ia;
1912 
1913 	IF_ADDR_RLOCK(ifp);
1914 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1915 		if (ifa->ifa_addr->sa_family != AF_INET6)
1916 			continue;
1917 		ia = (struct in6_ifaddr *)ifa;
1918 		if (ia->ia6_flags & IN6_IFF_TENTATIVE) {
1919 			/*
1920 			 * The TENTATIVE flag was likely set by hand
1921 			 * beforehand, implicitly indicating the need for DAD.
1922 			 * We may be able to skip the random delay in this
1923 			 * case, but we impose delays just in case.
1924 			 */
1925 			nd6_dad_start(ifa,
1926 			    arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz));
1927 		}
1928 	}
1929 	IF_ADDR_RUNLOCK(ifp);
1930 
1931 	/*
1932 	 * special cases, like 6to4, are handled in in6_ifattach
1933 	 */
1934 	in6_ifattach(ifp, NULL);
1935 }
1936 
1937 int
1938 in6if_do_dad(struct ifnet *ifp)
1939 {
1940 	if ((ifp->if_flags & IFF_LOOPBACK) != 0)
1941 		return (0);
1942 
1943 	if ((ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) ||
1944 	    (ND_IFINFO(ifp)->flags & ND6_IFF_NO_DAD))
1945 		return (0);
1946 
1947 	/*
1948 	 * Our DAD routine requires the interface up and running.
1949 	 * However, some interfaces can be up before the RUNNING
1950 	 * status.  Additionally, users may try to assign addresses
1951 	 * before the interface becomes up (or running).
1952 	 * This function returns EAGAIN in that case.
1953 	 * The caller should mark "tentative" on the address instead of
1954 	 * performing DAD immediately.
1955 	 */
1956 	if (!((ifp->if_flags & IFF_UP) &&
1957 	    (ifp->if_drv_flags & IFF_DRV_RUNNING)))
1958 		return (EAGAIN);
1959 
1960 	return (1);
1961 }
1962 
1963 /*
1964  * Calculate max IPv6 MTU through all the interfaces and store it
1965  * to in6_maxmtu.
1966  */
1967 void
1968 in6_setmaxmtu(void)
1969 {
1970 	unsigned long maxmtu = 0;
1971 	struct ifnet *ifp;
1972 
1973 	IFNET_RLOCK_NOSLEEP();
1974 	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1975 		/* this function can be called during ifnet initialization */
1976 		if (!ifp->if_afdata[AF_INET6])
1977 			continue;
1978 		if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
1979 		    IN6_LINKMTU(ifp) > maxmtu)
1980 			maxmtu = IN6_LINKMTU(ifp);
1981 	}
1982 	IFNET_RUNLOCK_NOSLEEP();
1983 	if (maxmtu)	/* update only when maxmtu is positive */
1984 		V_in6_maxmtu = maxmtu;
1985 }
1986 
1987 /*
1988  * Provide the length of interface identifiers to be used for the link attached
1989  * to the given interface.  The length should be defined in "IPv6 over
1990  * xxx-link" document.  Note that address architecture might also define
1991  * the length for a particular set of address prefixes, regardless of the
1992  * link type.  As clarified in rfc2462bis, those two definitions should be
1993  * consistent, and those really are as of August 2004.
1994  */
1995 int
1996 in6_if2idlen(struct ifnet *ifp)
1997 {
1998 	switch (ifp->if_type) {
1999 	case IFT_ETHER:		/* RFC2464 */
2000 	case IFT_PROPVIRTUAL:	/* XXX: no RFC. treat it as ether */
2001 	case IFT_L2VLAN:	/* ditto */
2002 	case IFT_IEEE80211:	/* ditto */
2003 	case IFT_BRIDGE:	/* bridge(4) only does Ethernet-like links */
2004 	case IFT_INFINIBAND:
2005 		return (64);
2006 	case IFT_FDDI:		/* RFC2467 */
2007 		return (64);
2008 	case IFT_ISO88025:	/* RFC2470 (IPv6 over Token Ring) */
2009 		return (64);
2010 	case IFT_PPP:		/* RFC2472 */
2011 		return (64);
2012 	case IFT_ARCNET:	/* RFC2497 */
2013 		return (64);
2014 	case IFT_FRELAY:	/* RFC2590 */
2015 		return (64);
2016 	case IFT_IEEE1394:	/* RFC3146 */
2017 		return (64);
2018 	case IFT_GIF:
2019 		return (64);	/* draft-ietf-v6ops-mech-v2-07 */
2020 	case IFT_LOOP:
2021 		return (64);	/* XXX: is this really correct? */
2022 	default:
2023 		/*
2024 		 * Unknown link type:
2025 		 * It might be controversial to use the today's common constant
2026 		 * of 64 for these cases unconditionally.  For full compliance,
2027 		 * we should return an error in this case.  On the other hand,
2028 		 * if we simply miss the standard for the link type or a new
2029 		 * standard is defined for a new link type, the IFID length
2030 		 * is very likely to be the common constant.  As a compromise,
2031 		 * we always use the constant, but make an explicit notice
2032 		 * indicating the "unknown" case.
