xref: /freebsd/sys/netinet6/in6.c (revision f18d3c411697ff46d85e579a72be54ca0cc67dd0)
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/syslog.h>
84 
85 #include <net/if.h>
86 #include <net/if_var.h>
87 #include <net/if_types.h>
88 #include <net/route.h>
89 #include <net/if_dl.h>
90 #include <net/vnet.h>
91 
92 #include <netinet/in.h>
93 #include <netinet/in_var.h>
94 #include <net/if_llatbl.h>
95 #include <netinet/if_ether.h>
96 #include <netinet/in_systm.h>
97 #include <netinet/ip.h>
98 #include <netinet/in_pcb.h>
99 #include <netinet/ip_carp.h>
100 
101 #include <netinet/ip6.h>
102 #include <netinet6/ip6_var.h>
103 #include <netinet6/nd6.h>
104 #include <netinet6/mld6_var.h>
105 #include <netinet6/ip6_mroute.h>
106 #include <netinet6/in6_ifattach.h>
107 #include <netinet6/scope6_var.h>
108 #include <netinet6/in6_pcb.h>
109 
110 VNET_DECLARE(int, icmp6_nodeinfo_oldmcprefix);
111 #define V_icmp6_nodeinfo_oldmcprefix	VNET(icmp6_nodeinfo_oldmcprefix)
112 
113 /*
114  * Definitions of some costant IP6 addresses.
115  */
116 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
117 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
118 const struct in6_addr in6addr_nodelocal_allnodes =
119 	IN6ADDR_NODELOCAL_ALLNODES_INIT;
120 const struct in6_addr in6addr_linklocal_allnodes =
121 	IN6ADDR_LINKLOCAL_ALLNODES_INIT;
122 const struct in6_addr in6addr_linklocal_allrouters =
123 	IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
124 const struct in6_addr in6addr_linklocal_allv2routers =
125 	IN6ADDR_LINKLOCAL_ALLV2ROUTERS_INIT;
126 
127 const struct in6_addr in6mask0 = IN6MASK0;
128 const struct in6_addr in6mask32 = IN6MASK32;
129 const struct in6_addr in6mask64 = IN6MASK64;
130 const struct in6_addr in6mask96 = IN6MASK96;
131 const struct in6_addr in6mask128 = IN6MASK128;
132 
133 const struct sockaddr_in6 sa6_any =
134 	{ sizeof(sa6_any), AF_INET6, 0, 0, IN6ADDR_ANY_INIT, 0 };
135 
136 static int in6_ifinit(struct ifnet *, struct in6_ifaddr *,
137 	struct sockaddr_in6 *, int);
138 static void in6_unlink_ifa(struct in6_ifaddr *, struct ifnet *);
139 
140 int	(*faithprefix_p)(struct in6_addr *);
141 
142 #define ifa2ia6(ifa)	((struct in6_ifaddr *)(ifa))
143 #define ia62ifa(ia6)	(&((ia6)->ia_ifa))
144 
145 void
146 in6_ifaddloop(struct ifaddr *ifa)
147 {
148 	struct sockaddr_dl gateway;
149 	struct sockaddr_in6 mask, addr;
150 	struct rtentry rt;
151 	struct in6_ifaddr *ia;
152 	struct ifnet *ifp;
153 	struct llentry *ln;
154 
155 	ia = ifa2ia6(ifa);
156 	ifp = ifa->ifa_ifp;
157 	IF_AFDATA_LOCK(ifp);
158 	ifa->ifa_rtrequest = nd6_rtrequest;
159 	ln = lla_lookup(LLTABLE6(ifp), (LLE_CREATE | LLE_IFADDR |
160 	    LLE_EXCLUSIVE), (struct sockaddr *)&ia->ia_addr);
161 	IF_AFDATA_UNLOCK(ifp);
162 	if (ln != NULL) {
163 		ln->la_expire = 0;  /* for IPv6 this means permanent */
164 		ln->ln_state = ND6_LLINFO_REACHABLE;
165 		/*
166 		 * initialize for rtmsg generation
167 		 */
168 		bzero(&gateway, sizeof(gateway));
169 		gateway.sdl_len = sizeof(gateway);
170 		gateway.sdl_family = AF_LINK;
171 		gateway.sdl_nlen = 0;
172 		gateway.sdl_alen = 6;
173 		memcpy(gateway.sdl_data, &ln->ll_addr.mac_aligned,
174 		    sizeof(ln->ll_addr));
175 		LLE_WUNLOCK(ln);
176 	}
177 
178 	bzero(&rt, sizeof(rt));
179 	rt.rt_gateway = (struct sockaddr *)&gateway;
180 	memcpy(&mask, &ia->ia_prefixmask, sizeof(ia->ia_prefixmask));
181 	memcpy(&addr, &ia->ia_addr, sizeof(ia->ia_addr));
182 	rt_mask(&rt) = (struct sockaddr *)&mask;
183 	rt_key(&rt) = (struct sockaddr *)&addr;
184 	rt.rt_flags = RTF_UP | RTF_HOST | RTF_STATIC;
185 	/* Announce arrival of local address to all FIBs. */
186 	rt_newaddrmsg(RTM_ADD, ifa, 0, &rt);
187 }
188 
189 void
190 in6_ifremloop(struct ifaddr *ifa)
191 {
192 	struct sockaddr_dl gateway;
193 	struct sockaddr_in6 mask, addr;
194 	struct rtentry rt0;
195 	struct in6_ifaddr *ia;
196 	struct ifnet *ifp;
197 
198 	ia = ifa2ia6(ifa);
199 	ifp = ifa->ifa_ifp;
200 	memcpy(&addr, &ia->ia_addr, sizeof(ia->ia_addr));
201 	memcpy(&mask, &ia->ia_prefixmask, sizeof(ia->ia_prefixmask));
202 	lltable_prefix_free(AF_INET6, (struct sockaddr *)&addr,
203 	            (struct sockaddr *)&mask, LLE_STATIC);
204 
205 	/*
206 	 * initialize for rtmsg generation
207 	 */
208 	bzero(&gateway, sizeof(gateway));
209 	gateway.sdl_len = sizeof(gateway);
210 	gateway.sdl_family = AF_LINK;
211 	gateway.sdl_nlen = 0;
212 	gateway.sdl_alen = ifp->if_addrlen;
213 	bzero(&rt0, sizeof(rt0));
214 	rt0.rt_gateway = (struct sockaddr *)&gateway;
215 	rt_mask(&rt0) = (struct sockaddr *)&mask;
216 	rt_key(&rt0) = (struct sockaddr *)&addr;
217 	rt0.rt_flags = RTF_HOST | RTF_STATIC;
218 	/* Announce removal of local address to all FIBs. */
219 	rt_newaddrmsg(RTM_DELETE, ifa, 0, &rt0);
220 }
221 
222 int
223 in6_mask2len(struct in6_addr *mask, u_char *lim0)
224 {
225 	int x = 0, y;
226 	u_char *lim = lim0, *p;
227 
228 	/* ignore the scope_id part */
229 	if (lim0 == NULL || lim0 - (u_char *)mask > sizeof(*mask))
230 		lim = (u_char *)mask + sizeof(*mask);
231 	for (p = (u_char *)mask; p < lim; x++, p++) {
232 		if (*p != 0xff)
233 			break;
234 	}
235 	y = 0;
236 	if (p < lim) {
237 		for (y = 0; y < 8; y++) {
238 			if ((*p & (0x80 >> y)) == 0)
239 				break;
240 		}
241 	}
242 
243 	/*
244 	 * when the limit pointer is given, do a stricter check on the
245 	 * remaining bits.
246 	 */
247 	if (p < lim) {
248 		if (y != 0 && (*p & (0x00ff >> y)) != 0)
249 			return (-1);
250 		for (p = p + 1; p < lim; p++)
251 			if (*p != 0)
252 				return (-1);
253 	}
254 
255 	return x * 8 + y;
256 }
257 
258 #ifdef COMPAT_FREEBSD32
259 struct in6_ndifreq32 {
260 	char ifname[IFNAMSIZ];
261 	uint32_t ifindex;
262 };
263 #define	SIOCGDEFIFACE32_IN6	_IOWR('i', 86, struct in6_ndifreq32)
264 #endif
265 
266 int
267 in6_control(struct socket *so, u_long cmd, caddr_t data,
268     struct ifnet *ifp, struct thread *td)
269 {
270 	struct	in6_ifreq *ifr = (struct in6_ifreq *)data;
271 	struct	in6_ifaddr *ia = NULL;
272 	struct	in6_aliasreq *ifra = (struct in6_aliasreq *)data;
273 	struct sockaddr_in6 *sa6;
274 	int carp_attached = 0;
275 	int error;
276 	u_long ocmd = cmd;
277 
278 	/*
279 	 * Compat to make pre-10.x ifconfig(8) operable.
280 	 */
281 	if (cmd == OSIOCAIFADDR_IN6)
282 		cmd = SIOCAIFADDR_IN6;
283 
284 	switch (cmd) {
285 	case SIOCGETSGCNT_IN6:
286 	case SIOCGETMIFCNT_IN6:
287 		/*
288 		 * XXX mrt_ioctl has a 3rd, unused, FIB argument in route.c.
289 		 * We cannot see how that would be needed, so do not adjust the
290 		 * KPI blindly; more likely should clean up the IPv4 variant.
291 		 */
292 		return (mrt6_ioctl ? mrt6_ioctl(cmd, data) : EOPNOTSUPP);
293 	}
294 
295 	switch(cmd) {
296 	case SIOCAADDRCTL_POLICY:
297 	case SIOCDADDRCTL_POLICY:
298 		if (td != NULL) {
299 			error = priv_check(td, PRIV_NETINET_ADDRCTRL6);
300 			if (error)
301 				return (error);
302 		}
303 		return (in6_src_ioctl(cmd, data));
304 	}
305 
306 	if (ifp == NULL)
307 		return (EOPNOTSUPP);
308 
309 	switch (cmd) {
310 	case SIOCSNDFLUSH_IN6:
311 	case SIOCSPFXFLUSH_IN6:
312 	case SIOCSRTRFLUSH_IN6:
313 	case SIOCSDEFIFACE_IN6:
314 	case SIOCSIFINFO_FLAGS:
315 	case SIOCSIFINFO_IN6:
316 		if (td != NULL) {
317 			error = priv_check(td, PRIV_NETINET_ND6);
318 			if (error)
319 				return (error);
320 		}
321 		/* FALLTHROUGH */
322 	case OSIOCGIFINFO_IN6:
323 	case SIOCGIFINFO_IN6:
324 	case SIOCGDRLST_IN6:
325 	case SIOCGPRLST_IN6:
326 	case SIOCGNBRINFO_IN6:
327 	case SIOCGDEFIFACE_IN6:
328 		return (nd6_ioctl(cmd, data, ifp));
329 
330 #ifdef COMPAT_FREEBSD32
331 	case SIOCGDEFIFACE32_IN6:
332 		{
333 			struct in6_ndifreq ndif;
334 			struct in6_ndifreq32 *ndif32;
335 
336 			error = nd6_ioctl(SIOCGDEFIFACE_IN6, (caddr_t)&ndif,
337 			    ifp);
338 			if (error)
339 				return (error);
340 			ndif32 = (struct in6_ndifreq32 *)data;
341 			ndif32->ifindex = ndif.ifindex;
342 			return (0);
343 		}
344 #endif
345 	}
346 
347 	switch (cmd) {
348 	case SIOCSIFPREFIX_IN6:
349 	case SIOCDIFPREFIX_IN6:
350 	case SIOCAIFPREFIX_IN6:
351 	case SIOCCIFPREFIX_IN6:
352 	case SIOCSGIFPREFIX_IN6:
353 	case SIOCGIFPREFIX_IN6:
354 		log(LOG_NOTICE,
355 		    "prefix ioctls are now invalidated. "
356 		    "please use ifconfig.\n");
357 		return (EOPNOTSUPP);
358 	}
359 
360 	switch (cmd) {
361 	case SIOCSSCOPE6:
362 		if (td != NULL) {
363 			error = priv_check(td, PRIV_NETINET_SCOPE6);
364 			if (error)
365 				return (error);
366 		}
367 		return (scope6_set(ifp,
368 		    (struct scope6_id *)ifr->ifr_ifru.ifru_scope_id));
369 	case SIOCGSCOPE6:
370 		return (scope6_get(ifp,
371 		    (struct scope6_id *)ifr->ifr_ifru.ifru_scope_id));
372 	case SIOCGSCOPE6DEF:
373 		return (scope6_get_default((struct scope6_id *)
374 		    ifr->ifr_ifru.ifru_scope_id));
375 	}
376 
377 	/*
378 	 * Find address for this interface, if it exists.