2033 		 */
2034 		printf("in6_if2idlen: unknown link type (%d)\n", ifp->if_type);
2035 		return (64);
2036 	}
2037 }
2038 
2039 #include <sys/sysctl.h>
2040 
2041 struct in6_llentry {
2042 	struct llentry		base;
2043 };
2044 
2045 #define	IN6_LLTBL_DEFAULT_HSIZE	32
2046 #define	IN6_LLTBL_HASH(k, h) \
2047 	(((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1))
2048 
2049 /*
2050  * Do actual deallocation of @lle.
2051  */
2052 static void
2053 in6_lltable_destroy_lle_unlocked(struct llentry *lle)
2054 {
2055 
2056 	LLE_LOCK_DESTROY(lle);
2057 	LLE_REQ_DESTROY(lle);
2058 	free(lle, M_LLTABLE);
2059 }
2060 
2061 /*
2062  * Called by LLE_FREE_LOCKED when number of references
2063  * drops to zero.
2064  */
2065 static void
2066 in6_lltable_destroy_lle(struct llentry *lle)
2067 {
2068 
2069 	LLE_WUNLOCK(lle);
2070 	in6_lltable_destroy_lle_unlocked(lle);
2071 }
2072 
2073 static struct llentry *
2074 in6_lltable_new(const struct in6_addr *addr6, u_int flags)
2075 {
2076 	struct in6_llentry *lle;
2077 
2078 	lle = malloc(sizeof(struct in6_llentry), M_LLTABLE, M_NOWAIT | M_ZERO);
2079 	if (lle == NULL)		/* NB: caller generates msg */
2080 		return NULL;
2081 
2082 	lle->base.r_l3addr.addr6 = *addr6;
2083 	lle->base.lle_refcnt = 1;
2084 	lle->base.lle_free = in6_lltable_destroy_lle;
2085 	LLE_LOCK_INIT(&lle->base);
2086 	LLE_REQ_INIT(&lle->base);
2087 	callout_init(&lle->base.lle_timer, 1);
2088 
2089 	return (&lle->base);
2090 }
2091 
2092 static int
2093 in6_lltable_match_prefix(const struct sockaddr *saddr,
2094     const struct sockaddr *smask, u_int flags, struct llentry *lle)
2095 {
2096 	const struct in6_addr *addr, *mask, *lle_addr;
2097 
2098 	addr = &((const struct sockaddr_in6 *)saddr)->sin6_addr;
2099 	mask = &((const struct sockaddr_in6 *)smask)->sin6_addr;
2100 	lle_addr = &lle->r_l3addr.addr6;
2101 
2102 	if (IN6_ARE_MASKED_ADDR_EQUAL(lle_addr, addr, mask) == 0)
2103 		return (0);
2104 
2105 	if (lle->la_flags & LLE_IFADDR) {
2106 
2107 		/*
2108 		 * Delete LLE_IFADDR records IFF address & flag matches.
2109 		 * Note that addr is the interface address within prefix
2110 		 * being matched.