379 	 *
380 	 * In netinet code, we have checked ifra_addr in SIOCSIF*ADDR operation
381 	 * only, and used the first interface address as the target of other
382 	 * operations (without checking ifra_addr).  This was because netinet
383 	 * code/API assumed at most 1 interface address per interface.
384 	 * Since IPv6 allows a node to assign multiple addresses
385 	 * on a single interface, we almost always look and check the
386 	 * presence of ifra_addr, and reject invalid ones here.
387 	 * It also decreases duplicated code among SIOC*_IN6 operations.
388 	 */
389 	switch (cmd) {
390 	case SIOCAIFADDR_IN6:
391 	case SIOCSIFPHYADDR_IN6:
392 		sa6 = &ifra->ifra_addr;
393 		break;
394 	case SIOCSIFADDR_IN6:
395 	case SIOCGIFADDR_IN6:
396 	case SIOCSIFDSTADDR_IN6:
397 	case SIOCSIFNETMASK_IN6:
398 	case SIOCGIFDSTADDR_IN6:
399 	case SIOCGIFNETMASK_IN6:
400 	case SIOCDIFADDR_IN6:
401 	case SIOCGIFPSRCADDR_IN6:
402 	case SIOCGIFPDSTADDR_IN6:
403 	case SIOCGIFAFLAG_IN6:
404 	case SIOCSNDFLUSH_IN6:
405 	case SIOCSPFXFLUSH_IN6:
406 	case SIOCSRTRFLUSH_IN6:
407 	case SIOCGIFALIFETIME_IN6:
408 	case SIOCSIFALIFETIME_IN6:
409 	case SIOCGIFSTAT_IN6:
410 	case SIOCGIFSTAT_ICMP6:
411 		sa6 = &ifr->ifr_addr;
412 		break;
413 	case SIOCSIFADDR:
414 	case SIOCSIFBRDADDR:
415 	case SIOCSIFDSTADDR:
416 	case SIOCSIFNETMASK:
417 		/*
418 		 * Although we should pass any non-INET6 ioctl requests
419 		 * down to driver, we filter some legacy INET requests.
420 		 * Drivers trust SIOCSIFADDR et al to come from an already
421 		 * privileged layer, and do not perform any credentials
422 		 * checks or input validation.
423 		 */
424 		return (EINVAL);
425 	default:
426 		sa6 = NULL;
427 		break;
428 	}
429 	if (sa6 && sa6->sin6_family == AF_INET6) {
430 		if (sa6->sin6_scope_id != 0)
431 			error = sa6_embedscope(sa6, 0);
432 		else
433 			error = in6_setscope(&sa6->sin6_addr, ifp, NULL);
434 		if (error != 0)
435 			return (error);
436 		if (td != NULL && (error = prison_check_ip6(td->td_ucred,
437 		    &sa6->sin6_addr)) != 0)
438 			return (error);
439 		ia = in6ifa_ifpwithaddr(ifp, &sa6->sin6_addr);
440 	} else
441 		ia = NULL;
442 
443 	switch (cmd) {
444 	case SIOCSIFADDR_IN6:
445 	case SIOCSIFDSTADDR_IN6:
446 	case SIOCSIFNETMASK_IN6:
447 		/*
448 		 * Since IPv6 allows a node to assign multiple addresses
449 		 * on a single interface, SIOCSIFxxx ioctls are deprecated.
450 		 */
451 		/* we decided to obsolete this command (20000704) */
452 		error = EINVAL;
453 		goto out;
454 
455 	case SIOCDIFADDR_IN6:
456 		/*
457 		 * for IPv4, we look for existing in_ifaddr here to allow
458 		 * "ifconfig if0 delete" to remove the first IPv4 address on
459 		 * the interface.  For IPv6, as the spec allows multiple
460 		 * interface address from the day one, we consider "remove the
461 		 * first one" semantics to be not preferable.
462 		 */
463 		if (ia == NULL) {
464 			error = EADDRNOTAVAIL;
465 			goto out;
466 		}
467 		/* FALLTHROUGH */
468 	case SIOCAIFADDR_IN6:
469 		/*
470 		 * We always require users to specify a valid IPv6 address for
471 		 * the corresponding operation.
472 		 */
473 		if (ifra->ifra_addr.sin6_family != AF_INET6 ||
474 		    ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6)) {
475 			error = EAFNOSUPPORT;
476 			goto out;
477 		}
478 
479 		if (td != NULL) {
480 			error = priv_check(td, (cmd == SIOCDIFADDR_IN6) ?
481 			    PRIV_NET_DELIFADDR : PRIV_NET_ADDIFADDR);
482 			if (error)
483 				goto out;
484 		}
485 		break;
486 
487 	case SIOCGIFADDR_IN6:
488 		/* This interface is basically deprecated. use SIOCGIFCONF. */
489 		/* FALLTHROUGH */
490 	case SIOCGIFAFLAG_IN6:
491 	case SIOCGIFNETMASK_IN6:
492 	case SIOCGIFDSTADDR_IN6:
493 	case SIOCGIFALIFETIME_IN6:
494 		/* must think again about its semantics */
495 		if (ia == NULL) {
496 			error = EADDRNOTAVAIL;
497 			goto out;
498 		}
499 		break;
500 
501 	case SIOCSIFALIFETIME_IN6:
502 	    {
503 		struct in6_addrlifetime *lt;
504 
505 		if (td != NULL) {
506 			error = priv_check(td, PRIV_NETINET_ALIFETIME6);
507 			if (error)
508 				goto out;
509 		}
510 		if (ia == NULL) {
511 			error = EADDRNOTAVAIL;
512 			goto out;
513 		}
514 		/* sanity for overflow - beware unsigned */
515 		lt = &ifr->ifr_ifru.ifru_lifetime;
516 		if (lt->ia6t_vltime != ND6_INFINITE_LIFETIME &&
517 		    lt->ia6t_vltime + time_uptime < time_uptime) {
518 			error = EINVAL;
519 			goto out;
520 		}
521 		if (lt->ia6t_pltime != ND6_INFINITE_LIFETIME &&
522 		    lt->ia6t_pltime + time_uptime < time_uptime) {
523 			error = EINVAL;
524 			goto out;
525 		}
526 		break;
527 	    }
528 	}
529 
530 	switch (cmd) {
531 	case SIOCGIFADDR_IN6:
532 		ifr->ifr_addr = ia->ia_addr;
533 		if ((error = sa6_recoverscope(&ifr->ifr_addr)) != 0)
534 			goto out;
535 		break;
536 
537 	case SIOCGIFDSTADDR_IN6:
538 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0) {
539 			error = EINVAL;
540 			goto out;
541 		}
542 		/*
543 		 * XXX: should we check if ifa_dstaddr is NULL and return
544 		 * an error?
545 		 */
546 		ifr->ifr_dstaddr = ia->ia_dstaddr;
547 		if ((error = sa6_recoverscope(&ifr->ifr_dstaddr)) != 0)
548 			goto out;
549 		break;
550 
551 	case SIOCGIFNETMASK_IN6:
552 		ifr->ifr_addr = ia->ia_prefixmask;
553 		break;
554 
555 	case SIOCGIFAFLAG_IN6:
556 		ifr->ifr_ifru.ifru_flags6 = ia->ia6_flags;
557 		break;
558 
559 	case SIOCGIFSTAT_IN6:
560 		if (ifp == NULL) {
561 			error = EINVAL;
562 			goto out;
563 		}
564 		COUNTER_ARRAY_COPY(((struct in6_ifextra *)
565 		    ifp->if_afdata[AF_INET6])->in6_ifstat,
566 		    &ifr->ifr_ifru.ifru_stat,
567 		    sizeof(struct in6_ifstat) / sizeof(uint64_t));
568 		break;
569 
570 	case SIOCGIFSTAT_ICMP6:
571 		if (ifp == NULL) {
572 			error = EINVAL;
573 			goto out;
574 		}
575 		COUNTER_ARRAY_COPY(((struct in6_ifextra *)
576 		    ifp->if_afdata[AF_INET6])->icmp6_ifstat,
577 		    &ifr->ifr_ifru.ifru_icmp6stat,
578 		    sizeof(struct icmp6_ifstat) / sizeof(uint64_t));
579 		break;
580 
581 	case SIOCGIFALIFETIME_IN6:
582 		ifr->ifr_ifru.ifru_lifetime = ia->ia6_lifetime;
583 		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
584 			time_t maxexpire;
585 			struct in6_addrlifetime *retlt =
586 			    &ifr->ifr_ifru.ifru_lifetime;
587 
588 			/*
589 			 * XXX: adjust expiration time assuming time_t is
590 			 * signed.
591 			 */
592 			maxexpire = (-1) &
593 			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
594 			if (ia->ia6_lifetime.ia6t_vltime <
595 			    maxexpire - ia->ia6_updatetime) {
596 				retlt->ia6t_expire = ia->ia6_updatetime +
597 				    ia->ia6_lifetime.ia6t_vltime;
598 			} else
599 				retlt->ia6t_expire = maxexpire;
600 		}
601 		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
602 			time_t maxexpire;
603 			struct in6_addrlifetime *retlt =
604 			    &ifr->ifr_ifru.ifru_lifetime;
605 
606 			/*
607 			 * XXX: adjust expiration time assuming time_t is
608 			 * signed.
609 			 */
610 			maxexpire = (-1) &
611 			    ~((time_t)1 << ((sizeof(maxexpire) * 8) - 1));
612 			if (ia->ia6_lifetime.ia6t_pltime <
613 			    maxexpire - ia->ia6_updatetime) {
614 				retlt->ia6t_preferred = ia->ia6_updatetime +
615 				    ia->ia6_lifetime.ia6t_pltime;
616 			} else
617 				retlt->ia6t_preferred = maxexpire;
618 		}
619 		break;
620 
621 	case SIOCSIFALIFETIME_IN6:
622 		ia->ia6_lifetime = ifr->ifr_ifru.ifru_lifetime;
623 		/* for sanity */
624 		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
625 			ia->ia6_lifetime.ia6t_expire =
626 				time_uptime + ia->ia6_lifetime.ia6t_vltime;
627 		} else
628 			ia->ia6_lifetime.ia6t_expire = 0;
629 		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
630 			ia->ia6_lifetime.ia6t_preferred =
631 				time_uptime + ia->ia6_lifetime.ia6t_pltime;
632 		} else
633 			ia->ia6_lifetime.ia6t_preferred = 0;
634 		break;
635 
636 	case SIOCAIFADDR_IN6:
637 	{
638 		int i;
639 		struct nd_prefixctl pr0;
640 		struct nd_prefix *pr;
641 
642 		/*
643 		 * first, make or update the interface address structure,
644 		 * and link it to the list.
645 		 */
646 		if ((error = in6_update_ifa(ifp, ifra, ia, 0)) != 0)
647 			goto out;
648 		if (ia != NULL)
649 			ifa_free(&ia->ia_ifa);
650 		if ((ia = in6ifa_ifpwithaddr(ifp, &ifra->ifra_addr.sin6_addr))
651 		    == NULL) {
652 			/*
653 			 * this can happen when the user specify the 0 valid
654 			 * lifetime.
655 			 */
656 			break;
657 		}
658 
659 		if (cmd == ocmd && ifra->ifra_vhid > 0) {
660 			if (carp_attach_p != NULL)
661 				error = (*carp_attach_p)(&ia->ia_ifa,
662 				    ifra->ifra_vhid);
663 			else
664 				error = EPROTONOSUPPORT;
665 			if (error)
666 				goto out;
667 			else
668 				carp_attached = 1;
669 		}
670 
671 		/*
672 		 * then, make the prefix on-link on the interface.
673 		 * XXX: we'd rather create the prefix before the address, but
674 		 * we need at least one address to install the corresponding
675 		 * interface route, so we configure the address first.
676 		 */
677 
678 		/*
679 		 * convert mask to prefix length (prefixmask has already
680 		 * been validated in in6_update_ifa().
681 		 */
682 		bzero(&pr0, sizeof(pr0));
683 		pr0.ndpr_ifp = ifp;
684 		pr0.ndpr_plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
685 		    NULL);
686 		if (pr0.ndpr_plen == 128) {
687 			break;	/* we don't need to install a host route. */
688 		}
689 		pr0.ndpr_prefix = ifra->ifra_addr;
690 		/* apply the mask for safety. */
691 		for (i = 0; i < 4; i++) {
692 			pr0.ndpr_prefix.sin6_addr.s6_addr32[i] &=
693 			    ifra->ifra_prefixmask.sin6_addr.s6_addr32[i];
694 		}
695 		/*
696 		 * XXX: since we don't have an API to set prefix (not address)
697 		 * lifetimes, we just use the same lifetimes as addresses.