2111 		 */
2112 		if (IN6_ARE_ADDR_EQUAL(addr, lle_addr) &&
2113 		    (flags & LLE_STATIC) != 0)
2114 			return (1);
2115 		return (0);
2116 	}
2117 
2118 	/* flags & LLE_STATIC means deleting both dynamic and static entries */
2119 	if ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC))
2120 		return (1);
2121 
2122 	return (0);
2123 }
2124 
2125 static void
2126 in6_lltable_free_entry(struct lltable *llt, struct llentry *lle)
2127 {
2128 	struct ifnet *ifp;
2129 
2130 	LLE_WLOCK_ASSERT(lle);
2131 	KASSERT(llt != NULL, ("lltable is NULL"));
2132 
2133 	/* Unlink entry from table */
2134 	if ((lle->la_flags & LLE_LINKED) != 0) {
2135 
2136 		ifp = llt->llt_ifp;
2137 		IF_AFDATA_WLOCK_ASSERT(ifp);
2138 		lltable_unlink_entry(llt, lle);
2139 	}
2140 
2141 	if (callout_stop(&lle->lle_timer) > 0)
2142 		LLE_REMREF(lle);
2143 
2144 	llentry_free(lle);
2145 }
2146 
2147 static int
2148 in6_lltable_rtcheck(struct ifnet *ifp,
2149 		    u_int flags,
2150 		    const struct sockaddr *l3addr)
2151 {
2152 	const struct sockaddr_in6 *sin6;
2153 	struct nhop6_basic nh6;
2154 	struct in6_addr dst;
2155 	uint32_t scopeid;
2156 	int error;
2157 	char ip6buf[INET6_ADDRSTRLEN];
2158 
2159 	KASSERT(l3addr->sa_family == AF_INET6,
2160 	    ("sin_family %d", l3addr->sa_family));
2161 
2162 	/* Our local addresses are always only installed on the default FIB. */
2163 
2164 	sin6 = (const struct sockaddr_in6 *)l3addr;
2165 	in6_splitscope(&sin6->sin6_addr, &dst, &scopeid);
2166 	error = fib6_lookup_nh_basic(RT_DEFAULT_FIB, &dst, scopeid, 0, 0, &nh6);
2167 	if (error != 0 || (nh6.nh_flags & NHF_GATEWAY) || nh6.nh_ifp != ifp) {
2168 		struct ifaddr *ifa;
2169 		/*
2170 		 * Create an ND6 cache for an IPv6 neighbor
2171 		 * that is not covered by our own prefix.
2172 		 */
2173 		ifa = ifaof_ifpforaddr(l3addr, ifp);
2174 		if (ifa != NULL) {
2175 			ifa_free(ifa);
2176 			return 0;
2177 		}
2178 		log(LOG_INFO, "IPv6 address: \"%s\" is not on the network\n",
2179 		    ip6_sprintf(ip6buf, &sin6->sin6_addr));
2180 		return EINVAL;
2181 	}
2182 	return 0;
2183 }
2184 
2185 static inline uint32_t
2186 in6_lltable_hash_dst(const struct in6_addr *dst, uint32_t hsize)
2187 {
2188 
2189 	return (IN6_LLTBL_HASH(dst->s6_addr32[3], hsize));
2190 }
2191 
2192 static uint32_t
2193 in6_lltable_hash(const struct llentry *lle, uint32_t hsize)
2194 {
2195 
2196 	return (in6_lltable_hash_dst(&lle->r_l3addr.addr6, hsize));
2197 }
2198 
2199 static void
2200 in6_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
2201 {
2202 	struct sockaddr_in6 *sin6;
2203 
2204 	sin6 = (struct sockaddr_in6 *)sa;
2205 	bzero(sin6, sizeof(*sin6));
2206 	sin6->sin6_family = AF_INET6;
2207 	sin6->sin6_len = sizeof(*sin6);
2208 	sin6->sin6_addr = lle->r_l3addr.addr6;
2209 }
2210 
2211 static inline struct llentry *
2212 in6_lltable_find_dst(struct lltable *llt, const struct in6_addr *dst)
2213 {
2214 	struct llentry *lle;
2215 	struct llentries *lleh;
2216 	u_int hashidx;
2217 
2218 	hashidx = in6_lltable_hash_dst(dst, llt->llt_hsize);
2219 	lleh = &llt->lle_head[hashidx];
2220 	LIST_FOREACH(lle, lleh, lle_next) {
2221 		if (lle->la_flags & LLE_DELETED)
2222 			continue;
2223 		if (IN6_ARE_ADDR_EQUAL(&lle->r_l3addr.addr6, dst))
2224 			break;
2225 	}
2226 
2227 	return (lle);
2228 }
2229 
2230 static void
2231 in6_lltable_delete_entry(struct lltable *llt, struct llentry *lle)
2232 {
2233 
2234 	lle->la_flags |= LLE_DELETED;
2235 	EVENTHANDLER_INVOKE(lle_event, lle, LLENTRY_DELETED);
2236 #ifdef DIAGNOSTIC
2237 	log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle);
2238 #endif
2239 	llentry_free(lle);
2240 }
2241 
2242 static struct llentry *
2243 in6_lltable_alloc(struct lltable *llt, u_int flags,
2244 	const struct sockaddr *l3addr)
2245 {
2246 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2247 	struct ifnet *ifp = llt->llt_ifp;
2248 	struct llentry *lle;
2249 	char linkhdr[LLE_MAX_LINKHDR];
2250 	size_t linkhdrsize;
2251 	int lladdr_off;
2252 
2253 	KASSERT(l3addr->sa_family == AF_INET6,
2254 	    ("sin_family %d", l3addr->sa_family));
2255 
2256 	/*
2257 	 * A route that covers the given address must have
2258 	 * been installed 1st because we are doing a resolution,
2259 	 * verify this.