698 		 * The (temporarily) installed lifetimes can be overridden by
699 		 * later advertised RAs (when accept_rtadv is non 0), which is
700 		 * an intended behavior.
701 		 */
702 		pr0.ndpr_raf_onlink = 1; /* should be configurable? */
703 		pr0.ndpr_raf_auto =
704 		    ((ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0);
705 		pr0.ndpr_vltime = ifra->ifra_lifetime.ia6t_vltime;
706 		pr0.ndpr_pltime = ifra->ifra_lifetime.ia6t_pltime;
707 
708 		/* add the prefix if not yet. */
709 		if ((pr = nd6_prefix_lookup(&pr0)) == NULL) {
710 			/*
711 			 * nd6_prelist_add will install the corresponding
712 			 * interface route.
713 			 */
714 			if ((error = nd6_prelist_add(&pr0, NULL, &pr)) != 0) {
715 				if (carp_attached)
716 					(*carp_detach_p)(&ia->ia_ifa);
717 				goto out;
718 			}
719 			if (pr == NULL) {
720 				if (carp_attached)
721 					(*carp_detach_p)(&ia->ia_ifa);
722 				log(LOG_ERR, "nd6_prelist_add succeeded but "
723 				    "no prefix\n");
724 				error = EINVAL;
725 				goto out;
726 			}
727 		}
728 
729 		/* relate the address to the prefix */
730 		if (ia->ia6_ndpr == NULL) {
731 			ia->ia6_ndpr = pr;
732 			pr->ndpr_refcnt++;
733 
734 			/*
735 			 * If this is the first autoconf address from the
736 			 * prefix, create a temporary address as well
737 			 * (when required).
738 			 */
739 			if ((ia->ia6_flags & IN6_IFF_AUTOCONF) &&
740 			    V_ip6_use_tempaddr && pr->ndpr_refcnt == 1) {
741 				int e;
742 				if ((e = in6_tmpifadd(ia, 1, 0)) != 0) {
743 					log(LOG_NOTICE, "in6_control: failed "
744 					    "to create a temporary address, "
745 					    "errno=%d\n", e);
746 				}
747 			}
748 		}
749 
750 		/*
751 		 * this might affect the status of autoconfigured addresses,
752 		 * that is, this address might make other addresses detached.
753 		 */
754 		pfxlist_onlink_check();
755 		if (error == 0 && ia) {
756 			if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) {
757 				/*
758 				 * Try to clear the flag when a new
759 				 * IPv6 address is added onto an
760 				 * IFDISABLED interface and it
761 				 * succeeds.
762 				 */
763 				struct in6_ndireq nd;
764 
765 				memset(&nd, 0, sizeof(nd));
766 				nd.ndi.flags = ND_IFINFO(ifp)->flags;
767 				nd.ndi.flags &= ~ND6_IFF_IFDISABLED;
768 				if (nd6_ioctl(SIOCSIFINFO_FLAGS,
769 				    (caddr_t)&nd, ifp) < 0)
770 					log(LOG_NOTICE, "SIOCAIFADDR_IN6: "
771 					    "SIOCSIFINFO_FLAGS for -ifdisabled "
772 					    "failed.");
773 				/*
774 				 * Ignore failure of clearing the flag
775 				 * intentionally.  The failure means
776 				 * address duplication was detected.
777 				 */
778 			}
779 			EVENTHANDLER_INVOKE(ifaddr_event, ifp);
780 		}
781 		break;
782 	}
783 
784 	case SIOCDIFADDR_IN6:
785 	{
786 		struct nd_prefix *pr;
787 
788 		/*
789 		 * If the address being deleted is the only one that owns
790 		 * the corresponding prefix, expire the prefix as well.
791 		 * XXX: theoretically, we don't have to worry about such
792 		 * relationship, since we separate the address management
793 		 * and the prefix management.  We do this, however, to provide
794 		 * as much backward compatibility as possible in terms of
795 		 * the ioctl operation.
796 		 * Note that in6_purgeaddr() will decrement ndpr_refcnt.
797 		 */
798 		pr = ia->ia6_ndpr;
799 		in6_purgeaddr(&ia->ia_ifa);
800 		if (pr && pr->ndpr_refcnt == 0)
801 			prelist_remove(pr);
802 		EVENTHANDLER_INVOKE(ifaddr_event, ifp);
803 		break;
804 	}
805 
806 	default:
807 		if (ifp == NULL || ifp->if_ioctl == 0) {
808 			error = EOPNOTSUPP;
809 			goto out;
810 		}
811 		error = (*ifp->if_ioctl)(ifp, cmd, data);
812 		goto out;
813 	}
814 
815 	error = 0;
816 out:
817 	if (ia != NULL)
818 		ifa_free(&ia->ia_ifa);
819 	return (error);
820 }
821 
822 
823 /*
824  * Join necessary multicast groups.  Factored out from in6_update_ifa().
825  * This entire work should only be done once, for the default FIB.
826  */
827 static int
828 in6_update_ifa_join_mc(struct ifnet *ifp, struct in6_aliasreq *ifra,
829     struct in6_ifaddr *ia, int flags, struct in6_multi **in6m_sol)
830 {
831 	char ip6buf[INET6_ADDRSTRLEN];
832 	struct sockaddr_in6 mltaddr, mltmask;
833 	struct in6_addr llsol;
834 	struct in6_multi_mship *imm;
835 	struct rtentry *rt;
836 	int delay, error;
837 
838 	KASSERT(in6m_sol != NULL, ("%s: in6m_sol is NULL", __func__));
839 
840 	/* Join solicited multicast addr for new host id. */
841 	bzero(&llsol, sizeof(struct in6_addr));
842 	llsol.s6_addr32[0] = IPV6_ADDR_INT32_MLL;
843 	llsol.s6_addr32[1] = 0;
844 	llsol.s6_addr32[2] = htonl(1);
845 	llsol.s6_addr32[3] = ifra->ifra_addr.sin6_addr.s6_addr32[3];
846 	llsol.s6_addr8[12] = 0xff;
847 	if ((error = in6_setscope(&llsol, ifp, NULL)) != 0) {
848 		/* XXX: should not happen */
849 		log(LOG_ERR, "%s: in6_setscope failed\n", __func__);
850 		goto cleanup;
851 	}
852 	delay = 0;
853 	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
854 		/*
855 		 * We need a random delay for DAD on the address being
856 		 * configured.  It also means delaying transmission of the
857 		 * corresponding MLD report to avoid report collision.
858 		 * [RFC 4861, Section 6.3.7]
859 		 */
860 		delay = arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz);
861 	}
862 	imm = in6_joingroup(ifp, &llsol, &error, delay);
863 	if (imm == NULL) {
864 		nd6log((LOG_WARNING, "%s: addmulti failed for %s on %s "
865 		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf, &llsol),
866 		    if_name(ifp), error));
867 		goto cleanup;
868 	}
869 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
870 	*in6m_sol = imm->i6mm_maddr;
871 
872 	bzero(&mltmask, sizeof(mltmask));
873 	mltmask.sin6_len = sizeof(struct sockaddr_in6);
874 	mltmask.sin6_family = AF_INET6;
875 	mltmask.sin6_addr = in6mask32;
876 #define	MLTMASK_LEN  4	/* mltmask's masklen (=32bit=4octet) */
877 
878 	/*
879 	 * Join link-local all-nodes address.
880 	 */
881 	bzero(&mltaddr, sizeof(mltaddr));
882 	mltaddr.sin6_len = sizeof(struct sockaddr_in6);
883 	mltaddr.sin6_family = AF_INET6;
884 	mltaddr.sin6_addr = in6addr_linklocal_allnodes;
885 	if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) != 0)
886 		goto cleanup; /* XXX: should not fail */
887 
888 	/*
889 	 * XXX: do we really need this automatic routes?  We should probably
890 	 * reconsider this stuff.  Most applications actually do not need the
891 	 * routes, since they usually specify the outgoing interface.
892 	 */
893 	rt = in6_rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL, RT_DEFAULT_FIB);
894 	if (rt != NULL) {
895 		/* XXX: only works in !SCOPEDROUTING case. */
896 		if (memcmp(&mltaddr.sin6_addr,
897 		    &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
898 		    MLTMASK_LEN)) {
899 			RTFREE_LOCKED(rt);
900 			rt = NULL;
901 		}
902 	}
903 	if (rt == NULL) {
904 		error = in6_rtrequest(RTM_ADD, (struct sockaddr *)&mltaddr,
905 		    (struct sockaddr *)&ia->ia_addr,
906 		    (struct sockaddr *)&mltmask, RTF_UP,
907 		    (struct rtentry **)0, RT_DEFAULT_FIB);
908 		if (error)
909 			goto cleanup;
910 	} else
911 		RTFREE_LOCKED(rt);
912 
913 	imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
914 	if (imm == NULL) {
915 		nd6log((LOG_WARNING, "%s: addmulti failed for %s on %s "
916 		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
917 		    &mltaddr.sin6_addr), if_name(ifp), error));
918 		goto cleanup;
919 	}
920 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
921 
922 	/*
923 	 * Join node information group address.
924 	 */
925 	delay = 0;
926 	if ((flags & IN6_IFAUPDATE_DADDELAY)) {
927 		/*
928 		 * The spec does not say anything about delay for this group,
929 		 * but the same logic should apply.
930 		 */
931 		delay = arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz);
932 	}
933 	if (in6_nigroup(ifp, NULL, -1, &mltaddr.sin6_addr) == 0) {
934 		/* XXX jinmei */
935 		imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, delay);
936 		if (imm == NULL)
937 			nd6log((LOG_WARNING, "%s: addmulti failed for %s on %s "
938 			    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
939 			    &mltaddr.sin6_addr), if_name(ifp), error));
940 			/* XXX not very fatal, go on... */
941 		else
942 			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
943 	}
944 	if (V_icmp6_nodeinfo_oldmcprefix &&
945 	     in6_nigroup_oldmcprefix(ifp, NULL, -1, &mltaddr.sin6_addr) == 0) {
946 		imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, delay);
947 		if (imm == NULL)
948 			nd6log((LOG_WARNING, "%s: addmulti failed for %s on %s "
949 			    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
950 			    &mltaddr.sin6_addr), if_name(ifp), error));
951 			/* XXX not very fatal, go on... */
952 		else
953 			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
954 	}
955 
956 	/*
957 	 * Join interface-local all-nodes address.
958 	 * (ff01::1%ifN, and ff01::%ifN/32)
959 	 */
960 	mltaddr.sin6_addr = in6addr_nodelocal_allnodes;
961 	if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) != 0)
962 		goto cleanup; /* XXX: should not fail */
963 	/* XXX: again, do we really need the route? */
964 	rt = in6_rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL, RT_DEFAULT_FIB);
965 	if (rt != NULL) {
966 		if (memcmp(&mltaddr.sin6_addr,
967 		    &((struct sockaddr_in6 *)rt_key(rt))->sin6_addr,
968 		    MLTMASK_LEN)) {
969 			RTFREE_LOCKED(rt);
970 			rt = NULL;
971 		}
972 	}
973 	if (rt == NULL) {
974 		error = in6_rtrequest(RTM_ADD, (struct sockaddr *)&mltaddr,
975 		    (struct sockaddr *)&ia->ia_addr,
976 		    (struct sockaddr *)&mltmask, RTF_UP,
977 		    (struct rtentry **)0, RT_DEFAULT_FIB);
978 		if (error)
979 			goto cleanup;
980 	} else
981 		RTFREE_LOCKED(rt);
982 
983 	imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
984 	if (imm == NULL) {
985 		nd6log((LOG_WARNING, "%s: addmulti failed for %s on %s "
986 		    "(errno=%d)\n", __func__, ip6_sprintf(ip6buf,
987 		    &mltaddr.sin6_addr), if_name(ifp), error));
988 		goto cleanup;
989 	}
990 	LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
991 #undef	MLTMASK_LEN
992 
993 cleanup:
994 	return (error);
995 }
996 
997 /*
998  * Update parameters of an IPv6 interface address.
999  * If necessary, a new entry is created and linked into address chains.
1000  * This function is separated from in6_control().
1001  */
1002 int
1003 in6_update_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
1004     struct in6_ifaddr *ia, int flags)
1005 {
1006 	int error = 0, hostIsNew = 0, plen = -1;
1007 	struct sockaddr_in6 dst6;
1008 	struct in6_addrlifetime *lt;
1009 	struct in6_multi *in6m_sol;
1010 	int delay;
1011 	char ip6buf[INET6_ADDRSTRLEN];
1012 
1013 	/* Validate parameters */
1014 	if (ifp == NULL || ifra == NULL) /* this maybe redundant */
1015 		return (EINVAL);
1016 
1017 	/*
1018 	 * The destination address for a p2p link must have a family
1019 	 * of AF_UNSPEC or AF_INET6.
1020 	 */
1021 	if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
1022 	    ifra->ifra_dstaddr.sin6_family != AF_INET6 &&
1023 	    ifra->ifra_dstaddr.sin6_family != AF_UNSPEC)
1024 		return (EAFNOSUPPORT);
1025 	/*
1026 	 * validate ifra_prefixmask.  don't check sin6_family, netmask
1027 	 * does not carry fields other than sin6_len.
1028 	 */
1029 	if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6))
1030 		return (EINVAL);
1031 	/*
1032 	 * Because the IPv6 address architecture is classless, we require
1033 	 * users to specify a (non 0) prefix length (mask) for a new address.
1034 	 * We also require the prefix (when specified) mask is valid, and thus
1035 	 * reject a non-consecutive mask.
1036 	 */
1037 	if (ia == NULL && ifra->ifra_prefixmask.sin6_len == 0)
1038 		return (EINVAL);
1039 	if (ifra->ifra_prefixmask.sin6_len != 0) {
1040 		plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
1041 		    (u_char *)&ifra->ifra_prefixmask +
1042 		    ifra->ifra_prefixmask.sin6_len);
1043 		if (plen <= 0)
1044 			return (EINVAL);
1045 	} else {
1046 		/*
1047 		 * In this case, ia must not be NULL.  We just use its prefix
1048 		 * length.
1049 		 */
1050 		plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
1051 	}
1052 	/*
1053 	 * If the destination address on a p2p interface is specified,
1054 	 * and the address is a scoped one, validate/set the scope
1055 	 * zone identifier.
1056 	 */
1057 	dst6 = ifra->ifra_dstaddr;
1058 	if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) != 0 &&
1059 	    (dst6.sin6_family == AF_INET6)) {
1060 		struct in6_addr in6_tmp;
1061 		u_int32_t zoneid;
1062 
1063 		in6_tmp = dst6.sin6_addr;
1064 		if (in6_setscope(&in6_tmp, ifp, &zoneid))
1065 			return (EINVAL); /* XXX: should be impossible */
1066 
1067 		if (dst6.sin6_scope_id != 0) {
1068 			if (dst6.sin6_scope_id != zoneid)
1069 				return (EINVAL);
1070 		} else		/* user omit to specify the ID. */
1071 			dst6.sin6_scope_id = zoneid;
1072 
1073 		/* convert into the internal form */
1074 		if (sa6_embedscope(&dst6, 0))
1075 			return (EINVAL); /* XXX: should be impossible */
1076 	}
1077 	/*
1078 	 * The destination address can be specified only for a p2p or a
1079 	 * loopback interface.  If specified, the corresponding prefix length
1080 	 * must be 128.
1081 	 */
1082 	if (ifra->ifra_dstaddr.sin6_family == AF_INET6) {
1083 		if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) == 0) {
1084 			/* XXX: noisy message */
1085 			nd6log((LOG_INFO, "in6_update_ifa: a destination can "
1086 			    "be specified for a p2p or a loopback IF only\n"));
1087 			return (EINVAL);
1088 		}
1089 		if (plen != 128) {
1090 			nd6log((LOG_INFO, "in6_update_ifa: prefixlen should "
1091 			    "be 128 when dstaddr is specified\n"));
1092 			return (EINVAL);
1093 		}
1094 	}
1095 	/* lifetime consistency check */
1096 	lt = &ifra->ifra_lifetime;
1097 	if (lt->ia6t_pltime > lt->ia6t_vltime)
1098 		return (EINVAL);
1099 	if (lt->ia6t_vltime == 0) {
1100 		/*
1101 		 * the following log might be noisy, but this is a typical
1102 		 * configuration mistake or a tool's bug.
1103 		 */
1104 		nd6log((LOG_INFO,
1105 		    "in6_update_ifa: valid lifetime is 0 for %s\n",
1106 		    ip6_sprintf(ip6buf, &ifra->ifra_addr.sin6_addr)));
1107 
1108 		if (ia == NULL)
1109 			return (0); /* there's nothing to do */
1110 	}
1111 
1112 	/*
1113 	 * If this is a new address, allocate a new ifaddr and link it
1114 	 * into chains.
1115 	 */
1116 	if (ia == NULL) {
1117 		hostIsNew = 1;
1118 		/*
1119 		 * When in6_update_ifa() is called in a process of a received
1120 		 * RA, it is called under an interrupt context.  So, we should
1121 		 * call malloc with M_NOWAIT.
1122 		 */
1123 		ia = (struct in6_ifaddr *)ifa_alloc(sizeof(*ia), M_NOWAIT);
1124 		if (ia == NULL)
1125 			return (ENOBUFS);
1126 		LIST_INIT(&ia->ia6_memberships);
1127 		/* Initialize the address and masks, and put time stamp */
1128 		ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
1129 		ia->ia_addr.sin6_family = AF_INET6;
1130 		ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
1131 		ia->ia6_createtime = time_uptime;
1132 		if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
1133 			/*
1134 			 * XXX: some functions expect that ifa_dstaddr is not
1135 			 * NULL for p2p interfaces.
1136 			 */
1137 			ia->ia_ifa.ifa_dstaddr =
1138 			    (struct sockaddr *)&ia->ia_dstaddr;
1139 		} else {
1140 			ia->ia_ifa.ifa_dstaddr = NULL;
1141 		}
1142 		ia->ia_ifa.ifa_netmask = (struct sockaddr *)&ia->ia_prefixmask;
1143 		ia->ia_ifp = ifp;
1144 		ifa_ref(&ia->ia_ifa);			/* if_addrhead */
1145 		IF_ADDR_WLOCK(ifp);
1146 		TAILQ_INSERT_TAIL(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
1147 		IF_ADDR_WUNLOCK(ifp);
1148 
1149 		ifa_ref(&ia->ia_ifa);			/* in6_ifaddrhead */
1150 		IN6_IFADDR_WLOCK();
1151 		TAILQ_INSERT_TAIL(&V_in6_ifaddrhead, ia, ia_link);
1152 		LIST_INSERT_HEAD(IN6ADDR_HASH(&ifra->ifra_addr.sin6_addr),
1153 		    ia, ia6_hash);
1154 		IN6_IFADDR_WUNLOCK();
1155 	}
1156 
1157 	/* update timestamp */
1158 	ia->ia6_updatetime = time_uptime;
1159 
1160 	/* set prefix mask */
1161 	if (ifra->ifra_prefixmask.sin6_len) {
1162 		/*
1163 		 * We prohibit changing the prefix length of an existing
1164 		 * address, because
1165 		 * + such an operation should be rare in IPv6, and
1166 		 * + the operation would confuse prefix management.
1167 		 */
1168 		if (ia->ia_prefixmask.sin6_len &&
1169 		    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) != plen) {
1170 			nd6log((LOG_INFO, "in6_update_ifa: the prefix length of an"
1171 			    " existing (%s) address should not be changed\n",
1172 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1173 			error = EINVAL;
1174 			goto unlink;
1175 		}
1176 		ia->ia_prefixmask = ifra->ifra_prefixmask;
1177 		ia->ia_prefixmask.sin6_family = AF_INET6;
1178 	}
1179 
1180 	/*
1181 	 * If a new destination address is specified, scrub the old one and
1182 	 * install the new destination.  Note that the interface must be
1183 	 * p2p or loopback (see the check above.)
1184 	 */
1185 	if (dst6.sin6_family == AF_INET6 &&
1186 	    !IN6_ARE_ADDR_EQUAL(&dst6.sin6_addr, &ia->ia_dstaddr.sin6_addr)) {
1187 		int e;
1188 
1189 		if ((ia->ia_flags & IFA_ROUTE) != 0 &&
1190 		    (e = rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST)) != 0) {
1191 			nd6log((LOG_ERR, "in6_update_ifa: failed to remove "
1192 			    "a route to the old destination: %s\n",
1193 			    ip6_sprintf(ip6buf, &ia->ia_addr.sin6_addr)));
1194 			/* proceed anyway... */
1195 		} else
1196 			ia->ia_flags &= ~IFA_ROUTE;
1197 		ia->ia_dstaddr = dst6;
1198 	}
1199 
1200 	/*
1201 	 * Set lifetimes.  We do not refer to ia6t_expire and ia6t_preferred
1202 	 * to see if the address is deprecated or invalidated, but initialize
1203 	 * these members for applications.
1204 	 */
1205 	ia->ia6_lifetime = ifra->ifra_lifetime;
1206 	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
1207 		ia->ia6_lifetime.ia6t_expire =
1208 		    time_uptime + ia->ia6_lifetime.ia6t_vltime;
1209 	} else
1210 		ia->ia6_lifetime.ia6t_expire = 0;
1211 	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
1212 		ia->ia6_lifetime.ia6t_preferred =
1213 		    time_uptime + ia->ia6_lifetime.ia6t_pltime;
1214 	} else
1215 		ia->ia6_lifetime.ia6t_preferred = 0;
1216 
1217 	/* reset the interface and routing table appropriately. */
1218 	if ((error = in6_ifinit(ifp, ia, &ifra->ifra_addr, hostIsNew)) != 0)
1219 		goto unlink;
1220 
1221 	/*
1222 	 * configure address flags.
1223 	 */
1224 	ia->ia6_flags = ifra->ifra_flags;
1225 	/*
1226 	 * backward compatibility - if IN6_IFF_DEPRECATED is set from the
1227 	 * userland, make it deprecated.
1228 	 */
1229 	if ((ifra->ifra_flags & IN6_IFF_DEPRECATED) != 0) {
1230 		ia->ia6_lifetime.ia6t_pltime = 0;
1231 		ia->ia6_lifetime.ia6t_preferred = time_uptime;
1232 	}
1233 	/*
1234 	 * Make the address tentative before joining multicast addresses,
1235 	 * so that corresponding MLD responses would not have a tentative
1236 	 * source address.
1237 	 */
1238 	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/* safety */
1239 	if (hostIsNew && in6if_do_dad(ifp))
1240 		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1241 
1242 	/* DAD should be performed after ND6_IFF_IFDISABLED is cleared. */
1243 	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)
1244 		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1245 
1246 	/*
1247 	 * We are done if we have simply modified an existing address.
1248 	 */
1249 	if (!hostIsNew)
1250 		return (error);
1251 
1252 	/*
1253 	 * Beyond this point, we should call in6_purgeaddr upon an error,
1254 	 * not just go to unlink.
1255 	 */
1256 
1257 	/* Join necessary multicast groups. */
1258 	in6m_sol = NULL;
1259 	if ((ifp->if_flags & IFF_MULTICAST) != 0) {
1260 		error = in6_update_ifa_join_mc(ifp, ifra, ia, flags, &in6m_sol);
1261 		if (error)
1262 			goto cleanup;
1263 	}
1264 
1265 	/*
1266 	 * Perform DAD, if needed.
1267 	 * XXX It may be of use, if we can administratively disable DAD.
1268 	 */
1269 	if (in6if_do_dad(ifp) && ((ifra->ifra_flags & IN6_IFF_NODAD) == 0) &&
1270 	    (ia->ia6_flags & IN6_IFF_TENTATIVE))
1271 	{
1272 		int mindelay, maxdelay;
1273 
1274 		delay = 0;
1275 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1276 			/*
1277 			 * We need to impose a delay before sending an NS
1278 			 * for DAD.  Check if we also needed a delay for the
1279 			 * corresponding MLD message.  If we did, the delay
1280 			 * should be larger than the MLD delay (this could be
1281 			 * relaxed a bit, but this simple logic is at least
1282 			 * safe).
1283 			 * XXX: Break data hiding guidelines and look at
1284 			 * state for the solicited multicast group.
1285 			 */
1286 			mindelay = 0;
1287 			if (in6m_sol != NULL &&
1288 			    in6m_sol->in6m_state == MLD_REPORTING_MEMBER) {
1289 				mindelay = in6m_sol->in6m_timer;
1290 			}
1291 			maxdelay = MAX_RTR_SOLICITATION_DELAY * hz;
1292 			if (maxdelay - mindelay == 0)
1293 				delay = 0;
1294 			else {
1295 				delay =
1296 				    (arc4random() % (maxdelay - mindelay)) +
1297 				    mindelay;
1298 			}
1299 		}
1300 		nd6_dad_start((struct ifaddr *)ia, delay);
1301 	}
1302 
1303 	KASSERT(hostIsNew, ("in6_update_ifa: !hostIsNew"));
1304 	ifa_free(&ia->ia_ifa);
1305 	return (error);
1306 
1307   unlink:
1308 	/*
1309 	 * XXX: if a change of an existing address failed, keep the entry
1310 	 * anyway.