2260 	 */
2261 	if (!(flags & LLE_IFADDR) &&
2262 	    in6_lltable_rtcheck(ifp, flags, l3addr) != 0)
2263 		return (NULL);
2264 
2265 	lle = in6_lltable_new(&sin6->sin6_addr, flags);
2266 	if (lle == NULL) {
2267 		log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
2268 		return (NULL);
2269 	}
2270 	lle->la_flags = flags;
2271 	if ((flags & LLE_IFADDR) == LLE_IFADDR) {
2272 		linkhdrsize = LLE_MAX_LINKHDR;
2273 		if (lltable_calc_llheader(ifp, AF_INET6, IF_LLADDR(ifp),
2274 		    linkhdr, &linkhdrsize, &lladdr_off) != 0) {
2275 			in6_lltable_destroy_lle_unlocked(lle);
2276 			return (NULL);
2277 		}
2278 		lltable_set_entry_addr(ifp, lle, linkhdr, linkhdrsize,
2279 		    lladdr_off);
2280 		lle->la_flags |= LLE_STATIC;
2281 	}
2282 
2283 	if ((lle->la_flags & LLE_STATIC) != 0)
2284 		lle->ln_state = ND6_LLINFO_REACHABLE;
2285 
2286 	return (lle);
2287 }
2288 
2289 static struct llentry *
2290 in6_lltable_lookup(struct lltable *llt, u_int flags,
2291 	const struct sockaddr *l3addr)
2292 {
2293 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2294 	struct llentry *lle;
2295 
2296 	IF_AFDATA_LOCK_ASSERT(llt->llt_ifp);
2297 	KASSERT(l3addr->sa_family == AF_INET6,
2298 	    ("sin_family %d", l3addr->sa_family));
2299 
2300 	lle = in6_lltable_find_dst(llt, &sin6->sin6_addr);
2301 
2302 	if (lle == NULL)
2303 		return (NULL);
2304 
2305 	KASSERT((flags & (LLE_UNLOCKED|LLE_EXCLUSIVE)) !=
2306 	    (LLE_UNLOCKED|LLE_EXCLUSIVE),("wrong lle request flags: 0x%X",
2307 	    flags));
2308 
2309 	if (flags & LLE_UNLOCKED)
2310 		return (lle);
2311 
2312 	if (flags & LLE_EXCLUSIVE)
2313 		LLE_WLOCK(lle);
2314 	else
2315 		LLE_RLOCK(lle);
2316 	return (lle);
2317 }
2318 
2319 static int
2320 in6_lltable_dump_entry(struct lltable *llt, struct llentry *lle,
2321     struct sysctl_req *wr)
2322 {
2323 	struct ifnet *ifp = llt->llt_ifp;
2324 	/* XXX stack use */
2325 	struct {
2326 		struct rt_msghdr	rtm;
2327 		struct sockaddr_in6	sin6;
2328 		/*
2329 		 * ndp.c assumes that sdl is word aligned
2330 		 */
2331 #ifdef __LP64__
2332 		uint32_t		pad;
2333 #endif
2334 		struct sockaddr_dl	sdl;
2335 	} ndpc;
2336 	struct sockaddr_dl *sdl;
2337 	int error;
2338 
2339 	bzero(&ndpc, sizeof(ndpc));
2340 			/* skip deleted entries */
2341 			if ((lle->la_flags & LLE_DELETED) == LLE_DELETED)
2342 				return (0);
2343 			/* Skip if jailed and not a valid IP of the prison. */
2344 			lltable_fill_sa_entry(lle,
2345 			    (struct sockaddr *)&ndpc.sin6);
2346 			if (prison_if(wr->td->td_ucred,
2347 			    (struct sockaddr *)&ndpc.sin6) != 0)
2348 				return (0);
2349 			/*
2350 			 * produce a msg made of:
2351 			 *  struct rt_msghdr;
2352 			 *  struct sockaddr_in6 (IPv6)
2353 			 *  struct sockaddr_dl;
2354 			 */
2355 			ndpc.rtm.rtm_msglen = sizeof(ndpc);
2356 			ndpc.rtm.rtm_version = RTM_VERSION;
2357 			ndpc.rtm.rtm_type = RTM_GET;
2358 			ndpc.rtm.rtm_flags = RTF_UP;