1311 	 */
1312 	if (hostIsNew) {
1313 		in6_unlink_ifa(ia, ifp);
1314 		ifa_free(&ia->ia_ifa);
1315 	}
1316 	return (error);
1317 
1318   cleanup:
1319 	KASSERT(hostIsNew, ("in6_update_ifa: cleanup: !hostIsNew"));
1320 	ifa_free(&ia->ia_ifa);
1321 	in6_purgeaddr(&ia->ia_ifa);
1322 	return error;
1323 }
1324 
1325 /*
1326  * Leave multicast groups.  Factored out from in6_purgeaddr().
1327  * This entire work should only be done once, for the default FIB.
1328  */
1329 static int
1330 in6_purgeaddr_mc(struct ifnet *ifp, struct in6_ifaddr *ia, struct ifaddr *ifa0)
1331 {
1332 	struct sockaddr_in6 mltaddr, mltmask;
1333 	struct in6_multi_mship *imm;
1334 	struct rtentry *rt;
1335 	struct sockaddr_in6 sin6;
1336 	int error;
1337 
1338 	/*
1339 	 * Leave from multicast groups we have joined for the interface.
1340 	 */
1341 	while ((imm = LIST_FIRST(&ia->ia6_memberships)) != NULL) {
1342 		LIST_REMOVE(imm, i6mm_chain);
1343 		in6_leavegroup(imm);
1344 	}
1345 
1346 	/*
1347 	 * Remove the link-local all-nodes address.
1348 	 */
1349 	bzero(&mltmask, sizeof(mltmask));
1350 	mltmask.sin6_len = sizeof(struct sockaddr_in6);
1351 	mltmask.sin6_family = AF_INET6;
1352 	mltmask.sin6_addr = in6mask32;
1353 
1354 	bzero(&mltaddr, sizeof(mltaddr));
1355 	mltaddr.sin6_len = sizeof(struct sockaddr_in6);
1356 	mltaddr.sin6_family = AF_INET6;
1357 	mltaddr.sin6_addr = in6addr_linklocal_allnodes;
1358 
1359 	if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) != 0)
1360 		return (error);
1361 
1362 	/*
1363 	 * As for the mltaddr above, proactively prepare the sin6 to avoid
1364 	 * rtentry un- and re-locking.
1365 	 */
1366 	if (ifa0 != NULL) {
1367 		bzero(&sin6, sizeof(sin6));
1368 		sin6.sin6_len = sizeof(sin6);
1369 		sin6.sin6_family = AF_INET6;
1370 		memcpy(&sin6.sin6_addr, &satosin6(ifa0->ifa_addr)->sin6_addr,
1371 		    sizeof(sin6.sin6_addr));
1372 		error = in6_setscope(&sin6.sin6_addr, ifa0->ifa_ifp, NULL);
1373 		if (error != 0)
1374 			return (error);
1375 	}
1376 
1377 	rt = in6_rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL, RT_DEFAULT_FIB);
1378 	if (rt != NULL && rt->rt_gateway != NULL &&
1379 	    (memcmp(&satosin6(rt->rt_gateway)->sin6_addr,
1380 		    &ia->ia_addr.sin6_addr,
1381 		    sizeof(ia->ia_addr.sin6_addr)) == 0)) {
1382 		/*
1383 		 * If no more IPv6 address exists on this interface then
1384 		 * remove the multicast address route.
1385 		 */
1386 		if (ifa0 == NULL) {
1387 			memcpy(&mltaddr.sin6_addr,
1388 			    &satosin6(rt_key(rt))->sin6_addr,
1389 			    sizeof(mltaddr.sin6_addr));
1390 			RTFREE_LOCKED(rt);
1391 			error = in6_rtrequest(RTM_DELETE,
1392 			    (struct sockaddr *)&mltaddr,
1393 			    (struct sockaddr *)&ia->ia_addr,
1394 			    (struct sockaddr *)&mltmask, RTF_UP,
1395 			    (struct rtentry **)0, RT_DEFAULT_FIB);
1396 			if (error)
1397 				log(LOG_INFO, "%s: link-local all-nodes "
1398 				    "multicast address deletion error\n",
1399 				    __func__);
1400 		} else {
1401 			/*
1402 			 * Replace the gateway of the route.
1403 			 */
1404 			memcpy(rt->rt_gateway, &sin6, sizeof(sin6));
1405 			RTFREE_LOCKED(rt);
1406 		}
1407 	} else {
1408 		if (rt != NULL)
1409 			RTFREE_LOCKED(rt);
1410 	}
1411 
1412 	/*
1413 	 * Remove the node-local all-nodes address.
1414 	 */
1415 	mltaddr.sin6_addr = in6addr_nodelocal_allnodes;
1416 	if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) != 0)
1417 		return (error);
1418 
1419 	rt = in6_rtalloc1((struct sockaddr *)&mltaddr, 0, 0UL, RT_DEFAULT_FIB);
1420 	if (rt != NULL && rt->rt_gateway != NULL &&
1421 	    (memcmp(&satosin6(rt->rt_gateway)->sin6_addr,
1422 		    &ia->ia_addr.sin6_addr,
1423 		    sizeof(ia->ia_addr.sin6_addr)) == 0)) {
1424 		/*
1425 		 * If no more IPv6 address exists on this interface then
1426 		 * remove the multicast address route.
1427 		 */
1428 		if (ifa0 == NULL) {
1429 			memcpy(&mltaddr.sin6_addr,
1430 			    &satosin6(rt_key(rt))->sin6_addr,
1431 			    sizeof(mltaddr.sin6_addr));
1432 
1433 			RTFREE_LOCKED(rt);
1434 			error = in6_rtrequest(RTM_DELETE,
1435 			    (struct sockaddr *)&mltaddr,
1436 			    (struct sockaddr *)&ia->ia_addr,
1437 			    (struct sockaddr *)&mltmask, RTF_UP,
1438 			    (struct rtentry **)0, RT_DEFAULT_FIB);
1439 			if (error)
1440 				log(LOG_INFO, "%s: node-local all-nodes"
1441 				    "multicast address deletion error\n",
1442 				    __func__);
1443 		} else {
1444 			/*
1445 			 * Replace the gateway of the route.
1446 			 */
1447 			memcpy(rt->rt_gateway, &sin6, sizeof(sin6));
1448 			RTFREE_LOCKED(rt);
1449 		}
1450 	} else {
1451 		if (rt != NULL)
1452 			RTFREE_LOCKED(rt);
1453 	}
1454 
1455 	return (0);
1456 }
1457 
1458 void
1459 in6_purgeaddr(struct ifaddr *ifa)
1460 {
1461 	struct ifnet *ifp = ifa->ifa_ifp;
1462 	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1463 	int plen, error;
1464 	struct ifaddr *ifa0;
1465 
1466 	if (ifa->ifa_carp)
1467 		(*carp_detach_p)(ifa);
1468 
1469 	/*
1470 	 * find another IPv6 address as the gateway for the
1471 	 * link-local and node-local all-nodes multicast
1472 	 * address routes
1473 	 */
1474 	IF_ADDR_RLOCK(ifp);
1475 	TAILQ_FOREACH(ifa0, &ifp->if_addrhead, ifa_link) {
1476 		if ((ifa0->ifa_addr->sa_family != AF_INET6) ||
1477 		    memcmp(&satosin6(ifa0->ifa_addr)->sin6_addr,
1478 		    &ia->ia_addr.sin6_addr, sizeof(struct in6_addr)) == 0)
1479 			continue;
1480 		else
1481 			break;
1482 	}
1483 	if (ifa0 != NULL)
1484 		ifa_ref(ifa0);
1485 	IF_ADDR_RUNLOCK(ifp);
1486 
1487 	/*
1488 	 * Remove the loopback route to the interface address.
1489 	 * The check for the current setting of "nd6_useloopback"
1490 	 * is not needed.
1491 	 */
1492 	if (ia->ia_flags & IFA_RTSELF) {
1493 		error = ifa_del_loopback_route((struct ifaddr *)ia,
1494 		    (struct sockaddr *)&ia->ia_addr);
1495 		if (error == 0)
1496 			ia->ia_flags &= ~IFA_RTSELF;
1497 	}
1498 
1499 	/* stop DAD processing */
1500 	nd6_dad_stop(ifa);
1501 
1502 	/* Remove local address entry from lltable. */
1503 	in6_ifremloop(ifa);
1504 
1505 	/* Leave multicast groups. */
1506 	error = in6_purgeaddr_mc(ifp, ia, ifa0);
1507 
1508 	if (ifa0 != NULL)
1509 		ifa_free(ifa0);
1510 
1511 	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1512 	if ((ia->ia_flags & IFA_ROUTE) && plen == 128) {
1513 		error = rtinit(&(ia->ia_ifa), RTM_DELETE, ia->ia_flags |
1514 		    (ia->ia_dstaddr.sin6_family == AF_INET6) ? RTF_HOST : 0);
1515 		if (error != 0)
1516 			log(LOG_INFO, "%s: err=%d, destination address delete "
1517 			    "failed\n", __func__, error);
1518 		ia->ia_flags &= ~IFA_ROUTE;
1519 	}
1520 
1521 	in6_unlink_ifa(ia, ifp);
1522 }
1523 
1524 static void
1525 in6_unlink_ifa(struct in6_ifaddr *ia, struct ifnet *ifp)
1526 {
1527 
1528 	IF_ADDR_WLOCK(ifp);
1529 	TAILQ_REMOVE(&ifp->if_addrhead, &ia->ia_ifa, ifa_link);
1530 	IF_ADDR_WUNLOCK(ifp);
1531 	ifa_free(&ia->ia_ifa);			/* if_addrhead */
1532 
1533 	/*
1534 	 * Defer the release of what might be the last reference to the
1535 	 * in6_ifaddr so that it can't be freed before the remainder of the
1536 	 * cleanup.
1537 	 */
1538 	IN6_IFADDR_WLOCK();
1539 	TAILQ_REMOVE(&V_in6_ifaddrhead, ia, ia_link);
1540 	LIST_REMOVE(ia, ia6_hash);
1541 	IN6_IFADDR_WUNLOCK();
1542 
1543 	/*
1544 	 * Release the reference to the base prefix.  There should be a
1545 	 * positive reference.
1546 	 */
1547 	if (ia->ia6_ndpr == NULL) {
1548 		nd6log((LOG_NOTICE,
1549 		    "in6_unlink_ifa: autoconf'ed address "
1550 		    "%p has no prefix\n", ia));
1551 	} else {
1552 		ia->ia6_ndpr->ndpr_refcnt--;
1553 		ia->ia6_ndpr = NULL;
1554 	}
1555 
1556 	/*
1557 	 * Also, if the address being removed is autoconf'ed, call
1558 	 * pfxlist_onlink_check() since the release might affect the status of
1559 	 * other (detached) addresses.
1560 	 */
1561 	if ((ia->ia6_flags & IN6_IFF_AUTOCONF)) {
1562 		pfxlist_onlink_check();
1563 	}
1564 	ifa_free(&ia->ia_ifa);			/* in6_ifaddrhead */
1565 }
1566 
1567 void
1568 in6_purgeif(struct ifnet *ifp)
1569 {
1570 	struct ifaddr *ifa, *nifa;
1571 
1572 	TAILQ_FOREACH_SAFE(ifa, &ifp->if_addrhead, ifa_link, nifa) {
1573 		if (ifa->ifa_addr->sa_family != AF_INET6)
1574 			continue;
1575 		in6_purgeaddr(ifa);
1576 	}
1577 
1578 	in6_ifdetach(ifp);
1579 }
1580 
1581 /*
1582  * Initialize an interface's IPv6 address and routing table entry.
1583  */
1584 static int
1585 in6_ifinit(struct ifnet *ifp, struct in6_ifaddr *ia,
1586     struct sockaddr_in6 *sin6, int newhost)
1587 {
1588 	int	error = 0, plen, ifacount = 0;
1589 	struct ifaddr *ifa;
1590 
1591 	/*
1592 	 * Give the interface a chance to initialize
1593 	 * if this is its first address,
1594 	 * and to validate the address if necessary.
1595 	 */
1596 	IF_ADDR_RLOCK(ifp);
1597 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1598 		if (ifa->ifa_addr->sa_family != AF_INET6)
1599 			continue;
1600 		ifacount++;
1601 	}
1602 	IF_ADDR_RUNLOCK(ifp);
1603 
1604 	ia->ia_addr = *sin6;
1605 
1606 	if (ifacount <= 1 && ifp->if_ioctl) {
1607 		error = (*ifp->if_ioctl)(ifp, SIOCSIFADDR, (caddr_t)ia);
1608 		if (error)
1609 			return (error);
1610 	}
1611 
1612 	ia->ia_ifa.ifa_metric = ifp->if_metric;
1613 
1614 	/* we could do in(6)_socktrim here, but just omit it at this moment. */
1615 
1616 	/*
1617 	 * Special case:
1618 	 * If a new destination address is specified for a point-to-point
1619 	 * interface, install a route to the destination as an interface
1620 	 * direct route.
1621 	 * XXX: the logic below rejects assigning multiple addresses on a p2p
1622 	 * interface that share the same destination.
1623 	 */
1624 	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1625 	if (!(ia->ia_flags & IFA_ROUTE) && plen == 128 &&
1626 	    ia->ia_dstaddr.sin6_family == AF_INET6) {
1627 		int rtflags = RTF_UP | RTF_HOST;
1628 		error = rtinit(&ia->ia_ifa, RTM_ADD, ia->ia_flags | rtflags);
1629 		if (error)
1630 			return (error);
1631 		ia->ia_flags |= IFA_ROUTE;
1632 		/*
1633 		 * Handle the case for ::1 .
1634 		 */
1635 		if (ifp->if_flags & IFF_LOOPBACK)
1636 			ia->ia_flags |= IFA_RTSELF;
1637 	}
1638 
1639 	/*
1640 	 * add a loopback route to self
1641 	 */
1642 	if (!(ia->ia_flags & IFA_RTSELF) && V_nd6_useloopback) {
1643 		error = ifa_add_loopback_route((struct ifaddr *)ia,
1644 		    (struct sockaddr *)&ia->ia_addr);
1645 		if (error == 0)
1646 			ia->ia_flags |= IFA_RTSELF;
1647 	}
1648 
1649 	/* Add local address to lltable, if necessary (ex. on p2p link). */
1650 	if (newhost)
1651 		in6_ifaddloop(&(ia->ia_ifa));
1652 
1653 	return (error);
1654 }
1655 
1656 /*
1657  * Find an IPv6 interface link-local address specific to an interface.
1658  * ifaddr is returned referenced.
1659  */
1660 struct in6_ifaddr *
1661 in6ifa_ifpforlinklocal(struct ifnet *ifp, int ignoreflags)
1662 {
1663 	struct ifaddr *ifa;
1664 
1665 	IF_ADDR_RLOCK(ifp);
1666 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1667 		if (ifa->ifa_addr->sa_family != AF_INET6)
1668 			continue;
1669 		if (IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa))) {
1670 			if ((((struct in6_ifaddr *)ifa)->ia6_flags &
1671 			    ignoreflags) != 0)
1672 				continue;
1673 			ifa_ref(ifa);
1674 			break;
1675 		}
1676 	}
1677 	IF_ADDR_RUNLOCK(ifp);
1678 
1679 	return ((struct in6_ifaddr *)ifa);
1680 }
1681 
1682 
1683 /*
1684  * find the internet address corresponding to a given interface and address.
1685  * ifaddr is returned referenced.
1686  */
1687 struct in6_ifaddr *
1688 in6ifa_ifpwithaddr(struct ifnet *ifp, struct in6_addr *addr)
1689 {
1690 	struct ifaddr *ifa;
1691 
1692 	IF_ADDR_RLOCK(ifp);
1693 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1694 		if (ifa->ifa_addr->sa_family != AF_INET6)
1695 			continue;
1696 		if (IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa))) {
1697 			ifa_ref(ifa);
1698 			break;
1699 		}
1700 	}
1701 	IF_ADDR_RUNLOCK(ifp);
1702 
1703 	return ((struct in6_ifaddr *)ifa);
1704 }
1705 
1706 /*
1707  * Find a link-local scoped address on ifp and return it if any.
1708  */
1709 struct in6_ifaddr *
1710 in6ifa_llaonifp(struct ifnet *ifp)
1711 {
1712 	struct sockaddr_in6 *sin6;
1713 	struct ifaddr *ifa;
1714 
1715 	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)
1716 		return (NULL);
1717 	if_addr_rlock(ifp);
1718 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1719 		if (ifa->ifa_addr->sa_family != AF_INET6)
1720 			continue;
1721 		sin6 = (struct sockaddr_in6 *)ifa->ifa_addr;
1722 		if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr) ||
1723 		    IN6_IS_ADDR_MC_INTFACELOCAL(&sin6->sin6_addr) ||
1724 		    IN6_IS_ADDR_MC_NODELOCAL(&sin6->sin6_addr))
1725 			break;
1726 	}
1727 	if_addr_runlock(ifp);
1728 
1729 	return ((struct in6_ifaddr *)ifa);
1730 }
1731 
1732 /*
1733  * Convert IP6 address to printable (loggable) representation. Caller
1734  * has to make sure that ip6buf is at least INET6_ADDRSTRLEN long.
1735  */
1736 static char digits[] = "0123456789abcdef";
1737 char *
1738 ip6_sprintf(char *ip6buf, const struct in6_addr *addr)
1739 {
1740 	int i, cnt = 0, maxcnt = 0, idx = 0, index = 0;
1741 	char *cp;
1742 	const u_int16_t *a = (const u_int16_t *)addr;
1743 	const u_int8_t *d;
1744 	int dcolon = 0, zero = 0;
1745 
1746 	cp = ip6buf;
1747 
1748 	for (i = 0; i < 8; i++) {
1749 		if (*(a + i) == 0) {
1750 			cnt++;
1751 			if (cnt == 1)
1752 				idx = i;
1753 		}
1754 		else if (maxcnt < cnt) {
1755 			maxcnt = cnt;
1756 			index = idx;
1757 			cnt = 0;
1758 		}
1759 	}
1760 	if (maxcnt < cnt) {
1761 		maxcnt = cnt;
1762 		index = idx;
1763 	}
1764 
1765 	for (i = 0; i < 8; i++) {
1766 		if (dcolon == 1) {
1767 			if (*a == 0) {
1768 				if (i == 7)
1769 					*cp++ = ':';
1770 				a++;
1771 				continue;
1772 			} else
1773 				dcolon = 2;
1774 		}
1775 		if (*a == 0) {
1776 			if (dcolon == 0 && *(a + 1) == 0 && i == index) {
1777 				if (i == 0)
1778 					*cp++ = ':';
1779 				*cp++ = ':';
1780 				dcolon = 1;
1781 			} else {
1782 				*cp++ = '0';
1783 				*cp++ = ':';
1784 			}
1785 			a++;
1786 			continue;
1787 		}
1788 		d = (const u_char *)a;
1789 		/* Try to eliminate leading zeros in printout like in :0001. */
1790 		zero = 1;
1791 		*cp = digits[*d >> 4];
1792 		if (*cp != '0') {
1793 			zero = 0;
1794 			cp++;
1795 		}
1796 		*cp = digits[*d++ & 0xf];
1797 		if (zero == 0 || (*cp != '0')) {
1798 			zero = 0;
1799 			cp++;
1800 		}
1801 		*cp = digits[*d >> 4];
1802 		if (zero == 0 || (*cp != '0')) {
1803 			zero = 0;
1804 			cp++;
1805 		}
1806 		*cp++ = digits[*d & 0xf];
1807 		*cp++ = ':';
1808 		a++;
1809 	}
1810 	*--cp = '\0';
1811 	return (ip6buf);
1812 }
1813 
1814 int
1815 in6_localaddr(struct in6_addr *in6)
1816 {
1817 	struct in6_ifaddr *ia;
1818 
1819 	if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
1820 		return 1;
1821 
1822 	IN6_IFADDR_RLOCK();
1823 	TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
1824 		if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
1825 		    &ia->ia_prefixmask.sin6_addr)) {
1826 			IN6_IFADDR_RUNLOCK();
1827 			return 1;
1828 		}
1829 	}
1830 	IN6_IFADDR_RUNLOCK();
1831 
1832 	return (0);
1833 }
1834 
1835 /*
1836  * Return 1 if an internet address is for the local host and configured
1837  * on one of its interfaces.
1838  */
1839 int
1840 in6_localip(struct in6_addr *in6)
1841 {
1842 	struct in6_ifaddr *ia;
1843 
1844 	IN6_IFADDR_RLOCK();
1845 	LIST_FOREACH(ia, IN6ADDR_HASH(in6), ia6_hash) {
1846 		if (IN6_ARE_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr)) {
1847 			IN6_IFADDR_RUNLOCK();
1848 			return (1);
1849 		}
1850 	}
1851 	IN6_IFADDR_RUNLOCK();
1852 	return (0);
1853 }
1854 
1855 int
1856 in6_is_addr_deprecated(struct sockaddr_in6 *sa6)
1857 {
1858 	struct in6_ifaddr *ia;
1859 
1860 	IN6_IFADDR_RLOCK();
1861 	LIST_FOREACH(ia, IN6ADDR_HASH(&sa6->sin6_addr), ia6_hash) {
1862 		if (IN6_ARE_ADDR_EQUAL(IA6_IN6(ia), &sa6->sin6_addr)) {
1863 			if (ia->ia6_flags & IN6_IFF_DEPRECATED) {
1864 				IN6_IFADDR_RUNLOCK();
1865 				return (1); /* true */
1866 			}
1867 			break;
1868 		}
1869 	}
1870 	IN6_IFADDR_RUNLOCK();
1871 
1872 	return (0);		/* false */
1873 }
1874 
1875 /*
1876  * return length of part which dst and src are equal
1877  * hard coding...
1878  */
1879 int
1880 in6_matchlen(struct in6_addr *src, struct in6_addr *dst)
1881 {
1882 	int match = 0;
1883 	u_char *s = (u_char *)src, *d = (u_char *)dst;
1884 	u_char *lim = s + 16, r;
1885 
1886 	while (s < lim)
1887 		if ((r = (*d++ ^ *s++)) != 0) {
1888 			while (r < 128) {
1889 				match++;
1890 				r <<= 1;
1891 			}
1892 			break;
1893 		} else
1894 			match += 8;
1895 	return match;
1896 }
1897 
1898 /* XXX: to be scope conscious */
1899 int
1900 in6_are_prefix_equal(struct in6_addr *p1, struct in6_addr *p2, int len)
1901 {
1902 	int bytelen, bitlen;
1903 
1904 	/* sanity check */
1905 	if (0 > len || len > 128) {
1906 		log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n",
1907 		    len);
1908 		return (0);
1909 	}
1910 
1911 	bytelen = len / 8;
1912 	bitlen = len % 8;
1913 
1914 	if (bcmp(&p1->s6_addr, &p2->s6_addr, bytelen))
1915 		return (0);
1916 	if (bitlen != 0 &&
1917 	    p1->s6_addr[bytelen] >> (8 - bitlen) !=
1918 	    p2->s6_addr[bytelen] >> (8 - bitlen))
1919 		return (0);
1920 
1921 	return (1);
1922 }
1923 
1924 void
1925 in6_prefixlen2mask(struct in6_addr *maskp, int len)
1926 {
1927 	u_char maskarray[8] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
1928 	int bytelen, bitlen, i;
1929 
1930 	/* sanity check */
1931 	if (0 > len || len > 128) {
1932 		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
1933 		    len);
1934 		return;
1935 	}
1936 
1937 	bzero(maskp, sizeof(*maskp));
1938 	bytelen = len / 8;
1939 	bitlen = len % 8;
1940 	for (i = 0; i < bytelen; i++)
1941 		maskp->s6_addr[i] = 0xff;
1942 	if (bitlen)
1943 		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
1944 }
1945 
1946 /*
1947  * return the best address out of the same scope. if no address was
1948  * found, return the first valid address from designated IF.
1949  */
1950 struct in6_ifaddr *
1951 in6_ifawithifp(struct ifnet *ifp, struct in6_addr *dst)
1952 {
1953 	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
1954 	struct ifaddr *ifa;
1955 	struct in6_ifaddr *besta = 0;
1956 	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
1957 
1958 	dep[0] = dep[1] = NULL;
1959 
1960 	/*
1961 	 * We first look for addresses in the same scope.
1962 	 * If there is one, return it.
1963 	 * If two or more, return one which matches the dst longest.
1964 	 * If none, return one of global addresses assigned other ifs.