2359 			ndpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
2360 			if (V_deembed_scopeid)
2361 				sa6_recoverscope(&ndpc.sin6);
2362 
2363 			/* publish */
2364 			if (lle->la_flags & LLE_PUB)
2365 				ndpc.rtm.rtm_flags |= RTF_ANNOUNCE;
2366 
2367 			sdl = &ndpc.sdl;
2368 			sdl->sdl_family = AF_LINK;
2369 			sdl->sdl_len = sizeof(*sdl);
2370 			sdl->sdl_alen = ifp->if_addrlen;
2371 			sdl->sdl_index = ifp->if_index;
2372 			sdl->sdl_type = ifp->if_type;
2373 			bcopy(lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
2374 			if (lle->la_expire != 0)
2375 				ndpc.rtm.rtm_rmx.rmx_expire = lle->la_expire +
2376 				    lle->lle_remtime / hz +
2377 				    time_second - time_uptime;
2378 			ndpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
2379 			if (lle->la_flags & LLE_STATIC)
2380 				ndpc.rtm.rtm_flags |= RTF_STATIC;
2381 			if (lle->la_flags & LLE_IFADDR)
2382 				ndpc.rtm.rtm_flags |= RTF_PINNED;
2383 			if (lle->ln_router != 0)
2384 				ndpc.rtm.rtm_flags |= RTF_GATEWAY;
2385 			ndpc.rtm.rtm_rmx.rmx_pksent = lle->la_asked;
2386 			/* Store state in rmx_weight value */
2387 			ndpc.rtm.rtm_rmx.rmx_state = lle->ln_state;
2388 			ndpc.rtm.rtm_index = ifp->if_index;
2389 			error = SYSCTL_OUT(wr, &ndpc, sizeof(ndpc));
2390 
2391 	return (error);
2392 }
2393 
2394 static struct lltable *
2395 in6_lltattach(struct ifnet *ifp)
2396 {
2397 	struct lltable *llt;
2398 
2399 	llt = lltable_allocate_htbl(IN6_LLTBL_DEFAULT_HSIZE);
2400 	llt->llt_af = AF_INET6;
2401 	llt->llt_ifp = ifp;
2402 
2403 	llt->llt_lookup = in6_lltable_lookup;
2404 	llt->llt_alloc_entry = in6_lltable_alloc;
2405 	llt->llt_delete_entry = in6_lltable_delete_entry;
2406 	llt->llt_dump_entry = in6_lltable_dump_entry;
2407 	llt->llt_hash = in6_lltable_hash;
2408 	llt->llt_fill_sa_entry = in6_lltable_fill_sa_entry;
2409 	llt->llt_free_entry = in6_lltable_free_entry;
2410 	llt->llt_match_prefix = in6_lltable_match_prefix;
2411  	lltable_link(llt);
2412 
2413 	return (llt);
2414 }
2415 
2416 void *
2417 in6_domifattach(struct ifnet *ifp)
2418 {
2419 	struct in6_ifextra *ext;
2420 
2421 	/* There are not IPv6-capable interfaces. */
2422 	switch (ifp->if_type) {
2423 	case IFT_PFLOG:
2424 	case IFT_PFSYNC:
2425 	case IFT_USB:
2426 		return (NULL);
2427 	}
2428 	ext = (struct in6_ifextra *)malloc(sizeof(*ext), M_IFADDR, M_WAITOK);
2429 	bzero(ext, sizeof(*ext));
2430 
2431 	ext->in6_ifstat = malloc(sizeof(counter_u64_t) *
2432 	    sizeof(struct in6_ifstat) / sizeof(uint64_t), M_IFADDR, M_WAITOK);
2433 	COUNTER_ARRAY_ALLOC(ext->in6_ifstat,
2434 	    sizeof(struct in6_ifstat) / sizeof(uint64_t), M_WAITOK);
2435 
2436 	ext->icmp6_ifstat = malloc(sizeof(counter_u64_t) *
2437 	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t), M_IFADDR,
2438 	    M_WAITOK);
2439 	COUNTER_ARRAY_ALLOC(ext->icmp6_ifstat,
2440 	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t), M_WAITOK);