1965 	 */
1966 	IF_ADDR_RLOCK(ifp);
1967 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1968 		if (ifa->ifa_addr->sa_family != AF_INET6)
1969 			continue;
1970 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
1971 			continue; /* XXX: is there any case to allow anycast? */
1972 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
1973 			continue; /* don't use this interface */
1974 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
1975 			continue;
1976 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
1977 			if (V_ip6_use_deprecated)
1978 				dep[0] = (struct in6_ifaddr *)ifa;
1979 			continue;
1980 		}
1981 
1982 		if (dst_scope == in6_addrscope(IFA_IN6(ifa))) {
1983 			/*
1984 			 * call in6_matchlen() as few as possible
1985 			 */
1986 			if (besta) {
1987 				if (blen == -1)
1988 					blen = in6_matchlen(&besta->ia_addr.sin6_addr, dst);
1989 				tlen = in6_matchlen(IFA_IN6(ifa), dst);
1990 				if (tlen > blen) {
1991 					blen = tlen;
1992 					besta = (struct in6_ifaddr *)ifa;
1993 				}
1994 			} else
1995 				besta = (struct in6_ifaddr *)ifa;
1996 		}
1997 	}
1998 	if (besta) {
1999 		ifa_ref(&besta->ia_ifa);
2000 		IF_ADDR_RUNLOCK(ifp);
2001 		return (besta);
2002 	}
2003 
2004 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2005 		if (ifa->ifa_addr->sa_family != AF_INET6)
2006 			continue;
2007 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_ANYCAST)
2008 			continue; /* XXX: is there any case to allow anycast? */
2009 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_NOTREADY)
2010 			continue; /* don't use this interface */
2011 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DETACHED)
2012 			continue;
2013 		if (((struct in6_ifaddr *)ifa)->ia6_flags & IN6_IFF_DEPRECATED) {
2014 			if (V_ip6_use_deprecated)
2015 				dep[1] = (struct in6_ifaddr *)ifa;
2016 			continue;
2017 		}
2018 
2019 		if (ifa != NULL)
2020 			ifa_ref(ifa);
2021 		IF_ADDR_RUNLOCK(ifp);
2022 		return (struct in6_ifaddr *)ifa;
2023 	}
2024 
2025 	/* use the last-resort values, that are, deprecated addresses */
2026 	if (dep[0]) {
2027 		ifa_ref((struct ifaddr *)dep[0]);
2028 		IF_ADDR_RUNLOCK(ifp);
2029 		return dep[0];
2030 	}
2031 	if (dep[1]) {
2032 		ifa_ref((struct ifaddr *)dep[1]);
2033 		IF_ADDR_RUNLOCK(ifp);
2034 		return dep[1];
2035 	}
2036 
2037 	IF_ADDR_RUNLOCK(ifp);
2038 	return NULL;
2039 }
2040 
2041 /*
2042  * perform DAD when interface becomes IFF_UP.
2043  */
2044 void
2045 in6_if_up(struct ifnet *ifp)
2046 {
2047 	struct ifaddr *ifa;
2048 	struct in6_ifaddr *ia;
2049 
2050 	IF_ADDR_RLOCK(ifp);
2051 	TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
2052 		if (ifa->ifa_addr->sa_family != AF_INET6)
2053 			continue;
2054 		ia = (struct in6_ifaddr *)ifa;
2055 		if (ia->ia6_flags & IN6_IFF_TENTATIVE) {
2056 			/*
2057 			 * The TENTATIVE flag was likely set by hand
2058 			 * beforehand, implicitly indicating the need for DAD.
2059 			 * We may be able to skip the random delay in this
2060 			 * case, but we impose delays just in case.
2061 			 */
2062 			nd6_dad_start(ifa,
2063 			    arc4random() % (MAX_RTR_SOLICITATION_DELAY * hz));
2064 		}
2065 	}
2066 	IF_ADDR_RUNLOCK(ifp);
2067 
2068 	/*
2069 	 * special cases, like 6to4, are handled in in6_ifattach
2070 	 */
2071 	in6_ifattach(ifp, NULL);
2072 }
2073 
2074 int
2075 in6if_do_dad(struct ifnet *ifp)
2076 {
2077 	if ((ifp->if_flags & IFF_LOOPBACK) != 0)
2078 		return (0);
2079 
2080 	if (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED)
2081 		return (0);
2082 
2083 	switch (ifp->if_type) {
2084 #ifdef IFT_DUMMY
2085 	case IFT_DUMMY:
2086 #endif
2087 	case IFT_FAITH:
2088 		/*
2089 		 * These interfaces do not have the IFF_LOOPBACK flag,
2090 		 * but loop packets back.  We do not have to do DAD on such
2091 		 * interfaces.  We should even omit it, because loop-backed
2092 		 * NS would confuse the DAD procedure.
2093 		 */
2094 		return (0);
2095 	default:
2096 		/*
2097 		 * Our DAD routine requires the interface up and running.
2098 		 * However, some interfaces can be up before the RUNNING
2099 		 * status.  Additionaly, users may try to assign addresses
2100 		 * before the interface becomes up (or running).
2101 		 * We simply skip DAD in such a case as a work around.
2102 		 * XXX: we should rather mark "tentative" on such addresses,
2103 		 * and do DAD after the interface becomes ready.
2104 		 */
2105 		if (!((ifp->if_flags & IFF_UP) &&
2106 		    (ifp->if_drv_flags & IFF_DRV_RUNNING)))
2107 			return (0);
2108 
2109 		return (1);
2110 	}
2111 }
2112 
2113 /*
2114  * Calculate max IPv6 MTU through all the interfaces and store it
2115  * to in6_maxmtu.
2116  */
2117 void
2118 in6_setmaxmtu(void)
2119 {
2120 	unsigned long maxmtu = 0;
2121 	struct ifnet *ifp;
2122 
2123 	IFNET_RLOCK_NOSLEEP();
2124 	TAILQ_FOREACH(ifp, &V_ifnet, if_list) {
2125 		/* this function can be called during ifnet initialization */
2126 		if (!ifp->if_afdata[AF_INET6])
2127 			continue;
2128 		if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
2129 		    IN6_LINKMTU(ifp) > maxmtu)
2130 			maxmtu = IN6_LINKMTU(ifp);
2131 	}
2132 	IFNET_RUNLOCK_NOSLEEP();
2133 	if (maxmtu)	/* update only when maxmtu is positive */
2134 		V_in6_maxmtu = maxmtu;
2135 }
2136 
2137 /*
2138  * Provide the length of interface identifiers to be used for the link attached
2139  * to the given interface.  The length should be defined in "IPv6 over
2140  * xxx-link" document.  Note that address architecture might also define
2141  * the length for a particular set of address prefixes, regardless of the
2142  * link type.  As clarified in rfc2462bis, those two definitions should be
2143  * consistent, and those really are as of August 2004.
2144  */
2145 int
2146 in6_if2idlen(struct ifnet *ifp)
2147 {
2148 	switch (ifp->if_type) {
2149 	case IFT_ETHER:		/* RFC2464 */
2150 #ifdef IFT_PROPVIRTUAL
2151 	case IFT_PROPVIRTUAL:	/* XXX: no RFC. treat it as ether */
2152 #endif
2153 #ifdef IFT_L2VLAN
2154 	case IFT_L2VLAN:	/* ditto */
2155 #endif
2156 #ifdef IFT_IEEE80211
2157 	case IFT_IEEE80211:	/* ditto */
2158 #endif
2159 #ifdef IFT_MIP
2160 	case IFT_MIP:	/* ditto */
2161 #endif
2162 	case IFT_INFINIBAND:
2163 		return (64);
2164 	case IFT_FDDI:		/* RFC2467 */
2165 		return (64);
2166 	case IFT_ISO88025:	/* RFC2470 (IPv6 over Token Ring) */
2167 		return (64);
2168 	case IFT_PPP:		/* RFC2472 */
2169 		return (64);
2170 	case IFT_ARCNET:	/* RFC2497 */
2171 		return (64);
2172 	case IFT_FRELAY:	/* RFC2590 */
2173 		return (64);
2174 	case IFT_IEEE1394:	/* RFC3146 */
2175 		return (64);
2176 	case IFT_GIF:
2177 		return (64);	/* draft-ietf-v6ops-mech-v2-07 */
2178 	case IFT_LOOP:
2179 		return (64);	/* XXX: is this really correct? */
2180 	default:
2181 		/*
2182 		 * Unknown link type:
2183 		 * It might be controversial to use the today's common constant
2184 		 * of 64 for these cases unconditionally.  For full compliance,
2185 		 * we should return an error in this case.  On the other hand,
2186 		 * if we simply miss the standard for the link type or a new
2187 		 * standard is defined for a new link type, the IFID length
2188 		 * is very likely to be the common constant.  As a compromise,
2189 		 * we always use the constant, but make an explicit notice
2190 		 * indicating the "unknown" case.
2191 		 */
2192 		printf("in6_if2idlen: unknown link type (%d)\n", ifp->if_type);
2193 		return (64);
2194 	}
2195 }
2196 
2197 #include <sys/sysctl.h>
2198 
2199 struct in6_llentry {
2200 	struct llentry		base;
2201 	struct sockaddr_in6	l3_addr6;
2202 };
2203 
2204 /*
2205  * Deletes an address from the address table.
2206  * This function is called by the timer functions
2207  * such as arptimer() and nd6_llinfo_timer(), and
2208  * the caller does the locking.
2209  */
2210 static void
2211 in6_lltable_free(struct lltable *llt, struct llentry *lle)
2212 {
2213 	LLE_WUNLOCK(lle);
2214 	LLE_LOCK_DESTROY(lle);
2215 	free(lle, M_LLTABLE);
2216 }
2217 
2218 static struct llentry *
2219 in6_lltable_new(const struct sockaddr *l3addr, u_int flags)
2220 {
2221 	struct in6_llentry *lle;
2222 
2223 	lle = malloc(sizeof(struct in6_llentry), M_LLTABLE, M_NOWAIT | M_ZERO);
2224 	if (lle == NULL)		/* NB: caller generates msg */
2225 		return NULL;
2226 
2227 	lle->l3_addr6 = *(const struct sockaddr_in6 *)l3addr;
2228 	lle->base.lle_refcnt = 1;
2229 	lle->base.lle_free = in6_lltable_free;
2230 	LLE_LOCK_INIT(&lle->base);
2231 	callout_init_rw(&lle->base.ln_timer_ch, &lle->base.lle_lock,
2232 	    CALLOUT_RETURNUNLOCKED);
2233 
2234 	return (&lle->base);
2235 }
2236 
2237 static void
2238 in6_lltable_prefix_free(struct lltable *llt, const struct sockaddr *prefix,
2239     const struct sockaddr *mask, u_int flags)
2240 {
2241 	const struct sockaddr_in6 *pfx = (const struct sockaddr_in6 *)prefix;
2242 	const struct sockaddr_in6 *msk = (const struct sockaddr_in6 *)mask;
2243 	struct llentry *lle, *next;
2244 	int i;
2245 
2246 	/*
2247 	 * (flags & LLE_STATIC) means deleting all entries
2248 	 * including static ND6 entries.
2249 	 */
2250 	IF_AFDATA_WLOCK(llt->llt_ifp);
2251 	for (i = 0; i < LLTBL_HASHTBL_SIZE; i++) {
2252 		LIST_FOREACH_SAFE(lle, &llt->lle_head[i], lle_next, next) {
2253 			if (IN6_ARE_MASKED_ADDR_EQUAL(
2254 			    &satosin6(L3_ADDR(lle))->sin6_addr,
2255 			    &pfx->sin6_addr, &msk->sin6_addr) &&
2256 			    ((flags & LLE_STATIC) ||
2257 			    !(lle->la_flags & LLE_STATIC))) {
2258 				LLE_WLOCK(lle);
2259 				if (callout_stop(&lle->la_timer))
2260 					LLE_REMREF(lle);
2261 				llentry_free(lle);
2262 			}
2263 		}
2264 	}
2265 	IF_AFDATA_WUNLOCK(llt->llt_ifp);
2266 }
2267 
2268 static int
2269 in6_lltable_rtcheck(struct ifnet *ifp,
2270 		    u_int flags,
2271 		    const struct sockaddr *l3addr)
2272 {
2273 	struct rtentry *rt;
2274 	char ip6buf[INET6_ADDRSTRLEN];
2275 
2276 	KASSERT(l3addr->sa_family == AF_INET6,
2277 	    ("sin_family %d", l3addr->sa_family));
2278 
2279 	/* Our local addresses are always only installed on the default FIB. */
2280 	/* XXX rtalloc1 should take a const param */
2281 	rt = in6_rtalloc1(__DECONST(struct sockaddr *, l3addr), 0, 0,
2282 	    RT_DEFAULT_FIB);
2283 	if (rt == NULL || (rt->rt_flags & RTF_GATEWAY) || rt->rt_ifp != ifp) {
2284 		struct ifaddr *ifa;
2285 		/*
2286 		 * Create an ND6 cache for an IPv6 neighbor
2287 		 * that is not covered by our own prefix.
2288 		 */
2289 		/* XXX ifaof_ifpforaddr should take a const param */
2290 		ifa = ifaof_ifpforaddr(__DECONST(struct sockaddr *, l3addr), ifp);
2291 		if (ifa != NULL) {
2292 			ifa_free(ifa);
2293 			if (rt != NULL)
2294 				RTFREE_LOCKED(rt);
2295 			return 0;
2296 		}
2297 		log(LOG_INFO, "IPv6 address: \"%s\" is not on the network\n",
2298 		    ip6_sprintf(ip6buf, &((const struct sockaddr_in6 *)l3addr)->sin6_addr));
2299 		if (rt != NULL)
2300 			RTFREE_LOCKED(rt);
2301 		return EINVAL;
2302 	}
2303 	RTFREE_LOCKED(rt);
2304 	return 0;
2305 }
2306 
2307 static struct llentry *
2308 in6_lltable_lookup(struct lltable *llt, u_int flags,
2309 	const struct sockaddr *l3addr)
2310 {
2311 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2312 	struct ifnet *ifp = llt->llt_ifp;
2313 	struct llentry *lle;
2314 	struct llentries *lleh;
2315 	u_int hashkey;
2316 
2317 	IF_AFDATA_LOCK_ASSERT(ifp);
2318 	KASSERT(l3addr->sa_family == AF_INET6,
2319 	    ("sin_family %d", l3addr->sa_family));
2320 
2321 	hashkey = sin6->sin6_addr.s6_addr32[3];
2322 	lleh = &llt->lle_head[LLATBL_HASH(hashkey, LLTBL_HASHMASK)];
2323 	LIST_FOREACH(lle, lleh, lle_next) {
2324 		struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)L3_ADDR(lle);
2325 		if (lle->la_flags & LLE_DELETED)
2326 			continue;
2327 		if (bcmp(&sa6->sin6_addr, &sin6->sin6_addr,
2328 		    sizeof(struct in6_addr)) == 0)
2329 			break;
2330 	}
2331 
2332 	if (lle == NULL) {
2333 		if (!(flags & LLE_CREATE))
2334 			return (NULL);
2335 		/*
2336 		 * A route that covers the given address must have
2337 		 * been installed 1st because we are doing a resolution,
2338 		 * verify this.
2339 		 */
2340 		if (!(flags & LLE_IFADDR) &&
2341 		    in6_lltable_rtcheck(ifp, flags, l3addr) != 0)
2342 			return NULL;
2343 
2344 		lle = in6_lltable_new(l3addr, flags);
2345 		if (lle == NULL) {
2346 			log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
2347 			return NULL;
2348 		}
2349 		lle->la_flags = flags & ~LLE_CREATE;
2350 		if ((flags & (LLE_CREATE | LLE_IFADDR)) == (LLE_CREATE | LLE_IFADDR)) {
2351 			bcopy(IF_LLADDR(ifp), &lle->ll_addr, ifp->if_addrlen);
2352 			lle->la_flags |= (LLE_VALID | LLE_STATIC);
2353 		}
2354 
2355 		lle->lle_tbl  = llt;
2356 		lle->lle_head = lleh;
2357 		lle->la_flags |= LLE_LINKED;
2358 		LIST_INSERT_HEAD(lleh, lle, lle_next);
2359 	} else if (flags & LLE_DELETE) {
2360 		if (!(lle->la_flags & LLE_IFADDR) || (flags & LLE_IFADDR)) {
2361 			LLE_WLOCK(lle);
2362 			lle->la_flags |= LLE_DELETED;
2363 #ifdef DIAGNOSTIC
2364 			log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle);
2365 #endif
2366 			if ((lle->la_flags &
2367 			    (LLE_STATIC | LLE_IFADDR)) == LLE_STATIC)
2368 				llentry_free(lle);
2369 			else
2370 				LLE_WUNLOCK(lle);
2371 		}
2372 		lle = (void *)-1;
2373 	}
2374 	if (LLE_IS_VALID(lle)) {
2375 		if (flags & LLE_EXCLUSIVE)
2376 			LLE_WLOCK(lle);
2377 		else
2378 			LLE_RLOCK(lle);
2379 	}
2380 	return (lle);
2381 }
2382 
2383 static int
2384 in6_lltable_dump(struct lltable *llt, struct sysctl_req *wr)
2385 {
2386 	struct ifnet *ifp = llt->llt_ifp;
2387 	struct llentry *lle;
2388 	/* XXX stack use */
2389 	struct {
2390 		struct rt_msghdr	rtm;
2391 		struct sockaddr_in6	sin6;
2392 		/*
2393 		 * ndp.c assumes that sdl is word aligned
2394 		 */
2395 #ifdef __LP64__
2396 		uint32_t		pad;
2397 #endif
2398 		struct sockaddr_dl	sdl;
2399 	} ndpc;
2400 	int i, error;
2401 
2402 	if (ifp->if_flags & IFF_LOOPBACK)
2403 		return 0;
2404 
2405 	LLTABLE_LOCK_ASSERT();
2406 
2407 	error = 0;
2408 	for (i = 0; i < LLTBL_HASHTBL_SIZE; i++) {
2409 		LIST_FOREACH(lle, &llt->lle_head[i], lle_next) {
2410 			struct sockaddr_dl *sdl;
2411 
2412 			/* skip deleted or invalid entries */
2413 			if ((lle->la_flags & (LLE_DELETED|LLE_VALID)) != LLE_VALID)
2414 				continue;
2415 			/* Skip if jailed and not a valid IP of the prison. */
2416 			if (prison_if(wr->td->td_ucred, L3_ADDR(lle)) != 0)
2417 				continue;
2418 			/*
2419 			 * produce a msg made of:
2420 			 *  struct rt_msghdr;
2421 			 *  struct sockaddr_in6 (IPv6)
2422 			 *  struct sockaddr_dl;
2423 			 */
2424 			bzero(&ndpc, sizeof(ndpc));
2425 			ndpc.rtm.rtm_msglen = sizeof(ndpc);
2426 			ndpc.rtm.rtm_version = RTM_VERSION;
2427 			ndpc.rtm.rtm_type = RTM_GET;
2428 			ndpc.rtm.rtm_flags = RTF_UP;
2429 			ndpc.rtm.rtm_addrs = RTA_DST | RTA_GATEWAY;
2430 			ndpc.sin6.sin6_family = AF_INET6;
2431 			ndpc.sin6.sin6_len = sizeof(ndpc.sin6);
2432 			bcopy(L3_ADDR(lle), &ndpc.sin6, L3_ADDR_LEN(lle));
2433 			if (V_deembed_scopeid)
2434 				sa6_recoverscope(&ndpc.sin6);
2435 
2436 			/* publish */
2437 			if (lle->la_flags & LLE_PUB)
2438 				ndpc.rtm.rtm_flags |= RTF_ANNOUNCE;
2439 
2440 			sdl = &ndpc.sdl;
2441 			sdl->sdl_family = AF_LINK;
2442 			sdl->sdl_len = sizeof(*sdl);
2443 			sdl->sdl_alen = ifp->if_addrlen;
2444 			sdl->sdl_index = ifp->if_index;
2445 			sdl->sdl_type = ifp->if_type;
2446 			bcopy(&lle->ll_addr, LLADDR(sdl), ifp->if_addrlen);
2447 			ndpc.rtm.rtm_rmx.rmx_expire =
2448 			    lle->la_flags & LLE_STATIC ? 0 : lle->la_expire;
2449 			ndpc.rtm.rtm_flags |= (RTF_HOST | RTF_LLDATA);
2450 			if (lle->la_flags & LLE_STATIC)
2451 				ndpc.rtm.rtm_flags |= RTF_STATIC;
2452 			ndpc.rtm.rtm_index = ifp->if_index;
2453 			error = SYSCTL_OUT(wr, &ndpc, sizeof(ndpc));
2454 			if (error)
2455 				break;
2456 		}
2457 	}
2458 	return error;
2459 }
2460 
2461 void *
2462 in6_domifattach(struct ifnet *ifp)
2463 {
2464 	struct in6_ifextra *ext;
2465 
2466 	/* There are not IPv6-capable interfaces. */
2467 	switch (ifp->if_type) {
2468 	case IFT_PFLOG:
2469 	case IFT_PFSYNC:
2470 	case IFT_USB:
2471 		return (NULL);
2472 	}
2473 	ext = (struct in6_ifextra *)malloc(sizeof(*ext), M_IFADDR, M_WAITOK);
2474 	bzero(ext, sizeof(*ext));
2475 
2476 	ext->in6_ifstat = malloc(sizeof(counter_u64_t) *
2477 	    sizeof(struct in6_ifstat) / sizeof(uint64_t), M_IFADDR, M_WAITOK);
2478 	COUNTER_ARRAY_ALLOC(ext->in6_ifstat,
2479 	    sizeof(struct in6_ifstat) / sizeof(uint64_t), M_WAITOK);
2480 
2481 	ext->icmp6_ifstat = malloc(sizeof(counter_u64_t) *
2482 	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t), M_IFADDR,
2483 	    M_WAITOK);
2484 	COUNTER_ARRAY_ALLOC(ext->icmp6_ifstat,
2485 	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t), M_WAITOK);
2486 
2487 	ext->nd_ifinfo = nd6_ifattach(ifp);
2488 	ext->scope6_id = scope6_ifattach(ifp);
2489 	ext->lltable = lltable_init(ifp, AF_INET6);
2490 	if (ext->lltable != NULL) {
2491 		ext->lltable->llt_prefix_free = in6_lltable_prefix_free;
2492 		ext->lltable->llt_lookup = in6_lltable_lookup;
2493 		ext->lltable->llt_dump = in6_lltable_dump;
2494 	}
2495 
2496 	ext->mld_ifinfo = mld_domifattach(ifp);
2497 
2498 	return ext;
2499 }
2500 
2501 void
2502 in6_domifdetach(struct ifnet *ifp, void *aux)
2503 {
2504 	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
2505 
2506 	mld_domifdetach(ifp);
2507 	scope6_ifdetach(ext->scope6_id);
2508 	nd6_ifdetach(ext->nd_ifinfo);
2509 	lltable_free(ext->lltable);
2510 	COUNTER_ARRAY_FREE(ext->in6_ifstat,
2511 	    sizeof(struct in6_ifstat) / sizeof(uint64_t));
2512 	free(ext->in6_ifstat, M_IFADDR);
2513 	COUNTER_ARRAY_FREE(ext->icmp6_ifstat,
2514 	    sizeof(struct icmp6_ifstat) / sizeof(uint64_t));
2515 	free(ext->icmp6_ifstat, M_IFADDR);
2516 	free(ext, M_IFADDR);
2517 }
2518 
2519 /*
2520  * Convert sockaddr_in6 to sockaddr_in.  Original sockaddr_in6 must be
2521  * v4 mapped addr or v4 compat addr
2522  */
2523 void
2524 in6_sin6_2_sin(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2525 {
2526 
2527 	bzero(sin, sizeof(*sin));
2528 	sin->sin_len = sizeof(struct sockaddr_in);
2529 	sin->sin_family = AF_INET;
2530 	sin->sin_port = sin6->sin6_port;
2531 	sin->sin_addr.s_addr = sin6->sin6_addr.s6_addr32[3];
2532 }
2533 
2534 /* Convert sockaddr_in to sockaddr_in6 in v4 mapped addr format. */
2535 void
2536 in6_sin_2_v4mapsin6(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2537 {
2538 	bzero(sin6, sizeof(*sin6));
2539 	sin6->sin6_len = sizeof(struct sockaddr_in6);
2540 	sin6->sin6_family = AF_INET6;
2541 	sin6->sin6_port = sin->sin_port;
2542 	sin6->sin6_addr.s6_addr32[0] = 0;
2543 	sin6->sin6_addr.s6_addr32[1] = 0;
2544 	sin6->sin6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
2545 	sin6->sin6_addr.s6_addr32[3] = sin->sin_addr.s_addr;
2546 }
2547 
2548 /* Convert sockaddr_in6 into sockaddr_in. */
2549 void
2550 in6_sin6_2_sin_in_sock(struct sockaddr *nam)
2551 {
2552 	struct sockaddr_in *sin_p;
2553 	struct sockaddr_in6 sin6;
2554 
2555 	/*
2556 	 * Save original sockaddr_in6 addr and convert it
2557 	 * to sockaddr_in.
2558 	 */
2559 	sin6 = *(struct sockaddr_in6 *)nam;
2560 	sin_p = (struct sockaddr_in *)nam;
2561 	in6_sin6_2_sin(sin_p, &sin6);
2562 }
2563 
2564 /* Convert sockaddr_in into sockaddr_in6 in v4 mapped addr format. */
2565 void
2566 in6_sin_2_v4mapsin6_in_sock(struct sockaddr **nam)
2567 {
2568 	struct sockaddr_in *sin_p;
2569 	struct sockaddr_in6 *sin6_p;
2570 
2571 	sin6_p = malloc(sizeof *sin6_p, M_SONAME, M_WAITOK);
2572 	sin_p = (struct sockaddr_in *)*nam;
2573 	in6_sin_2_v4mapsin6(sin_p, sin6_p);
2574 	free(*nam, M_SONAME);
2575 	*nam = (struct sockaddr *)sin6_p;
2576 }
2577