2441 
2442 	ext->nd_ifinfo = nd6_ifattach(ifp);
2443 	ext->scope6_id = scope6_ifattach(ifp);
2444 	ext->lltable = in6_lltattach(ifp);
2445 
2446 	ext->mld_ifinfo = mld_domifattach(ifp);
2447 
2448 	return ext;
2449 }
2450 
2451 int
2452 in6_domifmtu(struct ifnet *ifp)
2453 {
2454 	if (ifp->if_afdata[AF_INET6] == NULL)
2455 		return ifp->if_mtu;
2456 
2457 	return (IN6_LINKMTU(ifp));
2458 }
2459 
2460 void
2461 in6_domifdetach(struct ifnet *ifp, void *aux)
2462 {
2463 	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
2464 
2465 	mld_domifdetach(ifp);
2466 	scope6_ifdetach(ext->scope6_id);
2467 	nd6_ifdetach(ext->nd_ifinfo);
2468 	lltable_free(ext->lltable);
2469 	COUNTER_ARRAY_FREE(ext->in6_ifstat,
2470 	    sizeof(struct in6_ifstat) / sizeof(uint64_t));
2471 	free(ext->in6_ifstat, M_IFADDR);
2472 	COUNTER_ARRAY_FREE(ext->icmp6_ifstat,
2473 	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t));
2474 	free(ext->icmp6_ifstat, M_IFADDR);
2475 	free(ext, M_IFADDR);
2476 }
2477 
2478 /*
2479  * Convert sockaddr_in6 to sockaddr_in.  Original sockaddr_in6 must be
2480  * v4 mapped addr or v4 compat addr
2481  */
2482 void
2483 in6_sin6_2_sin(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2484 {
2485 
2486 	bzero(sin, sizeof(*sin));
2487 	sin->sin_len = sizeof(struct sockaddr_in);
2488 	sin->sin_family = AF_INET;
2489 	sin->sin_port = sin6->sin6_port;
2490 	sin->sin_addr.s_addr = sin6->sin6_addr.s6_addr32[3];
2491 }
2492 
2493 /* Convert sockaddr_in to sockaddr_in6 in v4 mapped addr format. */
2494 void
2495 in6_sin_2_v4mapsin6(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2496 {
2497 	bzero(sin6, sizeof(*sin6));
2498 	sin6->sin6_len = sizeof(struct sockaddr_in6);
2499 	sin6->sin6_family = AF_INET6;
2500 	sin6->sin6_port = sin->sin_port;
2501 	sin6->sin6_addr.s6_addr32[0] = 0;
2502 	sin6->sin6_addr.s6_addr32[1] = 0;
2503 	sin6->sin6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
2504 	sin6->sin6_addr.s6_addr32[3] = sin->sin_addr.s_addr;
2505 }
2506 
2507 /* Convert sockaddr_in6 into sockaddr_in. */
2508 void
2509 in6_sin6_2_sin_in_sock(struct sockaddr *nam)
2510 {
2511 	struct sockaddr_in *sin_p;
2512 	struct sockaddr_in6 sin6;
2513 
2514 	/*
2515 	 * Save original sockaddr_in6 addr and convert it
2516 	 * to sockaddr_in.
2517 	 */
2518 	sin6 = *(struct sockaddr_in6 *)nam;
2519 	sin_p = (struct sockaddr_in *)nam;
2520 	in6_sin6_2_sin(sin_p, &sin6);
2521 }
2522 
2523 /* Convert sockaddr_in into sockaddr_in6 in v4 mapped addr format. */
2524 void
2525 in6_sin_2_v4mapsin6_in_sock(struct sockaddr **nam)
2526 {
2527 	struct sockaddr_in *sin_p;
2528 	struct sockaddr_in6 *sin6_p;
2529 
2530 	sin6_p = malloc(sizeof *sin6_p, M_SONAME, M_WAITOK);
2531 	sin_p = (struct sockaddr_in *)*nam;
2532 	in6_sin_2_v4mapsin6(sin_p, sin6_p);
2533 	free(*nam, M_SONAME);
2534 	*nam = (struct sockaddr *)sin6_p;
2535 }
2536